Advances in Capillary Electrophoresis and the Implications for Drug Discovery

Abstract

Introduction

Many screening platforms are prone to assay interferences that can be avoided by directly measuring the target or enzymatic product. Capillary electrophoresis (CE) and microchip electrophoresis (MCE) have been applied in a variety of formats to drug discovery. CE provides direct detection of the product allowing for the identification of some forms of assay interference. The high efficiency, rapid separations, and low volume requirements make CE amenable to drug discovery.

Areas Covered

This article describes advances in capillary electrophoresis throughput, sample introduction, and target assays as they pertain to drug discovery and screening. Instrumental advances discussed include integrated droplet microfluidics platforms and multiplexed arrays. Applications of CE to assays of diverse drug discovery targets, including enzymes and affinity interactions are also described.

Expert opinion

Current screening with CE does not fully take advantage of the throughputs or low sample volumes possible with CE and is most suitable as a secondary screening method or for screens that are inaccessible with more common platforms. With further development, droplet microfluidics coupled to MCE could take advantage of the low sample requirements by performing assays on the nanoliter scale at high throughput.

1. Introduction

Capillary electrophoresis (CE), and its microfluidic counterpart microchip electrophoresis (MCE), have emerged as promising techniques with growing use in the pharmaceutical industry for characterizing biopharmaceuticals, quality control, and for drug discovery. In this review, we focus on the application of CE to drug discovery and screening. CE separates molecules based on their differential migration in an electric field. In free solution, the migration of a molecule in an applied field is dependent upon its charge and size, enabling separation by either property. Modification of the separation media can alter the separation selectivity, e.g. electrophoresis within gels enable separation on size only. Separated molecules are detected by a variety of methods including UV absorbance, capacitively coupled contactless conductivity, mass spectrometry (MS), and laser-induced fluorescence (LIF). The combination of separation and direct, on-line detection enable CE and MCE to be used for many kinds of assays useful for screening. Of relevance to drug screening, optical detection is well-suited for detecting enzyme activity and affinity interactions.

Compared to other drug screening platforms, such as fluorescent plate readers, CE offers a number of potential advantages including low sample volume requirements (nanoliter or less), rapid separations, and sensitive detection of analytes. Combining fractionation with direct detection of product can simplify assay development because it is not necessary to develop coupled reactions or engineer optical changes to generate a selective signal. CE also allows for resolution of confounding components in the assay such as interference from optically active test compounds [1–3], non-specific protein aggregation [3], and compound precipitation [3]. A comparison of methods for a kinase screen against fluorescent test compounds determined electrophoresis to be preferred over fluorescence polarization, amplified luminescent proximity homogenous assay, and enzyme fragment complementation for quantifying fluorescent inhibitors. The preference for CE was due to the tolerance of the CE assay to fluorescent compounds, the assay’s sensitivity, and comparatively low substrate and enzyme requirements. In the CE assay the fluorescent compound was tolerated because test compound was resolved from the substrate and product.[2] An example of similar benefits for a protein-protein interaction assay is illustrated in Figure 1.[3]

Figure 1.

Identification of assay interference in CE based on electropherograms. Electropherograms of Hsp70-488 and Bag3 interaction in the presence of (A) complex inhibitor epigallocatechin gallate, (B) hematoxylin which caused aggregation identified by loss of signal and sharp, unexpected peaks, (C) fluorescent test compound calcein causing optical interference.[3] Reprinted with permission from reference [3]. Copyright 2013 American Chemical Society.

MCE systems have been commercialized for screening. Perhaps the most popular system is the LabChip instrument (PerkinElmer), which uses vacuum to pull sample from a multiwell plate (MWP) into the microfluidic separation channel with fluorescence detection. This system has been used to screen 10 × 384 well plates in 10 h.[4] This platform has been applied to diverse targets including kinases, phosphatases, proteases, phosphodiesterases, epigenetic targets, and nucleic acid binding proteins.[1,2,4–9] Previous reviews on the use of CE in drug discovery have focused on the use of this system and specific enzyme target assays.[1,10]

Aside from this impressive technology, a variety of strategies for improving throughput and sample requirements have emerged, expanding the possibilities for using CE. In this review, we discuss new instruments and methods that have the potential to improve throughput and reduce sample consumption of CE or MCE for screening. Because one of the biggest limitations to throughput is sample injection onto the chip or capillary, many of these studies involve improvements in injection method. We also review assays, such as affinity assays for protein-protein interaction, where CE may be useful.

2. Improvements to Throughput

High throughput screening (HTS) using MWP with optical readouts can perform > 10⁴ assays per day. The throughput of CE and MCE assays is limited by the separation time required to resolve the molecules of interest. As commonly practiced, CE separation times are a few minutes per sample; however, it is possible to achieve separations in < 1 s. Throughput may also be improved by running separations in parallel or by multiplexing the assay.

2.1. Rapid separations

Strategies for increasing the speed of a separation can be appreciated with a brief review of electrophoresis principles. The migration time of an analyte (t_mig) in an electrophoresis separation channel with length (L) is dependent upon its electrophoretic mobility (μ_ep) under an applied voltage (V):[11]

$$t_{\text{mig}}=L^2/(V\times {\mu }_{\text{ep}})$$ (1)

Therefore, to decrease the separation time the applied voltage can be increased or the separation length can be decreased. For example, in an enzyme inhibition assay of metalloproteinase performed on a commercial CE instrument the separation time was decreased from about 250 s to 70 s by simply reversing the polarity of the applied electric field and injecting from the ‘outlet’ side of the capillary, near the detection window, resulting in a shorter separation length.[12] In general however, commercial CE instruments have limited accessible separation lengths so that it is difficult to reduce separations to shorter than this. Some custom-built CE systems have allowed sub-second separations;[13,14] however, with one exception [15,16], they have not been investigated for screening.

MCE allows for use of shorter separation channel lengths and smaller internal diameters than commercial CE systems. Increasing the applied voltage over a given separation length can increase the separation speed; however, this approach is limited because eventually Joule heating becomes significant enough to create mixing effects that destroy the separation. Effective heat dissipation can be achieved by using liquid cooling or by lowering separation channel diameters. These strategies enable higher voltages and faster separations. MCE with micrometer dimension channels enable separations in microseconds to seconds.[17] Decreasing the separation channel internal diameter places greater demands on the detector to achieve adequate sensitivity. LIF detection is often used when low internal diameters are employed because of the inherent sensitivity of fluorescence detection. Despite inherent challenges in coupling electrophoresis to MS recent advances in MS sensitivity[18] may eventually allow detection by this label-free, highly selective detector.

Another approach to improving throughput is to perform multiple injections rapidly so that multiple separations are overlapping in the separation channel at one time. This technique, which requires proper spacing of the injections relative to the separation times of peaks of interest, increases throughput by eliminating time between runs and taking advantage of the time between resolved peaks in a single separation. In one study, the total analysis time per sample was reduced by about half with sequential injection.[19] This approach is best used with CE where the separation time tends to be longer than MCE.

2.2. Sample Introduction

Although very fast separations are possible by CE and MCE, sample introduction and injection are critical parameters to achieving high throughput. Rapid separations in short, low volume separation channels require special considerations for sample injection to avoid overloading the separation channel with sample. Typical injection volumes are less than 1% of separation channel volume, corresponding to 120 pL volume on a 2 cm long x 20 μm tall x 30 μm wide channel. Rapid, small volume injection techniques include gated injection[20], optically gated injection[13,14,21,22], flow-gating interfaces[23], and spontaneous injection[24–26]. These methods allow adequate control to inject small volumes onto the channel and therefore achieve high quality separations in < 10 s; however, commonly used designs require that much larger samples (microliters) be loaded into a sample reservoir mounted on the chip. This approach therefore requires more sample than necessary for the actual electrophoresis separation and is not compatible with rapidly changing from one sample to the next. Recent advances in sample introduction methods are showing potential for improved throughput of screening.

2.2.1. Direct Injection from Multi-well Plate

Most screens are performed from MWP. Transferring samples rapidly and in series from the MWP to MCE requires specific fluidic handling components. The LabChip system uses “sippers” (capillaries attached to the chip) to dip into the MWP and pull as little as 10 nL of sample into the chip through a vacuum system. Interestingly, this system does not use the high efficiency injection methods described above. Instead, vacuum is used to pull samples into the separation channel. Therefore, a limitation of this commercial system is that the vacuum-induced flow within the separation channel causes band broadening, limiting the separation efficiencies that can be achieved. Nevertheless, this is the only commercial system that allows robust sampling for thousands of assays with MCE separations. Furthermore, high Z′-scores (>0.8), a measure of a screen’s data quality, have been reported using this platform.[1,7]

When assay reactions are performed in a MWP, the system does not reduce sample requirements even though only small volumes are removed from the MWP. The use of microfluidics, however, does allow enzymatic reactions to be performed on-chip. In on-chip mode, compounds are sipped from MWP and enzyme reagents are added in channels so that reactions occur on-chip. In one kinase screen, on-chip reactions reduced enzyme consumption 7-fold.[4] Due to short on-chip incubation times, however, the on-chip format requires high concentrations of enzymes and is limited to enzymes with rapid turnover (> 30% in 1 min).[6]

2.2.2. Injection from Small Sample Volumes

To reduce sample volumes required for injection, new designs for picoliter injections from low sample volumes have been developed.[24,27] One platform utilizes a “Slipchip.”[28] This design consists of two plates; one contains sample well and a discontinuous separation channel and the other contains a small volume sample well aligned with the bottom of the sample well on the other plate. The plates are aligned such that by moving one plate relative to the other discrete sub-nanoliter volume samples can be formed in the small volume sample well and alignment of this well with the separation channel allows for subsequent injection. The design is also parallelized for analysis from 10 discrete samples with 30 parallel separations.[27] Changing samples beyond the parallelized number, however, currently requires manual manipulations. In another platform, an array of nanoliter sample droplets was covered with immiscible oil and picoliter injections were achieved spontaneously by surface tension when the capillary tip was removed from the sample droplet. An array of 25 samples were injected with RSDs for peak height and migration time < 5%.[24] Potential exists to scale both of these systems to higher sample numbers to achieve higher throughput.

2.2.3. Integration of Segmented Flow

Another strategy for rapidly introducing new samples and miniaturizing sample requirements is use of droplet microfluidics. In droplet microfluidics, discrete aqueous samples are compartmentalized by an immiscible, often fluorinated, carrier fluid. Flow focusing and t-junctions can be used to make many droplets from one sample. [29–32] However, to make a few droplets from many samples, as is necessary in screening, different methods are required. In one approach, samples in MWPs are reformatted into segmented droplets of nanoliter volume inside tubing by using aspiration to sequentially draw up plugs of sample and carrier fluid.[15,16,33–36] Microfluidic droplets can then undergo further manipulations such as mixing,[37–41] merging,[42–46] splitting,[43,47–50] addition,[35,37,38,51,52] incubation and extraction,[15,53–59] which can enable entire assays to be performed at small scale. Indeed, this approach has been used with fluorescence detection to screen 704 compounds against protein tyrosine phosphatase and high-resolution dose response curves were obtained. Sample consumption per data point was reduced 25,000-fold.[60] Droplet microfluidics therefore is emerging as an exciting way to perform screens at much reduced volume and instrument overhead relative to MWP. Use of droplet microfluidics offers the potential to take better advantage of the throughput and miniaturization possible with MCE.

One hurdle associated with using droplet microfluidics in CE screening is automated injection of droplet samples into the CE separation channel. Injection of immiscible, non-conductive, segmenting liquid is not compatible with CE separation and has been observed to cause electroosmotic flow instability and plug formation in the separation channel leading to shorting and dielectric breakdown of the channel and device.[53,54,61] Therefore, extraction of aqueous sample from the segmented flow is necessary. Passive, active, whole, and partial droplet extraction and injection strategies have been reported.[15,53–59]

Active extraction uses an electric field to destabilize the fluorinated liquid-aqueous interface and merge the aqueous sample with a parallel aqueous stream providing robust and selective extraction.[55] Active extraction coupled to electrophoretic separation has yet to be reported.

Passive extraction is somewhat simpler to integrate, as it does not require external input. As seen in Figure 2, several passive droplet extraction strategies have been coupled to downstream CE separations. Pillar arrays have been applied for complete extraction of carrier phase[56] and injection of the whole droplet into an electrophoresis channel (Figure 2A).[57] Passive droplet extraction strategies often rely on surface modifications to extract the hydrophilic aqueous sample droplet from the hydrophobic carrier phase. In one strategy, whole droplets are injected into an electrophoresis channel by a multilayer device where a portion of the separation channel is open to the segmented flow channel. The droplet is simultaneously extracted and injected when the aqueous sample passes the junction and coalesces with the separation buffer (Figure 2B).[62] Interestingly, passive whole droplet extraction and injection assisted by a hydrophobic and oleophilic foam was also effectively coupled to capillary gel electrophoretic and free solution separations.[59] The limitation of whole droplet injection strategies is that large injection volumes limit the separation quality.

Figure 2.

Passive extraction and injection strategies for coupling segmented flow sample droplets to electrophoresis separations. (A) Pillar array extraction of oil.[57] (B) Intersecting segmented flow and separation channel geometry for simultaneous extraction and injection.[62] (C) Virtual wall used for extraction.[53] (D) Hybrid PDMS-glass device used for decoupling extraction and injection processes (http://pubs.acs.org/doi/full/10.1021/ac502758h).[15] Reprinted with permission from references listed in sub captions. Copyright Royal Society of Chemistry and American Chemical Society.

In another strategy, a microchip device with a ‘virtual wall’ between hydrophobic and hydrophilic channels was fabricated by derivatizing a channel surface to make it hydrophobic, allowing for aqueous sample to be discreetly injected (Figure 2C).[53,54] A hybrid polydimethylsiloxane (PDMS)-glass device passively extracted droplets from a segmented stream within the hydrophobic PDMS device into a hydrophilic capillary connected to a glass microchip for gated injection could be used to introduce small volume injections for electrophoretic separation (Figure 2D). For screening, droplets were catalogued by the presence of a positively charged, rapidly migrating analyte in every other droplet which was also used to visualize rinsing between droplets. High efficiency, rapid separations were reported with this platform (Figure 3).[15,16]

Figure 3.

Example of rapid enzyme assay separations achievable by MCE for screen of SIRT5 1,280 compounds. Separations of internal standard (R), product (P), and substrate (S) were achieved in 250 ms, #1117 is an enzyme inhibitor. Reprinted from reference [16], copyright 2016, with permission from Springer.

Several screens have been reported using this technology. Z′-scores of 0.8 were reported for protein kinase A and Sirtuin 5 screens using the hybrid PDMS-glass extraction approach. Throughputs of 0.2 to 0.5 samples/s have been reported with this platform.[15,16] To date, these approaches have only been used for up to 1,408 compounds. If these rates could be scaled to a large number of samples, then throughput could be 14,400 samples/8 h day on a single channel system.

2.3. Parallelization and Multiplexing Strategies

Operating CE or MCE in parallel can also improve the throughput. Challenges in developing platforms for running parallel separations include achieving multichannel detection, connections to peripheral power supplies, and attaining reliable simultaneous separations across parallel separation channels.[63] Many of these challenges were met when CE was being developed for DNA analysis. The necessity of high throughput sequencing led to using arrays of capillaries to run many parallel assays.[64] This concept has been adapted to drug discovery. A commercial instrumentation with a 96-capillary array was successfully applied to enzyme screening with UV absorption detection and throughputs of about 30 min per 96 samples without overlapping injections, high fields, or short capillaries.[65]

With MCE, parallel separation channel arrays can be made with small footprints at no additional fabrication cost. Up to 384 parallel separations have been reported for genotyping on a single microfluidic device.[66] Such highly parallel chips for screening have not yet been reported; however, the LabChip system can be used with 12 channels. Higher parallelization may be possible. A 36 channel microfluidic device with gated injection has been demonstrated for a model enzyme inhibitor screen with 36 parallel assays completed in 30 s.[67] Parallelization on chip with fast optically-gated injection has been used to perform for a high throughput enzyme assay with 4 parallel separations and 30 s separation time.[21] The throughput of a droplet extraction ‘virtual walls’ device was improved by parallelization with three extraction channels on one device and was demonstrated for an enzyme assay achieving throughputs of 120 samples in 10 min.[58] It seems likely that higher throughput could be achieved by parallelizing newer droplet extraction techniques.[15,16]

Another way to improve throughput is by test compound sample pooling. With sample pooling, versus assaying each test compound individually to find hits, a mixture of test compounds is assayed. If a hit is identified then each compound in the original test compound mixture is assayed individually. The low hit rates typical with screening allow far fewer assays to be done if sample pooling is used. Sample pooling for CE screening has been demonstrated with 10 test compounds assayed at a time.[68]

Multiplexing takes advantage of the separation power of CE to assay multiple targets simultaneously.[69,70] A four-plexed assay of protein kinases demonstrated simultaneous resolution of distinct substrates and products.[69] A study of protein-peptide interaction targets with src homology 2 (SH2) domains simultaneously assayed three proteins (Figure 4).[70] Such multiplexed assays also provide additional information about hit specificity by readily identifying selective and non-selective inhibitors.[69,70]

Figure 4.

Multiplexed assay of affinity interaction between SH2 domain proteins and phosphopeptides. Electropherograms identifying selective (middle) and non-selective inhibitors (bottom) of these interactions. Reprinted with permission from [70], copyright 2007, American Chemical Society.

2.4. Data Processing

A challenge of using CE for screening is that large numbers of electropherograms are generated which must be analyzed to determine peak areas used for quantification.[1] It has been shown that batch analysis of electropherograms allows for rapid data processing. For example, data processing software (available for download at http://kennedygroup.lsa.umich.edu/downloads/) allows for simultaneous analysis of hundreds of electropherograms to allow for data analysis on the time-scale of the rapid separations. In this strategy hundreds of electropherograms can be aligned based on a common peak, corrected for baseline drift, and peaks of interest can be defined.[71]

3. Targets

3.1. Enzymes

As previously mentioned, CE and MCE are amenable to performing enzymatic assays. Large screens of over 10,000 compounds have only been achieved using commercial microchip systems based on sipping from MWP.[8] The advent of droplet microfluidics interfaced to MCE may prove to have enough throughput, miniaturization, and robustness to provide a step forward for electrophoresis-based screening[15,16,53,54,58–62]; however, this has yet to be proven. The standard approach to enzyme assay by CE is to incubate enzyme with substrate and then separate substrate and product. The incubation may be performed in a MWP. Reduction in volume may be achieved by mixing on chip or in droplet samples as mentioned above. Other strategies, unique to CE, have also been demonstrated for analysis of enzyme modulation. These alternate methods include electrophoretically mediated microanalysis (EMMA), transverse diffusion of laminar flow profiles (TDLFP), and immobilized enzyme reactors (IMERs). Of the techniques, only “mix and separate” has been used for large scale screening with other techniques being demonstrated on small numbers of compounds (Table 1). Examples of screening using these techniques are described below.

Table 1.

Representative CE enzyme assay screens and demonstrations

Target	Assay Type	Compounds Assayed	Z-score	Reference
tyrosine phosphatase	commercial microchip platform	12,648	0.61	[8]
Sirt5	segmented flow coupled to MCE	1,280	0.8	[16]
glycerol kinase	EMMA	1	N.D.	[88]
aminopeptidase N	EMMA	30*	N.D.	[86]
neuraminidase	EMMA	24*	N.D.	[83]
β-N-acetylhexosaminidase	TDLFP	1	N.D.	[92]
four human kinases: GSK3β, DYRK1A, CDK5/p25, CDK1/cylcin B	TDLFP	13	N.D.	[93]
adenosine deaminase and xanthine oxidase	CE-based IMERs	20*	0.82, 0.74	[101]
L-glutamic dehydrogenase	CE-based IMERs	26*	0.95	[100]
glucose-6-phosphate dehydrogenase	CE-based IMERs	6	N.D.	[97]
alkaline phosphatases	CE-based IMERs	3	N.D.	[96]
acetylcholinesterase	CE-based IMERs	46*	0.9	[95]
angiotensin-converting enzyme	CE-based IMERs	34*	N.D.	[94]

Not determined (N.D.); electrophoretically mediated microanalysis (EMMA); transverse diffusion of laminar flow profiles (TDLFP); capillary electrophoresis-based immobilized enzyme reactors (CE-based IMERs)

^*includes crude product, natural extract, or chinese traditional herb

Sirtuins are deacetylases implicated in a number of human diseases including age-related diseases and various cancers making them biologically interesting targets.[72] Both fluorogenic and label free methods have been previously developed to screen modulators of sirtuin activity; however, many of these had drawbacks in regards to high-throughput screening applications.[73] One fluorogenic assay that has been used to assess the activity of Sirtuins 1, 2, and 3 involves the deacylation of a substrate peptide followed by a trypsin digest that releases a fluorescent molecule.[74] Although this format has high throughput, the design of these substrate peptides has resulted in false positive hits.[75] HPLC and MS assays that directly detect substrates and products have been developed; but, their application to high throughput screening campaigns is limited by relatively large sample consumption and analysis time.[76,77]

CE assays for measuring activities of various sirtuins have been reported.[7,16,78–82] An on-chip assay of SIRT1 inhibition by three known inhibitors was demonstrated with 10 min of on-chip reaction time.[78] In another study, a MCE SIRT5 assay was developed using a novel succinylated peptide substrate (GGQSLK[succ]FGKG) labeled with 5-carboxyfluorescein (5-FAM) at the N-terminus as the substrate for LIF detection. Assays were carried out using the previously discussed droplet system shown in Figure 2D.[16] In principle, the entire reaction can be performed in the droplet, as has been demonstrated with fluorescence[60] or MS[35] detection, further reducing sample requirements.

Several other CE assays that minimize sample consumption have been reported for small numbers of compounds. Performing enzymatic reactions within the capillary can minimize enzyme requirements. In EMMA, reactants are sequentially injected onto the capillary and mixed based on their differential electrophoretic mobilities allowing for the enzymatic reaction to occur on-line (Figure 5A).[12,83–88] In comparison to traditional enzyme screening assays, lower IC50 values have been determined using EMMA. Lower observed inhibition has been attributed to the decreased incubation time typical in EMMA.[86] EMMA was also demonstrated for a two substrate enzyme, glycerol kinase, with four reactant plugs: incubation buffer, enzyme and two distinct substrate plugs, mixed in-capillary.[88] Small-scale screening of Chinese herbs and other crude products against enzymatic targets by EMMA with UV detection demonstrated the utility of CE in screening of complex test compound mixtures, where potential optical interference is reduced compared to other optical platforms.[83–87] While EMMA requires very little enzyme per assay optimal conditions for the enzymatic reaction may be easier to achieve in well plate format. Off-line reactions can be completed in parallel, often with automated liquid handling, by carrying out the enzymatic reaction within the capillary, however, on-line incubation limits the throughputs achievable.

Figure 5.

Schematic of an EMMA strategy for enzyme inhibitor screening where (1) is depiction of sequential injection of plugs (2) polarity switching for mixing between plugs (3) separation of enzyme, substrate and product (A).[12] Schematic of CE-IMER for GAPDH.[97] Reprinted with permission from reference [12] and [97]. Copyright 2011 American Chemical Society and Elsevier.

In-line assays can also be performed with mixing of small volume pressure-injected plugs occurring by TDLFP.[89] Compared to EMMA, this method does not require prior knowledge of differential enzyme and substrate mobilities. The utility of this strategy first demonstrated with an assay of a farnesyltransferase target[90] and was later demonstrated for measuring inhibition of other enzymatic targets with UV detection.[91–93]

CE-based IMERs have also been demonstrated as useful for inhibition screening. Immobilized enzyme reactors can be fabricated within capillaries for on-line enzyme assays, saving enzyme in comparison to bulk assays (Figure 5B). CE-based IMERs for a wide variety of enzymes have been reported for characterizing enzyme inhibitors.[94–101] One study demonstrated the possibilities for multiplexing this CE-based IMER approach by fabricating an IMER of both immobilized adenosine deaminase and xanthine oxidase which was applied to screening of 20 natural extracts with a Z′-scores of 0.82 and 0.74, respectively. The separation was less than 3 min and the on-line incubation was 2.5 min. [101]

3.2. Affinity Interactions

Noncovalent binding of a partner can induce a mobility shift allowing electrophoretic separation. Many different assay schemes have been demonstrated for detecting and quantifying affinity interactions by MCE.[102–104] Noncovalent interactions assayed by CE include protein-nucleic acid [9,103,105], protein-peptide [70,68,106], protein-protein [3,107–110], nucleic acid-small molecule [111,112] and protein-small molecule interactions [113–117]. Of these, a “mix and separate” approach, sometimes called affinity probe CE [118], is perhaps the most amenable for high throughput screening. If the kinetics of dissociation are slow in comparison to the separation time, distinct free protein and protein complex peaks are observed enabling quantification of the bound to free ratio. Rapid separations are therefore preferable for both throughput and maintaining the complex of interest.

Affinity probe CE (APCE) on a microchip has been used to study protein-nucleic acid interactions.[9,104,105] In one example, the protein-nucleic acid interaction between human immunodeficiency virus 1 transactivator of transcription (Tat) and transactivation-responsive RNA (TAR) was studied using a commercial microchip platform. Inhibition of the Tat-TAR complex was demonstrated using a known inhibitor and dose dependent inhibition of the Tat-TAR complex peak was observed, suggesting the potential application of this platform to screening of affinity complexes.[9]

Protein-protein interactions (PPIs) represent a large class of targets that were, until recently, considered intractable. It can be difficult to predict small molecule modulators of these interactions as they often occur with high affinities over large, flat surfaces. Recent success in targeting these interactions has led to an increased interest in screening against them.[119,120] CE has potential as a useful screening technique because of the capability of high efficiency separation of large molecules.

CE has been used in studies of amyloid aggregation[121–123], protein-peptide interactions[70,68,106], and full length PPIs [3,107–110]. Most of these studies have been proof of concept assays; however, one assay did investigate a screen of 3,443 compound library against a target PPI. In this study, a fluorescently labeled heat shock protein 70 (Hsp70) in complex with its co-chaperone Bcl2-associated athanogene 3 (Bag3) was separated by CE. The resulting screen yielded a 1.4% hit rate. A 3.4% hit rate was achieved for the same screen using flow cytometry protein interaction assay (FCPIA). The lower hit rate in the CE assay was attributed to the identification of detection interfering compounds that may be false positives in other assays, such as fluorescent test compounds in the screening library and unexpected aggregation of proteins which could be readily identified by visual inspection of the electropherograms (Figure 1). Comparable Z′-scores of 0.78 for CE and 0.86 for FCPIA were found for these parallel screening platforms. The CE assay was not optimized for throughput however, as it used a commercial, single-channel CE system, and could only perform 220 assays/day. At this throughput CE is potentially more useful as a secondary screening platform.[3] In principle, such assays could be converted to parallel CE or MCE formats for higher throughput.

Although the APCE approach is most amenable to screening, CE-frontal analysis (CE-FA) has been used to measure affinity interactions. In CE-FA, relatively large volumes of equilibrated binding partner mixtures are injected. The large injection volume allows for maintenance of interaction during the separation because the free and complex zones are largely overlapping during the separation. The interaction between apoptosis regulatory, B-cell lymphoma 2 (Bcl-2) family proteins Bcl-X_L and BH3-interaction domain (Bid) was studied using a fluorophore-labeled peptide form of Bid. Quantification was achieved based on the plateau height of the free ligand. A high Z′-factor of 0.86 was determined for this method and a screen using sample pooling of 60 compounds with 10 compounds per sample was demonstrated with 10 min/sample separation times.[68]

A number of obstacles still exist for developing CE methods for screening affinity interactions; the protein complex may dissociate during the separation, many interactions induce only a small shift in mobility, and proteins tend to adsorb to the wall of the fused silica capillary which, depending on severity can cause shifts in migration time as well as loss of signal. The small mobility shift and protein adsorption issues must be overcome using conditions that maintain the non-covalent target interactions. One way to simplify method development for interacting proteins, therefore, is to covalently cross-link interacting proteins prior to electrophoresis. Once the proteins are cross-linked, the separation conditions can be optimized without concern of disrupting the interaction. For example, formaldehyde cross-linking followed by SDS-capillary gel electrophoresis separation (denaturing conditions) was shown to be effective for characterizing interaction affinities and small molecule inhibitors of PPI,[110] suggesting potential utility in screening.

Capillary coatings are useful for reducing protein adsorption. A number of recent studies have successfully characterized capillary coatings several of which are compatible with physiological pH and avoid adsorption that was observed on bare-fused silica capillary.[124–128] Unfortunately, a coating that prevents adsorption of one protein may not be as effective at preventing adsorption of a different protein[126,128] and replacing dynamic coatings between separations decreases throughput. Recently, a permanent coating for protein separation was observed to have an electroosmotic flow RSD of only 0.5% over one month.[128]

Assays for small molecule ligand interactions with proteins, peptides, and nucleic acids have been demonstrated by monitoring the mobility shift of the large molecule in the presence of small molecule.[111–117] Small molecule mobility shift assays can be challenging, with some binding events only inducing small changes in large molecule mobility.[114] Still, fragment-based screening has been demonstrated by affinity CE (CEfrag).[116] Successful fragment library screening relies on the ability to detect small amounts of inhibition as well as assay compatibility with high concentrations of test compound. A screen of heat shock protein 90α (Hsp90α) by a mobility shift assay with competitive inhibition was successful applied to a fragment library screen. In this method a small molecule affinity probe that is known to bind to the target molecule is used as an indicator of binding. When the affinity probe small molecule is unbound, it has a different migration time then when it is bound to its target. When another molecule competes with the affinity probe for target binding a mobility shift is observed and a hit is identified. For Hsp90α, using radicicol, a molecule known to interact with Hsp90α as the affinity probe, weaker affinity hits (>500 μM IC50 values) were identified using CE with UV detection versus with a fluorescence polarization assay in a screen of 609 compounds. The throughput of this screen was 100 samples/instrument/day using a 4-capillary instrument and a Z factor of 0.6 was determined.[117]

4. Conclusion

CE and MCE are capable of rapid analysis of small sample volumes making them natural tools for screening. The direct detection of product and resolving power allow for identification of assay interference, providing a powerful advantage relative to fluorescent assays that rely on a single channel optical readout. State of the art screening by CE is embodied in robust commercial microchip systems that sip samples from MWP and inject and separate on parallel arrays of channels. Such systems have been successfully used for a variety of enzyme assays.

Continued development of CE has revealed routes to potentially higher-throughput, lower sample consumption, and new types of assays. Microfluidic sample manipulation, especially droplet systems interfaced to MCE, has shown potential for high throughput, robust screening on nanoliter scale. Besides enzyme assay, affinity assays have been demonstrated that enable CE to be used for a wider range of targets. Although many new assays and formats using CE have been demonstrated, reports of using this new technology for screening large compound libraries are limited.

5. Expert Opinion

Although CE has demonstrated potential as a screening technique, high throughput screening is currently dominated by MWPs used with optical plate readers. This dominance is due to extremely high throughput and entrenchment of existing technology. Furthermore, the expertise of those researchers who develop screens often lies in chemical biology and engineering fluorescent or similar optical changes into desired biological assays rather than CE separations. Nevertheless, current commercial MCE platforms, such as the LabChip system, occupy an important niche in enabling screening when adequate fluorescent assays are challenging to develop, or as a secondary screening tool. Importantly, MCE and CE have demonstrated utility in identifying optical interferences including aggregation and fluorescent test compounds and lower test compound hit rates have been attributed to the identification of these interferences. The viability of CE has been shown with high Z-scores, demonstrating assay robustness on a variety of platforms. These advantages demonstrate the utility of further developing CE for screening.

While the applicability of CE to enzyme screening of diverse targets has been demonstrated using a variety of platforms, assays for affinity binding have yet to see widespread use in MCE screening. CE affinity assays offer significant advantages including sensitivity, ease of method development, multiplexing and information content. However, protein absorption to capillary wall and complex dissociation can complicate method development and have hindered progress in investigating affinity interaction targets by CE. As PPI and protein-nucleic acid interactions continue to grow as drug targets, affinity assays seem to be a promising area in need of further research to fully understand the utility of CE in these assays.

Although the current, commercially available, systems provide a robust platform for screening, they also impose limitations that do not fully support the possible high speeds of separation and therefore potential throughput of CE or MCE. Commercial instruments interface to MWP, containing microliters of each sample. While this is comparable to other screening technologies, interfacing to a MWP does not take advantage of the potential to analyze picoliter to nanoliter sample volumes in CE or MCE. Continued research into CE and MCE has identified promising in-line and microfluidic methods that use less sample volume and achieve higher throughputs. MCE has demonstrated advantages for improving throughput; therefore, the marriage of droplet microfluidics to MCE, so that entire assays are performed on a nanoliter scale, might be the most promising development. The coupling of this droplet technology with other techniques such as MS could also be furthered explored to take advantage of the power of the droplet technology and further the information content of the assay. Current droplet platforms have not demonstrated the robustness necessary for large scale screening efforts by CE and MCE. Further research should emphasize development of robust, integrated, multiplexed droplet and MCE system capable of screening > 10⁴ compounds routinely. Ultimately, such a system, based on MCE and droplet microfluidic technology, would open the door to next generation screening that is not limited by the well plate volume and need for expensive robotics. It would also fully unlock the power of CE screening.

Table 2.

Representative CE affinity interaction screens

Target	Assay type	Compounds Assayed	Z-score	Reference
Tat-TAR	Commercial microchip platform	1	N.D.	[9]
Hsp70-Bag3	APCE	3,443	0.78	[3]
Bcl-XL-Bid	CE-FA	105*	0.86	[68]
Hsp90α	CEfrag	609	0.6	[117]

Not determined (N.D.); affinity probe capillary electrophoresis (APCE); capillary electrophoresis-frontal analysis (CE-FA); fragment screening using capillary electrophoresis (CEfrag)

^*includes crude product, natural extract, or chinese traditional herb

Article Highlights.

• Capillary electrophoresis (CE) and microchip electrophoresis (MCE) offer rapid, low volume separations, well suited for drug discovery and screening.

• Few large-scale screens have been reported using CE; many of these have been performed on commercial microchip platform.

• Sub-second CE separations are possible in some cases and the throughput of CE can also be improved by parallelization and multiplexing.

• A variety of assays are accessible by CE including enzyme assays and biomolecular interactions.

• Novel strategies for reducing sample consumption have been reported including droplet microfluidics and on-line reactions demonstrating a future potential of low sample consumption by CE.