Rapid isolation of antigen-specific B-cells using droplet microfluidics

Monoclonal antibodies are powerful tools for scientific research and are the basis of numerous therapeutics. However, traditional approaches to generate monoclonal antibodies against a desired target, such as hybridoma-based techniques and display library methods, are laborious and suffer from fusion inefficiency and display bias, respectively. Here we present a platform, featuring droplet microfluidics and a bead-based binding assay, to rapidly identify and verify antigen-binding antibody sequences from primary cells. We used a defined mixture of hybridoma cells to characterize the system, sorting droplets at up to 100 Hz and isolating desired hybridoma cells, comprising 0.1% of the input, with a false positive rate of less than 1%. We then applied the system to once-frozen primary B-cells to isolate rare cells secreting target-binding antibody. We performed RT-PCR on individual sorted cells to recover the correctly paired heavy- and light-chain antibody sequences, and we used rapid cell-free protein synthesis to generate single-chain variable fragment-format (scFv) antibodies from fourteen of the sorted cells. Twelve of these showed antigen-specific binding by ELISA. Our platform facilitates screening animal B-cell repertoires within days at low cost, increasing both rate and range of discovering antigen-specific antibodies from living organisms. Further, these techniques can be adapted to isolate cells based on virtually any secreted product.

population. Following a week in culture the cells were harvested via centrifugation at 300xG, resuspended in 10% DMSO in fetal calf serum (FCS), aliquotted into cryovials and frozen in a controlled freezing container placed in a -80c freezer. These cryovials were then transferred to liquid nitrogen tank for storage. To determine the frequency of cells that secrete IgG-and antigen-specific IgG, we thaw an aliquot, remove dead cells using either Miltenyi dead cell removal kit (Miltenyi Biotec 130-090-101) or Lympholyte®-Rat density separation medium (Cedarlane CL5040) according to the manufacturer's instructions, and use the fluorescent foci method (Tickle et al., 2015) to characterize the cells. Prior to use in microfluidic experiments, frozen primary cell stocks are thawed at 37°C and added to 10ml pre-warmed complete DMEM. The cells are pelleted for 5min at 400g and resuspended in media. Resuspended primary cells and cultured hybridoma cells are passed through a 40μm nylon strainer (VWR 10199-654). We add 0.25 µl DNAseI (New England Biolabs N0303) and 1.2 µl 1M MgCl2 per 0.5ml cell solution and incubate at 37°C for 15min. Cells are then pelleted, resuspended in complete DMEM and purified. Then the cells are chilled on ice for >10min to prevent antibody secretion prior to encapsulation into droplets.

Droplet making
Cells and beads are resuspended in separate media solutions containing 40% (v/v) Percoll (Sigma P4937), 30ng/µl labeled antigen (~1.7 µM) and 2% (v/v) RNAse inhibitor (RNAsin® Plus Promega N2611). We use HFE7500 (3M, USA) containing 2% (w/v) surfactant (008-FluoroSurfactant, Ran Biotechnologies, USA) as the continuous phase. Harvard PHD2000/2200 pumps (Harvard Apparatus, Holliston, MA) are used for liquid injection. Typical flow speed is 150µl/hr for each of the aqueous solutions and 700µl/hr for the oil phase. Typically, we have an average of 3-5 capture beads per droplet and one cell per 3-5 droplets, so that most droplets have at least one capture bead and few droplets have more than one cell. We track droplet production by monitoring the fluorescence downstream of the droplet-forming nozzle via PMT. Droplets are collected into 1mL syringes (BD Luer-Lok™ 309628) and moved to 37°C cell culture incubator for desired period, typically 2 hours. We assume that cells will secrete at different rates, ranging from ~4 to ~10,000 molecules per second (3). In a droplet containing a cell that secretes at this maximum rate, secreted antibody will reach 1.7 µM after 1.4 hr incubation at 37 °C. Because cells do not secrete at their maximum rate immediately upon shift from 4 °C to 37 °C, and because most cells will have maximum secretion rate <10,000 antibody molecules per second, we believe that in almost all droplets the labeled antigen (alexa-488 labeled TNF-α protein, 1.7 µM ) will be in excess of cell-secreted antibody. However, this high concentration of fluorescent antigen also reduces detection sensitivity, which is related to the ratio of bead-bound to unbound fluorescent antigen.

Droplet sorting
After droplet incubation, we use PEEK tubing (Western Analytical Products, Tub PEEK Nat Tubing Sleeve PEEK Natural 1/32" OD x .008" ID) to attach the droplet-collection syringe to a sorting device. HFE7500 with 0.2% surfactant is used as the oil in the filter and downstream spacing. The sorting device (Fig. S2) can be operated at speeds over 100 Hz; however, sorting at the maximum rates may result in increased collection of undesired cells. Thus, when performing longer sorting experiments to isolate rare events, we use a flow rate of 15µl/hr for droplets, 200µl/hr for filter oil and 250µl/hr for spacing oil, and droplets pass through the sorting junction at ~70Hz. We use the "observation trap" on the microfluidic device (Fig.  S2, element j) to tune and verify sorting thresholds. We use our sorting instrument to analyze a fraction of input droplets and generate a scatterplot of droplet fluorescence intensity. All droplets contain the same concentration of antigen. For example, in Figure 2E, the vast majority of droplets are in the yellow region, having a droplet peak intensity of ~0.5 and droplet average intensity of ~0.45. Anything with droplet average intensity significantly higher than ~0.5 is assumed to be an abnormal droplet, e.g., shrunken during incubation or containing very high fluorescence due to particulate fluorescent antigen. Droplets lacking a fluorescent bead will have pedestal-shaped peaks where the droplet peak intensity is essentially the same as the droplet average intensity. Only the desired droplets have a droplet peak intensity significantly higher than the droplet average intensity. Thus, we sort based on droplet peak intensity relative to droplet average intensity. The droplet peak intensity may be from a single bright bead, or the cumulative signal of several beads detected simultaneously. In practice, we select a sorting threshold in two steps. We monitor droplets, usually >20,000, and select a preliminary sorting threshold based on the signal distribution. We then perform a limited sort and collect a small number of droplets into the observation trap. We then disconnect the device and image the trap at a confocal microscope to confirm sorting. We then re-connect the device and plug the trap outlet (Fig.S2, element k) to direct sorted droplets through the collection outlet (Fig.S2, element i). After the flows equilibrate, we sort into a collection tube. Use of the observation trap greatly improves sorting accuracy, though overall sorting rate is reduced. In a typical sort, which included several breaks to fine-tune sorting thresholds, we screened ~600,000 droplets in ~5hr, giving a rough average speed of 30Hz. ~33% of the droplets contained a cell and at least one bead, giving an overall rate of ~10 cells per second. To enable use of a pipetman to dispense sorted droplets, we collect sorted droplets into a ~20µL emulsion of 50 µm diameter "dilution" droplets generated by droplet microfluidics and consisting of 75.5% water, 20% Percoll, 2.5% PBS and 2% (v/v) RNAse inhibitor (Promega N2611). Droplets are then gently mixed and dispensed by pipetting into PCR tubes.

Identification of sorted hybridoma cells
Analysis of sorted droplets: We perform reverse transcription and nested PCR (RT-nPCR) on sorted droplets to identify the encapsulated cells as either anti-cMyc-or anti-TNF-alpha-hybridoma cells. Primer sequences are given in (SI Appendix Table S2). Prior to performing RT-nPCR on sorted droplets, we demonstrate the sensitivity and specificity of the assay through analysis of samples containing anti-cMycand anti-TNF-α-hybridoma cell mRNA mixed at equal ratios (mRNA concentration is expressed in "cell equivalents"). We assemble on ice reverse transcription reactions containing 0.5µM "mouse IgG Heavy chain RT" reverse primer, 0.25µM "TSO-mouse" template switching oligo, 0.5ul Superscript TM IV reverse transcriptase, 4µl 5x Superscript TM IV reverse transcription buffer, 1µl RNAsin® Plus, 5 mM DTT, the indicated number of cell equivalents as template, and dh20 to final volume of 20µl. To ensure efficient template-switching, we perform the RT in two steps. Initially, we incubate for ten minutes at 55°C, then lower the temperature to 42°C, add 0.5µl MuLV enzyme (NEB M0253S) and incubate 40 min at 42°C, then ramp to 4°C. Samples are frozen at -20°C or used immediately for PCR. One microliter of each RT mixture is used for 20µl initial PCRs (iPCR) containing: 0.25µM forward primer "O-LA1" (anneals to the sequence added during template switching), 0.25µM reverse primer "Mouse IgG 1st PCR" (anneals perfectly to both the anti-cMyc and anti-TNF-α heavy chain-encoding cDNA), 0.4µl Taq polymerase (ThermoFisher EP0401), 2µl 10x (NH 4 ) 2 SO 4 PCR buffer, 2 mM MgCl 2 , 0.2 mM dNTPs (each), and dh20 to 20µl total volume. The samples are cycled using a touchdown protocol: Initial denaturation at 94°C for 3 min; then 9 cycles of: 94°for 30 sec, anneal for 30 sec using temperature decreasing by 1°C per cycle (62°C to 54°C), extend 1 min 72°C; then 21 cycles of 94°C for 30 sec, anneal at 54°C for 30 sec, and extend at 72°C for 1 min; then a final extension at 72°C for 5 min, followed by a hold at 4°C. A fraction, 2.5%, of each iPCR is then used as template in 10ul nested PCRs (nPCR) containing a primer pair specific for anti-cMyc heavy chain cDNA (antiMyc Forward, antiMyc Reverse, 0.25µM each) and a pair specific for anti-TNF-α heavy chain cDNA (antiTNFforward, antiTNFreverse, 0.25µM each), 0.2µl Taq polymerase (ThermoFisher EP0401), 1µl 10x (NH 4 ) 2 SO 4 PCR buffer, 2 mM MgCl 2 , 0.2 mM dNTPs (each), and dh20 to 10µl total volume. The cycling protocol uses an initial denaturation at 94°C for 3 min; then 30 cycles of 94°C for 30 sec, anneal at 54°C for 30 sec, and extend at 72°C for 1 min; then a final extension at 72°C for 5 min, then samples are held at 4°C. The nPCR samples were separated on 3% agarose gel, then visualized by ethidium bromide staining and UV illumination.
To analyze sorting, reverse transcription is performed essentially as described above, but in small volumes to allow use of the entire RT reaction product as template for iPCR. We perform reverse transcription in sets of 8-well PCR strips. We prepare 100µl of reverse transcription cocktail comprising 0.5µM mouse IgG Heavy chain RT reverse transcription primer, 0.25µM TSO-mouse template switching oligo, 0.5 mM each dNTP, 5µl RNAsin® Plus RNAase inhibitor, 20µl 5x Superscript TM IV reverse transcription buffer, 5 mM DTT, 2.5µl Superscript TM IV reverse transcriptase, 2.5µl MuLV reverse transcriptase, 1 mM MnCl 2 , and dh20 to final volume of 100µl. We add 1µl cocktail to each well of dispensed droplets (see "Droplet sorting" section above for droplet dispensing details) and place the strips at -80C for >80 min to freeze-break the droplets. We transfer the strips to a thermocycler, incubate ten minutes at 55°C, then lower the temperature to 42°C and incubate for 40 minutes, and either freeze the samples at -20C or place on ice for immediate use in iPCR. The iPCR master mix contains 0.25µM O-LA1 forward primer, 0.25µM Mouse IgG 1st PCR reverse primer, 9µl Taq polymerase, 45µl 10x (NH 4 ) 2 SO 4 PCR buffer, 2 mM MgCl 2 , 0.2 mM dNTPs (each), and dh20 to 360µl total volume. We add 4ul of this mix directly to the reverse transcription reaction wells and cycle using the touchdown protocol detailed above. The nPCR master mix contains primers specific for the anti-cMyc heavy chain chain (antiMyc Forward, antiMyc Reverse, 0.25µM each) and primers specific for the anti-TNF-α heavy chain (antiTNFforward, antiTNFreverse, 0.25µM each), 6.5µl Taq polymerase, 45µl 10x (NH 4 ) 2 SO 4 PCR buffer, 2.5 mM MgCl 2 , 0.16 mM dNTPs (each), and dh20 to 360µl total volume. We aliquot 4µl nPCR master mix into PCR tubes, then use a multi-channel pipette to transfer 1µl of each iPCR to wells containing the nPCR mix. We cycle using the above nPCR conditions. The PCR products are electrophoresed on 3% agarose gel and bands are visualized by ethidium bromide staining and UV illumination. For some PCR analyses, we multiplexed the nested PCR by combining the anti-cMyc and anti-TNF-α primer sets in each reaction. Because one of the anti-cMyc primers can bind with mismatches to DNA encoding the anti-TNF-α heavy-chain, multiplex reactions performed on samples containing anti-TNF-α cells often generated a 300bp product in addition to the authentic 166 bp product. When this occurred, we re-tested the samples by performing the nested PCR in a non-multiplex fashion.
Gene recovery from sorted Rat splenocytes Reverse transcription was performed in a modified Superscript TM IV reaction (ThermoFisher 18090050). Briefly, ~0.5µl sort-emulsion droplets are added to ~3.8µl reaction mixture consisting of 6.64 units RNAsin, 10nmol dNTP, 50nmol dCTP, 0.4pmol Rat IgG Heavy RT primer, 0.1pmol Rat IgG Light RT primer, and 0.1pmol Rat B2M primer (SI Appendix Table S2). The mixture is frozen at -80°C for > 40min, then heated to 72°C for 3min and immediately incubated on ice for more than 1 min. These steps break the droplets and allow primer annealing. We next add 5.7µl of solution containing 10µmol Betaine, 2µl Superscript TM IV buffer, 0.25µl Superscript TM IV enzyme, 50nmol DTT, 60µmol MgCl 2 and 20pmol TSO DNA is added to each tube, mixed, and spun down. Reverse transcription is carried out at 37°C for 30min, 4°C for 45min then 37°C for another 30min, finally the enzyme is deactivated at 80°C for 10min. We PCR-amplify VH and VL DNA using Platinum TM SuperFi TM (Thermo Fisher 12358010) according to manufacturer's specifications. When amplifying cDNA from template-switching RT reactions, we use O-LA1 and I-LA1 as the forward primers for the initial and nested PCR, respectively. We use RatHeavyFirst and RatHeavyNest primers and RatLightFirst and RatLightNest primers (SI Appendix Table S2) for first round and nested PCR of V heavy and V light genes, respectively. Alternatively, we perform PCR using primer mixtures provided by UCB Pharma for initial and nested PCR(4). DNA is gel purified (Sigma NA1111-1KT) followed by Zero-blunt TOPO™ cloning (ThermoFisher K283520). Transformed colonies are incubated overnight at 37°C on agar plates with appropriate antibiotic (Ampicillin or Kanamycin). At least 4 colonies from each transformation are picked and plated onto stock plates. We identify clones that contain plasmids with VH-and VLencoding inserts by performing colony PCR using M13 forward and reverse sequence primers, and then we purify (QIAquick PCR cleanup kit, part 28106) and sequence the VH-and VL-products using Sanger sequencing with primers annealing to the T7 and T7 terminator sequences. We compare the sequences with the IMGT database to confirm inserts encode VH-and VL-domains. Although we occasionally isolated more than one distinct putative heavy-or light-chain sequence from a single well, we used only one heavy-chain and one light-chain to create that well's scFv expression construct.

Cell-free synthesis of single-chain antibody
We PCR-amplified the IMGT-verified VH-and VL-sequences using forward primers that add pCOAT-DB (5) sequence to the 5' end of the VH DNA (pc-VH Forward primer) and linker-encoding sequence to the 5' end of the VL DNA (pc-VL Forward primer); the reverse primers add linker-encoding sequence to the 3' end of the VH DNA (pc-VH reverse primer) and pCOAT-DB sequence to the 3' end of the VL DNA (pc-VL reverse primer) (SI Appendix Table S2). We gel-purify these PCR products and combine paired VH-and VLamplicons with a double-stranded linker DNA and linearized pCOAT-DB vector in NEBuilder HiFi Assembly reaction mixture (E2621) to generate circular plasmids. The sequence for the positive control is derived from adalimumab (CAS registry number 331731-18-1, sequence available at https://www.drugbank.ca/drugs/DB00051). We transform these reactions into competent E. coli and identify insert-containing clones by colony-PCR using primers flanking the pCOAT-DB insertion site. Clones with full-length insert are cultured overnight in ~5ml media with antibiotics. DNA is isolated from these cultures by Miniprep (Qiagen 27106) and further verified by Sanger sequencing. To minimize the effect of PCR-induced mutagenesis, we tolerate one amino acid difference between the pCOAT-DB inserts and the original IMGT-verified VH and VL sequences. The double-stranded linker DNA encodes the sequence GGGGSGGGGSGGGGS to link heavy and light chain domains (6). We perform sequential PCR amplifications using "first mp primers" and then "second mp primers" to generate linear cell-free protein synthesis constructs from these sequence-verified plasmids. These reactions add a T7 promotor and DNA encoding a myc epitope tag and transcription terminator to the 5' and 3' ends of each amplicon, respectively. The presence of the myc epitope in synthesized scFv molecules enables use of HRP-conjugated anti-myc antibody as the detection reagent in ELISA and dot blot experiments. The product is verified using gelelectrophoresis and gel-purified (Qiagen 27106). The concentration is measured with spectrophotometer (NanoDrop Technologies, Inc. ND-1000). We produce scFv proteins by using 300ng each PCR product in 25µl cell-free protein synthesis reactions (PURExpress® protein synthesis kit, New England Biolabs E6800). The mixtures are incubated at 37°C for 4hr and then stored at 4˚C ice. Although the PURExpress® buffer includes the reducing reagent DTT, protein synthesis in these conditions generates target-binding scFvs, as shown by target-specific binding of the positive control, an scFv derived from adalimumab. Further, we made power law correlations using DTT half-life data (provided by Sigmahttps://www.sigmaaldrich.com/content/dam/sigmaaldrich/docs/Sigma/Product_Information_Sheet/d9163pis.pdf) to estimate a 30 minute half-life in our cell-free synthesis conditions. Thus, at the start of the third hour of incubation, >95% of the input DTT should be degraded and not provide reducing activity.

scFv ELISA
We use ELISA to test binding of cell-free synthesized scFv proteins to the proteins TNF-α (from UCB), mouse Interferon-γ (BioLegend 575306), recombinant human TNFRII (Peprotech 310-12) and recombinant Histone 2B (Active Motif 31492). All are dissolved at 8ng/µl in PBS (Corning 21-040-CV). We add 47µl antigen solution to bottom of each well (Universal Binding plate, Thermo 9502227) and incubate at room temperature for >3 hours. We then wash 3 times with wash buffer consisting of PBS with 0.1% BSA (Sigma A7906) and 0.05% Tween 20 (Sigma P9416) and add 400µl blocking buffer (wash buffer containing 5% (w/v) powdered non-fat milk) to each well and block overnight at 4˚C. We then wash wells 3 times with wash buffer and add 4µl cell-free protein synthesis mixture diluted in 50µl wash buffer to each well. We rock the plate at room temperature for >45min, then wash 3 times with 180µl wash buffer and then add 100µl HRP-conjugated rabbit anti-myc antibody (Genscript A00173) at 0.1ng/µl in wash buffer. We incubate at room temperature with rocking for >30min, then wash 5 times using wash buffer, and then add 100µl HRP substrate (Biolegend 421501) to each well. We allow the reaction to proceed for several minutes at room temperature and stop by adding 100µl stop solution (Biolegend 77316). We determine the colorimetric signal intensity with a Synergy H1m plate reader (BioTek).

Dot Blot
Cell-free synthesis of scFv proteins was confirmed by dot blot. We pipetted 1µl of each cell-free synthesis reaction onto nitrocellulose membrane and air-dried for >30min. "Int. Ctrl." = internal control, a sample known to be detected by anti-c-Myc dot blot. The membrane was incubated at 4˚C overnight and washed using wash buffer. The membrane was then covered in HRP-conjugated mouse anti-myc antibody (Genescript A00863 diluted to 0.1ng/µl in wash buffer), rocked at room temperature for >30min and washed wash buffer (three washes, each at least 15min). PierceTM ECL substrate (Thermo 32109) was used for chemiluminescence signal generation and membrane was imaged using an AlphaInnotech ChemiImager (ProteinSimple).   . Abnormally large droplets are excluded by droplet filter (c) and exit the device from merged droplet outlet (d). Excess oil between irregularly spaced droplets is removed through a filter upstream of device outlet (e). Oil is added from oil inlet (f) to uniformly space the droplets. This packing and uniform re-spacing of droplets improves sorting speed and accuracy. After spacing, droplets are exposed to slitlaser excitation and the resulting fluorescence is detected (g). To sort droplets, activated electrodes (h) generate a temporary electrical field and steer desired droplets to the upper channel. The droplets can then be collected for optical observation by plugging the collection outlet (i) and imaging the observation trap (j). Alternatively, the trap outlet (k) can be plugged and sorted droplets are collected from collection outlet (i). Discarded droplets exit the device through the waste channel (l).

Figure S3. Binding assay in bulk.
Anti-mouse-antibody-coated beads and fluorescently labeled TNF-α protein are incubated without (a) or with (b) anti-TNF-α-secreting hybridoma cells. Images are taken 6 hours after reagents are mixed. (c) Capture beads and fluorescent TNF-α incubated in the absence or presence of anti-TNF-α secreting hybridoma cells. Bead fluorescence intensity was measured at indicated time points over 30 minutes. Each data point is the average intensity of 4 beads measured using confocal microscopy. Scale bars: 75μm. Figure S4. Sensitivity and specificity of hybridoma RT-PCR. mRNA was purified from counted anti-cMyc and anti-TNF-α hybridoma cells. The mRNA was combined to form a sample containing one thousand cell equivalents of each cell type per microliter. Reverse transcription was performed in 20µl reactions using the indicated number of cell equivalents as template. One microliter of each RT mixture (5% of the total RT reaction) was used for initial PCR (iPCR) using the template-switching forward primer O-LA1 and the reverse primer Mouse IgG 1st PCR, which anneals perfectly to both the anti-cMyc and anti-TNF-α heavy chain cDNA. A fraction, 2.5%, of each iPCR was then used as template in 10ul nested PCRs (nPCR) containing a primer pair specific for anti-cMyc heavy chain cDNA and a pair specific for anti-TNF-α heavy chain cDNA. The marker lane, M, was loaded with 50bp ladder. The predicted band size for each nPCR primer pair is indicated at left. Marker band sizes are indicated in basepairs at right. Animal spleen cells are immortalized by fusion to myeloma cells. Clones are isolated by sequential rounds of expansion and selection.

Full-length IgG
Low Low Some loss of antibody repertoire during cell fusion. Requires repeated rounds of expansion and selection. Hundreds of clones are screened. Accelerated commercial services (10) require ~40 days from cell fusion step to completion of hybridoma subcloning. Display methods using randomly paired heavy and light chains (11)(12)(13)(14)(15) IgG heavy-and light-chain genes are isolated from cells from naïve or immunized animals and assembled into DNA constructs to display the single-chain format molecules (scFv), typically on phage or yeast cells.

Low
High Library construction and screening can be accomplished in 2-4 weeks (15). Mispairing of heavy-and lightchain genes can result in low-affinity antibodies that may require affinity maturation. (16) Display methods pairing heavy and light chains in droplets (17)(18)(19) Single-cell reverse transcription and linking PCR are performed in droplets to maintain correct pairing of IgG heavy-and light-chain genes. Display and panning are performed as in standard display methods.

High
Initial publications demonstrate isolation of high-affinity antibodies with no affinity maturation required. Droplet-making equipment is inexpensive and simple to use. Library construction requires molecular biology expertise. Peptide barcoding ("Nestlink") of potential binders (20) Peptide barcodes ("Flycodes") are genetically fused to the potential binders and next-gen sequencing assigns each flycode to a potential binder. Library is expressed as a pool of barcoded proteins and binders are isolated. Flycodes are then identified by liquid chromatography-tandem mass spectrometry (LC-MS/MS).  (29) Next-Gen sequencing is used to determine total immune repertoires and bioinformatics is applied to identify, e.g., clones that expand in response to exposure or immunization. Ig-seq is very powerful when combined with functional data: total repertoires can be bioinformatically compared with sequences identified through functional screening of repertoire subsets to identify potential binders with improved features.

Studydependent
Low to medium High Does not suffer from biases due to cell immortalization. Requires bioinformatics expertise. Single-cell platforms such as nanowells or droplet-based barcoding are required for true heavy-and light-chain pairing.
Nanometer-sized, magnetically-aligned capture-bead-indroplet assays described in Gerard et al (30) Fluorescence-concentration assays are performed in microfluidic droplets using ~1300 magnetic capture beads/droplet. Assay determines amount of antibody secreted and amount bound, enabling sorting based on affinity. After droplet sorting, cells are released from droplets and reencapsulated for single-cell barcoding sequencing to identify correctly paired heavy-and light-chain genes.

High
The use of nanometer-sized magnetically alignable capture beads improves signal-to-noise and reduces chances of beads clogging channels but adds some complexity to device fabrication. Barcode-based identification of V-region sequences requires technical and bioinformatics expertise, and recombinant constructs must be synthetically generated.
Micron-sized capture-bead-indroplet assays (this study) Fluorescence-concentration assays are performed in microfluidic droplets to identify cells that secrete target-binding antibody. This system, as currently described, does not allow for affinity measurements. Droplets in which cell-secreted antibody binds target protein are sorted and dispensed into individual wells. Single-well RT-PCR is used to isolate paired heavy-and light-chain V-genes from single droplet-sorted cells.

Secreted full-length IgG
Medium

Mediumhigh
Bead-based concentration assay is easily modified for different protein targets. Some droplets lack a detection bead, resulting in some inefficiency. Unlike Gerard et al.,   Table S3: Summary of separate low-and high-threshold sorting experiments. We generated droplets containing cultured rat splenocytes a and anti-TNF-α fluorescence detection reagents, incubated to allow antibody secretion, and performed sorting using low and high fluorescent signal thresholds. In the low threshold sort, ~0.04% of the observed droplets had an above-threshold fluorescent signal, resulting in electrode activation to sort 117/320,000 observed droplets. We recovered both the heavy-and the lightchain-encoding DNA from fourteen of these droplets and used cell-free synthesis to make scFv constructs from nine of the recovered heavy-and light-chain pairs. Six of these scFv proteins clearly bound to TNF-α b . In the high threshold sort, ~0.01% of the observed droplets had an above-threshold fluorescent signal, resulting in electrode activation to sort 38/380,000 observed droplets. We recovered both the heavy-and the light-chain-encoding DNA from four of these droplets and used cell-free synthesis to make the corresponding four scFv proteins. All four scFv proteins clearly bound TNF-α b . The low-and highthreshold sorts showed ~25-fold and ~50-fold enrichment, as judged by the % of sorted cells converted into ELISA positive scFv/the % positive cells in the input. Superscript meanings -a: Cells secreting anti-TNF-α antibody comprised ~0.2% of the input cell population, as determined by fluorescent foci analysis (Clargo, 2014, mAbs). b: Produced a signal in TNF-α ELISA that was >175% the signal generated by a negative control scFv. c: The fold enrichment by the overall process is the % sorted cells converted into ELISA-positive scFvs / % of input cells that secreted anti-TNF-α antibody.