Assessing the allogenic realness of the Cw1/12/15 pattern occurring in the LABScreen single antigen assay

Several technical limitations of Luminex single antigen (LSA) assays have been described so far. This study focused on a reactivity pattern observed in many sera that cannot be explained by eplets described in the Epitope Registry database and sometimes appearing against a self‐HLA allele or antigen. In most cases, this pattern is revealed by a discrepant result when compared with other assays (Luminex PRA, cell‐binding assays such as flow cytometry cross match, LSA from another manufacturer…). We focus here on the Cw1/12/15 pattern appearing on the LABScreen class I LSA provided by One Lambda. We documented its behavior using this LSA after acid denaturation of the beads, using Lifecodes LSA from Immucor, and adsorption of sera either on spleen mononuclear cells from deceased donors or on single HLA transfected cell clones. We studied 33 sera from different patients positive for the three Cw beads, selected from our routine patients' LSA database. Nine patients had transplants from a Cw12 or Cw15 donor without any pejorative evolution of the graft, nor post‐transplant MFI (mean fluorescence intensity) increase of the Cw1/12/15 beads. A significant increase of MFI was observed after acid denaturation of the LABScreen beads. All sera tested by Lifecodes LSA were negative for these Cw beads. Finally, we found no significant difference of MFI after adsorption on cells from either origin. Therefore, the Cw1/12/15 pattern appears to be a false positive reactivity of the LABScreen single antigen assay.

be public that is, shared by several to multiple antigens. Nevertheless, the determination of compatibility between a recipient and a donor still relies on the antigen level in organ allocation programs worldwide. The deleterious clinical implication of eplet mismatching has been clearly demonstrated over the last years, at least for some eplets, in patients involved in organ transplantation programs. 2,[4][5][6][7] Donor HLA typing cannot be obtained at an eplet level with certainty for deceased organ donors due to time and technological constraints. Therefore, in order to take into account eplets, HLA laboratories use the HLA typing performed for recipient and donor, to extrapolate manually, or by using dedicated software, the eplet mismatch content for a given donor. Besides this, the profile of immunization of the recipient is now generated from the multiplex serum analysis of nearly 200 common HLA alleles in a Luminex based "single antigen" flow beads anti-HLA antibody assay (LSA). After nearly 15 years of use of LSA, several technical limitations were highlighted, and among them the presence of unexplainable serum reactivity patterns that may or may not encompass at least one of the recipient's own HLA alleles. Situations were described, of sera displaying high LSA MFI signals, that were negative in FlowPRA assay or flow cytometry cross match for the same antigens. [8][9][10][11] Other situations revealed discrepant results when comparing the two existing platforms of LSA assays, LABScreen by One Lambda ® , and Lifecodes by Immucor ® , 12 suggesting that one or the other assay could provide false positive or false negative results.
False positivity can also be due to antibodies (Abs) targeting denatured HLA proteins displayed on the surface of the LSA beads besides native HLA molecules targeted by clinically relevant DSA. 13 Acid-treatment of the beads fully denatures conformation-dependent native HLA epitopes. This process enriches the denatured targets on these LSA beads thereby increasing their MFI after incubation with serum. 14,15 Both providers are concerned to different degrees, 16,17 since denatured proteins are generated throughout the different steps between protein production and binding to Luminex beads due to the intrinsic fragility of the HLA molecule. However, erroneously identifying positive allele or antigen specificities in a serum will have negative consequences on patient management, either through excess detection of DSA posttransplant and possibly unjustified immunosuppression, or through a decrease in access to a compatible organ pre transplant.
In the present study, we carried out a detailed analysis of the reactivity pattern targeting the three C*01:02, C*12:03, and C*15:02 HLA-Cw beads in the class I LABScreen assay, herein named Cw1/12/15. This pattern has already been reported 18,19 but to our knowledge, it is the first experimental study exploring it with several complementary approaches. We indeed used denaturation of LSA, comparison of the two vendors' LSA assays, and patients' serum adsorption on cells from organ donors and from single HLA molecule transfected cell clones. To us, such an exhaustive analysis of "suspect" immunization patterns represents the most appropriate strategy to validate or invalidate any eplet and serum reactivity pattern in order to appreciate with the highest accuracy both the potential clinical relevance of suspected eplets, and the consequences of identified anti-HLA antibodies on immunological risk assessment in organ transplantation.

| Selection of patients' sera
Sera were selected from the MFI results of 135,225 LSA performed with LABScreen assay (One Lambda, Canoga Park, CA) in the course of routine patient follow-up between September 15th, 2015 and February 29th, 2020 at the Histocompatibility Laboratory of the Saint-Louis Hospital in Paris, France. September 15th is the day when systematic EDTA pre-treatment of serum was implemented, in order to circumvent the complement interference phenomenon. A first selection step retained the sera with a mean MFI value for the Cw1/12/15 beads between 2000 and 15,000 to avoid saturation effects due to the lack of linearity of the technique at high MFI values. A second selection step removed the sera with more than three other Cw beads above 500 to avoid samples containing complex profiles but also to take into account samples reacting against other beads besides the Cw1/12/15 restricted pattern. The remaining sera were from 80 different patients that were classified in five groups: (A) patients waiting for transplantation who tested negative for HLA-Cw1/12/15; (B) patients waiting for transplantation who tested positive for at least one of the three Cw1/12/15 antigens; (C) patients who have received a transplant from a donor who was negative for Cw1/12/15 antigens; (D) patients who have received a transplant from a donor who was positive for at least one of the three Cw1/12/15 antigens; (E) transplanted patients who were positive for at least one of the three Cw1/12/15 antigens and have received a transplant from a donor who was also positive for at least one of the three Cw1/12/15 antigens. Among these 80 patients, we retained those with at least five sera tested by LSA (33 patients). For each patient, the most recent serum that tested positive for the Cw1/12/15 pattern was selected for further experimental analysis. HLA class II LSA reactivity was not considered.

| Spleen mononuclear cells from deceased donors
Spleens from deceased organ donors were used within the 48 h that followed the pre transplant cross matches for which they were sent to the laboratory. Selected donors had to express at least one of the C*01:02, C*12:03, and C*15:02 alleles, without considering the other HLA loci. Spleen mononuclear cells (SMCs) from deceased donors were obtained by Ficoll density gradient (CMSMSL01-01, Eurobio, France). Intermediate resolution typing for the 11 HLA genes of donors were performed with LinkSeq SureTyper RT-PCR (One Lambda, Canoga Park, California).

| HLA transfected cell lines
We used EL4 S3-Rob mouse cell clones expressing HLA-C*01:02, C*12:02, or C*15:02 as targets. The EL4 S3-Rob cell is deficient in murine beta-2-microglobulin and was transfected with beta-2-microglobulin/HLA fusion constructs as already described, 20,21 Lemonnier. The C*12:02 allele was not the one displayed on LSA beads, but the C*12:03 allele was not part of our available cDNAs. EL4 S3-Rob cells were electroporated with 20-μg HLA-C plasmid linearized DNA at 225 V, 400 μF using a Gene Pulser Xcell (Biorad France, Marnes-La-Coquette, France). Then the cells were transferred in RPMI completed with 10% of fetal calf serum and Penicillin/Streptomycin. After 24 h, selection medium containing 40 μg/mL of blasticidine was added and cells were cloned by limiting dilution. Blasticidine-resistant clones expressing the HLA-Cw allele of interest were analyzed by flow cytometry using anti-HLA class I (W6/32) APC-Cy7 conjugated antibody (Biolegend #311426; isotype control #400230). Nontransfected cells were used as the negative control.

| LSA assays
Both LABScreen and Lifecodes (Immucor, Norcross, Georgia) LSA assays include the C*01:02 and C*15:02 beads. Only the Cw12 allele differed, with C*12:03 for LABScreen and C*12:02 for Lifecodes. All samples were run in the routine assay with LABScreen Single Antigen HLA Class I (reference LSA1A04) according to the manufacturer's instructions with half of the recommended amount of LSA beads. Routine follow up of the patients included in the study covered four different lots of LSA beads, lots 9-12 ( Table 1). The Lifecodes LSA class I single antigen (LSA1 lot 3011733) was modified as follows: 15 μL of serum were incubated with 15 μL of beads leading to a 1/2 serum dilution and the conjugate was used at a 1/15 dilution. For both LSA assays, microtitration filter plates were used for incubation and washing steps (Merck, MSBVN1250, 1.2 μm), and were read on a Luminex platform (LABScan 200, Luminex, Austin, TX). For LABScreen LSA, raw MFI data were normalized for each HLA bead by subtracting the MFI of the negative HLAfree control bead. All sera were pre-treated with EDTA at a final concentration of 0.01 M, and incubated 10 min at room temperature before running the assays.
For the acid-treatment assay, a saturation step was performed by incubating the LABScreen beads with 20 μL of the patient's serum during 30 min. Then, plates were washed five times and acid buffer (Glycine HCl 0.1 M with 1% BSA, pH = 2.2) was added for 1 min, removed by filtration, and added again for 15 min at room temperature. After three washes, serum was added again for 30 min and the LSA assay was performed as usual.

| Adsorption of patients' sera on cells
Twenty microliters of pre-treated serum were incubated with a cell pellet containing 20 million of spleen mononuclear cells (SMCs) or 5 million of HLA transfected cells, for 2 h at +4 C. The mix is then centrifuged and supernatant is collected to perform the LABScreen assay. Each LSA run included the serum before and after the adsorption step, in order to compare MFI values and adsorption efficiency in the most rigorous setting, given the well-documented and relatively high interassay variability of the technique.

| Statistical analysis
Statistical analysis and graphs were performed using GraphPad software version 9.0 (San Diego, CA). Unpaired-t-tests were used as statistics for comparing LABScreen assays with acid-treated and non-treated beads. One-way ANOVA investigated statistical differences of MFI values before and after adsorption on cells. Fluctuating Preformed 3408, 4014, and 3039 against the Cw1, Cw12, and Cw15 beads respectively (Figure 1), and a MFI range from 1738 to 10,963. Among the 33 samples, 9 also displayed reactivities against additional HLA-C alleles: C*04:01, C*05:01, C*06:02, C*07:02, C*14:02, C*17:01, and/or C*18:02 ( Figure 1). Characteristics of the selected sera are summarized in Table 1 Table 1). The trajectory defined as "de novo" was for patients with at least the first serum found negative then becoming positive and remaining positive until the last serum collected (six patients among the 33). The trajectory defined as "fluctuating" was for patients with combinations of negative and positive sera (nine patients among the 33). The trajectory defined as "transient" was for patients starting with positive sera then followed by a persistent disappearance of the pattern (two patients among the 33). For the 16 remaining patients, the Cw1/12/15 trajectory was considered as "stable" as it was present in all the sera of the immunological follow up.
Among the 33 sera, 24 had also been tested with Labscreen mixed antigens beads (LSM12 kit) in the course of the laboratory routine. Fourteen sera were negative with these beads, and the 10 others were positive but these sera contained antibody reactivities against HLA-A and/or B antigens (data not shown). When analyzing class II LABScreen LSA for the 33 sera, beads patterns already known as false positive were observed for 24 of them (i.e., 73%). Indeed, 17 sera displayed the pan-DR reactivity pattern previously reported, and 15 sera were positive for at least one of the two beads of the DP1 and DP5 antigens (i.e., the DPA1*02:01/DPB1*01:01, and DPA1*02:02/DPB1*05:01 beads) 8,11,12,16,22 (Table 1).
We used the Immucor Lifecodes LSA assay to explore this pattern with a concurrent LSA assay. The protocol of this assay was modified by using sera diluted 1/2 instead of 1/5 as recommended by the provider, in order to increase the MFI values to a level closer to the LABScreen LSA assay, which uses non-diluted serum. To validate our adapted protocol, 20 random sera, selected in the course of our laboratory routine, were tested with LSA class I and II kits from both providers. Correlations between the two assays were considered satisfying, as illustrated in Figure S1, with both a hyperbola curve (R 2 = 0.78) and a linear regression curve (R 2 = 0.75). Criteria for positivity of a Immucor LSA bead were those recommended by the manufacturer that is serum's raw MFI value above 750 and the ratio between serum's raw MFI and the lowest bead raw MFI for the same locus greater than the manufacturer's threshold. We tested 14 sera among the 33 with this modified protocol, picked randomly among each category. None was positive for any of the Cw1/12/15 beads (Table 2 and Figure 2). For patient #2, the C*03:03, C*04:01, and C*07:01 beads were positive with the Lifecodes LSA assay although they were negative with LABScreen LSA assay.

| Acid denaturation of LSA beads
To assess whether the Cw1/12/15 pattern was caused by an antibody targeting denatured Cw antigens, we performed acid treatment of the LABScreen beads. If the serum antibody targeted a denatured epitope, the number of these target epitopes per bead was expected to increase upon acid treatment with therefore a significant increase of the measured MFI. In contrast, a native epitope should disappear upon acid treatment, leading to complete MFI loss. Intermediate behaviors suggest a at least partial resistance to acid treatment. 15 To validate the assay, two control sera were used. MFI values of the serum containing antibodies targeting denatured epitopes were indeed significantly higher after acid treatment, as expected (p = 0.02, Figure S2A). In contrast, the MFI values of a serum containing antibodies targeting the Epregistry antibody-verified 62GE eplet drastically diminished while MFI values of 62GE-negative beads increased, after acid treatment ( Figure S2B). Regarding the 33 patients, we overall observed a highly significant increase of the Cw1/12/15 MFI values after acid treatment (p ≤ 0.0001) (Figure 3). We could categorize sera into three groups: (1) sera displaying a mean MFI increase above 100% (20 sera), (2) sera displaying a 1%-100% mean MFI increase (12 sera), and (3) sera displaying a decrease in MFI after acid treatment (1 serum) ( Table 2). Acid treatment was not performed with Lifecodes LSA as none of the sera was positive on non-treated beads.

| Adsorption of the sera on cells
The 33 sera were adsorbed on SMCs. Among them, due to limited availability of suitable donors, seven sera were adsorbed on Cw1 cells, 13 on Cw12 cells, and 13 on cells expressing Cw1 and Cw12 (Table 2). No Cw15 donor was used, Cw15 being less frequent than Cw1 and Cw12 in our local donor population. We found no significant differences of MFI after adsorption on every cell with mean MFI of 3154 (±1758) before adsorption, 2185 (±961) after incubation with Cw1 cells, 2773 (±1855) with Cw12 cells and 3073 (±2373) with Cw1/Cw12 cells (p = 0.07, p = 0.66 and p = 0.99 respectively) ( Figure 4A). A patient's serum containing an antibody targeting the Epregistry antibody-verified 80 N eplet present on HLA-Cw1 and Cw12, among other Cw antigens, was used as a positive control and was significantly adsorbed on SMCs (p ≤ 0.0001 for Cw1, Cw12, and Cw1/Cw12 cells) ( Figure S3A). Among the 33 sera, only 21 were also adsorbed with transfected cell clones due to insufficient leftover volumes for the remaining 12. We did not notice any nonspecific binding on the HLA-negative control cell clone as the mean MFI remained stable, at 2438 (±1703) before adsorption versus 2583 (±2155) after adsorption (p = 0.99) ( Figure 4B). For the clones expressing HLA-Cw molecules, mean MFI was 2847 (±2317) for sera incubated with Cw1 cells, 2990 (±2719) with Cw12 cells, and 2772 (±2290) with Cw15 cells. No significant differences were observed between MFI of all sera before and after adsorption with the three transfected-cells (p = 0.83, 0.60 and 0.91 for the Cw1, Cw12, and Cw15 cells respectively) ( Figure 4C). The same serum targeting the 80 N eplet was also used as a positive control for Cw1 and Cw12 cells, and was completely adsorbed (p ≤ 0.0001) ( Figure S3B). We used another patient's serum reacting against the Epregistry antibody-verified eplet 80 K present on HLA-Cw15 among others Cw antigens, and we also observed a significant adsorption after incubation with Cw15 transfected cell clones (p = 0.0022) ( Figure S3B).

| DISCUSSION
The identification of anti-HLA antibodies by LSA assays is very efficient, fast and sensitive, allowing monitoring of allosensitization of patients. HLA molecules are noncovalent heterotrimeric complexes that are highly dependent on their overall conformation for their function, as well as for their recognition by allogenic Abs raised in transplant recipients. However, LSA assays use recombinant HLA proteins and consequently, such in vitro assays suffer from technical limitations associated with the use of recombinant proteins, that need first to be understood, then solved. Unexplained LSA positive beads patterns commonly occur and often include self-alleles present in patients who developed the reactive Abs. False positivity interferes with the confidence that a user may have in the assay, and may also have important consequences on the clinical management of patients. Indeed, pre-transplantation, it may strongly reduce access to a compatible donor, and post-transplantation, may lead to excess immunosuppression because of erroneously suspected DSA. The advantage of using a recombinant system is that it allows strong overexpression, in order to maximize protein yield for production of large batches. Nevertheless, other assays exist relying on HLA proteins naturally produced by cells under physiological conditions. These are the FlowPRA (One Lambda) assay based on beads (i.e., with several HLA alleles per bead, purified from normal human cells), and the classical cross match assay performed on cells, especially in its flow cytometry design that can reach a detection level close to that of LSA assays. One limit of our study could be that we did not perform these assays. However, FlowPRA is limited in the panel of alleles available, and flow cross matches are not suitable for large series of samples and depend on the availability of a large panel of cells to analyze a potentially wide variety of patterns.
In this study, we reported an exhaustive characterization of the Cw1/12/15 pattern, using several complementary approaches. Although this pattern was reported in the literature, this study is unique in having taken so many approaches for its analysis. An abstract published in 2012 reported only one case with this peculiar Cw pattern for which the FlowPRA panel assay and the flow crossmatch were negative. 18 In another report, Zhang et al studied 66 cases with various patterns considered as falsely positive because of a discrepancy between LSA F I G U R E 3 Consequences of acid denaturation of the LABScreen Luminex single antigen (LSA) beads. LSA beads were treated with acid in order to denature HLA proteins, then sera were incubated with the beads. We compared Raw-NC MFI obtained for the Cw1, Cw12 and Cw15 beads in non-treated (N) and treated conditions (Ac). Symbol **** represent statistical significance p ≤ 0.0001 (unpaired t test).
and FlowPRA results, and reported the Cw1/12/15 pattern as the most frequent denatured antigen pattern encountered. 19 The comparison of both LSA assays from One Lambda and Immucor manufacturers is of great importance, as both assays are widely used for routine purposes. In addition, it is crucial that both assays concur on each pattern and eplet that may be encountered. 23 Conversely, discrepancy of positivity may lead to further exploration of this result and raise concerns on the differences between the recombinant HLA proteins produced by the suppliers. In their report, Sullivan et al suspected 11 different patterns with the LABScreen LSA assay to be false positive, among which the Cw1/12/15 pattern occurred for 2 sera among a total of 42. 12 These two samples were negative with the Lifecodes LSA assay, as we found for our sera in our study. As a whole, our findings are in complete agreement with the conclusions raised before us by other authors that this Cw1/12/15 pattern is a false positive one that should not be taken into account.
When compared to the literature reporting on the Cw1/12/15 pattern, our findings rely on a higher number of samples and patients, and are strengthened by a strategy combining additional approaches to the classical two LSA assays that is, the acid denaturation assay and two cellular assays based on two distinct cell sources. Importantly, the cellular assays of our study used serum adsorption and not merely flow cytometry binding as in a classical cross match that only provides information on the ability of the serum to bind to the expressed antigen by comparison with a negative control cell not expressing it. Indeed, the adsorption design allows to define precisely which alleles are recognized (or not) by the antibody of the adsorbed serum by performing LSA assay after the adsorption step. This strategy therefore provides additional important information beyond the mere capacity to interact.
Our results lead to confidently conclude to the absence of true (i.e., with a potential clinical relevance) anti HLA antibody against the Cw1/12/15 pattern. Importantly, more generally, we believe that our strategy could and even should be applied to any suspect LSA reactivity pattern that can be encountered with any manufacturer's LSA commercial assay, in order to definitely conclude regarding the pathological relevance of such profiles and antibodies. The combination of strategies may even facilitate the analysis of complex serum profiles that is, profiles that associate among the positive beads several that do not belong to the suspect pattern, and therefore may be explained by other coexisting antibodies targeting relevant or not relevant eplets and epitopes. However, complex patterns where beads targeted by both the suspect pattern and a concomitant different antibody might remain difficult or impossible to explain, due to the summed origin of the measured MFI, rendering e.g. the interpretation of the denaturation assay quite tricky. The limitation of the acid denaturation assay is that it can only be used for class I, as its relevance has not yet been proven for class II.
We also believe that this strategy should also be applied to situations besides the unexplainable LSA patterns, and especially to the eplets defined in the reference Epregistry list. 3 Indeed, a recent report clearly showed that a majority of the so-called "antibody-verified" eplets that is, those present on the cell surface of normal cells, which define the authentic targets of allogenic Abs with clinical relevance, has not been sufficiently well defined experimentally. 24 A recent review article also warned about the use of Epregistry to define eplet matching for organ allocation. 25 A combination of approaches such as the one we designed for the Cw1/12/15 pattern would largely facilitate the establishment of an accurate registry. Moreover, this is a very useful approach in order to precisely define new eplets, beyond the mere ability to demonstrate the capacity of an antibody to bind a donor antigen or allele.