Microfluidic Western blotting

Alex J. Hughesa,b and Amy E. Herra,b,1

a
Department of Bioengineering and bUniversity of California at Berkeley–University of California at San Francisco Graduate Program in Bioengineering,
University of California, Berkeley, CA 94720

Edited by David A. Tirrell, California Institute of Technology, Pasadena, CA, and approved November 7, 2012 (received for review May 7, 2012)

Rapid, quantitative Western blotting is a long-sought bioanalytical (e.g., BSA), incubation with antibodies to accomplish the prob-
goal in the life sciences. To this end, we describe a Western blotting ing step, and washing before obtaining assay readout. Several
assay conducted in a single glass microchannel under purely fundamental limitations impede performance, including slow
electronic control. The μWestern blot is comprised of multiple steps: mass transport. In fact, antibody probing often requires an
sample enrichment, protein sizing, protein immobilization (blot- overnight incubation period to compensate for diffusional limi-
ting), and in situ antibody probing. To validate the microfluidic as- tations on antibody equilibration with antigen captured on the
say, we apply the μWestern blot to analyses of human sera (HIV blotting material. Moreover, material and reagent consumption
immunoreactivity) and cell lysate (NFκB). Analytical performance is extensive. A single 12-lane Western blot requires ∼300 mL of
advances are achieved, including: short durations of 10–60 min, mul- buffer and, most importantly, 1 μg of each detection antibody per
tiplexed analyte detection, mass sensitivity at the femtogram level, analyte of interest. Because of the power of the assay and in light
high-sensitivity 50-pM detection limits, and quantitation capability of these deficiencies, Western blotting requires continued in-
over a 3.6-log dynamic range. Performance gains are attributed to novation to improve throughput, minimize resource use, and
favorable transport and reaction conditions on the microscale. The advance analytical sensitivity and dynamic range (7).
multistep assay design relies on a photopatternable (blue light) and Alternative single-step assays have emerged to overcome con-
photoreactive (UV light) polyacrylamide gel. This hydrophilic poly- ventional Western blotting drawbacks, yet none afford the speci-
mer constitutes both a separation matrix for protein sizing and, after ficity of the two-pronged blotting assay (8–13). Primarily relying on
brief UV exposure, a protein immobilization scaffold for subsequent spatially encoded antibody-probing, single-readout assays have
antibody probing of immobilized protein bands. We observe protein been developed for ultrahigh sensitivity yet remain inherently
capture efficiencies exceeding 75% under sizing conditions. This vulnerable to nonspecific bias. Cross-reactivity is especially chal-
compact microfluidic design supports demonstration of a 48-plex lenging for biological matrices and availability of specific anti-
μWestern blot in a standard microscope slide form factor. Taken bodies can be limiting (10, 14, 15). The inability of single-step
together, the μWestern blot establishes a foundation for rapid, tar- assays to guarantee specificity in complex biological samples
geted proteomics by merging exceptional specificity with the underpins the need for targeted immunoblotting methods coupled
throughput advantages of multiplexing, as is relevant to a broad to protein separation processes (16).
range of biological inquiry. Recent innovation in protein analysis tools recognizes the
crucial protein separation stage and seeks to retain separation-
|
immunoblotting medical diagnostics | protein microarrays | systems based information. In an approach using slab gels, Ciaccio et al.
|
biology electrophoresis achieved remarkable miniaturization and scale-up of the West-
ern blot workflow through a novel combination of highly multi-
plexed fluid handling and a large format slab gel (14). However,
W estern blotting is an indispensable analytical tool, benefit-
ing applications from clinical diagnostics to fundamental
questions in the life sciences (1–3). The broad relevance of
the approach conceded a loss in PAGE separation resolution
because of elimination of sample stacking by isotachophoresis
blotting stems from its highly specific results. Unlike separations (ITP), a sample preconcentration step standard in conventional
or immunological probing alone, blotting reports two physico- slab-gels that requires pore size and buffer chemistry dis-
chemical characteristics: molecular mass and immunoaffinity. continuities in the polyacrylamide sieving matrix. Furthermore,
Because of this specificity, numerous immunoblotting variants the workflow retained the conventional membrane transfer and
have emerged for measurements of proteins to RNA to bio- antibody-probing paradigm. In a commercial approach using
molecular interactions (2, 4–6). Nevertheless, despite being an capillary electrophoresis, downstream membrane electrotransfer
information-rich and widely used assay, the Western blot has was replaced with photoactivated capture of proteins onto the
crucial inadequacies. In particular, limitations in data density and inner wall of the capillary (17, 18). The capillary platform
throughput hinder progress for emerging pursuits, including sys- streamlined and automated Western blotting, but suffered from
tems approaches to biology. A major shortcoming of Western low protein-capture efficiencies of ∼0.01% (104-fold lower than
blotting is the resource-intensive nature of the assay, being com- membrane electrotransfer) and 3- to 5-h run times. In our own
prised of several steps requiring disparate pieces of equipment approaches using microfluidics, integration overcame some of
(Fig. 1A). the macroscale shortcomings, but we accepted either substantial
In the first stage of Western blotting, protein sizing, samples complexity in interfacing and device architecture or did not im-
are analyzed by denaturing PAGE. During PAGE, proteins plement the most widely used separation approach—protein
electromigrate through a polyacrylamide sieving gel, allowing sizing—as the first assay stage (7, 19).
determination of molecular mass. Once protein size is de-
termined, the protein separation is incubated with antibodies
(probing), thus allowing detection of interactions. Target identity Author contributions: A.J.H. and A.E.H. designed research; A.J.H. performed research; A.J.H.
is then established by linking immunoaffinity information to contributed new reagents/analytic tools; A.J.H. and A.E.H. analyzed data; and A.J.H. and
molecular mass. Although conceptually straightforward, sub- A.E.H. wrote the paper.

stantial preparation and manual intervention are required. To The authors declare no conflict of interest.
prepare for probing, the sized proteins are transferred from the This article is a PNAS Direct Submission.
sieving gel to a blotting membrane. The blotting membrane Freely available online through the PNAS open access option.
immobilizes the protein separation and is then subjected to 1
To whom correspondence should be addressed. E-mail: [email protected].
Downloaded by guest on April 28, 2021

a series of handling steps, including blocking of nonspecific This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
membrane interactions through coating with a dummy protein 1073/pnas.1207754110/-/DCSupplemental.

21450–21455 | PNAS | December 26, 2012 | vol. 109 | no. 52 www.pnas.org/cgi/doi/10.1073/pnas.1207754110

 A B realized previously, including development of rapid 48 concurrent
μWesterns on a standard microscope slide footprint, multiplexed
analysis of three protein targets per blot, and quantitation over
a linear dynamic range of 3.6 logs with 50-pM lower limits of
detection. We apply the μWestern to multiplexed protein analyses
of complex proteinaceous samples, including crude cell lysate and
crude human sera. Results suggest that purely microfluidic tech-
nologies are a viable means to imbue core analytical tools with
automation, quantitative capability, and speed; thus paralleling
advances that have positioned protein microarrays for high-
throughput proteomics duty.
Results and Discussion
μWestern Device Design and Assay Operation. Performance advan-
ces in slab gel-based Western blotting are limited by fundamental
processes, including high transport timescales, variable blotting
transfer, and diffusion-limited kinetics (Fig. 1A). To overcome
these limits on macroscale performance, we explored a microscale
Western blot (Fig. 1B, μWestern blot). We hypothesized that
miniaturization applied in conjunction with photopatternable,
light-responsive polymers would enable high analytical perfor-
mance. We used this combined approach to integrate the distinct
stages of the canonical Western blot into a simple, passively actu-
ated microchannel. Specifically, the μWestern blot is comprised of
isotachophoretic sample stacking during sample injection, weight-
based separation of denatured protein analytes through the widely
used SDS-PAGE, and in situ immunoblotting with fluorescently
labeled primary and secondary antibodies (Fig. 1B). We first de-
scribe the microfluidic design strategy followed by details of the
μWestern assay operation, experimental observations, and result-
ing performance.
We used single glass microchannels to facilitate high-density
integration of μWestern blots within the footprint of a standard
microscope slide (Fig. 1C). As such, each μWestern is comprised of
a pair of access wells linked by three parallel microchannels
(technical triplicates, each 70-μm wide × 10-μm deep) with elec-
trical connectivity provided by an electrode array. To facilitate
C integration of the μWestern stages in the single microchannel
geometry, we designed a unique photoactive gel with tunable po-
rosity (PACTgel) that is both photopatternable and light-re-
sponsive. Photopatterning of the sieving gel is needed for
reproducible control of the gel interface position along the channel
axis and, hence, assay repeatability (Fig. 1B). A light-responsive
material functionality was sought to allow the resulting photo-
patterned gel to switch from a molecular sieving gel during siz-
ing to a blotting polymer for subsequent antibody probing. To
introduce these new functionalities, we used two spectrally
distinct chemical mechanisms. First, a riboflavin-driven photo-
Fig. 1. Single-microchannel μWestern assay design enables high device
density formats. Aspects of scale, reagent use, blotting efficiency, and probe-
polymerization strategy was used to photopattern the material
binding kinetics are illustrated by comparative schematics for the conven- (using 470-nm light, see Materials and Methods). Second, a spec-
tional (A) and μWestern (B) assays (δ indicates a diffusion boundary layer trally distinct benzophenone-driven photo-immobilization strategy
thickness). The microfluidic workflow is comprised of: (i) analyte stacking allows the sieving matrix to form covalent bonds with proteins in
and SDS-PAGE within the PACTgel matrix; (ii ) band capture (“blotting”) the gel (via UV excitation) (Fig. 1B and Materials and Methods).
onto the benzophenone-decorated PACTgel in response to UV light (as Benzophenone is incorporated in the polyacrylamide gel via
MEDICAL SCIENCES

opposed to transfer to a separate sheet of hydrophobic material in con- a methacrylamide comonomer (BPMAC; N-[3-[(4-benzoylphenyl)
ventional Western blotting); (iii) removal of SDS by brief electrophoretic formamido]propyl] methacrylamide). As a corollary outcome, the
washing and electrophoretic introduction of fluorescently labeled primary
light-activated mechanism of the benzophenone-functionalized,
and (optionally) secondary detection antibodies specific to the target.
Finally, excess probe is electrophoretically driven out of each device and
hydrophilic gel means that no separate blocking steps are needed
peak intensities determined by fluorescence micrograph analysis. (C) Mod- after protein immobilization and before antibody probing. Owing
ular interfacing of standard microscope slide-sized chips with a scal- to the PACTgel functionality, the μWestern design is compatible
able electrode array accommodating 48 blots per chip in triplicate (144 with 48-sample throughput within a standard 1-inch by 3-inch
microchannels). microscope slide footprint (Fig. 1C).
ENGINEERING

Stage 1: Single Microchannel Protein Sizing. In seeking a broadly

In this study, we introduce scalable, automated μWesterns relevant protein separation assay, we adapted the widely used
uniting protein sizing and antibody probing in a single microfluidic Tris•glycine SDS-PAGE system of Laemmli (1, 20). Importantly,
Downloaded by guest on April 28, 2021

platform. The precision and control offered by microfluidic in- the assay requires a transient-ITP buffer arrangement and a
tegration and photoresponsive materials achieves advances not large-to-small pore-size discontinuity a short distance along the

Hughes and Herr PNAS | December 26, 2012 | vol. 109 | no. 52 | 21451
 distributions spanning hundreds of micrometers to several milli-
meters can be stacked into final zones of equivalent width in the
completed separations, achieving stacking factors over a broad
range that can be tuned simply by manipulating the length of the
open channel region upstream of the separation gel. In seeking to
understand the sample enrichment limits in this system, the length
of the sample loading region was extended to 4 mm and the sample
loading times were increased to 30 s, yielding as high as ∼100-fold
stacking factors (Fig. S1). These ITP-enabled enrichment and
stacking qualities are important to obtaining high-separation res-
olution in subsequent sizing, as well as competitive limits of antigen
detection. Interestingly, we observe that band ordering in the ITP
stack is not necessarily governed by molecular mass, thus dynamic
band reordering is often detected during the brief transition from
ITP to PAGE (see band “x” in Fig. 2A).
Given this observation, we next sought to confirm that the
separation mechanism in the PACTgel is indeed governed by
protein molecular mass differences. Analysis of protein migration
in the PACTgel yields a log-linear molecular mass versus migra-
tion distance relationship over the 20- to 150-kDa analyte range
(R2 > 0.98 for ladder proteins) (Fig. S2), thus confirming this
characteristic of SDS-PAGE (21). Protein peaks with molecular
mass differences of >19% were resolvable (separation resolution
Rs ≥ 1) (Fig. S3), with resolution similar to conventional slab-gel
Western blotting. Although only uniform pore-size gels are studied
here, the PACTgel pore size distributions are tunable, thus
allowing enhanced resolution over specific weight ranges of in-
terest (22–24). Importantly, the first sizing stage is observed to
complete in compact 3-mm separation distances in separation
times of ∼60 s. Consequently, the duration of the sizing step in
μWestern blotting compares favorably to the 40–90 min required
for macroscale protein sizing. The high-performance of protein-
sizing benefits from the favorable scaling of electrophoretic
transport with miniaturization.

Fig. 2. Compact μWestern with integrated high-resolution SDS-PAGE, blot, Stage 2: In-Chip Protein Blotting by Photocapture. Next, to permit
and detection. (A) SDS-PAGE of fluorescently labeled six protein ladder antibody-based probing of the sized protein bands, we immobilized
(black), complete in 60 s (4× magnification; band weights are 155, 98, 63, 40, each species on the PACTgel polymer using UV irradiation of the
32, and 21 kDa). Channel aspect ratios are adjusted to produce gel-like images entire separation channel (Fig. 2B and Fig. S4). UV exposure
(see dimensions). (B) Capture efficiency of BSA (± SD, n = 3) for PACTgels activates benzophenone groups to undergo hydrogen abstraction
fabricated chemically or photochemically. (C, Left) Multiplexed μWestern
and covalent coupling to nearby biomolecules by a free radical
readout (red) in 40-min total assay times using primary antibodies for (i) OVA,
and (ii) β-gal, OVA, and TI; all at 1 μM. (Right) Fluorescence micrographs and
mechanism (19, 25). First, we compared protein capture efficien-
plot of SNR (± SD, n = 3) for electrophoretic introduction of red fluorescent cies for both chemically and photochemically fabricated PACTgels
primary antibody (Ab*) to OVA band at 4 min total assay time (arrow). (D) (i.e., gels without and with riboflavin) to determine the impact of
Forty-eight concurrent μWesterns of the four-protein fluorescent ladder the riboflavin-driven polymerization mechanism on the UV-initi-
probed for OVA and β-gal targets (1 μM each) with unlabeled primary and red ated protein capture. In both cases, characterization of fluores-
fluorescent secondary antibodies in 60-min total assay time. At top, total cence retained on the PACTgel after photocapture and
injected (“stack”) fluorescence on weight marker spectral channel at the end electrophoretic washout reveals a sigmoidal dependence of fluo-
of the ITP phase of SDS-PAGE acts as loading control. rescently labeled BSA capture efficiency on UV exposure time
(Fig. 2B). The capture time courses for chemically and photo-
chemically initiated PACTgel formulations show ∼100% BSA
microchannel axis to yield high-resolution protein sizing (Fig. 2A).
capture when UV exposure times are > 45 s. As negative controls,
For the gel discontinuity, we used an open channel-to-7.5%T PACTgel formulations lacking benzophenone were studied and
sieving PACTgel interface at ∼400 μm into the microchannel (SI exhibited negligible protein blotting (Fig. 2B).
Materials and Methods). Transparency mask lithography of the Interestingly, PACTgel protein capture efficiencies are signifi-
photopatternable PACTgel was used to define the gel interface. cantly higher than those previously reported by our group for gels
Photopatterning yields fine control of the interface position (co- operating under nondenaturing isoelectric focusing conditions
efficient of variation, CV, of 3.5%, n = 60). During the ITP (1.3–13%) (19). We hypothesized that the capture efficiency im-
stacking phase, a diffuse plug of protein injected at the micro- provement for SDS-PAGE stems from the denatured state of the
channel entrance is electrophoretically compacted into a ∼200- target proteins. Denaturation likely exposes buried protein residues
μm zone before electromigration across the sharp sieving gel in- to the sieving matrix, thus promoting hydrophobic interactions
terface. As is also shown in Fig. 2A, protein electromigration between the unfolded analytes and the PACTgel benzophenone
through the gel interface transitions ITP to SDS-PAGE, as the groups (19). Furthermore, the reduced requirement for protein
protein stack slows down substantially and the trailing glycine solubilizing agents (especially detergents) in SDS-PAGE compared
electrolyte overspeeds the stack (1, 20). In characterizing ITP, we with isoelectric focusing likely reduces steric (among other) barriers
observe reduced injection dispersion and >twofold sample to efficient coupling between proteins and the PACTgel. Second,
we observed low capture efficiencies for exposure times of <20 s.
Downloaded by guest on April 28, 2021

stacking factors. The total mass of injected sample scales linearly

with the length of the sample loading region. Thus, initial sample We hypothesized that this is an initial inhibitory phase, perhaps

21452 | www.pnas.org/cgi/doi/10.1073/pnas.1207754110 Hughes and Herr

 caused by scavenging of reactive benzophenone sites by dissolved with immobilized protein targets. Building on the single protein
oxygen before a productive phase of analyte capture onto the probing result, Fig. 2C reports simultaneous probing of three
PACTgel (26). In sum, we conclude that photopatterning using analytes in a single microchannel using a three-antibody mixture
riboflavin and blue light does not compromise subsequent UV- applied in one electrophoretic step. Again, negligible off-target
mediated protein photocapture by the benzophenone-decorated signal was detected. As the number of antibodies present in the
polyacrylamide gels. probing mixture increases (see five-probe mixture in Fig. S5),
Next, we examined capture efficiencies for a wider range of multiplexing on one spectral channel becomes limited by in-
proteins under the 45-s UV exposure conditions that lead to highly creasing background arising from overlap of minor components in
productive capture. We now considered only the photochemically each target protein. The use of spectrally distinct dyes is supported
fabricated gels of interest. We observed appreciable immobiliza- by the current platform, pushing the conceptual multiplexing limit
tion of a set of test proteins on the PACTgels: 97.5 ± 0.7%, 93.1 ± to ∼5n analytes per blot, where n is the number of dyes that can be
3.4%, and 75.2 ± 0.8% for β-galactosidase (β-gal, 116 kDa), oval- imaged without significant spectral bleed-through. Between-de-
bumin (OVA, 45 kDa), and trypsin inhibitor (TI, 21 kDa), re- vice peak area CVs for identical samples probed simultaneously
spectively (all ± SD, n = 3). These capture efficiencies rival for OVA and β-gal were 25% each, with the ratio of peak areas
conventional electrotransfer blotting efficiencies on polymer varying with a CV of 14.7%. Use of internal migration controls
membranes (27). Furthermore, the nearly complete protein cap- allows data comparison across channels and chips.
ture in the photoactive bulk polyacrylamide gels is an orders-of- Next, we sought to develop an ultrarapid μWestern blot. In
magnitude improvement over reported capture efficiencies for designing this assay, we used dynamic imaging of fluorescent an-
photoactive inner capillary surfaces [0.01% for GFP (17)]. We tibody probe accumulation at the site of captured analytes (Fig.
observed similar capture efficiencies for protein concentrations up 2C). This dynamic imaging mode yields a primary antibody prob-
to 100 pg·nL−1 (∼109 proteins nL−1 or 0.1 mg·mL−1), a capacity ing time constant of 8.2 min for a 1-μM OVA band captured on the
attributed to an estimated benzophenone site density of ∼1012 nL−1 PACTgel after SDS-PAGE. A probe band signal-to-noise ratio
of the gel structure. Thus, the 3D reactive gel offers a high volu- (SNR) of >10 was recorded for a 10-min total assay time. We
metric density of binding sites. This large number of binding sites ascribed the rapid probing kinetics to the electrokinetic through-
distributed throughout the channel volume enables efficient pho- pore probe delivery strategy that leverages favorable overall re-
tocapture in < 60 s. In contrast, membrane electrotransfer in action kinetics that are minimally impeded by surface boundary
conventional bench top Western blotting requires 90 min to com- layer diffusion resistances (19). Dynamic monitoring of target peak
plete. This rapid capture kinetic of the μWestern is critical, first to SNR enables the assay to trade-off between readout signal and
yield low overall assay durations. Second, the rapid kinetic is es- assay time. Finally, we sought to introduce a broadly relevant
sential to maintain performance; this is because diffusional band μWestern using unlabeled primary antibodies and fluorescently
broadening erodes both SDS-PAGE separation resolution and labeled secondary antibodies, as is common in conventional slab-
analytical sensitivity of subsequent probing given the small inter- gel Western blotting (Fig. 2D). In the 48-channel μWestern device,
peak displacement distances and peak widths in the microfluidic we probed both OVA and β-gal with unlabeled primary antibodies
format (19, 28). specific to the target. We also probed with secondary fluorescently
labeled antibody probes and obtained the endpoint readout after
Stage 3: Probing. The final assay stage is in situ antibody probing of electrophoretic washout of excess probes. This dual antibody ap-
the immobilized, sized proteins. In the μWestern, probes are proach gave higher SNRs over the full separation range (19) but
electrophoresed through and along the length of the microchannel still maintained relatively rapid assay times of less than 60 min.
by an applied electric field. The approach ensures that probes
sample positions along the entire length of the protein-decorated High-Sensitivity and Quantitative Measurements for Proteinaceous
PACTgel. Results for probing of OVA with a red fluorescently Biospecimens. Next, we sought to ascertain the robustness of
labeled antibody are shown in Fig. 2C. Antibody was electropho- the μWestern to analysis of biological samples. In the first study,
retically introduced 4 min after the start of the assay and required we applied the μWestern to measurement of the transcription
∼1 min to migrate through the gel pores to the immobilized OVA factor NFκB (p105, p50) in lysate from an NFκB-transfected
band. We observed negligible red signal away from the OVA peak, 293T cell line (Fig. 3A). Western blotting is commonly applied to
suggesting that probe retention arises from specific interactions protein measurement in lysate (e.g., protein signaling studies).

Fig. 3. Validation of μWestern for cell lysate and

purified proteins. (A) Sixty-minute μWestern of 0.5
mg/mL transfected 293T lysate probed for NFκB with
MEDICAL SCIENCES

unlabeled primary and fluorescently labeled sec-

ondary antibodies (red). Untransfected negative
control lysate and loading controls (GAPDH and to-
tal injected fluorescence, “stack”) are included.
(Right) The corresponding conventional 6- to 8-h
Western blot readouts for visual comparison. Note
relative dimensions of the conventional blot. (B)
Forty-minute μWestern of purified HIV proteins (re-
verse-transcriptase, 200 nM; gp120, 200 nM; p24, 1
ENGINEERING

μM) after probing targets with fluorescently labeled

primary antibodies (red). (C) Standard curve for
gp120 over the 50 pM to 200 nM range (± SD, n = 3)
constructed from peak areas of the band indicated
Downloaded by guest on April 28, 2021

by an arrow in B. See Fig. S6 for standard curve of

the NFκB p105 peak indicated by an arrow in A.

Hughes and Herr PNAS | December 26, 2012 | vol. 109 | no. 52 | 21453
 An unlabeled primary and a fluorescently labeled secondary
antibody were used for immunoprobing. Assays of NFκB trans-
fected lysate and untransfected negative controls yielded similar
probing patterns in on-chip and conventional formats (Fig. 3A).
We included conventional GAPDH probing as μWestern loading
and biological controls; the total injected zone fluorescence of
the ladder proteins serves as a convenient alternative loading
control. With important implications to antibody screening
frameworks and systems biology, we observed total assay oper-
ation to consume <1 ng of each antibody, in contrast with ∼1 μg
consumed in conventional Western blotting.
In a second study to assess assay relevance, we applied our
μWestern assay to several purified HIV proteins (Fig. 3B). HIV
confirmatory diagnosis—presently a process that requires a cen-
tral laboratory and hours to complete—relies on Western blotting.
The μWestern and slab gel Western blot results agree to within
12% for the mass of the major bands of viral reverse transcriptase
and the envelope glycoprotein gp120 (Table S1). For the smallest Fig. 4. Sixty-minute μWestern for HIV antibody detection in human sera.
protein, capsid protein p24, a 25% error in the measured weights (A) Conventional confirmatory HIV diagnostic assay schematic. (B) Reactivity
is attributed to the reduced performance of SDS-PAGE at the of 1:100-diluted strongly reactive (++), weakly reactive (+), and nonreactive
lower mass end of the sizing range, which is incurred regardless of control (–) human sera to gp120 (200 nM) and p24 (1 μM) “bait” proteins
format. Nevertheless, precision in molecular mass prediction on- revealed by fluorescently labeled secondary antibody to human IgG (red). At
right, the conventional 6- to 18-h HIV Western blot, with gp120- and p24-
chip gives within-device CVs of <2% (n = 3 for each) across the
reactive bands indicated by arrows. The conventional blot employs whole
entire mass range. Minor bands for both gp120 (56 kDa; 40 HIV lysate, whereas the μWestern uses specific HIV antigens, accounting for
kDa) and p24 (49 kDa) observed only on conventional Western the additional reactive bands visible in the conventional blot.
blots are attributed to differences between the macro- and
microscale workflows, including differences in blotting effi-
ciency, in the SDS-PAGE and probing buffer systems, and in performed with 1:100 diluted human serum. Specific serum re-
the degree of analyte renaturation before immunoprobing. activity to each “bait” protein was determined using a fluo-
Potentially relevant to point-of-care μWestern applications, the rescently labeled secondary antibody directed to human IgG on
total wash and transfer buffer requirement is ∼300 μL, com- the PACTgel. The resulting dose–response was consistent with
pared with 300 mL consumed in the conventional assay. the expected antibody titer in each of three sera (strongly re-
In both of the above studies we sought to realize high-sensi- active, weakly reactive, nonreactive), in accordance with guide-
tivity measurement while enabling a capacity for quantitation in lines for determining HIV infection in humans (15).
μWestern blotting. For both NFκB and gp120 we observe Although promising, ongoing development of integrated serum
quantitative antibody signal readout over a linear dynamic range processing (31) and optimization of assay conditions is underway
of up to 3.6 logs, on par with expected performance for macro-
to reduce nonspecific background in the HIV μWestern. Electro-
scale counterparts (Fig. 3C and Fig. S6). Using ITP to obtain
a high stacking-factor enrichment, we measured a lower limit of phoretic transport of antibody probes through the dense PACTgel
detection of 50 pM for the latter analyte. This lower detection matrix produces nonspecific staining in the low-nanomolar probe
limit is comparable to enzyme-amplified chemiluminescent concentration range. This staining is strongly dependent on
readouts in conventional blots. Considered another way, the PACTgel pore size, and to a lesser extent on the degree of UV
detection capacity translates into measurement of 12 pg per 2-μL exposure during capture of separated antigens (Fig. S7). As the gel
sample or a total mass of 17 fg of gp120 per sample injection pore size decreases, large antibody probes can become irreversibly
volume. These results compare favorably to recent low-mass immobilized. Favorable to assay performance, however, we ob-
sensitivity (340 fg) slab gel Western blots (14). The mass limits served little impact of pore size on blotting background in the 7–
also suggest that we can detect the equivalent of ∼4,000 virus 8%T range that realizes optimal SDS-PAGE performance. A less
particles on a gp120 basis (4–35 copies per virion) or as few as dramatic increase in probing background as a function of UV ex-
∼20 particles for p24 (5,000 copies per virion) (29). As relevant posure time during antigen capture may be caused by decreased
to single-cell proteomics, the mass detection limit of ∼80,000 effective pore size because of the formation of accessory gel-gel
molecules we demonstrate is within the 104–106 molecule range cross-links, or because of the production of a hydrophobic ben-
expected for signaling proteins in single mammalian cells (30). zopinacol reaction product (26) that may encourage antibody
Building on the above HIV antigen study, we conducted a third immobilization within activated PACTgels. In any case, non-
study in which we sought to measure HIV antibodies directly in specific probing is only two- to threefold higher in BPMAC+
human sera. Currently, HIV diagnosis employs a conventional PACTgels than in BPMAC− control gels of the same nominal
Western blot as the final (confirmatory) assay, following a positive
porosity following UV exposure sufficient to cause near-quanti-
ELISA-based screening result (15). In a 6- to 18-h workflow, an
tative antigen capture.
HIV viral lysate is subjected to SDS-PAGE and immunoblotting
Specific to the HIV assay, we primarily attribute the background
(Fig. 4A). Diluted patient serum is incubated with a nitrocellulose
strip carrying the HIV protein bands. Any HIV-reactive anti- evident in the micrograph data of Fig. 4B to this basal nonspecific
bodies in the serum bind to specific HIV proteins on the strip. A probing caused by off-target immobilization of human IgGs pres-
positive result is indicated if two or more of the p24, gp41 and ent in serum at high concentrations (∼10–15 mg/mL), as well as
gp120/160 bands exhibit reactivity at least as intense as that of the by immobilization of the secondary antibody probe within the
p24 band on a blotting strip subjected to a weakly reactive control PACTgel matrix. To further enhance the assay performance, on-
serum (15). We translated the confirmatory HIV diagnostic assay going efforts are focused on minimizing nonspecific background
to the μWestern by assaying human sera against purified gp120 through on-chip sample cleanup, optimization of probe antibody
Downloaded by guest on April 28, 2021

and p24 HIV proteins (Fig. 4B). A mixture of these antigens was concentrations, and alternative approaches for probe introduction
subjected to the μWestern assay, and the first probing step and washout. After fully scrutinizing clinical specificity and sensitivity

21454 | www.pnas.org/cgi/doi/10.1073/pnas.1207754110 Hughes and Herr

 performance, we see potential for a 60-min stand-alone, rapid con- protein posttranslational modifications (32), characterization that
firmatory HIV diagnostic for near-patient application. is difficult with spatially encoded immunoreagents alone. Intrigu-
ingly, stable PACTgel analyte capture offers the ability to archive
Conclusions
and reanalyze μWestern chips. Taking these data together, we see
Protein measurement tools that are high-throughput yet afford promise for challenging analytical applications in systems biology,
high-specificity quantitation hold great promise for advances
cancer biology, and infectious disease diagnostics.
across a swath of inquiry, from systems biology to clinical med-
icine. The studies detailed herein introduce a fully microfluidic Materials and Methods
Western blot that modernizes and automates conventional Microfluidic Assay Instrumentation. Wet etching of microchannels in boro-
Western blotting. Because of the precision and control offered silicate glass, high-voltage control, fluorescence microscopy, and UV exposure
by microfluidic integration, we achieve advances in four key system details are in SI Materials and Methods. Microfluidic channels were
aspects of analytical performance: exceptional protein blotting functionalized with acrylate-terminated self-assembled monolayers and
efficiency with near complete analyte capture, accelerated run PACTgels fabricated as detailed in SI Materials and Methods.
times as all steps from sample separation to probing are com-
pleted in 10–60 min, small device footprint (800-fold smaller μWestern Protocol. Sample preparation, loading, separation, capture, and
device area compared with conventional gel lane), and superb probing along with specific imaging protocols are detailed in SI Materials
reagent economy with a 103-fold reduction in antibody and and Methods.
buffer requirements over conventional Western blot. In an ad-
vance over conventional capability, the μWestern yields quanti- Reagents and Samples. BPMAC monomer was synthesized in-house and
tative readouts from multiplexed analyte probing in a single verified by 1H NMR and mass spectrometry as previously described (19).
sample and from 48-blot microchips. As is important to myriad Purified proteins, antibodies and fluorescence labeling protocols are de-
protein measurements, the μWestern achieves desired limits of scribed in SI Materials and Methods.
detection across numerous specifications, including starting
sample concentration (low picomolar), starting sample total Benchmark Analysis. Conventional bench top Western blotting was per-
mass and volume (picograms per 2 μL of sample) and, finally, formed as detailed in SI Materials and Methods.
total detected mass (tens of femtograms of material per injected
volume). To ensure relevance, we validated our μWestern for Data Acquisition and Analysis. Whole-channel imaging at 10× was conducted
purified proteins, crude cell lysate, and crude human sera. via stitching of adjacent, overlapping CCD images in ImageJ (National
Looking forward, success in microfluidic integration of dis- Institutes of Health, Bethesda, MD) to produce full gel-channel images and
parate separation, blotting, and immunoprobing stages into electropherograms (SI Materials and Methods). Assay limits of detection
a unified workflow presents an exciting opportunity for “quan- were defined as the nominal antigen concentrations at which probed peak
titative western blot microarrays.” Such approaches may even- SNRs were no smaller than 5 under optimal fluorescence imaging conditions.
tually rival the throughput capacity of protein microarrays yet
retain a currently missing and crucial separation step. As rele- ACKNOWLEDGMENTS. A.J.H. is a Department of Defense National Defense
Science and Engineering Graduate research fellow and a 2013 Siebel Scholar.
vant to personalized medicine, our flexible electrophoretic A.E.H. is an Alfred P. Sloan research fellow (chemistry). This study was
blotting strategy is likely amenable to diverse probing (e.g., lec- supported in part by National Institutes of Health’s New Innovator Award
tin) and gel-staining strategies needed for characterization of 1DP2OD007294 (to A.E.H.).

1. Laemmli UK (1970) Cleavage of structural proteins during the assembly of the head of 17. O’Neill RA, et al. (2006) Isoelectric focusing technology quantifies protein signaling in
bacteriophage T4. Nature 227(5259):680–685. 25 cells. Proc Natl Acad Sci USA 103(44):16153–16158.
2. Towbin H, Staehelin T, Gordon J (1979) Electrophoretic transfer of proteins from 18. Fan AC, et al. (2009) Nanofluidic proteomic assay for serial analysis of oncoprotein
polyacrylamide gels to nitrocellulose sheets: Procedure and some applications. Proc activation in clinical specimens. Nat Med 15(5):566–571.
Natl Acad Sci USA 76(9):4350–4354. 19. Hughes AJ, Lin RKC, Peehl DM, Herr AE (2012) Microfluidic integration for automated
3. Kurien BT, Scofield RH (2009) Protein Blotting and Detection: Methods and Protocols targeted proteomic assays. Proc Natl Acad Sci USA 109(16):5972–5977.
(Springer, New York). 20. Ornstein L (1964) Disc electrophoresis—I Background and theory. Ann N Y Acad Sci
4. Southern E (2006) Southern blotting. Nat Protoc 1(2):518–525. 121:321–349.
5. Alwine JC, Kemp DJ, Stark GR (1977) Method for detection of specific RNAs in agarose 21. Herr AE, Singh AK (2004) Photopolymerized cross-linked polyacrylamide gels for on-
gels by transfer to diazobenzyloxymethyl-paper and hybridization with DNA probes. chip protein sizing. Anal Chem 76(16):4727–4733.
Proc Natl Acad Sci USA 74(12):5350–5354. 22. Hou C, Herr AE (2010) Ultrashort separation length homogeneous electrophoretic
6. Wu Y, Li Q, Chen XZ (2007) Detecting protein-protein interactions by Far western immunoassays using on-chip discontinuous polyacrylamide gels. Anal Chem 82(8):
blotting. Nat Protoc 2(12):3278–3284. 3343–3351.
7. He M, Herr AE (2010) Automated microfluidic protein immunoblotting. Nat Protoc 5 23. Hughes AJ, Herr AE (2010) Quantitative enzyme activity determination with zepto-
mole sensitivity by microfluidic gradient-gel zymography. Anal Chem 82(9):
(11):1844–1856.
3803–3811.
8. Lequin RM (2005) Enzyme immunoassay (EIA)/enzyme-linked immunosorbent assay
24. Lo CT, Throckmorton DJ, Singh AK, Herr AE (2008) Photopolymerized diffusion-defined
(ELISA). Clin Chem 51(12):2415–2418.
polyacrylamide gradient gels for on-chip protein sizing. Lab Chip 8(8):1273–1279.
9. MacBeath G, Schreiber SL (2000) Printing proteins as microarrays for high-throughput
25. Dormán G, Prestwich GD (1994) Benzophenone photophores in biochemistry. Bio-
function determination. Science 289(5485):1760–1763.
MEDICAL SCIENCES

chemistry 33(19):5661–5673.
10. Spurrier B, Ramalingam S, Nishizuka S (2008) Reverse-phase protein lysate microarrays
26. Schneider MH, Tran Y, Tabeling P (2011) Benzophenone absorption and diffusion in
for cell signaling analysis. Nat Protoc 3(11):1796–1808.
poly(dimethylsiloxane) and its role in graft photo-polymerization for surface modi-
11. Gaster RS, et al. (2009) Matrix-insensitive protein assays push the limits of biosensors
fication. Langmuir 27(3):1232–1240.
in medicine. Nat Med 15(11):1327–1332. 27. Jungblut P, Eckerskorn C, Lottspeich F, Klose J (1990) Blotting efficiency investigated
12. Rissin DM, et al. (2010) Single-molecule enzyme-linked immunosorbent assay detects by using two-dimensional electrophoresis, hydrophobic membranes and proteins
serum proteins at subfemtomolar concentrations. Nat Biotechnol 28(6):595–599. from different sources. Electrophoresis 11(7):581–588.
13. Zhu H, et al. (2001) Global analysis of protein activities using proteome chips. Science 28. Tia SQ, He M, Kim D, Herr AE (2011) Multianalyte on-chip native Western blotting.
293(5537):2101–2105. Anal Chem 83(9):3581–3588.
14. Ciaccio MF, Wagner JP, Chuu CP, Lauffenburger DA, Jones RB (2010) Systems analysis 29. Swanson CM, Malim MH (2008) SnapShot: HIV-1 proteins. Cell 133(4):742–, 742.e1.
of EGF receptor signaling dynamics with microwestern arrays. Nat Methods 7(2): 30. Milo R, Jorgensen P, Moran U, Weber G, Springer M (2010) BioNumbers—The data-
ENGINEERING

148–155. base of key numbers in molecular and cell biology. Nucleic Acids Res 38(Database
15. Centers for Disease Control (CDC) (1989) Interpretation and use of the Western blot issue):D750–D753.
assay for serodiagnosis of human immunodeficiency virus type 1 infections. MMWR 31. Apori AA, Herr AE (2011) Homogeneous immunosubtraction integrated with sample
Morb Mortal Wkly Rep 38(Suppl 7):1–7. preparation enabled by a microfluidic format. Anal Chem 83(7):2691–2698.
16. Ramaswamy A, et al. (2005) Application of protein lysate microarrays to molecular 32. He M, Novak J, Julian BA, Herr AE (2011) Membrane-assisted online renaturation for
Downloaded by guest on April 28, 2021

marker verification and quantification. Proteome Sci 3:9. automated microfluidic lectin blotting. J Am Chem Soc 133(49):19610–19613.

Hughes and Herr PNAS | December 26, 2012 | vol. 109 | no. 52 | 21455