Quantification of Nanoscale Density Fluctuations by Electron Microscopy: Probing Cellular Alterations in Early Carcinogenesis
Quantification of Nanoscale Density Fluctuations by Electron Microscopy: Probing Cellular Alterations in Early Carcinogenesis
Quantification of Nanoscale Density Fluctuations by Electron Microscopy: Probing Cellular Alterations in Early Carcinogenesis
Abstract
Most cancers are curable if they are diagnosed and treated at an early stage. Recent studies
suggest that nanoarchitectural changes occur within cells during early carcinogenesis and that
such changes precede microscopically evident tissue alterations. It follows that the ability to
comprehensively interrogate cell nanoarchitecture (e.g., macromolecular complexes, DNA,
RNA, proteins and lipid membranes) could be critical to the diagnosis of early carcinogenesis.
We present a study of the nanoscale mass-density fluctuations of biological tissues by
quantifying their degree of disorder at the nanoscale. Transmission electron microscopy
images of human tissues are used to construct corresponding effective disordered optical
lattices. The properties of nanoscale disorder are then studied by statistical analysis of the
inverse participation ratio (IPR) of the spatially localized eigenfunctions of these optical
lattices at the nanoscale. Our results show an increase in the disorder of human colonic
epithelial cells in subjects harboring early stages of colon neoplasia. Furthermore, our findings
strongly suggest that increased nanoscale disorder correlates with the degree of tumorigenicity.
Therefore, the IPR technique provides a practicable tool for the detection of nanoarchitectural
alterations in the earliest stages of carcinogenesis. Potential applications of the technique for
early cancer screening and detection are also discussed.
supramicron scales in tissues/cells that are prominent in the inverse area) and (ii) the average IPR value increases from
later stages of carcinogenesis (i.e. dysplasia) resulting from a uniform lattice to a disordered lattice with an increase in
multiple genetic and epigenetic alterations [9–12]. However, the degree of the disorder. Moreover, in 2D, the value of
understanding the earlier initiating morphological events in IPR not only depends on the mass-density fluctuations, but
carcinogenesis requires the ability to detect nanoarchitectural also the correlation length of these fluctuations (Lc). The
changes, which, at best, remains challenging. In particular, average value of the IPR is proportional to the refractive
the ability to comprehensively interrogate the nanoarchitecture index fluctuation (n) of optical media and corresponding
of cells and tissues is critical for applications that require fluctuation correlation decay length (Lc), or IPR ∝ n × Lc.
an understanding of the role of cell nanoarchitecture in This has been well studied in condensed-matter physics for
early carcinogenesis [5, 13]. To date, the nanoarchitectural characterizing the disorder properties of localization and the
properties of cells/tissues have not been well understood or quantum Hall effect [20–23].
studied. In a recent short letter [24] we showed that applying
As suggested by several recent studies, the progression the IPR technique we could measure and differentiate the
of cancer in the early stages is accompanied by nanoscale degree of disorder at the nanoscales in microscopically similar
morphological or architectural changes in the internal cell cells, but with different malignant potential. In particular,
structure that precede histological abnormalities. Such we used human colon cancer cell line HT29 and its genetic
changes result in nanoscale mass-density fluctuations in cells. variant to study the degree of disorder via IPR analysis
While conventional visible-light microscopy techniques have of their TEM images. The IPR technique was enabled to
been widely used to characterize biological systems, their quantify the degree of disorder at the nanoscale and their
ability to detect changes at the nanoscale is fundamentally relative variation in otherwise microscopically similar cells
impeded by their diffraction-limited resolution [14]. However, with different malignant potential for human colon cancer cell
recent spectral optical microscopy studies have shown that line HT29 and its genetic variant (CSK). In this study, we
such nanoscale fluctuations manifest themselves prominently also validated the IPR technique of TEM image analysis by
in early carcinogenesis [15–17]. Furthermore, these studies using model disordered nanoparticle systems, that is, using
indicate that changes in nanomorphology occur at length scales nano-disordered dielectric samples of known parameters.
of approximately 10–100 nm, which reflect the length scale Furthermore, rigorous numerical simulations were performed
associated with the cellular building blocks (e.g., DNA, RNA, for cell like weakly disordered media to establish the relation
proteins and lipids). However, far-field optical techniques are between IPR and the disorder parameters of the samples to
unable to probe the detailed nature of these nanoscale changes. confirm IPR ∝ n × Lc. In the context of the biological
This obstacle led us to explore the potential of transmission study, the significance of this relation is related to the degree
electron microscopy (TEM) with its nanoscale (∼1 nm) of disorder which is related to the morphological condition
resolution. TEM imaging techniques have been widely used in a cell in carcinogenesis. Specifically, IPR provides
to visualize nano- and micro-structures in biological samples the measure of the degree of disorder of refractive index
[18]. Still, the quantitative information embedded within fluctuations in biological media due to light scattering in this
a TEM image in the context of biological sub-structure is closed media.
poorly understood, in particular, subtle short-range nanoscale In this paper, we first discuss the methodology for
fluctuations and correlations of the grayscale, as well as the
quantifying nanoscale mass-density fluctuations in detail. A
changes of such fluctuations with the pathogenesis of disease
short description was presented in [24] about the methodology,
states [19].
but here we present the methodology in detail (section 2).
To quantify the mass-density fluctuations, we utilized
Then we review in brief our previous results [24] to provide
an idea from condensed-matter physics and applied it to
a background for the present study (section 3). Finally,
biological systems. This concept involves construction of
for the first time, we report the results of the quantification
optical lattices from the mass-density fluctuations derived
of the nanoscale mass-density fluctuations in human
from TEM images of cells/tissues, and then statistically
colon tissues and their alterations in early carcinogenesis
analyzing the localization characteristics of the eigenfunctions
(section 4.1). We further study the correlation between the
of these disordered lattices via the average value of the inverse
nanoscale fluctuations and tumorigenicity (or the degree of
participation ratio (IPR) of these eigenfunctions. The IPR of
malignancy) in colon early carcinogenesis (section 4.2). We
an eigenfunction E is defined as
conclude the results and discuss the potential application of
IPR = |E(r)|4 dr (1) the technique for early cancer detection and screening at the
end (section 5).
(in units of inverse area in two dimensions). The optical-
localization properties of these lattices are directly related
2. Inverse participation ratio analysis technique for
to the statistical average value of the IPR. The statistical
biological cells/tissues
properties of IPR are important quantitative measures of the
spatial localization of the eigenfunctions in optical lattices (i.e. 2.1. Construction of optical lattices
either ordered or disordered). In two-dimensional (2D) optical
lattices, (i) the average value of IPR for a uniform lattice is a A short description of the methodology was presented in
fixed universal number (approximately 2.5 in the unit of the [24], but here we present the methodology in detail. The
2
Phys. Biol. 8 (2011) 026012 P Pradhan et al
(A) (B) (E )
(C ) (D)
(G )
(F )
Figure 1. (A) and (B) Representative TEM images of tissues from normal patients and patients harboring colonic adenometous polyps.
(C) and (D) Corresponding IPR-pixel images for the pixel size 154 nm × 154 nm (unit grid length is 7.7 nm). (E) Relative IPR(L)Pixel
distributions (ensemble) of IPR-pixel sizes L × L = 154 nm × 154 nm for rectal tissues from normal patients and early precancerous
patients. (F) Ensemble-averaged values of the IPR-pixel IPR(L)Pixel versus L (in nm) plots for (i) uniform background, (ii) normal
tissues and (iii) early precancerous tissues. (G) Corresponding standard deviation σ (IPR(L)Pixel ) versus L (in nm) plots for normal tissues
and early precancerous tissues. (Because of the large number of samples, error bars are negligible in (F) and (G).)
TEM grayscale intensity, ITEM, decays exponentially with It has been well studied and shown that the optical
the thickness of the sample, where the decay constant is a refractive index (n) is linearly proportional to the local density
function of the mass density [9]. This exponential decay can (ρ) of intracellular macromolecules, such as proteins, lipids,
be approximated as a linear decay for a very thin sample, when DNA and RNA, i.e. n = n0 + n = n0 + αρ, where n0 is
the sample length is much smaller than the decay length scale. the refractive index of the medium surrounding a scattering
TEM studies of thin layers of nanoscale dielectric beads have structure, ρ is the local concentration of solids, and α is a
shown that ITEM is linearly proportional to the mass density of proportionality constant. The studies further showed that the
the beads [25, 26]. For a very thin biological sample, one can majority of the scattering substances found in living cells have
assume that the grayscale TEM image intensity at any lattice approximately the same value of the proportionality constant α
point around the point (x, y), ITEM (x, y) (see figures 1(A)– ∼ 0.18 [27, 28]. Furthermore, we consider that the absorption
(D)), is linearly proportional to the total scattering strength of the contrast agent by the biomass of the thin tissue voxel
and, hence, the mass M(x, y) (total biomass of the biological is linearly proportional to the total mass present in the voxel
samples) present in the corresponding tissue voxel around the and that the lattice is an effective optical lattice. We consider
lattice point (dimensions: 10 nm × 10 nm in the x–y plane; that the refractive index at the voxel around the point (x, y)
70 nm along the z-direction) of the sample slices. In this paper, is n(x, y). Therefore n(x, y) is functionally proportional to its
all the TEM studies are related to the biological samples of biomass and we can write
thin films, therefore we will consider them as 2D samples. n(x, y) = f (M(x, y)) = f (ITEM (x, y)). (2)
3
Phys. Biol. 8 (2011) 026012 P Pradhan et al
We further assume that the form of ITEM at (x, y) is given mean is the same as the hopping parameter, i.e. t = mean(εi ),
by and we further consider (I0 (x, y)f0 /f0 )/t = constant = 1,
ITEM (x, y) = I0 + ITEM (x, y), (3) without any loss of generality.
The average value of the IPR of a pixel of side
where I0 is the mean background of the whole TEM sample, length L, IPRPixel , (area L × L) can then be written as
and ITEM (x, y) is the fluctuating part of the intensity around a [20–22]
spatial point (x, y) of the pixel. Then, the refractive index of a N
tissue voxel from the corresponding TEM pixel can be written 1 L L 4
IPR(L)Pixel = Ei (x, y) dx dy, (7)
as N i=1 0 0
n(x, y) = n0 + n(x, y) = f (ITEM (x)) where Ei is the ith eigenfunction of the Hamiltonian in
= f0 + f0 × ITEM (x, y), (4) equation (6) of an optical lattice (i.e. an IPR pixel) of size L ×
L; N = L2a (La = L/a (lattice size), a = dx = dy) is the total
where n0 is the constant-background part of the full sample,
number of the eigenfunctions; and Pixel denotes the average
and n(x, y) is the fluctuating part of the refractive
over all the N eigenfunctions of the IPR pixel. Importantly,
index n(x, y). It can be shown that the effective optical
this derivation process mainly considers the fluctuating part of
potential of an optical lattice, εi, has the following form
n relative to its average background as the rescaled potential.
[29, 30]:
The statistical analyses of the IPR were done by taking the
εi ∝ n(x, y)/n0 = (ITEM (x, y)/I0 ) × (I0 × f0 /f0 ), (5) ensemble averaging and std of the samples of a fixed size over
where fluctuating TEM intensity ITEM (x, y) I0 and biological tissues from a single patient and then averaging
n(x, y) n0 . This is a good approximation, this is from several patients.
because for tissue the range of n0 = 1.33–1.38 and the range of
n = 0.01–0.1. The exact eigenfunctions (Ei (x, y), i = 1−N) 3. Previous studies using the IPR analysis
of each two-dimensional pixel of optical sample size of area
L × L can be calculated by solving the wave equation for the In this section we briefly review our previous results reported
electric field in the lattice using a disorder tight-binding model in [24] to provide a background of the present study of the
[29, 30]. precancerous tissues. In the study, it was shown that the
IPR technique can be used to quantify minute changes in
nanoscale disordered dielectric (optical) media comprising a
2.2. Tight-binding model and IPR statistics
known, controlled disordered model system. Then we showed
To quantify the disorder properties of the TEM images, that the IPR technique can be used to differentiate between
we have carried out numerical calculations of the Anderson two types of cell lines with different malignancy.
disorder tight-binding model (TBM). The TBM has been well
studied, and it has proven to be a good model for describing 3.1. Validation of the IPR technique using nano-disordered
single-optical states or localized optical states of systems of media
any geometry and disorder. In our study, we consider one
optical state of a photon per lattice site, and the interlattice site In order to investigate the hypothesis that IPR technique
hoppings are restricted to the nearest neighbors only. Such a can accurately quantify nanoscale disorder changes, we
Hamiltonian can be written as [29, 30] performed TEM studies on model experimental systems of
dielectric nanoparticles (average diameter ∼6 nm, standard
H = εi |ii| + t |ij | + |j i|, (6) deviation ∼2 nm) [24]. Dielectric nanoparticles, which act
i ij as disordered scatterers in a concentration-dependent manner,
where εi ∝ n(x, y)/n0 is the ith lattice site energy; |i and were deposited on a formvar thin dielectric film. The goal
|j are the optical eigenfunctions at the ith and j th lattice sites, of this study was to determine whether the short-length-scale
respectively; i j indicates the nearest neighbors; and t is the disorder strength could be measured by the IPR technique and
overlap integral between sites i and j . how the average value of IPR changes with an increase in the
In our analyses, for short-length refractive index nanoparticle density or the scattering mean-free path. Given
fluctuations, a large TEM micrograph (in our study 15.8 μm × the ultrafine size of the nanoparticles used in these experiments
15.8 μm as shown in figures 1(A) and (B) is virtually cut into and their random orientation and spatial distribution, the TEM
smaller ((77 nm × 77 nm)–(308 nm × 308 nm)) samples image contrast is dominated by mass thickness, with minimal
or IPR pixels. To project the TEM image to the tight- average contribution from diffraction and phase contrast under
binding model, the fluctuating part of every grayscale point of the imaging conditions employed. While much simpler than
the TEM image is first considered proportional to the onsite a biological system, this model allowed us to control and
optical potential energy εi (i.e. using equation (5)). The characterize the disorder strength. Both, the model and
optical potential is obtained by first projecting the optical wave biological systems, have an average uniform refractive index
equation to the Schrödinger equation, and then equating the background (n0 ) and weaker refractive index fluctuations (n)
optical potential of the optical wave equation with the potential above the uniform background (n/n0 1).
of the Schrödinger equation. Finally, in the tight-binding For the main result, we performed length scale
model, the optical potential εi is then rescaled such that its dependence of IPR(L)Pixel . The result showed that
4
Phys. Biol. 8 (2011) 026012 P Pradhan et al
IPR(L)Pixel value is proportional to the product of the technique is applicable in measuring the early signatures of
amplitude of the refractive index fluctuation and its correlation. nanoscale mass density changes in tissues via the ‘field-effect’
Overall, the validation study in [24] showed that the nanoscale phenomenon. Many early cancer screening techniques are
disorder can be quantified by the IPR technique, which can designed to exploit the ‘field effect’ of carcinogenesis. The
distinguish statistically significant differences between the two ‘field effect’ is a proposition that the genetic/environmental
disordered systems of minute difference in disorder. milieu that results in neoplasia in one region of an organ
should be detectable throughout the organ [17, 31, 32]. In
3.2. IPR study of cell lines: nanoscale mass-density particular, we wanted to test this ‘field effect’ model for colon
fluctuation analysis of HT29 cells and their CSK-knockdown carcinogenesis by measuring the nanoarchitectural changes
genetic variant in rectal cells/tissues of patients who have premalignant
adenomatous polyps in their colon.
To investigate the changes in nanoscale mass-density
fluctuations with the progression of carcinogenesis, we 4.1.1. Tissue sample collection. To this end, we carried
performed experiments on the well-characterized HT29 colon out a pilot study involving IPR analyses of 10 human
cancer cell line and its CSK-knockdown genetic variant, subjects. Biopsies from endoscopically normal rectal mucosa
which was engineered via knockdown of the tumor suppressor were acquired from these subjects at the time of their
gene c-terminus src kinase, leading to more aggressive colonoscopies in accordance with standard clinical practice.
neoplastic behavior of these cells [24]. The increase of the The colonoscopies indicated that five subjects harbored
nanoscale disorder (in particular, increase of the disorder precancerous adenomatous polyps (the size of adenomatous
strength parameter Ld = n2 × Lc) with the progression polyps ranged from 2 to 10 mm) in their colon, while the
of carcinogenesis was shown in our prior optical experiments other five subjects were free from adenomas. All tissue
using partial wave spectroscopic microscopy (PWS) for the samples appeared histopathologically normal. TEM images
same cell line study [15–17]. HT29 cells, which were were then acquired following the same protocol as described
used as a control in the experiment, were actually a human below.
colonic adenocarcinoma cells that were able to express
differentiation features which are characteristic of mature
4.1.2. Sample preparation for TEM and TEM imaging. The
colonic cells. The knockdown of the tumor-suppressor
biopsy samples were first placed in Karnovsky’s fixative
CSK, a gene which is knocked down in most colon cancers,
for 2 weeks to preserve structure. The fixative consists of
increases the growth/proliferation of HT29 cells. In TEM
0.1 M phosphate buffered solution containing 5%
experiments on these cells, we were interested to observe
glutaraldehyde with pH between 7.2 and 7.4. Following a
the effect of carcinogenesis in terms of intracellular mass-
standard protocol, the samples were stained with osmium
density fluctuations at the nanoscale. We noted that all cells
tetraoxide (OsO4 ), dehydrated, and then embedded in
used for the experiment were otherwise cytologically (i.e.
resin containing 36% ERL 4221, 12% diglycidyl ether
microscopically) indistinguishable.
of polypropyleneglycol (DER 736), 51% nonenyl succinic
The length scale-dependent analyses of the average and
anhydride (NSA), and ∼1% dimethylaminoethanol (DMAE)
std of IPR with the increasing sample length for the HT29
by mass. Samples were then sectioned with an ultra-
and CSK cells were studied. Our results showed that the
microtome to a thickness of 70 nm.
average IPR(L)Pixel and the std σ (IPR(L)Pixel ) values are
higher for the more malignant CSK cells relative to the less
malignant HT29 cells, with a significant p-vale (<.05). Both 4.1.3. TEM imaging. Finally, TEM micrographs were
IPR(L)Pixel and σ (IPR(L)Pixel ) increased with increasing obtained (JEM-1400, JEOL, Tokyo, Japan) for each of
L for the CSK cells relative to the HT29 cells, indicating a the prepared samples from different patients (control and
relatively greater degree of disorder or nanoscale fluctuations precancerous). A 200 keV electron beam with a fixed
for the CSK cells, relative to HT29 cells. magnification (40 K) was used for each micrograph.
The above results of the cell line study motivate us to study
the nanoscale fluctuations in precancerous human tissues for 4.1.4. Statistical analysis. For the statistical analysis, we
early cancer detection using IPR technique. took tissue samples from n = 5 control patients and n = 5
precancerous patients. We collected ∼15–20 separate tissue
samples from the different parts of the rectum of each patient,
4. Results of the human precancerous colonic tissue
and TEM micrographs were prepared for each tissue sample.
study
We randomly chose 10 TEM micrographs of independent
4.1. Nanoscale mass-density fluctuation of normal and early tissue samples for each patient. In particular, an ensemble
precancerous colon tissues of 50 independent TEM micrographs of independent tissues
(i.e. 50 independent measurements) from control patients
To investigate the hypothesis that nanoscale mass-density (n = 5, 10 micrographs per patient) and 50 independent
fluctuations increase with the progression of carcinogenesis, micrographs (i.e. 50 independent measurements) of early
we performed experiments on colonic tissues from control and precancerous patients (n = 5, 10 micrographs per patient)
precancerous patients. Based on the results reviewed in the were taken for the statistical significance of the IPR difference
previous section, in this paper, we investigate whether the IPR study.
5
Phys. Biol. 8 (2011) 026012 P Pradhan et al
Figures 1(A) and (B) show representative TEM grayscale indicate that the average nanoscale fluctuations within a cell,
images of rectal tissue samples from adenoma-free subjects in part of a cell, or in a patient are quite randomly distributed,
and from subjects harboring precancerous adenomatous i.e. having weaker correlations.
polyps in their colon. Figures 1(C) and (D) show the
corresponding IPR images (IPR pixel dimension = 154 nm ×
4.2. Nanoscale mass-density fluctuation and tumorigenicity
154 nm). The IPR images clearly indicate different disorder
correlation
states for tissues obtained from the normal and the early
precancerous subjects. In this section we study the correlation of nanoscale mass-
Figure 1(E) shows the distribution P (IPR(L)Pixel ) for density fluctuations with tumorigenicity. In particular, to
rectal tissues from the normal and the early precancerous quantify the correlation between the degrees of developing
subjects for IPR pixel sizes L × L = 154 nm ×154 nm. This malignant phenotype (i.e. tumorigenicity) and increase in the
shows a distinct separation between the two groups of subjects. degree of nanoscale mass-density fluctuations (i.e. the average
Figure 1(F) shows plots of IPR(L)Pixel versus L IPR value), we further studied the relative IPR values for
for three different cases: (i) uniform lattice, (ii) rectal the above five adenoma patients according to their clinically
tissues from the normal subjects, and (iii) rectal tissues from classified tumorigenicity. Most colon cancers progress
the early precancerous subjects. Here, < > denotes the through a precursor lesion, the premalignant adenomatous
ensemble averaging (averaged over ∼500 000 IPR pixels for polyp [33]. The degree of tumorigenicity of colonic adenomas
the sample size L × L = 77 nm × 77 nm and averaged over (i.e. the risk of progression into cancer) depends on their
∼125 000 IPR pixels for the sample size L × L = 154 nm × size and histology and increases from diminutive adenomas
154 nm). Figure 1(G) shows plots of the corresponding (polyps size < 5 mm) to 5–9 mm adenomas to advanced
standard deviation σ (IPR(L)Pixel ) versus L.
adenomas (polyp size 10 mm, high-grade dysplasia or 25%
The three curves in figure 1(F) clearly show that the
villous features) [34]. Accordingly, in our study, the patients
IPR(L)Pixel value is highest for rectal tissues from the
with adenomas were divided into three categories: patients
early precancerous subjects. For example, IPR(L)Pixel
with diminutive adenoma (n = 2 patients), patients with
values for the uniform background, the normal-subject rectal
non-diminutive, non-advanced adenoma (n = 2) and patients
tissues, and the early precancerous-subject rectal tissues are
with advanced adenoma (n = 1). Given that IPR increase
2.5, 3.053 and 3.196, respectively. Student’s t-test, two-
parallels the degree of malignant aggressiveness of cell lines,
tailed unequal variance p-value = 0.021, which is statistically
we hypothesized that IPR would not only be increased in the
significant.
field of carcinogenesis (patients harboring adenomas), but that
Higher values of the average IPR correspond to larger
its value would also correlate with the malignant potential of
disorder strengths by increased nanoscale fluctuations in the
these premalignant lesions.
tissues. Importantly, figure 1(F) also shows that the ratio
σ (IPR(L)Pixel )/IPR(L)Pixel increases much faster with In figure 2 we plotted the average IPR for patients without
increasing L for the early precancerous-subject rectal tissues adenomas and patients harboring adenomas of progressively
relative to the normal-subject rectal tissues. The rapid increasing tumorigenicity. Each average IPR value
increase of this ratio is attributed to the long tails in the IPR was calculated over ∼50 000 IPR pixels, with sample size
distributions. 154 nm × 154 nm. The figure clearly indicates an increasing
Figure 1 provides substantial evidence that trend in the average IPR value, and the data show that such
microscopically normal-appearing colonic epithelial cells in trend correlates well with the increase of tumorigenicity. The
early carcinogenesis exhibit a higher degree of nanoscale same trends were obtained for other IPR pixel sizes (i.e. for
disorder than the cells from a control patient. These results any pixel size 30 nm×30 nm to 308 nm×308 nm). These
suggest that IPR has the potential to detect early carcinogenic results underscore the potential of using the IPR technique and
alterations in the human colon. the average IPR value obtained as a biomarker of colorectal
Importantly, the data given in figures 1 (F) and (G) show carcinogenesis at the early stages.
that the difference in average and standard deviation of IPR We further calculated the statistical significance of
between control and adenoma patients appears to be around the difference in the average IPR values over the tissue
L ∼ 75 nm, but more prominent around L ∼ 100 nm. This micrographs of control (n = 5, 50 independent tissue
is the building block of the cell/tissue (DNA, RNA, Lipids, measurements/micrographs) and the IPR values of the three
proteins, etc). different adenoma sub-categories: diminutive adenoma (n =
Further, we have calculated the intratissue/ 2, 30 different tissue micrographs), non-diminutive adenoma
intramicrograph correlation for IPRPixel , which shows (n = 2, 30 different tissue micrographs) and advanced adenoma
∼0.148 as the correlation coefficient (50 micrographs of (n = 1, 20 different tissue micrographs). Student’s t-test, two-
independent tissues from control patients and 50 micrographs tailed unequal variance p-values are as follows: (i) between
of independent tissues from five adenoma patients for a total control and diminutive adenoma patients, p-value = 0.67
of 100 independent tissue micrographs). Furthermore, we (confirming less clinical significance of diminutive adenoma
have calculated the intrapatient correlation for IPRPixel , [34]). (b) Between control and non–diminutive adenoma, p-
showing that the correlation coefficient is ∼0.031 (for five value = 0.03. (c) Between control and advanced adenoma,
control patients and five adenoma patients). These results p-value = 0.05. Overall, the results show a correlation between
6
Phys. Biol. 8 (2011) 026012 P Pradhan et al
7
Phys. Biol. 8 (2011) 026012 P Pradhan et al
We anticipate that IPR analyses of TEM, as well [15] Subramanian H et al 2008 Optical methodology for detecting
as scanning transmission electron microscopy (STEM) histologically unapparent nanoscale consequences of
images, will have potential applications for tissue/cell genetic alterations in biological cells Proc. Natl Acad. Sci.
USA 105 20118–23
characterizations in basic biological research, as well as [16] Subramanian H et al 2009 Partial-wave microscopic
medical applications in detecting early-stage cancers. This spectroscopy detects subwavelength refractive index
also demands a larger population of patients, a study which is fluctuations: an application to cancer diagnosis Opt. Lett.
currently underway. 34 518–20
[17] Subramanian H et al 2009 Nanoscale cellular changes in field
carcinogenesis detected by partial wave spectroscopy
Acknowledgments Cancer Res. 69 5357–63
[18] Bozzola J J and Russell L D 1999 Electron Microscopy:
This work was supported by NIH grants (nos R01EB003682, Principles and Techniques for Biologists 2nd edn (Boston,
R01CA128641 and U54CA143869) and NSF grant no CBET- MA: Jones and Bartlett)
[19] Leapman R D 1986 Scanning-transmission
0937987. VPD acknowledges support from NIH/NCI PS-
electron-microscope (stem) elemental mapping by electron
OC grant no DMR-0603184 and NIH-CCNE (Northwestern) energy-loss spectroscopy Ann. N Y Acad. Sci.
grant no U54CA119341. Parts of the experiments were done 483 326–38
at the EPIC/NIFTI facility of the NUANCE Centre (supported [20] Pradhan P and Kumar N 1994 Localization of light in
by NSF NSEC, NSFMRSEC, Keck Foundation, the State coherently amplifying random-media Phys. Rev. B
of Illinois, and Northwestern University) at Northwestern 50 9644–47
[21] Pradhan P and Sridhar S 2002 From chaos to disorder:
University. PP thanks S Sridhar (Northeastern University, statistics of the eigenfunctions of microwave cavities
Boston) for many encouraging and insightful discussions. Pramana-J. Phys. 58 333–41
[22] Schwartz T, Bartal G, Fishman S and Segev M 2007 Transport
and Anderson localization in disordered two-dimensional
References photonic lattices Nature 446 52–5
[23] Prigodin V N and Altshuler B L 1998 Long-range spatial
[1] Jemal A et al 2009 Cancer Statistics, 2009 CA Cancer J. Clin.
correlations of eigenfunctions in quantum disordered
59 225–49
systems Phys. Rev. Lett. 80 1944–7
[2] Delcuve G P, Rastegar M and Davie J R 2009 Epigenetic
control J. Cell. Physiol. 219 243–50 [24] Pradhan P, Damania D, Joshi H, Turzhitsky V, Subramanian
[3] Mohn F and Schubeler D 2009 Genetics and epigenetics: H, Roy H K, Taflove A, Dravid V and Backman V 2010
stability and plasticity during cellular differentiation Trends Quantification of nanoscale density fluctuations using
Genet. 25 129–36 electron microscopy: light-localization properties of
[4] Adams P D 2007 Remodeling of chromatin structure in biological cells Appl. Phys. Lett. 97 243704–6
senescent cells and its potential impact on tumor [25] Zeitler E and Bahr G F 1962 Photometric procedure for weight
suppression and aging Gene 397 84–93 determination of submicroscopic articles quantitative
[5] Zuckerkandl E and Cavalli G 2007 Combinatorial epigenetics, electron microscopy J. Appl. Phys. 33 847–53
‘junk DNA’, and the evolution of complex organisms Gene [26] Loferer-Krossbacher M, Klima J and Psenner R 1998
390 232–42 Determination of bacterial cell dry mass by transmission
[6] Lawlis S J, Keezer S M, Wu J R and Gilbert D M 1996 electron microscopy and densitometric image analysis Appl.
Chromosome architecture can dictate site-specific initiation Environ. Microbiol. 64 688–94
of DNA replication in Xenopus egg extracts J. Cell Biol. [27] Barer R, Ross K F A and Tkaczyk S 1953 Refractometry of
135 1207–18 living cells Nature 171 720–4
[7] Fox C A and Weinreich M 2008 Beyond heterochromatin [28] Davies H G and Wilkins M H F 1952 Interference microscopy
SIR2 inhibits the initiation of DNA replication Cell Cycle and mass determination Nature 169 541
7 3330–4 [29] Lee P A and Ramakrishnan T V 1985 Disordered electronic
[8] Ellis R J 2001 Macromolecular crowding: obvious but systems Rev. Mod. Phys. 57 287–337
underappreciated Trends Biochem. Sci. 26 597–604 [30] Schmitt J M and Kumar G 1996 Turbulent nature of
[9] Beuthan J, Minet O, Helfmann J, Herrig M and Muller G 1996 refractive-index variations in biological tissue Opt. Lett.
The spatial variation of the refractive index in biological 21 1310–2
cells Phys. Med. Biol. 41 369–82 [31] Bernstein C, Bernstein H, Payne C M, Dvorak K and Garewal
[10] Perelman L T et al 1998 Observation of periodic fine structure H 2008 Field defects in progression to gastrointestinal tract
in reflectance from biological tissue: a new technique for cancers Cancer Lett. 260 1–10
measuring nuclear size distribution Phys. Rev. Lett. [32] Dakubo G D, Jakupciak J P, Birch-Machin M A and Parr R L
80 627–30 2007 Clinical implications and utility of field cancerization
[11] Mourant J R et al 1995 Spectroscopic diagnosis of bladder Cancer Cell Int. 7 2
cancer with elastic light scattering Lasers Surg. Med. [33] Atkin W S and Saunders B P 2002 Surveillance guidelines
17 350–7 after removal of colorectal adenomatous polyps Gut
[12] Sokolov K, Drezek R, Gossage K and Richards-Kortum R 51 V6-9
1999 Reflectance spectroscopy with polarized light: is it [34] Butterly L F, Chase M P, Pohl H and Fiarman G S 2006
sensitive to cellular and nuclear morphology Opt. Express Prevalence of clinically important histology
5 302–17 in small adenomas Clin. Gastroenterol. Hepatol.
[13] Coffey D S 1998 Self-organization, complexity and chaos: the 4 343–8
new biology for medicine Nat. Med. 4 882–5 [35] Richter K, Nessling M and Lichter P 2007 Experimental
[14] Born M and Wolf E 1999 Principles Of Optics: evidence for the influence of molecular crowding on nuclear
Electromagnetic Theory of Propagation, Interference and architecture J. Cell Sci. 120 1673–80
Diffraction of Light vol 952 7th edn (Cambridge: [36] Damania D, Subramanian H, Tiwari A K, Stypula Y, Kunte D,
Cambridge University Press) p xxxiii Pradhan P, Roy H K and Backman V 2010 Role of
8
Phys. Biol. 8 (2011) 026012 P Pradhan et al
cytoskeleton in controlling the disorder strength of cellular [38] Wax A et al 2002 Cellular organization and substructure
nanoscale architecture Biophys. J. 99 989–96 measured using angle-resolved low-coherence
[37] Kopelovich L, Conlon S and Pollack R 1977 Defective interferometry Biophys. J. 82 2256–64
organization of actin in cultured skin fibroblasts from [39] Wax A et al 2005 Prospective grading of neoplastic change in
patients with inherited adenocarcinoma Proc. Natl Acad. rat esophagus epithelium using angle-resolved
Sci. USA 74 3019–22 low-coherence interferometry J. Biomed. Opt. 10 051601