ABSTRACT. To investigate the impact of lncRNA expression variability during acute renal allograft rejection a total of 82 ESRD patients who underwent renal transplantation were enrolled. Patients with acute allograft rejection and those with a well-functioning graft were evaluated at 3 different time points (pre transplant, at the time of rejection, and post anti-rejection therapy). RNA-Seq was performed on the samples collected at the 3 stated time points. Data obtained was utilized for identification of differentially expressed (DE) RNAs using Wald t test and their related pathways. Gene ontology (GO) analysis was applied to analyse the main function of DE genes according to the Database for Annotation, Visualization and Integrated Discovery (DAVID). Statistical significance of enrichment was determined using a 2-side Fisher exact test. GO categories with a false discovery rate (FDR) <0.05 are considered. The differential expression of RNAs were evaluated by segregating the cohort in two groups: i) Group I-Well-functioning Vs. Rejection and ii) Group II- Rejection Vs. Post anti-rejection therapy. In the group I, 2218 RNAs were observed to be up-regulated and 887 down-regulated whereas, in the group II, 360 RNAs were up-regulated and 147 down-regulated (based on fold change and significant p value). 79 RNAs were down-regulated during rejection but up-regulated post-therapy whereas, 45 were up-regulated during rejection but down-regulated post-therapy. Majority of these GO terms were associated with leukocyte activation, degranulation and regulation of cytokine secretion. Kyoto Encyclopedia of Genes and Genomes pathway (KEGG) analysis revealed involvement of pathways like cytokine-cytokine receptor interaction and TNF signalling. Select RNAs are identified and presently under qPCR validation. Few novel lncRNAs identified are yet to be annotated. Our preliminary analysis indicates the role of lncRNA in immune pathways with possible roles in acute renal allograft rejection.

ABSTRACT. HLA mismatches between donor and recipient may facilitate Natural Killer (NK) cell alloreactivity upon kidney transplantation though the lack of interaction between donor HLA molecules and inhibitory killer immunoglobulin-like receptors on the NK cells of the recipient, a condition called “missing-self”. In this study, we investigated the effect of NK alloreactivity on kidney transplant failure . From the Dutch PROCARE consortium database, containing extended patient and donor information from kidney transplantations with at least 10 years follow-up, we selected 4024 patients without pretransplant donor-specific antibodies. We determined NK alloreactivity based on the Bw4, Bw6 and C1, C2 typing of patients and donors. The patients were divided in a non NK-alloreactive group (group 0, N=2693), an alloreactive group with 1 missing-self HLA ligand (group 1, N=1179) and 2 missing-self HLA ligand (group 2, N=152). We checked for potential confounders such as age, sex and type of kidney donor but these did not influence the results we found. NK cell alloreactivity protected against graft failure during the first year post-transplant, and the protective effect was stronger for group 2 (HR=0.3, p=0.002) than for group 1 (HR=0.7, p=0.001). However, after 5 years, this protection disappeared. The effect was most pronounced for group 2, where HRs (group 2 vs. 0) increased from 0.3 (T=1y) to 0.6 (T=5y) or 1.2 (T=10y). After 15 years, the risk for graft failure was even higher in group 2 than in the non-alloreactive group, though this was not significant (HR=1.5 , p=0.3, group 2 vs. 0). In the group with 1 missing-self HLA ligand, the risk for graft failure after >5 years was almost comparable to non-alloreactive group (HR (T=5y) = 1.0, p=0.9; HR (T=10y)= 1.1, p=0.5; HR (T=15y) = 0.8 p=0.3). In conclusion, we found that NK alloreactivity temporarily protects against graft failure, but after 5 years, this effect disappears and the hazard ratio of failure slowly increases.

ABSTRACT. The HLA Epitope Registry has documented all theoretically defined HLA eplets along with their antibody verification status, and has been the foundation for many clinical studies investigating eplet mismatch loads in clinical transplantation. Despite a uniform use of antibody-verified (AbVer) eplets in these studies, verification methods for eplets in the Registry range from polyclonal sera from multi- and uniparous women to murine and human monoclonal antibodies (mAbs), and eluted antibodies from sera of immunized individuals. The classification of antibody verification based on different methods for validation is problematic, since not all approaches represent the same level of evidence. We introduced a classification system to evaluate the level of evidence for antibody verification and demonstrate that for a considerable number of eplets, the AbVer status is solely based on polyclonal serum reactivity of multiparous women, or on reactivity of murine mAbs. The prior does not exclude multiple antibody specificities contributing to the reaction pattern, and the latter does not prove reactivity in the human situation. Furthermore, we noted that a substantial proportion of patient sera analyses and human mAb data presented in the HLA Epitope Registry Database had never been published in a peer-reviewed journal. Therefore, we tested several unpublished human HLA-specific mAbs by luminex single antigen bead assay to analyze their HLA reactivity for eplet antibody verification. Although the majority of analyzed mAbs indeed verified their assigned eplets, this was not the case for a number of eplets. Based on extensive literature review and re-testing of human mAbs, we have compiled a new list of AbVer eplets. This comprehensive overview is instrumental for future investigations towards eplet immunogenicity and clinical studies considering AbVer eplet mismatch in transplantation and warrants further standardization of antibody verification using high quality data.

ABSTRACT. Background Transplantation (tx) is the treatment of choice for chronic kidney disease, but also disrupts immune function. We use immunophenotyping to characterize immune cell subsets of tx patients, to understand the interaction between allograft-recognition and immune cell reconstitution after induction. Methods Immunophenotyping panels for CD4 T-, CD8 T-, B-, NK-cells, and monocytes were run on freshly isolated PBMCs of 35 kidney tx patients at day 0 and at five timepoints during the early post-tx phase. Data were analyzed using an R/Bioconductor pipeline utilizing the neural network algorithm FlowSOM and visualized using the dimensionality reduction tool UMAP. Results We observed striking longitudinal changes in patients with basiliximab (BXM, n=27) induction and no rejection episodes. Expression of the BXM target CD25 was reduced on CD4 central- (CM) and effector memory (EM) T cell from day 7 to 60. We also detected a downregulation of HLA-DR on monocytes immediately post-tx and its recovery after three months. In the anti-thymocyte globulin (n=8) induced cohort the frequency of a NK cell subset (CD16+ CD57+ CD8-) was reduced post-tx until the end of follow-up. The T cell rejectors (n=4) receiving BXM exhibited increased counts of CD4 and CD8 CM as well as CD8 EM T cells at the time of the episode (day 5-7) and for several weeks thereafter. Strikingly, while most quiescent patients showed a de- and subsequent increase of CD226 expression on NK cells, the rejectors had consistently high levels, including pre-tx. These changes were also visualized in dimensionality-reduced UMAP plots for groups of - or individual patients, allowing for a more intuitive tracking of subset – and marker expression changes over time and during rejection. Conclusion Our study shows how immunophenotyping paired with machine learning tools can helps us reveal the impact of tx and different induction therapies on immune cell subsets and will help to identify new biomarkers for rejection.

ABSTRACT. Patients transplanted across donor-specific HLA antibodies (DSA) experience various clinical outcomes. Several research efforts have been initiated to better identify features that predict the hazardous potential of circulating DSA. A conceivable informative value in this context offers DSA affinity. To date, several biophysical technologies allow the assessment of antibody binding properties, but require prior knowledge of antibody concentration. Therefore, the objective of the current study was to develop an approach that includes affinity measures as well as DSA concentration determinations for patient sample evaluation. In a preliminary validation step, we investigated several platforms including bio-layer interferometry (BLI), surface plasmon resonance (SPR), flow induced dispersion analysis (FIDA), and a Luminex Single Antigen Bead (SAB) Titration Assay, testing representative sets of HLA-specific monoclonal antibodies (mAbs) by incubating them with various soluble HLA molecules. Binding affinities were assessed either based on real time antibody on- (ka) and off-rate (kd) determinations (BLI, SPR) or by end-point measurements using the steady state equilibrium dissociation constant (steady state KD) (FIDA, SAB). All four technologies delivered comparable results, with affinities ranging between 0.028 – 7.23 nM. Focusing on the FIDA approach including affinity and DSA concentration, a correlation with SAB-derived MFI signals was observed. Values from the twenty investigated pre-transplant patient samples, all with negative CDC-crossmatch and SAB signals ranging between 571 and 14899 MFI, varied 110-fold in concentration (1.68 – 184 ug/mL) and their apparent affinities were between 0.199 – 24.7 nM (124-fold difference). Overall, the approach presented has substantial diagnostic potential and holds great promise to provide a novel clinically relevant assessment of transplant patients.

ABSTRACT. In this study we aimed to investigate the role of 82 different non-HLA antibodies in the occurrence of histology of ABMR after kidney transplantation. We included all patients who underwent kidney transplantation between 2004-2013 in a single center and had biobanked serum. Pre- and post-transplant sera (n=2870) were retrospectively tested for the presence of 82 different non-HLA antibodies. 874 patients had available pretransplant sera and were included in this analysis. In total, 204 (23.3%) patients developed histology of ABMR after kidney transplantation. In 79 patients (38.7%) the histology of ABMR was explained by pretransplant or de novo HLA-DSA. The multivariable Cox analysis revealed that only the broadly non-HLA sensitized (number of positive non-HLA antibodies) patients and those with the highest total strength of the non-HLA antibodies (total ratios of the positive non-HLA antibodies) were independently associated with increased rates of histology of ABMR after transplantation. Additionally, independent associations were found for antibodies against TUBB, Collagen III, VCL and STAT6. This study shows that patients highly and broadly sensitized against non-HLA targets are associated with an increased risk of ABMR histology after kidney transplantations in the absence of HLA-DSA.

ABSTRACT. T-helper cells play an important role in alloimmune reactions following transplantation by stimulating humoral and cellular responses, which might lead to failure of the allograft. Previously, it has been shown that graft failure is more frequent in patients with non-donor-specific HLA antibodies as compared to patients without HLA antibodies. Possibly, this could be explained by the presence of donor-HLA-reactive memory T-helper cells from a previous immunizing event, which can potentially be reactivated by exposure to donor HLA mismatches that share T-cell epitopes with the initial immunizing HLA. In this study, the effect of pre-transplant donor HLA-reactive memory T-helper cells on kidney allograft failure was investigated using the predicted overlap of T-cell epitopes between immunizing and donor HLA. For 190 kidney transplant recipients with non-donor specific HLA antibodies, the immunizing HLA alleles were identified by means of a Luminex Single Antigen Bead assay. The potential T-cell epitopes originating from these immunizing HLA alleles as well as from the donor HLA were determined using the Predicted Indirectly ReCognizable HLA Epitopes (PIRCHE-II) algorithm, and the overlapping PIRCHE-II load was calculated for each recipient. In a multivariable Cox proportional hazards model, the ln-transformed PIRCHE-II overlap score was significantly associated with the 10-year risk of kidney graft failure (HR 1.48, p=0.015). In addition, recipients with a low number of shared T-cell epitopes had a significantly higher cumulative incidence of graft failure as compared to recipients with a higher number of shared T-cell epitopes (p=0.014). Our findings suggest that the PIRCHE-II overlap score might be a strong indicator for the risk of allograft failure and that the presence of pre-transplant donor-HLA reactive memory T-helper cells might play an important role in the development of graft failure.

ABSTRACT. BACKGROUND De-novo donor-specific antibodies(dnDSA) are key in the diagnosis of antibody-mediated rejection(ABMR) and are related to premature graft loss(GL). This study was designed to investigate the impact of dnDSA mean fluorescence intensity (MFI) value evolution as detected by Luminex® on kidney allograft survival.

METHODS A retrospective analysis including 400 kidney transplant recipients with dnDSA until 31/05/2021 was performed. Only patients with a functioning kidney and dnDSA against the last graft were included. dnDSA were assessed annually after transplantation and determined by Luminex®(n=691).

RESULTS ABMR occurred in 24.8% of patients and GL in 33.3% during 9.7 years (IQR 7.2-12.2) of follow-up. Patients with a higher count of dnDSA and those with both class I and II had inferior graft survival at 5 years post-dnDSA(log rank p=0.008 and <0.001). There was a greater proportion of MFI≥3000 at first occurrence in class II(p<0.001). MFI reduction≥50% was frequent during follow-up(36.9%), and a greater MFI negativity(MFI<500) was observed in class I(p<0.001). 5-year allograft survival post-dnDSA was 74.0% when reduction ≥50% MFI, 62.4% when fluctuating MFI and 52.8% when doubling MFI(log rank p<0.001).

CONCLUSIONS MFI evolution is associated with graft failure, being fluctuating and doubling MFI significantly related to worse allograft survival at 5 years post-dnDSA.

ABSTRACT. Inhibitory killer cell immunoglobulin-like receptors (iKIRs) are a family of inhibitory receptors that are expressed by natural killer cells and late-stage differentiated T cells. There is accumulating evidence that iKIRs regulate T cell-mediated immunity. Recently, we reported that T cell-mediated control was enhanced by iKIRs in chronic viral infections. We hypothesize that in the context of autoimmunity, where an enhanced T cell response might be considered detrimental, iKIRs would have an opposite effect. We choose type 1 diabetes (T1D) as a paradigmatic example to investigate the effect of functional iKIR genes in autoimmunity. T cells have a major role in T1D etiopathogenesis as indicated by the strong HLA associations with disease. To investigate if iKIRs impact these autoreactive T cell responses we ask whether HLA associations in T1D are modified by iKIR genes via immunogenetic analysis of a case-control T1D dataset (N= 11,961). An independent dataset consisting of family trios is used for validation. We use mathematical modeling to investigate possible mechanisms of the iKIR effect. We found that functional iKIR genes, i.e. iKIR genes in the presence of genes encoding their ligands, have a consistent and significant effect on protective HLA class II genetic associations. No effect was found in detrimental HLA class II or class I genetic associations. Single cell RNAseq analysis of PBMCs from 4 T1D cases revealed very low numbers of CD4+ T cells expressing iKIRs. Finally, using a mathematical model of the interaction between T cells and beta cells, we formulate hypotheses about the iKIR gene effect during autoreactive T cell responses in T1D. We present different scenarios that can recapitulate our immunogenetic findings. We conclude that iKIRs significantly decrease HLA class II protective associations and suggest that iKIRs regulate CD4+ T cell responses in the context of T1D, probably through an NK cell modulatory effect on CD4+ T cells.

ABSTRACT. The KIR region reorganized rapidly during primate evolution and displays species-specific haplotype architecture. Haplotype diversity is reflected by differential expansion of gene lineages, and of centromeric and telomeric segments. In humans, two haplotype groups, A and B, are distinguished based on differential KIR gene content. Group A contains a fixed set of seven genes, with KIR2DS4 representing the only activating member. More variable gene content, with multiple activating entities, is characteristic for group B. Both haplotype groups are encountered in all human populations, however, with different relative frequencies. In other primate species, similar haplotype groups are not yet defined, which might indicate a unique feature for humans. However, this observation could also relate to the lack of KIR haplotype studies in large cohorts of non-human primates. In this study, we roughly quadrupled the number of rhesus macaques analyzed for their KIR transcriptome and distinguished functional haplotype profiles similar to the human situation. Several genes, like Mamu-KIR3DS04, seem to be restricted to one functional profile, like KIR2DS4 on human group A haplotypes. Thus, balancing selection on more inhibitory (A) and more activating (B) profiles seems to be shared by different primate species and might reflect an important functional relevance.

ABSTRACT. The prediction of NK cells alloreactivity toward leukemic cells provides promising perspectives, although the underlying biological processes remain unclear. Many prediction models based on KIR and MHC genotyping have been designed and used across studies, bringing ambiguity in its utility and application for clinicians.

We applied 25 models divided into five categories of alloreactivity prediction (ligand-ligand, receptor-ligand, educational, KIR haplotype-based and KIR matching) as well as three models based on KIR polymorphism on 78 NGS-typed donor/recipient pairs undergoing allogeneic haematopoietic stem cell transplantation in genoidentical (n=43) or haploidentical (n=35) matchings.

We highlighted that the different scoring strategies impact the assessment of alloreactivity within the pairs; e.g. two broadly used scoring strategies – educational models and missing-ligand models – show clear opposite predictions. Some scoring strategies are better adapted to genoidentical or haploidentical cohorts, whereas others are robust across both. Regarding clinical-biological correlations, we identified that (i) each scoring strategy is differentially associated with the different outcomes (ii) the different scoring strategies predict particular outcomes with different efficacy (iii) the donor/recipient compatibility impacts the pertinence of the scoring strategy. The number of centromeric B-motifs was the one model associated with the strongest correlation with the incidence of acute graft-versus-host disease in our set of patients. The effect was apparent in both the genoidentical and the haploidentical cohorts, suggesting that KIR-based alloreactivities, not MHC mismatches, are responsible for it.

Our work contributes to unravelling the KIR-based alloreactivity prediction of NK cells in transplantation. This study will help resolve the current literature discrepancies in this field and develop our understanding of NK alloreactivity for its use in medical practice.

ABSTRACT. Highly polymorphic killer cell immunoglobulin like receptors (KIR) modulate effector functions of cytotoxic immune cells and have important roles in human health and tissue transplantation. Study of KIR has been hampered by complex structural polymorphism and homology among genes. Countering these challenges are our targeted short-read next generation sequencing (NGS) method, and Pushing Immunogenetics to the Next Generation (PING), a bioinformatic pipeline for the analysis of KIR. This combined approach has been used to genotype KIR in over 30,000 individuals across worldwide populations. Although 1,535 KIR alleles have been named to date, discovery and confirmation of further alleles requires extensive manual verification, especially at a high-throughput scale. We developed an automated pipeline that mines large datasets output from PING to determine whether unresolved genotypes are a product of low-quality sequence data, misaligned reads from another KIR gene, or new alleles. The pipeline is called Prioritizing Unresolved New Genotypes (PUNG) and has three main modules. First, it performs quality control of new variants. To determine if the novel variant is specific to the gene being analyzed, the second step aligns reads to a local database containing all KIR variants. Finally, PUNG verifies whether the resulting sequences have already been deposited in IPD-KIR. We applied PUNG to a dataset of 2,000 European individuals who had been sequenced using our targeted NGS method and analyzed using PING. PUNG was able to completely resolve 101 of the 284 unresolved genotypes in this dataset, a 35.6% reduction in sequences to be analyzed. We further investigated KIR3DL2 using the Sanger method and confirmed 81.1% of PUNG suggestions. This work allowed us to validate 21 newly discovered KIR3DL2 alleles, with minimal bioinformatic efforts. Thus, we suggest that PING be used in conjunction with PUNG to enable state-of-the-art study of KIR polymorphism.

ABSTRACT. Natural killer (NK) cell activity is modulated by an array of highly variable receptors encoded by more than 90 genes within the LRC (leukocyte receptor complex, 19q13.42) and NKC (natural killer complex, 12p13.1). The LRC encodes the immunoglobulin superfamily receptors, which include the highly polymorphic killer cell immunoglobulin-like receptors (KIR) and leukocyte immunoglobulin-like receptors (LILR). In contrast, NKC encodes the C-type lectin-like receptors. The complex structural variation in combination with the sequence similarity observed for several homologous genes within the LRC and NKC has imposed technical barriers that have systematically precluded analysis of these complexes at high resolution. Our previous analysis has shown that specific genes share up to 98% of sequencing identity when considering 150-base-pair fragments, which is a typical size for short-read next-generation sequencing (NGS) technology. Here, we developed and validated a capture-based short-read NGS method specifically to sequence these complex regions. We designed 10,635 120-nt double-stranded probes targeting the 1.5 Mb LRC from upstream PRKGC to GP6 and 1.6 Mb NKC from KLRG1 to LY49L. Libraries were prepared with 100 ng DNA using the KAPA HyperPlus kit (Roche), captured with hybridization kits from Twist Bioscience, and sequenced using the paired-end 2x150 pb sequencing protocol at the Illumina platform. Our results showed coverage of 98.8% of the targeted regions, with 113 average fold enrichment and duplication rate of 3%. We ran the raw fastq files of 26 samples that were previously sequenced for KIR with a validated method and observed a concordance of 100% for all allele calls (260/260). Our results show that our probes successfully captured these unusually complex genomic regions and that we precisely determined KIR genotypes from our data, demonstrating that this method can assess high-resolution genotypes and copy number variation of the entire NK cell receptome.

ABSTRACT. Killer-cell immunoglobulin-like receptors (KIR) serve a critical role in regulating natural killer cell function. They are encoded by highly polymorphic genes within a complex genomic region that has proven difficult to interrogate owing to structural variation and extensive sequence homology. Previously, we introduced the PING bioinformatic pipeline for high-throughput KIR sequence analysis from targeted short-read sequencing data that provided accurate KIR copy number and genotype determination. Here, we expand that pipeline to process whole genome sequence (WGS) data, greatly increasing the utility of the workflow. To demonstrate the pipeline effectiveness, we performed an initial analysis on 215 30x WGS samples of European ancestry from the 1000 Genomes Project. The results of this analysis showed PING determined copy numbers were in line with expected values for that population and PING gene content determination is concordant with an independent published KIR gene content determination workflow. Average ambiguity of genotypes determined by PING were zero for KIR2DL3, KIR2DL4, KIR2DP1, KIR2DS4, KIR3DP1 and KIR3DS1. All other KIR genes were at or below an average genotype ambiguity of two except for KIR2DS5 at five. Unresolved genotype frequencies, a proxy for measuring genotype accuracy with low unresolved frequencies being ideal, were below 6% for KIR2DL3, KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS4, KIR2DS5, KIR3DL2, KIR3DP1, and KIR3DS1, and below 10% for KIR2DL2, KIR2DL4, KIR2DL5, and KIR3DL3. Together, these results demonstrate that the PING bioinformatic pipeline provides accurate copy number and genotype results for WGS data.

ABSTRACT. Proper invasion of the semi-allogeneic fetal trophoblast cells in the maternal-fetal interface (decidua) is essential for healthy placentation and pregnancy. Determining the characteristics and interplay of the maternal immune cells with fetal trophoblast cells at the maternal-fetal interface is key to understand their role in healthy and complicated pregnancy. To comprehend the composition and spatial orientation of maternal immune cells and fetal trophoblasts in situ, we applied imaging mass cytometry (IMC) on the decidua of the three trimesters of healthy pregnancy. We visualized the composition and interaction profiles of maternal immune cell populations, stromal cells, and fetal trophoblasts using a 41-marker IMC panel. We analyzed 148,177 segmented cells of first (n=3), second (n=5), and third trimester (n=5) placentas using phenotyping tools and interaction analysis programs including ImaCytE. We found a higher frequency of myeloid cells within the CD45+ population in the decidua using our in-situ approach compared to approaches using cell suspensions following enzymatic cell isolation from decidua (1st trim.: median 35.8% vs 17.6%, p=0.004; 2nd trim.: 52.5% vs 26.8%, p=0.004; Term: 60% vs 9.4%, p=0.001). Moreover, myeloid cells were the most pronounced cell type in the microenvironment of other decidual cells including fetal trophoblasts. In first trimester but not at term, HLA-DR-negative myeloid cells interact with CD4+ T cells, decidual NK cells, and trophoblasts. At term, T cell interactions with other immune cells became more prominent. By using IMC we confirmed previously described data but most notably further extended the comprehension of the decidual landscape. Our results highlight the dynamic role of myeloid cells at the human maternal-fetal interface throughout gestation. Furthermore, the results can be used to set up substantiated in vitro culture experiments to identify the consequences of cell interactions as found in the current study.

ABSTRACT. Introduction: Oocyte donation (OD) is a method for achieving pregnancy in women with reproductive disorders, and also is a risk factor for pre-eclampsia. OD pregnancies are related to a higher extent of fetal-maternal HLA mismatches compared to naturally-conceived pregnancies. Due to the higher HLA dissimilarity the immune response at the fetal-maternal interface is probably divergent and affects clinical outcome. Placental histologic assessment may provide a clue on the pathophysiologic mechanisms. Methods: In this retrospective case-control study placentas were collected after delivery. A total of 21 pathology lesions were scored in four groups: OD with pre-eclampsia (OD/PE; n=16), uncomplicated OD (OD/healthy; n=37), naturally conceived with pre-eclampsia (NC/PE; n=45), and uncomplicated non-OD in vitro fertilization pregnancies (IVF/healthy; n=17). All combinations were genotyped for HLA-A, -B, -C, -DR, and -DQ. Analyses were performed on individual lesions and combined categories. The OD/healthy group was separated into semi-allogeneic (≤5 fetal-maternal HLA mismatches) and fully allogeneic (>5 mismatches) subgroups to study the association between morphology and HLA dissimilarity. Results: Inflammatory pathology lesions were more frequent in OD/healthy group than in other groups (p<0.05). As for HLA class I, a higher total inflammatory score was found in the fully-allogeneic OD/healthy group than in the IVF/healthy group (p<0.05). As for HLA class II, the fully-allogeneic OD/healthy group showed a higher percentage of chronic deciduitis with plasma cells than the IVF/healthy group (p<0.05). These differences were not found in the semi-allogeneic OD/healthy group. Conclusion: Placental inflammatory lesions are specifically detected in healthy OD pregnancies, and their presence is associated with a higher extent of fetal-maternal HLA mismatching. Further studies will characterize the nature of this inflammation.

ABSTRACT. Pancreatic cancer (PC) is associated with poor prognosis and immunotherapy may provide new treatment perspectives. Predictive preclinical models can help to identify successful immunotherapy combinations. Tumor organoids are in vitro 3D organ-like structures that retain many pathophysiological characteristics of the in vivo tumor. We aimed to establish a co-culture system to study the interaction between NK cells and patient-derived PC-organoids in which strategies to boost NK cells can be tested. We developed a cytotoxicity assay with PC-organoids and healthy donor-derived IL-2-activated NK cells. We observed that the assay should be performed in medium with 10% Geltrex (Gx) seeded onto a pre-solidified Gx scaffold in order to prevent organoid 2D outgrowth. Moreover, NK cells and organoids should be seeded simultaneously, as NK cells could not migrate towards the organoids if the 10% Gx in the culture medium was pre-solidified. For most donors, NK cells effectively target PC-organoids, showing an increase in organoid death at higher effector-to-target (E:T) ratios. After 24-hours of co-culture, the apoptotic signal within organoids plateaued, and organoid 2D growth and NK cell death increased. Some organoids were relatively resistant to NK-induced apoptosis, indicating heterogeneity between donors and among organoid lines. To determine if NK-anti-tumor responses could be enhanced in this 3D setting, antibody-dependent cellular cytotoxicity (ADCC)-inducing antibodies were added. Trastuzumab (anti-Her2) and avelumab (anti-PD-L1) increased NK cell-induced organoid cell death, especially for the 5:1 E:T ratio (p=0.001 and p=0.003). Importantly, in organoid-NK combinations with no or low kill, ADCC-inducing antibodies also triggered NK cell mediated apoptosis. Our results thus support the use of organoids as a relevant model for in vitro studies of tumor response to NK cell based immunotherapy and the use of ADCC-inducing antibodies to enhance NK cell effector function.

ABSTRACT. Human Cytomegalovirus (HCMV) reactivation causes frequent complications in immunocompromised patients after solid organ or hematopoietic cell transplantation (HCT). T-cell immunity to HCMV has been extensively studied regarding the role of HLA class I and HLA-DR, in particular for the viral proteins pp65, gB, and IE1. Since HLA-DP is an important determinant in HCT, we sought to characterize HLA-DP-restricted HCMV immunity and its dependency on the DEAV/GGPM polymorphism at amino acid positions 84-87, which has a strong impact on the immunopeptidome. Using a selected set of 9 HLA-DPB1 allotypes (5 DEAV and 4 GGPM) validated on the NetMHCIIpan4.0 platform, we screened the complex HCMV translatome, arose from the 235kb genome, to predict strong binding peptides (pSBP; rank <2%). We found better predicted HCMV peptide binding capacities for DEAV compared to GGPM allotypes, both in numbers (1865 vs 1378) and binding scores (0.58±0.17 vs 0.31±0.18, p<0.001). Only 19, 27, and 7 pSBP to at least one of the studied HLA-DP allotypes were derived from pp65, gB, and IE1, respectively, while the remaining 2977 pSBP were distributed across the entire HCMV genome with no apparent hotspots. In line with this, we found significantly higher HLA-DP-restricted in vitro CD4+ T-cell responses from HCMV sero-positive individuals (N=10) in CD137 up-regulation assays against HCMV lysates compared to pp65 peptide pools (11.9±6.0% vs 2.2±2.0%, p<0.0001). In these HLA-DP heterozygous individuals, CD4+ T-cell responses were significantly higher in the context of DPB1*03:01 (DEAV) than of DPB1*04:01 (GGPM) for HCMV lysates (16.8±3.8% vs 7.1±3.4%, p=0.016) but not for pp65 (3±2.4% vs 1.4±1.1%, p=0.48). Our data support the notion that HLA-DP-restricted immunity to HCMV is directed against a broad variety of viral proteins and is modulated by the DEAV/GGPM polymorphism. These findings have potential implications for clinical diagnostics and for HCMV-related risk associations in transplantation.

ABSTRACT. In hematopoietic cell transplants (HCT) from unrelated donors, ~40% of the permissive HLA-DPB1 mismatches involve exclusively 3 high-frequency “core” alleles (i.e. DPB1*02:01, 04:01, or 04:02), which differ at 3 amino acid positions (i.e. 55, 56, and 69). Due to their structural and functional closeness, these alleles have similar peptide-binding motifs, potentially allowing for cross-reactivity by T-cell receptors (TCR) recognizing minor histocompatibility antigens (mHAg). We investigated this hypothesis in an in vitro system using HeLa cells transduced with HLA-DP, HLA-DM, Ii, and CD80 to stimulate isolated CD4+ T cells from DPB1*04:01-self individuals (N=12). Stimulators expressed matched DPB1*04:01 (mHAg responses), permissive DPB1*04:02 or DPB1*02:01, or non-permissive DPB1*09:01. Responses were quantified by CD137 upregulation assays, against the original stimulator and all other stimulators as control of cross-reactivity. mHAg responses and permissive alloresponses to DPB1*04:02 were similar in magnitude and cross-reactivity patterns, the latter mainly mediated by the naïve CD4+ T-cell subset. In contrast, permissive alloresponses to DPB1*02:01 and non-permissive alloresponses diverged in magnitude and showed less cross-reactivity. Next-generation TCR sequencing revealed similar diversity in mHAg responses and permissive or non-permissive alloresponses (mean Inverse Simpson Index TRA 8.6-31.4; TRB 8.7-26.1). However, alloreactive TCR repertoires elicited against mHAg showed higher overlaps (>40%) and similarity (mean Jaccard index TRA 0.1396; TRB 0.1035) to those responding against permissive DPB1*04:02 than to those against permissive DPB1*02:01 and non-permissive DPB1*09:01. Our results support the hypothesis of cross-reacting mHAg and permissive alloreactivity for DPB1*04:01 and DPB1*04:02, and highlight the relevance of polymorphism at position 69 in peptide selection for DPB1*02:01, suggesting different mechanisms for permissiveness among “core” alleles.

ABSTRACT. Alloreactive donor T cells recognizing mismatched HLA-DPB1 can mediate both toxic graft-versus-host disease and beneficial graft-versus-leukemia effects in hematopoietic cell transplantation depending on whether the alloantigens are expressed on non-hematopoietic or hematopoietic patient tissues (e.g. leukemia). It is therefore desirable to identify alloreactive T-cell receptors (TCR) that recognize HLA-DPB1 selectively on hematopoietic tissues. To this end, we used a stable, single HLA-DP-expressing hematopoietic B-lymphoblastoid cell line after CRISPR/Cas9-mediated HLA class II beta-chain knockout and subsequent reconstitution with DPB1*03:01 (BLCL-DP3) to stimulate CD4+ T cells from a DPB1*03:01-negative healthy donor. In parallel, we used single DPB1*03:01-expressing HeLa cells transduced with the HLA class II processing machinery and CD80 as non-hematopoietic stimulators (HeLa-DP3). In CD137 upregulation assays, polyclonal CD4+ T cells raised against BLCL-DP3 showed stronger recognition of BLCL-DP3 (43.1%) or dendritic cells from a DPB1*03:01-positive third party individual (19.4%), than of HeLa-DP3 (7.06%). In contrast, recognition of hematopoietic and non-hematopoietic target cells was similar by polyclonal CD4+ T cells stimulated with HeLa-DP3 (38.9% and 54.9%). Two TRBV-19-bearing T cell clones carrying TRBJ2-2 (C1) and TRBJ1-6 (C2) and different TRA rearrangements were obtained from the BLCL-DP3 culture by limiting dilution. Both clones responded to BLCL-DP3 in IFN-g ELISpot (mean spot-forming cell (SFC): C1 360.5±2.1 vs C2 210.5±85.6), but differed for recognition of HeLa-DP3 (mean SFC: C1 2.5±0.7 vs C2 166±62.2). C1 was 20x more abundant in the polyclonal T cells raised against BLCL-DP3 than C2 (0.12% vs 0.006%). This and other potentially hematopoietic-tissue specific alloreactive TCR under characterization by our approach might be of translational use in cellular immunotherapy of leukemia.

ABSTRACT. Natural killer (NK) cells are lymphocytes that mediate cytotoxic and cytokine producing effector functions, which are modulated via interactions of HLA class I molecules and their ligands of the KIR family. The alloreactivity of donor-derived NK cells could be utilized as immunotherapy for leukemic patients, whereas these cells could initiate graft rejection in solid organ transplantation. Accurate and high resolution characterization of the polymorphic HLA and KIR genes is essential for clinical applications.

We developed a rapid and cost-efficient method to characterize HLA and KIR haplotypes. An amplification-free and Cas9-mediated enrichment of the target regions, combined with adaptive sampling of whole genome fragments using Oxford Nanopore sequencing, allowed the construction of haplotypes at an allele level resolution. The captured fragments, ranging from 10 to 22 kb, complemented with sampled fragments that reached over 1 Mb, resolved even complex haplotypes.

This method to characterize the HLA and KIR regions without PCR-induced errors is a feasible and accessible strategy, using an relatively cheap sequencing platform. Compared to the currently used methods, which mainly rely on determining the presence or absence of genes, our approach offers opportunities for a substantial improvement in the direction of immunotherapy and transplantation biology.

ABSTRACT. Xenotransplantation of porcine organs offers a promising alternative to human transplants. Generation of transgenic pigs with knockout of specific antigens and expression of human protective genes significantly improved histocompatibility of xenografts, but they remain a target for rejection. So far, potential human immune responses towards xenografts can only be tested in vitro using specific cell types or after xenotransplantation in non-human primate (NHP) models. However, data obtained with in vitro assays is insufficient, and transplantation of NHPs raises several ethical, logistic and economic concerns. Normothermic ex vivo kidney perfusion (EVKP) emerged as a promising technology to assess organ quality and function ex vivo. Here, we have evaluated the feasibility to evaluate xenogeneic T-cell responses towards pig kidneys during EVKP. A protocol for the normothermic perfusion of pig kidneys with human T-cells was established. Briefly, kidneys were perfused during 4h without or with 4x107 human T-cells. Transmigration of the T-cells into the renal tissue and their localization was evaluated by immunohistochemistry. Human T-cell responses were analyzed by determining the transcript and protein levels of human cytokines such as TNF-α and IFN-γ as well as Granzyme B and Perforin. Pig kidneys were perfused during 4h at 37°C and a pressure between 50-70mmHg. T-cell counts in the perfusate decreased in mean by 50% of the initial T-cell numbers and were further reduced to 15% after 4h perfusion. Human T-cells showed to specifically respond to the porcine renal tissue as demonstrated by an upregulation of TNF-α and IFN-γ by up to 174- and 120-fold, respectively. Human Granzyme B expression was increased by 19-fold and Perforin by 16-fold. This study shows that it is possible to assess human T-cell responses during EVKP, and may help to predict risk of rejection of new genetically engineered xenografts as well as reduce the otherwise required NHP trials.

ABSTRACT. Recently, immunotherapy has gained relevance in Oncology due to its great therapeutic potential. Nevertheless, tumor cells often develop several mechanisms of immune escape which make them resistant to it, such as the loss of HLA-I expression. When HLA-I expression is altered, tumor cells cannot be recognized by T lymphocytes and the efficiency of immunotherapy is reduced. Therefore, the analysis of tumor HLA-I expression is crucial to predict the response of each patient to immunotherapy and effectively personalize cancer treatment. We proposed the study of tumor HLA-I using tumor-derived-exosomes (TDEs) and cell-free DNA (cfDNA) derived from patients as a novel non-invasive method. We purified exosomes and exosomal DNA from different cancer cell lines. As a model of cancer progression, we also isolated TDEs from three cell lines derived from a melanoma patient and stablished from one stage III and two consecutive stage IV metastatic lymphatic nodes. HLA-I alterations were studied by flow cytometry, HLA-I genotyping, B2M sequencing and the analysis of loss of heterozygosity (LOH) in chromosome 6 and 15. Exosomal HLA-I genotype and expression profiles, B2M mutations and LOH patterns exactly coincided with their cellular counterparts. Thus, TDEs may be employ to characterize tumor HLA-I phenotypes and identify alterations during the progression of the disease. Then, we purified exosomes and cfDNA from two serum samples which were representative of the first and the last metastatic melanoma lesions and we observed that these TDEs showed the same HLA-I phenotypes and LOH patterns as TDEs derived from the corresponding cell lines. In conclusion, although further studies in a larger cohort of patients are required, preliminary results strongly support the potential of liquid biopsy for the analysis of tumor HLA-I alterations, which may enhance the selection of cancer treatment and bring us one step closer to personalized medicine.

ABSTRACT. HLA-mediated drug hypersensitivities represent a great challenge in the human health system. The variability of HLA precludes the extensive analysis of HLA/drug interactions and their impact on patient specific immune responses. To enable side effect predictions for uncharacterized drugs and to prevent delay in the licensing process one widely used action is the approach to analyze drug-protein interactions. Drug targeted proteins are clearly disturbed or even enabled in their protein-protein-interaction networks. The ability of a drug to alter peptide presentation by an HLA molecule could be impressively demonstrated using the example of HLA-B*57:01 and Abacavir (ABC). ABC occupies the PBR of B*57:01, resulting in the recruitment of a different peptide subset. Yet, not all HLA-B*57:01+ patients are affected by severe ABC hypersensitivity, implying the involvement of further patient-specific factors. Following the classification of B*57:01 carriers in ABC sensitive or ABC tolerant, TCR Vβ regulation and proteomic content has been analyzed. Upregulation of TCR Vβ24 following drug exposure could be clearly assigned to B*57:01+/ABC responders. The elaborated proteomic MS based analysis enabled the detection of differentially abundant proteins in B*57:01+/ABC responder vs. non-responder. Protein abundance significantly discriminates responder and non-responder pre-drug treatment; this facilitates the reasonable potentiality to identify a public biomarker. Network analysis visualized the contribution of proteins involved in antigen presentation and peptide processing. CD247 could be determined as one of the main upstream regulators in responders post-drug treatment. CD247 plays a crucial role in adaptive immune responses and TCR:HLA engagement and explains the severe B*57:01/ABC-mediated immune reaction for responders. These data will open the door to comprehend the molecular basis of HLA-mediated adverse drug reactions.

ABSTRACT. Current sequencing methods allow for detailed samples of T cell receptors (TCR) repertoires. To determine from a repertoire whether its host had been exposed to a target, computational tools that predict TCR-epitope binding are required. Currents tools are based on conserved motifs and are applied to peptides with many known binding TCRs. We employ new Natural Language Processing (NLP) based methods to predict whether any TCR and peptide bind. We combined large-scale TCR-peptide dictionaries with deep learning methods to produce ERGO (pEptide tcR matchinG predictiOn), a highly specific and generic TCR-peptide binding predictor. A set of standard tests are defined for the performance of peptide-TCR binding, including the detection of TCRs binding to a given peptide/antigen, choosing among a set of candidate peptides for a given TCR and determining whether any pair of TCR-peptide bind. ERGO reaches similar results to state of the art methods in these tests even when not trained specifically for each test. The software implementation and data sets are available at https://github.com/louzounlab/ERGO. ERGO is also available through a webserver at: http://tcr.cs.biu.ac.il/.

ABSTRACT. Genotyping and sequencing improvements confirmed the important role of Human Leukocyte Antigen (HLA) in genome-wide association studies, as 20% of traits have an association within this locus. To cut the time and cost of HLA typing, algorithms such as HIBAG, allow predicting HLA alleles from Single Nucleotide Polymorphisms and a reference panel. However, HLA genes high polymorphism and reference panels of mainly European composition are a challenge for studying more diverse populations. The SNP-HLA Reference Consortium (SHLARC) aims to strengthen HLA association studies by improving HLA imputation with custom models.

We built models using the 1,000 Genomes (1KG) dataset (n=2,504 from diverse ancestry) as training and using African-American individuals (n=880) as a test sample. They consisted of the full dataset (n=2,504), and subsets of 200 individuals: super-population, or individuals genetically close to the target population. We defined genetically close individuals with dimension reduction (PCA or UMAP). Among subset models, we identified custom conditions that outperformed other reference panels of 1KG in terms of F1 score, a metric averaging precision and recall. For HLA-B, custom datasets have a mean F1 score near 0.45, whereas other studied models are between 0.10 and 0.35 (except the full dataset, 0.58). We identified alleles that were correctly imputed by our custom models, but absent from the full model. Unfortunately, such HLA alleles have lower imputation probabilities in custom models: it is impossible to merge both to improve results.

We highlight the need to create a larger reference panel. We conclude that custom models from different, but genetically close populations are not to be preferred to large multi-ethnic panels. The crucial point is the overlap of the HLA allele between training and target data. Relying on more individuals, and therefore HLA alleles, may slightly decrease HLA imputation of rare alleles but is overall more accurate.

ABSTRACT. The major histocompatibility complex (MHC), containing classical and non-classical HLA, is widely associated with transplant alloimmunity and human diseases. Study of immunogenetic disease associations has been greatly facilitated by single-nucleotide polymorphism (SNP) based imputation methods for the classical HLA genes. However, methods for imputation of the non-classical/non-HLA genes are sparse. The objective of the present study is to construct imputation models for five non-classical/non-HLA class I genes MICA, MICB, HLA-E, HLA-F and HLA-G including HLA-G UTR haplotypes using HIBAG imputation algorithm in a Finnish reference data set (MICA/MICB n=761, HLA-E n=441, HLA-F n=221, HLA-G n=435). The number of different alleles in the reference were 21, 17, 4, 3 and 20, respectively. The imputation results were compared with the sequence-based typing results to evaluate imputation accuracy at four-digit level for all genes except for HLA-G, for which imputation accuracy was evaluated at eight-digit level. The frequency of incorrectly imputed MICA, MICB, HLA-E, HLA-F and HLA-G alleles in the cross-validation varied between 0 and 0.7%. The models were further evaluated using an additional independent validation set of 214 Finnish individuals genotyped using a different genotyping platform than what was used for the reference set. Imputation of this validation set yielded slightly higher error rates (0.2 to 2.2%) due to the differences in the SNP composition of the genotyping platforms. The high accuracy of the models enables reliable imputation of class I non-classical HLA and non-HLA genes e.g. for large-scale association studies.

ABSTRACT. Platelet transfusion is a common treatment for patients with low platelet counts due to depletion caused by cancer chemotherapy or preparation for stem cell transplantation. Repeated transfusions can decrease in effectiveness, leading to a state of refractoriness, a serious complication causing bleeding and decreased survival. In many cases, the refractoriness is caused by the presence of alloantibodies to HLA Class I. The role of HLA mismatches in antibody development in platelet transfusions is known but incompletely explored. Transfusions are often matched between donor and recipient at a low resolution for HLA-A and HLA-B to reduce antibody development, but insights can be drawn from epitope-focused HLA mismatch simulations that predict the immunogenicity in pooled thrombocyte transfusions.

In this study, simulated populations were created based on established HLA haplotype frequencies. Simulated patients with a variety of HLA genotypes were selected, and donors were randomly selected and pooled into groups of 5 as per clinical practice and assigned sequentially to the patients. The T-Cell and B-cell epitope load for these transfusions were calculated using PIRCHE (I and II) and HLAMatchmaker algorithms, respectively, and the cumulative epitope load was summarized to define patient epitope load profiles. These profiles indicate that subsequent transfusions lead to increasing cumulative epitope load, which eventually reaches a plateau. The epitope load plateau varies widely between recipients, indicating the presence of high-and low-risk patients. These simulations demonstrate the immunogenicity of HLA mismatches in the context of platelet transfusion, allows the identification of high- and low-risk patients and donors, and enables in silico analysis of different allocation techniques.

ABSTRACT. The IPD-IMGT/HLA Database is the official repository for sequences of the human major histocompatibility complex named by the WHO Nomenclature Committee for Factors of the HLA System. The database provides a highly curated dataset of HLA sequences and metadata, extensively adopted by the HLA community and the wider research and clinical communities. The advent of Next-Generation Sequencing and its application for high-throughput HLA sequence-based typing has resulted in a greater need for tools for the analysis of genomic variation. Furthermore, the increasing volume of data has impacted how data is consumed, shifting the usage from web-based graphical user interfaces to programmatic access. The database has grown substantially since it’s release in 1998, with curators regularly processing over 1,000 submissions per month, compared to the first release which contained less than 900 alleles. The sequences submitted are also now predominately gDNA, which means submission length has increased from 270-546 bps to between 3-16 kb. This increase in submissions now means that some genes have in excess of 8,000 alleles and the traditional methods for searching and visualising the data are less appropriate. To facilitate its continuous development, and to provide better fruition of data, the IPD database has developed a REST API for querying the database. As a result, a new version of the database has been released, allowing users to access both the classical tools and the programmatic access for advanced users. The new API provides powerful search options for the allele and cell data, as well as new programmatical interfaces for the matching tools. The traditional search tools have been updated to run using the API whilst retaining a more traditional web form allowing the user to build queries visually, from simple searches to complex queries using a variety of computational operators. The IPD-MHC, IPD-NHKIR and IPD-KIR Databases have also benefited from these improvements.

ABSTRACT. This genetically determined diversity affects severity of infections, immune-mediated diseases, and some cancers, and impacts the course of immunotherapies, including transplantation. KIR3DL1 is an inhibitory receptor, and KIR3DS1 is an activating receptor encoded by the KIR3DL1/S1 gene that has more than 200 diverse and divergent alleles. Determination of KIR3DL1/S1 genotypes for medical application is hampered by complex sequence and structural variation, requiring targeted approaches to generate and analyze high-resolution allele data. To overcome these obstacles, we developed and optimized a model for imputing KIR3DL1/S1 alleles at high-resolution from whole-genome SNP data. We designed the model to represent a substantial component of human genetic diversity. Our Global imputation model is effective at genotyping KIR3DL1/S1 alleles with an accuracy ranging from 88% in Africans to 97% in East Asians, with mean specificity of 99% and sensitivity of 95% for alleles >1% frequency. We used the established algorithm of the HIBAG program, in a modification named Pulling Out Natural killer cell Genomics (PONG). Because HIBAG was designed to impute HLA alleles also from whole-genome SNP data, PONG allows combinatorial diversity of KIR3DL1/S1 with HLA-A and -B to be analyzed using complementary techniques on a single data source. The use of PONG thus negates the need for targeted sequencing data in very large-scale association studies where such methods might not be tractable. All code, imputation models, test data and documentation are available at https://github.com/NormanLabUCD/PONG

ABSTRACT. The genes C4A and C4B, located on chromosome 6, encode for Complement 4 (C4), a key protein in the classical and lectin pathways of the complement system. The C4A and C4B loci exhibit copy number variation, with each gene varying between 0-4 copies. These genes also vary in size depending on the presence of a retrovirus in intron 9 that modulates expression, resulting in long (L) and short (S) forms of C4A and C4B. Additionally, human blood group antigens Rodgers (Rg) and Chido (Ch) are located on the C4 protein, with the Rg epitope generally found on C4A and the Ch epitope generally found on C4B. Despite the central role of C4 in immunity and their implication in numerous diseases, sequence analysis of C4 variants has been impeded by the high degree of sequence similarity and their complex structural variation. To address these limitations, we have developed a novel bioinformatic pipeline for comprehensive, high-throughput characterization of human C4A and C4B sequence from short-read sequencing data, named C4Investigator. Using paired-end data from targeted or whole-genome sequencing, C4Investigator determines copy numbers for overall C4, C4A, C4B, C4(Rg), C4(Ch), C4(L), and C4(S). Additionally, C4Ivestigator identifies sequence variants, including frame-shift mutations and recombinations. Our analysis of 2998 whole genome sequence (WGS) samples from the 1000 Genomes Project showed high frequencies of novel C4A-C4B recombinants in certain populations, for example the Mende population with 41% frequency (39/95) of the C4A(Ch) recombinant, and the Tamil population with 11% frequency (12/113) of the C4B(Rg) recombinant. Identifying these common variants was so far intractable with digital droplet PCR, the current gold standard method for C4 copy number, highlighting the benefit of using our sequence analysis method to start the new era of C4 interrogation.

ABSTRACT. Abacavir hypersensitivity syndrome (AHS) resulting in multiorgan reactions occurs in approximately 2%-8% of the patients with human immunodeficiency virus-1 (HIV-1) infection. The association of AHS with HLA-B*57:01 is very strong and testing for HLA-B*57:01 is advised before prescription of abacavir. However, with the increased use of high-resolution-typing-techniques many new HLA-B*57 alleles are reported and currently 234 alleles and 226 proteins (alleledatabase 3.46) are described. Recently we identified a new non-HLA-B57:01 allele in a clinical sample and the allele name B*57:145 has been assigned by the WHO Nomenclature Committee for Factors of the HLA System in August 2021. Unfortunately, the frequency of occurrence of most HLA-B*57 alleles is not sufficient to enable disease association studies. Since the abacavir binding capacity of some HLA-B*57 alleles is described, abacavir binding can be predicted. Based on physical and biochemical properties of abacavir in HLA-B*57, we grouped HLA-B*57 alleles according to permissiveness to abacavir binding. N=102 alleles were grouped as permissive, 4 as probably-permissive and 4 as non-permissive. The B*57:145 allele was grouped as “permissive” and it was advised not to prescribe abacavir to this patient. Using this prediction model, AHS can be prevented in patients with HLA-B*57 alleles not tested in association studies.