Allo-Specific Humoral Responses: New Methods for Screening Donor-Specific Antibody and Characterization of HLA-Specific Memory B Cells

Antibody-mediated allograft rejection (AMR) causes more kidney transplant failure than any other single cause. AMR is mediated by antibodies recognizing antigens expressed by the graft, and antibodies generated against major histocompatibility complex (MHC) mismatches are especially problematic. Most research directed towards the management of clinical AMR has focused on identifying and characterizing circulating donor-specific HLA antibody (DSA) and optimizing therapies that reduce B-cell activation and/or block antibody secretion by inhibiting plasmacyte survival. Here we describe a novel set of reagents and techniques to allow more specific measurements of MHC sensitization across different animal transplant models. Additionally, we have used these approaches to isolate and clone individual HLA-specific B cells from patients sensitized by pregnancy or transplantation. We have identified and characterized the phenotypes of individual HLA-specific B cells, determined the V(D)J rearrangements of their paired H and L chains, and generated recombinant antibodies to determine affinity and specificity. Knowledge of the BCR genes of individual HLA-specific B cells will allow identification of clonally related B cells by high-throughput sequence analysis of peripheral blood mononuclear cells and permit us to re-construct the origins of HLA-specific B cells and follow their somatic evolution by mutation and selection.


INTRODUCTION
HLA sensitization remains a barrier to transplantation of all organ types (1,2). The etiology of alloantibodies stem from exposure to foreign HLA through pregnancy, transfusion, and previous transplantation and is driven by the high polymorphism among HLA genes. Advances in histocompatibility testing have improved the sensitivity and specificity of HLA antibody detection and has revealed broader HLA sensitization among transplant candidates. Transplantation in the presence of donor-specific HLA antibody (DSA) for mismatched HLA determinants in the donor's phenotype increases the risk of rejection and allograft loss, while locating an HLA compatible donor can increase waiting time and associated mortality (3).
Screening for HLA antibodies is routinely performed prior to transplantation to avoid DSA and the increased risk for AMR and graft failure. In this way, donor selection and transplant outcomes are optimized. HLA single antigen bead (SAB) Luminex ® assays allow sensitive screening for antibodies specific for hundreds of HLA alleles and are used widely in clinical practice. These assays, however, are prohibitively expensive for non-clinical studies and unavailable for animal species frequently used to study humoral responses to MHC antigens, notably macaques and swine. This lack of comparable diagnostic tools limits detailed study of the breadth and specificity of MHC-specific humoral responses in non-human transplant model systems. Crossmatch assays using donor lymphocytes have been used to detect the presence or absence of species-specific anti-MHC antibodies in experimental transplantation studies with rhesus macaques (4) and swine (5). In a series of recent studies (6,7), swine SLA class II single antigen expressing HEK293 cell lines were generated to monitor xenoreactive antibody responses with human serum samples, albeit, with singleplex capability and limited MHC allele coverage.
Sensitized patients are vulnerable to AMR when transplanted in the presence of circulating DSA, but some may also experience early AMR as a result of post-transplant activation of cryptic HLAspecific B memory. Unfortunately, current methods of assessing AMR risk at time of transplant relies solely on detection of circulating DSA and are unable to detect HLA-primed memory B (Bmem) cells that can rapidly proliferate and differentiate to antibody secreting cells on restimulation. In order to reveal this hidden humoral memory, a number of groups have developed in vitro methods to evaluate the presence of HLA-specific B cells capable of alloantibody production on stimulation. Luque et al. have demonstrated "Flourospot" techniques that permit semiquantitative, multiplex detection of HLA specific B cells that secrete antibody following B cell stimulation (8). Karahan et al. demonstrated a bulk-culture method stimulating PBMC with TLR8/9 ligands and IL-2 to elicit HLA antibody production and screening culture supernatant using a commercially available SAB Luminex ® platform (9). These methods demonstrate the ability to detect HLA-specific humoral memory, likely residing in the Bmem cell compartments; yet the exact in vivo frequency and phenotype of DSA B cells, and the characteristics of DSA B-cell antigen receptors (BCRs) remain obscure. A recent study by Heidt and colleagues has moved the field a step forward by using HLA tetramers to sort single HLA-specific B cells to allow more accurate characterization of their frequency and BCR structural analysis (10).
Here we present new tools developed to aid in the identification and characterization of B cells sensitized by allo-MHC exposures and to advance research in human and animal transplantation models. In our previous studies with influenza-vaccinated subjects, a single B-cell culture method was established to profile hemagglutinin specific Bmem from PBMC samples (11). This culture system supports the proliferation and differentiation of single naïve and Bmem cells and secretion of clonal antibody into culture supernatants. The amount of antibody secreted is sufficient so that antibody specificities are readily determined with the culture supernatants, while the V(D)J DNA rearrangements encoding BCRs can be recovered from expanded B-cell clones in culture wells for sequencing and re-expression. For antibody screening of the culture supernatants, we developed a cell-based immunoassay platform for the detection of antibody binding to cell-surface expressed antigens in a multiplex way (submitted). In this study, we further adapted this cellular platform and established an economic, versatile, and multiplex reporter cell assay for the detection of antibodies specific for MHC molecules from different mammalian species. Coupled with the established single Bmem cell culture system, we identified and characterized an anti-HLA B memory clone from one allo-sensitized patient.

Mice, Rhesus Macaques 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 mouse for RNA extraction. All experiments involving animals were approved by the Duke University Institutional Animal Care and Use Committee (IACUC A128-20-06).
Serum samples from rhesus macaques were obtained from animals sensitized by serial skin transplantation as part of an established transplant model (12). Naïve juvenile male rhesus macaques received two swapping skin transplants from their maximally Mamu mismatched donor, at eight-week intervals. Day 0 samples refer to naïve sera, while "Day 54-70 represent 'peak' samples, taken 2 weeks after the second skin transplant (IACUC A153-18-06).
Peripheral blood mononuclear cells (PBMCs) were obtained from one human subject under Duke Institutional Review Board Committee guidelines (IRB Pro00062495) and archived sera from transplant recipients obtained under IRB Pro00104220.

HLA Typing and HLA Antibody Analyses
High-resolution HLA typing was performed by next generation sequencing using the MIA FORA kit (Immucor, Inc., Norcross, GA, USA) on an Illumina Miseq platform (Illumina, Inc., San Diego, CA, USA). HLA specific antibodies were evaluated using HLA single antigen Luminex ® beads (SAB) according to manufacturer's instructions and acquired on a LABScreen 200 instrument (One Lambda, Canoga Park, CA). HLA antibody positivity was determined by HLA pattern analysis incorporating serological cross-reactive patterns and HLA matchmaker eplet analysis FUSION version 4.3 (One Lambda, Canoga Park, CA) and HLA Epitope Registry (https://www.epregistry.com.br/). Recombinant antibodies: P3E7 and P3E7-UA were tested at 0.16 mg/ml and P3E7 was also tested at 0.96 mg/ml (~6x) and analyzed as described above. Three-dimensional structures of the HLA-A molecules were created using Cn3d software (https:// www.ncbi.nlm.nih.gov/Structure/CN3D/cn3d.shtml).

Plasmid Modification and Gene Cloning
Standard molecular cloning procedures were followed for plasmid modification and gene cloning. Endotoxin-free plasmids were prepared (E.Z.N.A. ® Endo-free Plasmid DNA Mini Kit II, Omega Bio-tek) for mammalian cell transfection. All plasmids were verified by DNA Sanger sequencing (Duke University DNA Analysis Facility).

Cell Surface and Intracellular Staining, Flow Cytometry Analysis and Single-Cell Sorting and Cloning
For cell surface staining, cultures of K530 and derivative cells were harvested, centrifuged at 300× g for 2 min at 4°C and resuspended in staining buffer (PBS supplemented with 2% heatinactivated 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 intracellular staining, Fixation/Permeabilization Solution Kit (BD 554714) was used and manufacturer's instructions were followed.
Flow cytometry analysis was carried out using BD FACSCanto II cytometer (Duke Cancer Institute Flow Cytometry Shared Resource). Single-cell sorting was performed with BD Aria II (The Duke Human Vaccine Institute Research Flow Cytometry Facility). The bulk cell line after transfection or transduction were labeled 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, robustly 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.

Multiplex Reporter Cell Assay
Fluorescence-barcoded reporter cell lines expressing unique MHC antigens were pooled and labeled with reference antibodies, diluted serum samples or single B-cell culture supernatants. After a secondary labeling with PE-conjugated secondary antibody, the binding activities were detected by flow cytometry as described above. To increase the throughput, 96-well V-bottom plates were used for staining and a highthroughput sampler was used during flow cytometer detection.
Single antigen-specific B-cells were sorted into 96-well culture plates with MS40L low feeder cells (19,20) and in the presence cytokines, including human IL-2, IL-4, IL-21 and BAFF (all Peprotech). After 25 days' culture, supernatants were harvested for screening the reactivity of clonal IgG Abs. Expanded clonal B cells were frozen for V(D)J sequence analyses. IgG + culture supernatants were tested for HLA binding activities with HLAexpressing cell lines in a multiplex reporter cell assay.
Rhesus IgG1 and human IgG1 versions of W6/32 and human IgG1 version of BB7.2 antibody expression vectors were prepared by cloning the heavy and light chain V(D)J sequences (16) from hybridoma cell line cDNA samples to corresponding expression vectors mentioned above. V(D)J sequences of HLA-specific Bmem cell clone H02P3E7 and the inferred UA (synthesized, Gene Universal Inc.) were cloned into human IgG1 and IgK expression vectors as described (15). Recombinant antibodies were produced and purified as described (11).

Luminex ® Assay
Luminex ® beads conjugated with anti-human IgG were incubated with serial dilutions of IgG + culture supernatants or recombinant human IgG1 antibodies. After washing, either PE-conjugated goat anti-human IgG or PE-conjugated HLA-A2 tetramers were added as the detection reagents. The concentrations of bead-bound human IgG1 antibodies were normalized to a commercial isotype control antibody and the antigen binding activities of the bound antibodies were revealed by corresponding MFI values in the tetramer-PE staining group.

Generation of a Versatile Cell Line for Surface Expression of MHC Class I and II Molecules From Different Mammalian Species
We recently developed (submitted) the K530 cell line, a CD32Adeficient derivative of K562 cells, which was subsequently used as the parental cell line for endogenous barcoding by fluorescent proteins (FPs) and cell surface antigen expression. The parental K562 line was derived from a chronic myelogenous leukemia and does not express HLA class I and class II products (23) but is positive for b2M.
Concerned that this endogenous human b2M could interact with transduced MHC class I a chain from non-human species and generate chimeric class I molecules with altered structure and epitopes (24)(25)(26), we further modified the K530 cell line by knocking out the human B2M gene using a standard CRISPR-Cas9 mediated gene targeting strategy. Briefly, K530 cells were transfected with a CRISPR-Cas9 targeting vector co-expressing Cas9 nuclease and single-guide RNAs towards exon 1 region of human B2M gene ( Figure S1A). After puromycin selection, more than 70% of surviving cells showed decreased intracellular b2M expression comparable to that of K530 cells stained with an isotype control antibody ( Figure S1B). A monoclonal cell line, K528, was generated by single-cell flow cytometry sorting. The lack of b2M expression in K528 cells was confirmed by intracellular staining (Figure S1B), and the inactivation of B2M genes validated by genomic DNA sequencing ( Figure S1C).
The resultant clonal cell line, K528, is negative for HLA class I, class II and b2M but expresses low levels of cytoplasmic CD74 ( Figure 1A, Figure S2). The K528 line can readily be made to express MHC molecules from any mammalian species. To demonstrate this versatility, the mouse class I molecule H2Kb was transduced in either K530 or K528 cells, i.e., in the presence or absence of human b2M, respectively. The b2M-independent H2Kb mAb (clone 5041.16.1), recognizes transduced H2Kb comparably on K530 (human b2M+) cells transduced with H2Kb alone or with co-expressed mouse b2M; on transduced K528 (human b2M-) cells, the 5041.16.1 mAb binds only to cells expressing both H2Kb and mouse b2M ( Figure 1B). In contrast, the mouse b2M-dependent H2Kb mAb (clone AF6-88.5), recognizes only cells co-transduced with H2Kb and mouse b2M. Significantly, AF6-88.5 mAb binding to co-transduced K530 was significantly reduced by the presence of human b2M, suggesting that the endogenous human b2M efficiently competes with transgenic mouse b2M to generate chimeric class I molecules (compare 5041.16.1 binding to AF6-88.5 binding). Only K528 cells co-expressing mouse b2M supported normal expression of surface H2Kb molecules as determined by the mouse b2M-dependent H2Kb-specific antibody ( Figure 1B). In the absence of endogenous human b2M, pMHC molecules from other species with fixed peptide can also be expressed by K528 cells as single-chain trimers with disulfide traps (dtSCT) (18) ( Figure 1C). For a mouse class II molecule IAb, co-expression of mouse CD74 increased the surface expression level dramatically. A chimeric mouse CD74 molecule with the core CLIP peptide replaced with an exogenous antigenic peptide drove the presentation of corresponding peptide by IAb (27), suggesting that the natural class II molecular processing and antigen presentation pathways can be simulated in these cells ( Figure 1D). As for human class II molecules, co-expression of human CD74 results in much higher surface expression of human DQ molecules ( Figure 1E). We conclude that K528 is a versatile parental cell line for surface expression of MHC class I and II molecules from different mammalian species.

Multiplex Detection of Anti-HLA Antibodies in Serum Samples From Allo-Sensitized Patients
Using the newly generated K528 cell line, a basic panel of 16 fluorescence-barcoded sub-lines were generated in the same way as for K530 cell line (submitted). Briefly, K528 cells were transduced to express different combinations of four FPs which can be discriminated from each other by flow cytometry, resulting in 16 sub-lines with unique fluorescence barcodes. To establish a general tool for anti-HLA antibody screening, we designed a basic panel consisting of ten HLA class I and four class II molecules which are highly prevalent in our patient population. Class I genes were transduced into barcoded cell lines along with human b2M whereas class II genes were incorporated along with human CD74. The resultant HLAexpressing reporter-cell panel was subsequently validated for specificity using a panel of HLA-specific monoclonal antibodies. All 14 HLA-expression cell lines were pooled along with two internal control cell lines without HLA transgene but express all or none of the barcoding FPs, respectively. The pooled reporter cells were then labeled with HLA-specific reference antibodies individually followed by a PE-conjugated secondary antibody. The labeled and washed cells were then examined for FP and PE signals by flow cytometry. Individual reporter cell lines were demultiplexed by gating for specific combinations of FP expression ( Figure S3). Binding activity by each HLA reference antibody is proportional to the median fluorescence intensity (MFI) for each reporter cell line in the PE detection channel. In consequence, the specificity of each reference HLA antibody can be determined within the reporter cell line panel used for screening. In all cases, the binding patterns of each reference antibody matched its published specificity (Figure 2A).
This reporter cell panel permitted detection of anti-HLA antibodies present in serum samples from allo-sensitized patients. Ten blinded, serum samples were tested ( Figure 2B, Figure S4) and the FACS MFI readouts compared to corresponding Luminex ® readouts from a clinical HLA SAB assay ( Figure 2C). No HLA antibodies detected by the reporter cell assay were absent in the clinical SAB assay, confirming the high specificity of the reporter cell assay. In contrast, for 59 true positive (TP) readouts in the SAB assay, only 33 were detected in the reporter cell assay. Most (22/26, 85%) of the discrepancies correlated with lower bead MFI values when compared to HLA alleles that tested positive with the reporter cell assay. These discrepant results may reflect lower sensitivity of the reporter cell assay due to the lower density of HLA antigen on the cell surface compared to beads. This lower density would result in lower signals, especially when the relevant antibody is at low concentration in the serum (28). It is also possible that these "false negatives" in the reporter cell assay reflect spurious antibody binding in the SAB assay due to the presence of denatured HLA antigens (29)(30)(31). However, for this study and in clinical practice, this type of spurious background binding is mitigated by the use HLA pattern analysis to determine TP SAB reactions. We conclude that the multiplex reporter cell assay likely exhibits lower sensitivity than the clinical SAB assay. If this is the case, its application should be limited to conditions when the concentration of relevant antibody is not limiting, e.g., with monoclonal antibodies or single B-cell culture supernatants. In the case of single B-cell cultures, the clonal antibodies in most (≥ 90%) culture supernatants reach concentrations above 1 µg/ml (11). Under these conditions, the reporter cell assay is an economic and effective alternative to clinical SAB assay in high-throughput screening.

Multiplex Detection of Mamu Antibodies in Serum Samples From Sensitized Rhesus Macaques
Rhesus macaques (RMs) are a valuable animal model in many pre-clinical studies, including transplantation. However, specific reagents for studies with RMs are limited. To our knowledge, only one allele-specific Mamu antibody has been reported (32) and there is no single Mamu antigen bead panel available commercially. The multiplex reporter cell platform we established for HLA is readily adaptable to Mamu antigen expression. To demonstrate this potential, a panel of class I Mamu molecules and rhesus b2M were expressed in FPbarcoded reporter cell lines by transduction. Robust surface expression of Mamu proteins was confirmed using the crossreactive antibody, W6/32 ( Figure 3). To validate the specificity and utility of this reporter cell panel, we tested 15 blinded serum samples from allo-sensitized RMs following two sequential skin transplants (12) for the presence of allo-Mamu antibodies. We predicted the elicited antibody reactivities based on donor Mamu haplotypes that were mismatched with recipient haplotypes, and compared these to the binding activities observed in the reporter cell assay. As shown in Figure 3 and Table 1, except for a single case in RM H73E, we detected binding activities for all the predicted antibody responses (23/24, 96%). This single negative result is either a false negative due, perhaps to the relative low density of Mamu proteins expressed by the reporter cells, or to the non-responsiveness of the particular allo-antigen in this host RM.
On the other hand, 12 binding signals beyond the predicted reactivities were detected. These positive readouts could be false positive results due to expression artifacts in the reporter cells, orand we believe more likely -the result of cross-reactive epitopes shared between Mamu allo-antigens. Unfortunately, validated reagents to rule out either of these two possibilities do not exist; however, in view of the high specificity of the HLA reporter cell assay above (Figures 2A, B), and considering the well documented cross-reactive epitopes shared between HLA molecules ( Figure 2C), we believe that these positive readouts represent true cross-reactivities in allo-sensitized RMs. The potential identification of novel, cross-reactive Mamu epitopes, highlights the utility of this multiplex reporter cell assay in experimental models where multiplex bead reagents are unavailable or difficult to generate.
Of note, serum samples from RM HADV at Day 0 and from H73E at Days 0 and 70 exhibited high background staining of all reporter cell lines including the internal control lines bearing no MHC molecules; with dilution this polyreactivity became negligible (compare serum dilutions of 1/50 to 1/500). Our records show no significant differences in pre-sampling treatments for these animals and we were able to exclude technical issues in the binding assay. Consequently, we consider this serum polyreactivity to reflect intrinsic properties of the animals or samples. Possible explanations include 1) presence of polyreactive, presumably low affinity, serum IgG, 2) serum IgG antibody specific for an unknown antigen(s) present on the parental K528 cells, and 3) serum components, e.g., complement-decorated antibody/antigen complexes, which mediate non-specific binding of antibodies to the reporter cells.

Evidence for Affinity Maturation in an HLA-Specific Bmem Cell From a Sensitized Patient
We established an antigen-specific single-cell culture system recently (11,20,33,34), which supports robust proliferation of single Bmem cells (to ≈90,000 daughter cells by day 25) followed by plasmacytic differentiation and IgG production, yielding an  The specificities were predicted based on donor Mamu haplotypes that mismatch with host ones. Positive readouts matched with prediction were indicated with green arrows. Positive readouts beyond prediction were either false positive detections or due to cross-reactivity and were indicated with black arrows. Predicted but undetected specificities were either false negative detections or due to non-allo-reactivity of the donor Mamu molecules in the corresponding host, and were indicated with blue arrows. B012 stands for B012a or B012b.
average 50 µg/ml of clonal antibody in culture supernatants. This robust production of clonal IgG enables high-throughput screening for MHC-specific clones using the reporter cell assay described in this report. With this culture system, the V(D)J rearrangements encoding MHC-specific BCRs can be readily recovered by PCR amplification, sequenced, and expressed as recombinant antibodies (rAbs) for further characterization.
We employed this single-cell culture method to isolate and characterize HLA-specific Bmem cells and to generate HLA rAbs from an allo-sensitized patient. The patient was sensitized by three pregnancies and a history of transfusion. Serological HLA-A*02:01 reactivity was detected by SAB as the immunodominant or strongest antibody as measured by mean fluorescent intensity (MFI). PBMCs from this patient were labeled with A*02:01 tetramer along with surface markers (CD19 + CD27 + CD24 hi IgD -IgM -IgG + ) to identify Bmem cells. Tetramer + IgG + Bmem cells ( Figure 4A) were sorted into single B-cell cultures as described (11). After 25 days of culture, a cloning efficiency of 40% (270/672) was determined by screening for secreted IgG in culture supernatants. IgG + supernatants were then tested for HLA binding activity using the HLA reporter cell assay. One clone, H02P3E7, exhibited avid binding to A*02:01 but not to A*01:01 ( Figure 4B).
We sequenced the V(D)J rearrangements present in this HLA-A*02:01 specific clone by amplifying the human IgG heavy-and light-chain genes from the expanded, clonal B-cell population in the corresponding culture (15,35). The recovered and sequenced BCR genes carried mutations (>1%) in both the VH and VL gene segments ( Figure 4C), consistent with this IgG + Bmem cell clone being derived from a germinal center (GC) response. A recombinant antibody (rAb), P3E7, was generated from the V(D)J sequences present in the cloned Bmem cell H02P3E7 and a second, P3E7-UA, from the inferred, germline V (D)J sequences expressed by the unmutated ancestor [UA, (36)] of the H02P3E7 Bmem cell. The binding of both rAbs to A*02:01 was confirmed in a Luminex ® assay using A*02:01 tetramer as the antigen target ( Figure 4D). The binding avidity of rAb P3E7 was comparable to that of BB7.2, a hybridoma derived A*02:01specific antibody, and has a calculated Kd value of 68.5 pM. Interestingly, the binding activity of the rAb P3E7-UA to A*02:01 was decreased by 4-logs compared to the mutated rAb P3E7. In a reporter cell assay ( Figure 4E), we confirmed that both rAbs can bind to cell-surface expressed A*02:01 antigens, with P3E7 showing much more avid binding activity than P3E7-UA does. Moreover, P3E7 also shows binding activity towards B*57:01, suggesting that this antibody binds to a shared epitope in these two alleles. This observation supports the notion that in the sensitized patient, this Bmem cell resulted from HLA-specific affinity maturation in a GC prior to its entry into the Bmem compartment.
In addition, we determined the HLA allele binding patterns of rAbs P3E7 and P3E7-UA using a clinically validated HLA class I 100-plex SAB panel. The threshold for positivity was determined by examining MFI values for the negative controls and the point at which the HLA allele binding pattern was interrupted by nonrelated HLA-B and -C alleles. Tests of the unmutated P3E7-UA rAb displayed strong reactivity for beads bearing HLA-A*02 and B*57 alleles (23,679 -19,171 MFI), lower binding to B*15:16 (B63) and B*58:01 (15,685 -9,377 MFI), and the lowest binding to A*43:01 and A*29 alleles (3,815-1,866 MFI) ( Figure 5A). Using the HLA Epitope Registry (https://www.epregistry.com. br/), we identified a common HLA amino-acid motif or "eplet" shared across the HLA alleles to which the rAbs bound. The 10 bound HLA alleles share homology within a potential 15 angstrom conformational epitope surrounding an arginine residue at amino acid position 65 defined by HLA eplet 65RA (Figures 5B, E). As expected, stronger reactivity was observed for the mutated P3E7 rAb compared to the unmutated P3E7-UA ( Figures 5A, D) along with an extended binding pattern to include 13 additional HLA alleles also containing HLA eplet 65RA.
Bordering amino acid differences influence antibody binding to HLA conformational epitopes; binding variability, however, may also reflect the quality and density of HLA molecules on individual Donor haplotypes mismatching with recipient haplotypes, which had a matched reporter cell lines in the detection panel and would therefore be considered as a predicted positive readout, are highlighted in bold. B012 stands for B012a or B012b.
Luminex ® beads. Simultaneous comparison of P3E7 and P3E7-UA binding at identical concentrations allowed us to control for bead differences and examine how BCR mutation affected BCR affinity and specificity. Our analysis revealed that affinity maturation of P3E7 increased binding to non-HLA-A*02 alleles with variable residues at position 45 and 62 increasing MFI values (>15,000) for HLA-B*58, A*29, B*15:16 and A*43 from that observed for the progenitor P3E7-UA antibody ( Figure 5B). This expanded reactivity also allowed binding to additional HLA alleles bearing the 65RA eplet but with substantial variability about this conformational epitope. Interestingly, in the course of its somatic evolution, the H02P3E7 Bmem cell acquired the capacity to recognize HLA alleles (A*11:01 and A*03:01) present within the patient; this auto-reactivity was not present in the germline P3E7-UA rAb. When P3E7 was tested at a higher concentration (6x) the same reactivity pattern was maintained and extended to include A*68, A*69, A*01, and A*33:01 and correspondingly higher (> 6,000) MFI values were observed for the autologous HLA   Figure 5E). Therefore, the process of affinity maturation in this patient was coupled to the generation of humoral autoreactivity.  (39). Attempts to use HLA-specific SAB reagents to characterize NHP alloantibody responses have been made (40), albeit with limited resolution of specificity. In most NHP research, therefore, the standard assay for alloantibody detection relies on a donor-cell antibody capture method analogous to the flow cytometric crossmatch used in clinical transplantation (12,41,42). Unfortunately, antigen-specific alloantibody quantitation, such as the HLA SAB Luminex ® assay provides, has not been realized, despite conserved epitopes permitting detectable positivity (43). Our use of FP barcoded cells expressing Mamu antigens ( Table 1) allows for multiplex detection of multiple alloantibody specificities within the same sample, with good detection of alloantibody directed against sensitizing antigens. The additional positivity detected in the Mamu reporter assay may reflect false positives but, as the single-cell human rAb data suggest, more likely represents unknown cross-reactive binding across similar epitopes. Since this degree of alloantibody specificity has not been widely available these new tools allow for a greater understanding of shared epitopes, binding patterns, hierarchy among Mamu (or Mafa) antigens.

DISCUSSION
In human studies, the utility of HLA expressing reporter cells may be limited to that of a screening tool, given the apparent greater sensitivity of the SAB Luminex ® ; assay. For research purposes, however, SAB testing with commercially available products can be prohibitively expensive. For example, in the case of single-cell cultures in which hundreds or thousands of individual culture wells contain clonal IgG, SAB bead analysis of each well could present a daunting expense. As for the analogous Luminex ® ; bead assay, the barcoding of each HLA or Mamu expressing cell line with a unique pattern of FPs permits multiplexing of the assay in a single tube, limiting the volume of sample necessary for antibody detection.
In conjunction with our barcoded reporter cell lines, we used a single human B-cell culture system (11,20,33,34) to isolate and characterize HLA-specific Bmem cells from an allo-sensitized patient. In previous studies, we sorted hemagglutinin-binding Bmem cells from vaccinated subjects into single-cell cultures and about half of the cultured Bmem cells produced hemagglutinin-specific IgG (11). In this study, however, from 270 cultures of Bmem cells identified by HLA-A*02:01 tetramer binding only a single clonal culture produced HLA-specific IgG ( Figure 5B). The low frequency of HLA-specific clones among tetramer-binding B cells is likely the consequence of the low specificity of the antigen-tetramer "hook" and perhaps, the absolute frequency of HLA-specific B cells. As HLA-specific Bmem cells are usually at low frequency in allo-sensitized patients (10,44), development of high-quality HLA antigen hooks will be critical for efficient isolation of specific Bmem cells.
With the single HLA-specific B cell clone identified, we determined the BCR V(D)J sequences and identified mutation frequencies of about 1% in the V genes of both the heavy-and light chain gene rearrangements ( Figure 4C). We confirmed the avid binding of the A*02:01 tetramer by secreted clonal IgG with a recombinant antibody prepared from the recovered BCR gene sequences. In addition, the inferred germline progenitor (UA) of the recovered Bmem clone was also generated and comparison of P3E7 and the UA recombinant antibodies provided evidence for an avidity increase of some 10 4 -fold as a consequence of somatic hypermutation ( Figure 4D). These data, and the Bmem phenotype recovered by flow sorting (Figure 4A), suggest strongly that this B cell clone was a typical class-switched Bmem cell derived from a GC response after successful rounds of affinity maturation. This is an example, albeit a single one, of HLA-specific Bmem cells in sensitized patients. Moreover, the strong binding avidity of the mature rAb implies that the H02P3E7 Bmem cell would likely be capable of differentiating into a plasma cell (45), and to secrete HLA antibody on rechallenge. Antibody of this affinity would impose a threat of an acute AMR after transplantation even if HLA serum antibody titers were negligible prior to transplant (44). This observation highlights the significance of monitoring of HLA Bmem cells in pre-transplant patients in clinic.
An interesting observation with the HLA Bmem recovered is that its mutated and affinity-matured BCR acquired autoreactivity absent in its germline precursor. Whereas the inferred germline antibody was specific for the products of several allo-HLA-A*02 alleles, the more avid BCR expressed by the Bmem H02P3E7 daughter cell acquired the capacity to bind autologous HLA molecules in the course of its somatic evolution ( Figure 5). The acquisition of self-reactivity is known to be associated with human Ig-class recombination (35) and, presumably, clonal maturation in GCs. Nonetheless, most current models for GC responses posit negative selection to limit the generation of autoimmunity by AID-mediated hypermutation (46)(47)(48)(49). Indeed, we observed the opposite of what Goodnow and colleagues have named clonal redemption (47,48). Rather than the recruitment of autoreactive B cells into GCs where they lose autoreactivity in the process of affinity maturation to an exogenous antigen, the naïve P3E7-UA cell was presumably recruited to a GC on exposure to allo-HLA and there its progeny acquired reactivity to self-HLA in conjunction with allo-specific affinity maturation ( Figure 5). Surprisingly, this de novo autoreactivity was not against some rare or cryptic selfantigen but to universally expressed HLA class I determinants.
While it is possible that the binding activities to self HLA molecules detected in the SAB assay are not physiologically relevant, it will be interesting to determine if the relevant membrane associated HLA molecules trigger BCR signaling in B cell lines harboring the matured P3E7 BCR. Additional study will be needed to determine whether autoreactivity is a common phenomenon among allo-HLA Bmem cells.
These new cell tools permit multiplex detection of allo-MHC antibody in humans and non-human primates. The utility of this approach is limited only by the number of FP barcoded cell lines and has the potential of multiplexing up to 256 distinct cells for antibody detection in single samples. Additionally, we have demonstrated the application of barcoded reporter cells in screening for HLA-specific antibody production by single-cell cultures of human Bmem cells. In this way, we characterized and expressed rAb derived from a human Bmem cell, demonstrating its somatic evolution by mutation and selection for increasing affinity. We rapidly identified the 65RA eplet as the putative epitope for the Bmem BCR using SAB HLA-Luminex ® assays. These combinations of new and established techniques offer increased understanding of humoral responses to allo-MHC.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: GenBank accessions MZ152901, MZ152902, MZ152903, MZ152904.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Duke Institutional Review Board Committee. The patients/participants provided their written informed consent to participate in this study. The animal study was reviewed and approved by Duke University Institutional Animal Care and Use Committee.

AUTHOR CONTRIBUTIONS
SS, SK, and GK contributed to conception and design of the study. SS, MM, JK, AJ performed the experiments described in this study. SK and GK helped to review and trouble-shoot all experiments. SS, MM, JK, and AJ wrote specific sections of the initial manuscript. All authors contributed to the article and approved the submitted version.