ABSTRACT: - EFI 2017

DOI: 10.1111/tan.13035

MINI ORAL PRESENTATIONS Solid Organ Transplantation MO1

MO2

ELEVATED SERUM PENTERAXIN IN CARDIAC ALLOGRAFT ANTIBODY-MEDIATED REJECTION

PREVALENCE AND CLINICAL RELEVANCE OF ANGIOTENSIN II TYPE 1 RECEPTOR ANTIBODIES IN PEDIATRIC RENAL TRANSPLANT RECIPIENTS

Medhat Askar1, Minoo Kavaran2, Katherine Twoembley2, Omar Moussa3 1

Baylor College of Medicine, Huston, USA, 2Medical University of South Carolina, Charleston, USA, 3University of Michigan, Ann Arbor, MI, USA Correspondence: [email protected] Pentraxin 3 (PTX3), also known as tumor necrosis factorinducible protein-14 (TSG-14) is a glycoprotein that can be secreted by several cell types including endothelial and mononuclear cells following stimulation with inflammatory cytokines including TNF-alpha and IL-1 beta. PTX3 plays an important role in innate immune system by acting as a pattern recognition receptor. Up-regulation of PTX3 increased the pro-inflammatory responses in several cell types. The role of PTX3 in cardiac allograft antibodies mediated rejection is investigated in this study. To investigate the downstream effects mediated by HLA antibodies, we utilized a protein microarray approach to study the alteration in protein expression following treatment of primary HUVEC endothelial cells with humanized HLA class I antibodies. The secreted PTX3 levels were determined in serum samples using ELISA. Endothelial cells treated with HLA class I antibodies showed increased PTX3 protein levels using the protein membrane arrays. Protein arrays results were further confirmed with ELISA. Higher serum PTX3 levels were observed in 18 AMR patients (Mean 15 SD 7.2) compared to 15 non-AMR patients (mean 33.14 SD 16.69), and 6 healthy controls (mean 1.7 SD 1.3). The serum levels of PTX3 are elevated in cardiac allograft patients with AMR. The use of PTX3 as possible biomarker to monitor AMR will need further evaluation on a larger scale. The biological significance of PTX3 secretion during AMR requires further investigation.

Claudia Schroeder1, Alexander Fichtner1, Caner Süsal2, Britta Höcker1, Susanne Rieger1, Jens Westhoff1, Rüdiger Waldherr3, Burkhard Tönshoff1 1

University Children’s Hospital Heidelberg, Heidelberg, Germany, 2Institute of Immunology, University of Heidelberg, Heidelberg, Germany, 3Institute of Clinical Pathology Heidelberg, Heidelberg, Germany Correspondence: [email protected]

The predominant role of antibody-mediated rejection (ABMR) for premature graft dysfunction is well established in adult patients. The main cause is donor-specific HLA-antibodies (HLA-DSA); but not all cases could be attributed to HLADSA. The aim of this cross-sectional study was to describe the prevalence of angiotensin II type 1 receptor antibodies (AT1RAB) and analyze their association with biopsy results and graft function in 62 pediatric renal transplant recipients. Serum samples at the time of an indication biopsy performed ≥ 1 year post-transplant were analyzed for the presence of HLA-DSA and AT1R-AB. HLA-DSA with an MFI value above 500 were classified as positive (LABScreen Luminex Kits (One Lambda)). The cut-off value for AT1R-AB was set at ≥10 U/mL (ELISA, CellTrend GmbH). 32 of 62 patients (52%) were positive for AT1R-AB at the time of indication biopsy. After stratification according to histologic entities, the 28 patients with ABMR showed significantly elevated AT1RAB levels (12, IQR 10–17 U/mL) compared to the 15 patients with T-cell mediated rejection (9, IQR 7–10 U/mL, P = 0.016, Dunn post-hoc) or the 19 patients without rejection (8, IQR 6–14 U/mL, P = 0.009, Dunn post-hoc). ROC-curve analyzes revealed 10 U/mL as an optimal cut-off to differentiate between ABMR and non-ABMR (ROC-AUC 0.76). AT1R-AB positive patients more often reached the end-point “graft function deterioration”, defined as an eGFR-decrease more than 50% of baseline at the time before indication biopsy (P = 0.012, Kaplan-Meier analysis). Subdividing the group by

© 2017 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd HLA. 2017;89:373–383.

wileyonlinelibrary.com/journal/tantan

373

Abstract

374

HLA-DSA and AT1R-AB status revealed no significant difference between AT1R-AB-positive and AT1R-AB and HLADSA double-positive groups (P = 0.33). In conclusion, our results indicate that AT1R-AB positivity at the time of an indication biopsy is associated with a significantly increased risk of deterioration of graft function.

MO3

SOLUBLE CD16 (SCD16) IS AFFECTED BY IMMUNOSUPPRESSIVE DRUGS AND ASSOCIATES SIGNIFICANTLY WITH ACUTE RENAL REJECTION Maria Paula D.P. Xavier1, Susana Sampaio Norton2, José Gerardo Oliveira3 1

Instituto Português do Sangue e da Transplantação, Porto, Portugal, 2CHSJ, Porto, Portugal, 3Faculdade Medicina Porto, Hospital S João, Porto, Portugal Correspondence: [email protected] Serum levels of sCD16 depend mainly on the number of CD16 positive cells and on the density of CD16 expression at the cell membrane. CD16 expression is expressed mainly by neutrophils and monocyte lineage cells, natural killer cells and can mark a dendritic cell subtype with higher antigen presenting capacity. Other cell types seem to be able to express it, including capillary cells during transplant heart rejection. sCD16 was reported to be unaffected by calcineurin inhibitor (CNI), rapamycin (Rapa) and mycophenolate mofetil. Previously we reported that CD16 expression is significantly higher in aspiration biopsy of acutely rejecting kidney transplants. We studied sCD16 in cadaver kidney allograft recipients (KTx). KTx receiving different treatments, CNI (n=13), RAPA (n=8), thymoglobulin induction (n=10), anti-IL-2 receptor antibody (n=15), all stable patients and patients with acute (n=18) and chronic (n=7) rejection were confirmed by biopsy. Aspiration biopsy was done on day 7 post-transplant in stable cases and on rejection day. Samples were cultured and supernatants were collected at day two of incubation and analyzed for sCD16 by ELISA. No significant differences for patients’ demographics were observed among groups. Results are expressed in ng/mL, quartiles. CNI stable 21–25 ng/mL, anti-IL-2 receptor antibody 24–27, Rapa 27–40, thymoglobulin induction 21–23, acute rejection 25–38 and chronic rejection 26–28. The value of sCD16 was significantly lower among thymoglobulin as compared to CNI (P = 0.008), as compared to anti-IL-2R (P = 0.035) and anti-IL-2R was lower as compared to Rapa (P = 0.028). sCD16 was significantly higher in acute rejection than in stable cases (P = 0.014). Our results display for the first time a significant association between sCD16 and acute rejection in human kidney transplant and contrary to others we saw different effects of immunosuppressive drugs, the higher down-modulation observed with thymoglobulin.

MO4

PIRCHE-II: A NOVEL TOOL TO IDENTIFY PERMISSIBLE HLA MISMATCHES IN KIDNEY TRANSPLANTATION Kirsten Geneugelijk1, Matthias Niemann2, J. Drylewicz1, Arjan D. van Zuilen1, Irma Joosten3, Wil A. Allebes3, Arnold van der Meer3, Luuk B. Hilbrands3, Marije C. Baas3, C. Erik Hack1, Franka E. van Reekum1, Marianne Verhaar1, Elena G. Kamburova1, Michiel L. Bots1, Marc A.J. Seelen4, J.S.F. Sanders4, Bouke G. Hepkema4, Annechien J. Lambeck4, Laura B. Bungener4, C. Roozendaal4, Marcel G.J. Tilanus5, Joris Vanderlocht5, Christina E.M. Voorter5, Lotte Wieten5, Elly M. van Duijnhoven5, Mariëlle Gelens5, Maarten H.L. Christiaans5, Frans J. van Ittersum6, Azam Nurmohamed6, Neubury M. Lardy7, Wendy Swelsen7, Karlijn A. van der Pant8, Neelke C. van der Weerd8, Ineke J.M. ten Berge8, Frederike J. Bemelman8, Andries Hoitsma9, Paul J.M. van der Boog10, Johan W. de Fijter10, Michiel G.H. Betjes11, Sebastiaan Heidt10, Dave L. Roelen10, Frans H.J. Claas10, Henny G. Otten1, Eric Spierings1 1

University Medical Center Utrecht, Utrecht, the Netherlands, PIRCHE AG, Berlin, Germany, 3Radboud University Medical Center, Nijmegen, the Netherlands, 4University of Groningen; University Medical Center Groningen, Groningen, the Netherlands, 5Maastricht University Medical Center, Maastricht, the Netherlands, 6VU University Medical Center, Amsterdam, the Netherlands, 7Sanquin, Amsterdam, the Netherlands, 8Academic Medical Center, Amsterdam, the Netherlands, 9Dutch Transplant Foundation (NTS), Leiden, the Netherlands, 10Leiden University Medical Center, Leiden, the Netherlands, 11 Erasmus Medical Center, Rotterdam, Rotterdam, the Netherlands Correspondence: [email protected] 2

Individual HLA mismatches may have differential effects on graft survival after kidney transplantation. Therefore, there is a need for a reliable tool to define permissible HLA mismatches in kidney transplantation. We previously demonstrated that donor-derived Predicted Indirectly ReCognizable HLA epitopes presented by recipient HLA class II (PIRCHEII) play a role in de novo DSA formation after kidney transplantation. In the present Dutch multi-center study we evaluated the possible association between PIRCHE-II and kidney graft failure in 2,918 donor-recipient couples that were transplanted between 1995 and 2005. For these donor-recipient pairs, PIRCHE-II was determined and was related to graft survival in both univariate and multivariate analyzes. Adjusted for confounders, the natural logarithm of PIRCHE-II was associated with a higher risk for graft failure (HR:1.13, 95% CI:1.04-1.23, P = 0.003). Univariately analyzed, patients with low PIRCHE-II numbers had a better 10-years graft survival than patients with higher PIRCHE-II numbers (P = 0.006; PIRCHE-II strata: 500 IU/mL) and 41% (prophylaxis, PCR CMV-DNA > 40 IU/mL). According to our preliminary data, in the preemptive group a cutoff of 19 spot-forming cells (SFC) to the TTrack CMV IE1 and of 130 SFC to the T-Track CMV pp65

corresponded with protection against reactivation. At the end of prophylaxis the T-SPOT.CMV pp65 was the best marker to predict protection (cutoff of 54 SFC). QuantiFERON CMV was not predictive in any of the two groups. An excellent positive agreement was obtained between T-SPOT.CMV or TTrack CMV and CMV IgG (kappa = 1.000 or 0.905, respectively) whilst a moderate agreement was obtained for QuantiFERON CMV (kappa = 0.648). In conclusion, the two ELISpot assays were superior to functionally assess the CMVspecific immunity in transplant recipients.

MO6

INFLUENCE OF IMMUNOSUPPRESSION ON SCD30 LEVELS AND ASSOCIATION OF SCD30 LEVELS WITH CMV INFECTION/DISEASE IN KIDNEY TRANSPLANT RECIPIENTS UNDER DIFFERENT IMMUNOSUPPRESSIVE REGIMENS Patricia C. Grenzi1, Erika F. Campos1, Hélio Tedesco-Silva2, Claudia R. Felipe2, Marcello F. Franco3, José O. MedinaPestana2, Hinrich P. Hansen4, Maria Gerbase-DeLima1 Immunogenetics Institute – AFIP/Federal University of Sao Paulo, São Paulo, Brazil, 2Hospital do Rim e Hipertensão, São Paulo, Brazil, 3Federal University of São Paulo, São Paulo, Brazil, 4University Clinic Cologne, Cologne, Germany Correspondence: [email protected]

1

CD30, a member of the tumor necrosis receptor superfamily, is strongly expressed on activated lymphocytes. The ectodomain shedding of CD30 results in soluble CD30 (sCD30) that may serve as a prognostic serum marker of outcomes in kidney transplantation (Tx). In a previous study (Grenzi et al, 2016), we observed that de novo kidney transplant recipients (R), converted at month-3 post-Tx from tacrolimus (TAC) to sirolimus (SRL), had significant lower sCD30 levels in comparison with patients maintained with TAC, in different samples collected up to 24 months post-Tx. To further investigate the influence of these drugs on sCD30 serum levels, in the present study we tested samples collected on days 7, 90, 180 and 365 post-Tx in 288 R, randomized to receive antithymocyte globulin, TAC (low dose) and everolimus (r-ATG/EVR, n=85); basiliximab, TAC (low dose) and EVR (BAS/EVR, n=102); or basiliximab, TAC and mycophenolate (BAS/MPS, n=101). All patients received prednisone. Serum sCD30 levels were significantly lower in R treated with EVR, even when combined with reduced doses of TAC (r-ATG/EVR and BAS/EVR), in comparison with BAS/MPS-treated R (p92.99, respectively. Overall, allele concordance against previous KIR typing was 92% (86/93 alleles) and CDS concordance to existing KIR alleles was 99.99% of nucleotides (107002/107006), demonstrating the legitimacy of our technique. The discordant alleles (8%, 7/93 alleles) include three with novel polymorphisms (3%) and five completely discrepant types (5%), four of which form examples of allele combinations that require full-length, isolated sequencing to determine polymorphism phase, thereby highlighting the advantage that SMRT DNA sequencing has over other KIR allele typing methods. Sanger sequencing is being performed to confirm all other polymorphisms. We have been able to specifically target and simultaneously sequence multiple full-length KIR alleles to definitively assign KIR types. By increasing the resolution at which we study KIR, we aim to better understand their clinical importance as NK cell receptors.

MO11

OPTIMAS: A MATCHING SERVICE FOR THE NEW BMDW SEARCH AND MATCH SERVICE Hans-Peter Eberhard1, Werner Bochtler1, Markus Beth1, Andreas W. Vogt1, Daniel Freund1, Hans-Georg Rist1, Lydia Foeken2, Carlheinz R. Müller1 ZKRD – German National Bone Marrow Donor Registry, Ulm, Germany, 2World Marrow Donor Association (WMDA), Leiden, the Netherlands Correspondence: [email protected] 1

MO10

KILLER-CELL IMMUNOGLOBULIN-LIKE RECEPTOR (KIR) ALLELE TYPING USING SINGLE MOLECULE REAL-TIME (SMRT) DNA SEQUENCING FROM FULL LENGTH PCR AMPLICONS Will P. Bultitude1, Arthur W. Gymer2, James Robinson1, Neema P. Mayor1, Steven G.E. Marsh1

In 2006, the German National Bone Marrow Donor Registry (ZKRD) introduced the OptiMatch® search engine for the probabilistic matching of volunteer unrelated stem cell donors to patients in need of a transplant. In subsequent years, the core

Abstract

378

parts of OptiMatch® were isolated into a deployable standalone matching service called OptiMaS (OptiMatch® as a Service). OptiMaS has been used in the production environments of the Canadian OneMatch registry since 2012 and of the Australian Bone Marrow Donor Registry since 2013 and has meanwhile demonstrated to be a sound option for registries to benefit from the advantages of the OptiMatch® matching service locally. In 2016 OptiMatch® was selected as a matching algorithm for the new Bone Marrow Donors Worldwide (BMDW) Search & Match Service. In addition to the OptiMatch® search engine, the OptiMaS framework uses solely open-source software. Altogether, OptiMaS is a black box computer giving access to the capabilities of OptiMatch® via a small set of high-level web service functions. It fully supports the European Marrow Donor Information System (EMDIS) matching preferences and is compliant with the World Marrow Donor Association (WMDA) HLA Nomenclature Guidelines. For this study, we have investigated the live system’s performance for 60 consecutive days as well as for a complete recalculation of the search report for all patients. The analysis is based on over 29 million donors and cord blood units and on over 4000 for the daily and over 3000 patients for the complete calculation. We have focused on average matching times for typical search settings and highlighted special constellations leading to extreme (low and high) system burden. We have also studied the effects of input parameters on parallel processing. Daily average match time is 2 minutes with 50% of the searches returning within 25 seconds on a machine with 4 dedicated matching processes. In conclusion, OptiMaS has shown for two representative scenarios to provide an appropriate matching service for BMDW’s global patient and donor load.

MO12

VALIDATION OF AMPLICON-BASED NEXT GENERATION SEQUENCING OF HUMAN LEUKOCYTE ANTIGENS BY MORE THAN 17,000 CONFIRMATORY TYPING RESULTS

1

DKMS gGmbH, Tübingen, Germany, 2DKMS Life Science Lab, Dresden, Germany Correspondence: [email protected] Since 2013, typing of DKMS donors at recruitment level has been performed by an in-house developed amplicon based NGS approach. Since this release, over 1.8 million potential unrelated stem cell donors have been typed at recruitment for HLA-A,-B,-C,-DRB1,-DQB1 and -DPB1 with over 98% highresolution results by DKMS Germany. For these donors, more than 95,000 HLA loci have been re-typed in the context of over 17,000 CT requests. In total we found 129 discrepant typing reports of which 115 (89%) were due to genotyping errors during confirmatory typing. Only for 11% (n=14) the typing at recruitment turned out to be erroneous. 11 of these cases referred to HLA-DPB1, 2 cases to HLA-DQB1 and one to HLA-B. Issues with homozygosity handling were responsible for most DPB1 discrepancies. We observed this effect in a preliminary analysis in 2015 and optimized our software neXtype to cope with homozygosity by adjusting the detection parameters for hetero- and homozygosity. Since this optimization, no additional errors due to false detection of homozygosity have been observed. Out of 115 errors in CT, 98 (85%) comprised HLA class II loci. About half (60) of these 115 discrepancies were outside the transplantation relevant antigen recognition domain and might be due to inadequate usage of multi-allele codes for reporting typing results. 34 of the 39 DRB1 errors were related to DRB1*14:01 vs. DRB1*14:54 that could be distinguished by typing exon 3 to recognize the relevant SNP. If exon 3 is not typed, the result should be reported as DRB1*14:01:01G (or DRB1*14: BCAD). One error was related to ambiguities remaining in Sanger sequencing which are resolved by next generation sequencing. In conclusion, we were able to show that using an NGS based typing strategy is of exceptional high quality. We observed an error rate below 0.15 ‰ (14 out of 95,876) with the maximum error rate per locus not exceeding 1.6‰ for DPB1 (11 out of 6,921).

Daniel M. Baier1, Heike Fischer1, Jennifer Wuchter1, Annett Heidl2, Irina Böhme2, Vinzenz Lange2, Jürgen Sauter1, Jan A. Hofmann1, Alexander H. Schmidt1

New Technologies

MO13 WHOLE MHC SEQUENCING USING SEQCAP TARGET ENRICHMENT AND ILLUMINA HISEQ 2500 Kathrin Starz, Barbara Bangol, Riccardo Brumm, Daniel Becker, Kaimo Hirv Center for Human Genetics and Laboratory Diagnostics Dr. Klein, Dr. Rost and Colleagues, Martinsried, Germany

Correspondence: [email protected] The major histocompatibility complex (MHC) on chromosome 6 is a region with the highest genetic variability in a human genome. In addition to the undisputed importance of the human leukocyte antigens for the selection of matched solid organ or blood stem cell donors, this region encompasses many other genes with immune-related function,

Abstract

which are involved in etiopathogenesis of various diseases. High-resolution MHC sequencing can be utilized to elucidate the molecular immunopathology of MHC-associated diseases or to avoid severe complications after blood stem cell transplantation. Sequencing the whole MHC is challenging due to the genetic complexity and extensive polymorphism of this region. We used Roche Nimblegen SeqCap EZ library, which targets the whole MHC and surrounding regions (1.6 Mb) with a total design of 4.97 Mb. For 17 gDNA samples containing three related and two unrelated blood stem cell transplantation patient/donor pairs, the first step was fragmentation using the KAPA HyperPlus Kit. Target regions were enriched and captured by hybridizing the gDNA with the SeqCap EZ library before using Roche Nimblegen adapter sequences to multiplex gDNA samples. After a second amplification, the library was sequenced on one flow cell lane using Illumina HiSeq 2500. Run output for one lane was 184.7 million paired-end reads and 163.8 million passing filter reads, resulting in 81.6 Gb of sequence. We used CLC Genomic Workbench 9.0 (Qiagen) to analyze the results and generate a coverage report as well as variant charts for the related and unrelated patient/donor pairs. Mean sequence coverage over the whole MHC was 205 for 17 gDNAs. A total of 85-93% of the target region was sequenced with a coverage >20. The combination of the SeqCap target enrichment and massively parallel sequencing on HiSeq is a feasible approach for whole MHC sequencing and can be used in many clinical and research settings such as transplantation medicine, cancer, inflammatory or autoimmune disorders.

MO14

ADENOVIRUS-MEDIATED GENE TRANSFER OF HLA-B*44:02 INDUCES PROTEIN EXPRESSION ON CELL SURFACE OF CERVICAL CANCER CELL LINE SIHA Josefa A. Rodríguez1, Nataly Cruz1, Lina M. Martinez1, Johana A. Lineros1, Alba Lucía. Cómbita1, Jessica C. Ruiz1, Diana M. Palacios2, Vilma L. Medina1, Diana R. Tovar3 1

National Cancer Institute, Bogotá, Colombia, 2Fundación Santa Fe de Bogotá, Bogotá, Colombia, 3District Hemocenter, Bogotá, Colombia Correspondence: [email protected] Altered HLA expression on tumor cell surface is the most widespread mechanism used by cancer cells to avoid tumor immune-detection and destruction by T-cells. In cervical cancer, the HLA-B*44 allele is lost more frequently than other HLA class I alleles. This specific allele loss may play an important role in immune escape mechanisms. Here we standardize a methodology to restore HLA-B*44:02 expression by gene transfer using an adenoviral vector. HLA status in various

379

cancer cell lines were determined using PCR-SSO and luminex analysis to decide which one can be useful as control for positive gene transfer of the HLA-B*44 gene. SiHA cell line was chosen due to its homozygous HLA phenotype (HLA-A*24:02/ A*24:02; HLA-B*40:02/B*40:02 and HLA-C*03:04/C*03:04). The homozygous phenotype was confirmed by LOH analysis at 6p and 15q. HLA-B*44 gene was amplified and cloned on the p-Shuttle-CMV transfer vector and the resultant p-ShuttleCMV-HLA-B*44:02 was linearized by digestion with Pme I. Then, the linearized vector and adenoviral backbone plasmid were used to co-transform E. coli BJ5183 to obtain recombinant adenoviral vectors. It was amplified, purified and confirmed by duplex PCR. Then, Pac I linearized recombinant plasmid was transfected into adenovirus packaging cell line Ad293 and primary recombinant virus stock was obtained and amplified. Further, SiHa cell line was infected with the recombinant virus and surface cell protein expression was analyzed by flow cytometry using a specific antibody. We found HLAB*44 expression in 58.9% of infected cells at 72 hours compared to 1.99% of uninfected control. In conclusion we achieved HLA-B*44 expression in SiHa cell line, however further analysis are required to determine if HLA-B*44 restoration improve the cytotoxic activity of heterologous PBMCs against transfected tumor cells. This approach could have important implications for designing immunotherapeutic strategies for cancer.

MO15

HLA TYPING WITH THE NANOPORE DEVICE Gottfried F. Fischer, Sabine Wenda, Ingrid Faé Medical University Vienna, Vienna, Austria Correspondence: [email protected] Next generation sequencing (NGS) has drastically changed our ability to resolve HLA typing ambiguities. While for Sanger sequencing, the analysis was confined to exons that encode the peptide binding regions of the HLA molecules, NGS for the first time allowed the characterization of the whole gene on a routine basis as duration and costs of the analysis have become comparable to the classical approach. The real flaw of the second generation NGS approach is, however, the need to assemble the sequenced fragments (with a length of some hundred base pairs) to the full gene by either aligning to reference sequences or by de novo assembly. Both approaches do not always maintain the phase of a single allelic sequence across its whole length, resulting in typing ambiguities. A technically simple alternative to the second generation NGS devices at present is the MinION device from Oxford Nanopore. Here, sequencing happens in a device of USB stick size. We tested the applicability of this approach by typing HLA class I and class II genes. Preparation of the sequencing templates was similar to the work flow for our routine sequencing on the Ion Torrent device. The method was in fact considerably simpler since fragmentation and clonal

Abstract

380

amplification could be omitted, resulting in a two hour hands-on time after long range PCR. Base calling took place on cloud based software provided by Nanopore. Eighty percent of the clonal sequences could be aligned to the reference sequence using the BLAST software. Errors comprised insertions, deletions, and nucleotide exchanges and exceeded 15%. A satisfactory amount of full length sequences, however, was available. In conclusion, the Nanopore approach needs high coverage and specialized analysis tools to allow unequivocal assignment. It could, however, provide a solution for the phasing problem.

MO16

NORMAL KARYOTYPE FLT3 ITD AML PATIENTS HAD LOWER NUMBER OF GAINS IDENTIFIED BY ARRAY CGH IN THE 17Q21.31 REGION THAN THOSE LACKING THE MUTATION Emilia Jaskuła1,2, Janusz Lange2, Mariola Sedzimirska2, Agnieszka Tarnowska2, Anna Sobczynska-Konefal2, Andrzej Lange1,2 1

L. Hirszfeld Institute of Immunology and Experimental Therapy, Polish Academy of Sciences, Wrocław, Poland, 2Lower Silesian Center for Cellular Transplantation with National Bone Marrow Donor Registry, Wrocław, Poland Correspondence: [email protected] Genetic dissection of AML patients helps in tailoring treatment according to the identified mutations and/or risk factors associated with genetic hallmarks of the disease. Normal karyotype (NK) patients represent a poorly defined subgroup of AML patients. In this study, we used array based comparative genomic hybridization (CGH) in the group of patients with NK to find out whether the presence of FLT3 ITD associates with CGH and identify gains and losses as candidates for more precise description of NK FLT3 ITD negative (FLT3 ITD-) patients. 91 patients with AML (median age: 57.5 years old (range 21–81 years), 87 primary and 4 secondary AML) were investigated. 55 patients had NK and/or were FISH negative for abnormalities as follow: XY, inv(3), −5/5q-, −7/7q-,+8, MLL, RUNX1, PML/RARA or RARA, inv(16) and MLL. 27 patients in this group were FLT3 ITD positive (FLT3 ITD+). It appears that NK FLT3 ITD+ patents had a lower number of gains (two or more copies: 2.43 0.71 vs. 4.78 0.65, P = 0.029) and one copy loss (2.13 0.80 vs. 4.97 2.54, P = 0.048). For further comparisons, we weaved out two subgroups (NK FLT3 ITD- vs. NK FLT3 ITD +) well balanced regarding the number of patients and their demographic and disease features. Array CGH was performed in all patients and the results are as follow: NK FLT3 ITD- patients characterized with a higher number of gains in the 17q 21.31 sub-region: 41.57-41.71Mb, encompassing the DDX8 and E1AF genes, than NK FLT3 ITD+ individuals (21.9% vs. 4.8%, P = 0.126). NK FLT3 ITD+ patients had a higher number of gains in the region 17q21.31:41.80-42.14Mb (23.8% vs. 3.12%, P =

0.032) covering the genes: CDD (involved in process: negative regulation of Wnt signaling pathway), VHR (inactivation of MAPK activity), DLG3 (CD300L, immune system process), CLM9, DLG2, APR-2, PP, PYY1, AGAS, PNAS-135, SCM4, transmembrane protein 101 (positive regulation of I-kappaB kinase/NF-kappaB signaling). 20 patients were examined with the both techniques (CGH and FISH for 17q21.31 41.5741.71Mb region); 18 out of 20 results were concordant. In conclusion, FLT3 ITD NK patients have a lower number of CNV than those lacking this mutation if the mutation is present, which is associated with a higher number of gains in chr17q21.31, possibly affecting the response to FLT3 kinase inhibitors.

MO17

APPROACHING HLA GENOTYPING USING NANOPORE SEQUENCING: TECHNOLOGY EVOLUTION DRIVES PERFORMANCE IMPROVEMENTS Vineeth Surendranath1, Gerhard Schöfl1, Kathrin Lang1, Hanneke W.M. van Deutekom2, Erik H. Rozemuller2, Alexander H. Schmidt3, Vinzenz Lange1 1

DKMS Life Science Lab, Dresden, Germany, 2GenDx, Utrecht, the Netherlands, 3DKMS, Tübingen, Germany Correspondence: [email protected] The recent advent of Nanopore Sequencing allows sequencing of single molecules with lengths in the order of hundreds of kilobases. It represents a promising platform for the generation of full-length HLA allele sequences and ameliorating SNP phasing issues which currently challenge Second Generation Sequencing platforms due to limitations in read length. At the DKMS Life Science Lab, the MinION sequencer from Oxford Nanopore Technologies has been in use as a test bed for fulllength HLA sequencing since March 2015. During this time, the MinION technology has progressed through multiple iterations with regards to the biological pore used, the associated chemistry, and the algorithms underlying the basecaller. We have used a benchmarking set of 30 samples, 10 each for HLA-A, HLA-B and HLA-C to explore 3 successive Nanopore chemistries R7.3, R9 and R9.4 using both template-only (1D) reads and template-complement (2D) reads. Concomitant with improvements in chemistry, read yield has progressively increased from approximately 3000 reads to 8000 reads per sample. We observed a concordant increase in mappability against the benchmarking reference sequences from 90 to 98% for 1D reads and 97 to 99.8% for 2D reads. Comparison of MinION-derived consensus sequences with benchmarking reference sequences shows an increase in sequence accuracy from 90% to 98% for 1D reads, and from 98% to 99.8% for 2D reads. The only errors remaining in the consensus sequences derived from the latest MinION chemistry (R9.4) are due to gaps introduced by systemic undercalling of polynucleotide stretches. Therefore, homopolymer accuracy remains an open

Abstract

381

issue, although progress to resolve this issue is being made. These data establish Nanopore sequencing as a viable platform for HLA mapping and consensus calling with high accuracy, which is a first step towards successful HLA genotyping.

MO18

FROM WHOLE GENE SEQUENCING TO WHOLE GENOME SEQUENCING IN HUMANS Nezih Cereb, HwaRan Kim, Jaejun Ryu, Eunsil Kim, Soo Young Yang Histogenetics, Ossining, New York, USA Correspondence: [email protected] We started sequencing full gene HLA class I and Long Range (exon 2 and 3) HLA class II at large scale at Histogenetics on PacBio RS II® platform in the beginning of July 2016. To date we have sequenced over 150,000 samples for whole gene class I and Long Range class II. With this approach we are able to provide allele level results and rarely using NMDP codes. NMDP Codes were used for certain class I alleles where the variations were at the outside of primer binding sites at 5’ and 3’ UTRs. We were also able to full gene sequence certain KIR genes (3DL1) on PacBio RS II® platform. Most recently we sequenced a batch of amplicons that were already sequenced on PacBio RS II® platform on PacBio Sequel® platform using

v2 chemistry and v4 software. We found no difference in quality and base calling between the two platforms. The advantage of Sequel is that the SMRTcells have 1,000,000 ZMW as compared to 150,000 on PacBio RSII® platform, that means more a than 6-fold increase in capacity and multiplexing ability. We wanted to use this new breakthrough in sequencing technology to sequence the whole genome of a known human subject. All HLA genes of this subject been sequenced at whole gene level for class I and long range (exon 2 - 3’UTR) for class II. KIR gene content was determined using in house method on Illumina platforms. Full or partial sequences of KIR genes were determined with amplicon and/or fosmid cloning. We prepared a 40 kb Library with 35 kb cut-off size selection and followed manufacturer’s protocol thereafter. Sequencing performed on PacBio Sequel® platform using v2 chemistry and v4 software. Movie time for these large fragments were set to 600 minutes and secondary analysis was performed with SMRTLink’s (v4.0.0) SMRT Analysis Application: de novo Assembly (HGAP 4). In order to get 50 x coverage, we have sequenced the same library in 30 different SMRT® Cell 1M v2. In this presentation we will focus on genomic regions coding the immune response genes and analyze their concordance with some known genotypes established by long amplicon sequencing. To our knowledge this is also one of the first human genome sequenced using v2 chemistry on Sequel.

Miscellaneous

MO19

ALLELIC DIVERSITY AND HAPLOTYPE STRUCTURE OF HLA-DPB1 BASED ON THE PHASED FULL-LENGTH CHARACTERISATION OF 247 DISTINCT ALLELES Gerhard Schöfl1, Kathrin Lang1, Marie Günther1, Grit Schober1, Carolin Massalski1, Alexander H. Schmidt2, Vinzenz Lange1 DKMS Life Science Lab, Dresden, Germany, 2DKMS, Tübingen, Germany Correspondence: [email protected]

1

Our current understanding of sequence variation in HLA-DPB1 is largely limited to exon 2, as only few alleles have been characterized at full-length. To close these gaps and better understand patterns of polymorphisms in the HLA-DPB1 gene, we selected samples from the DKMS donor pool that represent all common and well-documented (CWD) alleles and performed phased full-length sequence characterization. In addition we characterized alleles that are currently not defined as CWD but occur at moderate to high frequencies in the DKMS donor populations. Primers were developed flanking the UTR-regions of DPB1 resulting in a 12 kb amplicon. After long-range PCR

two redundant sequencing strategies were employed: Shotgun sequencing on Illumina MiSeq instruments and Single Molecule Real Time (SMRT) sequencing on PacBio® RS II instruments. Phase-defined consensus sequences were generated from the long-read data and polished with the high-fidelity short reads using the DR2S software (DKMS Life Science Lab). We successfully sequenced the full-length DPB1 sequences of 331 donors (662 haplotypes) covering 39 common, 14 well-documented, and 40 not-CWD-defined alleles. 247 distinct allelic variants were found if considering variation outside of exons 2 and 3. 12 previously known full-length DPB1 alleles were each confirmed at least once. A phylogenetic analysis of the 247 DPB1 alleles showed the existence of two major evolutionary branches characterized by two haplotype groups downstream of exon 2 with little internal variation and no evidence of historical recombination. These haplotype groups are tightly linked to the expression marker rs9277534 in the 3’ UTR. In conclusion, we present full-length sequences of 247 unique DPB1 alleles covering all common and all welldocumented (CWD) alleles. We observe distinct differences in recombination rates and patterns of nucleotide diversity/divergence within DPB1 indicating differential evolutionary forces acting on different parts of the gene.

Abstract

382

MO20

COLORING HLA TREES: BAYESIAN ESTIMATION OF HLA ASSOCIATIONS Christiaan H. van Dorp, Can Kesmir Utrecht University, Utrecht, the Netherlands Correspondence: [email protected] The human leukocyte antigen (HLA) is associated with many (infectious) disease outcomes. These associations are best documented for HIV-1, e.g. the HLA-B*81:01 allele is associated with control of the virus, while HLA-B*18:01 is considered detrimental. In finding these associations, it is often ignored that certain HLA alleles are very similar to other alleles. For instance, HLAB*15:03 differs only at 6 positions that interact with the peptide from HLA-B*18:01, and not surprisingly, HLA-B*15:03 is also associated with fast progression to AIDS. Here, we present a Bayesian method for finding HLA associations with quantitative traits such as HIV-1 virus load, that takes functional HLA similarities into account. The method is based on the phylogenetic mixed model, and can easily be modified to answer a wide range of research questions, like the role of the heterozygote advantage, or KIR ligands on disease outcomes. We show that for in the case of HIV-1, our model is significantly better at predicting set-point viral load than a model that ignores HLA similarities altogether. Furthermore, our method provides a comprehensible visualization of similarity-aware associations.

MO21

PREDICTING DISEASE COURSE FOR MULTIPLE SCLEROSIS PATIENTS: MACHINE LEARNING APPROACHES Antoine M. Lizée1, Rex Shihao1, Riley Bove1, Sergio E. Baranzini1, Pierre-Antoine Gourraud2 UCSF, San Francisco, USA, 2INSERM U1064 – CHU de Nantes, Nantes, France Correspondence: [email protected]

1

Multiple Sclerosis (MS) is a complex, chronic, autoimmune disease of the Central Nervous System affecting over 2.5 million people worldwide. MS phenotypes vary significantly across patients and over time. The large amounts of data required to investigate the various dimensions of the disease are challenging to integrate. Machine learning (ML) methods are emerging data analysis techniques that can help understand MS data and predict progression for individual patients. We used data collected from 2,977 visits of 484 patients from the 10-year-long UCSF EPIC cohort. Data included patient-reported outcomes, brain MRI metrics, demographics, and genetic information. Individuals’ Extended Disability Severity Scale (EDSS) at each visit was compared over time to classify disease progression as “worsening” or “stable” over two years. Several ML algorithms were compared: Bayesian classifiers, K-nearest neighbors, random

forests, regressions and support vector machines. Multiple combinations of data types were tested within a rigorous crossvalidation framework. The best performing algorithm and features achieved an AUROC of 0.76, suggesting that problemagnostic ML approaches perform better than traditional heuristic models. Performance varied greatly by data type: the EDSS or its breakdown along different components (MS Functional Scores) contributed significantly to performance (AUROC < 0.57 if not present, > 0.68 if present). Data on patient-reported quality of life and clinical tests for motor abilities best complemented this baseline information (+0.03 and +0.04 AUROC). Surprisingly, brain MRI metrics have limited contribution to prediction of clinical progression (