Biol Blood Marrow Transplant 22 (2016) 27e36

Biology of Blood andMarrow Transplantationjournal homepage: www.bbmt.org

Biology: Genetic Markers

Analysis of a Genetic Polymorphism in theCostimulatory Molecule TNFSF4 with HematopoieticStem Cell Transplant Outcomes

Peter T. Jindra 1, Susan E. Conway 1, Stacy M. Ricklefs 2, Stephen F. Porcella 2, Sarah L. Anzick 2,Mike Haagenson 3, Tao Wang 4, Stephen Spellman 3, Edgar Milford 1, Peter Kraft 5,David H. McDermott 6, Reza Abdi 1,*1 Transplant Research Center, Renal Division, Brigham and Women’s Hospital and Children’s Hospital, Boston, Massachusetts2Genomics Unit, Research Technologies Branch, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Hamilton, Montana3Center for International Blood and Marrow Transplant Research, Minneapolis, Minnesota4Division of Biostatistics, Medical College of Wisconsin, Milwaukee, Wisconsin5Departments of Epidemiology and Biostatistics, Harvard School of Public Health, Boston, Massachusetts6Molecular Signaling Section, Laboratory of Molecular Immunology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda,Maryland

Article history:

Received 28 April 2015Accepted 31 August 2015

Key Words:Gene polymorphismCostimulationHematopoietic stem cell

transplantationOutcomesInfectionGVHD

Financial disclosure: See Acknowle* Correspondence and reprint r

Avenue, Transplantation ResearchHarvard Medical School, Boston, M

E-mail address: rabdi@rics.bwh

http://dx.doi.org/10.1016/j.bbmt.201083-8791/� 2016 American Socie

a b s t r a c tDespite stringent procedures to secure the best HLA matching between donors and recipients, life-threatening complications continue to occur after hematopoietic stem cell transplantation (HSCT). Study-ing single nucleotide polymorphism (SNP) in genes encoding costimulatory molecules could help identifypatients at risk for post-HSCT complications. In a stepwise approach we selected SNPs in key costimulatorymolecules including CD274, CD40, CD154, CD28, and TNFSF4 and systematically analyzed their associationwith post-HSCT outcomes. Our discovery cohort analysis of 1157 HLA-A, -B, -C, -DRB1, and -DQB1 matchedcases found that patients with donors homozygous for the C variant of rs10912564 in TNFSF4 (48%) had betterdisease-free survival (P ¼ .029) and overall survival (P ¼ .009) with less treatment-related mortality (P ¼ .006).Our data demonstrate the TNFSF4C variant had a higher affinity for the nuclear transcription factor Myb andincreased percentage of TNFSF4-positive B cells after stimulation compared with CT or TT genotypes. How-ever, these associations were not validated in a more recent cohort, potentially because of changes in stan-dard of practice or absence of a true association. Given the discovery cohort, functional data, and importanceof TNFSF4 in infection clearance, TNFSF4C may associate with outcomes and warrants future studies.

� 2016 American Society for Blood and Marrow Transplantation.

INTRODUCTIONAlthough hematopoietic stem cell transplantation

(HSCT) has become the gold standard therapy for hemato-logic disorders and malignancies, the full therapeuticpotential of HSCT has been limited because of its compli-cations. Although the standardization of pretransplantdonorerecipient matching for HLAs has greatly improvedpost-HSCT outcomes, the mortality rate still remains twicethat of the general population even in patients who

dgments on page 35.equests: Reza Abdi, MD221 LongwoodCenter, Brigham and Women’s Hospital,A 02115..harvard.edu (R. Abdi).

15.08.037ty for Blood and Marrow Transplantation.

survived 2 or more years after allogenic HSCT [1]. Thisexcess mortality is caused by a number of complicationspost-transplantation such as graft-versus-host disease(GVHD), infection, and relapse of the primary disease [1,2].

There is growing evidence to support the importanceof genetic variability outside the HLA system that iscontributing to the heterogeneity in HSCT outcomes [3-5].Responses to alloantigen, tumor surveillance, and infectiouscomplications post-HSCT rely heavily on a functional im-mune system. Costimulatory molecules represent an essen-tial regulatory component of the immune system, whichmay be functionally affected by gene polymorphisms [5,6].For T cell activation 2 activation signals are required. The firstsignal occurs after the T cell receptor and a coreceptorinteract with the antigen peptide presented on an MHC

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e3628

molecule by an antigen-presenting cell. The second positivecostimulatory signal occurs when 1 or more T cell surfacereceptors engage with their specific ligands on antigen-presenting cells [7-10]. Conversely, negative costimulatorymolecules decrease T cell proliferation and cytokine pro-duction, promote T cell anergy or apoptosis, and induce theactivity of regulatory T cells [11].

For this study we selected single nucleotide poly-morphisms (SNPs) based on a stepwise criteria from cos-timulatory genes, which have an established association inhuman and animal HSCT outcomes [12-14]. SNPs needed tohave a strong linkage disequilibrium to additional taggedSNPs within their gene to maximize a subset of informativeSNPs. Each SNP required an adequate minor allelic frequencyof polymorphism above 5% to ensure the least commonvariant was present in our population. Finally, every SNPrequired enough subjects in our population to detect a sig-nificant association at 80% power. From these guidelines wecreated a panel of SNPs with a high likelihood of discoveringan association in our HSCT population. Determining novelpredictors could help develop a pretransplant matchingsystem to identify patients at risk and modify immunosup-pressive regimens based on their risk assessment.

The tumor necrosis factor ligand superfamily member 4(TNFSF4) and tumor necrosis factor receptor superfamilymember 4 (TNFRSF4) pathway represent key positive cos-timulatory signals required for cell activation. TNFRSF4 ispresent on both activated CD4þ and CD8þ T cells and itscognate ligand, TNFSF4, is expressed on dendritic cells, Bcells, and activated endothelial cells [15]. Signaling throughthe TNFSF4/TNFRSF4 pathway facilitates Th 2 differentiation,enhances effector CD8þ T cell memory commitment andpromotes cytokine production [16,17]. Gene polymorphismsin TNFSF4 have been associated with atherosclerosis andsystemic lupus erythematosus [18-20]. These studies postu-late that TNFSF4 is a major component in the T celleantigen-presenting cell interaction, leading to activation of immunecells to produce proinflammatory cytokines and chemokines,resulting in active disease. The role of TNFSF4 in determiningthe post-HSCT outcomes remains to be explored.

Infectious complications are a contributing source ofsevere morbidity and nonrelapse-related mortality in unre-lated donor allogeneic HSCT [21]. They account for a higherpercentage of mortality compared with GVHD in bothHLA-identical sibling and unrelated donor transplants stud-ied over a 5-year period [22]. Although the early prophylacticregimens reduce the incidence of early infection, the riskof late infection remains [23]. Deficiencies in the functionof immunoregulatory genes that activate the cellular andhumoral immune responses can be the underlying cause ofan increased risk of infection. As part of the immune system’sresponse to infection, activation of T cells through TNFSF4costimulation has been shown to effectively clear pathogens[15,24]. Genetic variation may influence the timing andstrength of TNFSF4 signaling to effectively respond to in-fectious pathogens.

In this study we carefully chose candidate SNPs foundwithin a group of extensively studied costimulatorymolecules that might associate with HSCT outcomes.We analyzed genetic data from discovery (n ¼ 1157) andvalidation (n ¼ 1188) cohorts using HLA-matched (at theHLA-A, -B, -C, -DRB1, and -DQB1 loci) HSCT recipients andtheir respective donors and then searched for associationswith important transplant outcomes.

METHODSPatient Population

A discovery cohort of 1157 and a validation cohort of 1188 recipi-entedonor pairs from unrelated HLA-A, -B, -C, -DRB1, and -DQB1 allele-matched transplantations facilitated by the National Marrow DonorProgram (NMDP) were included in the study. A detailed description can befound in Supplementary Methods. Patient data were acquired from theCenter for International Blood and Marrow Transplant Research, a researchaffiliation between the Medical College of Wisconsin, and the NMDP.Observational studies conducted by the Center for International Blood andMarrow Transplant Research were performed in compliance with thePrivacy Rule under the Health Insurance Portability and Accountability Actof 1966, as a Public Health Authority, and in compliance with all applicablefederal regulations pertaining to the protection of human research partici-pants and the Declaration of Helsinki as determined by continuous review ofthe Institutional Review Board of the NMDP.

Definition of OutcomeThe primary endpoints analyzed in the study were overall survival (OS),

disease-free survival (DFS), treatment-related mortality (TRM), relapse,acute GVHD (aGVHD) grades II to IV and III to IV occurring within the first100 days post-transplant, and chronic GVHD (cGVHD). Our analysis of OStreated death from any cause as the event, and surviving patients werecensored at the date of last contact. For analysis of DFS, failures were relapseor death from any cause with patients who were alive and in completeremission censored at time of last follow-up. TRM was defined as deathduring a continuous complete remission. Relapse was defined as clinical orhematologic relapse of primary disease with death without evidence ofdisease as a competing risk. For chronic myeloid leukemia patients ourdefinition of relapse included cytogenetic, molecular, and hematologicrelapse as an event. Assessment of aGVHD grades II to IV and III to IV weredefined using the Glucksberg scale [25], and extensive cGVHD was definedaccording to the Seattle criteria [26].

GenotypingWe genotyped 9 SNPs located in 5 immunoregulatory genes: CD274,

CD40, CD154, CD28, and TNFSF4 (Supplementary Table 1). These SNPs wereselected based on strong linkage disequilibrium to tagged SNPs withintheir gene and power calculations suggested in our population a high likeli-hood of discovering an association. For the discovery set, genomic DNA wasisolated from cryopreserved leukocytes from donor and recipient bloodsamples provided by the NMDP Research Sample Repository followingthe manufacturer’s protocol (Promega, MAdison, WI). Isolated DNA wasquantitated using a Nanodrop (Thermo Scientific, Waltham, MA) and wholegenome amplified with Phi29 DNA polymerase. For the validation set,genomic DNA was isolated from frozen blood samples using the QIAmp 96DNA Blood Kit and the manufacturer’s protocol (Qiagen, Valencia, CA). Gen-otyping was then performed using a Taqman SNP genotyping assay (AppliedBiosystems, Foster City, CA) and automated genotype calling software. Sam-ples in the discovery (n ¼ 213) and validation (n ¼ 12) cohorts where a SNPgenotype was not obtained due to DNA degradation or minimal isolatedgenomic DNA were excluded from further analysis.

SNP Selection and Identification of Tag SNPsSNPs were investigated using the Hapmap Genome Browser Phase 1 and

2 full dataset (International Hapmap Project, http://www.hapmap.org/).Gene SNP data from the white population was imported into Haploview(http://www.broad.mit.edu/mpg/haploview/) using the Hapmap formatfunction. The uploaded SNP data were further modeled by setting thefollowing Haploview parameters: Hardy-Weinberg P cutoff of .001, a mini-mum genotype percent of 80, a maximum number Mendel error of 1, and aminimumminor allele frequency of .05. The minor allele frequency cutoff of5% was selected for compatibility with HapMap, which includes only thoseSNPs with allele frequencies greater than .05. The tagger function in Hap-loview was performed using pair-wise tagging only with an r2 equal to .8.Identified tag SNPs and the corresponding predicted common SNPs werecompiled.

Statistical AnalysisDiscovery set

Univariate probabilities of DFS and OS were calculated using the Kaplan-Meier method. The log-rank test was used for comparing survival curves.Probabilities of TRM, relapse, aGVHD, and cGVHD were calculated usingcumulative incidence estimates. The cumulative incidence calculated foraGVHD and cGVHD treated death as a competing risk [27]. Relapse wastreated as a competing risk for TRM and vice versa.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e36 29

Multivariate analyses of OS, DFS, TRM, relapse, aGVHD, and cGVHDwereperformed using the Cox proportional hazards regression models with amultivariate analysis performed to identify clinical variables that wereassociated with particular outcomes. Clinical variables considered in themodels include disease, disease status, conditioning regimen, cytomegalo-virus match, Karnofsky score, graft type, GVHD prophylaxis, year of trans-plant, recipient age, donor age, and sex. For each outcome, all variableswere examined for proportional hazards. Factors violating the proportionalhazards assumption were adjusted by stratification. The stepwise modelbuilding approach was then used in developing models for all the outcomes.OS was adjusted for disease, disease stage, conditioning regimen, patientage, and GVHD prophylaxis and stratified by graft type and Karnofsky score.DFS was adjusted for disease, disease stage, cytomegalovirus match, patientage, and conditioning regimen and stratified by disease type and Karnofskyscore. Relapse was adjusted for disease, disease stage, conditioning regimen,and donor sex. TRM was adjusted for disease stage, conditioning regimen,and patient age and stratified by disease, graft type, and Karnofsky score.aGVHD grades II to IV was adjusted for disease, disease stage, graft type, andGVHD prophylaxis. aGVHD grades III to IV was adjusted for disease, diseasestage, conditioning regimen, and year of transplant. cGVHDwas adjusted fordisease and patient age and stratified by GVHD prophylaxis.

Potential interactions between the genetic variants and clinical cova-riates were tested and were not present. Because of the possibility of Type Ierror occurring from multiple comparisons, the genetic associations werereferred to as statistically significant when P � .01. SAS software, version 9.2(SAS Institute, Cary, NC), was used in all the analyses.

Weused apriori power calculations todetermine the powerof our cohort(1157 donorerecipient pairs) to detect significant associations between eachgeneSNPandHSCToutcomes. Setting the odds ratio to equal 1.5 for each geneSNP with minor allelic frequency � 5%, assuming a Type 1 error level of .05,calculations were performed using the Quanto software (http://hydra.us-c.edu/GxE/). Using our cohort we calculated the minimum number of in-dividuals required to attain 80% power in our study to identify significantassociations between HSCT outcomes and 9 costimulatory gene SNPs.

For additional statistical analyses, the association between TNFSF4 SNPrs10912564 genotypes and causes of death were performed using a chi-square test, and TNFSF4 in vitro data were analyzed by Student’s t-test.Results were considered statistically significant when P � .05.

Validation setTo validate the SNP rs10912564 observations in the discovery cohort, we

assembled an independent cohort of 1200 unrelated donorerecipient pairsusing the same covariates from the discovery cohort.

DNA Protein Interaction ELISADNA protein interaction ELISA has been described previously [28].

Briefly, we created 50 biotin linked 30-mer double-stranded DNA oligomersequences corresponding to the rs10912564 region of TNFSF4 gene withrs10912564C/T located in the center position in bold (50-TAGCAGTATGTTAACGGAAGCATGTTCATG-30) and (50-TAGCAGTATGTTAATGGAAGCATGTTCATG-30) (Integrated DNA Technologies, Coralville, IA). Ninety-sixewellplates were coated with streptavidin (15520; Pierce Biotechnologies, Rock-ford, IL) and incubated with 2 pmol DNA oligo probes for 1 hour at 37�C,blocked with 5% milk for 30 minutes, and incubated with 5 mg of Jurkat cellnuclear extract (36110; Active Motif, Carlsbad, CA). After 1 hour c-myb wasdetected by primary c-myb antibody (1:500, D-7, sc-74512) and anti-mousehorseradish peroxidase linked secondary (1:1500, sc-2031) antibody (SantaCruz Biotechnologies, Santa Cruz, CA). After each step thewells werewashed3 times with PBS Tween-20. An o-phenylenedianime dihydrochloride sub-strate (P9187; Sigma, St. Louis, MO) produced a peroxidase reaction that wasquantitated by measuring absorbance at 450 nm on an ELISA reader.

Flow CytometryPeripheral blood mononuclear cells (available from the NMDP Research

Repository for in vitro studies) from 12 HSCT donors were tested with 4 eachfrom the CC, CT, and TT rs10912564 genotypes. Cells were plated in a 96-wellplate at .25 � 106 and were stimulated with 10 mg/mL F(ab’)2 anti-humanIgM Fc5m (Jackson Immunoresearch, West Grove, PA) and 5 mg/mL CD40(S2C6) (Mabtech, Cincinnati, OH) monoclonal antibody for 72 hours shownto upregulate TNFSF4 and c-myb in B cells [29,30]. Cells were stained for30 minutes on ice with CD19-FITC (HIB19), TNFSF4-PE (11C3.1) (Biolegend,San Diego, CA), and 7AAD (BD Biosciences, San Jose, CA) and run on aFACScan flow cytometer (BD Biosciences). A PE mouse IgG1 kappa MOPC-21isotype control (Biolegend) was used to estimate nonspecific binding of theTNFSF4-PE antibody. Dead cells were excluded by 7AAD staining, and aforward and side scatter lymphocyte gate was applied. The percentage ofTNFSFþCD19þ high cells was analyzed using FloJo version 10 (Tree Star,Ashland, OR) and Prism version 4.0a (GraphPad Software, Inc., La Jolla, CA).

RESULTSGenotyping of Costimulatory Molecule Genetic Variants

Costimulatory molecule SNPs were genotyped usingautomated technology in discovery (n¼ 1370) and validation(n ¼ 1200) cohorts totaling 2570 recipients and theirrespective donors. Genotyping was successful for 84.4%(n¼ 1157) and 99% (n¼ 1188) in our discovery and validationcohorts, respectively. The allele frequencies observed in ourpredominately white transplant population were compara-blewith those obtained in the Hapmap database phase IIþ IIIon NCBI B36 assembly, dbSNP b126 [31]. We found eachgenetic variant to be in Hardy-Weinberg equilibrium, con-firming no evolutionary selection and validating the accu-racy of our genotyping (Supplementary Table 1). Focusing onthe rs10912564 TNFSF4 genotypes, the characteristics (age attransplant, Karnofsky score, sex, disease, stage, etc.) of C/Crecipients shown in Table 1 were not significantly differentcompared with the general population tested by thechi-square test.

Outcome Analysis by Donor and Recipient CostimulatoryMolecule Genotypes

Initially, we examined the association between cos-timulatory molecule SNPs and transplant outcomes in ourdiscovery cohort. We found a significant association withincreased OS (hazard ratio, .823; 95% confidence interval,.718 to .953; P ¼ .0085; Table 2, Figure 1) in donors carryingthe TNFSF4 CC genotype compared with TT or CT. Interest-ingly, we found a significant decrease in TRM (hazard ratio,.778; 95% confidence interval, .650 to .931; P ¼ .006; Table 2,Figure 1) in CC rs10912564 genotyped donors. When TRMwas analyzed using T cell depletion status we observedsimilar hazard ratios, suggesting T cell depletion status didnot affect TRM (Supplementary Table 2). There was no as-sociation between rs10912564 with other HSCT outcomessuch as relapse, aGVHD, or cGVHD (Table 2, Figure 1). SNPs ingenes CD274 and CD40 of recipients were associated withHSCT outcomes, but this requires additional validationtesting (Supplementary Table 3).

To determine if SNP rs10912564 had any effect onmortality in our patient population, we analyzed causes ofdeath comparing patients transplanted with rs10912564 CCor TT/TC genotyped donors (Supplementary Table 4). Wefound a significant increase in mortality due to infection inthe TT/TC group compared with CC in our discovery cohort(P ¼ .014). This association was present in patients with noGVHD and in contrast was absent in the GVHD cohort(Supplementary Table 5). A cause of death analysis is diffi-cult to define a specific variable, and these results requirefurther investigation into the rate of infection. Although alarger and well-designed study is needed to further supportthese data, TNFSF4 has been shown to be critically requiredto maintain T cell responses during persistent viralinfections [32].

To validate the associations observed with TNFSF4SNP rs10912564, we identified a much more recent cohortavailable to us through the NMDP. We found no associationbetween SNP rs10912564 and transplant outcomes or causeof death in this validation cohort (Table 2 and SupplementaryTable 6). Not surprisingly, we noted a number of contributingfactors that were significantly different between the cohorts,including age, disease type, graft type, GVHD prophylaxis,donor age, and year of transplant (Table 3). These factorscould potentially impact the clinical relevance of SNP

Table 1Characteristics of Patients Receiving a First Myeloablative Transplant forAML, ALL, CML, or MDS analyzed by rs10912564 SNP Donor Typing in theDiscovery Cohort*

Characteristic TT þ CT CC P

Total number of patients 606 551Number of centers 97 96Median age, yr (range) 38 (<1-64) 36 (1-65) .93Age at transplant, yr .680-9 47 (8) 38 (7)10-19 62 (10) 59 (11)20-29 94 (16) 96 (17)30-39 165 (27) 135 (25)40-49 161 (27) 140 (25)50 and older 77 (13) 83 (15)

Male sex 352 (58) 317 (58) .85Karnofsky score before transplant >90 442 (73) 393 (71) .22Disease at transplant .74AML 160 (26) 131 (24)ALL 114 (19) 111 (20)CML 237 (39) 224 (41)MDS 95 (16) 85 (15)

Disease status at transplant .62Early 285 (47) 244 (44)Intermediate 137 (23) 142 (26)Advanced 154 (25) 136 (25)Other MDS 30 (5) 29 (5)

Graft type .44Bone marrow 551 (91) 508 (92)Peripheral blood 55 (9) 43 (8)

Conditioning regimen .75Cy/TBI � other 464 (77) 415 (75)TBI � other 25 (4) 20 (4)Bu/Cy � other 98 (16) 103 (19)Bu þ melphalan � other 12 (2) 8 (1)Other 7 (1) 5 (1)

GVHD prophylaxis .80Cyclosporine þ MTX � other 354 (58) 334 (61)Cyclosporine � other (no MTX) 18 (3) 21 (4)Tacrolimus � other 124 (20) 107 (19)MTX � other (no cyclosporine) 7 (1) 5 (1)Ex vivo T cell depletion 100 (17) 83 (15)Other 3 (<1) 1 (<1)

Donorerecipient sex match .98Maleemale 244 (40) 223 (40)Maleefemale 142 (23) 128 (23)Femaleemale 108 (18) 94 (17)Femaleefemale 112 (18) 106 (19)

Donorerecipient CMV match .91Negativeenegative 215 (35) 205 (37)Negativeepositive 166 (27) 151 (27)Positiveenegative 105 (17) 93 (17)Positiveepositive 103 (17) 84 (15)Unknown 17 (3) 18 (3)

Median donor age, yr (range) 37 (18-60) 35 (18-60) .20Donor age, yr .5518-19 6 (1) 4 (1)20-29 150 (25) 150 (27)30-39 226 (37) 215 (39)40-49 183 (30) 143 (26)50 and older 41 (7) 39 (7)

Year of transplant .151990 11 (2) 17 (3)1991 16 (3) 21 (4)1992 35 (6) 21 (4)1993 30 (5) 26 (5)1994 56 (9) 37 (7)1995 47 (8) 37 (7)1996 43 (7) 36 (7)1997 48 (8) 46 (8)

(Continued)

Table 1(continued)

Characteristic TT þ CT CC P

1998 70 (12) 54 (10)1999 46 (8) 69 (13)2000 87 (14) 83 (15)2001 83 (14) 69 (13)2002 34 (6) 35 (6)

Median follow-up of survivors, mo(range)

95 (3-194) 96 (22-196) .12y

AML indicates acute myeloid leukemia; ALL, acute lymphoblastic leukemia;CML, chronic myeloid leukemia; MDS, myelodysplastic syndromes; Cy,cyclophosphamide; TBI, total body irradiation; BU, busulfan; MTX, metho-trexate; CMV, cytomegalovirus.Values are total number of cases with percents in parentheses, unlessotherwise noted.

* Data were Corrective action plan (CAP) -modeled.y Log-rank P value.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e3630

rs10912564 TNFSF4 and transplant outcomes. We furtherperformed a subset analysis of bone marrow graft typecomparing OS, DFS, and TRM outcomes in the validationcohort. There was no association, whether the model was

Table 2Discovery and Validation Cohort Multivariate Analysis of Donor rs10912564Genotypes TT þ CT versus CC

Outcome rs10912564Genotype

n HazardRatio

95% ConfidenceInterval

P

OSDiscovery TT þ CT 606

CC 551 .823 (.718-.953) .0085Validation TT þ CT 614

CC 571 1.036 (.893-1.202) .6405DFSDiscovery TT þ CT 606

CC 551 .855 (.743-.984) .0291Validation TT þ CT 604

CC 561 1.059 (.915-1.226) .4408TRMDiscovery TT þ CT 606

CC 551 .778 (.65-.931) .006Validation TT þ CT 606

CC 561 1.013 (.823-1.247) .9039RelapseDiscovery TT þ CT 606

CC 551 .979 (.777-1.233) .857Validation TT þ CT 604

CC 561 1.083 (.883-1.329) .4448aGVHD grades

II-IVDiscovery TT þ CT 606

CC 551 .903 (.769-1.06) .2117Validation TT þ CT 614

CC 572 .995 (.841-1.177) .951aGVHD grades

III-IVDiscovery TT þ CT 606

CC 551 1.062 (.838-1.346) .618Validation TT þ CT 606

CC 564 .948 (.731-1.230) .6888cGVHDDiscovery TT þ CT 606

CC 551 .931 (.777-1.114) .4339Validation TT þ CT 600

CC 557 .986 (.838-1.160) .8648

Comparison for all categories is (TT þ CT) vs. CC. Covariates used includeddisease, disease status, conditioning regimen, CMV match, Karnofsky score,graft type, GVHD prophylaxis, year of transplant, recipient age, donor age,and sex.

Figure 1. Donor TNFSF4 rs10912564 C/C genotype versus C/T and T/T in a discovery cohort. Kaplan-Meier curves for (A) DFS, (B) patient OS, (C) cumulative incidenceof TRM, and (D) cumulative incidence of relapse in the 10 years after transplantation. (E) Cumulative incidence curves for aGVHD grades II to IV (E) and III to IV (F) inthe first 100 days after transplantation.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e36 31

adjusted for covariates identified from the discovery(all P > .44) or validation (all P > .46) cohorts.

Analysis of Linkage of Genetic Variants in TNFSF4In the TNFSF4 gene encoded on chromosome 1, we iden-

tified rs10912564C/T as a potential SNP marker based on itsminor genotypic frequency TT > .05 within the white pop-ulation and ability to tag in linkage 5 other SNPs (rs7525284,rs4081545, rs3861950, rs11811856, and rs7514229) eachlocated in the first intron of TNFSF4, except rs7514229, whichis in the 30 untranslated region (Figure 2). Because of thestrong linkage disequilibrium, the haplotype created by the6 SNPs had a minor allelic frequency (29.4%) in whites from

the CEU Hapmap population that was similar to the fre-quency of the rs10912564 T allele (31.8%) in our patientpopulation. Each tagged TNFSF4 SNP was scored for func-tional relevance using the F-SNP bioinformatics algorithmwhich cross-references 16 bioinformatics tools and data-bases to generate a SNP score on a 0 to 1 scale where highscoring SNPs are known to be disease associated [33,34]. Thealgorithm used TFSearch version 1 [3] to score potentialtranscription factor binding sites in the C and T rs10912564SNP variant sequence with a threshold of 85. A score of90.9 was observed between the C variant and myeloblastosisviral oncogene homolog (v-myb), which was not present inthe T variant. The F-SNP readout indicated v-myb was the

Table 3Characteristics of Patients Receiving Myeloablative Transplants for AML,ALL, CML, or MDS Reported to the NMDP for the Discovery and ValidationCohorts

Variable DiscoveryCohort

ValidationCohort

P

Number of patients 1301 1200Number of centers 119 125Median age, yr (range) 36 (<1-65) 38 (<1-70) .003Age at transplant, yr <.0001<1-9 99 (8) 90 (8)10-19 133 (10) 152 (13)20-29 206 (16) 212 (18)30-39 337 (26) 186 (16)40-49 344 (26) 263 (22)50 and older 182 (14) 297 (25)

Recipient raceWhite N/A 1059 (90)African American N/A 31 (3)Asian/Pacific Islander N/A 12 (1)Hispanic N/A 66 (5)Native American N/A 3 (<1)Other/missing/unknown/TBD N/A 29

Male sex 747 (57) 673 (56) .5Karnofsky score before

transplant.51

<90 321 (25) 280 (23)�90 936 (72) 767 (64)Missing/unknown 44 153

Disease <.0001AML 326 (25) 571 (48)ALL 251 (19) 350 (29)CML 526 (40) 175 (15)MDS 198 (15) 104 (9)

Disease stageEarly 589 (45) 500 (42)Intermediate 325 (25) 377 (31)Advanced/Late 323 (25) 323 (27)Other 64 (5) 0

Graft Type <.0001Marrow 1195 (92) 490 (41)PBSC 106 (8) 710 (59)

Donor SNP rs10912564 .88TT 111 (10) 119 (10)CT 495 (43) 497 (42)CC 551 (48) 572 (48)Missing/Unknown 144 12

Donorerecipient sex match .53Maleemale 527 (41) 475 (40)Maleefemale 309 (24) 315 (26)Femaleemale 220 (17) 198 (17)Femaleefemale 245 (19) 212 (18)

Donorerecipient CMV match <.0001Donor�/recipient� 477 (37) 387 (32)Donor�/recipientþ 357 (27) 440 (37)Donorþ/recipient� 215 (17) 140 (12)Donorþ/recipientþ 215 (17) 213 (18)Unknown 37 (3) 20 (2)

GVHD prophylaxis <.0001FK506 þ (MTX or MMF or

steroids) � other281 (22) 681 (57)

FK506 � other 3 (<1) 72 (6)CSA þ MTX � other 760 (58) 288 (24)CSA � other 38 (3) 53 (4)MTX � other 4 (<1) 0Ex vivo T cell depletion 212 (16) 83 (7)Other 3 (<1) 23 (2)

Median donor age, yr (range) 36 (18-60) 33 (18-60) <.0001Donor age in decades <.000118-19 12 (1) 18 (2)20-29 336 (26) 410 (34)

(Continued)

Table 3(continued)

Variable DiscoveryCohort

ValidationCohort

P

30-39 505 (39) 449 (37)40-49 354 (27) 265 (22)50 and older 94 (7) 58 (5)

Donor raceWhite N/A 1033 (89)African American N/A 30 (3)Asian/Pacific Islander N/A 11 (1)Hispanic N/A 17 (1)Native American N/A 13 (1)Other/declined/unknown N/A 62 (5)Missing/TBD N/A 34

Year of transplant <.00011990 30 (2) 01991 46 (4) 01992 58 (4) 01993 61 (5) 01994 105 (8) 01995 98 (8) 01996 86 (7) 01997 118 (9) 01998 144 (11) 01999 137 (11) 02000 181 (14) 02001 161 (12) 02002 76 (6) 02003 0 165 (14)2004 0 248 (21)2005 0 324 (27)2006 0 354 (30)2007 0 109 (9)

Median follow-up of survivors,mo (range)

96 (3-196) 71 (11-109) <.0001*

N/A indicates not available; TBD, to be determined; PBSC, peripheral bloodstem cell; FK506, tacrolimus; MMF, mycophenolate mofetil; CSA, cyclo-sporine.Values are number of cases with percents in parentheses, unless otherwisenoted.

* Log-rank P value.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e3632

only transcription factor that differed between the 2 SNPrs10912564 variants. F-SNP analysis of the SNPs tagged withrs10912564 showed rs4081545 as a potential binding site fornuclear factor-1 and rs7514229 as a site for caudal typehomeobox 1 (Cdx-1) with scores above threshold (Figure 2).

Functional Relevance of TNFSF4 SNP rs10912564To determine whether rs10912564 had any effect on the

function of TNFSF4, we initially used the SNP bioinformaticsdatabase F-SNP to search for potential functional changes inprotein coding, splicing regulation, transcriptional regula-tion, transcription regulation, or post-translation involvingSNP rs10912564. We found rs10912564 was located within apotential Myb binding site. To further explore the possibilitythat the rs10912564 T/C variant could augment the bindingof transcription factor c-myb, we created a DNA proteininteraction ELISA assay to assess the ability of c-myb to bind abiotinylated 30mer sequence of TNFSF4 containing either theT or C rs10912564 variant. We used Jurkat cells shown toexpress c-myb in nuclear extracts as our source of c-mybprotein [35]. Upon comparison, we found c-myb boundsignificantly stronger to the C biotinylated probe comparedwith the T variant (Figure 3A, P ¼ .0024). To check thespecificity of the assay we performed a competition experi-ment. As shown in Figure 3B, adding excess amount of

Figure 2. TNFSF4 gene organization as depicted in the NCBI database and analysis of SNPs in linkage with rs10912564. The TNFSF4 gene is on chromosome 1q25 andis a 3-exon, 2-intron gene. Filled boxes denote exons, unfilled boxes indicate untranslated regions, and solid lines between boxes represent introns in the TNFSF4 genediagram. The relative positions of the TNFSF4 variants tagged with rs10912564 are indicated along with their percent of linkage disequilibrium calculated in Hap-loview. A haplotype frequency is noted for the linked major alleles compared with minor alleles. The assessment of each SNPs functional relevance was calculatedusing F-SNP bioinfomatics algorithm. Potential transcription factor binding sites altered by the SNP polymorphism are listed.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e36 33

nonbiotin labeled probe was able abrogate c-myb bindingto the plate bound rs10912564C probe. To functionallydetermine if these variants effect cellular responses, wecollected peripheral blood mononuclear cells from 4 donorsof each rs10912564 genotype and stimulated themwith IgMand anti-CD40 antibodies for 72 hours and assessed theirability to upregulate TNFSF4 on B cells. Stimulation led to anincrease in the percentage of TNFSF4þCD19þ B cells in donorscarrying the CC genotype compared with CT and TT, sug-gesting the rs10912564C allele may augment the expressionlevel of TNFSF4 (Figure 4, P ¼ .049). These data highlight thepotential functional capacity of SNP rs10912564 to regulatetranscription factor binding to TNFSF4 and the expressionlevel on the cell surface of B cells.

DISCUSSIONEfforts to identify risk factors such as HLA mismatch

have had a significant impact on improving the outcomes ofHSCT [36,37]. These studies allowed for better matching ofrecipients and donors. However, the rate of post-HSCTcomplications remains high. A main focus has been toinvestigate the importance of genetic polymorphisms inimmunoregulatory genes in determining HSCT outcomespost-transplant [5,38-40]. Virtually all post-HSCT complica-tions are immune response driven; therefore, we focused onthe role of costimulatory molecules by examining the asso-ciation between 9 SNPs in 5 costimulatory genes and HSCToutcomes in unrelated HLA-matched donorerecipient pairs.

These costimulatory molecules play a key role in deter-mining the outcome of immune responses [41-43]. Ourdataset of 2570 donorerecipient pairs collected from 2 in-dependent cohorts represents 1 of the largest SNP studiesassembled focused on TNFSF4, which is expressed on B cells,vascular endothelial cells, natural killer cells, and mast cellsand provides clonal expansion of antigen-specific T cells,expansion of regulatory T cells, and generation of T cellmemory [15,44].

The TNFSF4 rs10912564CC genotype of donors was asso-ciated with a higher likelihood of OS with less TRM in ourdiscovery cohort. This association between rs10912564CCand less TRM may reflect the costimulatory capacity of thereconstituted immune system by the donor leukocytes [9]. Insupport of our discovery findings, it was demonstrated in acohort of patients that low responders to hepatitis B vacci-nation were more likely to have the rs10912564 T allele [45].The effect of TNFSF4 on infection clearance has been heavilysupported in a number of studies. For instance, prolongedimmunity has been observed against mycobacterium tuber-culosis when vaccination is delivered in conjunction with aTNFSF4:Ig fusion protein, generating a strong Th1 cytokineresponse [16,46,47]. Furthermore, T cells lacking TNFRSF4 failto accumulate in response to a chronic viral infection that isassociated with decreased survival proteins Bcl-2 and Bcl-xL[32]. In future studies a correlation between infection ratewith genetic variants in marrow grafts could identify in-dividuals at higher risk for infection after HSCT and would

Figure 3. Specific binding of c-myb to a 30-oligomer sequence of TNFSF4containing the rs10912564 polymorphism. (A) Binding of c-myb to a 30-meroligonucleotide probe containing either the rs10912564 C or T variant in thecenter. Binding was measured as a function of absorbance at 450 nm after aperoxidase reaction. Data represent 3 independent experiments. (B) Thespecific binding of c-myb to the rs10912564C probe was shown by a compe-tition experiment with nonbiotinylated dsDNA. Different concentrations ofprobe (100, 1000 pmol) were added to Jurkat nuclear extract and incubatedon an ELISA plate coated with 2 pmol of biotinylated rs10912564C probeand absorbance read at 450 nm after a peroxidase reaction. Data represent2 independent experiments. B indicates blank; Bio, biotinylated; c-myb,myeloblastosis viral oncogene homolog.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e3634

allow physicians to create a more stringent plan for moni-toring and treating this subset of patients to optimize clinicaloutcomes.

Because we evaluated a clinical association betweenrs10912564 genotypes and HSCT outcomes, we strived todevelop a possible mechanism of action. Functional SNPs areknown to mainly interact in gene promoter regions; how-ever, noncoding intronic sequences may contain functionalSNPs that can enhance the transcriptional level of genesleading to disease associations [48,49]. Regulatory elementslocated in the first intron have been identified as sitesof transcriptional regulation [50]. Examination of SNPrs10912564 using bioinformatics tools helped us to pinpointthe c-myb binding site consensus sequence as a possiblefunctionally relevant site inwhich the C nucleotide would bepredicted to be a critical base required for c-myb binding[51]. We speculate that rs10912564C creates a stable bindingsite for c-myb that could potentially augment the expressionof TNFSF4 in comparison with the T allele. Many genes areupregulated through the binding of c-myb, which is inducedafter activation of T and B cells and is required for the dif-ferentiation and survival of the B cell lineage [52-54]. Thelocation of the rs10912564 polymorphism in the intron

1 region suggests it affects the creation of a transactivatorenhancer or interacts with regulatory elements required forgene expression.

This study has strengths and limitations. First, thisstudy represents 1 of the largest non-HLA genetic associationstudies using a well-defined discovery cohort of 1157donorerecipient pairs. We meticulously chose SNPs incandidate immunoregulatory genes, which were known tohave pivotal roles in various outcomes post-transplantation.The main limitation in this study was the inability to repli-cate our association data in a more recent validation cohortwhere the significance of TNFSF4 SNP rs10912564 wasnegligible, suggesting the absence of a true association orindicative of a small size effect. The latter is an importantaspect of gene association studies for complex diseases,because most individual disease-associated SNPs exhibitmodest effects on relative risk. A typed SNP may serve as animperfect surrogate for the true causal SNP, suggesting a jointeffect of multiple SNPs. Multiple comparisons in clinicaloutcome-driven SNP studies might require a more stringentP value; however, our TRM P value (.006) met rigidityrequired to adjust for multiple comparisons. Furthermore,we tried to reduce the burden of multiple comparisons byhaving high statistical power with prior literature supportedknowledge of the effect these candidate genes had on HSCToutcomes. If the ever-evolving field of clinical bone marrowtransplantation obscured the effect of our SNP, it sheds lighton a challenge for genetic association studies in this field andvirtually for all other fields that therapies and standard ofpractice change over time. Furthermore, the actual contri-bution of a SNP to the biology of a gene and novel insightsdrawn from the data, even obscured, facilitates better un-derstanding of the mechanisms by which the gene mightaffect disease.

The lack of validation could also be due to the significantdifferences between the discovery and validation cohorts.The validation cohort included an increased number of olderpatients than the discovery cohort, and this older patientpopulation was associated with more acute leukemias.Furthermore, the newer validation cohort had access to amuch younger donor population, highlighting improvementin donor options via registry growth in recent years. Addi-tionally, the validation group benefited from recent changesin HSCT technique in that graft types have transitioned toperipheral blood stem cell transplants. Finally, the validationcohort used more FK506/tacrolimus for GVHD prophylaxis,which has been associated with less aGVHD and better out-comes compared with the cyclosporine Aebased treatmentsused in the discovery cohort [55]. Moreover, ex vivo T celldepletion was used in the discovery cohort and is associatedwith poor engraftment and higher relapse rates [56].Conversely, the validation cohort had little to no ex vivo T celldepletion. Although we made an effort to incorporate thedifferences between the discovery and validation cohorts inour analysis, we cannot completely rule out that they did notsignificantly impact the effect of SNP rs10912564 on HSCToutcomes.

In conclusion, although we found associations betweenTNFSF SNP rs10912564 and HSCT outcomes in our discoverycohort and were able to show the functional contribution ofthe SNP, our findings were not validated in a more recentcohort. We believe a larger prospective study is warranted toexamine the influence of TNFSF4 genetic variants on post-transplant outcomes as well as indepth functional studieson specific subsets of immune cells to determine the

Figure 4. TNFSF4 surface expression on CD19þ B cells from HSCT donor peripheral blood mononuclear cells. Peripheral blood mononuclear cells (n ¼ 12) from eachTNFSF4 rs10912564 genotype (n ¼ 4) were stimulated with 10 mg/mL IgM and 5 mg/mL anti-CD40 antibodies for 72 hours. (A) Overlaid histograms were normalizedand depicted as the % of max for the CC (i), TC (ii), and TT (iii) genotypes. Black histograms indicate TNFSF4 staining of stimulated B cells, and the gray histogramsindicate TNFSF4 staining of unstimulated B cells. A negative baseline cutoff was created based on isotype staining demarcated where the negative and positivehistogram markers overlap. (B) The percentage of TNFSF4 positive B cells was calculated in the presence and absence of stimulation. The average difference in thepercentage of TNFSF4 positive B cells with and without stimulation for each genotype was plotted with standard error of mean.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e36 35

potential impact of TNFSF4 genetic variants regulating theirfunction.

ACKNOWLEDGMENTSFinancial disclosure: This work was partly supported by

the Division of Intramural Research of the National Instituteof Allergy and Infectious Diseases, National Institutes ofHealth, Bethesda, Maryland. The Center for InternationalBlood and Marrow Transplant Research is supported byPublic Health Service Grant/Cooperative Agreement U24-CA76518 from the National Cancer Institute (NCI), theNational Heart, Lung and Blood Institute (NHLBI), and theNational Institute of Allergy and Infectious Diseases; Grant/Cooperative Agreement 5U01HL069294 from the NHLBI andNCI; a contract (HHSH234200637015C) with the HealthResources and Services Administration; 2 grants (N00014-12-1-0142 and N00014-13-1-0039) from the Office of NavalResearch; and grants from Allos, Inc.; Amgen, Inc.; Angio-blast; Anonymous donation to the Medical College ofWisconsin; Ariad; Be the Match Foundation; Blue Cross andBlue Shield Association; Buchanan Family Foundation; Cari-dian BCT; Celgene Corporation; CellGenix, GmbH; Children’sLeukemia Research Association; Fresenius-Biotech NorthAmerica, Inc.; Gamida Cell Teva Joint Venture Ltd.; Gen-entech, Inc.; Genzyme Corporation; GlaxoSmithKline; His-toGenetics, Inc.; Kiadis Pharma; The Leukemia & LymphomaSociety; The Medical College of Wisconsin; Merck & Co, Inc.;Millennium: The Takeda Oncology Co.; Milliman USA, Inc.;Miltenyi Biotec, Inc.; National Marrow Donor Program;Optum Healthcare Solutions, Inc.; Osiris Therapeutics, Inc.;Otsuka America Pharmaceutical, Inc.; RemedyMD; Sanofi;

Seattle Genetics; Sigma-Tau Pharmaceuticals; Soligenix, Inc.;StemCyte, A Global Cord Blood Therapeutics Co.; StemsoftSoftware, Inc.; Swedish Orphan Biovitrum; Tarix Pharma-ceuticals; Teva Neuroscience, Inc.; THERAKOS, Inc.; andWellpoint, Inc. The content of this publication does notnecessarily reflect the views or policies of the US Departmentof Health and Human Services and mention of trade names,commercial products, or organizations does not implyendorsement by the US Government.

Conflict of interest statement: There are no conflicts of in-terest to report.

Authorship statement: Research was designed by P.T.J.,S.E.C., E.M., P.K., D.H.M., and R.A. Samples were genotyped byS.M.R., S.L.A., and S.F.P. Data generated for tables and figureswas done by P.T.J., S.E.C., S.S., and T.W. Statistical data analysiswas done by T.W., M.H., and S.S. All the authors contributedto the writing of the manuscript and analysis of the data.

SUPPLEMENTARY DATASupplementary data related to this article can be found

online at 10.1016/j.bbmt.2015.08.037.

REFERENCES1. Bhatia S, Francisco L, Carter A, et al. Late mortality after allogeneic

hematopoietic cell transplantation and functional status of long-termsurvivors: report from the Bone Marrow Transplant Survivor Study.Blood. 2007;110:3784-3792.

2. Bieri S, Roosnek E, Ozsahin H, et al. Outcome and risk factors for late-onset complications 24 months beyond allogeneic hematopoieticstem cell transplantation. Eur J Haematol. 2011;87:138-147.

3. Hansen JA, Chien JW, Warren EH, et al. Defining genetic risk for graft-versus-host disease and mortality following allogeneic hematopoieticstem cell transplantation. Curr Opin Hematol. 2010;17:483-492.

P.T. Jindra et al. / Biol Blood Marrow Transplant 22 (2016) 27e3636

4. Dickinson AM, Holler E. Polymorphisms of cytokine and innateimmunity genes and GVHD. Best Pract Res Clin Haematol. 2008;21:149-164.

5. Conway SE, Abdi R. Immunoregulatory gene polymorphisms and graft-versus-host disease. Exp Rev Clin Immunol. 2009;5:523-534.

6. Ting C, Alterovitz G, Merlob A, Abdi R. Genomic studies of GVHD-lessons learned thus far. Bone Marrow Transplant. 2013;48:4-9.

7. Sayegh MH. Looking into the crystal ball: kidney transplantation in2025. Nat Clin Pract Nephrol. 2009;5:117.

8. Linsley PS, Ledbetter JA. The role of the CD28 receptor during T cellresponses to antigen. Annu Rev Immunol. 1993;11:191-212.

9. Bluestone JA. New perspectives of CD28-B7-mediated T cell cos-timulation. Immunity. 1995;2:555-559.

10. Rothstein DM, Sayegh MH. T-cell costimulatory pathways in allograftrejection and tolerance. Immunol Rev. 2003;196:85-108.

11. Boenisch O, Sayegh MH, Najafian N. Negative T-cell costimulatorypathways: their role in regulating alloimmune responses. Curr OpinOrgan Transplant. 2008;13:373-378.

12. Yu XZ, Martin PJ, Anasetti C. Role of CD28 in acute graft-versus-hostdisease. Blood. 1998;92:2963-2970.

13. Tsukada N, Akiba H, Kobata T, et al. Blockade of CD134 (OX40)-CD134Linteraction ameliorates lethal acute graft-versus-host disease in amurine model of allogeneic bone marrow transplantation. Blood. 2000;95:2434-2439.

14. Blazar BR, Carreno BM, Panoskaltsis-Mortari A, et al. Blockade of pro-grammed death-1 engagement accelerates graft-versus-host diseaselethality by an IFN-gamma-dependent mechanism. J Immunol. 2003;171:1272-1277.

15. Hori T. Roles of OX40 in the pathogenesis and the control of diseases.Int J Hematol. 2006;83:17-22.

16. Wythe SE, Dodd JS, Openshaw PJ, Schwarze J. OX40 ligand and pro-grammed cell death 1 ligand 2 expression on inflammatory dendriticcells regulates CD4 T cell cytokine production in the lung during viraldisease. J Immunol. 2012;188:1647-1655.

17. Mousavi SF, Soroosh P, Takahashi T, et al. OX40 costimulatory signalspotentiate the memory commitment of effector CD8þ T cells.J Immunol. 2008;181:5990-6001.

18. Ria M, Lagercrantz J, Samnegard A, et al. A common polymorphism inthe promoter region of the TNFSF4 gene is associated with lower allele-specific expression and risk of myocardial infarction. PLoS One. 2011;6:e17652.

19. Cunninghame Graham DS, Graham RR, Manku H, et al. Polymorphismat the TNF superfamily gene TNFSF4 confers susceptibility to systemiclupus erythematosus. Nat Genet. 2008;40:83-89.

20. Bossini-Castillo L, Broen JC, Simeon CP, et al. A replication study con-firms the association of TNFSF4 (OX40L) polymorphisms withsystemic sclerosis in a large European cohort. Ann Rheum Dis. 2011;70:638-641.

21. Parody R, Martino R, Rovira M, et al. Severe infections after unrelateddonor allogeneic hematopoietic stem cell transplantation in adults:comparisonof cordblood transplantationwithperipheral blood andbonemarrow transplantation. Biol BloodMarrow Transplant. 2006;12:734-748.

22. Pasquini M, Wang Z. Current uses and outcomes of hematopoietic stemcell transplantation. CIBMTR summary slides 2010. Available at: http://www.cibmtr.org.

23. Copelan EA. Hematopoietic stem-cell transplantation. N Engl J Med.2006;354:1813-1826.

24. Akiba H, Miyahira Y, Atsuta M, et al. Critical contribution of OX40ligand to T helper cell type 2 differentiation in experimental leish-maniasis. J Exp Med. 2000;191:375-380.

25. Glucksberg H, Storb R, Fefer A, et al. Clinical manifestations ofgraft-versus-host disease in human recipients of marrow from HL-A-matched sibling donors. Transplantation. 1974;18:295-304.

26. Shulman HM, Sullivan KM, Weiden PL, et al. Chronic graft-versus-hostsyndrome in man. A long-term clinicopathologic study of 20 Seattlepatients. Am J Med. 1980;69:204-217.

27. Gooley TA, Leisenring W, Crowley J, Storer BE. Estimation of failureprobabilities in the presence of competing risks: new representationsof old estimators. Stat Med. 1999;18:695-706.

28. Brand LH, Kirchler T, Hummel S, et al. DPI-ELISA: a fast and versatilemethod to specify the binding of plant transcription factors to DNAin vitro. Plant Methods. 2010;6:25.

29. Xiao C, Calado DP, Galler G, et al. MiR-150 controls B cell differentiationby targeting the transcription factor c-Myb. Cell. 2007;131:146-159.

30. Takasawa N, Ishii N, Higashimura N, et al. Expression of gp34 (OX40ligand) and OX40 on human T cell clones. Jpn J Cancer Res. 2001;92:377-382.

31. Frazer KA, Ballinger DG, Cox DR, et al. A second generation humanhaplotype map of over 3.1 million SNPs. Nature. 2007;449:851-861.

32. Boettler T, Moeckel F, Cheng Y, et al. OX40 facilitates control of apersistent virus infection. PLoS Pathog. 2012;8:e1002913.

33. Lee PH, Shatkay H. F-SNP: computationally predicted functional SNPsfor disease association studies. Nucleic Acids Res. 2008;36(DatabaseIssue):D820-D824.

34. Heinemeyer T, Wingender E, Reuter I, et al. Databases on transcrip-tional regulation: TRANSFAC, TRRD and COMPEL. Nucleic Acids Res.1998;26:362-367.

35. Quintana AM, Zhou YE, Pena JJ, et al. Dramatic repositioning of c-Mybto different promoters during the cell cycle observed by combining cellsorting with chromatin immunoprecipitation. PLoS One. 2011;6:e17362.

36. Petersdorf EW. Genetics of graft-versus-host disease: the major his-tocompatibility complex. Blood Rev. 2013;27:1-12.

37. Woolfrey A, Klein JP, Haagenson M, et al. HLA-C antigen mismatch isassociated with worse outcome in unrelated donor peripheral bloodstem cell transplantation. Biol Blood Marrow Transplant. 2011;17:885-892.

38. Dickinson AM, Harrold JL, Cullup H. Haematopoietic stem cell trans-plantation: can our genes predict clinical outcome? Exp Rev Mol Med.2007;9:1-19.

39. Ambruzova Z, Mrazek F, Raida L, et al. Association of IL6 and CCL2 genepolymorphisms with the outcome of allogeneic haematopoietic stemcell transplantation. Bone Marrow Transplant. 2009;44:227-235.

40. Perez-Garcia A, De la Camara R, Roman-Gomez J, et al. CTLA-4polymorphisms and clinical outcome after allogeneic stem cell trans-plantation from HLA-identical sibling donors. Blood. 2007;110:461-467.

41. Ambruzova Z, Mrazek F, Raida L, et al. Association of IL-6 gene poly-morphism with the outcome of allogeneic haematopoietic stem celltransplantation in Czech patients. Int J Immunogenet. 2008;35:401-403.

42. Azarian M, Busson M, Lepage V, et al. Donor CTLA-4 þ49 A/G*GGgenotype is associated with chronic GVHD after HLA-identical hae-matopoietic stem-cell transplants. Blood. 2007;110:4623-4624.

43. Tseng LH, Storer B, Petersdorf E, et al. IL10 and IL10 receptor genevariation and outcomes after unrelated and related hematopoietic celltransplantation. Transplantation. 2009;87:704-710.

44. Hippen KL, Harker-Murray P, Porter SB, et al. Umbilical cord bloodregulatory T-cell expansion and functional effects of tumor necrosisfactor receptor family members OX40 and 4-1BB expressed on artificialantigen-presenting cells. Blood. 2008;112:2847-2857.

45. Pan LP, Zhang W, Zhang L, et al. CD3Z genetic polymorphism in im-mune response to hepatitis B vaccination in two independent Chinesepopulations. PLoS One. 2012;7:e35303.

46. Snelgrove RJ, Cornere MM, Edwards L, et al. OX40 ligand fusion proteindelivered simultaneously with the BCG vaccine provides superiorprotection against murine mycobacterium tuberculosis infection.J Infect Dis. 2012;205:975-983.

47. Walch M, Rampini SK, Stoeckli I, et al. Involvement of CD252 (CD134L)and IL-2 in the expression of cytotoxic proteins in bacterial- or viral-activated human T cells. J Immunol. 2009;182:7569-7579.

48. Ozaki K, Ohnishi Y, Iida A, et al. Functional SNPs in the lymphotoxin-alpha gene that are associated with susceptibility to myocardialinfarction. Nat Genet. 2002;32:650-654.

49. Bassuny WM, Ihara K, Sasaki Y, et al. A functional polymorphism in thepromoter/enhancer region of the FOXP3/Scurfin gene associated withtype 1 diabetes. Immunogenetics. 2003;55:149-156.

50. Reddy CD, Reddy EP. Differential binding of nuclear factors to theintron 1 sequences containing the transcriptional pause site correlateswith c-myb expression. Proc Natl Acad Sci U S A. 1989;86:7326-7330.

51. Ogata K, Morikawa S, Nakamura H, et al. Solution structure of a specificDNA complex of the Myb DNA-binding domain with cooperativerecognition helices. Cell. 1994;79:639-648.

52. Fahl SP, Crittenden RB, Allman D, Bender TP. c-Myb is required for pro-B cell differentiation. J Immunol. 2009;183:5582-5592.

53. Golay J, Cusmano G, Introna M. Independent regulation of c-myc,B-myb, and c-myb gene expression by inducers and inhibitors ofproliferation in human B lymphocytes. J Immunol. 1992;149:300-308.

54. Quintana AM, Liu F, O’Rourke JP, Ness SA. Identification and regulationof c-Myb target genes in MCF-7 cells. BMC Cancer. 2011;11:30.

55. Jacobson P, Uberti J, Davis W, Ratanatharathorn V. Tacrolimus: a newagent for the prevention of graft-versus-host disease in hematopoieticstem cell transplantation. Bone Marrow Transplant. 1998;22:217-225.

56. Marmont AM, Horowitz MM, Gale RP, et al. T-cell depletion ofHLA-identical transplants in leukemia. Blood. 1991;78:2120-2130.