The precipitation of candidate drugs on

mixing DMSO stock solutions with

aqueous buffers

By

Cheska Gillespie
A thesis submitted to the University of Strathclyde in partial fulfilment of the
requirements for the degree of Doctor of Philosophy.

Department of Pure and Applied Chemistry

Thomas Graham Building

295 Cathedral Street

Glasgow

G1 1XL
 This thesis is the results of the authors’ original research. It has been
composed by the author and has not previously been submitted for
examination which has led to the award of a degree.

The copyright of this thesis belongs to the author under the terms of
the United Kingdom Copyright Acts as qualified by University of
Strathclyde Regulation 3.50. Due acknowledgement always be made in
use of any material contained in, or derived from, this thesis.

Signed………………………………………………. Date………………………………

i
Acknowledgements

With thanks to;

My supervisor, Professor Peter Halling, for his time, guidance and support
throughout the past three years.

My industrial supervisor, Dr Pierre Daublain, MSD, for taking the time to give an
industrial opinion on the research.

Dr Darren Edwards, Univ. Strathclyde, for his knowledge and practical training on
the Tecan instrument, and his comments on the research.

Dr Alan Kennedy and Professor Alastair Florence for their comments on papers
drafts.

Lee Dowden (MSD Hoddesdon) for his help and training on the Biomek instrument.

Scott McKellar (formally Univ. Strathclyde) for assistance with the XRPD.

Craig Irvine (Univ. Strathclyde) and Ken Cameron (formally MSD Newhouse) for
performing NMR analysis.

Iain Larmour (Univ. Strath) for assistance with the Raman instrumentation.

Jonathan Smith and Patrick Hole (NanoSight) for their technical advice on the
instrument.

MSD and the University of Strathclyde for funding contributions.

My colleagues in TG 103 for their support and companionship.

ii
Abbreviations, Acronymns and Definitions

Ami - amiodarone HCl

AFM – Atomic Force Microscopy

BSA – Bovine Serum Albumin

CAC – Criticial Aggregate Concentration

Clot - clotrimazole

DLS – Dynamic Light Scattering

DMSO – Dimethyl Sulfoxide

DSC – Differential Scanning Calorimetry

FeSSIF – Fed State Simulated Intestinal Fluid

FIA-MS - Flow Injection Analysis-Mass Spectrometry

g - earth’s gravitational force, 9.80665 m/s2 (see RCF)

G.I Tract – Gastro-Intestinal Tract

GSE – General Solubility Equation

HCl - hydrochloride

HPLC – High Performance Liquid Chromatography

HTS – High Throughput Screening

LogD – octanol/water distribution co-efficient

M – molar concentration, mols per litre

mm – millimetre, 1 x 10-3 metres

mM – millimolar, 1x10-3 mols per litre

MSD – Merck Sharp and Dohme

NMR – Nuclear Magnetic Resonance

NTA – Nanoparticle Tracking Analysis

iii
ρ – density, g/cm3

PCMCs – Protein Coated Microcrystals

pI – isoelectric point, defined as the pH at which a protein carries no net charge.

PLM – Polarised Light Microscopy

RCF – Relative Centrifugal Force, quoted as value x g

SEM – Scanning Electron Microscopy

‘Shake-Flask’ Solubility – the solubility value obtained from dissolving a crystalline

solid in a given solvent system, at a given temperature, over an extended time
period with agitation until the system is believed to have reached equilibrium.

TEM – Transmission Electron Microscopy

μM – micromolar, 1x10-6 mols per litre

μm – micrometre, 1 x 10-6 metres

μL – microlitre, 1 x 10-6 litres

UV – Ultra-Violet

XRPD – X-Ray Powder Diffraction

% - percent

iv
 Abstract

The precipitation of candidate drugs from dimethylsulfoxide (DMSO) stock solutions

and buffer mixtures has practical importance within the pharmaceutical industry. At
the drug discovery stage, mixing of drug DMSO stocks and aqueous buffer is a
common procedure for many analyses, such as bioassay screening and some
solubility screens. Precipitation from these mixtures, whilst a recognised issue, has
little in the way of published work. This thesis explores in detail the precipitation of
three poorly water soluble, commercially available drug molecules, namely
amiodarone HCl, clotrimazole and tolnaftate. The work presented here assesses the
process from a physical chemistry point of view.

Upon mixing the DMSO drug stock and aqueous buffer, all three compounds form
particulates in the nanometre size region, which grow over time. The growth of
these particles can be monitored using nanoparticle tracking analysis (NTA) and the
NanoSight. We show that the presence of these particles can interfere with the
results of ‘kinetic solubility’ measurements, depending on the exact analysis used. It
is already known that these types of particles can interfere with bioassay results,
causing false positives. Variables such as the mixing employed, the exact percentage
of DSMO present in the samples and the concentration of protein present in the
samples all affect the formation and growth of these particles. The results were
correlated to high performance liquid chromatography (HPLC) measurements. It can
therefore be concluded that there are several, controllable variables which can
affect the precipitation of the three test compounds investigated here, and thus
potentially the results of some discovery stage pharmaceutical screening assays.

v
Contents

Abbreviations, Acronymns and Definitions............................................................... iii

Abstract .................................................................................................................... v
1. Theory and Literature....................................................................................... 6
1.1 Introduction ............................................................................................... 6
1.2 Theory ....................................................................................................... 8
1.2.1 Solubility ............................................................................................. 8
1.2.2 Crystallisation ................................................................................... 13
1.3 Literature Review..................................................................................... 18
1.3.1 Overview of literature explored ........................................................ 18
1.3.2 Storage in DMSO and its shortcomings ............................................. 18
1.3.3 Utilisation of the DMSO stock solution – bioassays ........................... 21
1.3.4 Aggregation based non-stoichiometric inhibition in HT activity screens
.............................................................................................................22
1.3.5 Utilisation of DMSO stock – kinetic solubility assay ........................... 26
1.3.6 Previous unpublished work in this area ............................................. 30
1.4 Aims ........................................................................................................ 33
1.5 Model Compound Selection ..................................................................... 35
1.5.1 Compilation of initial compound list ................................................. 35
1.5.2 Model compounds used in detailed experiments – general information
.............................................................................................................38
1.6 References ............................................................................................... 42
2. Monitoring of particle growth at a low concentration of a poorly water soluble
drug using the NanoSight LM20 ............................................................................. 45
2.1 Abstract........................................................................................................ 45
2.2 Introduction ............................................................................................. 47
2.3 Materials and Methods ............................................................................ 49
2.3.1 Tolnaftate Sample Preparation ......................................................... 49
2.3.2 NanoSight – Sample Procedure ......................................................... 49

1
 2.3.3 Video Capture Settings ..................................................................... 50
2.3.4 Processing Parameters...................................................................... 51
2.3.5 NanoSight – Preparation and analysis of a blank sample ................... 52
2.3.6 DLS Procedure – Sample ................................................................... 53
2.3.7 Preparation and analysis of a blank – DLS ......................................... 53
2.4 Results and Discussion ............................................................................. 54
2.4.1 NanoSight Results ............................................................................. 54
2.4.2 DLS results ........................................................................................ 61
2.5 Conclusions .............................................................................................. 67
2.6 References ............................................................................................... 69
3. Detailed study of precipitation of a poorly water soluble test compound using
methodologies as in activity and solubility screening - mixing and automation
effects.................................................................................................................... 70
3.1 Introduction ............................................................................................. 70
3.2 Materials and Methods ............................................................................ 73
3.2.1 Stock and Buffer Preparation ............................................................ 73
3.2.2 Instrumentation................................................................................ 73
3.2.3 Sample Preparation using automated systems.................................. 75
3.2.4 Sample preparation by manual pipetting .......................................... 75
3.2.5 Mixing............................................................................................... 75
3.2.6 Centrifugation and Dilution............................................................... 76
3.2.7 NTA and NanoSight LM10 Analysis .................................................... 77
3.2.8 Variation present in separation and dilution steps ............................ 77
3.2.9 Kinetics of precipitation .................................................................... 78
3.2.10 Dissolution from solid ................................................................... 78
3.3 Results and Discussion ............................................................................. 79
3.3.1 Dissolution from solid/kinetics of precipitation ................................. 79
3.3.2 Variation present in separation and dilution steps ............................ 79
3.3.3 Decrease in supernatant amount measured as more experiments
performed ...................................................................................................... 80
3.3.4 Results and comparison of data between Labs ................................. 85

2
 3.3.5 Comparison of variability seen in robotically and manually pipetted
samples .............................................................................................................87
3.3.6 Comparison of mixing variables ........................................................ 87
3.3.7 NTA Particle size analyses ................................................................. 88
3.4 Conclusions .............................................................................................. 91
3.5 References ............................................................................................... 93
4. Investigating the effects of nanoparticle formation on the kinetic solubility
determination of 3 hydrophobic test compounds .................................................. 94
4.1 Introduction ............................................................................................. 94
4.2 Materials and methods ............................................................................ 97
4.2.1 DMSO stock and buffer preparations ................................................ 97
4.2.2 ‘Shake-flask’ based method of solubility determination .................... 98
4.2.3 GSE ................................................................................................... 98
4.2.4 Precipitation method 1 - NTA determination of drug concentration at
which nanoparticles are first detected ........................................................... 99
4.2.5 Precipitation Method 2 - Analysis by HPLC and subsequent
investigation of separation and kinetics by NTA ............................................. 99
4.2.5 Instrument Details ...........................................................................101
4.3 Results and discussion ............................................................................103
4.3.1 ‘Shake-flask’ solubility and GSE results.............................................103
4.3.2 Methods originating from DMSO stock solutions – precipitation
methods ........................................................................................................103
4.3.3 Precipitation methods and varying incubation times .......................110
4.4 Conclusions .............................................................................................116
4.5 References ..........................................................................................117
5. The effect of % DMSO on solubility measurements and nanoparticle behaviour
.........................................................................................................................119
5.1 Introduction ............................................................................................119
5.2 Materials and Methods ...........................................................................121
5.3 Results and Discussion ............................................................................122
5.3.1 ‘Shake-Flask’ solubility .....................................................................122
5.3.2 Kinetic Solubility Assay Results ........................................................123

3
 5.3.3 1 and 5% DMSO and incubation times .............................................128
5.4 Conclusions .............................................................................................132
5.5 References ..........................................................................................133
6. Non-Stoichiometric, aggregate based ............................................................134
inhibition: Bioassay trends explained via nanoparticle growth measurements ......134
6.1 Abstract.......................................................................................................134
6.1 Introduction ............................................................................................135
6.2 Experimental ..........................................................................................136
6.2.1 Material s and Methods ...................................................................136
6.2.2 Stock and Buffer preparations .........................................................136
6.2.3 Sample preparation .........................................................................137
6.3.4 Instrument Details ................................................................................138
6.3 Results and Discussion ............................................................................139
6.4 Conclusions .............................................................................................147
6.5 References ..............................................................................................148
7. Attempts at nanoparticulate crystalline form determination.........................149
7.1 Introduction ............................................................................................149
7.2 Experimental ..........................................................................................151
7.2.1 Materials .........................................................................................151
7.2.2 DMSO stock, buffer and drug suspension preparations....................151
7.2.3 Instrumental Details ........................................................................152
7.3 Results and Discussion ............................................................................154
7.3.1 1 mL scale analysis ...........................................................................154
7.3.2 1 mL scale analysis - Raman .............................................................156
7.3.3 Scale up (500 mL) and XRPD analysis ...............................................160
7.4 Conclusions .............................................................................................167
7.5 References ..........................................................................................169
8. Overall Conclusions and Future Work ............................................................170
9. Appendix ........................................................................................................... I
9.1 Chapter 1 Additional Data ........................................................................... I
9.2 Chapter 2 Additional Data ....................................................................... VIII

4
9.3 Chapter 3 Additional Data ........................................................................ XV
9.4 Chapter 5 Additional Data ..................................................................... XXIII
9.5 Chapter 6 Additional Data ...................................................................... XXV

5
 1. Theory and Literature
1.1 Introduction

High Throughput Screening (HTS) is commonplace within the pharmaceutical

industry, and is applied to many different processes, for example solubility
measurements,1,2,3 bioassays, 4 and salt and polymorph screens.5 The need for HTS
within the industry evolved from the sheer volume of compounds being
synthesised, and the need to rapidly but efficiently evaluate their properties and
activity, in order to streamline the number of compounds from thousands to the
most promising few. HTS occurs early in the drug discovery process and involves
miniaturization of the technique in question, allowing it to be automated, resulting
in rapid testing and data production.6,7 It is important to maintain low cost and use
the minimal amount of compound, especially at the early stages of drug discovery. 6
Whilst speed is an important part of these screening methods, they must also be
robust and reliable. Promising, possibly even lifesaving drugs could potentially be
disregarded if the screening process is not efficient.

In order to maximise the efficiency of the HTS process, compounds are stored as
stock solutions dissolved in dimethylsulfoxide (DMSO). This allows easier
automation compared to dry compound storage, as liquid handling is readily
performed by robots.1,8 DMSO is a dipolar aprotic solvent with a large dielectric
constant, meaning interaction with a molecule containing a dipole is favourable. 9
Since most drugs contain a dipole, the majority of pharmaceutical compounds will
dissolve in DMSO, making it ideal for use in an HTS set up. Aqueous stock solutions
are not used, due to the poor water solubility of many drug molecules. Different
assays have different tolerance for DMSO, and generally the percentage of solvent
is kept to a minimum to limit these effects. 4

For both bioassays and solubility analyses at the early discovery stage, the stock
solution in DMSO is introduced to an aqueous buffer system. As the aqueous
solubility of many of the compounds under test is poor, precipitation may occur.

6
Low aqueous solubility is common, as drug molecules are optimised with a view to
increasing their activity against a particular target, and the motifs introduced for
this purpose tend to increase the lipophilicity of the drug. 10 In solubility assays,
precipitation is indicative that the solubility limit has been reached, and is an
anticipated phenomenon. In a bioassay setting, precipitation is undesirable, and
results in a poor assessment of the compound’s activity towards a particular target.
There is a growing realisation within the industry that precipitation may occur
during activity analyses.4 Pope-Burke et al note that whilst long term stability of the
aqueous solutions used in assays is not a relevant issue, the initial solubility and
stability can be.11 The concentrations used in HTS are low, typically in the µM range.
Literature on the crystallisation of organic compounds, which has been written from
a preparative viewpoint, generally deals with higher concentrations. Despite the
implications for HTS, little is known about the factors which affect crystallisation
and precipitation at these low concentrations.

7
1.2 Theory

Solubility is a widely studied topic, particularly in the pharmaceutical industry. Since

aqueous solubility is directly related to absorption and bioavailability in vivo, it is an
important physical property of drugs which are intended to be administered
orally.12 Oral dosing remains the most popular method of drug delivery, with
advantages such as self administration and low cost of manufacture. 13
Crystallisation has also been studied extensively, and theory and principles of
crystallisation are widely available in literature. Both crystallisation and solubility
underpin the work in this project, with an understanding of the theory in the
literature crucial in order to determine whether the systems which will be studied
here behave as would be expected. These two processes will now be discussed in
turn.

1.2.1 Solubility

The solubility of a compound is the maximum amount of solid that can be dissolved
in a certain volume of liquid, under known solution conditions. The ‘true’
thermodynamic solubility of a compound is regarded as that of the most stable solid
form which has been achieved under equilibrium conditions i.e. the amount
dissolved does not vary over time, and is equal to amount left molecularly dissolved
in solution after a precipitation has reached full completion. From the perspective
of the pharmaceutical industry, the aqueous solubility of a compound is directly
related to absorption and bioavailability in vivo.12 The rate at which a compound
dissolves is known as the dissolution rate, and can be described by the Noyes-
Whitney equation;

Equation 1

8
Where;

J is the dissolution rate

k is a constant consisting of the diffusion coefficient divided by the diffusion length

A is the surface area of the solute

Cs is the solubility

C is the concentration of solute in the medium

The dissolution rate is related to the solubility of the compound.14

Many factors affect the experimental solubility value of a molecule, and can be
broadly grouped into three categories;

i. Experimental conditions used

The medium used, the temperature, and the time that the sample has been left to
equilibrate for all affect the solubility value seen, and it is important to note these
conditions when quoting a solubility value. Once the equilibrium solubility has been
reached, time does not affect it – however whether or not equilibrium is reached
within the time frame of the experiment will affect the result obtained. Any
filtration processes performed, and even the type of instrument used to analyse the
sample can also affect the result of a solubility experiment.12 In addition, the mixing
process employed contributes. The addition of salts to the medium can have a
‘salting in’ or salting out’ effect, resulting in a higher or lower solubility,
respectively.9,14

ii. Properties of the solute

Structural properties of the molecule, such as the pKa, lipophilicity and size, can also
affect the solubility, and modifications to the structure can change the solubility.
The pKa of the molecule, and the pH of the solution, have a large impact on the
solubility seen. In general, the ionised form of the molecule is more soluble than the
neutral form (intrinsic solubility), and so the solution pH should be stated when

9
quoting solubility values. At pH = pKa, 50% of the molecules in the solution are
ionised, and the solubility is increased compared to the intrinsic solubility.14 The
solubility of salts is also dependent on the counter ions present. Modifications to
improve aqueous solubility include addition of ionisable groups (e.g. –COOH), or
addition of hydrogen bond donors and acceptors (e.g. –NH2, -OH). Alteration of the
structure of the molecule, and the pH, affect the equilibrium solubility value (due to
a change to a different but related molecule, or due to changes in the percent of
the drug ionised).12

iii. Physical form of the solute

The physical form the solute is presented in, e.g. whether crystalline or amorphous,
and what polymorph is used, affects both solubility and dissolution rate. Crystalline
solids tend to have lower solubility values than their amorphous counterparts.15 The
physical state can also include the particle size of the solid, and whether a salt form
of the drug is used. It should be noted that the physical state in which the
compound is introduced into the medium affects the dissolution rate and the
apparent solubility of the compound.14 The apparent solubility is the solubility of a
solute in a form which is not the most thermodynamically stable, or has been
altered in some way, for example by particle size reduction.14 The equilibrium
solubility is not affected. Although amorphous particles tend to be more soluble
than crystalline ones, if left for sufficiently long enough, the material will convert to
its most stable form, and the solubility approaches that of the thermodynamic
value. Use of different crystalline forms can only increase dissolution rate and give a
temporary change in solubility, however these effects may be useful in the
pharmaceutical industry to increase absorption within a certain time period, for
example to allow absorption within the gastro-intestinal (G.I.) tract to occur.14

Particle size reduction can be achieved by micronisation, milling or micro-

crystallisation, and increases the dissolution rate and apparent solubility of the
compound. From the Noyes-Whitney equation, it can be seen that an increase in

10
surface area (per unit volume) due to smaller particles would result in an increase in
dissolution rate. From the Kelvin equation;

( )

Equation 2

Where;

γ is the surface energy of the solute

V is the molecular volume of the solute

S0 is the solubility of an infinity large particle

R is the universal gas constant

T is temperature (Kelvin)

R is particle radius of the solute

S is the solubility of the solute

the apparent solubility of the compound also increases as particle size (radius)
decreases. The processes used to reduce particle size can also result in
imperfections within the solid - it is less ‘crystalline’ than before, is not the most
stable form of the solid, and therefore only it’s apparent solubility is altered. 14

Another common way of increasing solubility and dissolution rate of a drug is salt
formation. Again, the equilibrium solubility is not affected - it does not matter
whether the compound is introduced as the salt or the free acid/base, the solubility
achieved by both is the same at the same pH. The salt form will, however reach this
value faster and can achieve supersaturation, whereas the free acid/base would not
normally dissolve to supersaturation. This means, for example, that the salt form of
a drug could fully dissolve to its maximum solubility or even achieve and maintain
supersaturation, prior to transit through the gastro-intestinal (G.I.) tract, allowing

11
absorption to occur. The free acid/base, on the other hand, may not even reach
maximum dissolution during that time.12

12
1.2.2 Crystallisation

The process of crystallisation can be broken down into three main stages;

i. The formation of a supersaturated (non-equilibrium conditions) solution

ii. Nucleation
iii. Growth 16

These will now be briefly described in turn.

i. Supersaturation

Once the maximum amount of solute has been dissolved in a solvent, the solution is
said to be saturated, and is in equilibrium with the solid phase. It is possible, e.g. by
evaporation or cooling, to achieve a solution where there is more dissolved solute
than would be seen under equilibrium conditions. This is known as supersaturation.
In relation to screening, supersaturation can occur when the stock solution in DMSO
is added to the aqueous buffer – it is possible to achieve a higher concentration of
solubilised drug in the (mainly) aqueous solution than would be seen under normal
conditions. The relationship between saturation, supersaturation and crystallisation
can be depicted as follows;

13
Figure 1: Supersaturation diagram

In region A, the solution is undersaturated with respect to the solute. The line
between regions A and B represents the saturation solubility. Both regions B and C
are supersaturated, however there are differences between the two in terms of
nucleation. In region B, no spontaneous crystallisation can occur, and only seed
crystals can grow. Spontaneous nucleation cannot occur until Region C. 9 The terms
‘metastable’ and ‘labile’ were introduced by Ostwald, who was the first to specify
these conditions exactly.17

ii. Nucleation

A nucleus can be defined as ‘the minimum amount of a new phase capable of

independent existence’.16 Experimentally, the existence of a critical nucleus was
first demonstrated by Oswald.17 Erdemir et al compare and contrast two nucleation
theories, classical nucleation theory, and the two step model of nucleation. Classical

14
nucleation theory is the simplest and most commonly used theory to explain the
process of nucleation – here, a critical radius must be reached before it is
energetically favourable for the nucleus to remain as a cluster and grow rather than
re-dissolve. There is a competition between the surface free energy, which favours
dissolution, and the volume free energy, which favours growth – only at nuclei of a
certain size will growth become possible. This theory assumes that the nuclei are
already ordered in a manner that is identical to the bulk crystal, and that the
nucleation rate is time independent. It also assumes that growth occurs by addition
of one monomer at a time. The theory has a number of shortcomings, for example
it cannot predict absolute nucleation rates, no information on the structure of
aggregates or the pathway from solution to solid is provided, and the only criterion
of whether aggregates form nuclei or not is radius. These shortcomings, along with
experimental discrepancies, have prompted the development of other theories,
such as the two step nucleation theory. This theory proposes that nucleation is a
two step process, with the first step involving the formation of a highly disordered
liquid droplet, then re-arrangement of the droplet to a crystalline nucleus beyond a
certain critical size. Here, the second, crystalline forming step is the slow, rate
determining step. Computational, theoretical and experimental studies are
described in the work of Erdemir et al which support the two step nucleation
model.16

Factors which affect nucleation include impurities present in the system,

temperature, the amount of supersaturation in the system, and mechanical shock
e.g. sonication, agitation. It has been found that agitation can induce nucleation at
lower degrees of supersaturation compared to a solution left unperturbed, however
it is thought that there may be circumstances where agitation can in fact lower the
tendency to nucleate. Impurities can increase or decrease nucleation rates,
depending on the system.9

15
 iii. Growth

Many growth theories exist, and can broadly be categorised into three groups;
surface energy theories, diffusion theories and adsorption layer theories. Surface
energy theories have the basis that the shape of a growing crystal is that with the
minimum surface energy – Gibbs suggested that ‘the total free energy of a crystal in
equilibrium with its surroundings at constant temperature and pressure would be a
minimum for a given volume’. The main limitation of this type of theory is its
inability to explain the effects of supersaturation and solution movement on
growth, and so surface energy theories are rarely used.

Diffusion theories presume that material is deposited onto the surface of the crystal
continuously, at a rate proportional to the concentration difference between the
point of deposition and the bulk solution, and initial theories regarded
crystallisation as the inverse of dissolution. This is not the case, as most solids will
dissolve faster at the same temperature and pressure conditions than they will
crystallise. Later modifications to the theory suggested that a two step process
occurred; a diffusion process where molecules are transferred from the bulk to the
solid, then a rearrangement process where the molecules arrange themselves into
the crystal lattice. These processes are driven by differences in concentrations.

Adsorption layer theories suggest that crystal growth takes place by layer-by-layer
adsorption onto the surface, arising from imperfections. The theory is based on
thermodynamic reasoning. When the units of crystallising substance arrive at the
crystal face, they do not immediately form part of the lattice, but instead lose one
degree of freedom and are adsorbed loosely onto the surface. This ‘third phase’ is
in dynamic equilibrium with the bulk solution. Some imperfections in the layer-by-
layer adsorption are seen, and these dislocations, the most important of which is
the screw dislocation, cause steps to be formed on crystal faces, promoting
growth.9

After formation and growth of particles, other processes can occur, particularly in
systems where fast crystallisation has occurred. Small particles have a tendency to

16
aggregate in solution and form clusters. This is most common at high
supersaturation.9 This has been discussed by McGovern et al in relation to the
effect of aggregates on bioassays results (see section 1.3.4).18, 19 Particles can also
age in solution, by both ripening and phase transformations. As discussed
previously, smaller particles are more soluble than larger particles – in a sample of
particles in solution, the smaller particles tend to dissolve, and then the solute re-
deposits itself onto the larger particles, which is more energetically favourable. This
process is called Ostwald ripening, and has the overall result of increasing the size of
the larger particles and decreasing the size of the smaller particles.14 In theory, the
particle size distribution of a sample undergoing ripening should change over time
to become more monodisperse, however the distribution will initially become more
polydisperse until ripening is completed. Ripening tends to occur at low
supersaturation.9 There is also the possibility of the initial solid being a metastable
(e.g. amorphous) precipitate, and then phase transformation to the most stable
form over time occurs. This follows the principle of Ostwald’s rule of stages – an
unstable system does not always transform into the most stable state, but the one
which is closest in energy to its own. This means that the form which precipitates
fastest is not always the most thermodynamically stable form.9

17
1.3 Literature Review

1.3.1 Overview of literature explored

As discussed in the introduction, the precipitation of drugs during screening upon

mixing DMSO stock and aqueous buffer is an important issue. Much of the literature
on this topic deals with practical steps to minimise precipitation4 and the effects of
any small particles on the results of an enzyme assay,18-26 rather than fully
investigating the underlying process of solid formation or parameters affecting
precipitation. There is also literature available on the stability of the DMSO stock
solutions,27-34 which could also affect the result from the assay. Kinetic solubility
assay methods, where precipitation is utilised as an indicator of the end point of the
assay, are also explored within the literature.2, 3, 10, 15, 35-38 Since the protocols for
both activity and solubility screening have similarities in both concentration and use
of DMSO stock, the methods used for solubility screens and observations from such
experiments are useful in obtaining information on the factors which affect
formation of solid at these low concentrations.

1.3.2 Storage in DMSO and its shortcomings

DMSO is regarded as a ‘universal’ solvent, and will readily dissolve most

hydrophobic drug compounds. It is miscible with water, has a reasonably high
polarity, low volatility and is relatively cheap and non-toxic.32 It is also ideal for use
in biological assays as it does not interfere with the biological components in low
concentrations (typically 1-5% by volume), unlike other organic solvents. 30 Dry
DMSO cannot form hydrogen bonds with itself as it only has acceptor capability,
and so solvent cavities are easily formed around the drug molecules. Cavity
formation cannot occur in water due to hydrogen bonding between the water
molecules, which the drug molecules cannot disrupt. DMSO and water have very
pronounced interactions, which is the drive for DMSO’s hygroscopicity. In a
DMSO/water mixture, the organic solvent acts as a hydrogen bond acceptor and
forms hydrogen bonds with the donor water molecules, giving a more structured

18
solvent compared to DMSO alone. This makes it harder for cavities to form, and so
the solubility of a molecule is decreased compared to dry DMSO. At 33% water (by
weight) in DMSO, the structure of the solvent is more ordered than that of pure
water.29

Typically, stock solutions are stored at concentrations of 10-30 mM,4,12 with the
mother stock solution being diluted as required for subsequent analyses. Whilst
DMSO has several advantages, it has been reported that 10-20% of compounds are
in fact insoluble in DMSO.4 There are also concerns over the uptake of water into
the DMSO stock solution which can result in the precipitation of the drug from the
stock solution,34 changing the concentration of the solubilised compound, and any
further dilutions will therefore be erroneous.12 Stocks are usually stored frozen at
low temperatures to minimise degradation, with Blaxill et al demonstrating that
compound integrity (affected by either precipitation or degradation) is related to
both temperature and storage time. Based on extrapolations from experimental
stability results (from 530 structurally diverse samples) at room temperature and 4
°C, they recommend samples stored at 4 °C were stable for 3 months, and samples
stored at -20 °C stable for up to five years.33 Kozikowski et al used statistical analysis
of experimental FIA-MS data to calculate the probability of observing the correct
compound at a given time point. They concluded that after 6 months storage at
ambient conditions, there is still a good chance (83 %) that the mass spectrum
matches that produced at time 0. The work done here looked at around 7200
compounds stored at room temperature for a year.28 Cheng et al performed an
accelerated stability study at 40 °C, and from the results calculated that compounds
stored in water free (under inert gas atmosphere) DMSO are stable (≥ 80 %
compound retention) for up to a year at room temperature. The use of low
temperatures to store the DMSO stock solutions has also been explored – this
however lowers the compound’s solubility, and can result in the same issue as
water uptake i.e. precipitation.4 Both Blaxill et al and Ilouga et al have reported that
lower storage temperatures lead to a reduction in precipitation/degradation
compared to higher temperatures – if precipitation is the main cause of issues in

19
the stock solution, it would be expected that lower temperatures would exacerbate
the issue as the solubility would be lower – it appears that lower temperatures have
minimised water uptake in these cases. Cheng et al also state the importance of
controlled humidity when storing the stock solutions.34

Frozen stock solutions are put through a series of ‘freeze-thaw’ cycles, again
increasing the risk of the compound ‘crashing out’,27,29 with Oldenburg et al
reporting a synergistic effect between water uptake and freeze thaw cycles. Cheng
et al report no precipitation or degradation after 11 freeze-thaw cycles.34 As well as
avoiding water uptake and minimising freeze thaw cycles, other practical measures
to avoid precipitation include the use of sonication post freeze thaw,29 and the use
of an additive to retard crystallisation.39 Lowering stock solution concentrations is
also a recommendation, with the optimum concentration reported to be 2-5 mM.11
Results of studies of compound storage and stability are difficult to compare due to
the differing conditions used, including different experimental/statistical analysis
methods used and different criteria/definitions of ‘stable’. These all tend to be
based on the authors’ current practice.31 This means that whilst there are several
papers on this issue, none give a definitive answer with regards to which storage
conditions are optimum, and apparently contradictory results and time limits are
seen. There also seems to be a balance which needs to be achieved between the
use of low temperatures to minimise degradation, and ensuring that no
precipitation is occurring due to low solubility at decreased temperatures, whilst all
the while minimising water uptake. Water uptake by DMSO is a major issue, with all
authors agreeing that increased water content enhances precipitation and
degradation in the stock solution.

20
1.3.3 Utilisation of the DMSO stock solution – bioassays

The DMSO stock solution is used as appropriate in a number of different analyses,

for example ‘kinetic solubility’ assays and bioassays. In both analyses, the stock is
added to an aqueous buffer system, with the pH used dependent on the analysis.
The percent DMSO used in bioassays is dependent on the individual bioassay
tolerance for DMSO, although overall it has limited biological effect at low
concentrations.30 The small percentage of organic solvent can also help solubilise
the drug in the aqueous media, by disrupting the hydrogen bonding network in
water. Since this would not give a true ‘aqueous’ solubility, the percent DMSO in
solubility screens is kept low to try and minimise this solubilisation effect. Even at
the low concentrations used for drug screening, there are significant differences
between the solubility of the drug in DMSO and that in an aqueous solution. 11

The addition of the DMSO stock solution to the aqueous system can cause
supersaturation, which can be desirable, but can also lead to precipitation of the
drug. In a bioassay setting, if the drug precipitates upon addition of the aqueous
media due to low solubility, its activity against the target cannot be properly
assessed due to the fact that some of the drug added initially is not present in
solution. This is demonstrated by the fact that low solubility libraries have lower
HTS hit rates.4 It is estimated that around 30 % of discovery compounds have
aqueous solubility lower than 10 µM, which is within the concentration range
typically used in screening (although the exact concentrations used depend on the
assay in question).40 There is also the observation that activity can vary greatly
depending on the exact assay composition.12 Popa-Burke et al concluded that the
solubility of the drug in the assay medium is more of an issue than stability – the
plates are kept for around 24 hours on average. The group also note that the actual
amount of drug present in the assay samples is significantly different to that
expected, even at µM concentrations.11

21
A number of recommendations have been made in the literature to either prevent
precipitation or to re-dissolve any solid material formed. 4, 29 These include
performing serial dilutions in DMSO rather than diluting with aqueous buffer, using
kinetic solubility data to adjust results from bioassays, reducing screening
concentrations, 4 and the use of in-well sonication to re-dissolve any solid material
formed. 29

1.3.4 Aggregation based non-stoichiometric inhibition in HT activity

screens

In 2002, McGovern et al reported that a number (15) of diverse compounds,

introduced into aqueous media via DMSO stocks, all inhibited a series of unrelated
model enzymes. The compounds did not behave as stoichiometric inhibitors, and in
a screening context, they would give a false positive result. Non-stoichiometric
inhibition is an undesirable characteristic, and means the compound is
demonstrating non-classical interaction behaviour with the enzyme. The group
found that these compounds all had common, distinctive behaviours – inhibition
was time dependent, there was a change in the IC50 value when enzyme
concentration was increased, the compounds all had unusually steep Hill curves,
and showed sensitivity to detergent.22,20 All of the compounds formed aggregates
(diameter 95 – 400 nm) in solution, which were detectable by DLS. 18

It was proposed that these aggregates were responsible for target inhibition.
Several properties of the aggregates themselves were investigated – the aggregates
were ‘destroyed’ upon the addition of detergent, and formed at a ‘critical aggregate
concentration’ (CAC). TEM images of the aggregates were obtained, showing
spherical particles. Further investigation of these aggregates suggested micelle like
behaviour, due to the CAC, the fact that the dissolved content remained constant
above the CAC, and that a fairly constant size distribution was obtained, despite
increasing total compound amount.26 Note that whilst this is consistent with
micelles, it is also consistent with precipitation. None of these papers mentions the
equilibrium solubility of the compounds used, or if the concentrations used are in

22
fact above this. The authors however, state that they do not think the aggregates
are an early stage of precipitate, due to the ease with which the aggregates re-
dissolve upon dilution to below the CAC. They refer to the difficulty at driving
precipitated organic material back into aqueous solution, and reason that since
dissolution of the particles after dilution is rapid, the particles are not precipitate. 26
Another way to think of this is that, typically, solution in contact with an organic
precipitate is likely to be saturated, and so it is thermodynamically unfavourable for
the solid to re-dissolve. In the case of the aggregates, dissolution only occurs when
the compounds are diluted to below the CAC i.e. below the solubility limit of the
aggregates. Since the particles are in the nm region, their solubility will be slightly
higher than larger particles due to higher surface area. The small size of the
particles also allows for much faster dissolution compared to larger particles.
Compounds which formed these aggregates were present in both screening
libraries and marketed drugs.19

Of major interest was exactly how these aggregates inhibited the enzyme.
Denaturation was ruled out, however some small scale unfolding was found. 24 It
was hypothesised that the enzyme was sequestered on the aggregate surface, with
Coan et al demonstrating that there is more than enough surface available on these
aggregates to accommodate the inhibited enzyme.26 Exactly what interactions
govern this surface adsorption is as of yet unknown, though they are likely to be of
a hydrophobic nature. The phenomenon has some features in common with PCMC
formation – both processes involve the surface adsorption of proteins to small
molecule clustered species.41 Several differences are of course also present (the
solubility of the protein in the respective media used and the crystalline form and
size of the coated clusters being the main ones), but a similar overall effect is seen.
Several attempts have been made at both prediction of compounds which would
form these aggregates 19, and high throughput methods of detection of aggregates .
From a prediction point of view, Seidler et al found that whilst structural similarity
was not related to aggregating capability (based on 5 related azole antifungals),
hydrophobicity was. Physical properties such as solubility and clogP values were

23
fairly reliable indicators, however they were still not accurate enough to classify
compounds definitively (clogP cut off applied – 81% correctly classified from 111
compounds, solubility cut off applied - 87% correctly classified from 111
compounds). Using more complicated prediction methods incorporating molecular
descriptors, the success rate was increased to 94% of 111 compounds correctly
predicted as aggregating or non-aggregating.19 From a detection point of view, DLS
has been used directly to detect these aggregates, as have TEM/SEM – however
these two microscopy techniques cannot be used in a high throughput manner. Not
mentioned in bioassay screening literature, but of potential use are the light
scattering detectors used in kinetic solubility type analysis. These can be utilised in
multi-well plates, and could potentially offer some use here for detection purposes
in a high throughput manner. These methods are said to have a lower limit of 20
uM.35 More indirect methods of determining aggregators include testing for
detergent and enzyme concentration sensitivity. Habig et al found that some non-
aggregating compounds could be sensitive to detergent, and that sensitivity to
enzyme concentration was not only a more reliable indicator of an aggregating
inhibition, it also eliminated non-stoichiometric inhibition in the same way as the
addition of detergent would.21 Kerns et al recommend lowering screening
concentrations within bioassays to e.g. 5 μM, where aggregation is less common.12
Owen et al42 explored the effect of aggregate formation of anti-cancer drugs within
cell culture. They found that three compounds have the ability to form colloidal
aggregates in cell culture media and that efficacy of the aggregates is lowered. The
presence of aggregates in cell based assays is therefore likely to results in a false
negative effect – interestingly this is the opposite of the false positive effect that
aggregates have on enzyme based assay.

Whilst several papers have documented thoroughly the effect of these aggregates
on the results of activity assays, little is known about the aggregates themselves.
Their physical form, behaviour under assay conditions, kinetics of formation/growth
and thermodynamic stability in assay medium are all unknown.

24
An interesting follow on from aggregation in vitro affecting bioassay results is the
possibility of it occurring in vivo and directly affecting drug absorption into the
body. Frenkel et a43l looked at this possibility by studying a series of non-nucleoside
reverse transcriptase inhibitors (NNRTI’s, used as anti- AIDs drugs) and correlating
their adsorption in rats and humans (from blood plasma studies) with their
aggregation properties. These drugs tend to be very potent and extremely
hydrophobic. In media which mimicked the in vivo gastrointestinal environment, it
was found that the compounds formed aggregates which were measurable by DLS.
The aggregating compounds fell into two groups; those which formed aggregates
30-110 nm in radius and those which formed aggregates >250 nm in radius. Those
forming smaller sized aggregates showed greater absorption, with those forming
bigger sized aggregates having poorer absorption behaviour. If a compound could
maintain these smaller particle sizes for the duration of the GI tract transit time
(and while encountering continuous pH increase), it is likely more drug will be
absorbed. The particle size and absorption data suggest that this class of drug is
taken up in the absorption tests in aggregate form, and it is suggested that this may
occur for many other hydrophobic compounds where aggregates are formed.
Whilst this work provides a basis for comprehending aggregate formation and
absorption, many factors are still unknown, such as the sustainability of these
aggregates as they move along the GI tract, and exactly how the aggregates
themselves are absorbed.43 Other work in this area includes that of Doak,44 who
found that 6 of 22 compounds which formed aggregates in buffer also had the
capability to form aggregates in simulated intestinal fluids (FeSSIF). The compounds
were tested at drug concentrations typical of these seen in vivo in the intestine. DLS
measurements demonstrate the ability of some compounds to form colloidal
aggregates in media mimicking in vivo conditions, again suggesting the possibility of
these compounds being absorbed in this form. It was hypothesized that the drug is
able to assume surfactant like properties under physiologically relevant conditions,
which in turn leads to the aggregation effects attributed to their high potency.45
Coan et al25 looked at another aspect of the in vivo environment, namely high

25
protein concentrations, and found that aggregates are not disrupted by milligram
per mL protein concentrations. This adds further evidence to the possibility of
aggregation in vivo, for at least some compounds. The possibility of in vivo
formation and subsequent absorption of aggregates is a very interesting hypothesis,
which will no doubt be investigated further in the literature in the coming years.

1.3.5 Utilisation of DMSO stock – kinetic solubility assay

Analyses which evaluate a compound’s solubility whilst starting from DMSO stock
solutions are generally known as ‘kinetic solubility’ assays. A compound’s kinetic
solubility can be said to represent the maximum solubility of the fastest
precipitating form, and is used in the early stages of discovery to give an indication
of potential solubility issues; its thermodynamic solubility is the maximum solubility
of the most stable form, and is regarded as the true solubility of the compound.
Kinetic solubility measurements are generally regarded as sufficient at the early
stages of the drug discovery process to give an idea of aqueous solubility and are
indicative of any potential solubility issues early on. They are also developed to
mimic the procedure employed during screening bioassays, with the solubility
results used to aid in interpretation of HTS results. 38 They are rapid, automated and
used to give an early indication of potential solubility issues. 15 Thermodynamic
measurements are typically performed during the development stages. These
solubility measurements do not use compound pre-dissolved in DMSO – the
solubility is determined from crystalline solid which has been left to equilibrate with
a given solvent over a period of time. These solubility assays are therefore initially
affected by the dissolution rate of the compound, until equilibrium has been
achieved – for kinetic measurements, the compound is already dissolved in DMSO,
so this effect is not important in the assay. Thermodynamic solubility assays also
use more compound compared to the kinetic assay, which at a discovery stage is
important as there is not a vast quantity of drug available. 8 The measured
thermodynamic solubility and kinetic solubility of a drug can be very different, with
the kinetic solubility being either equal or higher – for the majority of compounds

26
analysed by Hoelke et al,15 the kinetic solubility values are higher compared to
thermodynamic measurements. The difference can occur due to many factors – as
stated before, the kinetic solubility can be said to be the maximum solubility of the
fastest precipitating form, which may have a different solubility from the most
stable form. Bard et al 38 also found differences between kinetic and
thermodynamic solubility values, and attributed these to both the ability of certain
compounds to remain supersaturated when starting from DMSO stocks, and a co-
solvent effect of DMSO. The DMSO present in the kinetic solubility assay may help
solubilise the drug, and although the percentage is generally kept to a minimum,
there still may be some effects. Bard et al demonstrated that while co-solvent
effects due to DMSO were fairly general amongst a set of test compounds, the
ability to remain supersaturated was more compound specific. The work looked at
the effect of shaking intensity on the kinetic solubility value obtained, and
determined that less vigorous shaking resulted a in supersaturated solution with
less tendency to nucleate. This effect was more pronounced for low solubility
compounds. Only kinetic solubility measurements were performed, with no analysis
of the precipitate. Chen et al found that solubility enhancement by DMSO was
highly compound specific, and, for poorly soluble compounds, increasing the %
DMSO present can give significantly higher results. 46

There is also the issue of the time that the experiment has been performed over –
typically thermodynamic assays are performed over extended time periods to allow
the system to reach equilibrium. Kinetic assays tend to be shorter, although some
work has looked at prolonging the experimental time periods used. It was found
that extension of the incubation period resulted in solubility values closer to the
thermodynamic solubility value.38

Two widely used, general approaches of performing kinetic solubility are used;
precipitation, separation of the suspension and finally analysis of the
supernatant/filtrate content, or determination of the concentration at which
precipitation first occurs. Measurement of the dissolved content involves either

27
filtration or centrifugation to separate the liquid and the solid, after a period of
incubation. UV 35, 15 or HPLC 15, 47 analysis is then performed. Using the alternative
approach, turbidity 10 can be utilised to detect the presence of precipitate,
indicating the solubility limit of the compound – nephelometry 1,2, 15 is also used in a
similar way. Analyses are typically performed in HTS 96- or 384 well plates. For a
kinetic solubility assay determined by nephelometry, the stock solution is added
incrementally to an aqueous buffer until precipitation occurs. This can, however,
result in varying final % DMSO between different drug samples. Alternatively, an
excess of compound can be added such that precipitation occurs, and dilution
performed until the detected precipitate dissolves. This approach allows the
percent DMSO present to be kept constant.2 Kinetic solubility measurements using
nephelometry typically determine the drug concentration at which light scattering
become significantly different to background scattering – this is a similar concept to
the CAC described in the previous section. Nephelometric kinetic solubility methods
have been quoted as a having a lower limit of 20 μM - compounds with kinetic
solubility values below this value cannot be accurately assessed.15

Alelyunas et al8 have described the removal of DMSO prior to addition to the buffer,
to give a solid for use in solubility measurement while still utilising the liquid storage
of the drug. This approach gives results which are closer to the thermodynamic
solubility, whilst still retaining the advantages of having the compound dissolved in
DMSO, and a ‘solid state normalisation’ is seen. This suggests that the solid film
formed after removal of DMSO resulted in the same solid state of the material
being produced every time, and would remove any variation in the measurement
from the formation of different solid forms. The work also suggests that for most of
the compounds studied, equilibrium is reached during 24 hours of stirring – if this is
the case, the solubility value seen should be independent of the starting form of the
compound. This is of course assuming equilibrium has been reached between both
the solid and liquid phases (dissolution now constant) and between different solid
forms (conversion to most stable polymorph has been completed). The paper also
notes that different methods of stirring and mixing give varying solubility results for

28
a model compound, glyburide. Replicate analyses were performed using both the
dried DMSO method, and the compound predissolved in DMSO. Even the very high
supersaturated solutions which formed after one particular mixing method were
reproducible, demonstrating the effect that mixing has on these systems. There is
however no mention of whether the glyburide solid which was obtained after the
drying process was crystalline or amorphous – the group do however state that
most compounds form a ‘dry film’ after DMSO removal, which is indicative of an
amorphous material.

Typically, no characterisation of the solid produced at the end of kinetic solubility

experiments is performed in screening,37 although there has been a move towards
incorporating this into the analysis,48 and a few characterisation studies have been
reported in the literature. Hoelke et al analysed the precipitate formed after
addition of DMSO stock to aqueous buffer, and found that the precipitate was
amorphous in both cases. They suggest that the conditions used in HTS solubility
assays, i.e. rapid addition of the compound in DMSO to the buffer, do not promote
crystallisation, and that amorphous material is more likely to be formed.15 The
precipitation of amorphous material would be associated with higher solubility
values as amorphous materials have a higher solubility than the crystalline solid,
although there is not necessarily a large difference between them. Solubility can
also vary between polymorphs. Sugano et al also explore the nature of the
precipitate formed during kinetic solubility experiments using polarised light
microscopy (PLM). The group analysed kinetic solubility after a ten minute and 20
hour incubation, and compared these values to the thermodynamic solubility from
solid. Characterisation of the solid was performed for both 10 minute and 20 hour
incubation times. Many of the drugs precipitated as crystalline, even after only ten
minutes. They also found that those compounds which precipitated as amorphous
solids showed kinetic experimental solubility values more than five times higher
than the solubility from the solid. The final drug concentrations were, however,
fairly high in order to allow adequate amounts of solid for analysis. 3

29
Other methods have been reported, such as the use of Raman or powder x-ray
diffraction for analysis of the precipitate during HTS. Specialised well plates have
been developed allowing for direct analysis in the plates. 47,37 However, due to the
low concentrations used in HTS, the absolute amount of compound present in a
well plate must still pose some issue.

1.3.6 Previous unpublished work in this area

From unpublished work in the Halling lab based on three drug-like compounds,
many factors in the sample preparation process influence the amount of solid
formed upon mixing of the DMSO and buffer solutions. These variables are not
necessarily considered in current experimental set up of either bioassay or kinetic
solubility assays, nor is there a huge amount of detailed literature on the subject.
These include the way in which the DMSO stock and buffer solutions are added
together initially, and the way in which the resulting solutions are subsequently
mixed. 49,50 As mentioned, some published work has also demonstrated the effect of
shaking and stirring on kinetic solubility results,37, 38 but no detailed mixing
investigation has been performed. The use of sonication post dilution of the DMSO
into an aqueous buffer, as recommended in the literature,29 was also explored, but
was found to actually increase the amount of precipitate formed, contradicting the
general literature claims.49 The experiment in the literature29 investigated the effect
of sonication on DMSO stock solutions which were then spiked with 30 % water to
induce precipitation, at concentrations of 10 mM – although both DMSO and drug
concentration is increased here compared to that used in the work from the Halling
lab, the paper does recommend the use of sonication post dilution into aqueous
buffer. The results from the literature29 were that the majority of compounds which
had precipitated after dilution of the stock with water were driven back into
solution by sonication. The authors did note that 7 compounds showed further
precipitation on sonication, and suggested that precipitation was still occurring
when the solutions were sonicated, i.e. equilibrium had not been reached.
Sonication then increased the speed at which the process was occurring or gave the

30
system enough energy to overcome nucleation barriers. The paper also states that
sonication would only drive the compound back into solution to the extent allowed
for by thermodynamics – however due to the input of energy into the system, a
stable solid form may be converted to a metastable or amorphous form, which is
likely to have a higher solubility than that of the lowest energy solid form, meaning
the higher energy form of the compound can then re-dissolve. The samples which
did show the ability to be re-dissolved by sonication could have precipitated as e.g.
a crystalline solid, which was in equilibrium with the solution, and no further
precipitation would occur over time. Then, due to the input of sonic energy, this
was transformed to a higher energy form such as an amorphous solid or a higher
energy polymorph, which was then not at its equilibrium solubility and so could
dissolve into solution. This does suggest that equilibrium is reached quickly for most
compounds after dilution of the DSMO stock with water – this contradicts results
from the model compound investigated in the Halling lab (erthythrosin), which
demonstrated slow precipitation occurring over several days.49 Sonication would
then be expected to increase the amount of solid seen for erthythrosin, as
equilibrium behaviour was not demonstrated, in a similar manner to the seven
anomalous compounds in the literature. If sonication was seen to drive a particular
compound back into solution, the length of time that this effect could last for would
depend on the time required for conversion back to the most stable form.

No detailed investigation has been published on the nature of any solid material
produced upon mixing DMSO drug stock solutions with buffers. Determination of
whether the solid is amorphous or crystalline, and particle size and distribution
analyses, would allow further insight into the process which is occurring in these
systems. Some limited characterisation has been performed on the nature of the
solid formed in kinetic solubility assays (see section 1.3.5). From the previous work
carried out in the Halling lab, characterisation of the solid after scale up was
performed on two test compounds - one test compound formed an amorphous
solid, with another forming crystalline material. 50,51

31
Although precipitation is a recognised issue in this area, no detailed work on the
factors affecting it has been completed, with only guidelines to minimise its
occurrence being seen in the literature. The work from the Halling lab is limited in
the number of compounds it has covered, and a wider sample set is need. More
detailed investigation into the solid produced is also required in order to fully
understand the formation of solid at these low concentrations.

32
1.4 Aims

The aim of this work is to investigate in detail the factors which influence the
precipitation of drug molecules from DMSO and aqueous buffer systems at very low
concentrations. Several (typically uncontrolled) variables which may affect the
precipitation process are present within these experiments, and these will be
explored to ascertain any effect present. These variables include the exact way in
which the DMSO stock and aqueous buffer are added together (pipetting variables
such as speed, height, and angle of dispension of the liquid), whether the samples
are mixed further after initial addition of the organic and aqueous component, the
time the DMSO and buffer solutions are incubated for prior to testing, the amount
of protein present in the final sample mixture, and the % DMSO present in the final
sample. The precipitate formed will be investigated in detail, with regards to the
size and number of particles formed, and how these behave kinetically. As there is
already published literature on the effects of compound aggregate/particle
formation on bioassay results, the work here will investigate the precipitation from
a physical chemistry point of view. Performing the work in such a way will allow an
understanding of the process itself, as well complementing the already published
literature and aiding in interpretation of trends seen in bioassay screening results.

A selection of 3 poorly soluble test compounds will be explored in detail, with the
overall aim of rationalising the precipitation for that selection, rather than
performing mass screening of tens of compounds. The results will therefore give an
indication of the types of behaviour seen, as a starting point for potential further
investigation by the industry itself. Both solubility and precipitate analysis will be
performed. Various sample preparation methods and techniques for separation of
the liquid and solid phases and their effects on the results of kinetic solubility type
experiments will be explored. HPLC will be used in determining supernatant and
precipitate concentrations. Light scattering, Raman, microscopy and x-ray powder
diffraction are some techniques which could be utilised in analysis of the precipitate
formed. The form of the precipitate, as well as particle size and growth over time,

33
are important areas to investigate, and will be linked to both the sample
preparation method, and what behaviour is expected based on crystallisation
theory at higher concentrations, and the literature available. This work has practical
relevance to the pharmaceutical industry, as well as the novelty of exploring
precipitation and/or crystallisation of organic compounds at these low
concentrations.

The main difficulty in this work is the low concentration of drug (in the micromolar
concentration region) present in the samples – this makes any precipitation difficult
to see, and any material formed difficult to analyse.

As there is little published experimental data of the type performed here, where an
effect was seen, the results were written up as a series of research articles with a
view for publication. Since various different aspects were explored which were
suited to be published separately, the results are presented in this thesis as a
compilation of research papers (chapter 2 onwards). One article is already
published (chapter 2).52 The author of this thesis performed all of the experimental
work detailed, and undertook the preparation of the majority of the first draft of
every paper. Subsequent corrections and addition of comments to the draft papers
from the other contributing authors were also performed by the author of this
thesis.

34
1.5 Model Compound Selection
1.5.1 Compilation of initial compound list

Model compounds were required to meet four basic, practically relevant criteria;

1. The compounds must be poorly water soluble, as only low solubility

compounds are likely to precipitate at the low concentrations relevant to
screening.
2. The compounds should be commercially available, and known to be
biologically active. This latter criterion gives the results relevance to the
pharmaceutical industry.
3. Following on from 2, whilst it is desirable for the compound to be a known
pharmaceutical, it should not be too toxic. This criterion was included for
safety purposes.
4. The compounds should have a UV absorbance. This allows straightforward
analysis.

In discussion with MSD, a shortlist of eleven commercially available compounds was

compiled. All compounds met the criteria set out above. The compounds were
judged to have low solubility based on results of in-house kinetic solubility
experiments and those available online at the Physchem forum website.53 The
original eleven compounds were as follows;

35
Table I: 11 compounds on which preliminary experiments were performed

Compound Name MW (g/mol) Product No. Lot No. Storage Requirements

Amiodarone HCl 681.8 Sigma A-8423 87H0597 2-8°C

Clotrimazole 344.8 Sigma C-6019 118H1359 Room Temp

Thioridazine HCl 407.0 Sigma T-9025 87H0803 Room Temp

Tolnaftate 307.4 Sigma T-6638 43H0109 2-8°C

Imipramine HCl 316.9 Sigma I-7379 48H0362 Light Sensitive. Room Temp.

Hydrocortisone 21-Acetate 404.5 Sigma H-4126 78H0468 Light Sensitive. Room Temp.

Promethazine HCl 320.9 Sigma P-4651 128H1474 Light Sensitive. Room Temp.

Bumetanide 364.4 Sigma B3023 021K1134 No special storage requirements

Carbamazepine 236.3 Sigma C-4024 121K1510 Light Sensitive. 2-8°C

Metroprolol Tartrate 684.8 Sigma M-5391 101K1517 Light Sensitive. Room Temp.

Tolbutamide 270.3 Sigma T-0891 061K1629 Room Temp

36
All eleven compounds were subjected to several scoping precipitation experiments,
with the intention of having a first look at how the compounds behaved under
conditions relevant to screening. The results from these experiments were then
used to narrow the list to only those compounds which showed precipitation and
were stable under the conditions investigated. The eight excluded compounds and
the reasoning behind this exclusion are as follows –

 Four compounds (carbamazepine, tolbutamide, metroprolol tartrate and

hydrocortisone acetate) showed no precipitation under the conditions
trialled here. Variables investigated include initial drug concentration (up to
500 μM), pH and incubation time. None of these four compounds
demonstrated any tendency to precipitate during the course of the scoping
experiments. They were therefore disregarded.
 Two compounds (imipramine HCl and promethazine HCl) demonstrated
decreasing recoveries over short time periods, suggesting a tendency to
decompose rapidly in the buffer used. Due to this possible degradation, no
in depth experiments were performed with these compounds.
 Thioridazine HCl and bumetanide did not show any of the undesirable
characteristics mentioned above. The reason for their exclusion in further
experiments was merely due to the fact that more data had been collected
on some of the remaining compounds during the initial trials, and it made
more sense to continue on with those.

This streamlining process left three compounds (amiodarone HCl , clotrimazole and
tolnaftate) for detailed investigations.

37
1.5.2 Model compounds used in detailed experiments – general
information

Amiodarone HCl is the salt of a weak base, and has use as an antiarrythmic agent.
The pKa of the tertiary amine has been reported as 10.254 and 8.73 ± 0.05.55

Figure 2: Amiodarone HCl structure

Figure 3: Amiodarone HCl starting material, viewed under a microscope with a polarising filter

38
Clotrimazole is a weak base, with a pKa of 6.1.56 The drug has anti fungal
applications, and is typically used commercially as a topical treatment. 57

Figure 4: Clotrimazole structure

Figure 5: Clotrimazole starting material, viewed under a microscope with a polarising filter

39
Tolnaftate is a commercially available, neutral compound. It is a squalene epoxidase
inhibitor and also has applications as an anti-fungal agent, typically as an athlete’s
foot treatment.

Figure 6: Tolnaftate Structure

Figure 7: Tolnaftate starting material, viewed under a microscope with a polarising filter

All starting materials used in the studies described were purchased from Sigma
Aldrich, and used without further purification. NMR was used to confirm compound
purity, and all three compounds either matched or exceeded manufacturer’s
specifications (≥ 98%). NMR spectra are available in the appendix. XRPD spectra of
all starting material were also obtained, for use as reference spectra for any XRPD
analysis. These are also available in the appendix.

40
Table II: Detailed compound information

Compound Merck Index Solubility Description Literature Solubility Range Polymorphism seen previously?

3 μM, < 2 μM, pH 7.453

‘0.07 g/100 mL water’ (~1 mM)
Amiodarone HCl ~0.2 mM (<0.125 mg/mL) H2O, 20 °C58 None found documented in literature.

~1 mM (0.716 g/litre) H2O, 25 °C59

1.45 μM (0.5 μg/mL), ‘aqueous’56

‘practically insoluble in water'
Clotrimazole 5.5 μM, ‘aqueous’ 60 Yes – but only from melt. 61

15 μM,53 0 μM53 and 3 μM, pH 7.453

<3 and 3 μM, pH 7.453

Tolnaftate ‘practically insoluble in water’ None found documented in literature.

41
1.6 References
1. K. A. Dehring, H. L. Workman, K. D. Miller, A. Mandagere and S. K. Poole,
Journal of Pharmaceutical and Biomedical Analysis, 2004, 36, 447-456.
2. C. D. Bevan and R. S. Lloyd, Analytical Chemistry, 2000, 72, 1781-1787.
3. K. Sugano, T. Kato, K. Suzuki, K. Keiko, T. Sujaku and T. Mano, Journal of
Pharmaceutical Sciences, 2006, 95, 2115-2122.
4. L. Di and E. H. Kerns, Drug Discovery Today, 2006, 11, 446-451.
5. P. J. Desrosiers, Modern Drug Discovery, 2004, 7, 3.
6. M. Hoever and P. Zbinden, Drug Discovery Today, 2004, 9, 358-365.
7. D. A. Pereira and J. A. Williams, British Journal of Pharmacology, 2007, 152,
53-61.
8. Y. W. Alelyunas, R. Liu, L. Pelosi-Kilby and C. Shen, European Journal of
Pharmaceutical Sciences, 2009, 37, 172-182.
9. J. Mullin, Crystallisation, Butterworth-Heinemann, 2001.
10. C. A. Lipinski, F. Lombardo, B. W. Dominy and P. J. Feeney, Advanced Drug
Delivery Reviews, 1997, 23, 3-25.
11. I. G. Popa-Burke, O. Issakova, J. D. Arroway, P. Bernasconi, M. Chen, L.
Coudurier, S. Galasinski, A. P. Jadhav, W. P. Janzen, D. Lagasca, D. Liu, R. S.
Lewis, R. P. Mohney, N. Sepetov, D. A. Sparkman and C. N. Hodge, Analytical
Chemistry, 2004, 76, 7278-7287.
12. E. H. Kerns and L. Di, Drug-like Properties:Concepts, Structure Design and
Methods From ADME to Toxicity Optimization, Elsevier Inc, 2008.
13. H. D. Chirra and T. A. Desai, Advanced Drug Delivery Reviews, 2012, 64,
1569-1578.
14. S. H. Yalkowsky, Solubility and Solubilization in Aqueous Media, Oxford
University Press, Inc, New York, 1999.
15. B. Hoelke, S. Gieringer, M. Arlt and C. Saal, Analytical Chemistry, 2009, 81,
3165-3172.
16. D. Erdemir, A. Y. Lee and A. S. Myerson, Accounts of Chemical Research,
2009, 42, 621-629.
17. A. Van Hook, Crystallization, Reinhold Publishing Corporation, 1961.
18. S. L. McGovern, E. Caselli, N. Grigorieff and B. K. Shoichet, Journal of
Medicinal Chemistry, 2002, 45, 1712-1722.
19. J. Seidler, S. L. McGovern, T. N. Doman and B. K. Shoichet, Journal of
Medicinal Chemistry, 2003, 46, 4477-4486.
20. A. J. Ryan, N. M. Gray, P. N. Lowe and C.-w. Chung, Journal of Medicinal
Chemistry, 2003, 46, 3448-3451.
21. M. Habig, A. Blechschmidt, S. Dressler, B. Hess, V. Patel, A. Billich, C.
Ostermeier, D. Beer and M. Klumpp, Journal of Biomolecular Screening,
2009, 14, 679-689.
22. S. L. McGovern and B. K. Shoichet, Journal of Medicinal Chemistry, 2003, 46,
1478.
23. K. E. D. Coan, J. Ottl and M. Klumpp, Expert Opinion on Drug Discovery, 2011,
6, 405-417.

42
24. K. E. D. Coan, D. A. Maltby, A. L. Burlingame and B. K. Shoichet, Journal of
Medicinal Chemistry, 2009, 52, 2067-2075.
25. S. L. McGovern, B. T. Helfand, B. Feng and B. K. Shoichet, Journal of
Medicinal Chemistry, 2003, 46, 8.
26. K. E. D. Coan and B. K. Shoichet, Molecular BioSystems, 2007, 3, 208-213.
27. K. E. D. Coan and B. K. Shoichet, Journal of the American Chemical Society,
2008, 130, 9606-9612.
28. B. A. Kozikowski, T. M. Burt, D. A. Tirey, L. E. Williams, B. R. Kuzmak, D. T.
Stanton, K. L. Morand and S. L. Nelson, Journal of Biomolecular Screening,
2003, 8, 210-215.
29. B. A. Kozikowski, T. M. Burt, D. A. Tirey, L. E. Williams, B. R. Kuzmak, D. T.
Stanton, K. L. Morand and S. L. Nelson, Journal of Biomolecular Screening,
2003, 8, 205-209.
30. K. Oldenburg, D. Pooler, K. Scudder, C. Lipinski and M. Kelly, Combinatorial
Chemistry & High Throughput Screening, 2005, 8, 499-512.
31. K. V. Balakin, Y. A. Ivanenkov, A. V. Skorenko, Y. V. Nikolsky, N. P. Savchuk
and A. A. Ivashchenko, Journal of Biomolecular Screening, 2004, 9, 22-31.
32. P. E. Ilouga, D. Winkler, C. Kirchhoff, B. Schierholz and J. Wolcke, Journal of
Biomolecular Screening, 2007, 12, 21-32.
33. T. J. Waybright, J. R. Britt and T. G. McCloud, Overcoming Problems of
Compound Storage in DMSO: Solvent and Process Alternatives, Arlington,
VA, 2006.
34. Z. Blaxill, S. Holland-Crimmin and R. Lifely, Journal of Biomolecular Screening,
2009, 14, 547-556.
35. X. Cheng, J. Hochlowski, H. Tang, D. Hepp, C. Beckner, S. Kantor and R.
Schmitt, Journal of Biomolecular Screening, 2003, 8, 292-304.
36. C. Teng-Man, S. Hong and Z. Chegnyue, Combinatorial Chemistry and High
Throughout Screening, 2002, 5, 575-581.
37. S. Martel, m.-E. Castella, F. Bajot, G. Ottaviani, B. Bard, Y. Henchoz, B. G.
Valloton, M. Reist and P.-A. Carrupt, Chimia, 2005, 59, 308-314.
38. J. Alsenz and M. Kansy, Advanced Drug Delivery Reviews, 2007, 59, 546-567.
39. B. Bard, S. Martel and P.-A. Carrupt, European Journal of Pharmaceutical
Sciences, 2008, 33, 230-240.
40. C. Lipinski, Samples in DMSO: What an end user needs to know, Accessed
April, 2010.
41. C. A. Lipinski, Current Drug Discovery, 2001, 17-19.
42. M. Kreiner, B. D. Moore and M. C. Parker, Chemical Communications, 2001,
1096-1097.
43. S. C. Owen, A. K. Doak, P. Wassam, M. S. Shoichet and B. K. Shoichet, Acs
Chemical Biology, 2012, 7, 1429-1435.
44. Y. V. Frenkel, A. D. Clark, K. Das, Y.-H. Wang, P. J. Lewi, P. A. J. Janssen and E.
Arnold, Journal of Medicinal Chemistry, 2005, 48, 1974-1983.
45. A. K. Doak, H. Wille, S. B. Prusiner and B. K. Shoichet, Journal of Medicinal
Chemistry, 2010, 53, 4259-4265.

43
46. Y. V. Frenkel, E. Gallicchio, K. Das, R. M. Levy and E. Arnold, Journal of
Medicinal Chemistry, 2009, 52, 5896-5905.
47. T.-M. Chen, H. Shen and C. Zhu, Combinatorial Chemistry and High
Throughput Screening, 2002, 5, 575-581.
48. J. Alsenz, E. Meister and E. Haenel, 2007, 96, 1748-1762.
49. H. van de Waterbeemd and B. Testa, Methods and Principles in Medicinal
Chemistry, WILEY-VCH Verlag GmbH & Co, 2009.
50. R. Eadie, unpublished work.
51. R. Eadie, unpublished work.
52. B. Hall, unpublished work.
53. C. Gillespie, P. Halling and D. Edwards, Colloids and Surfaces A:
Physicochemical and Engineering Aspects, 2011, 384, 233-239.
54. physchem forum, http://physchem.org.uk/solubility_data.htm, Accessed 01
May 2013.
55. A. Avdeef, Absorption and drug development; Solubility, permeability and
charge state, John Wiley and Sons Inc., 2012.
56. A. Llinàs, R. C. Glen and J. M. Goodman, Journal of Chemical Information and
Modeling, 2008, 48, 1289-1303.
57. P. B. Tarsa, C. S. Towler, G. Woollam and J. Berghausen, European Journal of
Pharmaceutical Sciences, 2010, 41, 23-30.
58. P. Sawyer, R. Brogden, R. Pinder, T. Speight, G. Avery and n. Clotrimazole,
Drugs, 1975, 9, 424.
59. Y. Bouligand, F. Boury, J. M. Devoisselle, R. Fortune, J. C. Gautier, D. Girard,
H. Maillol and J. E. Proust, Langmuir, 1998, 14, 542-546.
60. M. Bonati, F. Gaspari, V. D'Aranno, E. Benfenati, P. Neyroz, F. Galletti and G.
Tognoni, Journal of Pharmaceutical Sciences, 1984, 73, 829-831.
61. E. Bilensoy, M. Abdur Rouf, I. Vural, M. Šen and A. Atilla Hincal, AAPS
PharmSciTech, 2006, 7, E54-E60.
62. L. Borka and S. Valdimarsdottir, Acta Pharamceutica Suecica, 1975, 12, 479-
484.

44
 2. Monitoring of particle growth at a low
concentration of a poorly water soluble drug using
the NanoSight LM20

Cheska Gillespiea Peter Hallinga Darren Edwardsb

a. WestCHEM, The Department of Pure and Applied Chemistry, University of

Strathclyde, Thomas Graham Building, 295 Cathedral Street, Glasgow, G1 1XL
b. Department of Chemistry, MSD, Newhouse, Lanarkshire ML1 5SH, UK

Corresponding author contact information; [email protected]

Tel; +44 (0)141 548 2482

(Published in; Colloids and Surfaces A: Physicochemical and Engineering Aspects,

384, 1-3, p. 233-239)

2.1 Abstract

The purpose of this work was to investigate the precipitation of a poorly water
soluble drug (tolnaftate) from low, µM concentration solutions. This was to test the
applicability of nanoparticle tracking analysis (NTA; the NanoSight instrument), with
comparison to results from dynamic light scattering (DLS).

Samples containing 30 µM of tolnaftate, 1% DMSO (by volume) were prepared by

mixing a concentrated DMSO stock solution and an aqueous buffer. The samples
were then analysed over time either using the NanoSight instrument or by DLS.
Obtaining meaningful results from the former required careful attention to
instrument settings.

From NTA there was initially a fairly narrow size distribution around 200 nm, with a
concentration of around 4x108 nanoparticles/mL. Over 3 hours, the particles grew,

45
with increasing polydispersity, and a skewed distribution up to 800 nm, while the
concentration fell to around 1x 108 particles/mL. DLS was consistent in showing the
size increase, but could not detect the remaining smaller particles and
polydispersity.

Conclusions: The growth of particles of a poorly water soluble drug was successfully
monitored using NTA, which gives additional information not offered by DLS.
Nanoparticle precipitation at the concentrations used here is of relevance to high
throughput screening in early drug discovery.

Keywords; HTS; solubility; nanoparticle growth; NTA; DLS; Tolnaftate

46
 2.2 Introduction

The growth of organic and inorganic crystals has been studied extensively using a
variety of techniques.1-7 This work has been motivated in relation to preparative
crystallisation, and hence has used high solute concentrations, typically 0.1 to 1 M.
In comparison, almost nothing is known about growth of crystals or other
precipitate particles from very low dissolved concentrations. Precipitation or
crystallisation of poorly soluble organics from µM concentrations is however of
practical importance, for example in studies of candidate pharmaceuticals. 8
Solutions in dimethylsulfoxide (DMSO) are added to aqueous media during testing
for biological activity 9,10 or solubility.11,12 In the latter case precipitation is expected,
while in the former it may occur.9 We now demonstrate how formation and growth
of precipitate particles in such low concentration systems can be monitored using a
novel optical technique that detects and characterises individual particles.

The technique of nanoparticle tracking analysis is a relatively new approach to

particle size and concentration measurement, implemented in the commercial
NanoSight LM20 instrument. Recent papers have looked at several different
applications of the instrument.13-16 The sample is introduced into a viewing unit, and
an image of the particles’ scattering of the laser light is captured by a CCD camera,
set on a microscope. A video of the sample is recorded, and then processed. Each
individual particle seen is ‘tracked’ by the nanoparticle tracking analysis (NTA)
software. From the Brownian motion, particle diameter can be calculated using the
Stokes-Einstein equation.17 The displacement co-efficient (Dt) from the mean
squared displacement of the particle tracked is calculated, and substituting this
value into the Stokes-Einstein equation allows calculation of particle size. Estimation
of the concentration of the particles is based on the assumed scattering volume
calculated from the dimensions of the field of view (at a given magnification) and
the depth of the laser beam. The average number of particles per millilitre of
sample is then simply extrapolated from the assumed scattering volume. The
instrument differs to other light scattering techniques such as dynamic light

47
scattering (DLS) in that an image of the particles’ scattering (although not the
particle itself) is recorded. The concentration of particles present and their size
distribution can be estimated based on the tracking of individual particles, unlike
DLS.18 In contrast, in DLS the signal is dominated by the larger, more intensely
scattering particles, and the result is an intensity weighted mean size. It is also very
difficult to obtain information on the concentration of particles present in the
sample from DLS. The two techniques are complementary to one another, and
both offer their own advantages and disadvantages, as discussed extensively by
Filipe et al.17 In the work done here, the NanoSight LM20 and NTA 2.0 software
were used to measure the growth of particles of a poorly soluble drug compound,
tolnaftate (Tol), over time. The sample concentration of around 30 µM is within the
range of that used in pharmaceutical screening, and the data here provides an
insight into how poorly soluble drug molecules behave at low concentrations upon
addition of DMSO stock to an aqueous system. The effects of instrument settings
were investigated in detail, and sensible choices suggested. The results from the
NanoSight were then compared to those obtained by DLS.

48
 2.3 Materials and Methods
2.3.1 Tolnaftate Sample Preparation

Tolnaftate (T6638) was purchased from Sigma Aldrich, and the purity was confirmed
by 1H-NMR (>99%).

A 0.05 M phosphate buffer solution was prepared, using 0.025 moles of KH2PO4 and
0.025 moles of Na2HPO4.2H2O per litre of deionised (Millipore, Direct Q,
ultrafiltered) water. The measured pH was 7.0 ±0.1. The buffer was then filtered
using 0.2 µm polyethersulfone (Sartorius-Stedim, Vivaspin PES 0.2 µm) centrifuge
filters (3 mins, 1037 x g RCF) to remove any particles present.

The DMSO used was SAFC Methyl sulfoxide, >99%.

A 2.99 mM stock solution of the drug was prepared in DMSO. A volume (10 µL) of
this stock solution was pipetted into a 1.5 mL glass HPLC vial. The filtered buffer
(990 µL) was then added to the HPLC vial containing the tolnaftate solution. This
gave a solution at a concentration of 29.9 µM tolnaftate, with a 1% DMSO
concentration. Each sample was inverted by hand approximately ten times before
introduction into the NanoSight sample chamber or the cuvette used in DLS.

2.3.2 NanoSight – Sample Procedure

After introduction into the NanoSight sample chamber, the sample was left to
equilibrate to the temperature of the chamber for 30 seconds (as recommended in
the NanoSight LM20 and NTA 2.0 software manual. Software version used – release
version build 0132). The first video (t=0) was captured after this equilibration time,
with subsequent videos captured at x minutes after t=0. The instrument used did
not have the capacity to control temperature, and so each time course was
performed at room temperature. At time 0, the average temperature recorded
from the three replicates was 21.8 ± 2.4 °C, and at t=180 minutes the average
recorded temperature had risen to 23.7 ± 2.0 °C. The outputs from the NTA

49
software were strongly dependent on the instrument settings used, and the next
sections explain in detail our observations here and the rationale of choices made.

2.3.3 Video Capture Settings

The camera settings for the instrument were set using the ‘Autosettings’ option on
the software – this allows the software to optimise the shutter and gain settings
and was thought to give less chance of bias than optimising manually for every
sample. The shutter setting applied is stated in the NanoSight LM20 and NTA 2.0
software operating manual as affecting the length of time the camera shutter is
open for each frame, therefore controlling the amount of light captured from the
particles. The gain setting controls the sensitivity of the camera.

The focus used was judged by eye, and was adjusted so that the majority of
particles seen were in focus at any one time. A test of repeated de-focusing and re-
focusing on a sample no longer changing significantly with time showed that results
were reproducible. This reproducibility was demonstrated by performing seven
measurements of the same sample (a suspension of thioridazine HCl). The results of
this were an average diameter of 251 ± 22 nm, and an average concentration of
0.7x108± 0.2x108 particles/mL. This experiment therefore validated the focusing
used in the time course experiments. The position in the X and Y planes was chosen
in accordance with the description of the optimum viewing position in the
NanoSight manual. First, the intense patch of light corresponding to that of the laser
was located. The field of view was moved slightly to the right, and the Z-plane
adjusted to focus until the light to the right of the laser patch converged into a line.
The position was then moved one field of view to the right, so that the laser line
was not in view.18 The X-Y viewing position was not changed during the time course.

The accuracy of particle size distribution data is expected to increase with the total
number of particles whose size is determined. Hence each video was recorded for
the maximum time allowed for by the instrument (215 seconds).

50
 2.3.4 Processing Parameters

The processing parameters of brightness and gain were not changed due to the
video contrast already being optimised by the programme.

A blur (smoothing setting) of 5x5 was employed, following the recommendation in

the operating manual,18 which states that if automatic threshold detection was
employed (which it was for all samples), the blur setting should be increased by one
level than that normally used. The threshold detection value determines the
minimum grey scale value of any ‘blob’ image necessary for it to qualify as a particle
to be tracked.

The minimum expected particle size was set at 100 nm for all samples.
Reprocessing the data with the lowest minimum particle size (30 nm) did not
significantly change the results. The minimum track length was set to automatic,
with the software calculating this based on the particles in the video. The minimum
track length setting defines the minimum number of steps a particle must take
before its size value can be accepted for inclusion in the particle size distribution
plot. Figure 1 shows a snapshot of the software processing the video, and tracking
the particles. The red and blue lines represent the particles’ trajectory under
Brownian motion, with the lines which have turned red representing those particles
which have been tracked for a long enough duration to have their size included in
the size distribution data. The smoothed particle size distribution curve can also be
seen – the distribution changes dynamically as more particles are tracked and
added to the plot.

Reprocessing the same video would not give consistent results due to a bug present
in the software (shutter and gain settings were taken from the current instrument
state, rather than those recorded when the video was acquired). Therefore, if the
video was not processed immediately after recording, the shutter and gain were set
manually to those used in the capture of the video.

51
Figure 1: Still of software ‘tracking’ particles

The data obtained from the NanoSight was adjusted for e.g. drift and camera
settings by the instrument. These calculations occur ‘behind the scenes’ in the
software and no file is available which details the corrections made. After
adjustment, the data are then smoothed by the instrument using a moving average,
to give a plot similar to that seen in figure 1. In order to obtain data which were
adjusted as appropriate by the software, but which was not smoothed, ‘Ctrl + 0’ was
pressed prior to processing each video – this adjusted, but not smoothed, data are
that which will be discussed further with relation to the NanoSight. The mean
obtained from the NanoSight instrument is a number weighted average. From the
distribution obtained from the NanoSight output, volume and intensity weighted
means could be calculated for each time point.

2.3.5 NanoSight – Preparation and analysis of a blank sample

A blank sample was prepared by pipetting 10 µL of DMSO into an HPLC vial. 990 µL
of filtered pH 7 phosphate buffer was added, and the sample inverted
approximately ten times by hand. The blank was then loaded into the sample
chamber, and left to equilibrate for 30 seconds. The camera autosettings option

52
was used to locate any small particles in the blank, and these were then brought
into focus. The camera shutter and gain settings were then changed, first to the
higher, and then to the lower, limits of these setting which were used in recording
the actual samples.

2.3.6 DLS Procedure – Sample

The filtered phosphate buffer was used to prepare a 1% DMSO/99% buffer blank.
Analysis of this solution by the DLS showed that some particles were present, as the
autocorrelation function was not completely flat. The buffer was then filtered again
using a 0.2 µm PTFE hydrophilic (Millipore) syringe filter. These double filtered
blanks also showed some particles in solution. The buffer which had been double
filtered was used to prepare the sample.

The DLS instrument used was a Malvern High Performance Particle Sizer, (HPPS 3.3),
with non-invasive back scatter (NIBS) technology. The scattering angle was 173 °.
The laser wavelength used was 632 nm. A 20 nm calibration standard was checked
using the DLS, with the size given as 19.5 nm, demonstrating that the instrument
was working accurately.

The DLS instrument was programmed to make one measurement every ten
minutes, with the first measurement being t=0, and to perform 5x20 sec
acquisitions on each sample, with the average result quoted for that time point. The
instrument automatically optimised the position of the laser and the attenuation
depending on the sample. The instrument was set to 25°C for all experiments. The
refractive index used for tolnaftate was 1.697, the value for the crystalline solid. 19

2.3.7 Preparation and analysis of a blank – DLS

As described in section 2.3.6, the buffer was filtered, then filtered again. A 1%
DMSO/buffer solution was prepared using the double filtered buffer, and analysed
using the Malvern instrument.

53
 2.4 Results and Discussion

2.4.1 NanoSight Results

To obtain meaningful results it was necessary to pay careful attention to a number

of instrument settings that would significantly affect the outputs, as explained in
detail in section 2.2 above. The blank samples showed no particles visible at the
camera settings used, eliminating this as a source of any particles seen in the
samples.

In total, three time course experiments at the same sample concentration were
performed using the NanoSight. The results from all three showed similar trends
over time - particle size increased, distribution changed from relatively
monodisperse to polydisperse, and particle concentration decreased. Whilst
different particle diameters and concentrations were obtained for the three
experiments, the overall trends were similar and for a nucleation dependent
process such as this, variability between samples is expected. The results from one
of the tolnaftate experiments are shown in table I and discussed in more detail. The
data for the remaining two time course experiments is present in the appendix.

54
Table I: Mean diameter and concentration with time obtained from NTA

8
No. of
Number Mean Standard Concentration (10
Time (mins) Completed
(nm) Deviation (nm) particles per mL)
Tracks

0 206 49 4.41 1516

10 240 54 1.92 642

20 251 56 1.57 444

30 289 71 1.99 505

60 291 69 1.72 425

90 299 66 1.33 300

120 326 90 0.76 166

150 373 125 1.27 257

180 378 138 1.52 293

The standard deviation in column three is a measure of the spread of particle sizes
in the particle size distribution, and is not an indication of error in the mean.

The mean and standard deviation shown have been calculated from the raw data,
rather than the smoothed NanoSight output. The concentration and number of
completed tracks have been taken from the NanoSight output files. Each completed
track corresponds to a particle whose size has been included in the distribution.
Stills from the first frame of the videos for t=0 and t=180 minutes are shown in
figures 2 and 3.

55
Figure 2: Still of video from Tolnaftate 29.9 µM at t=0 minutes (38 visible particles)

Figure 3: Still from video of Tolnaftate 29.9 µM at t=180 minutes (10 visible particles)

At t=0, there is a large number of particles visible (38) with the NanoSight, which is
what would be expected due to the low solubility of the drug. It is not, however,
obvious from the solubility that the particles present are nanoparticles within the
size range which can be analysed by the NanoSight. Seidler et al have demonstrated
that some small organic molecules with poor solubility can form aggregates of
nanometre size in aqueous solution, which inhibit enzymes promiscuously during
screening, and have measured the size of these aggregates using DLS. 20 Here, the

56
NanoSight is used to monitor both the size of the particles formed and how they
change over time, and the results compared to DLS. No enzyme inhibition studies
were performed. The image of the same sample at t=180 minutes shows that there
are fewer particles. The red crosses shown on the images represent the particles
that the software ‘sees’ and will track. The number at the bottom left corner of the
image is the total number of particles the software can ‘see’ in this particular frame.

Although for the three replicate experiments performed the actual particle sizes
varied, all three demonstrated the same trend of increasing size over time. The
standard deviation increases with each time point, demonstrating the increasing
spread of data as the system becomes more polydisperse, as shown in table 1. This
can also be seen from the overlay of distributions at different time points (figure 4).
Particle concentration ( 106 particles/mL)

80
70 t=180 t=150
60
50
t=120 t=90
40 t=60 t=30
30 t=20 t=10
20
t=0
10
0
0 100 200 300 400 500 600 700 800 900

Particle diameter (nm)

Figure 4: Overlay of NTA particle size distributions obtained from tolnaftate samples at various
time points

The distribution at t=0 has a sharp monodisperse profile. As time increases, the
sample profiles gradually become more polydisperse and ill-defined.

57
The mean particle diameter was plotted as a function of time, resulting in the graph
shown in Figure 5. Mean particle diameter shows a clear upward trend over the
three hour time period.

400

350
Mean Particle Diameter (nm)

300

250

200

150

100
0 50 100 150 200
Time (mins)

Figure 5: Mean particle diameter versus time obtained from the NanoSight instrument – error bars
represent standard deviations (n=3)

Figure 5 shows the average diameter across the three repetitions performed at
each time point, plus or minus the standard deviation in the mean.

It can be seen from figure 4 that as well as mean particle size increasing, there are
populations of larger particles emerging over time. It is also possible to see that,
after an initial fall, the concentration of the smallest particles also increases again at
longer times: this was clearly seen in the raw data. This is consistent with Ostwald
ripening, where smaller particles become smaller, and larger particles become
larger. Ostwald ripening suggests that particle size distribution should tend to
become more uniform over time as the particles tend to larger sizes.3 Whilst this is
not observed in these samples, this could be because the latest time point here is
not the end point of the process, and if the sample were left for long enough, the

58
distribution would re-align towards that of a monodisperse population. It could also
be the case that at the initial time point, there are particles present which are too
small to be detected by the NanoSight. These would re-dissolve into the bulk
solution, and then precipitate back onto the larger particles, which is more
favourable. This is, however, speculative. If the particles were truly monodisperse,
little ripening would occur. If growth were the only mechanism responsible for the
trend seen, there would be no emergence of smaller particles over time. Another
possibility is that more drug precipitates over time, and due to the reduction in
super saturation, initial particle size is reduced. Since no information on the
crystallinity of the particles is obtained from the NanoSight, theoretically the initial
particles could be amorphous – over time they could re-dissolve, and then undergo
conversion to a more stable crystalline form, and this is what is present in the small
particle populations at later time points.3 One other possibility may be that a
metastable polymorph precipitates, then undergoes a solution mediated phase
transformation to a less soluble form, which then grows.21 There is also a possible
contribution from the artefact of particles settling out of the measurement volume
or becoming stuck to surfaces over time. From the data obtained here it is
impossible to know which of these is occurring, or whether several are in fact
occurring at once.

The mass of drug present in the particles per volume unit can be estimated by
calculating the volume fraction occupied by the particles, then multiplying by the
density of the solid (took to be 1.223 g/cm3 for tolnaftate, that of the crystalline
material).19 It should be noted that for this calculation the particles are assumed to
be solid, non porous spheres. This mass concentration can then be converted into a
molar concentration using the molecular weight of tolnaftate (307.4 g/mol). The
amount of drug present in the particles as a percent of the initial added mass (29.9
nanomoles) can also be calculated.

Equation 1

59
 M
Equation 2

Table II: Calculated number of nanomoles of drug present at each time interval based on
NanoSight results from one measurement, calculated using equations 1 and 2

No. of
Volume mean nanomoles of
Concentration (108 % of drug
Time (mins) (adjusted not tolnaftate
particles per mL) added
smoothed) (nm) present in the
particles

0 218 4.41 9.51 32

10 253 1.92 6.47 22

20 264 1.57 6.01 20

30 307 1.99 11.99 40

60 308 1.72 10.46 35

90 314 1.33 8.57 29

120 353 0.76 6.96 23

150 415 1.27 18.90 63

180 428 1.52 24.81 83

Overall, an increase in the number of nanomoles of drug present in the particles is

seen, in particular for the last two time points, suggesting that further precipitation
is occurring as well as growth and ripening of the particles already present. The
random nature of the results prior to the last two time points reflect the errors
within the concentration measurement. Since an increase is seen, it is likely that
further precipitation is occurring, rather than just ripening of the particles already
present. By the end of the three hour time period, the results from the NanoSight
suggest that around 83% of the drug added is present in the form of nanoparticles.

60
Full details of the calculation used here is seen the appendix (chapter 2 additional
data).

A decrease in the number of particles per mL (from the NanoSight output) over time
is also seen (figure 6), which would be expected due to particle growth. The
concentration does not drop steeply however, again giving support to the theory
that several processes are occurring at once, as discussed above.

4.5
Concnetration (108 particles per mL)

3.5

2.5

1.5

0.5

0
0 50 100 150 200
Time (mins)

Figure 6: Particle Concentration Vs Time

2.4.2 DLS results

Some particles were still present in the blank sample even after the double
filtration, however the autocorrelation function (the standard output of a DLS) for
the blanks and the samples are completely different, eliminating the particles in the
blank as a cause of the data recorded for the samples. This is shown in figure 7.

61
 1
1% DMSO Blank

Correlation Co-efficient ( G2 -1)

0.8
Tol 30uM t=0
mins
0.6
Tol 30uM
T=95mins
0.4 Tol 30uM t=185
mins

0.2

0
0.1 10 1000 100000 10000000 1E+09 1E+11

-0.2 Time (µseconds)

Figure 7: Autocorrelation functions

The blue line at the bottom of the chart is the autocorrelation function relating to
the blank, with the other lines being the autocorrelation functions at different time
points for one 29.9 µM sample.

As with the NanoSight, three replicate analyses were performed, and each time, the
Z-averaged (intensity weighted) hydrodynamic diameter increased over the course
of 3 hours. A decrease in mean count rate was seen for each sample. Differences
were seen in the mean sizes between replicates, however this is thought to be due
to the sample itself rather than the technique used to measure the particles as
discussed previously. As with the NanoSight, the results from one of the samples
are shown in more detail in table III.

62
Table III: DLS Results, Z average diameter, count rate and PDI

Z average Polydispersity
Time (minutes) Count rate (kcps) Attenuator Position (mm)
index (PDI)
hydrodynamic
diameter (nm)

0 269 453.4 8 2 0.144

12 345 448.7 8 2 0.019

25 411 449.1 8 2 0.118

37 460 451 8 2 0.096

62 475 432.9 8 2 0.078

99 545 419.6 8 2 0.036

123 558 400.3 8 2 0.042

157 543 375.6 8 2 0.126

185 603 367 8 2 0.053

A plot of hydrodynamic diameter versus time shows an upward trend, similar to

that seen for the NanoSight samples (figure 8).

63
 650

600

Z average particle diamater (nm) 550

500

450

400

350

300

250

200

150
0 50 100 150 200
Time (mins)

Figure 8: Mean particle diameter versus time, DLS

Although the mean particle diameters obtained from DLS are intensity weighted
averages, and those obtained from the Nanosight are number weighted averages,
the trend seen is similar. Since the attenuator and measurement position was the
same for all time points, the count rate can be compared between measurements.
The count rate is a measure of the scattering intensity displayed by the sample.
Since larger particles scatter more intensely, it would be expected that the count
rate increase with time, however this is not observed. The main difference in the
data obtained from the NanoSight and that obtained from DLS is that the DLS
results do not show the change from monodisperse to polydisperse with time. The
polydispersity index (PDI) shows only random variation with time, and even at the
end time point is still very low. Filipe et al17 found that DLS cannot resolve
polydisperse samples, whilst the NanoSight can - this is in line with the results
shown in the present work.

64
In order to compare the average particles sizes obtained from the NanoSight to
those from the DLS, an intensity weighted mean was calculated from the particle
distributions obtained from the NanoSight for each time point. Since DLS is an
intensity weighted mean, there is a bias in the mean towards larger, more intensely
scattering particles, with low concentrations of larger particles being easily
detected, and affecting the measurement. Filipe et al analysed a mixture of 100,
400 and 1000 nm polystyrene bead using DLS – the results showed a monodisperse
sample of 700 nm, due to the masking of the smaller by the larger particles.17 The
NanoSight was however able to resolve the three particle size populations – similar
results are seen here. The average from the DLS is likely to be biased towards the
larger particles, both due to the weighting of intensity, and possibly the masking of
the smaller particles in the sample by the larger ones. The DLS mean would
therefore still be expected to be larger than that of the calculated NanoSight
intensity mean. In order to represent the particles measured by the DLS, the
diameter of the 95th percentile of each NanoSight time point was calculated and
plotted in figure 9, along with the results from the DLS, the NanoSight number
mean and the NanoSight calculated intensity mean.

65
 800

700

600
Mean Diameter (nm)

500

400

300

Mean From DLS

200
NanoSight Volume Mean

100 NanoSight Intensity Mean

NanoSight 95th Percentile
0
0 50 100 150 200
Time (mins)

Figure 9: Overlay of various mean diameters over time

Converting the NanoSight number mean into an intensity mean utilises all the
diameters seen from the NanoSight – this is not the case for the DLS, where the
larger particles dominate. The 95th percentile of each time point from the NanoSight
data is closer to the mean given by DLS, highlighting the inherent bias of DLS
towards more strongly scattering particles. The NanoSight instrument therefore
gives more information on the particle size distribution compared to that obtained
from the DLS.

66
 2.5 Conclusions

The NanoSight instrument and software were successfully used to analyse the
particle growth of a poorly water soluble drug at a low (µM) concentration. The
instrument offers an advantage over DLS for this particular sample in that the
emergence of a polydisperse distribution over time can be seen. The results from
DLS do not show this level of detail, with mean particle diameter only seen to
increase over time. The NanoSight is however more difficult to use than DLS due to
the numerous parameters for both video capture and processing which can be
altered, and a certain amount of expertise is required. This is also acknowledged in
the work of Filipe et al.17

The results from both light scattering techniques demonstrate that precipitation
and particle growth of lipophillic molecules at low concentrations can be slow,
occurring over several hours. This information is of practical relevance to drug
screening.

67
Acknowledgements

With thanks to;

Alan Kennedy and Alastair Florence, (University of Strathclyde), for their comments
on the paper

Ken Cameron (MSD Research, Newhouse, Lanarkshire) for performing the 1H-NMR
analysis

Jonathan Smith (NanoSight Technology Ltd) for technical advice on the instrument

MSD Ltd and the University of Strathclyde for funding contributions

68
 2.6 References

1. A. Van Hook, Crystallization, Reinhold Publishing Corporation, 1961.

2. B. Y. Shekunov and P. York, Journal of Crystal Growth, 2000, 211, 122-136.
3. J. Mullin, Crystallisation, Butterworth-Heinemann, 2001.
4. M. Kitamura, Journal of Crystal Growth, 2002, 237-239, 2205-2214.
5. E. Verdurand, C. Bebon, D. Colson, J. P. Klein, A. F. Blandin and J. M.
Bossoutrot, Journal of Crystal Growth, 2005, 275, e1363-e1367.
6. D. Erdemir, A. Y. Lee and A. S. Myerson, Accounts of Chemical Research,
2009, 42, 621-629.
7. K. Maeda, Y. Yamamoto, Y. Asakuma and K. Fukui, Chemical Engineering and
Processing: Process Intensification, 2009, 48, 902-906.
8. K. Sugano, T. Kato, K. Suzuki, K. Keiko, T. Sujaku and T. Mano, Journal of
Pharmaceutical Sciences, 2006, 95, 2115-2122.
9. L. Di and E. H. Kerns, Drug Discovery Today, 2006, 11, 446-451.
10. T. J. Waybright, J. R. Britt and T. G. McCloud, Overcoming Problems of
Compound Storage in DMSO: Solvent and Process Alternatives, Arlington,
VA, 2006.
11. B. Hoelke, S. Gieringer, M. Arlt and C. Saal, Analytical Chemistry, 2009, 81,
3165-3172.
12. J. Alsenz and M. Kansy, Advanced Drug Delivery Reviews, 2007, 59, 546-567.
13. A. W. Wark, R. J. Stokes, S. B. Darby, W. E. Smith and D. Graham, The Journal
of Physical Chemistry C, 2010, 114, 18115-18120.
14. H. Saveyn, B. De Baets, O. Thas, P. Hole, J. Smith and P. Van der Meeren,
Journal of Colloid and Interface Science, 2010, 352, 593-600.
15. C. D. Sorrell and L. A. Lyon, Langmuir, 2008, 24, 7216-7222.
16. D. E. Zhunuspayev, G. A. Mun, P. Hole and V. V. Khutoryanskiy, Langmuir,
2008, 24, 13742-13747.
17. V. Filipe, A. Hawe and W. Jiskoot, Pharmaceutical Research, 2010, 27, 796-
810.
18. NanoSight LM20 nd NTA2.0 Analytical Software Operating Manual, 2009.
19. http://www.lookchem.com/Tolnaftate/ (accessed 3rd Feb 2013)
20. J. Seidler, S. L. McGovern, T. N. Doman and B. K. Shoichet, Journal of
Medicinal Chemistry, 2003, 46, 4477-4486.
21. D. Murphy, F. Rodríguez-Cintrón, B. Langevin, R. C. Kelly and N. Rodríguez-
Hornedo, International Journal of Pharmaceutics, 2002, 246, 121-134.

69
 3. Detailed study of precipitation of a poorly water
soluble test compound using methodologies as in
activity and solubility screening - mixing and
automation effects

Cheska Gillespiea, Alan R. Kennedya, Darren Edwardsb, Lee Dowdenc, Pierre

Daublaind and Peter Hallinga

a- WestCHEM, The Department of Pure and Applied Chemistry, University of

Strathclyde, Thomas Graham Building, 295 Cathedral Street, Glasgow, G1 1XL
b- Strathclyde Institute of Pharmacy and Biomedical Sciences, University of
Strathclyde, 161 Cathedral Street, Glasgow G4 0RE,
c- Merck Sharp & Dohme, Hertford Road, Hoddesdon, Hertfordshire, EN11 9BU
d- Merck Research Laboratories, 33 Avenue Louis Pasteur, Boston, MA 02115

3.1 Introduction

Within the pharmaceutical industry, storage of discovery compounds dissolved in

dimethylsulfoxide (DMSO) is widespread. The DMSO stock solution is then used as
appropriate in several different analyses, some of which involve addition of a small
aliquot of DMSO stock solution to an aqueous media. However, most drug
molecules synthesised by medicinal chemists are largely water insoluble;1 this trend
and the reasons behind it have been discussed by Lipinski et al.2 It has been
acknowledged that precipitation of the compound could occur upon addition into
the aqueous buffer.3 This precipitation process and how it proceeds has
implications for two important areas of analysis; activity assays in a high throughput
screening (HTS) setting and kinetic solubility assays. Both of these analyses involve
the addition of the test compound dissolved in DMSO to an aqueous buffer system.
In an activity screening context, it is generally assumed that the compound remains
in solution for the entirety of the analysis. However, given the low aqueous

70
solubility of many test compounds, precipitation is a strong possibility. This could
potentially affect the in-vitro assay in two ways;

1. A false negative result could be obtained if a compound which has biological

activity has not remained in solution and is therefore not available to
interact with the enzyme/cell. This negative result obtained is caused by
lack of solubility rather than a lack of biological activity.4
2. A false positive result could be obtained due to non-stoichiometric,
aggregate based inhibition,5 resulting from interaction of the precipitated
particles with the enzyme. This means the compound has no specific
molecular interaction with the enzyme in question, and the positive result is
not due to stoichiometric biological activity. This phenomenon been
discussed in depth by Seidler et al, McGovern et al and Feng et al.5-9

At the low concentrations and volumes used in HTS, it is difficult if not impossible
to visually detect precipitation, particularly if small aggregates are formed. This
means that potentially useful drugs could be discarded. These solubility issues in
HTS have led to the introduction of what is known as ‘kinetic’ solubility assays. 3 As
with activity screening, these tests also utilise the addition of DMSO stock solutions
of the drug to an aqueous buffer, however here precipitation is an expected
phenomenon. The assays are used to determine a guideline aqueous solubility value
for the compounds under test. They give an indication of potential solubility issues
early on and can be used to aid bioassay result interpretation. 3, 10 Investigation into
the precipitation that occurs on mixing DMSO stock solutions and buffer, and the
factors affecting the process, is therefore of interest to these two important areas
of drug discovery analysis.

Several controllable factors could potentially affect the rate at which a compound
precipitates from solution. The technique used for initial addition of the DMSO may
be important, as well as the method for subsequent mixing. Within an activity assay
setting, no particular importance is given to how the compound stock solutions in
DMSO are added into the aqueous buffer, with many papers either stating that the

71
solutions were ‘mixed’ with no specific details, or not mentioning mixing at all.
Within a kinetic solubility setting, some limited investigation into the effect of
mixing has been carried out. Bard et al investigated the effect of shaking on
precipitation, and found that more vigorous shaking resulted in lower solubility
values (so long as the compound hadn’t precipitated completely and some
compound was still dissolved in solution at time 0). 11 Alelyunas et al state that
‘stirring is one of the most important experimental variables in determining
solubility’ and noted stirring is likely to affect the supersaturation obtained during
kinetic solubility assays.12 Bevan et al however found that shaking is not required
for reproducible results using their light scattering kinetic solubility determination
method.13

The work reported here was conducted using a poorly soluble test compound
(tolnaftate) to evaluate precipitation when DMSO stock solution and an aqueous
buffer are mixed together at low concentrations (within the range used in HTS).
Whilst several publications have looked at the kinetic solubility of many different
compounds, the experiments here focus on exploring in detail the effect of different
mixing variables on the precipitation and kinetic solubility of one model compound.
The results here are intended as an initial indication of whether mixing and
pipetting variables have an effect on the process at low concentrations and small
scales. The use of robotics allows control over how the DMSO stock solutions and
buffer are pipetted, with respect to the angle, speed and height used, and also
mimics the procedure used in drug screening, where robotic liquid handling is
common. Different mixing methods were investigated, and a comparison of the
reproducibility of results obtained from robot and hand pipetting was performed.

72
 3.2 Materials and Methods

The experiments were deliberately performed in 3 separate laboratories; Lab 1

(MSD Newhouse), Lab 2 (MSD Hoddesdon) and Lab 3 (University of Strathclyde).

Tolnaftate was purchased from Sigma-Aldrich, and its purity confirmed by NMR
(>99%).

Dimethyl sulfoxide (DMSO) used in the experiments was HPLC grade from Sigma
Aldrich (Lab 1), ACS reagent from Sigma Aldrich (Lab 2) or UV Vis grade from Fluka
(Lab 3).

Buffer Reagents – KH2PO4 was purchased from AnalaR (Lab 1) or Fisher Scientific
(Lab 2 and 3). Na2HPO4 (anhydrous) and Na2HPO4.2H2O were purchased from Fisher
Scientific (all labs).

3.2.1 Stock and Buffer Preparation

Stock solutions of nominal concentrations 1 mM and 5 mM were prepared (actual

stock solution concentrations ranged from 0.99 to 1.03 mM, and 4.97 to 5.08 mM).
These solutions will be referred to by their nominal concentrations. An aliquot of all
1 mM stock solutions prepared was diluted with DMSO to give a 0.5 mM stock
solution. Aliquots of the 5 mM stock were diluted to 2, 3, and 4 mM using DMSO.
This gave a total of 6 stock solutions, originating from two weighed stocks. A pH 7.0
(±0.1) phosphate buffer was prepared, using 0.0025 moles of KHPO4 and 0.0025
moles of Na2HPO4 per litre of deionised, Milli-Q water. The pH adjustment was
conducted with 2 M NaOH if required.

3.2.2 Instrumentation

A summary of the instrumentation is given in Table I. Full details of individual HPLC

methods and robot system settings are present in the appendix (chapter 3
additional data).

73
Table I: Instrumentation Summary

Lab 1 Lab 2 Lab 3

Beckmann Coutler N/A – samples hand

Robot system Tecan Genesis RSP 200
Biomek 2000 pipetted

Robot system Bioworks run version

Genesis v. 4.217.304 N/A
software 3.5(2003)

HPLC system Agilent 1100 Agilent 1100 Waters Alliance

Zorbax SB-C18 Rapid Phenomenex Luna Thermo Scientific

HPLC column Resolution HT 2.1x50 C18(2) 50 x 4.6 mm 3 Hypersil Gold 50 x 4.6
mm 1.8 µm column µm column mm, 3 μm column.

Drug retention time

2.44 2.18 1.99
(mins)

Eppendorf
Vortex Genie 2, vortex
MSI Minishaker IKA thermomixer comfort,
Vortex used setting 4 (around 1900
2500 rpm 1.5 mL holder,
rpm)
1400rpm

Stirrer used Heidolph MR3002 N/A N/A

Eppendorf 5804, with Eppendorf 5804 R, Eppendorf 5254 R with

Centrifuge used
T60-11 rotor with F45-30-11 rotor F45-24-11 rotor.

Average lab
20 23 20.5
temperature (°C)

74
 3.2.3 Sample Preparation using automated systems

Suspensions (test samples): an aliquot (10 µL) of 1-5 mM DMSO stock solution was
pipetted by the robot into a vial. The pH 7 buffer was then added by the robot in
two 495 µL aliquots (Tecan) or one 990 μL aliquot (Biomek). This gave 5 sample
suspensions within a concentration range of 10-50 µmol/L (10-50 nanomoles/mL).

HPLC standards: an aliquot (10 μL) of 0.5-5 mM DMSO stock solutions was pipetted
by robot into a vial. DMSO was then added in two 495 µL aliquots of DMSO (Tecan)
or one (990 μL) aliquot (Biomek). This gave 6 calibration solutions in DMSO in the
concentration range 5-50 µmol/L (5-50 nanomoles/mL). Fresh calibration solutions
were prepared for every HPLC experiment.

The difference in pipetted volumes between the two instruments was due to the
volume capabilities of the robot – the Tecan instrument had a maximum volume of
500 μL.

3.2.4 Sample preparation by manual pipetting

The same procedure as for robotic pipetting with the Biomek was followed, but
with the samples being manually pipetted using Gilson pipetteman pipettes.

3.2.5 Mixing

After preparation, the samples were mixed as follows;

Vortex Mixing (VOR) – The samples were vortexed for 30 seconds.

Robot Aspiration Mixing (ASP-R) and Manual Aspiration Mixing (ASP-M) - After
dispension of buffer, the robot aspirated and dispensed 500 µL of the suspension 5
times. The manual version of this mixing (ASP-M) was performed in a similar way
i.e. around half the suspension was aspirated and dispensed 5 times after the
addition of the initial buffer aliquot.

75
No Additional Mixing (NO) – After addition of the final aliquot of buffer, no further
mixing was performed.

3.2.6 Centrifugation and Dilution

Immediately after the mixing stage (section 3.2.5), the samples were centrifuged for
5 minutes at 7200 x g rcf. Slightly different procedures were then used in each
laboratory:

Lab 1 (initial samples prepared in a 1.5 mL, glass HPLC vial). 0.5 mL of the
supernatant was removed and diluted to 1 mL with DMSO. The solution was
vortexed for around 5 seconds at 2500 rpm to ensure thorough mixing. This is the
supernatant sample. 1 mL of DMSO was added to the remaining 0.5 mL of solution
plus any solid precipitate, giving a total volume of 1.5 mL. The mixture was vortexed
for around 5 seconds at 2500 rpm, and placed on a rotator for 15 minutes at 20 rpm
to allow full dissolution of any precipitated solid. This is the precipitate sample.

Lab 2 and 3 (initial samples prepared in a 1.5 mL, plastic microcentrifuge tube). The
entire supernatant was removed from the tube, and 0.5 mL of supernatant was
diluted to 1 mL with DMSO to give the supernatant sample. To the microcentrifuge
tube containing the precipitate, 1 mL of DMSO was added, and the solution
vortexed for around 5 seconds to ensure full dissolution of the precipitated
material. This is the precipitate sample.

The samples were then analysed by HPLC. Lab 1 precipitate samples were adjusted
for dilution factors, and by subtracting the amount of material contained in 0.5 mL
of supernatant.

Recoveries were calculated by summation of the amount of drug present in

supernatant and precipitate sample, and comparison with the amount of drug
present in the initial sample. Experimental results with calculated recoveries lower
than 85% or larger than 115% were rejected.

76
Mean results from experiments performed under various conditions in each lab
were compared using t-tests in Microsoft Excel. This test returns a value of the
probability that the two means come from two underlying populations with the
same mean. 95% confidence level was used in all analysis. Equal/unequal variance t-
test was chosen based on f-test result, which was also performed in Excel. Boxplots
were created using MiniTab (v16).

3.2.7 NTA and NanoSight LM10 Analysis

For NTA (software version 2.1 build version 0316) analysis, the buffer was filtered
(0.2 um filter), but otherwise the samples were prepared as described above. After
mixing, samples were either incubated for 5 minutes or centrifuged for 5 minutes,
and then introduced into the NanoSight LM 10 chamber. No dilution was
performed. Video capture settings were shutter, gain 26, 0 and 153, 0 for incubated
and centrifuged samples respectively. Processing parameters were kept as ‘auto’,
with blur increased to 5 x 5. The instrument temperature was set to 25 °C. The
resulting particle size distribution was that which had not been smoothed, obtained
by pressing ‘Ctrl + 0’ during analysis. The distributions were then binned (20 nm).

3.2.8 Variation present in separation and dilution steps

A scaled up 50 µmoles of drug per litre of mixture suspension was prepared, by

manually pipetting 200 µL of the 5 mM stock solution into a 30 mL glass vial. 19.8
mL of the pH7 buffer was added, and solution stirred using an ‘x’ shaped magnetic
stirrer bar. One experiment was performed with a stirring speed of 375 rpm, and
two with a stirring speed of 750 rpm. After stirring for 15 minutes, several 1 mL
aliquots were removed (ranging in number from 10 to 18, depending on the
experiment), and were placed into HPLC vials. The solution was stirred continuously
during sampling. The samples were then centrifuged and diluted as those from the
mixing experiment (section 3.2.5).

77
 3.2.9 Kinetics of precipitation

Using the Tecan robot with the settings described previously, 20 test samples (at a
concentration of 30 nanomoles of drug per mL of suspension) were prepared. The
samples were all rotated for 5 minutes at 20 rpm. Five samples were centrifuged
immediately; these are t=0. The remaining samples were incubated at 20 °C (± 2 °C),
and at t=48 h, t=120 h and t=168 h, 5 samples were removed, centrifuged and
diluted. Here, an increased centrifugation speed was used (10,000 x g rcf, 5
minutes).

3.2.10 Dissolution from solid

100 mL of a 1% DMSO/99% pH 7 buffer solution was added to approximately 1 mg

of solid tolnaftate, resulting in a sample concentration of ca. 30 μM. The resulting
suspension was stirred for 40 minutes at 500 rpm. 5 x 1 mL aliquots were then
removed and centrifuged for 5 minutes at 10,000 x g rcf. 0.5 mL of the supernatant
was removed and diluted twofold with DMSO. Only the supernatant samples were
analysed in this experiment. The remaining mixture was kept stirring at a constant
speed and the sampling process repeated. The sampling process was repeated
after 24, 48 and 162.5 hours of stirring.

78
 3.3 Results and Discussion

3.3.1 Dissolution from solid/kinetics of precipitation

Detailed results from the amount of drug in the supernatant after dissolution from
solid and after prolonged incubation in the precipitation kinetics experiments are
shown in the appendix. Whilst no specific value has been obtained from these
particular experiments, both results sets agree that the true equilibrium solubility of
tolnaftate is likely to be extremely low (around <2 μmoles of drug per L of 1%
DMSO/99% buffer mixture). The initial time point for the precipitation experiment
is higher than that starting from the solid. However, as the sampling time is
increased, both solubility values tend to low values. Shorter timeframes are likely
to be necessary to fully study the kinetics of precipitation, as it occurs rapidly in this
compound. It is certain, however, that at the total drug amounts investigated here
(30 and 50 μmoles of drug per L), the drug is way in excess of the solubility from the
solid value, samples have high supersaturation, and precipitation is expected. It
should however be noted that it is known that tolnaftate forms nanoparticles in the
concentration range used here14 and it is likely that centrifugation has not
completely removed all of these nanoparticles from the supernatant. The amount
of drug detected in the supernatant is therefore the sum of molecularly dissolved
tolnaftate plus any nanoparticulate material which has not been centrifuged out.

3.3.2 Variation present in separation and dilution steps

To establish how much variability resulted from the variation inherent in a

nucleation process, as well as variations in how the robot pipetted and mixed, it
was necessary to first assess the reproducibility of the unavoidable manual steps
involved in the experiments. This included both the removal of the supernatant
after centrifugation and the dilution of samples for HPLC analysis. Solid tolnaftate
could stick to the polypropylene pipette tip used whilst sampling due to both the
compound and the tip being hydrophobic. This was in fact observed during

79
dissolution from solid experiments, and would affect the recoveries seen. For the
reproducibility experiment, all replicate supernatant values from the same solution
were used, even if total recoveries were low, as the adsorption of any solid drug to
the pipette tip would not affect the amount seen in solution, only the total
recovery. All samples came from the same homogeneous solution, and so the
amount present in solution should therefore be the same for each sample. The
results from these experiments show that, for a given sample where several
aliquots are removed, the results are reproducible (Table II).

Table II: Reproducibility of manual steps

Amount of tolnaftate in supernatant

(μmoles per L)
Treatment

Mean S.D.

Stir 375 rpm 10.8 1.4

Stir 750 rpm R1 17.3 2.5

Stir 750 rpm R2 5.2 2.5

When different treatments were used, it can be seen that the standard deviation is
of the order of a few nanomoles per mL.

3.3.3 Decrease in supernatant amount measured as more

experiments performed

Table III shows the results of some experiments obtained from labs 1, 2 and 3.
When performing the very first set of experiments, it was noticed that the standard
deviation was higher than that obtained for the results in section 3.3.2. As several
subsequent replicates were performed, the standard deviation further increased.
Upon analysis of the full dataset, it was found that experiments performed early in
the experimental time period had significantly higher amounts of tolnaftate present

80
in the supernatant with larger standard deviations than those performed later on
(regardless of the mixing used).

In order to investigate this further, and to confirm this to be a real effect for this
drug, the experiments were repeated in Lab 2 and Lab 3. Lab 2 had no recent
history of usage of tolnaftate, and the stock solutions were prepared in Lab 3, so
that no nuclei in the form of solid crystalline tolnaftate were introduced into the
local environment by opening the bottle to weigh out the compound. Again, a
similar trend was observed. Lab 3 also demonstrated this trend as for a group of
mixing experiments performed over a non-continuous 5 month period, the amount
of tolnaftate present in the supernatant in later experiments was lower. The
division of results into ‘initial’ and ‘further’ experiments was arbitrarily performed
by inspection of the data.

81
Table III: Comparison of Labs 1, 2 and 3, initial and further experiments amount of tolnaftate
present in the supernatant

Amount of tolnaftate in the supernantat ± S.D.

Treatment Details
(μmoles per L)
p value for t-test
comparing early and
Lab and initial
a b later experiments
amount Mixing Initial experiments Further experiments
added

NO 15.7 ± 5.6 6.1 ± 1.1 0.008

Lab 1, 30
μmoles per L
ASP-R 12.7 ± 5.4 6.5 ± 1.7 0.02

NO 19.2 ± 5.6 9.6 ± 1.2 0.008

Lab 1, 50
μmoles per L
ASP-R 14.6 ± 6.6 7.3 ± 1.2 0.07 (N.S)

NO 10.2 ± 1.1 9.2 ± 0.8 0.2 (N.S)

Lab 2, 30
μmoles per L
ASP-R 5.7 (only 2 reps) 7.7 ± 1.9 0.2 (N.S)

NO 17.5 ± 1.2 7.7 ± 1.4 0.00002

Lab 2, 50
μmoles per L
ASP-R 13.4 ± 5.1 5.7 ± 2.9 0.01

NO 27.0 ± 1.1 10.7 ± 4.4 0.0005

Lab 3, 50
VOR 8.2 ± 1.5 4.5 ± 1.4 0.008
μmoles per L*

ASP-M 15.0 ± 1.7 8.8 ± 2.9 0.01

a b
– Lab 1 - Days 1-5 –Lab 1 - Day 6 onwards

Lab 2 - Day 1 Lab 2 - Day 2 onwards

Lab 3 – Day 1-2 Lab 3 – around 5 months later

*All experiments performed at 25 ± 1 °C, and centrifuged at 9,300 x g rcf. These are the only experiments to be
performed under these conditions and were not used in comparisons to Labs 1 and 2.

NS – not significant, p value >0.05

8 out of 11 t-tests show significant differences between mean supernatant values

obtained at the start of the experimental time period and that obtained later in the

82
experimental time frame. The later values appeared to be fairly consistent, with less
variation (judged by standard deviation values) than for earlier repetitions. The later
values are also more in line with the standard deviation values seen in section 3.3.2.

Several aspects of the experimental procedure were examined for potential causes
to this unusual result. Water uptake from DMSO stocks was investigated. All
calibration solutions were prepared from the same stock solutions as the samples,
thus any change in stock concentration due to water uptake would be noticeable in
the peak areas of the calibration samples during HPLC analysis. This was not
observed. The calibration lines themselves were also examined, and no changes in
the slope and intercept were observed which would explain the effect. The trend
could not be correlated to the age of the buffer used, or whether a freshly prepared
stock solution was used – once the decrease had occurred, the supernatant
amounts remained low despite preparation of fresh buffer/stock solutions. The
experiments were performed at a constant temperature throughout. Potential
explanations include the generation of homogeneous nuclei of the drug as more
precipitation experiments were performed, resulting in increased levels of air borne
nuclei of the drug. Something within the particular lab environment could also have
changed, giving rise to heterogeneous nuclei (e.g. increased numbers of dust
particles, which could act as nuclei, although this would have to have occurred at all
three sites). Both types of nucleation would result in increased precipitation levels.
These explanations are, of course, speculative.

For the 50 μmol/L samples with no mixing, a boxplot gives a pictorial representation
of this trend over time for all three laboratories (Figure 1)

83
 Nanomoles of drug present in 1 mL supernatant
Labs 1,2 and 3; Initial and further repetition data
30

25

20

15

10

5
s s s ps ps ps
rep rep r ep re re re
l l l
iti
a ia iti
a er er er
in tit in r th r th rth
1 in 3 fu fu fu
b 2 b 1 2 3
La b La b b b
La La La La

Figure 1: Tolnaftate 50 μmoles per L (50 nanomoles per mL) samples with NO mixing – comparison
of initial and further reps for Labs 1, 2 and 3

In further comparison of the data only the results from the later experiments will be
considered as they have been judged to have reached a steady, low value which
remains unchanged as further experiments are performed. Although these later
data sets tend to have less variability, they also have lower amounts of drug present
in the supernatant. The difference between the earlier and later experiments can be
up to 15 nanomoles in a 1 mL total volume – in the context of the low
concentrations and volumes used in screening, this is a large portion of the total
drug present. From the results seen here, further repetitions of a bioassay/kinetic
solubility assay of this particular compound would give different results to those
seen from an initial analysis.

84
 3.3.4 Results and comparison of data between Labs

Presented in Table IV is the mean drug amount present in the supernatant (plus or
minus the standard deviations) for all three labs (only further experiments shown).
The measurement performed is essentially that of the kinetic solubility of tolnaftate
under various conditions. This data will be used in comparisons between the labs,
and also between different mixing procedures (section 3.3.6).

Table IV: Mean supernatant amounts (μmoles/L) with standard deviations

Lab 1 Lab 2 Lab 3

ASP-
NO ASP-R NO VOR ASP-R NO VOR
M

6.1 ± 6.5 ± 8.1 ± 5.9 ± 7.6 ±

30 μmol/L
1.1 1.7 1.3 2.4 0.9

9.6 ± 7.3 ± 7.2 ± 4.9 ± 7.5 ± 9.3 ± 5.9 ± 7.9 ±

50 μmol/L
1.2 1.1 1.5 1.8 3.0 0.9 1.0 1.4

Lab 1- 30 nmol/mL, NO and ASP-R, n =14 and 10 respectively. 50 nmol/mL, NO and ASP-R, n=5 and 16
respectively. Lab 2 – n = 24 for all treatments. Lab 3 – n=9 for all treatments

Performing t-tests on the means obtained across different laboratories gives the p-
values reported in table V.

85
Table V: Comparison of results between different labs; P values obtained after t-test

VOR
NO mixing ASP mixing
mixing

Lab 1 vs. Lab 1 Lab 2 Lab 2 Lab 1 Lab 1 Lab 2

Lab 2 vs. vs. Lab vs. Lab vs.Lab 2 vs. Lab vs. Lab
Lab 3 3 3 3* 3*

30 0.08
0.00006 - - - - -
μmol/L (NS)

50 0.6 0.6
0.002 0.002 0.1 (NS) 0.8 (NS) 0.3 (NS)
μmol/L (NS) (NS)

*ASP-R vs. ASP-M

NS – not significant at a 95% confidence interval

From table V, it can be seen that comparable supernatant amounts were observed
for all three labs when VOR or ASP mixing were used. NO mixing, however, results
in significant differences in 3 out of 4 comparisons (bold and italicised). This could
be attributed to external laboratory factors (air borne nuclei) and/or the exact
instrumentation used (e.g. initial dispension speed). It appears that the extent of
precipitation depends upon the lab environment or the exact pipetting parameters
used when no mixing is applied while any further mixing overrides this effect. This
means that mixing of the samples should result in less lab to lab variation, provided
they are mixed in a similar way – this result is somewhat intuitive.

Finally the results also show that manual aspiration mixing (ASP-M) is not
significantly different from automated aspiration mixing (ASP-R). Control of the
exact speed, height and angle that the buffer is dispensed/aspirated at does not
significantly affect supernatant levels and these are similar to those for manual
dispensing/aspiration.

86
 3.3.5 Comparison of variability seen in robotically and manually
pipetted samples

From the standard deviations present in the results (Table IV), there does not
appear to be a significant advantage in using a robotic system to reduce variability.
Obviously, this is not the reason why robotics are used for HTS, but it is an
interesting point to note, and suggests that the variability present does in fact
reflect the natural variation present in a precipitation process, rather than being
caused by a controllable pipetting variable.

3.3.6 Comparison of mixing variables

Using the data detailed in Table IV, mixing variables were compared to one another
and p values were obtained (Table VI).

Table VI: P values for mixing method comparison

Lab 1 Lab 2 Lab 3

NO vs NO vs NO vs VOR vs NO vs VOR vs
NO vs VOR
ASP-R VOR ASP-R ASP-R ASP-M ASP-M

30
- - -
μmol/L 0.5 (NS) 0.001 0.2 (NS) 0.007

50
μmol/L 0.0008 0.00004 0.7 (NS) 0.002 0.000007 0.02 0.002

NS – not significant at a 95% confidence interval

There is no particular correlation between the amount of drug present in the

supernatant and whether automated pipetting or manual pipetting is used (Table IV
and

87
VI): results for Lab 1 NO mix (robotic pipetting) and Lab 3 NO mix (manual
pipetting) are similar. Control of the exact pipetting parameters used therefore
does not significantly affect the results seen, and provided further mixing is
performed, lab to lab variation is minimal.

From the data, VOR mixing consistently leads to amounts of tolnaftate present in
the supernatant (bold and italicised) which are significantly lower from that for
other mixing types, for all three labs. Data at 25 °C from Lab 3 (shown in appendix)
also demonstrates this trend. Overall it can be concluded that NO and ASP-R/M
mixings tend to be fairly similar with significant differences not seen consistently
(i.e. 3/5 comparisons not significant, 2/5 comparisons significant). However VOR
mixing always results in a lower amount of tolnaftate present in the supernatant
than for other mixing types investigated. As mentioned previously, it is known that
these mixtures form nanoparticles of the drug.14 The nanoparticles are present
almost immediately upon addition of the buffer to the DMSO stock i.e. the
induction time for these particles to form is extremely short. After addition of the
buffer aliquot (be it by robotic means or manually), it is expected that these
nanoparticles are already present in fairly high abundance. Subsequent mixing
would therefore affect any further growth and ripening of these particles, rather
than initial nucleation. Vortex mixing is the most vigorous mixing method used, and
appears to be aiding in particle growth and/or ripening, facilitating the formation of
larger particles compared to those formed with other mixing types. A more vigorous
mixing method would also result in increased likelihood of collisions between
particles, and aggregation of the particles could potentially occur. All of these
scenarios would result in increased numbers of particles being removed during
centrifugation, and so leaving less in the supernatant.

3.3.7 NTA Particle size analyses

In order to test the hypothesis offered in section 3.3.6, 50 μmoles per L VOR and NO
mix samples were analysed using NTA. Samples were analysed both after 5 minute

88
centrifugation (analysis of the supernatant) and with a 5 minute incubation period
(analysis of the suspension without prior separation of solid and liquid phases).

VOR mix - after

centrifugation
Average particle concentration (E6 particles/mL)

No Mix - after
centrifugation
200 No Mix - no
centrifugation
150 VOR Mix - no
centrifugation

100

50

0
0 100 200 300 400 500 600 700 800
900
Particle diameter (nm)

Figure 2: NTA Data; samples incubated for 5 minutes or centrifuged for 5 minutes at 9,300 x g rcf

Prior to centrifugation, both NO and VOR mixed samples give very similar mean
particle diameters and particle concentrations. The particle size distributions are
also fairly similar, although it can be seen that the VOR mixed samples have slightly
higher concentrations of particles >300 nm. This is consistent with the view that
more vigorous mixing affects the particle size distribution, although the effect is
minimal. After centrifugation, however, a large difference is seen, suggesting
centrifugation itself consolidates the effect seen from mixing. NO mix samples have
a higher particle concentration remaining in the supernatant compared to VOR
mixed samples, and also have a greater concentration of larger particles (>200 nm).
Whilst the NTA concentration output is an estimate, based on 5 replicates, the
nanoparticle concentrations for NO mixed and VOR mixed after centrifugation are
significantly different (p = 0.02). This is consistent with the HPLC data, where NO

89
mixing results in a higher supernatant content – the NTA data suggests that this is
not due to an increase in molecularly dissolved content, but an increase in
nanoparticulate content which is not removed by centrifugation compared to those
mixed by vortex.

90
 3.4 Conclusions

Several conclusions can be drawn from the work presented. Experiments performed
early in the experimental time period have a significantly different (higher) mean
supernatant amount (with increased standard deviations) compared to those
performed after a given time. This decrease has been observed at three different
laboratories, confirming it as a real effect for this particular compound. Whilst it
cannot be completely rationalised, it is of practical relevance to screening – further
repetitions of this compound’s kinetic solubility/bioassay activity would result in
significantly different results to those seen from an initial screen.

No significant differences are seen between manual and robotic pipetting, in terms
of either the amount of drug present in the supernatant or the variability seen from
the samples. These data are useful in trying to understand the factors which affect
the precipitation of poorly soluble drugs from these types of samples, and gives
more insight into the process – the use of robotics does not decrease the standard
deviations seen.

It can be concluded that whilst the way in which the initial addition of the organic
and aqueous components occurs is not particularly important, subsequent mixing is.
More vigorous mixing consistently gives lower supernatant concentrations than the
other two mixing types investigated here, and affects the concentration of
precipitate nanoparticles remaining after centrifugation. Again, this has implications
for screening and kinetic solubility assays, where sometimes no particular
importance is given to how the samples/plates are mixed.

Performing similar experiments with other compounds would aid in further

understanding of the effect of mixing and pipetting variables on the precipitation
process as a whole, but initial results here suggest mixing is a potential source of
variation in the bioassay/solubility screening procedure which is not currently
controlled.

91
Acknowledgements:

Alastair Florence (Univ. Strathclyde) and Timothy Rhodes (Merck) for their useful
discussions on the experiments.

Ken Cameron (MSD Research, Newhouse, Lanarkshire) for performing the 1H-NMR
analysis

University of Strathclyde and MSD for funding contributions

92
 3.5 References

1. M. S. Ku, The AAPS Journal, 2008, 10, 208-212.

2. C. A. Lipinski, F. Lombardo, B. W. Dominy and P. J. Feeney, Advanced Drug
Delivery Reviews, 1997, 23, 3-25.
3. L. Di and E. H. Kerns, Drug Discovery Today, 2006, 11, 446-451.
4. E. H. Kerns and L. Di, Drug-like Properties:Concepts, Structure Design and
Methods From ADME to Toxicity Optimization, Elsevier Inc, 2008.
5. S. L. McGovern, E. Caselli, N. Grigorieff and B. K. Shoichet, Journal of
Medicinal Chemistry, 2002, 45, 1712-1722.
6. S. L. McGovern, B. K. Shoichet and n. Kinase Inhibitors, J. Med. Chem., 2003,
46, 1478.
7. S. L. B. T. McGovern, B. T. Helfand, B. Feng and B. K. Shoichet, J Med Chem,
2003, 46, 8.
8. J. Seidler, S. L. McGovern, T. N. Doman and B. K. Shoichet, Journal of
Medicinal Chemistry, 2003, 46, 4477-4486.
9. B. Y. Feng, A. Shelat, T. N. Doman, R. K. Guy and B. K. Shoichet, Nat Chem
Biol, 2005, 1, 146-148.
10. L. Lindfors, P. Skantze, U. Skantze, J. Westergren and U. Olsson, Langmuir,
2007, 23, 9866-9874.
11. B. Bard, S. Martel and P.-A. Carrupt, European Journal of Pharmaceutical
Sciences, 2008, 33, 230-240.
12. Y. W. Alelyunas, R. Liu, L. Pelosi-Kilby and C. Shen, European Journal of
Pharmaceutical Sciences, 2009, 37, 172-182.
13. C. D. Bevan and R. S. Lloyd, Analytical Chemistry, 2000, 72, 1781-1787.
14. C. Gillespie, P. Halling and D. Edwards, Colloids and Surfaces A:
Physicochemical and Engineering Aspects, 2011, 384, 233-239.

93
 4. Investigating the effects of nanoparticle formation
on the kinetic solubility determination of 3
hydrophobic test compounds

4.1 Introduction

One of the most important physical properties of a candidate drug within the
pharmaceutical industry is aqueous solubility. For a given compound, it is
investigated (to varying degrees) at several stages in the drug discovery and
development process.1 Whilst good aqueous solubility is a very desirable property,
generally compounds synthesised by the industry at the discovery stage are poorly
water soluble and hydrophobic. They are optimised structurally with a view to
maximising binding to the target in question, and this is usually at the expense of
aqueous solubility. Compounds tend to be selected for further investigation based
on their affinity for their particular target - physical property investigation at this
stage is minimal.2 Moving on to a development setting, one of the big challenges
facing scientists is to come up with the optimum salt/crystalline form/formulation
which will give the final marketed product the desired physical properties. Detailed
investigation of the physical properties of the compound and its formulations are
therefore required.3

In a discovery setting, higher throughput solubility methods are used to give an

initial indication of the solubility of the compounds under test. These analyses
typically utilise DMSO stock solutions of the drug, short incubation times, and are
said to be representative of the solubility of the fastest precipitating form. 4 It is not
expected that types of samples reach an equilibrium solubility value. Two main
methods of analyses are employed;

1. Using nephelometric3,5 or turbidity based4 methods to determine the

concentration at which precipitation occurs. Here, DMSO stock is added
incrementally to an aqueous phase until the solubility limit is reached.

94
 2. Addition of the DMSO stock and aqueous phase, followed by separation of
any solid formed (via centrifugation or filtration) and analysis of the filtrate
by UV or HPLC.3, 6, 7

Analyses which utilise DMSO stock solutions of the compound are known generally
as ‘kinetic solubility’ methods. These assays have several advantages; they utilise
the drug stock dissolved in DMSO, making automation possible and so can be
performed in 96, 384 etc well plates. Minimal compound is also used, which is
essential at the discovery stage.8 The samples do, however contain DMSO
(sometimes varying amounts) in the final solvent composition, and it cannot be said
for certain if the system has reached equilibrium. These analyses are, however,
successful in giving an estimate of aqueous solubility of the compound, which is
deemed sufficiently informative for the discovery stage.

In a development setting, solubility is determined thermodynamically, using a

classical ‘shake flask’ approach. The compound is introduced in a solid form (in
excess) into the solvent, and agitated for extended time periods, to ensure full
dissolution to equilibrium.1 Analysis is then performed after separation of the
excess solid, typically by HPLC. This approach, whilst considered to give the ‘true’
solubility of a compound under a given set of conditions, is time consuming and
requires larger amounts of compound compared to its kinetic counterpart.8

Typically the results from ‘kinetic’ solubility assays tend to be higher than those
obtained from their thermodynamic counterparts. This is attributed to several
different factors, such as the presence of small quantities of DMSO (which could
increase solubility compared to aqueous systems alone – a so-called ‘co-solvent’
effect), the shorter incubation times employed (the system may remain
supersaturated in the timescale employed), and the fact that the precipitating form
may not be the thermodynamic crystalline form (amorphous/higher energy
polymorphs can have increased solubility compared to the lowest energy
polymorph).1 All of these scenarios would lead to a higher solubility value than that
obtained from a solid starting material in a ‘shake-flask ‘experiment. Approaches

95
have been made to try and minimise discrepancies between kinetic and
thermodynamic solubility values, such as the use of longer incubation times 5 and
the removal of DMSO to give a solid prior to solubility measurements, 9 whilst still
retaining some of the advantages of ‘kinetic’ based analysis.

In the work described here, a detailed investigation into the thermodynamic and
‘kinetic’ solubility of 3 hydrophobic test compounds (amiodarone HCl ‘ami’,
clotrimazole ‘clot’ and tolnaftate ‘tol’) was carried out. The behaviour was
determined from the solid form using a ‘shake-flask’ type method and from DMSO
stock solutions after precipitation using 1) light scattering (NTA and NanoSight)
detection of particles and 2) HPLC analysis (after either centrifugation or filtration)
with various incubation periods. As stated, typically analyses which utilise the drug
dissolved in DMSO stock solution are referred to as ‘kinetic methods’. Here, we will
use the terminology ‘precipitation’ methods, to emphasise the physical process
occurring. Since the solubility from ‘shake-flask’ (starting from solid) experiments
used in comparison were obtained in solvent systems containing 1% DMSO, co-
solvent effects were already accounted for when comparing to precipitation
methods. The aqueous solubility was also predicted using the general solubility
equation (GSE).10

96
 4.2 Materials and methods

Amiodarone HCl (A-8423), clotrimazole (C-6019) and tolnaftate (T-6638) were

purchased from Sigma-Aldrich. Compound purity confirmed by NMR, and either
matched or exceeded the manufacturer’s specifications. (≥ 98%).

The DMSO used was UV-Vis grade, purchased from Fluka.

KH2PO4 and Na2HPO4.2H2O were purchased from Fisher Scientific.

Citric acid was purchased from Fisher Scientific.

4.2.1 DMSO stock and buffer preparations

Weighed stock solutions of nominal concentrations 1 mM and 5 mM were prepared

in DMSO for each individual compound (actual stock solution concentrations ranged
from 0.99 to 1.03 mM, and 4.97 to 5.08 mM). These solutions will be referred to by
their nominal concentrations. Note that amiodarone HCl DMSO solutions visually
began to degrade after ~10 days of storage at room temperature. Fresh stocks were
therefore prepared weekly for this compound. Clotrimazole and tolnaftate stocks
were monitored by HPLC and found to be stable for several weeks.

A pH 7.0 (± 0.05) phosphate buffer was prepared, using 0.0025 moles of KHPO4 and
0.0025 moles of Na2HPO4 per litre of deionised, Milli-Q water. The pH adjustment
was conducted with 2 M NaOH if required.

A pH 4 citric acid/phosphate buffer was prepared by mixing 147 mL of a 0.1 M citric

acid solution and 103 mL of 0.2 M sodium diphosphate solution and diluting to 500
mL with Milli-Q water. The buffer was adjusted using citric acid to pH 4.0 (±0.05).

97
 4.2.2 ‘Shake-flask’ based method of solubility determination

An excess of solid (~ 10 mg) was added to 100 mL of a 1% DMSO/99% buffer

mixture (pH 4 for amiodarone HCl, pH 7 for clotrimazole and tolnaftate) and stirred
at 25 °C using a Fisher scientific stirrer FB15001, setting 4.5, in a 250 mL vessel using
a 4 cm length stir bar. At various time points (1 hour up to one week), 3 x 1 mL
aliquots were removed and centrifuged at 9,300 x g rcf for 5 minutes using an
Eppendorf 5254 R with a F45-24-11 rotor, set at 25 ± 3 °C. 0.5 mL of supernatant
was removed and diluted to 1 mL with DMSO. The samples were then offered to
HPLC for analysis. For amiodarone HCl samples, the solid suspension appeared very
turbid, as if very small particles were present. Centrifugation gave variable results –
thus, for this compound, 0.22 μm Whatman 4 mm syringe tip filters were used
during ‘shake-flask’ experiments rather than centrifugation. Due to the low
concentration of drug in present in aqueous solution, HPLC injection volumes for
‘shake- flask’ method samples were typically increased from those documented in
table I. No trend was seen over time for any compound investigated – thus, an
average value of all timepoints was taken.

4.2.3 GSE

The general solubility equation as defined by Sanghvi et al;10

Log Sw = -0.01 (MP-25) – log Ko/w + 0.5

Equation 1

Where;

Sw = aqueous solubility, moles per litre.

MP – Melting point in °C

Log Ko/w = log octanol/water partition co-efficient.

98
The values used for each compound found from literature shown in table I. All
octanol/water co-efficient values are experimental LogD values at pH 7.4

Table I: Physical properties used in GSE calculation

Melting point Log

Compound
(° C) Ko/w
18 19
Ami 161 4.92

Clot 143.520 4.921

22 9
Tol 112 5.4

4.2.4 Precipitation method 1 - NTA determination of drug

concentration at which nanoparticles are first detected

Various diluted drug stock solutions in DMSO (0 to around 2 mM, depending on the
compound) were prepared from the 1 and 5 mM stock solutions detailed above. At
all drug stock concentrations, 10 μL of stock was added to an HPLC vial. 990 μL of
buffer of the appropriate pH (4 for amiodarone HCl, 7 for clotrimazole and
tolnaftate) was filtered and added. The % DMSO by volume was kept constant at
1%. The samples were removed from the vial and introduced into the NanoSight
chamber. The point at which the measured nanoparticle concentration became
significantly different from that of the blank was determined in Excel using a two
tailed t-test (95% confidence interval).

4.2.5 Precipitation Method 2 - Analysis by HPLC and subsequent

investigation of separation and kinetics by NTA

(i) Suspension Preparation

An aliquot (10 µL) of 5 mM DMSO stock solution was pipetted into a

microcentrifuge tube. A 990 μL aliquot of the appropriate buffer was then added.
This gave a total concentration of 50 μM with a 1% DMSO content by volume.

99
For NTA analysis, the pH 4 or pH 7 buffer was filtered using a 0.2 μm hydrophilic
PTFE filter (Millipore) before addition to the DMSO aliquot. Precipitations were also
performed in glass HPLC vials rather than plastic microcentrifuge tubes for NTA. If
centrifugation was required, the suspension was transferred using a glass pipette to
a microcentifuge tube after initial precipitation had been initiated. All samples were
prepared at 25 °C. A minimum of three replicates were performed for each variable.

(ii) Separation and Analyses

After incubation (if required), the samples were then either centrifuged (9,300 x g
rcf, 5 minutes) or filtered (0.45 μm PVDF Millex-HV 4 mm syringe filter, 0.22 μm
Whatman 4 mm syringe tip filter).

NTA – supernatant/filtrate samples were measured on the NanoSight as is. Note

samples were also analysed by NTA with no separation performed i.e. immediately
after addition of the buffer aliquot.

HPLC Analysis – Dilution and further preparation required.

For centrifuged samples - The entire supernatant was removed from the
microcentrifuge tube, and 0.5 mL of supernatant was diluted to 1 mL with DMSO to
give the supernatant sample. To the microcentrifuge tube containing the
precipitate, 1 mL of DMSO was added, and the solution vortexed for around 5
seconds to ensure full dissolution of the precipitated material. This is the
‘precipitate’ for centrifuged samples.

For filtered samples – The filtrate was diluted with DMSO by a factor of 2. The used
filter tip was retained, placed back into the tube where the initial precipitation had
taken place, and 1 mL of DMSO added. The tube was rotated for 15 mins to ensure
dissolution of the solid within the filter membrane. This is the ‘precipitate’ for
filtered samples.

The samples were then offered to HPLC for analysis.

100
Addition of the quantified amounts from both the supernatant and precipitate gave
the total amount recovered for each sample. Recoveries in the range 85-115 % were
deemed acceptable and included in the calculation of the average. Filtration
consistently gave total recoveries < 85% for all compounds – results are still
displayed with this noted.

4.2.5 Instrument Details

(i) HPLC Settings

Analysis was carried out using a Waters Alliance HPLC system, separation module
2695, dual wavelength detector 2487 and PDA detector 2996, used in conjunction
with Empower Pro, 2002 (build 1154) software. The column used was Thermo
Scientific Hypersil Gold 50 x 4.6 mm (or 50 x 4 mm) with a 3 μm particle size. Mobile
phase A comprised of Milli-Q ultra filtered water + 0.01 % TFA and Mobile phase B
comprised of HPLC Grade MeCN + 0.01% TFA. For all compounds, the column
temperature was set to 40 °C and the mobile phase flow rate to 1 mL/min. The total
data collection time was 3 minutes for all 3 compounds and an isocratic method
was used. Individual HPLC settings were as follows;

Table II: HPLC conditions

Detection
Mobile phase Injection Retention time
Compound wavelength
ratio A:B (%) volume (μL) (mins)
(nm)

Ami 60:40 270 10 1.82

Clot 60:40 265 10 1.77

1.99 (2.7 on 50
Tol 40:60 257 3
x 4 mm column)

Calibration procedure – 6 calibration solutions (100% DMSO) were prepared in the

concentration range 5 to 50 μM for each compound, originating from 1 and 5 mM
weighed stock solutions in DMSO as detailed previously. Fresh calibration solutions

101
were prepared and analysed for each experiment along with the samples under
test.

(ii) NTA Settings

Instrument - NanoSight LM10 with NTA software version 2.1 release build 0316
(2010).

Videos lengths were 90 seconds, unless otherwise stated, and all processing
parameters set to automatic. The blur was increased in all cases to 5 x 5. All
analyses were performed at 25 °C. A distribution which had been adjusted as
normal by the software, but had not been smoothed, was obtained. This non-
smoothed data is that shown or used further in calculations etc. Camera shutter and
gain settings for each compound and experiment are detailed in table III.

Table III: NTA Video capture settings

Kinetic Method 2 - Various separation techniques

Kinetic method 1 -
NTA
No separation Centrifugation 0.45 Filtration

Compound Shutter Gain Shutter Gain Shutter Gain Shutter Gain

Ami 1500 680 1500 500 1500 680 1500 680

Clot 1500 680 100 0 605 260 1500 680

Tol 700 690 66 0 153 0 300 459

The amount of drug present as particles can be estimated from the NanoSight
output, by calculation of the volume mean and utilisation of the NTA concentration
estimate. 23 Knowledge of the particle density is also required – here we used the
density of the crystalline solid in the calculations, although accept it is not known
whether the particles are in fact crystalline or not. (Density values used in the
calculation are in g/cm3 – Amiodarone HCl, 1.71,24 Clotrimazole, 1.31625 and
Tolnaftate, 1.22323).

102
 4.3 Results and discussion

4.3.1 ‘Shake-flask’ solubility and GSE results

The results from the ‘shake-flask’ solubility experiment and GSE calculations were
as follows (table IV);

Table IV: Solubility from solid and GSE values

‘Shake- GSE
flask’ predicted
Compound
solubility ± solubility
S.D ( μM) (μM)

Ami 1.9 ± 0.2 1.7

Clot 4.3 ± 1.3 2.6

Tol 1.5 ± 0.8 1.7

Ami – n =15 (average of 5 timepoints, 3 samples per timepoint), Clot – n=9 (average of 3 timepoints, 3 samples

per timepoint), Tol – n=15 (average of 5 timepoints, 3 samples per timepoint ).

No corrections were applied for % ionisation (relevant for amiodarone HCl and
clotrimazole) at the pH of solution used in the measurement. Amiodarone HCl and
clotrimazole would be expected to be positively charged at the pH values used here
(7 for clotrimazole and 4 for amiodarone HCl). All partition values used are LogD at
pH 7.4. The GSE gives the solubility of the compound in pure water, whereas the
‘shake-flask’ experiments contain 1% DMSO. Despite all of these factors, the GSE
gives solubility values which are in good agreement with those obtained
experimentally from the solid using a ‘shake-flask’ method.

4.3.2 Methods originating from DMSO stock solutions –

precipitation methods

(i) Precipitation Method 1 - NTA based

The first method used to determine solubility utilising DMSO stocks employed the
NTA and the NanoSight to measure the increase in particle concentration as drug

103
concentration is increased. Test compound concentrations above and below the
‘shake-flask’ solubility were tested until concentration at which nanoparticle
formation began could be determined (table V). This was performed by obtaining
the x- intercept of the slope of increasing particle concentrations, in a similar way to
that described by Hoelke et al.3

Table V: Kinetic solubility from NTA results

Drug concentration at
Compound which particle formation
occurs

Ami 5 μM

Clot 16 μM

Tol 2 μM

In order for this analysis to give a similar result to that obtained from section 4.3.1,
one main criterion would have to be met. The drug would have to precipitate as
soon as the equilibrium solubility value was exceeded i.e. not have the ability to
remain supersaturated for any length of time. The fulfilment of this criterion will of
course be compound specific. For tolnaftate, the result in table V is in reasonable
agreement with the results obtained from section 4.3.1, suggesting that despite
using a precipitation based method, a solubility value very close to the equilibrium
solubility has been obtained. Nucleation is thus initiated at total drug
concentrations very close to the equilibrium solubility for this compound. For
amiodarone HCl and, in particular, clotrimazole, the results are slightly higher than
those from the ‘shake-flask’ emthod and GSE, suggesting these compounds have
the ability to remain supersaturated for the short time period of the experiment.
For all three compounds, nanoparticle formation occurs rapidly at concentrations
fairly close to the equilibrium solubility value.

This approach to determining kinetic solubility has a similar underlying principle to

the use of nephelometric methods of kinetic solubility determination i.e. light

104
scattering is used to detect the total drug amount at which scattering from
precipitate becomes significant compared to background scattering. It has been
stated however that nephelometric methods have a lower limit of around 20 μM.7
Whilst the NTA method used here appears more sensitive, it does not have the
capability to be automated or performed in plates. The results presented here have
constant % DMSO, whereas in some other analyses, the % DMSO can vary as stock
solution is added incrementally to a buffer until scattering is detected. 2

(ii) Precipitation Method 2 - HPLC based

The second precipitation method involved adding a volume of drug dissolved in

DMSO to the buffer such that the total drug amount was in excess of that expected
to remain dissolved in solution. The same concentration (50 μM) was chosen for all
three compounds. Since the system has a total drug content above the equilibrium
solubility value, precipitation is expected to occur. After precipitation, the solid and
liquid phases are separated and quantification of the amount of drug in each phase
performed by HPLC. The results from these experiments are shown table VI;

Table VI: Kinetic solubility results for all three test compounds, with sample analysis performed by
HPLC

Amount present in
Amount present in Amount present in
supernatant ± S.D.
Compound supernatant ± S.D. (μM), supernatant ± S.D. (μM),
(μM), after 0.45 μm
after centrifugation (5 mins, after 0.22 μm filtration
filtration
9,300 x g rcf)

Ami 27.5 ± 1.53 (n=15) 27.3 ± 2.7 (n=6) *11.9 ± 0.8 (n=3)

Clot 20.2 ± 3.2 (n=21) *8.4 ± 0.6 (n=6) N/A

Tol 13.3 ± 5.6 (n=30) *No Peak Detected N/A

*poor total recovery (<85%) N/A – analysis not performed

Again, in order for the results obtained to be similar to those from section 4.3.1,
certain conditions must be met. The drug must precipitate (within the timescale of
the experiment) to its equilibrium solubility, and not remain supersaturated for any

105
extended time period. The separation process of the solid and the liquid phases
must be efficient.

From table VI, various different results were obtained depending on the exact
conditions used to separate the solid and liquid phases;

(i) Centrifugation

All three compounds show significantly more compound present in the supernatant
after centrifugation than would be expected from the results in section 4.3.1.
Tolnaftate shows the least amount of drug in the supernatant, and amiodarone HCl
the most. The total amount of drug present per mL (50 nanomoles) is constant
between compounds, however in order to fairly compare the relative drug amounts
in the supernatant after centrifugation treatment, the ‘apparent’ supersaturation
(total drug amount present/’shake-flask’ solubility) should also be accounted for.
This calculates as 33 for tolnaftate, 26 for amiodarone HCl and 12 for clotrimazole.
Fast precipitation would be expected from all three compounds from these
calculated values. Tolnaftate has by far the highest ‘apparent’ supersaturation, and
as expected has precipitated the most within the experimental timescale (lowest
supernatant amount). Clotrimazole has the lowest ‘apparent’ supersaturation, and
could therefore be expected to have the slowest kinetics, with the highest
supernatant amount. This is not the case, with amiodarone HCl having the highest
supernatant content.

(ii) Filtration

For 2 compounds (clot and tol), 0.45 μm PVDF filtration lowers both the drug
amount detected after liquid/solid separation and the % recovered compared to
centrifugation. For amiodarone HCl, no difference is seen in either drug amount
detected in the filtrate compared to the supernatant or in % recovery. Moving
down to a smaller pore size filter (but maintaining the same membrane type) results
in a lower amount of drug detected and lowered recoveries.

106
Overall, the results detailed here are significantly higher than those obtained from
the solubility determination methods discussed earlier. As mentioned, ‘kinetic
solubility’ assays can result in the formation of supersaturated solutions – this
would explain the increase in solubility value seen if it were assumed that the
compound had not reached equilibrium within the timescale of the experiment.
There also had to be sufficient separation of the solid and liquid phases – this was
investigated further (table VII). Note we can rule out immediately any co-solvent
(thermodynamic) increase in solubility from the presence of DMSO itself as the
‘shake-flask’ experiments used in comparison were also performed in a 1%
DMSO/buffer solvent system. Another complicating factor is that hydrophobic
compounds have a tendency to adsorb non-specifically to filter membranes – this
would account for the reduction in both detected drug amount and % recoveries
seen for clotrimazole and tolnaftate after use of 0.45 μm PDVF filters. Amiodarone
HCl however only showed a decrease in recovery and drug filtrate amount when
0.22 μm filters were used.

In order to further investigate these results, similar experiments were performed

but this time followed by NTA rather than HPLC analysis, with results as detailed in
table VII.

Table VII: NTA analysis of the supernatant/filtrate after sample suspension centrifugation/
filtration

5 mins 9,300 x g cent 0.45 um PVDF filter

Calculated Calculated
Particle nanomoles Particle nanomole
Compound concentration of drug per concentration of drug
Diameter (nm) Diamter (nm)
(106 mL (106 per mL
particles/mL) particles/mL)
Ami 111 ± 2 12 ± 1 3 ± 0.3 110 ± 14 10 ± 7 4±3

Clot <100 tracks* 0.4 ± 0.2* N/A < 100 tracks* 0.2 ±0.2* N/A

Tol 221 ± 9 5±2 14 ± 6 < 100 tracks* 0.03± 0.06* N/A

*<100 tracks – calculated diameter from NTA analysis inaccurate due to low number of particles
included in distribution. This is also reflected in the particle concentrations obtained, which are very
low.

107
It can now be seen from table VII that the centrifugation parameters used here do
not completely remove the nanoparticles for all compounds under test, and that
filtration has varied results. This helps explain the high results seen. Despite the
presence of nanoparticles and the kinetic nature of the measurement, the results
produced by HPLC were very reproducible as can be seen from the small standard
deviation values in table VI. It is worth mentioning that several replicates of the
centrifuged HPLC samples (table VI) were performed (15 for amiodarone HCl, 21 for
clotrimazole and 30 for tolnaftate), using various different stock solutions, buffer
solutions and on different days – the HPLC supernatant amounts are very consistent
despite this. The results can now be rationalised on a compound by compound
basis.

Table VIII summarizes the results from the various ‘solubility’ assessments detailed
so far for tolnaftate;

Table VIII: Summary of results from Tolnaftate solubility experiments – ‘shake-flask’, GSE and
precipitation based methods

‘Solubility’ value obtained (μM - mean plus/minus standard deviation)

Precipitation
Precipitation
Precipitation method 2,
‘Shake- method 2
GSE method 1 (0.45 μm
Flask’ (centrifugation
(NTA based) filtration, then
then HPLC)
HPLC)
no peak
Tol 1.5 ± 0.8 1.7 2 13.3 ± 5.6
detected
*poor recovery

Table VIII shows that tolnaftate’s equilibrium solubility is low; however the
‘solubility’ value obtained from precipitation method 2 (centrifugation then HPLC
analysis) however gives a much higher supernatant value. We can now attribute this
to nanoparticles (table VII) which have not been removed by centrifugation. Thus,
we are not seeing supersaturation in the usual sense, but a sample which contains
nanoparticles contributing to the HPLC supernatant amount. From NTA calculations,
the amount of drug present as nanoparticles more than accounts for the HPLC
supernatant amount. It should however be borne in mind that this calculated
amount is an estimate due to the nature of the NTA concentration measurement.

108
This also agrees with the precipitation and NTA analysis results, where tolnaftate
shows little tendency to remain supersaturated. Upon filtration, the particles are
completely removed and the HPLC response obtained comes from molecularly
dissolved solution only. The measured response after filtration is below that of the
solubility from solid value, likely due to non-specific adsorption of the drug to the
filter membrane, which also explains the lower recovery values.

A summary of clotrimazole’s behaviour is detailed in table IX;

Table IX: Summary of results from Clotrimazole solubility experiments – ‘shake-flask’, GSE and
precipitation based methods

Solubility value obtained, (μM - mean plus/minus standard deviation)

Precipitation,
Precipitation,
From Precipitation, 0.45 μm
GSE centrifugation
Solid by NTA filtration, then
then HPLC
HPLC
Clot 4.3 ± 1.3 2.6 16 20.2 ± 3.2 8.4 ± 0.6

Again, the precipitation method with HPLC analysis shows a much higher value
compared to that obtained from solid. From table VII, clotrimazole shows very few
particles present, either after centrifugation or filtration. A reduction in the
detected drug and in the total recovery by HPLC is, however, seen after filtration,
again suggesting some adsorption. Both filtered and centrifuged samples have a
drug response higher than the solubility from solid value. From the filtered results,
the compound shows the capability to remain supersaturated in the conventional
sense at this concentration. This also agrees with the results in table V, where
clotrimazole demonstrated a capability to remain supersaturated.

Amiodarone HCl’s behavior can be summarized as follows;

109
Table X: Summary of results from Amiodarone HCl solubility experiments – ‘shake-flask’, GSE and
precipitation based methods

Solubility value obtained, (μM - mean plus/minus standard deviation)

Precipitation, Precipitation,
Precipitation,
From Precipitation, 0.45 μm 0.22 μm
GSE centrifugation
Solid by NTA filtration, filtration, then
then HPLC
then HPLC HPLC
Ami 1.9 ± 0.2 1.7 5 27.5 ± 1.5 27.3 ± 2.7 *11.9 ± 0.8
*poor recovery

Again, the compound has low solubility, but this is not apparent from the
precipitation-HPLC based experiments. However, whilst high concentrations of
nanoparticles are present for amiodarone HCl, their small size means they are
estimated to contribute to only a few μM of the measured HPLC response, (Table
VII, NTA data). This is in contrast to tolnaftate, where the particles accounted for a
substantial portion of the HPLC supernatant value. Filtration through 0.45 μm filters
does not remove the particles formed by amiodarone HCl, and no loss in measured
drug amount or total recovery is seen from HPLC is seen (table IX). Moving to 0.22
μm filters, most nanoparticles are removed (calculated to account for < 1 μM of
drug from NTA, results not shown). A lower amount of drug measured in the filtrate
by HPLC is seen, combined with lower recoveries, again suggesting some non-
specific adsorption to the filter. However, the 0.22 μm filtrate response (containing
no/few nanoparticles) is still higher than that obtained when starting from solid
(table IX). This compound is demonstrating true supersaturation i.e. the samples
appear to contain molecularly dissolved content. An interesting point to note is that
a decrease in % recovery is only seen when the nanoparticles are removed, despite
use of the same filter membrane. This suggests that the loss in recovery could
potentially be due to nanoparticle adsorption rather than adsorption of the
molecularly dissolved content.

4.3.3 Precipitation methods and varying incubation times

Intuitively, solubility values obtained by precipitation should depend on the

incubation time employed. For all of the previous results, no incubation time was

110
employed – separation of the liquid and solid phases took place immediately after
the aqueous phase was added to the DMSO stock. Both HPLC (with centrifugation)
and NTA analysis (no separation) were employed to investigate the effect of time on
the solubility results (table XI and XII).

Table XI: Amount of drug present in the supernatant after centrifugation for tolnaftate,
clotrimazole and amiodarone HCl, as a function of time

Amount of drug present (μM) in the supernatant

after centrifugation (9,300 x g rcf, 5 mins)

Incubation time Amiodarone

Clotrimazole Tolnaftate
(mins) HCl

0 27.5 ± 1.5 20.2 ± 3.2 13.3 ± 5.6

10 23.8 ± 0.3 - 9.4 ± 6.7

60 22.9 ± 0.5 16.5 ± 0.5 -

180 - 18 ± 1 3.0 ± 0.5

1440 24.0 ± 0.3 2.9 ± 0.2 *5.1 ± 1.8

*poor total recovery value obtained

As expected, the solubility values obtained by HPLC for all 3 compounds decreases
with time. The time taken to become close to the solubility from solid value
however varies dramatically between compounds (values bold and italicised).
Tolnaftate is very close to the solubility from solid value after only ten minutes
incubation; clotrimazole is still well in excess of the solubility from solid value after
3 hours incubation, but has reached a similar value after 24 hour incubation;
amiodarone HCl is still well in excess of the solubility from solid even after a 24 hour
incubation period. What remains unknown is why clotrimazole, having a lower
‘apparent’ supersaturation than amiodarone HCl, reaches an equilibrium value
faster. The NTA results in table XII add further insight;

111
Table XII: NTA analysis of sample suspensions for all three model compounds, incubated for
various amounts of time at room temperature

Calculated
Average particle
Average particle nanomoles of drug
concentration (E6
diameter (nm) present as
particles /mL)
nanoparticles

*Initial rate End of

End of fast End of fast
of increase fast
initial initial
Compound of mean t0 initial t0 t0
particle particle
diameter particle
growth growth
(nm/min) growth

19.8 ±
Ami 2 118 ± 3 270 ± 38 6.4 ± 1 6±1 27 ± 10
2.5

>100 % of
Clot 1 215 ± 3 450 ± 24 9.9 ± 2.8 0.5 ± 0.1 26 ± 8
added

185 ± 19.1 ± >100 % of

Tol 7 635 ± 26 2.8 ± 0.4 36 ± 16
19 3.0 added

*initial rate of mean diameter increase taken as the initial slope in the graph of mean particle
diameter against time; for tolnaftate and amiodarone HCl, the fast initial rate increase occurred
between t0 and t 60. For clotrimazole it occurred between t 0 and t 180. Note this is a rough
indication of the average particle diameter increase rate to give a comparable variable between
compounds. Diameter and concentration results are for the start and end of this fast initial growth
period.

As detailed previously, the compounds can be ranked in terms of ‘apparent’

supersaturation (Tolnaftate>amiodarone>clotrimazole). In terms of the initial mean
diameter increase rate seen, this ranking holds true. Having now looked at several
methods of solubility behaviour determination, some overall conclusions for each
compound and its individual precipitation behaviour can be deduced.

Tolnaftate has fairly straightforward precipitation behaviour. It ‘immediately’ forms

nanoparticles upon mixing DMSO and buffer solutions. As expected from a system
with high apparent supersaturation, it reaches equilibrium solubility rapidly. The
enhanced HPLC response is merely due to the difficulties in completely removing
the nanoparticles present – no molecularly dissolved drug above the solubility from
solid value appears to be present. The ‘supersaturation’ seen is therefore not ‘true‘

112
supersaturation. Fast particle growth kinetics are seen; after 3 hours, the particles
have grown to a large enough size to allow removal by centrifugation.

Clotrimazole has slightly more complicated precipitation behaviour. As with

tolnaftate, it forms substantial numbers of nanoparticles upon mixing the aqueous
and organic components together. In contrast to tolnaftate, these particles do not
account for the full HPLC supernatant response (based on calculations). We have
also shown (from filtrate HPLC response where filtrate is known from NTA to
contain few particles) that there is dissolved drug content above the solubility from
solid value. ‘True’ supersaturation is therefore seen. This is despite the fact that
nanoparticles have already formed and thus initial nucleation has already begun.
This means while initial nucleation is rapid, growth appears to be slow. Slow
growth/precipitation to equilibrium is also seen from tables XI and XII. From table
XI, clotrimazole precipitation does eventually ‘finish’ at the 24 hr time point i.e. the
particles have grown to a size that can be centrifuged out and the amount of drug
remaining in solution is close to the solubility from solid value.

Amiodarone HCl also forms nanoparticles as the first measurable step in the
precipitation process. As with the other two compounds, precipitation methods
with HPLC analysis give the highest solubility results. Particles are still present in the
supernatant, although they are calculated to only contribute to a few μM.
Amiodarone HCl therefore has behaviour similar to clotrimazole; despite the
presence of nanoparticles, there is still a significant amount of drug present
molecularly dissolved. It can be seen from table XII that amiodarone HCl initially
forms very small (and weakly scattering relative to the other two compounds)
particles. This is one of the main differences between this compound and the other
two - amiodarone HCl particles are around half the size of those formed by
tolnaftate and clotrimazole. Another difference is that they cannot be removed by
0.45 μm PVDF filters. This compound also unexpectedly, based on ‘apparent’
supersaturation, shows the highest HPLC supernatant amounts, both initially and
after a 24 hour incubation period. This can now be linked to the small size of the

113
particles formed. Despite a slowing of the particle growth rate at t60, amiodarone
HCl particles are still < 300 nm in diameter after an hour’s incubation. The small
size of the particles means that there is little change in the amount of compound
measured in the supernatant over time by HPLC - despite growing at a reasonable
rate, they do not reach a large enough size to be removed fully by the
centrifugation employed and are counted in the HPLC response along with any
molecularly dissolved content. It should also be noted that a similar time course
experiment on the HPLC was performed with amiodarone HCl only, using 0.22 μm
filters (results not shown). Even after 24 incubation, the filtrate (containing no/few
nanoparticles) was still in excess of the solubility from solid value (12.0 ± 0.1 μM),
meaning the compound was still supersaturated, despite having a large number of
nanoparticulate nuclei present. A combination of amiodarone HCl’s small initial
particle size and the stability of the samples at concentrations above the
equilibrium mean that it shows the slowest overall kinetics of the three compounds,
despite having a high sample ‘apparent’ supersaturation.

The determination of the kinetic solubility, using HPLC, of all three compounds
investigated here was complicated by the presence of nanoparticles. The
centrifugation parameters employed were not sufficient to remove all of the
particles; filtration however gave compound dependent results, and for one
compound, the nanoparticles could pass through the filter membrane completely.
Thus determining whether the samples were truly supersaturated or whether the
HPLC response was partly due to the presence of nanoparticles involved several
other experiments. Of course, should a compound form larger, micron-sized
particles ‘immediately’ or at least on a fairly short time scale (like tolnaftate), these
problems are largely avoided. Several papers have compared different methods of
kinetic solubility determination for a large number of compounds;3-8 crucial
differences in the work performed here include the use of a light scattering method
with a lower limit of detection, and a detailed investigation into the type of particles
formed and their behaviour.

114
As mentioned previously, it is known, that a large number of compounds form
‘aggregates’ of a few hundred nm in size upon mixing DMSO stock solutions and
aqueous buffers at the μM concentrations relevant here. 12, 15, 13, 14 This is well
documented in bioassay and screening literature, but the impact of the sort of
particles on the results of experiments similar to those performed during solubility
screening has not been assessed until now. Here we show that the separation
method used can affect the results of the solubility assay, with centrifugation
potentially resulting in incomplete removal of nanoparticulate material, and
filtration giving compound dependent results with regard to nanoparticle removal
and adsorption. As expected, time has an effect on the samples –this change in
response can be due to both supersaturation effects and growth rate of
nanoparticles to larger, removable sizes. We also show that, while initial particle
formation for all three compounds investigated here was rapid, growth was much
slower.

115
4.4 Conclusions

Several conclusions can be drawn from the work presented. In terms of the
analytical method used, determination of the ‘kinetic solubility’ by NTA using a
precipitation method gives results closest to that obtained by solubility from solid.
In terms of determining the solubility via precipitation using HPLC analysis, several
factors can dramatically affects the results of the experiment. On a sample
preparation level, the separation method used can change the results of the
solubility assay due to the small particle sizes encountered here. Centrifugation can
result in incomplete separation of the nanoparticles and the molecularly dissolved
content, which can lead to an increase in kinetic solubility solely due to nanoparticle
presence. The extent of this increase depends on the centrifugation parameters
employed, the density of the particles and the size of the initial particles. Of course,
the analysis of compounds which formed larger sized particles would not encounter
these problems. Filtration using a 0.45 μm filter membrane can actually remove
nanoparticles smaller than expected (~200 nm), likely due to adsorption. At smaller
sizes (~100 nm), the particles can pass through the filter and a 0.22 μm filter is
required. For all three compounds, low total % recovery was obtained using filters,
suggesting some type of non-specific adsorption.

From the point of view of the precipitation process itself, fast nucleation and
subsequently relatively slower growth to larger sizes is a common theme for all
three compounds. As expected, incubation time decreases the kinetic solubility
value obtained by HPLC. The rate at which this decreased occurred varied for all
three compounds. This was rationalised by looking at the supersaturation of the
sample, the initial nanoparticle growth rate, the initial size of the particles formed
and the ability of the compound to demonstrate some supersaturation. Here we
have used a limited selection of 3 structurally unrelated, poorly water soluble test
compounds – further and more complete understanding of the process would
involve similar experiments with a larger sample set.

116
4.5 References

1. J. Alsenz and M. Kansy, Advanced Drug Delivery Reviews, 2007, 59, 546-567.
2. C. A. Lipinski, F. Lombardo, B. W. Dominy and P. J. Feeney, Advanced Drug
Delivery Reviews, 1997, 23, 3-25.
3. B. Hoelke, S. Gieringer, M. Arlt and C. Saal, Analytical Chemistry, 2009, 81,
3165-3172.
4. A. Blasko, A. Leahy-Dios, W. O. Nelson, S. A. Austin, R. B. Killion, G. C. Visor
and I. J. Massey, Monatshefte Fur Chemie, 2001, 132, 789-798.
5. C. D. Bevan and R. S. Lloyd, Analytical Chemistry, 2000, 72, 1781-1787.
6. B. Bard, S. Martel and P.-A. Carrupt, European Journal of Pharmaceutical
Sciences, 2008, 33, 230-240.
7. C. Teng-Man, S. Hong and Z. Chegnyue, Combinatorial Chemistry and High
Throughout Screening, 2002, vol. 5, pp. 575-581.
8. K. A. Dehring, H. L. Workman, K. D. Miller, A. Mandagere and S. K. Poole,
Journal of Pharmaceutical and Biomedical Analysis, 2004, 36, 447-456.
9. Y. W. Alelyunas, R. Liu, L. Pelosi-Kilby and C. Shen, European Journal of
Pharmaceutical Sciences, 2009, 37, 172-182.
10. T. Sanghvi, N. Jain, G. Yang and S. H. Yalkowsky, QSAR & Combinatorial
Science, 2003, 22, 258-262.
11. K. E. D. Coan, J. Ottl and M. Klumpp, Expert Opinion on Drug Discovery, 2011,
6, 405-417.
12. S. L. McGovern, E. Caselli, N. Grigorieff and B. K. Shoichet, Journal of
Medicinal Chemistry, 2002, 45, 1712-1722.
13. S. L. McGovern, B. K. Shoichet and n. Kinase Inhibitors, J. Med. Chem., 2003,
46, 1478.
14. S. L. B. T. McGovern, B. T. Helfand, B. Feng and B. K. Shoichet, J Med Chem,
2003, 46, 8.
15. J. Seidler, S. L. McGovern, T. N. Doman and B. K. Shoichet, Journal of
Medicinal Chemistry, 2003, 46, 4477-4486.
16. B. Y. Feng, A. Shelat, T. N. Doman, R. K. Guy and B. K. Shoichet, Nat Chem
Biol, 2005, 1, 146-148.
17. K. E. D. Coan and B. K. Shoichet, Journal of the American Chemical Society,
2008, 130, 9606-9612.
18. T. A. Plomp, in Analytical Profile Of Drugs and Excipients, ed. K. Florey, vol.
20.
19. K. M. Waldhauser, K. Brecht, S. Hebeisen, H. R. Ha, D. Konrad, D. Bur and S.
Krähenbühl, British Journal of Pharmacology, 2008, 155, 585-595.
20. L. Borka and S. Valdimarsdottir, Acta Pharamceutica Suecica, 1975, 12, 479-
484.
21. Y. Dohta, T. yamashita, S. Horiike, T. Nakamura and T. Fukami, Analytical
Chemistry, 2007, 79.
22. A. K. Dash, Journal of Microencapsulation, 1997, 14, 101-112.
23. C. Gillespie, P. Halling and D. Edwards, Colloids and Surfaces A:
Physicochemical and Engineering Aspects, 2011, 384, 233-239.

117
24. V. Cody and J. Luft, Acta Crystallographica Section B, 1989, 45, 172-178.
25. H. Song and H. S. Shin, Acta Crystallographica Section C, 1998, 54, 1675-
1677.

118
 5. The effect of % DMSO on solubility measurements
and nanoparticle behaviour

5.1 Introduction

The use of DMSO as a compound storage or stock medium is commonplace within

the pharmaceutical industry. Aliquots of the DMSO stock solution are then utilised
for various different analyses, such as high throughput screening (HTS) biological
activity assays, and solubility screens at the discovery stage. These analyses both
typically involve the addition of the DMSO stock to an aqueous buffer system - since
most compounds are hydrophobic, there is the possibility of precipitation on the
timescale of the analysis. Within a kinetic solubility setting, precipitation is an
anticipated phenomenon, and is even used as a way of detecting when the
solubility limit has been reached.1 Within a bioassay setting, it is assumed
precipitation does not occur, and the test compound is fully dissolved and able to
interact with whatever target is being used (e.g. cell, enzyme). The precipitation of
the compounds from the bioassay media is a recognised issue, and can often
interfere with the results of the assay.2 Both false negative (due low amount of drug
dissolved and available to interact with target) and false positive (aggregates of the
compound non-stoichiometrically inhibit the target) results could potentially be
obtained.

Various controllable factors within the experimental set up for both types of assay
could theoretically affect the precipitation process, which in turn would affect the
results obtained. One of these factors is the percent DMSO present in the final
sample composition. Within a bioassay setting, the amount of DMSO present in
samples is kept to a minimum (maximum 5% by volume), in order to reduce any
interference effects on the assay. This is the main concern with the DMSO present,
rather than its effect on any compound precipitation. In kinetic solubility analyses,
varying percentages of DMSO can be used depending on the protocol, although as

119
with bioassays, it is generally kept to a minimum. DMSO is reported as having both
‘co-solvent’ and ‘supersaturation’ effects on kinetic aqueous solubility values i.e.
affects both the equilibrium solubility and the ability of the system to
supersaturate. Bard et al 3 reported that enhanced kinetic solubility values were
mainly due to a compound’s ability to form stable supersaturated solutions when
prepared from DMSO stocks, with co-solvent effects only being modest. Chen et al
reported an increase in solubility when samples had DMSO present (starting from
the stock solution, using a precipitation method) compared to those with no DMSO
present (starting from the solid, using a ‘shake-flask’ type dissolution method). The
group noted that poorly soluble compounds could be affected significantly. 4 Martel
et al noted that increasing the percent DMSO resulted in solubility values which had
less correlation to those obtained with no DMSO present. They also noted the effect
was highly compound specific.5 Here we explore in detail the effect of the % DMSO
on the precipitation of 3 poorly water soluble test compounds (amiodarone HCl,
clotrimazole and tolnaftate), at concentrations typically used in bioassay and kinetic
solubility screening. Previously (chapter 4), the kinetic solubility of all three
compounds was explored in detail at 1% DMSO by volume. Here, we increase this to
5% DMSO, while retaining the same total drug amount added, and compare the
results of the experiments.

120
 5.2 Materials and Methods

The sample preparation details, instrumental details, methods and materials are all
as used for 1% samples in chapter 4. All that differed was the volumes of DMSO and
buffer, and the concentration of the DMSO stock used. For example; to prepare a
50 μM (total volume 1 mL), 5% DMSO by volume suspension of drug, 50 μL of a 1
mM stock solution in DMSO was pipetted. 950 μL of the appropriate buffer was
then added.

121
 5.3 Results and Discussion

5.3.1 ‘Shake-Flask’ solubility

The following results were obtained from the ‘Shake flask’ solubility experiments;

Table I: ‘Shake-Flask’ solubility results

‘Shake-Flask’ solubility ± S.D (μM)

p value for t-test comparing

Compound 1% DMSO/Buffer 5% DMSO/Buffer
1 and 5% DMSO solubility values

Amiodarone HCl 1.9 ± 0.2 3.7 ± 0.8 Sig (0.008)

Clotrimazole 4.3 ± 1.3 6.6 ± 1.5 *NS (>0.05)

Tolnaftate 1.5 ± 0.8 1.6 ± 0.8 *NS (>0.05)

Ami – n =15 (average of 5 time points, 3 samples per time point), Clot – n=9 (average of 3 time points, 3 samples

per time point), Tol – n=15 (average of 5 time points, 3 samples per time point ).*Where NS = not significant,
95% confidence interval used

As can be seen from Table I, all three compounds show a slight increase in going
from 1 and 5% DSMO/buffer mixtures. However, only amiodarone HCl samples
show any statistically significant enhancement from a 5% DMSO/buffer mixture
compared to a 1% DMSO mixture, and in terms of solubility, both 1 and 5% DMSO
results would still be thought of as poorly soluble. Thus while a co-solvent effect is
present for these samples, it is fairly modest. However, amiodarone HCl
precipitation samples containing 1 and 5% DMSO will, for the same amount of drug
added, have slightly different ‘apparent’ supersaturations (total amount of drug
added/’shake-flask’solubility value) – this will be borne in mind in later discussions.
Both tolnaftate and clotrimazole have very similar ‘apparent’ supersaturations, for
the same total amount of drug added, regardless of the % DMSO present.

122
 5.3.2 Kinetic Solubility Assay Results

(i) Precipitation method 1 – NTA based

As with Chapter 4, 2 methods were used to determine the kinetic solubility of the 3
compounds. Method 1 involved using NTA (light scattering) to determine the
compound concentration at which the detected nanoparticle concentration
becomes significantly different from a blank sample, as described in chapter 4. The
results obtained for the 1% DMSO samples are documented in that chapter; here
the equivalent 5% DMSO samples were tested using the same settings as a
comparison (Table II).

Table II: Kinetic solubility determination by NTA

Drug concentration at which particle formation

occurs

Compound 1% v/v DMSO 5% v/v DMSO

Amiodarone HCl 5 μM 4 μM

Clotrimazole 16 μM 16 μM

Tolnaftate 2 μM 2 μM

The results from both 1 and 5% DMSO are very similar, regardless of the compound
investigated. As discussed in chapter 4, these are also fairly similar to the solubility
from solid values obtained. This suggests no effect of the % DMSO present on the
concentration at which nanoparticle formation is initiated. Initial nucleation at
concentrations fairly close to the equilibrium solubility is therefore not affected.

123
 (ii) Precipitation method 2 – HPLC based

In this method, an excess of compound is added via the aliquot of DMSO stock
solution such that the total drug amount is greater than the solubility from solid
amount. Separation of the precipitate from the molecularly dissolved compound is
then performed. Shown are the results of HPLC analysis where either centrifugation
or filtration has been used as the solid/liquid separation method (table III).

Table III: Comparison of results obtained from samples containing 1 and 5% DMSO, using
precipitation method 2, analysed by HPLC after either centrifugation or filtration

Centrifugation (5 mins, 0.22 μm PVDF

0.45 μm PVDF filtration
9,300 x g rcf) filtration

5% DMSO supernatant
5% DMSO supernatant 5% DMSO supernatant
value as a % of 1%
Compound value as a % of 1% DMSO value as a % of 1% DMSO
DMSO supernatant
supernatant value supernatant value
value

Amiodarone HCl 145% 140% 213%

Clotrimazole 154% 232% Not performed

Tolnaftate 235% *N/A Not performed

*1% DMSO samples showed no peak after filtration, 5% DMSO samples showed a small peak, but
inconsistently.

Note that in-depth discussion of causes of variance between separation methods

has already been performed in chapter 4, and will not be repeated here. Only the
results of 1 and 5% DMSO samples for any given method will be compared. All of
the results obtained for 1% DMSO samples are also given in chapter 4; where
possible, ratios or percentages will be presented to avoid repetition. 5% absolute
values are available in the appendix.

All compounds show significantly higher 5% values compared to 1% values,

regardless of the separation method used. Standard deviations for each sample set

124
are small, and experimental error does not account for the difference (full results
with standard deviations are detailed in the appendix). As mentioned in chapter 4,
all 1% results are in excess of the solubility from solid values. Since the results in
section 5.3.1 showed no thermodynamic increase in solubility for tolnaftate and
clotrimazole when moving to a 5% DMSO solvent system, the effect must be kinetic.
For samples where centrifugation was employed as the solid/liquid separation
method, tolnaftate shows the largest difference between 1 and 5% samples. The
samples containing 5% DMSO have concentrations of tolnaftate measured in the
supernatant that are more than double that measured from 1% DMSO samples.
Filtration results for this drug with a 5% sample composition were irreproducible;
sometimes a small peak was detected, sometimes not.

Investigating the nanoparticle/liquid separation method employed for the 5%

DMSO samples using NTA gives the following results (Table IV).

Table IV: NTA results of samples containing 5% DMSO, after employment of centrifugation or
filtration

5 mins 9,300 x g cent 0.45 um PVDF filter

Particle Particle
Calculated Calculated
Diameter concentration Diameter concentration
Compound nanomoles nanomoles
(nm) (106 (nm) (106
of drug of drug
particles/mL) particles/mL)

Amiodarone HCl 116 ± 3 22 ± 1 6±1 110 ± 3 23 ± 8 7±3

Clotrimazole 212 ± 14 6±2 14 ± 8 < 100 tracks*

particles detected inconsistently, sometimes

Tolnaftate 197 ± 18 9±7 15 ± 9
<100 tracks*

*<100 tracks – low track numbers mean inaccurate mean diameter and very low particle
concentrations

125
From Table IV, we can see that, as for the 1% DMSO samples, centrifugation does
not remove all of the particles formed. Filtration has more success in performing
this, and, as for 1% DMSO samples, amiodarone HCl requires the use of a 0.22 μM
PVDF filter for removal of the majority of particles. Whilst similar trends are
therefore seen between the 1 and 5% DMSO samples, what should be highlighted
are the calculated nanomoles of drug present in the form of nanoparticles left after
centrifugation (Table V);

Table V: Calculated μmoles of drug per L suspension in the form of nanoparticles after
centrifugation for 1 and 5% DMSO samples, based on NTA data

Calculated μM of drug after centrifugation

Compound 1% DMSO 5% DMSO

Amiodarone HCl 3 ± 0.3 (n=3) 6 ± 1 (n=3)

Clotrimazole unable to quantify (n=4) 14 ± 8 (n=4)

Tolnaftate 14 ± 6 (n=6) 15 ± 9 (n=4)

Tolnaftate 1 and 5% DMSO samples show similar amounts of drug in remaining in

the supernatant after centrifugation in the form of nanoparticles from NTA, but
different supernatant drug concentrations after HPLC analysis. The 5% DMSO
samples had increased drug amounts detected in the supernatant by HPLC
compared to the 1% DMSO samples – the NTA results shown in table V suggest that
this cannot be attributed to increased concentrations of drug present as
nanoparticles. No ‘true’ supersaturation was seen for tolnaftate 1% DMSO samples
(chapter 4) but here the results suggest some ‘true’ supersaturation in the 5%
DMSO samples. Amiodarone HCl and clotrimazole 5% DMSO samples have
increased calculated nanomoles of drug present as nanoparticles in the supernatant
after centrifugation compared to 1% DMSO samples.

126
The use of filters (0.45 or 0.22 μm PVDF, depending on the compound) has been
shown by NTA to completely remove the drug nanoparticles, but HPLC analysis still
shows an increase in the concentration of drug measured in the filtrate drug
amiodarone HCl and clotrimazole 5% DMSO samples, again leading to the
conclusion that 5% DMSO samples have an increased ability to remain
supersaturated compared to 1% samples. This effect is in addition to having
differences in the amount of particles formed. For tolnaftate, 5% DMSO HPLC
filtration results were sporadic; sometimes a peak was detected in the filtrate,
sometimes not. This was also reflected in the NTA results, where a fairly large range
of nanoparticle concentrations were seen.

127
 5.3.3 1 and 5% DMSO and incubation times

Since it seemed likely that the effect of 5% DMSO was one of a kinetic nature,
experiments involving various incubation times were performed. Both HPLC and
NTA were employed to analyse changes in the samples as incubation times were
varied.

Table VI shows the difference in measured HPLC supernatant amounts for 1 and 5%
DMSO samples after 24 hours incubation. These are presented as a % of the
solubility from solid value in Table I. As before, full details of the 1% incubation
results are available in chapter 4; full details for 5% results are available in the
appendix.

Table VI: 1 and 5% DMSO samples with varying incubation times, precipitation method 2 followed
by centrifugation and HPLC analysis of the supernatant

Kinetic solubility after 24 hr as % of ‘shake-flask’

solubility

Amiodarone HCl Clotrimazole Tolnaftate

1% 1263% Close to equil. Close to equil.

5% 1000% 374% 1088%

As can be seen from the table, all of the 5% DMSO samples are still well in excess of
the ’shake-flask’ solubility value, even after 24 hour incubation. Two of the three 1%
DMSO samples have reached the equilibrium value – the difference in kinetics
between the three compounds has already been discussed in the previous chapter.
What is key to note here is the difference between the 1 and 5% DMSO samples.
For amiodarone HCl, whilst the percentage detailed in table VI is slightly higher for
the 1% DMSO sample, the absolute value for the 5% sample is higher (see appendix
for full details of 5% DMSO samples). This is due to the slight difference in ‘shake-
flask’ solubility values between amiodarone samples containing 1 and 5% DMSO

128
solvent systems. All 5% samples show increased drug concentrations present in the
supernatant by HPLC analysis compared to the 1% DMSO samples, even after 24
hours. Although the ‘shake-flask’ solubility experiments performed for these
compounds showed that the % DMSO does not cause any thermodynamic increase
in solubility for 2 of the three compounds (and only a modest increase for the 3rd),
5% DMSO precipitation method 2 samples still show increased concentrations of
drug detected by HPLC in the supernatant after 24 hours. This suggests the higher %
DMSO slows the kinetics of the precipitation down dramatically. Investigating this
further via NTA (Table VII);

Table VII: NTA, comparison of 1 and 5% DMSO results showing evidence of slower particle
formation kinetics for 5% DMSO samples

*Initial rate of Mean particle

Initial particle
increase in mean diameter at the Initial calculated
concentration (106
nanoparticle end of fast initial nanomoles (no
particles/mL) (no
diameter particle growth centrifugation)
centrifugation)
(nm/min) period* (nm)

1% 5% 1% 5% 1% 5% 1% 5%
Compound
DMSO DMSO DMSO DMSO DMSO DMSO DMSO DMSO

Amiodarone 270 ± 19.8 ± 9.3 ±

2 2 237 ± 8 6±1 2±0
HCl 38 2.5 1.9

450 ± 9.9 ± 5.2 ±

Clotrimazole 1 0.3 278 ± 8 26 ± 8 18 ± 4
24 2.8 1.4

635 ± 307 ± 19.1 ± 7.4 ±

Tolnaftate 7 2 36 ± 16 15 ± 6
26 18 3.0 4.0

* initial rate of mean diameter increase taken as the initial slope in the graph of mean particle
diameter against time; for tolnaftate and amiodarone HCl, the fast initial rate increase occurred
between t0 and t 60. For clotrimazole it occurred between t 0 and t 180. Note this is a rough
indication of the average particle diameter increase rate to give a comparable variable between
compounds. Mean particle diameter results are for the end of this fast initial growth period,

129
indicating differences in growth over time. Particle concentration results are those obtained initially,
and highlight immediate differences between the samples.

For all three compounds investigated, precipitation from a 5% DMSO sample system
results in;

1. Lower initial particle concentrations

2. Lower calculated nanomoles of drug present as nanoparticles (initially)
3. Smaller particle diameters at the end of the rapid particle diameter growth
period
compared to precipitation from a 1% DMSO sample system at the same total
amount of drug added. For two of the three compounds, the initial rate of
nanoparticle mean diameter increase (Table VII, column one) is lower for the 5%
DMSO samples. Clotrimazole shows a slight lowering; the effect for tolnaftate is
dramatic due to the rapid rate at which the 1% nanoparticles grow. This is also
highlighted in the second column of table VII. The % DMSO present in the samples
therefore has the greatest effect on the growth of tolnaftate nanoparticles. The rate
of mean diameter increase for amiodarone HCl samples is unaffected by the
percentage of DMSO present in the precipitation experiment – the difference in
initial supersaturation could be a reason for this. NTA measured nanoparticle
concentrations are affected by the % DMSO present; all 5% samples show lower
initial particle concentrations. All of the initial calculated nanomoles are lower for
5% samples. Less drug is therefore present as particles at time 0 for 5%, meaning
more is present in the molecularly dissolved solution. For amiodarone HCl and
clotrimazole, some supersaturation is seen from 1% samples – increasing the DMSO
present is therefore enhancing this effect. For tolnaftate, very little true
supersaturation is seen for 1% samples – the drug appears to precipitate to
equilibrium solubility almost immediately, as discussed in chapter 4. Here, however,
the 5% samples show supersaturation, thus having a dramatic effect on the
behaviour of the compound.

130
Looking at trends in particle concentrations over time, all 1% DMSO samples show a
steady decrease at the measured time points. 5% DMSO samples, however, show
an initial increase then decrease for all three compounds (data not shown). This
again leads to the conclusion that 5% DMSO slows the precipitation kinetics of all
three compounds – compared to the 1% DMSO samples, more time is taken for the
maximum particle concentration to be reached.

131
 5.4 Conclusions

Increasing the % DMSO present in the final sample composition has, for these three
compounds, varying results depending on the method used to assess solubility. In
terms of solubility from solid, only one compound showed any statistically
significant enhancement with the increased percentage of DMSO present. This
increase was also fairly modest, around a factor of 2, but with both values
remaining below 5 μM. So with this method, overall 1 and 5% samples were similar.
Using the NTA to determine the concentration at which precipitation began also
yielded similar results from 1 and 5% samples. This method of course uses various
concentrations which are fairly close to the solubility from solid value – it was not
necessary to add in a large excess of compound in order for nanoparticle formation
to be initiated (for these three compounds). Moving to the precipitation-HPLC
based method, a large excess of compound is added, so that the amount of drug
present in the system is way above that obtained from the solubility from solid
experiments. It would therefore be expected that rapid precipitation to equilibrium
would occur. This is not the case – whilst initial nanoparticle formation is rapid, the
samples still have dissolved drug in excess of the equilibrium solubility value. Thus,
subsequent growth/ripening/ further formation of the particle is the slow, rate
determining step, rather than initial nucleation. In samples containing 5% DMSO,
this effect is exacerbated - nanoparticle formation and growth is slowed. Although
particles are present from time 0, there are fewer compared to the 1% DMSO
samples. Samples with 5% DMSO are therefore supersaturated for longer compared
to their counterpart 1% samples.

132
 5.5 References
1. B. Hoelke, S. Gieringer, M. Arlt and C. Saal, Analytical Chemistry, 2009, 81,
3165-3172.
2. L. Di and E. H. Kerns, Drug Discovery Today, 2006, 11, 446-451.
3. B. Bard, S. Martel and P.-A. Carrupt, European Journal of Pharmaceutical
Sciences, 2008, 33, 230-240.
4. C. Teng-Man, S. Hong and Z. Chegnyue, Combinatorial Chemistry and High
Throughout Screening, 2002, vol. 5, pp. 575-581.
5. S. Martel, E. Castella, F. Bajot, G. Ottaviani, B. Bard, Y. Henchoz, B. G.
Valloton, M. Reist and P.-A. Carrupt, Chimia, 2005, 59, 308-314.

133
 6. Non-Stoichiometric, aggregate based
inhibition: Bioassay trends explained via
nanoparticle growth measurements

Cheska Gillespiea*, Pierre Daublainb and Peter Hallinga

a
WestCHEM, The Department of Pure and Applied Chemistry, University of
Strathclyde, Thomas Graham Building, 295 Cathedral Street, Glasgow, G1 1XL

b
Merck Research Laboratories, 33 Avenue Louis Pasteur, Boston, MA 02115

*correspondence to [email protected]

Keywords – Non-stoichiometric aggregate based inhibition, bioassays, HTS, particle

growth

6.1 Abstract

Non-stoichiometric, aggregate based inhibition has been a much studied

phenomenon in recent years. Most of the work has focused on enzyme inhibition by
the aggregator/nanoparticle former. Here, we look at the effects of varying
amounts of dissolved protein on the nanoparticle precipitation of 3 model
compounds (amiodarone HCl, clotrimazole and tolnaftate). The work demonstrates
how protein content changes both the initial precipitation and further growth
kinetics of the system. Using tolnaftate, we also show that the enzyme-nanoparticle
interaction (hypothesized to be a surface adsorption type interaction) occurs to a
greater extent when the protein is present whilst the particles are forming. These
results are of importance to HTS and interpretation of discovery bioassay results.

134
 6.1 Introduction

Recently, several publications have explored the phenomenon of non-

stoichiometric, aggregate based inhibition as a source of false positive results in
high throughput bioassay screening. 1-5 The aggregates interact with the target in a
non-specific, non-stoichiometric manner, resulting in inhibition despite the lack of
specific, drug like interaction.2 Whilst in depth analysis of how these aggregates can
affect enzyme inhibition and the results of activity screens has been performed
already,1-7 little is known about how the aggregates themselves behave under test
conditions.8 Here, we demonstrate that the dissolved protein content present in the
assay medium significantly hinders the nanoparticle growth kinetics of 3 model
compounds, amiodarone HCl (ami), clotrimazole (clot) and tolnaftate (tol). All drugs
have low aqueous solubility (<10 μM), and form particles in the nanometre size
range upon mixing DMSO stock solutions of the drugs with aqueous buffers.
Clotrimazole has been shown previously to be a non-stoichiometric, aggregate
based inhibitor.2 Whilst the inhibition effects of tolnaftate and amiodarone HCl are
unknown, they form particles within the size range typically seen for non-
stoichiometric, aggregate based inhibitors. Over time, clear differences were seen
between samples precipitated in the presence of dissolved protein, and those
precipitated in buffer only. Using tolnaftate, it was also found that addition of
enzyme to a nanoparticle suspension after precipitation had already been initiated
by an initial aliquot of buffer alone resulted in no difference in kinetics of particle
growth. This suggests the enzyme molecules must be present during the
aggregate/nanoparticle formation process in order to adsorb to the nanoparticle
surface. All of the factors studied here are likely to be variable in HTS bioassay
protocols. The results are therefore directly relevant to this area of drug discovery.

135
 6.2 Experimental
6.2.1 Material s and Methods

Amiodarone HCl (A-8423), clotrimazole (C-6019) and tolnaftate (T-6638) were

purchased from Sigma-Aldrich. Compound purity confirmed by NMR, and either
matched or exceeded the manufacturer’s specifications. (≥ 98%).

DMSO used was UV-Vis grade, purchased from Fluka.

KH2PO4 and Na2HPO4.2H2O were purchased from Fisher Scientific.

Citric Acid was purchased from Fisher Scientific.

α-chymotrypsin (type II, bovine pancreatic, lyophilised) was purchased from Sigma
(C-4129), protein content ≥ 85% (value obtained from supplier).

6.2.2 Stock and Buffer preparations

A 5 mM weighed stock of either amiodarone, clotrimazole or tolnaftate in 10 mL

DMSO was prepared.

A pH 7.0 (± 0.05) phosphate buffer was prepared, using 0.0025 moles of KHPO 4 and
0.0025 moles of Na2HPO4 per litre of deionised, Milli-Q water. The pH adjustment
was conducted with 2 M NaOH if required. The buffer was filtered using a 0.2 μm
hydrophilic PTFE filter (Millipore), and then used to prepared a 1 mg/mL solution of
α-chymotrypsin. Aliquots of this solution were then diluted to give 0.5, 0.1, 0.05 and
0.01 mg/mL solutions of α-chymotrypsin in filtered pH 7 buffer.

A pH 4.0 citrate-phosphate buffer was also used for amiodarone HCl samples. This
was prepared by mixing 147 mL of a 0.1 M citric acid solution and 103 mL of 0.2 M
sodium diphosphate solution and diluting to 500 mL with Milli-Q water. The buffer
was adjusted to pH 4.0 (±0.05) using citric acid. The buffer was filtered and then
used to prepared a 1 mg/mL solution of α-chymotrypsin. This was then diluted to
give a 0.1 mg/mL solution of α-chymotrypsin in pH 4 filtered buffer.

136
 6.2.3 Sample preparation

i. Compound precipitation initiated using buffer containing dissolved protein

10 μL of the 5 mM DMSO stock solution was pipetted into an HPLC vial. 990 μL of
filtered pH 7 buffer containing various concentrations (0 to 1 mg/mL) of α-
chymotrypsin was added to the vial. This gave a total of 50 nanomoles of drug in 1
mL of DMSO/buffer/protein mixture, with 1% DMSO by total volume. No further
mixing was performed. The resulting suspension was transferred to the NanoSight
chamber, either immediately, or after a 1 hour incubation period. Both precipitation
and incubation steps were performed at 25 °C. For each time point and each α-
chymotrypsin content, 3 replicate samples were prepared and analysed.

ii. Compound precipitation initiated using buffer only, with secondary dilution to
introduce dissolved protein

10 μL of the tolnaftate 5 mM DMSO stock solution was pipetted into an HPLC vial.
990 μL of filtered pH 7 buffer was added to the vial, giving 50 nanomoles of
tolnaftate per mL of suspension. After twenty seconds, the suspension was diluted
using 0.5 mL of filtered pH 7 buffer, containing either no protein or 1 mg/mL
protein, to give a total sample volume of 1.5 mL. After dilution, the total drug
content was 33 nanomoles per mL and the total protein content was either 0
mg/mL or 0.33 mg/mL. After dilution, the suspensions were analysed using the
NanoSight either immediately, or after one hour incubation. As before, precipitation
and incubation were performed at 25 °C. For each time point and each α-
chymotrypsin content, samples were prepared in triplicate.

137
 6.3.4 Instrument Details

The NanoSight LM10 with NTA software version 2.1 was used in all experiments.

Video recording settings - Fixed camera settings were employed for each compound
(table 1), to allow fair comparison of particle concentration estimate between
samples. The NanoSight chamber was set to 25 °C. 90 second videos were recorded
for all samples.

Table I: NanoSight camera settings used

Compound Shutter Gain

Amiodarone HCl pH
100 0
7

Amiodarone HCl pH
1500 500
4

Clotrimazole 100 0

Tolnaftate 66 0

Processing settings – All processing settings were set to automatic, with blur
increased to ‘5x5’. During processing, ‘Ctrl +0’ were pressed to obtain a non-
smoothed particle size distribution.

Blank -A blank of 1% DMSO and 99% buffer containing 1 mg/mL chymotrypsin

showed no detectable particles at the camera settings used for sample analysis.

Data handling –Number mean diameters and estimate of particle concentration

obtained for three replicate samples were averaged and standard deviations
obtained. The particle size distributions were binned (20 nm), and the distributions
of 3 replicate samples averaged. If error bars are shown, they represent the
standard deviation of the 3 replicate samples.

138
 6.3 Results and Discussion

Table II shows particle sizes found immediately after addition of the aqueous buffer
to the DMSO stock, and after 60 min further incubation.

Table II: Average particle diameter for samples containing 0 and 0.1mg/mL α-chymotrypsin

Number Mean Diameter (nm) average of 3 reps ± S.D

0 mg/mL protein, 0 mg/mL protein 0.1 mg/mL 0.1 mg/mL

Compound
t0 t60 protein, t0 protein, t60

Ami 158 ± 26 524 ± 32 168 ± 33 210 ± 23.9

Clot 190 ± 4 367 ± 3 221 ± 15 227 ± 4

Tol 186 ± 19 635 ± 26 200 ± 12 225 ± 17

From Table II, regardless of the amount of protein added, the particles formed at
time zero are very similar in diameter. However, the particles formed by all three
compounds grow substantially over the incubation period in the absence of protein,
but in the presence of 0.1mg/mL protein, growth is inhibited and only very
increases in diameter are obtained. Table III shows the results of NTA concentration
analysis.

Table III: Average particle concentrations for samples containing 0 and 0.1 mg/mL protein

Particle concentration (108 particles/mL) average of 3 reps ± S.D.

0 mg/mL 0 mg /mL 0.1 mg/mL 0.1 mg/mL

Compound
protein, t0 protein, t60 protein, t0 protein, t60

Ami 25 ± 4 5 ± 0.1 10 ± 8 15 ± 1

Clot 14 ± 2 2±1 1 ± 0.4 7±1

Tol 19 ± 3 3 ± 0.4 7±2 18 ± 4

139
From table III, the particles formed by all three compounds grow substantially over
the incubation period in the absence of protein, but in the presence of 0.1mg/mL
protein, growth is inhibited. Samples containing no protein show a decrease in
concentration over time as well as growth, whereas samples containing protein
show little to no growth over time and an increase in particle concentration from
time 0 to 60 mins (particle concentration data in appendix). A potential explanation
is that the protein ‘forces’ the compound to form more particles in order to
‘complete’ the precipitation process. This would allow the samples to tend towards
an equilibrium solubility value by forming more particles, rather undergoing
ripening/growth of the particles already formed (as seems to be the case without
protein present). This correlates well with the enzyme molecules adsorbing to the
surface and in essence ‘blocking’ particle growth/ripening. An interesting effect was
noted with amiodarone HCl – at pH4, no effect on particle concentration or size was
seen from the presence of 0.1 mg/mL protein (data available in appendix). A
possible explanation is that the compound, being a weak base, is more positively
charged at pH 4 compared to 7 (pKa 10.29) and thus has even greater repulsion
towards the also positively charged chymotrypsin (pI 8.67). However, even at pH 7
the molecule would carry a substantial charge, and this may only partially explain
the effect.

In order to explore the effect of enzyme concentration on precipitation, tolnaftate

was precipitated in the presence of varying concentrations of protein (table IV
which shows average particle diameter results, figure 1 which plots particle
concentration results and figure 2 which shows particle size distribution results.).

140
Table IV: Tolnaftate and varying concentrations of chymoptrypsin, average particle diameters

Number mean diameter (nm) average of 3 reps ± S.D.

Alpha-chymotrypsin concentration,
time 0 time 60 min
mg/mL

0 186 ± 19 635 ± 26

0.01 186 ± 9 264 ± 7

0.05 185 ± 6 259 ± 5

0.1 200 ± 12 225 ± 17

0.5 191 ± 13 237 ± 3

1 195 ± 6 199 ± 31

141
 30.00

time 60 mins

Particle concentration (108 particles per mL)

25.00
time 0
20.00

15.00

10.00

5.00

0.00
0 0.2 0.4 0.6 0.8 1
alpha-chymotrypsin content (mg/mL)

Figure 1: Tolnaftate - particle concentration Vs protein content at time zero and time 60 mins

Even as little as 0.01 mg/mL present results in samples which differ dramatically
compared to 0 mg/mL protein. It can also be seen from the concentration results
that, in general, as the protein content is increased (whilst maintaining the same
amount of drug), the effect on the particle concentration formed changes. 0.01 and
1mg/mL show significantly different behaviours, both immediately and after a 60
minute incubation. Further investigation into the effect of protein concentrations
below 0.01 mg/mL could be of use, and could determine the minimum amount of
protein required for an effect.

The amount of drug present in particles can be estimated from their concentrations
and sizes:10 this indicates that at the highest protein concentration, only around
13% is particulate initially, but nearly all after 60 min. It is clear that all of the
protein concentrations explored here slows both initial nanoparticle formation, and
subsequent growth.

At time 0, all samples show monodisperse particle size distributions (see appendix)
Figure 2 shows the particle size distributions obtained after 60 min.

142
Figure 2: Average particle size distributions for tolnaftate samples containing various protein
amounts, time 60 mins, with insert of close up 350-950 nm section.

Samples with 1 mg/mL α-chymotrypsin present have remained fairly monodisperse,

with little ripening or growth seen, in stark contrast to those with no protein
present. Higher concentrations of protein induce the formation of a population of
smaller particles (~100 nm) after 60 minutes– these are not present at time 0 and
likely form after the initial nanoparticle precipitation.

All 3 compounds explored here are form nanoparticles in the size range typically
seen from non-stoichiometric, aggregation-based inhibitors. The effects seen here
with regards to the presence of protein during precipitation initiation help explain
some of the trends seen in previous inhibition studies. Time dependent inhibition
was one of the distinguishing behaviours. 1-3 IC50 values typically decreased
(improved) when the inhibitor and enzyme were pre-incubated for five minutes. As
the formation of nanoparticles is a kinetic process, there is likely a change in particle
concentration/size between zero and five minutes. If more particles are formed

143
after 5 minutes compared to at time 0, more surface is available for enzyme
attachment, increasing non-stoichiometric inhibition and decreasing IC50. Enzyme
concentration is also known to influence non-stoichiometric inhibition. 1-3, 11 A ten-
fold increase in enzyme concentration was observed to significantly increase
(worsen) the IC50 value, and such an effect has been suggested as a means of
detecting non-stoichiometric inhibition.11 From the results presented earlier, an
increase in protein concentration inhibits initial particle formation, so there are less
particles available to interact with and inhibit the enzyme. Note however that
particle formation is not prevented completely but merely slowed – so the extent of
inhibition may be time and compound dependent. Similar effects may help explain
the observation that excess protein (0.1 mg/mL BSA,1 lysozyme and trypsin7)
decreased inhibition by the nanoparticles. Finally, demonstration of the effects of
protein on the nanoparticles is entirely consistent with the view that protein
molecules interact with the particle surfaces. Note this interaction has some
features in common with protein coated microcrystals (PCMCs). PCMC formation is
a method of enzyme immobilization by which proteins are coated onto the surface
of a micron-sized ‘carrier’ crystal during crystal formation. 12 Coan et al have
previously shown that aggregates are not disrupted by high protein concentrations
in the same way as they are with surfactants. The results shown here are consistent
with this - even at protein concentrations as high as 1 mg/mL, nanoparticles are still
present.

With the aim of further investigating the mechanism of the effects, protein was
added to the system after precipitation had been initiated. Buffer was added to the
DMSO stock solution of tolnaftate as before, initiating the precipitation process, but
after 20 s, an additional 0.5 mL of filtered pH 7 buffer was added, containing either
0 or 1 mg/mL protein. Dilution alone reduces the subsequent growth of the
nanoparticles (table V).

144
Table V: Diameter and concentration results, protein added after precipitation of tolnaftate

8
Particle Concentration (10
Number mean diameter (nm -
particles/mL - average of 3 reps ±
average of 3 reps ± S.D.)
S.D.)

Sample diluted with; Time 0 Time 60 Time 0 Time 60

Buffer only 206 ± 16 297 ± 43 5.0 ± 0.8 10.0 ± 2.8

1 mg/mL chymotrypsin 204 ± 10 260 ± 11 3.4 ± 1.6 10.8 ± 0.8

Dilution had a large effect on particle growth for the control (no protein) samples
(table V), and has slowed the process. Both the total amount of drug added and the
incubation volume have changed here compared to previous samples. However,
now there is little effect of protein, either immediately after its addition or upon
further incubation. It could be expected that this would slow the process even
further if the samples behaved as described previously. The data here, however,
shows no synergistic or even additive effect– samples diluted with buffer containing
protein now behave exactly as the control samples do. The particle size distributions
confirm the similar behaviour.

145
 140

Diluted with buffer only, 1 hr inc

particle concentration (106 particles /mL)

120
Diluted with chymotrypsin dissolved
100 in buffer, 1 hr inc
Diluted with buffer only
80
Diluted with chymotrypsin dissolved
in buffer
60

40

20

0
0 100 200 300 400 500 600 700 800 900 1000
Particle Diameter (nm)

Figure 3: Particle size distributions, protein added after precipitation

The findings could also be of immediate practical relevance to screening. It should

be relatively easy to adopt a two-stage mixing procedure i.e. addition of an initial
aqueous portion (free from protein) to the DMSO stock, to begin compound
precipitation, then subsequent addition of a further aqueous portion containing the
enzyme, to begin the activity assessment. This would allow initial precipitation to
occur before the enzyme is present. Of course, in an ideal world compound
precipitation would not occur, but realistically it is a likely event for a large number
of screening compounds due to their hydrophobicity. This approach, however,
could potentially reduce the interaction of the enzyme and aggregate, thus reducing
the number of false positive results, with only minor changes to the dilution
procedure. Obviously more work is needed, but this idea may warrant some further
investigation by those involved in bioassay screening.

146
 6.4 Conclusions

In conclusion, we show that protein significantly affects the precipitation kinetics of

3 model compounds, with 0.1 mg/mL of α-chymotrypsin slowing both formation
and ripening of drug nanoparticles. As little as 0.01 mg/mL has an effect. The
slowing effect of protein on compound nanoparticle precipitation kinetics may
explain some effects noted in studies of non-stoichiometric aggregate-based
enzyme inhibition. These effects (sensitivity to enzyme concentration and time
dependent inhibition) correlate well with the results seen here, in that the size and
concentration of compound nanoparticles formed are dependent on both the
amount of protein present and time. Interactions appear reduced if protein is
added after precipitation is initiated, so this might be a valuable strategy in
screening, although dilution alone alters the kinetics as well. Further investigation
to see if the results obtained here translate into a practical recommendation would
be of use. Better understanding of non-stoichiometric inhibition requires study of
effects on particle precipitation itself, and extension of this work to other
compounds and proteins.

ACKNOWLEDGMENTS

With thanks to MSD and University of Strathclyde for funding contributions

147
 6.5 References

1. S. L. McGovern, E. Caselli, N. Grigorieff and B. K. Shoichet, Journal of

Medicinal Chemistry, 2002, 45, 1712-1722.
2. J. Seidler, S. L. McGovern, T. N. Doman and B. K. Shoichet, Journal of
Medicinal Chemistry, 2003, 46, 4477-4486.
3. S. L. McGovern and B. K. Shoichet, Journal of Medicinal Chemistry, 2003, 46,
1478.
4. S. L. McGovern, B. T. Helfand, B. Feng and B. K. Shoichet, Journal of
Medicinal Chemistry, 2003, 46, 8.
5. K. E. D. Coan and B. K. Shoichet, Journal of the American Chemical Society,
2008, 130, 9606-9612.
6. B. Y. Feng, A. Shelat, T. N. Doman, R. K. Guy and B. K. Shoichet, Nature
Chemical Biology, 2005, 1, 146-148.
7. K. E. D. Coan and B. K. Shoichet, Molecular BioSystems, 2007, 3, 208-213.
8. K. E. D. Coan, J. Ottl and M. Klumpp, Expert Opinion on Drug Discovery, 2011,
6, 405-417.
9. A. Avdeef, Absoprtion and drug development; Solubility, permeability and
charge state, John Wiley and Sons Inc., 2012.
10. C. Gillespie, P. Halling and D. Edwards, Colloids and Surfaces A:
Physicochemical and Engineering Aspects, 2011, 384, 233-239.
11. M. Habig, A. Blechschmidt, S. Dressler, B. Hess, V. Patel, A. Billich, C.
Ostermeier, D. Beer and M. Klumpp, Journal of Biomolecular Screening,
2009, 14, 679-689.
12. M. Kreiner, B. D. Moore and M. C. Parker, Chemical Communications, 2001,
1096-1097.

148
 7. Attempts at nanoparticulate crystalline form
determination
7.1 Introduction

The crystalline form of a substance is directly related to physical properties such as

solubility, melting point and dissolution rate. 1 Knowledge of the crystal form of a
compound is therefore very important within the pharmaceutical industry.

Solids can exist in a variety of different crystalline forms – these are known as
polymorphs, and differ in the arrangement of molecules within the crystal lattice.
Solids can also be amorphous in character, show no long range order in the
molecular arrangement within the solid, and thus do not exhibit Bragg diffraction.
Compared to the most stable crystalline state, amorphous materials are a higher
energy solid form, and tend to be more soluble than crystalline material.2 Of course,
as amorphous material is not the most stable solid form of the drug, conversion to
the crystalline form occurs over time. This is one of the difficulties in fully assessing
the solubility advantage for amorphous solids. 2

With regards to compound precipitation from mixing DMSO stock with aqueous
buffers, little published work is available on the solid form of the precipitate
formed. Hoelke et al3 used x-ray diffraction (XRD) to analyse the solid formed after
precipitation from DMSO and buffer mixtures. Both model compounds investigated
precipitated in an amorphous form. They conclude that the addition of DMSO stock
solutions to aqueous buffers results in conditions which are too harsh for
crystallisation, and thus fast precipitation of amorphous solid occurs. However,
Sugano et al used polarised light microscopy (PLM) to investigate the crystallinity of
26 structurally diverse, low solubility drug molecules after precipitation from DMSO
and buffer solutions. They found that 58% of their test compounds precipitated as
crystalline after a 10 minute incubation period, and that this increased to 73% after
20 hours incubation. The compounds which precipitated as amorphous solids had
solubility values in excess of those obtained from the solid. The technique (PLM)

149
was fully automated and so useful in solubility screening. The total drug
concentration used here was however, fairly high in terms of screening (0.3 mM).4

Both theory and some limited results from the literature show that the solid form of
the precipitate can affect the amount of drug left dissolved in solution. This would
therefore affect the results of both an HTS activity screen and kinetic solubility
experiments. All three of the drug molecules investigated in this thesis form
nanoparticles upon precipitation from DMSO and buffer solutions. These particles
then grow and ripen over time. This chapter documents one of the most challenging
aspects of this work – attempts at establishing the crystalline form of the
nanoparticulate precipitate. In order to do so, one compound (tolnaftate) was
picked to trial various approaches. Due to the nature of the samples, the low
concentrations used here (μM) and the small size of the initial precipitate, analysis
was not straightforward. Several different techniques and approaches were
considered, with only a few modest successes. Raman analysis of tolnaftate gave no
conclusive results with regards to crystallinity, however could obtain a signal from a
1 mL scale sample. Scale up and subsequent XRPD analysis showed that the
tolnaftate precipitate, after a given stir time and growth to larger sizes, was
crystalline and matched that of the starting material. None of the methods
investigated were found to be suitable for determining the crystalline form of the
nanoparticles themselves. The work is, therefore, deemed incomplete and as such is
not written with a view for publication.

150
 7.2 Experimental
7.2.1 Materials

Tolnaftate (T-6638) was purchased from Sigma-Aldrich. Compound purity confirmed

by NMR (>99%).

The DMSO used was UV-Vis grade, purchased from Fluka.

KH2PO4 and Na2HPO4.2H2O were purchased from Fisher Scientific.

7.2.2 DMSO stock, buffer and drug suspension preparations

Weighed stock solutions of nominal concentrations of 1 mM and 5 mM were

prepared in DMSO for each individual compound (actual stock solution
concentrations ranged from 0.99 to 1.03 mM, and 4.97 to 5.08 mM). These
solutions will be referred to by their nominal concentrations. For Raman analysis,
stocks in acetone were prepared in a similar manner.

A pH 7.0 (± 0.05) phosphate buffer was prepared, using 0.0025 moles of KHPO 4 and
0.0025 moles of Na2HPO4 per litre of deionised, Milli-Q water. The pH adjustment
was conducted with 2 M NaOH if required. For this analysis, the buffer was then
filtered through a 0.22 μm hydrophilic PTFE (Millipore) or a 0.1 μm VVLP filter
(Millipore).

For a 1% DMSO (by volume) 1 mL suspension of drug concentration 50 μM, 10 μL of

the 5 mM tolnaftate DMSO was pipetted into a microcentrifuge tube. 990 μL of
filtered pH buffer was then added. The next step in sample preparation was
dependent on the analysis technique. For samples (1 mL volume) analysed using the
Raman microscope, a strip of aluminium foil was placed inside the microcentrifuge
tube prior to addition of the solutions. After addition of the buffer aliquot, the
samples were centrifuged at 9,300 x g rcf for 5 minutes. The tubes were placed

151
inside the centrifuge such that the precipitate was centrifuged onto the piece of
film. This was then analysed under the Raman microscope. For scale up samples for
XRPD, the appropriate larger volumes of DMSO stock and buffer were used. The
scale up samples were stirred on either a Fisher FB15001 or a Stuart CC 162 stir
plate. After the appropriate stir time, the samples were filtered through a 0.1 μm
VVLP filter, and the dried powder offered to analysis.

7.2.3 Instrumental Details

i. Raman
Raman Microscopy (solid and precipitated onto foil samples)

The analysis was performed using a Leica DM/LM microscope equipped with a Leica
50 x/N.A. 0.75 objective lens. A Renishaw inVia with a 514.5 nm Ar+ Modu laser was
used as the excitation laser source. Spectra were acquired for 5 or 120 s depending
on the sample.

Raman probe (liquid samples - tolnaftate in DMSO and acetone, solid samples –
tolnaftate starting material)

Avalon Instruments Ramanstation R3, 532 nm excitation, with a fibreoptic probe.

ii. NTA and NanoSight

I mL aliquots of the scale up 500 mL samples were removed during stirring and,
prior to final filtration of the sample, analysed at various time points by NTA. The
camera settings used were ‘autosettings’. All processing settings were set to auto
apart from ‘blur’ which was increased to 5x5.

iii. XRPD
A small quantity (10-50 mg) of the sample was analyzed using transmission foil
XRPD data collected on a Bruker AXS D8-Advance transmission diffractometer
equipped with θ/θ geometry, primary monochromated radiation (Cu Kα λ=
1.54056Å), a Braun 1D position sensitive detector (PSD) and an automated multi-
position x-y sample stage. The sample was mounted on a 28 position sample plate

152
supported on a polyimide (Kapton, 7.5 μm thickness) film. Data was collected from
the sample in the range 4-35o 2θ with a 0.015o 2θ step size and 1s step-1 count time.
Count time was increased to 5s step-1 if very small (less than 10 mg) amounts of
sample were analysed.

153
 7.3 Results and Discussion

7.3.1 1 mL scale analysis

It became apparent whilst performing scoping experiments that the analysis was
not going to be straightforward. Several factors contributed to this. The
concentrations used here are very low – in a 1 mL sample, there is roughly
micrograms of solid material precipitating. This is, for many techniques, way below
the limit of detection. Since we were particularly interested in how the process
occurs at these low concentrations, increasing sample concentration was not an
option. Table I documents the various analytical techniques considered for
determination of the crystallinity of the drug nanoparticles formed for a typical 1
mL scale sample.

154
Table I: Analytical techniques considered for form analysis

Analytical Technique Pros Cons Attempted? Outcome

Gives definitive answer on Could not separate nanoparticles

XRD crystallinity and polymorphic Requires mg of sample Yes from solution in large enough
form. quantities for analysis on a mL scale.

Can analyse sample in mother Advised that sample conc. is too low,
Environmental-SEM Only gives morphology. No
liquor, no need for separation. and water to drug ratio too high.

Works in nm size range, can get

Atomic Force Microscopy down to atomic level Only gives surface topography. Yes No signal from tolnaftate sample.
resolution.

Gives information on glass

Differential Scanning Advised by expert it would be unlikely
transitions/phase Requires roughly mg of sample. No
Calorimetry that a signal would be obtained.
transformations.

Small amounts difficult to analyse, cannot

Can distinguish amorphous
PLM see individual nanoparticles in Yes No nanoparticulate results obtained.
from crystalline.
conventional light microscope.

Can potentially distinguish Requires fairly large difference in spectra

Some data obtained, however not
crystalline and amorphous between molecules in a solution
Raman Yes conclusive (further details given
material due to shifts in environment and those in crystalline
later).
spectra. lattice.

155
Both the small scale and low concentrations used contribute to the difficulty in
finding a suitable analytical technique, as for many of the instruments, the amount
of compound present in these samples is below the detection limit. An obvious
solution to this problem would be to retain the concentration, but increase sample
volume, to give an increase in drug – this will be discussed in detail in section 7.3.3
and presented its own difficulties.

7.3.2 1 mL scale analysis - Raman

An approach which had some (modest) success was the use of Raman spectroscopy.
The application of Raman to investigate polymorphic forms of various drug
molecules is well documented in literature.5, 6, 7 The idea behind the use of Raman
here was to firstly obtain reference samples of tolnaftate completely dissolved in
solvent (adjusting for solvent scattering) and samples of the solid, crystalline
starting material. It would then be possible to determine whether the precipitated
material spectra matched that obtained from the starting material (crystalline) or
was closer to that obtained from the dissolved drug reference spectra
(representative of an amorphous sample). 1 mL scale samples of tolnaftate were
used, and the initial precipitate was analysed. The nanoparticles were centrifuged
onto a piece of metal foil, which concentrated the particles onto the surface and
allowed a signal to be obtained using Raman microscopy. Removal of the mother
liquor after centrifugation ensured the particles were kept as close to their original
precipitating form as possible (i.e. lowered the risk of solution mediated phase
transformation). The following spectra were obtained from the nanoparticulate
material and the solid starting material reference;

156
 4500
Solid tolnaftate, 50x (mag) 5 sec
4000
Precipitated tolnaftate (foil), 50x
3500
(mag), 120 secs tinfoil
3000
Main peaks visible in
Intensity (AU)

2500 both solid starting

material (red) and
precipitated material
2000

1500

1000

500

0
400 600 800 1000 1200 1400
Raman Shift (cm-1)

Figure 1: Overlay of precipitate and starting material Raman spectra. Large green peak is tinfoil
peak.

From figure 1, the position of the four major peaks (indicated on the spectrum)
from the crystalline starting material (red) match the weak peaks seen from the
precipitated material (green). Note the very intense green peak is scattering from
the foil. This demonstrates that Raman has the sensitivity to analyse the small,
microgram amounts of material produced by precipitation on a 1 mL scale. In order
to obtain crystallinity information from the results obtained here, at least one of the
four main peaks indicated on figure 1 must show a shift when comparing solution
and solid state Raman spectrum. Shown in figure 2 is an overlay of solution and
solid state Raman spectra for this compound.

157
Figure 2: Overlay of Raman spectra of solution phase and solid drug

From figure 2, we can see clearly that the solution and solid Raman spectra of the
drug were actually very similar. The four major peaks, one of which we needed to
shift in order to obtain crystallinity information, have not moved. In fact, only one
minor peak shifts very slightly between the two phases (see figure 3 for details). The
solution spectra were obtained in both DMSO and acetone, with the solvent
background spectra being subtracted from the respective sample spectra. Figure 3
shows a close up of the one minor peak which shifts slightly between the solid and
solution phase. Note this is only visible when comparing the drug dissolved in
acetone and the solid form (with drug samples dissolved in DMSO, an interfering
peak from the DMSO masks the drug peak shift).

158
Figure 3: Overlay of Raman spectra showing shift between solution (acetone) and solid tolnaftate
samples

Even with copious amount of sample, differentiation between the two states of
matter would be difficult due to the similarity of the spectra of two phases.
Therefore, using the small amounts typical of the experiments here makes it near
impossible to obtain the required crystallinity information for this compound, and
as expected, this minor peak was not visible in the precipitated sample spectrum
(figure 1). An attempt was made to monitor the precipitate as it forms in
suspension using the Raman probe, but no signal from the drug could be obtained.
Whilst the results obtained do not address the question of the crystallinity of the
nanoparticles, Raman itself does have the capability to analyse these small amounts
of material (roughly micrograms for a 1 mL scale). Different compounds may be
more successful, depending on the Raman shifts present and how different the
solution and solid spectra were.

159
 7.3.3 Scale up (500 mL) and XRPD analysis

As demonstrated, use of 1 mL volumes was problematic when attempting to

perform physical form analysis of the precipitate, as only micrograms of material
are present within a 1 mL volume. Larger scale samples at the same concentration
offer the advantage of increased amounts of precipitate for analysis, potentially
facilitating analysis. There is, however, the potential for the scale of the suspension
prepared to affect the result obtained, due to varying surface to volume ratios.
Jadhav et al noted that changing from a HTS analysis (1536 well plates) to a 96 well
plate format affected the ability of certain compounds to form aggregates. They
partly attributed this to the high surface to volume ratio for the 1536 well plate
analysis. 8 The potential difference between differing sample scales was therefore
kept in mind when performing the experiments and interpreting the results.

The samples were scaled up to a 500 mL sample volume, with a 50 μM

concentration, as shown in Figure 4.

Figure 4: Tolnaftate 500 mL suspension (scaled up sample)

The scaled up samples were immediately (visually) turbid upon addition of the
buffer to the DMSO stock solution. As with the lower volume samples, the scale up

160
samples contain nanoparticles suspended in the mixture, which grew over time.
NTA analysis was performed on the scale up samples - we can compare the results
obtained here to those detailed in chapter 4, to see the effects of scale on particle
properties.

Table II: Comparison of tolnaftate 50 μM samples at 1 and 500 mL volumes, NTA data

Mean
Mean Mean Mean calculated
Mean particle Mean particle calculated conc.
Sample particle particle 6 6 conc. of tol in
conc. (10 conc. (10 of tol in
Scale diameter diameter nanoparticulate
particles/mL) particles/mL) nanoparticulate
(mL) (nm) time (nm) time form, (μM) time
time 0 time 60 form, (μM)
0 60 60
time 0

1 185 ± 19 635 ± 26 19.1 ± 3.0 2.8 ± 0.4 36 ± 16 >50

500 255 ± 17 326 ± 24 11.4 ± 2.0 8.0 ± 2.7 46 ± 6 >50

From table II, it can be seen that the 1 mL and 500 mL samples do give slightly
different results from NTA analysis. Differences in both particle diameter and
concentration are obtained; however, the amount of drug calculated to be present
as nanoparticles is similar. The scale up samples do not seem to grow as much as
the 1 mL samples, and have higher concentrations of smaller particles after the
incubation period; however it should be borne in mind that the larger scale samples
were stirred constantly whereas the 1 mL samples were not agitated during the
incubation period. So this could be affecting the samples, rather than scale alone. It
can be said that the calculated drug as particulate content is very similar for both
sample volumes – the amount of drug which has precipitated does not appear to be
different. The scale up samples, after a given stir time, eventually lead to large,
individually visible particles. Due to the large volume used, this change could be
monitored visually as well as using the NanoSight. Smaller volumes are not afforded
this luxury. Whilst scale up meant that absolute compound quantities no longer
presented any issue, it was not without its own difficulties. Ideally, once the
particles were formed, they would be removed from the mother liquor
immediately, dried, and then analysed by XRPD or any of the other techniques.

161
However, no filter membrane was found which would both remove the nanometre
sized material from solution and allow recovery and removal of the precipitate from
the filter paper. Since the particles were in the nm range, centrifugation of larger
samples aliquots did not completely separate a sufficient quantity for analysis by
any of the techniques. Thus whilst there was plenty of nanoparticulate material in
the vial, the ratio of nanoparticle to liquid was such that the particles could not be
isolated for analysis. However, over time, as the particles grew and ripened, there
was a change from nanometre sized particles to micron sized particles. After the
samples had reached this point, it was possible to filter off the solid material, and
recover it for XRD analysis. This meant that whilst the initial aim of the scale up
(crystallinity analysis of the nanosized precipitate) could not fulfilled using this
method, analysis of the larger precipitate which developed over time was
straightforward. The dried precipitate was removed from the filter paper of four
samples and analysed using XRD. In all four cases, the samples exhibited diffraction,
indicating crystallinity, and the powder pattern obtained matched that of the
starting material. An example of the diffraction pattern overlaid with the starting
material is shown in Figure 5.

162
Figure 5: Powder XRD pattern from starting material (pink) and example scale up samples (green)

This means that, after several hours stirring, the precipitate is crystalline, and in the
most stable form.

The kinetics of particle growth (from nanosized to larger, micron (visible) sized)
showed large variation between samples, despite constant stirring. This could be
observed as demonstrated by Figures 6 and 7.

163
 Figure 6: Sample 1, 2.66 hours stirring Figure 7: Sample 2, 2.66 hours stirring

The samples were prepared simultaneously, and the images captured after nearly 3
hours stirring, however sample 1 is visibly more cloudy/turbid than sample 2. It
should be noted that while sample 2 looks ‘clearer’ than sample 1, it is still a
suspension and has particles present. However, the particle sizes present in the two
samples were different - sample 2 had larger, individually visible particles whereas
sample 1’s particles were still small (few hundred nm) at the point the image was
captured. This gave the impression the sample looked ‘clearer’. From 8 replicate
samples, the following times for a visible improvement in sample ‘clarity’ were seen
(table II);

164
Table II: Range of timescales for large particle formation

Sample Time until ‘clear’ (hours)

A 2.5

B 3.3

C 5

D 5

E 7

F 4.5

G 3.3

Note D and E were prepared simultaneously.

As can be seen from table II, there is substantial variation in the amount of time
taken for larger, visible particles to form, ranging from 2.5 to 7 hours. This large
variability in the time taken for the nanoparticles to transform into larger particles is
unusual; whilst nucleation is a random and chance event, this has already occurred
as the nanoparticles have already formed, and in large numbers. Further growth or
ripening of the particles would be expected to occur fairly consistently between
samples – this is not the case. This suggests that there may be a second energy
barrier to overcome, i.e. a second nucleation event/phase transition is required to
allow growth into the larger, crystalline particles – and since nucleation is a
stochastic process, this would go some way in explaining the variation in the data.
From a phase transition point of view, it is possible the samples could contain
varying low levels of larger, crystalline particles, which nucleate after the
appearance of the bulk nanoparticle precipitate. The number of these rare, larger
particles could poetentially govern the kinetics of the phase change –those samples
with higher numbers of these larger particles would have higher rates of
transformation, with phase change occurring more slowly in those samples with
comparatively smaller numbers of crystalline particles. Following Ostwald’s Rules, if

165
a phase change were to occur, the initial precipitate would be of a higher energy,
such as an amorphous material or a higher energy polymorph. The two-step theory
of nucleation could also be applied here; this theory states that nucleation is a two
step process, with the first stage being that of the formation of ‘clusters’, which are
highly disordered and liquid like. These clusters then rearrange to form a crystalline
nucleus, although the sizes are likely to be smaller than those of the nanoparticles
seen in this work. 9

166
 7.4 Conclusions

The form analysis of the nanoparticulate material formed by tolnaftate was not
straightforward. Several different analytical techniques were considered, and two
(Raman and scale up followed by XRPD analysis) implemented.

Raman microscopy demonstrated the capability to obtain a signal from a 1 mL

sample of nanoparticles. Information on the crystal form of the particles could not
be obtained. This was due to the extremely similar reference spectra obtained from
crystalline solid and dissolved tolnaftate solution samples. Only one minor peak
showed any shift between solid and solution samples; this small peak could not be
resolved from the baseline in the precipitated sample. This is of course an issue with
the compound rather than solely the limits of the instrumentation; other
compounds could potentially show more success. Whilst this approach did not fulfil
the initial aims of the experiment, further work on Raman with different
compounds could be worth investigating.

Scale up of the samples overcame the problem of the low absolute amount of
precipitated compound. It did, however, come with its own issues. Whilst the
samples were visually immediately cloudy and had high concentrations of
nanoparticles present, these could not be isolated from the mother liquor for
immediate analysis. Due to their small size, neither centrifugation nor filtration was
effective. It was noticed that, after several hours stir time, the particles had grown
to a size which allowed them to be filtered, dried and then analysed by XRPD. These
larger particles were the same polymorph as the solid tolnaftate starting material.
An interesting observation was that the time taken for the scale up sample particles
to grow to visible micron size particles was not consistent between samples. Even
those prepared simultaneously could look visibly different. A few possible causes
were discussed, however the underlying mechanism is not fully understood.

167
Future work could include further investigation of Raman to fully explore its
potential for analysis of other compounds, and investigation of scale up with other
compounds to see if the same issues are obtained with them as with tolnaftate.

168
 7.5 References

1. J. Mullin, Crystallisation, Butterworth-Heinemann, 2001.

2. B. C. Hancock and M. Parks, Pharmaceutical Research, 2000, 17, 397-404.
3. B. Hoelke, S. Gieringer, M. Arlt and C. Saal, Analytical Chemistry, 2009, 81,
3165-3172.
4. K. Sugano, T. Kato, K. Suzuki, K. Keiko, T. Sujaku and T. Mano, Journal of
Pharmaceutical Sciences, 2006, 95, 2115-2122.
5. R. Vehring, Applied Spectroscopy, 2005, 59, 286-292.
6. A. P. Ayala, Polymorphism in drugs investigated by low wavenumber Raman
scattering, Ceara, BRAZIL, 2006.
7. A. Docoslis, K. L. Huszarik, G. Z. Papageorgiou, D. Bikiaris, A. Stergiou and E.
Georgarakis, Aaps Journal, 2007, 9.
8. A. Jadhav, R. S. Ferreira, C. Klumpp, B. T. Mott, C. P. Austin, J. Inglese, C. J.
Thomas, D. J. Maloney, B. K. Shoichet and A. Simeonov, Journal of Medicinal
Chemistry, 2009, 53, 37-51.
9. D. Erdemir, A. Y. Lee and A. S. Myerson, Accounts of Chemical Research,
2009, 42, 621-629.

169
 8. Overall Conclusions and Future Work

The work presented in this thesis has laid foundations for further detailed work into
compound precipitation from DMSO stock and buffer mixtures. Whilst only three
compounds have been studied here, several common behaviours are seen. All three
compounds readily form nanoparticles upon mixing the stock and aqueous portion,
even at concentrations close to their equilibrium solubility (at low
supersaturations). Complete separation of the drug nanoparticulate and dissolved
drug phases for quantification is problematic due to the small size of the particles
(chapter 4). The particles grow and ripen over time, with the rate at which this
occurs depending on the compound, the % DMSO present in the samples and the
concentration of dissolved protein present (chapters 5 and 6). For one compound,
the exact way in which the two solvents are added together has no effect, but
subsequent mixing does (chapter 3). The particles formed by this particular
compound eventually grow to large, micron sized particles which have been
confirmed to be fully crystalline. No crystallinity information was obtained for any
compound on the initial nanoparticle precipitate (chapter 7).

The experiments were performed with industrial relevance in mind. The factors
investigated here are all variables in screening bioassays/kinetic solubility assays –
thus the results here have direct relevance to these areas of drug discovery.
Recommendations based on this work however differ based on the specific area in
question. While screening bioassays and kinetic solubility assays have similar
procedures and both utilise DMSO stocks, their ultimate aims are slightly different.

In a bioassay setting, it is desirable for the test compound to remain solubilised for
as long as possible in order to fully assess its activity against the target. Precipitation
on the timescale of the assay can result in both false positives and negatives; thus
avoidance of precipitation completely is ideal. In accordance with this, it would be
sensible, based on the results presented here, to recommend using higher % DMSO
and protein concentrations, which slow down particle formation kinetics. It should
however be kept in mind that these measures only have a kinetic effect and do not

170
prevent particle formation; as such analyses could still be susceptible to false
positives due to non-stoichiometric inhibition. However, compared to lower
percentages of DMSO and samples with no protein, more compound remains
dissolved, and this could be an advantage. As mentioned in chapter 6, a possible
way of eliminating false positives due to aggregation could be alteration of the
dilution procedure – though further work is needed to ascertain whether the result
seen here is mirrored in screening assay results.

The aim of a kinetic solubility experiment is to obtain an early indication of a

compounds aqueous solubility without performing laborious thermodynamic shake
flask experiments. While it is primarily used as a rough guide to solubility, the closer
the results obtained are to the compounds’ thermodynamic values, the better.
From this point of view, the more rapid the precipitation, the better. So for example
using lower percentages of DMSO and ensuring no protein is present would allow
this to happen more quickly. Due to the small size of the particles which can
potentially be formed, extra care should be given to the separation process
employed (if required). The results of kinetic solubility assays can also be used to aid
interpretation of bioassay results – ideally, for this purpose the experimental
conditions should be as similar as possible. This could prove difficult based on the
recommendations here, due to the differing aims of the experiments.

In terms of future work, detailed studies of this type on more compounds would be
beneficial in determining the prevalence of the effects in a larger samples set.
Several other factors which were only considered in scoping experiments here could
also be looked at in detail. These include the size of the precipitation container
used, the effect of varying drug concentrations, and performing the precipitation in
96/384 -well-plates. In terms of understanding non-stoichiometric aggregate based
inhibition, an investigation into the effects of different proteins and more
compounds, investigated from a more physical chemistry point of view, would be
useful. It would also be interesting to see if changing the dilution procedure

171
affected the results of an inhibition experiment in a similar manner to how it affects
the precipitation experiment seen here.

Whilst there is ample literature on various kinetic solubility analysis and on HT

bioassays, there is a need to combine the underlying principles of both to fully
understand the precipitation. It would also be beneficial to find an analytical
method suitable for determining the crystalline form of the nanoparticles – this
would give further insight into the process as a whole.

172
 9. Appendix

9.1 Chapter 1 Additional Data

Figures 1 and 2 - Amiodarone HCl starting material NMR and XRPD analysis

Figures 3 and 4 - Clotrimazole starting material NMR and XRPD analysis

Figures 5 and 6 - Tolnaftate starting material NMR and XRPD analysis

I
Figure 1: Amiodarone HCl, starting material NMR with close up insert of 0.9 to 1.9 ppm. Solvent used – CDCl3

II
Figure 2: Amiodarone HCl XRPD trace of starting material
III
Figure 3: Clotrimazole starting material NMR, with close up insert of 7.00 to 7.90 ppm. Solvent used - CDCl3

IV
Figure 4:8 Clotrimazole starting material XRPD trace

V
Figure 5: Tolnaftate starting material NMR, with close up insert of 6.90 to 8.00 ppm. Solvent used - CdCl3

VI
Figure 6: Tolnaftate starting material XRPD trace

VII
 9.2 Chapter 2 Additional Data

Table I and II – Amount of drug as particle calculation and Data, Rep 1

Table III and IV - Amount of drug as particle calculation and Data, Rep 2

Table V and VI - Amount of drug as particle calculation and Data, Rep 3

VIII
Table I: Amount of Drug in Particle Calculation, NanoSight Rep 1

Volume
Concentration
Mean Concentration Vol Fraction X
Diameter (cm) Volume of drug in % Drug
Time (mins) (adjusted (E8 ρ X 1000 No of Nanomoles
^3 Fraction particles Added
not particles/mL) (g/litre)
(moles/litre)
smoothed)

0 127 13.08 2.05E-15 1.40E-06 1.71E-03 5.58E-06 5.58 18.66

5 145 14.15 3.05E-15 2.26E-06 2.76E-03 8.98E-06 8.98 30.04

10 147 10.73 3.18E-15 1.78E-06 2.18E-03 7.10E-06 7.10 23.73

20 158 11.35 3.94E-15 2.34E-06 2.87E-03 9.32E-06 9.32 31.17

30 190 7.39 6.86E-15 2.65E-06 3.24E-03 1.06E-05 10.55 35.30

60 238 5.51 1.35E-14 3.89E-06 4.75E-03 1.55E-05 15.47 51.73

90 260 5.09 1.76E-14 4.68E-06 5.73E-03 1.86E-05 18.63 62.30

120 282 3.66 2.24E-14 4.30E-06 5.25E-03 1.71E-05 17.09 57.16

190 322 2.16 3.34E-14 3.77E-06 4.62E-03 1.50E-05 15.01 50.22

IX
Table II: Data From NanoSight Rep 1

Number Mean
Error in
Particle Size Concentration (E8 No of Completed
Time(mins) Standard Deviation Concentration (E8
adjusted not particles/mL) Tracks
particles/mL)
smoothed (nm)

0 108 46 13.08 756 0.48

5 126 50 14.15 630 0.56

10 124 56 10.73 579 0.45

20 140 53 11.35 733 0.42

30 170 60 7.39 602 0.30

60 213 76 5.51 447 0.26

90 229 86 5.09 374 0.26

120 244 96 3.66 288 0.22

190 299 85 2.16 136 0.19

X
Table III: Amount of Drug in Particle Calculation, NanoSight Rep 2

Volume
Mean Vol Concentration
(Adjusted Concentration/108 Diameter Volume Fraction X of drug in % Of Drug no of
Time (Mins)
not particles per mL (cm) ^3 Fraction ρ X 1000 particles Added nanomoles
smoothed)/ (g/litre) (moles/litre)
nm

0 218 4.41 1.04E-14 2.39E-06 2.92E-03 9.51E-06 31.82 9.51

10 253 1.92 1.62E-14 1.63E-06 1.99E-03 6.47E-06 21.65 6.47

20 264 1.57 1.84E-14 1.51E-06 1.85E-03 6.01E-06 20.12 6.01

30 307 1.99 2.89E-14 3.01E-06 3.69E-03 1.20E-05 40.10 11.99

60 308 1.72 2.92E-14 2.63E-06 3.22E-03 1.05E-05 35.00 10.46

90 314 1.33 3.10E-14 2.15E-06 2.64E-03 8.57E-06 28.67 8.57

120 353 0.76 4.40E-14 1.75E-06 2.14E-03 6.96E-06 23.28 6.96

150 415 1.27 7.15E-14 4.75E-06 5.81E-03 1.89E-05 63.21 18.90

180 428 1.52 7.84E-14 6.24E-06 7.63E-03 2.48E-05 82.99 24.81

XI
Table IV: Data From NanoSight Rep 2

Number Mean
Adjusted Data Error in
not smoothed Concentration (E8 No of Completed Concentration (E8
Time (Mins) nm Std Dev particles/mL) Tracks particles/mL)

0 206 49 4.41 1516 0.11

10 240 54 1.92 642 0.08

20 251 56 1.57 444 0.07

30 289 71 1.99 505 0.09

60 291 69 1.72 425 0.08

90 299 66 1.33 300 0.08

120 326 90 0.76 166 0.06

150 373 125 1.27 257 0.08

180 378 138 1.52 293 0.09

XII
Table V: Amount of Drug in Particle Calculation, NanoSight Rep 3

Concentration
Volume Concentration Vol Fraction X of drug in
Mean (E8 Diameter (cm) Volume ρ X 1000 particles no of
Time (Mins) (raw) particles/mL) ^3 Fraction (g/litre) (moles/litre) % Of Drug Added nanomoles

0 155 14.5 3.724E-15 2.826E-06 3.456E-03 1.12425E-05 37.48 11.24

10 243 1.92 1.435E-14 1.442E-06 1.763E-03 5.73616E-06 19.12 5.74

20 238 1.57 1.348E-14 1.108E-06 1.355E-03 4.40688E-06 14.69 4.41

30 281 1.99 2.219E-14 2.311E-06 2.826E-03 9.19333E-06 30.64 9.19

60 309 1.72 2.950E-14 2.656E-06 3.248E-03 1.05659E-05 35.22 10.57

90 315 1.33 3.126E-14 2.176E-06 2.661E-03 8.65534E-06 28.85 8.66

120 395 0.76 6.163E-14 2.451E-06 2.998E-03 9.75227E-06 32.51 9.75

150 389 1.27 5.886E-14 3.912E-06 4.785E-03 1.55652E-05 51.88 15.57

180 421 1.52 7.462E-14 5.936E-06 7.259E-03 2.36152E-05 78.72 23.62

XIII
Table VI: Data From NanoSight Rep 3

Number Mean
Adjusted Data Error in
not smoothed Concentration (E8 No. of Completed Concentration (E8
Time (mins) (nm) Std Dev particles/mL) Tracks particles/mL)

0 134 55 14.5 2598 0.28

10 225 54 6.05 1357 0.16

20 215 74 5.12 878 0.17

30 260 76 2.55 620 0.10

60 282 90 2.23 549 0.10

90 284 97 2.07 400 0.10

120 355 120 1.62 326 0.09

150 373 125 1.27 257 0.08

180 383 138 1.52 293 0.09

XIV
 9.3 Chapter 3 Additional Data

With both instruments, optimisation of pipetting settings was required for accurate

dispension of the 10 μL DMSO aliquot. Slightly different optimisation was used for

each instrument, as detailed below. Once optimised, reproducible 10 μL dispension

was achieved.

1. Detailed robot settings

Lab 1 – Tecan Genesis

10 μL DSMO stock solution dispension -The setting used was the default ‘DMSO

low volume’ setting (see appendix for details), with one change. The robot was set to

aspirate 30 µL of the solution, discard 10 µL to waste, and then pipette 10 µL into

the sample vial. This gave more reproducible results than pipetting 10 µL from a dry

tip.

Aspiration settings;

Speed – 20 μL/s

Delay – 200 ms

Liquid detection used.

Aspirate position – z max ± offset, no tracking.

Dispension settings;

Speed – 600 μL/s

Break off speed – 400 μL/s

XV
Delay – 0ms

Dispense position – z dispense ± offset, no tracking.

990 μL of buffer - default ‘water’ setting. For samples mixed by the robot, the

instrument was

programmed to aspirate and dispense 500 μL of suspension 5 times.

Aspiration settings;

Speed – 150 μL/s

Delay – 300 ms

Aspirate position – z max ± offset, with tracking, 2mm X centre Y centre. Use liquid

detection.

Dispension settings

Speed – 600 μL

Break off speed – 400 μL/s

Delay – 0ms

No liquid detection.

Dispension position- z dispense ± offset, no tracking.

990 μL of DMSO - the default ‘DMSO’ setting.

Aspiration settings;

Speed – 150 μL/s

XVI
Delay – 400 ms

Aspiration position – z max ± offset, no tracking.

Dispension settings;

Speed – 600 μL/s

Break off speed – 400 μL/s

Delay – 0 ms

Dispension position – z dispense ± offset, no tracking.

Lab 2 – Biomek 2000

10 μL DSMO stock solution dispension – The robot was programmed to aspirate at

height 25%, using liquid level tracking, with an aspiration rate of 10. Pre-wet and tip

touch settings were applied. The dispension settings were to dispense at 25% height,

and dispension type was set to ‘to deliver’. The rate used was 10, and the tip touch

setting was applied.

990 μL of buffer – The robot was programmed to aspirate at height 50%, using liquid

level tracking, with an aspiration rate of 10. Pre-wet settings were applied. The

dispension settings were to dispense at 70% height, and dispension type was set to

‘to contain’, with a ‘blowout’. It should be noted that Lab 2 experiments for ‘Early’

and ‘Later’ experiments (see section 3.2) were all performed with a dispension

speed setting of 10 (estimated to be around 0.5 mL/second). This is the

recommended setting for this instrument, however was not suitable for preparation of

XVII
large numbers of samples as it caused buffer to ‘splash’ out of the tube around 50%

of the time. For this reason, the rate was lowered to setting 7 (estimated to be around

0.35 mL/second), and all other Lab 2 experiments (mixing investigation experiments,

section 3.3 onwards) use rate setting 7990 μL of DMSO - The robot was

programmed to aspirate using liquid ‘sensing’, with an aspiration rate of 10. Pre-wet

settings were applied. The dispension settings were to dispense at 70% height, and

dispension type was set to ‘to contain’, with a ‘blowout’. The rate used was 10.

2. Detailed HPLC Conditions

Lab 1

Chromeleon software (version 6.80 SR8 Build 2623(156243)) was used with the

HPLC system.

A 2 mL/min flow rate was used, and the column temperature was 60 °C. An injection

volume of 10 µL was used, and the detector wavelength set to 267 nm. The total run

time was 3.5 mins. Mobile phase A consisted of 5 % MeCN in water + 0.1 % TFA,

and mobile phase B consisted of 5 % water in MeCN + 0.1 % TFA. The retention

time of the drug using this method was 2.44 minutes.

XVIII
Table III; Gradient Conditions, Lab 1

Time (mins) %B

0.5 40

2.5 95

3 95

3.01 40

3.5 40

Lab 2

Atlas software (version 1.6.0) was used with the HPLC system.

A 2 mL/min flow rate was used, and the column temperature was 40 °C. An injection

volume of 10 µL was used, and the detector wavelength set to 260 nm. The total run

time was 3.5 mins. Mobile phase A consisted of water + 0.02 % TFA, and mobile

phase B consisted of MeCN + 0.02 % TFA. The retention time of the drug using this

method was 2.18 minutes.

XIX
Table IV; Gradient conditions Lab 2

Time (mins) %B

0 40

0.5 40

1.5 95

2.5 95

2.51 40

3.5 40

Lab 3

Empower Pro, 2002 (build 1154) software was used with the HPLC system.

A 1 mL/min flow rate was used, and the column temperature was 40 °C. An injection

volume of 3 µL was used, and the detector wavelength set to 257 nm. The total run

time was 3 mins. Mobile phase A consisted of water + 0.1 % TFA, and mobile phase

B consisted of MeCN + 0.1 % TFA. An isocratic method was used. The retention

time of the drug using this method was 1.99 minutes.

XX
 3. Dissolution from solid and kinetic solubility – detailed results

Table V: Dissolution from solid results

Time (hrs) 0.66 24 48 162.5

4/5 samples
2/5 samples - no
4/5 samples - no - no analyte
Amount of drug analyte peak
analyte peak peak 5/5 samples - no
present in detected. Mean
detected. detected. analyte peak
supernatant from remaining
Remaining 1 Remaining detected.
(nmoles/mL) 3 samples -0.8
sample - 2.3 1 sample -
± 0.6]
0.3

Table VI: Kinetics of precipitation results

Time (hrs) 0 48 120 168

2/5 samples - no 3/5 samples

analyte peak - no analyte 3/5 samples = no

Amount of drug
detected.1/5 peak analyte peak
present in
2.7 ± 1.9 samples - poor detected. detected. Mean from
supernatant
recovery. Mean Mean from remaining two
(nmoles/mL)
from other two 2/5 samples samples - 0.6

samples - 1.7 - 0.7

XXI
 4. Lab 3 25 °C data

As with the other data sets, VOR mixing is significantly different to ASP-M and NO

mixing, and returns the lowest supernatant amount.

Table VII: Lab 3 25 °C supernatant amounts and p-values

Treatment Amount in supernatant, nanomoles ± S.D. p value

50 nanomoles per mL VOR, 9.9 ±3.0 NO, 13.8 ± 2.4 0.01

50 nanomoles per mL VOR, 9.9 ±3.0 ASP-M, 15.1 ± 2.9 0.003

50 nanomoles per mL NO, 13.8 ± 2.4 ASP-M, 15.1 ± 2.9 0.3 (NS)

XXII
 9.4 Chapter 5 Additional Data

Table XIV -1 and 5% DMSO, supernatant/filtrate amounts after various separation

procedures

Table XV- 1 and 5% DMSO samples, incubation times (supernatant analysis after
centrifugation)

XXIII
Table XIV: 1 and 5% DMSO, supernatant/filtrate amounts after various separation procedures

Amount present in supernatant ± S.D., μM, after Amount present in filtrate ± S.D., Amount present in filtrate ± S.D., μM, 0.22
centrifugation (5 mins, 9,300 x g rcf) μM, 0.45 μm PVDF filtration μm PVDF filtration

Compound 1% v/v DMSO 5% v/v DMSO 1% v/v DMSO 5% v/v DMSO 1% v/v DMSO 5% v/v DMSO

Amiodarone HCl 27.5 ± 1.53 40.0 ± 1.9 27.3 ± 2.7 36.9 ± 1.0 11.9 ± 0.8 25.3 ± 1.9
Clotrimazole 20.2 ± 3.2 31.2 ± 2.9 8.4 ± 0.6 19.5 ± 0.3 N/A N/A
No Peak Peak Detected
Tolnaftate 13.3 ± 5.6 31.2 ± 3.3 N/A N/A
Detected Inconsistently
Table XV: 1 and 5% DMSO samples, incubation times (supernatant analysis after centrifugation)

Incuba
tion Amiodarone HCl, amount of compound present in Clotrimazole, amount of compound present in Tolnaftate, amount of compound present in
time, supernatant ± S.D., μM supernatant ± S.D., μM supernatant ± S.D., μM
mins
1% DMSO 5% DMSO 1% DMSO 5% DMSO 1% DMSO 5% DMSO
0 27.5 ± 1.5 39.8 ± 2.1 20.2 ± 3.2 31.2 ± 2.9 13.3 ± 5.6 31.2 ± 3.3
10 23.8 ± 0.3 41.0 ± 0.5 - - 9.4 ± 6.7 20.8 ± 3.3
60 22.9 ± 0.5 38.6 ± 0.5 16.5 ± 0.5 29.4 ± 0.5 - -
180 - - 18 ± 1 29 ± 2 3.0 ± 0.5 16.8 ± 0.2
1440 24.0 ± 0.3 37.0 ± 0.4 2.9 ± 0.2 *24.7 ± 0.6 *5.1 ± 1.8 *17.4 ± 4.0
*poor recovery (<85 %)

XXIV
 9.5 Chapter 6 Additional Data

Figure 9 - Average particle size distributions for samples containing various protein
amounts, time zero

Figure 10 - Amiodarone HCl, pH 4 samples, time zero, with and without protein
present

Figure 11 - Amiodarone HCl pH 4 samples, time 60 mins, with and without protein
present

300

0 mg/mL 0.01 mg/mL 0.05 mg/mL

250
particle concentration (106 particles/mL)

200 0.1 mg/mL 0.5 mg/mL 1 mg/mL

150

100

50

0
0 200 400 600 800 1000 1200 1400

Particle diameter (nm)

Figure14: Average particle size distributions for samples containing various protein amounts, time
zero

XXV
 450

Particle concentration (106 particles/mL)

400

350
0.1mg/ml pH 4 time 0
300

250 no protein pH 4 time 0

200

150

100

50

0
0 50 100 150 200 250 300 350 400

Particle diameter (nm)

Figure15: Amiodarone HCl, pH 4 samples, time zero, with and without protein present

70
Particle cocnentration (106 particles/mL)

60
pH 4 0.1 mg/mL chymo t60
50

40 pH 4 no protein t60

30

20

10

0
0 200 400 600 800 1000
Particle diameter (nm)

Figure 16: Amiodarone HCl pH 4 samples, time 60 mins, with and without protein present

XXVI