A cell-based multiplex immunoassay platform using fluorescent protein-barcoded reporter cell lines

Multiplex immunoassays with acellular antigens are well-established based on solid-phase platforms such as the Luminex® technology. Cell barcoding by amine-reactive fluorescent dyes enables analogous cell-based multiplex assays, but requires multiple labeling reactions and quality checks prior to every assay. Here we describe generation of stable, fluorescent protein-barcoded reporter cell lines suitable for multiplex screening of antibody to membrane proteins. The utility of this cell-based system, with the potential of a 256-plex cell panel, is demonstrated by flow cytometry deconvolution of barcoded cell panels expressing influenza A hemagglutinin trimers, or native human CCR2 or CCR5 multi-span proteins and their epitope-defining mutants. This platform will prove useful for characterizing immunity and discovering antibodies to membrane-associated proteins.

M embrane proteins are often critical antigens in humoral responses to infection and physiologically significant in health and disease. They are, therefore, valuable targets in vaccine design and therapeutic antibody development 1 . For many multi-subunit or multi-pass transmembrane proteins, the native conformation is preserved only when the protein is properly expressed on the cell membrane 2 . For this reason, cellbased binding assays using protein antigens expressed on the cell surface are widely used in antibody screening and characterization 3 . A major limitation in these assays is that they are single analyte, or singleplex methods, and consequently timeconsuming, sample-depleting, and costly when used to screen samples with high numbers (>10) of antigens. Although multiplex immunoassays based on the Luminex ® technology are wellestablished, only acellular antigens can be immobilized on detection beads with this platform.
Fluorescent cell barcoding (FCB) was developed as a cell-based multiplexing technique [4][5][6][7] . Distinct cell populations are labeled with unique signatures of amine-reactive fluorescent dyes of different emission wavelengths and intensities before pooling for exposure to potentially reactive antibody followed by analysis by flow cytometry. The FCB technique improves the throughput of flow cytometry experiments and minimizes staining variability and antibody consumption. Although this FCB approach offers flexibility in barcoding cell samples from different sources, especially primary cells, barcoding, labeling, and quality assurance procedures are required prior to every experiment. In contrast, a panel of reporter cell lines stably barcoded by endogenous expression of fluorescent proteins (FPs) yet capable of expressing select membrane-associated antigens by transduction, would establish a stable and highly reproducible immunoassay platform. Once established, endogenously barcoded reporter cell lines would greatly simplify the FCB approach.
Here we describe a cell-based, multiplex immunoassay platform using reporter cell lines with stable barcoding by FPs. Briefly, a basal cell line was engineered to express different combinations of FPs that can be reliably detected and distinguished in different channels on flow cytometers. This resulted in a panel of FP-barcoded reporter cell lines analogous to barcoded Luminex ® beads. Individual reporter cell sub-lines, each bearing a unique combination of FPs, were subsequently engineered to express different cell surface proteins to permit their pooling in a cell-based multiplex immunoassay. Flow cytometry was then used to de-multiplex signals from pooled reporter cell lines by serial gating to identify FP combinations. In this proofof-concept study, we established a 16-plex basic panel and tested the feasibility of expanding to a 256-plex panel of reporter cell lines. We successfully applied this platform to multiplex detection of antibody binding to cell panels expressing various influenza A hemagglutinin (HA) trimers or human CCR2b and CCR5, and several domain-swap or point mutants that define specific protein domains/epitopes.

Results
Generation of a basal reporter cell line suitable for immunoassays. The human chronic myelogenous leukemia cell line K562 was selected as a parental cell line; as a suspension cell line, detachment or digestion treatment is unnecessary, removing the possibility of denaturation of expressed surface antigens. K562 cells also exhibit a high efficiency of transfection and transduction, and are resilient to apoptosis 8 , simplifying the processes of genetic modification and assay processing. A drawback of K562 as a reporter cell line is its expression of the Fcγ receptor, CD32A, which binds most human IgG subclasses with sufficient affinity to raise background thresholds ( Supplementary Fig. 1). We therefore knocked out CD32A by CRISPR-Cas9-mediated gene targeting ( Supplementary Fig. 1). Using genomic DNA sequencing to ensure disabling all CD32A alleles, we generated a cloned, basal cell line, K530 . Compared with the parental K562 cells,  K530 cells are negative for all Fcγ receptors CD16, CD32, and  CD64, and show minimal nonspecific binding by human IgG1 (Supplementary Fig. 1).
Stable barcoding of reporter cell lines with unique combinations of FPs. To introduce stable barcodes into the reporter cells for multiplex detection, we conceived a strategy to use combinations of FPs expressed in the cytosol of reporter cells. The growing toolbox of FPs 9 coupled with widely used multicolor flow cytometry allows for marking and detecting multiple FPs simultaneously. The number of unique FP combinations is 2 n where n is the number of different FPs. This exponential result provides the "power" for multiplexity. After careful selection and experimental tests, we selected eight FPs, including EBFP2 10 , mTurquoise2 11 , mNeonGreen 12 , mCardinal 13 , mKate2 14 , miRFP703 15 , LSSmOrange 16 , and hmKeima8.5 17 (Supplementary  Table 1). These FPs are characterized by the following: (1) bright fluorescence for good separation and limited spill-over into other fluorescence channels; (2) good photostability and low cytotoxicity; and (3)  We validated the four-color basic panel in supporting 16-plex detection of cell surface molecules. K530 cells were engineered to express all 16 combinations of four FPs from the basic panel, resulting in 16 uniquely FP-barcoded reporter cell lines ( Fig. 1a and Supplementary Fig. 8). Pooled cell lines can be demultiplexed by flow cytometry based on patterns of FP expression ( Fig. 1b and Supplementary Fig. 8). The growth rates of these 16 FP-barcoded reporter cell lines were determined individually and after pooling; similar growth rates were observed in both conditions (Supplementary Fig. 9). Although proliferation rates of all barcoded cells are similar, variation in growth rates is sufficient such that expansion of pooled cell lines should be limited to three to four doublings or about 3 days of culture to preserve comparable and adequate numbers for the reliable detection of each subpopulation. To validate the deconvolution of multiplexed cell populations, we generated 16 cell lines expressing human CD4, CD8a, CD86, or CD154 so that each protein was associated with four unique FP patterns. The pooled cells were stained in single tubes with monoclonal antibody for one of the four human antigens, followed by a common phycoerythrin (PE)-conjugated secondary antibody. Deconvolution by flow cytometry showed high resolution of bound and unbound cells and patterns of binding consistent with antigen expression by barcoded cells before multiplexing ( Fig. 1c and Supplementary Figs. 8 and 10).
Cell surface expression of influenza HA antigens and multiplex detection of binding by reference antibodies. Influenza HAs are diverse trimeric membrane proteins. Recombinant HA protein preparation can be challenging as some HA ectodomains are unstable and cell surface expression can be a surer option for obtaining native structure when working with unknown or novel influenza strains. Multiplex immunoassays with collections of cell surface-expressed HA antigens may be useful in influenza vaccine design and immunological monitoring. To demonstrate, we prepared a panel of 12 reporter cell lines expressing the HA trimers of diverse influenza A subtypes and successfully validated their structural integrity with several reference antibodies ( Fig. 2 and Supplementary Table 2). The pan-influenza A stem-binding antibody FI6 18 bound to all the HA antigens tested. The S5V2-29 antibody specific for a conserved interface epitope 19 bound to a majority of group 1 and group 2 HAs tested. CH67, a neutralizing antibody for H1N1 influenza 20 21 , bound to that HA but not to others. This assay may be an alternative to Luminex bead assays in highthroughput screening of antigen-specific and cross-reactive antibodies secreted in clonal cultures of human B cells from influenza-vaccinated subjects 22 .
Application of the multiplex immunoassay in domain/epitope mapping of antibodies specific for human CCR2 and CCR5. G protein-coupled receptors (GPCRs) are key to many biological functions and are an important class of drug targets, constituting some 34% of US Food and Drug Administration approvals 23 . For these seven-pass transmembrane proteins, mammalian cell surface expression is the most reliable way to preserve native conformation. Domain/epitope mapping of GPCR-specific antibodies can be helpful for structure-function studies of the receptors and for the screening antibodies for therapeutic, diagnostic, and research applications 24 . A cell-based multiplex immunoassay that could accommodate the large number of protein variants needed for domain/epitope mapping in a single-tube could be useful.
As an example, a panel of reporter cell lines was prepared, expressing human CCR2b or CCR5 and 12 domain-swapped chimeric mutants (Fig. 3a). Two CCR2-specific and five CCR5specific antibodies were tested for their binding patterns to these domain-swapped mutants in a multiplex immunoassay (Fig. 3b,

FP0010-hCD86
Anti-hCD154 Anti-hCD86 Anti-hCD8a Anti-hCD4 Isotype Ctrl c The 16 barcoded reporter cell lines were transduced to express human CD4, CD8a, CD86, and CD154 molecules in a shifted pattern relative to FP expression. These cells were pooled and stained with corresponding mouse monoclonal antibodies (as indicated on the top of each histogram) followed by a PE-conjugated anti-mouse IgG antibody. Signals from individual reporter cell lines were demultiplexed as shown in b and the binding by corresponding antibodies were plotted as half-offset histograms. In all cases, the detected expression patterns were consistent with antigen expression by barcoded cells before multiplexing as shown on the right of histograms for each individual cell line. Isotype Ctrl, mouse IgG1, κ-isotype control antibody. Data from one experiment are shown. Data from another independent repeat experiment are shown in Supplementary Fig. 8. Supplementary Fig. 11, and Supplementary Table 3). We found the two uncharacterized CCR2 antibodies to be specific for the CCR2 N1 domain (Fig. 3b, c). Of the previously characterized CCR5 antibodies, CTC8, 2D7, and 45529 bound to the N1, ECL2A, and ECL2B domains, respectively, whereas binding of 45523 and 45549 involved multiple domains (Fig. 3b, c). These observations matched prior reports 25,26 . The combination of gain-and loss-of-binding readouts across all the ectodomains allowed a semi-quantitative determination of the contribution of each individual ectodomain (Fig. 3c) and revealed more minorcontributing domains unappreciated previously 25,26 . Epitope mapping with another panel of reporter cell lines expressing CCR5 mutants with alanine scanning (or glycine-toserine mutation) in ECL2A domain (Fig. 3d, 26 , K171 and E172 were critically involved in the binding by 2D7 and 45523. With this scanning panel, another five residues were also determined to be engaged in antibody binding, which were not covered or appreciated in previous studies. The only discrepancy with previous studies 25 is the role of K171 and E172 in the binding by 45549 and awaits further confirmation. In practice, a full scanning panel can be established to cover every residue in all ectodomains. This multiplex immunoassay can enable high-throughput screening of antibodies with desired specificity and potential function at singleresidue resolution in a single-tube manner to expedite therapeutic or diagnostic antibody discovery. Potential for expanding the 16-plex reporter cell line panel. To expand the established 16-plex reporter cell line panel, additional transductions for more FPs are necessary. Potentially, the accumulations of genetic insertions in the reporter cell genome could impact the expression and processing of preexisting FPs, other transduced proteins (antigens), and impair cell metabolism or proliferation. To address this concern, we generated a reporter expressing a fifth FP, LSSmOrange. Reporter cell lines FP0000 and FP1111, with none or four preexisting FPs, were pooled and transduced to express LSSmOrange. Similar transduction efficiencies and LSSmOrange intensities were achieved in both parental lines ( Supplementary Fig. 12). Monoclonal reporter cell lines were generated from these bulk transductants, with FP10000 expressing only LSSmOrange and FP11111 expressing all five FPs (Fig. 4a). The fluorescent intensities of LSSmOrange in FP10000 and FP11111 were similar; importantly, the fluorescent intensities of the preexisting four FPs remained the same in cell line FP11111 as compared with the parental line FP01111 (Fig. 4a). Similar growth rates were also detected for these four cell lines ( Supplementary Fig. 12), within the same range for the 16-plex reporter cell lines (Supplementary Fig. 9).
To confirm that the cell line expressing five FPs (FP11111) remained capable of the expression, folding, and processing of analyte proteins, a complex membrane protein, human CCR5, was tested as an example analyte. Similar transduction efficiencies and surface CCR5 expression levels were detected based on the four cell lines established above ( Supplementary Fig. 12). After antibiotic selection, the vast majority (>97%) of successful transductants expressed surface CCR5 at similar levels across all the four cell lines tested (Fig. 4b). These results showed that simultaneous expression of five FPs in the reporter cells did not detectably impair surface expression of a complex membrane

Discussion
We describe and present here a cell-based multiplex immunoassay platform. A basal cell line was stably barcoded with unique combinations of four endogenous FPs to generate a stable and adaptable panel of 16 distinct reporter cell lines. Analyte protein antigens were then expressed on these reporter cell lines and detected in a multiplex binding assay by flow cytometry. We validated its feasibility in multiplexing and reliability in demultiplexing by flow cytometry. We demonstrated its broad application in multiplex detection of antibody binding to influenza HA and domain/epitope mapping against human CCR2b/CCR5 mutant antigens. The four reporter cell lines generated above were transduced to express human CCR5. After puromycin selection, the bulk transductants were detected for cell surface express of CCR5. CCR5+ populations were gated and the expression levels were compared in overlaid histogram plot on the right.
A transient, cell-based multiplex fluorescent barcoding technique, FCB, based on the use of amine-reactive fluorescent dyes was reported previously [4][5][6][7] . Although this method enables barcoding of primary cell samples for multiplex detection in singletube assays, the labeling procedure of each cell reporter must be carried out shortly before every experiment. Obviously, cells labeled in this way cannot be allowed to proliferate as doing so dilutes the label. In contrast, the multiplex platform we have developed uses cell lines with stable FP barcodes to establish membrane antigen-expressing reporter cell line panels for multiplex immunoassays. This platform is especially suitable for the generation of biological or experimental variant libraries based on a same parental cell line background. The downside of the latter platform is that it's applicable only to cell lines and the initial preparation of reporter cell line panels is a time-consuming process. Once the reporter cell line panel is made, it becomes an unlimited ready-to-use reagent.
In practice, there are several considerations which affect the potential utility of our approach. First, the selection of a proper founder cell line is the principal determinant for generating reporter cell panels that provide both ease of use and robust binding data. Although in theory, any cell line can be used, a suspension cell line with high transfection/transduction efficiency is optimal. In the event that the founder cell line expresses analyte antigen endogenously, an isogenic antigen-negative control cell line must be generated. Second, the stability of introduced transgenes is crucial for the maintenance of reporter cell lines and their reproducibility in immunoassays. In our experience, lentiviral vectors with genetic elements known to prevent epigenetic silencing [27][28][29] result in stable and uniform expression of transgenes in the absence of selection pressure. Third, monoclonal reporter cell lines with similar growth rates should be selected to ease challenges in the maintenance of panels with large numbers of reporter cell lines. Reporter cell lines with similar growth rates can be expanded individually and pooled as a single-tube panel and frozen down as single-use stocks. A short-term expansion of the pool after thawing preserves sufficient portions and allows reliable detection for each sub-population ( Supplementary Fig. 9). In the maintenance of panels of larger number of reporter cell lines, the portion of each individual cell line for pooling can be adjusted prospectively so that to uniformize the differences after expansion.
In summary, we have developed a cell-based multiplex immunoassay platform that is analogous to the well-established Luminex ® platform but for the expression of membrane-associated, not acellular, antigens. The cell-based property of this platform offers flexibility in multiplexing, superiority in supporting surface display of complex membrane proteins in their native conformation, and economy in production once the requisite cell lines are established. These unique features make this platform a valuable tool in humoral response monitoring and antibody characterization. Pooled reporter cell line panels are amenable to high-throughput screening with flow cytometry and in concert can facilitate existing antibody isolation techniques such as hybridoma, single-cell PCR, or single B-cell cultures to aid in antibody discovery and engineering.

Methods
Mice and human subjects. Female C57BL/6 mice were obtained from the Jackson Laboratory and maintained under specific pathogen-free conditions at the Duke University Animal Care Facility. Splenocytes were isolated from one 12-week-old mice for RNA extraction. All experiments involving animals were approved by the Duke University Institutional Animal Care and Use Committee (IACUC A128-20-06).
Peripheral blood mononuclear cells (PBMCs) were obtained from one human (Homo sapiens sapiens) subject under Duke Institutional Review Board Committee guidelines (IRB Pro00062495). Written informed consent was obtained.
Culture, transfection, and transduction of mammalian cell lines. HEK 293T and K562 cell lines were purchased from ATCC. HEK 293T cells were cultured in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% heatinactivated HyClone fetal bovine serum (FBS) (Cytiva), 10 mM HEPES buffer, and 55 µM 2-Mercaptoethanol (all Gibco). K562 and derivative cell lines were maintained in RPMI-1640 medium (Gibco) supplemented with 10% heat-inactivated HyClone FBS, 10 mM HEPES buffer, 1 mM sodium pyruvate, 1× MEM NEAA, 55 µM 2-Mercaptoethanol, 100 units/ml penicillin, and 100 µg/ml streptomycin (all Cell surface staining, flow cytometry analysis, and single-cell sorting and cloning. Cultures of K562 and derivative cells were collected, centrifuged at 300 × g for 2 min at 4°C, and resuspended in staining buffer (phosphate-buffered saline supplemented with 2% heat-inactivated FBS). After incubation with antibodies at 4°C in the dark for 30 min, cells were washed with staining buffer and resuspended in staining buffer for either secondary staining following the same procedure above or stored on ice for flow cytometry analysis or single-cell sorting. For cell counting, CountBright™ Absolute Counting Beads (Invitrogen) were added into cell suspensions prior to flow cytometry analysis.
Flow cytometry analysis was carried out using either BD FACSCanto II cytometer (Duke Cancer Institute Flow Cytometry Shared Resource) or BD LSR II cytometer (The Duke Human Vaccine Institute (DHVI) Research Flow Cytometry Facility). Single-cell sorting was performed with BD Aria II (The DHVI Research Flow Cytometry Facility). The bulk cell line after transfection or transduction were stained with corresponding antibodies and single cells expressing FP or antigen of interest were sorted into 96-well flat-bottom plates containing 100 µl/well of the complete RPMI medium above supplemented with 20% heat-inactivated FBS. Nine days after sorting, healthily proliferating cell clones were transferred into 24-well plates for further expansion. Three days later, individual monoclonal cell lines were validated for the expression of FP or antigen of interest and a single clone with a uniform expression level was selected for further engineering or used as a reporter cell line in immunoassays. Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All raw data generated during this project are available from the authors upon request. Licensing from Allele Biotechnology and Pharmaceuticals, Inc., is needed prior to material transfer of plasmids containing mNeonGreen sequences. All other plasmids are available from the authors upon request.