HLA Haplotypes and Relapse After Hematopoietic Cell Transplantation

Abstract

PURPOSE

Recurrence of blood malignancy is the major cause of hematopoietic cell transplant failure. HLA class II molecules play a fundamental role in antitumor responses but the role of class II haplotypes is not known.

METHODS

HLA-DR, -DQ, -DM, and -DO allele variation was determined in 1,629 related haploidentical transplants to study the clinical significance of individual molecules and haplotypes.

RESULTS

Outcome correlated with patient and donor variation for HLA-DRβ residue 86 (Gly/Val), HLA-DQ (G1/G2) heterodimers, and donor HLA-DM (DM11,11/nonDM11,11) molecules, and depended on patient-donor mismatching. Risks of relapse were lower for DRβ-86 GlyGly patients when the donor was GlyVal (hazard ratio [HR], 0.46 [95% CI, 0.30 to 0.68]; P < .001); GlyVal patients benefited from HLA-DRB1–matched donors, whereas no donor was superior to another for ValVal patients. G1G2 patients with G1G2-mismatched donors had lower relapse. Transplantation from donors with DMα residue 184 ArgHis was associated with higher risk of relapse (HR, 1.60 [95% CI, 1.09 to 2.36]; P = .02) relative to ArgArg. Relapse and mortality risks differed across HLA-DR-DQ-DM haplotypes.

CONCLUSION

HLA class II haplotypes may be functional constituents of the transplantation barrier, and their consideration in patients and donors may improve the success of transplantation.

INTRODUCTION

Haploidentical hematopoietic cell transplantation incorporating post-transplant cyclophosphamide (PTCy) has emerged as an effective transplant strategy for patients with life-threatening hematologic malignancies, notably acute leukemia, for whom timely transplantation is of the essence. Haploidentical donors increase the availability of transplantation to patients who lack HLA-matched related or unrelated donors.^1-3 Advances that lower transplant-associated risks of graft failure, graft-versus-host disease (GVHD), and infections have increased safety,^4-11 while a more complete understanding of the mechanisms of relapse and the development of novel therapies have increased efficacy.^12-18 These transformative advancements have propelled haploidentical transplantation as the fastest growing strategy worldwide.^2,19,20

CONTEXT

• Key Objective

• Disease relapse is the main cause of failure of haploidentical related donor hematopoietic cell transplantation for hematologic malignancies. We studied the clinical significance of HLA-DR, -DQ, and -DM class II molecules and haplotypes in relapse after haploidentical related donor hematopoietic cell transplantation.

• Knowledge Generated

• The HLA-DRβ-86 Val/Gly dimorphism, HLA-DQ molecules, and HLA-DM molecules each influence the risk of relapse. Relapse and mortality depend on the patient and donor's HLA-DR-DQ-DM haplotypes as defined by the unique functional features of each gene.

• Relevance (C.F. Craddock)

• These data, which require confirmation in future studies, have the potential to inform donor selection in recipients of haploidentical transplantation and represent an important strategy to reduce the risk of disease relapse which remains the most important cause of transplant failure.*

  *Relevance section written by JCO Associate Editor Charles F. Craddock, MD.

Relapse is the chief cause of transplant failure and once established is difficult to eradicate.^13,21 Two landmark discoveries shed light on the mechanisms of post-transplant relapse.^22,23 Loss of heterozygosity involving the nonshared patient HLA haplotype results in the absence of the mismatched class I and class II molecules on target cells, rendering donor-mediated graft-versus-leukemia ineffective.^22,24,25 A second pathway for leukemia escape occurs with downregulation of the patient's class II molecules.^23,26 Both models suggest a central role of HLA class II in relapse. An enduring challenge is to better understand the immunologic mechanisms of relapse, which could inform new prevention and treatment strategies.

Information on the HLA system in hematopoietic cell transplantation has been limited to individual loci and the effects of mismatching between HLA allele sequences²⁷; however, there is mounting evidence that germline features of HLA molecules affect immunogenicity.^27-29 In this way, structural features of HLA molecules function as transplantation determinants, much in the same way as they affect susceptibility to autoimmune or infectious diseases.^30,31 A limitation of current knowledge is the role of HLA haplotypes. Genetic variation across the human genome is inherited as haplotypes of linked polymorphisms.³² Some of the best characterized haplotypes are those encoded by HLA genes because of their role in the immune response and disease susceptibility.^33,34 A hallmark of HLA haplotypes is the physical organization of genes with closely related function.^35,36 In the class II region, genes for the antigen-presenting molecules HLA-DR, -DQ, and -DP reside side-by-side genes encoding peptide-exchange molecules HLA-DM and -DO (Fig 1).³⁷ In this way, the inheritance en bloc of molecules with coordinated function in the immune response has far-reaching implications in health and disease.^30,38-42

FIG 1.

HLA class II genes, molecules, and haplotypes. (A) The HLA-DRA, -DRB1, -DQA1, -DQB1, -DOB, -DMB, -DMA, and -DOA genes are encoded within a 570-kilobase region of chr 6p21.3. Each class II HLA-DR, -DQ, -DM, and -DO molecule is an αβ heterodimer composed of an α chain product of an A gene (DRA, DQA1, DMA, and DOA) and a β chain product of a B gene (DRB1, DQB1, DMB, and DOB). Each diploid individual has two HLA haplotypes (shown is a single haplotype). Only genes examined in the current study are shown. Map is not to scale. (B) A patient and related haploidentical transplant donor share one complete HLA haplotype (gray) and are variably matched for the nonshared haplotypes (blue patient, purple donor).

The clustering of antigen-presentation and peptide-exchange molecules in the HLA class II region provides a framework for understanding the biological implications of haplotype-linked effects of this region of the genome. For a haplotype to provide maximal information on function, each gene marker is ideally described by a feature that itself has clinical significance. To achieve this, we identified the features of HLA-DR, -DQ, and -DM molecules that individually inform transplant outcome. We then applied these features to identify shared and nonshared class II haplotypes in related haploidentical patients and donors. The risks of post-transplant complications associated with HLA-DR-DQ-DM haplotypes were studied.

METHODS

Patients and Donors

We retrospectively studied 1,629 patients who received a related haploidentical transplant between 2007 and 2019 in US centers who reported clinical data to the Center for International Blood and Marrow Transplant Research (CIBMTR) and for whom research biospecimens for both patient and donor were available from the CIBMTR Research Repository (Data Supplement, Table S1, online only). There were no exclusion criteria. The HLA-DRA, -DM, and -DO polymorphisms under study were not considered at the time of transplantation. The hypotheses were formulated independent of data collection. Protocols were approved by the institutional review boards of the National Institutes of Health Office for Human Research Protections, Fred Hutchinson Cancer Center, and the National Marrow Donor Program. Written informed consent was obtained from participants. The funding agencies had no role in study design, data collection and analysis, the decision to submit the manuscript for publication, or manuscript preparation.

HLA

HLA-A, -B, -C, -DRB1, -DQB1, -DPB1 were typed pretransplant at two-field resolution.⁴³ In the current study, DNAs for the patient and donor were characterized for HLA-DRA, -DMA, -DMB, -DOA, and -DOB loci (Data Supplement, Methods; Data Supplement, Tables S2 and S3). Hypotheses focused on the contribution of single-nucleotide polymorphisms, residues, individual molecules, and haplotypes of molecules to transplant outcome. HLA-DQ G1 and G2 heterodimers were defined as described.²⁹ The shared, patient nonshared, and donor nonshared haplotypes were defined for HLA-DRβ-86, HLA-DQ, and HLA-DM heterodimers through familial segregation. Mismatching was defined as a polymorphism in the patient that is absent in the donor (graft-v-host vector).

Biostatistical Methods

The primary outcome measurements were relapse (studied only in patients with a malignant disorder), disease-free survival (DFS, malignant disorders), and survival after transplantation. Secondary end points were acute (grades II-IV and III-IV) and chronic GVHD. The associations of patient genotype, donor genotype, and patient/donor mismatching with DFS and overall mortality were studied using Cox regression models. Associations with relapse were assessed using Fine-Gray regression, where death without relapse was regarded as a competing risk. Logistic regression was used for acute GVHD. Patients who did not fail by last contact were censored at last contact. Day 0 for all time-to-event outcomes was taken as the day of transplantation. Failure for DFS was taken to be the earlier date of death or relapse. Models adjusted for patient age, donor age, patient sex, donor sex, year of transplant, intensity of conditioning regimen, disease status, cytomegalovirus serostatus, source of stem cells, use of total-body irradiation, use of PTCy, comorbidity index (0, 1, 2, ≥3), patient race, family relationship of donor, HLA-B-leader, and HLA-A, -B, -C, -DR, -DQ, and -DP match status as appropriate.^27,43 Covariates with missing data were included in models by creating an additional category to reflect the missing value of the appropriate covariate. Individual patients were excluded from regression analysis if outcome data were missing for the particular end point. Two-sided P values from regression models were obtained from the Wald test. No adjustments were made to the P values associated with the fitted regression models, but a global test was conducted for comparisons involving more than two groups; select formal pairwise comparisons were made only if the global P value was ≤5%.

RESULTS

DRβ-86

In haploidentical HCT, relapse is lower in patients mismatched for HLA-DRB1 exon 2 compared with matched patients.⁴³ A protective effect of DRB1 allele-mismatching relative to -mismatching on relapse was also observed in the current study (hazard ratio, 0.75 [95% CI, 0.60 to 0.94]; P = .01; Fig 2).

FIG 2.

Clinical significance of HLA-DRβ-86 in DFS. (A) Effect of HLA-DRB1 allele (mis)matching. (B) Effect of HLA-DRβ-86 GlyGly, GlyVal, and ValVal genotypes among HLA-DRB1–matched transplants. (C) Effect of donor DRβ-86 genotype and (mis)matching among HLA-DRB1–mismatched DRβ-86 GlyGly patients. (D) Effect of donor DRβ-86 genotype and (mis)matching among HLA-DRB1–mismatched DRβ-86 GlyVal patients. (E) Effect of donor DRβ-86 genotype and (mis)matching among HLA-DRB1–mismatched DRβ-86 ValVal patients. DFS, disease-free survival.

Each patient and donor HLA-DRB1 allele encoded either glycine (DRβ-86Gly; 1,655/3,152 [53%] patient haplotypes and 1,708/3,190 [54%] donor haplotypes) or valine (DRβ-86Val). The hazards of mortality, relapse, and DFS associated with each patient DRβ-86 genotype differed between HLA-DRB1–matched and HLA-DRB1–mismatched transplants (interaction P = .04 mortality, 0.05 relapse, 0.01 DFS) but not with donor DRβ-86 genotype (interaction P = .41 mortality, 0.38 relapse, 0.16 DFS). The effect of patient DRβ-86 genotype on clinical outcome was therefore examined among HLA-DRB1–matched patients separately from HLA-DRB1–mismatched patients (Table 1; Fig 2). Patient genotype correlated significantly with relapse and DFS among HLA-DRB1–matched but not HLA-DRB1–mismatched transplants. No associations with GVHD and DRβ-86 genotype were appreciated, and outcome was similar whether or not DRβ-86 mismatching was accompanied by mismatching at other missense positions of exon 2 (Data Supplement, Table S4). These results suggest that the DRβ-86 dimorphism may be a determinant of transplantation outcome.

TABLE 1.

The Clinical Significance of HLA-DRβ-86 in Relapse and Survival After Haploidentical Transplantation

Group	DRβ-86 Genotype	Overall Mortality			Relapse			Disease-Free Survival
		No.	HR (95% CI); P	Global P	No.	HR (95% CI); P	Global P	No.	HR (95% CI); P	Global P
HLA-DRB1–matched transplantsa	GlyGly	97	1	.04	90	1	.02	90	1	.01
	GlyVal	87	0.53 (0.31 to 0.90); .02		80	0.49 (0.30 to 0.80); .004		80	0.39 (0.25 to 0.63); <.001
	ValVal	86	1.09 (0.68 to 1.74); .72		74	0.72 (0.45 to 1.15); .17		74	0.88 (0.57 to 1.35); .56
HLA-DRB1–mismatched transplantsb	GlyGly patients	338	1	.36	311	1	.73	311	1	.30
	GlyVal patients	683	1.12		606	1.08		606	1.12
	ValVal patients	273	1.32		244	1.11		244	1.32
DRβ-86 GlyGly patientsa	Matched	97	1	.60	90	1	.0005	90	1	.03
	Mismatched GlyGly donorc	184	1.07		169	0.60 (0.42 to 0.87); .007		169	0.85 (0.61 to 1.20); .36
	Mismatched GlyVal donor	153	0.81		141	0.46 (0.30 to 0.68); <.001		141	0.57 (0.39 to 0.82); .003
DRβ-86 GlyVal patientsa	Matched	87	1	.11	80	1	.32	80	1	.26
	Mismatched GlyGly donor	203	1.54		178	1.21		178	1.46
	Mismatched GlyVal donorc	304	1.75		271	0.92		271	1.33
	Mismatched ValVal donor	176	1.57		157	0.96		157	1.32
DRβ-86 ValVal patientsa	Matched	86	1	.77	74	1	.66	74	1	.75
	Mismatched GlyVal donor	157	1.04		142	0.86		142	0.88
	Mismatched ValVal donorc	115	1.04		101	0.80		101	0.91

NOTE. Clinical outcome depends on both the patient DRβ-86 genotype and the donor HLA-DRB1 match status. A mismatch is defined as the presence of a polymorphism in the patient that is absent in the donor (graft-v-host vector).

Abbreviations: CMV, cytomegalovirus; GVHD, graft-versus-host disease; HR, hazard ratio; TBI, total-body irradiation.

^a Models adjusted for the year of transplant, preparative regimen, disease stage, patient age, donor age, stem-cell source, patient sex, and patient race.

^b Models adjusted for the year of transplant, preparative regimen, disease stage, CMV serostatus, patient age, donor age, stem-cell source, patient sex, donor sex, use of TBI, GVHD prophylaxis, comorbidity index, family relationship, HLA-B leader mismatching, patient race, and patient nonshared HLA-DPB1 rs9277534 allele.

^c Patients and donors are mismatched for two different HLA-DRB1 alleles at sites other than DRβ-86.

A patient's genotype cannot be modified but patients may have a choice of matched and mismatched donors. To examine potential differences among haploidentical donors, we assessed outcomes for each patient genotype according to donor HLA-DRB1 match status and DRβ-86 (Table 1; Fig 2). GlyGly patients had significantly lower risks of relapse and improved overall survival and DFS with mismatched GlyVal donors. By contrast, no particular donor was obviously superior to another for GlyVal and ValVal patients. These results suggest that a preferred donor may be defined by both the DRβ-86 genotype and HLA-DRB1 match status.

HLA-DRA rs8084 generates transcripts of different lengths⁴⁴ and rs7192 encodes DRα-217Val/Leu⁴⁵ but neither polymorphism was demonstrably correlated with outcome in the patient, the donor, or patient/donor mismatching, and no further consideration was given to HLA-DRA (Data Supplement, Table S5). In summary, the DRβ-86 dimorphism, together with donor (mis)matching, could influence the biological effects of HLA-DR variation.

HLA-DQ Heterodimers

The effect of patient HLA-DQ heterodimers showed some evidence of a differential effect in HLA-DQB1–matched and HLA-DQB1–mismatched transplants for mortality (interaction P = .13) and DFS (P = .18) but there was no statistically significant difference among donor heterodimers between HLA-DQB1–matched and –mismatched transplants (interaction P = .73 mortality, 0.85 relapse, 0.58 DFS).

The potential dependency of patient heterodimer on match status motivated examination of the effect of heterodimer genotypes with HLA-DQB1–matched and –mismatched donors among patients of a given heterodimer genotype (Table 2). Among patients with HLA-DQB1–matched and –mismatched donors, donor heterodimer (mis)match status affected DFS for G1G2 patients. Among patients with only HLA-DQB1–mismatched donors, G1G2 patients benefited from knowledge of donor heterodimer (mis)match status. Heterodimers did not obviously inform GVHD risk (Data Supplement, Table S6). These data suggest that in select cases, DQ heterodimers may be functional in haploidentical transplantation. Not all HLA-DQB1 mismatches confer the same risks. The combination of heterodimer genotype with HLA-DQB1 match status may provide more complete information on outcome than either alone.

TABLE 2.

HLA-DQ Heterodimers and HLA-DQB1 Allele Match Status in Relapse and Survival After Haploidentical Transplantation

Donor Availability	Patient HLA-DQ Heterodimer Genotype	Donor Match Status and HLA-DQ Heterodimer Genotype	Overall Mortality			Relapse			Disease-Free Survival
			No.	HR (95% CI); P	Global P	No.	HR (95% CI); P	Global P	No.	HR (95% CI); P	Global P
Matched and mismatched donors available	G1G2	Match	115	1	.18	101	1	.31	101	1	.02
		Mismatched G1G1	194	1.45		172	1.20		172	1.46 (1.04 to 2.05); .03
		Mismatched G1G2a	251	1.22		221	0.91		221	1.04 (0.74 to 1.45); .84
		Mismatched G2G2	141	1.40		126	1.10		126	1.37 (0.95 to 1.99); .09
	G1G1	Match	232	1	.20	211	1	.67	211	1	.50
		Mismatched G1G1b	157	1.21		134	1.06		134	1.21
		Mismatched G1G2	126	1.44		118	0.89		118	1.14
	G2G2	Match	89	1	.76	83	1	.78	83	1	.71
		Mismatched G1G2	109	0.75		102	0.93		102	0.74
		Mismatched G2G2c	63	0.80		56	0.82		56	0.86
Only mismatched donors available	G1G2	Mismatched G1G1	194	1	.50	172	1	.17	172	1	.05
		Mismatched G1G2a	251	0.86		221	0.75		221	0.71 (0.54 to 0.93); .01
		Mismatched G2G2	141	0.97		126	0.92		126	0.95 (0.70 to 1.29); .74
	G1G1	Mismatched G1G1b	157	1	NA	134	1	NA	134	1	NA
		Mismatched G1G2	126	1.16 (0.81 to 1.67); .41		118	0.88 (0.58 to 1.32); .53		118	0.95 (0.68 to 1.32); .76
	G2G2	Mismatched G1G2	109	1	NA	102	1	NA	102	1	NA
		Mismatched G2G2c	63	1.10 (0.63 to 1.91); .74		56	0.88 (0.48 to 1.62); .69		56	1.10 (0.67 to 1.82); .70

NOTE. Clinical outcome depends on the HLA-DQ heterodimer and donor match status. Models show donor choices according to the HLA-DQ heterodimer genotype of the patient. When patients have both matched and mismatched donors, models assess hazards relative to matched donors. When patients have only mismatched donors, models assess hazards relative to a mismatched donor group. Models adjusted for year of transplantation, preparative regimen, disease stage, patient age, donor age, stem-cell source, patient sex, and patient race.

Abbreviations: HR, hazard ratio; NA, not applicable.

^a G1G2 patients with mismatched G1G2 donors are mismatched for two different HLA-DQB1 alleles within the same heterodimer group.

^b G1G1 patients with mismatched G1G1 donors are mismatched for two different HLA-DQB1 alleles within the G1 group.

^c G2G2 patients with mismatched G2G2 donors are mismatched for two different HLA-DQB1 alleles within the G2 group.

HLA-DM and HLA-DO

HLA-DM was polymorphic, with a total of 42 unique genotypes in patients and donors. The most common genotype was DM11,11 in 598 (37%) patients and 587 (36%) donors. Because of the extensive diversity of nonHLA-DM11 genotypes, genotypes were grouped together (nonDM11). Possession of nonDM11 genotypes in the donor lowered DFS relative to DM11,11 donors (Table 3). The patient's genotype and patient/donor (mis)matching did not obviously inform outcome; neither HLA-DM genotype nor (mis)matching correlated with GVHD (Data Supplement, Table S7).

TABLE 3.

Clinical Significance of Donor HLA-DM Variation in Haploidentical Transplantation

Donor Modela	Donor Genotype	Overall Mortality			Relapse			Disease-Free Survival
		No.	HR (95% CI); P	Global P	No.	HR (95% CI); P	Global P	No.	HR (95% CI); P	Global P
Genotype	DM11/DM11	576	1	NA	522	1	NA	522	1	NA
	NonDM11/DM11	872	1.06 (0.89 to 1.25); .53		782	1.09 (0.91 to 1.30); .35		782	1.21 (1.04 to 1.42); .01
DMα-184	ArgArgb	1,466	1	.04	1,315	1	.02	1,315	1	.03
	ArgCysb	42	1.69 (1.10 to 2.59); .02		40	1.36 (0.92 to 2.01); .13		40	1.59 (1.09 to 2.32); .02
	ArgHisb	69	1.43 (1.01 to 2.02); .04		62	1.60 (1.09 to 2.36); .02		62	1.41 (1.01 to 1.95); .04
DMα-155	GlyGlyb	1,508	1	NA	1,355	1	NA	1,355	1	NA
	AlaGlyb	69	1.40 (0.99 to 1.98); .06		62	1.58 (1.08 to 2.33); .02		62	1.38 (1.00 to 1.92); .05
DMβ-10	ThrThrb	1,508	1	NA	1,355	1	NA	1,355	1	NA
	AlaThrb	69	1.33 (0.94 to 1.88); .11		62	1.58 (1.08 to 2.33); .02		62	1.30 (0.94 to 1.81); .12

Abbreviations: CMV, cytomegalovirus; GVHD, graft-versus-host disease; HR, hazard ratio; NA, not applicable; TBI, total-body irradiation.

^a Models adjusted for the year of transplant, preparative regimen, disease stage, CMV serostatus, patient age, donor age, stem-cell source, patient sex, donor sex, use of TBI, GVHD prophylaxis, comorbidity index, family relationship, HLA-B leader mismatching, patient race, and patient nonshared HLA-DPB1 rs9277534 allele.

^b Arg, arginine; Cys, cysteine; His, histidine; Gly, glycine; Ala, alanine; Thr, threonine.

The association of donor HLA-DM genotype with DFS motivated examination of the role of specific missense substitutions in the HLA-DM αβ heterodimer. For DMα residue 184 (DMα-184), the presence of a histidine (His) or a cysteine (Cys) substitution was associated with higher risks of relapse and lower DFS and overall survival relative to arginine (Arg), notably for DMα-184ArgHis donors relative to ArgArg donors (Table 3). Missense changes at DMα-155 and DMβ-10 were also each clinically significant (Table 3). Phasing of DMα-155Ala with DMα-184His defines the DMA*01:03 allele, while DMβ-10Ala is a marker for DMB*01:07. The αβ heterodimer DMA*01:03-DMB*01:07 (“DM37”) was found in a total of 40 donors. DMα-184Cys is a marker for DMA*01:04 and was found in 43 donors. These data suggest that donor HLA-DM variation may have biological relevance in transplantation.

HLA-DO was also polymorphic, with 20 unique genotypes in patients and 21 in donors. The major genotype DO11,11 was found in 962 (59%) patients and 946 (58%) donors. Because of the extensive diversity of nonHLA-DO11 genotypes, genotypes were grouped together (nonDO11). Patients and donors with DO11,11 and nonDO11,11 genotypes had similar outcome; patient/donor matching for HLA-DO genotype did not show a demonstrable correlation with outcome (Data Supplement, Table S8).

HLA Class II Haplotypes

The studies described above show that HLA-DR, HLA-DQ, and HLA-DM may each influence relapse and mortality. We hypothesized that specific haplotypes of molecules together with haplotype (mis)matching are important for understanding their immunobiology. Among the 1,629 pairs in the study, shared and nonshared HLA-DR-DQ-DM haplotypes could be defined for 945 pairs. Of these, the nonshared haplotypes were identical in 125 (matched) and different in 820 pairs (mismatched). The role of haplotype matching was evaluated without regard to the specific molecules of the shared and nonshared haplotypes. Relative to matched pairs, mismatched pairs had similar overall mortality, relapse, and DFS; risks did not appear to depend on the total number of mismatched loci on the nonshared haplotypes (Data Supplement, Table S9).

The mismatch models described above do not consider the specific HLA-DR-DQ-DM haplotypes. Given the sentinel role of individual HLA-DR, -DQ, and -DM molecules in transplant outcome, we distinguished the shared haplotype from the nonshared haplotype to evaluate the effects contributed by specific haplotypes defined by DRβ-86 Gly/Val, HLA-DQ G1/G2, and HLA-DM11/nonDM11. An interaction between the shared and nonshared haplotypes was evident when the shared haplotype is specified. Global tests were performed on the four most common shared haplotypes (Gly-G1-DM11; Val-G1-DM11; Val-G2-DM11; and Gly-G2-DM11) to determine if effects varied for each nonshared patient haplotype and nonshared donor haplotype. The global tests suggested that the patient's nonshared haplotype effects varied for relapse when the shared haplotype is Val-G2-DM11 (P = .04) and the donor's nonshared haplotype effects varied for DFS when the shared haplotype is Gly-G2-DM11 (P = .04). Specifically, when Val-G2-DM11 is shared, a nonshared Gly-G1-nonDM11 in the patient conferred higher risk of relapse and lower DFS relative to a nonshared Gly-G1-DM11 (Fig 3; Data Supplement, Table S10). When the shared haplotype is Gly-G2-DM11, a donor nonshared Val-G2-DM11 conferred lower DFS relative to a nonshared Gly-G1-DM11 (Fig 3). In both scenarios, the presence of Val-G2-DM11 as a matched or a mismatched haplotype increased risk. Haplotypes did not appear to be correlated with GVHD (Data Supplement, Table S10). These results suggest that clinical outcome may be shaped by the specific constellation of HLA-DR, -DQ, and -DM molecules inherited as haplotypes, and whether the haplotype is matched (shared) or mismatched (nonshared). This novel approach combines the concepts of matching together with specific sequence features to define the potential clinical significance of class II haplotypes.

FIG 3.

Clinical significance of HLA-DR-DQ-DM haplotypes in disease-free survival. Class II haplotypes are defined by DRβ-86 Gly/Val, DQ G1/G2 heterodimer, and DM11/nonDM11 heterodimer. (A) Effect of the patient's nonshared haplotype when the shared haplotype is Val-G2-DM11. (B) Effect of the donor's nonshared haplotype when the shared haplotype is Gly-G2-DM11. The x-axis is log scale. Arrowhead denotes that the lower limit of the 95% CI spans further than indicated by the x-axis. Models adjusted for year of transplantation, disease stage, patient age, and patient race.

DISCUSSION

Relapse remains an enduring challenge in haploidentical transplantation. The clustering of class II molecules with shared function in the HLA class II region led us to hypothesize that HLA-DR, -DQ, and -DM molecules and haplotypes have biological relevance. Haploidentity between the patients and donors permitted unequivocal phasing of A genes with B genes and across loci that define HLA-DR-DQ-DM haplotypes. Haplotypes describe a series of sequence features. We prioritized the DRβ-86 Gly/Val dimorphism because of its role in autoimmunity and peptide repertoire, and HLA-DQ heterodimers because of their effects in unrelated donor transplantation.^26,46-49 Models for HLA-DM and -DO are lacking, and we used a classic approach to define αβ variation. Once the functional feature of each molecule was identified, then the locus-specific features were linked as shared or nonshared haplotypes. The association of transplant outcome with haplotypes of physically linked accessory and antigen-presenting molecules illustrates the potential biological significance of this highly unique region of the genome and the integrated roles of molecules with coordinated immunologic function. This new paradigm for transplantation integrates models of allorecognition of at-risk epitopes as seen in autoimmunity, with models based on HLA mismatching. Our findings require validation in independent transplant populations because of the large number of comparisons currently made, which increased the chances of false-positive findings. If confirmed, they offer strategies to optimize donor selection^50-54 and isolate graft-versus-leukemia from GVHD.

We observed that the effects of the mismatched haplotype may depend on the matched haplotype. In the broadest sense, such interactions between multiple genes of haplotypes might provide new evidence for genetic and molecular epistasis, in which the effects of one variant may be modified by another variant.^55,56 The HLA-DQ–associated risks are consistent with those observed in unrelated donor transplantation without PTCy.²⁹ HLA-DRβ-86, HLA-DM, and class II haplotypes are newly recognized components of the HLA barrier, and their associations (and lack thereof for HLA-DRA and HLA-DO) require validation. The inclusion of HLA-DP into the definition of extended haplotypes will be feasible when a much larger transplant experience is available.⁵⁷

The DRβ-86 dimorphism alters the peptide-groove, HLA-DQ heterodimers differ for their peptide-binding region, and HLA-DM facilitates class II peptide-exchange and presentation.^37,46,58 These concerted functions implicate peptide-HLA complexes (pHLAs) in graft-versus-leukemia effects, which may depend on the specific constellation of antigen-presenting and peptide-exchange determinants specified by the haplotypes. In this regard, DM37 has been predicted to be a less efficient peptide-exchange catalyst than DM11⁵⁹ and our results provide a clinical model in which natural HLA-DM variation results in very different transplant outcomes. We observed significant allelic diversity of HLA-DM molecules, and the functional importance of nonDM11 molecules will enhance understanding of class II function. Although we identified potential risks associated with Val-G2-DM11, haplotypes also provide information on untyped variation, including regulatory variants that influence the expression of class II molecules.⁶⁰ Future studies using functional assays to describe the peptide repertoires of DRβ-Val86Gly and G1/G2 heterodimers, the impact of specific HLA-DM molecules on those repertoires,^37,59 and the magnitude and specificity of T-cell responses against pHLAs^61,62 will inform the biological models. It is intriguing to hypothesize that graft-versus-leukemia effects might hinge upon the interplay of haplotypes, loss of heterozygosity, and downregulation, and introduce combined mechanisms to be explored.

New frontline clinical strategies may be envisioned to reduce post-transplant relapse for future patients, including, but not limited to, patients with acute leukemia. Germline variation cannot be modified, but patients may have choices among related donors. The data suggest that the choice of an optimal haploidentical donor may depend on the patient's DRβ-86 and HLA-DQ heterodimers. Patients with DRβ-86GlyGly may benefit from mismatched GlyVal donors, whereas GlyVal patients may have lower relapse if their donor is HLA-DRB1–matched. Noteworthy is that GVHD risk in this cohort of patients treated with PTCy was apparently not increased despite DRβ disparity. G1G2 patients may benefit from G1G2 donors. These findings suggest that prospective selection of donors may be optimized with knowledge of HLA-DRβ-86, HLA-DQ heterodimer, and donor HLA-DM genotype. HLA-DRβ-86 and the HLA-DQ heterodimers may be determined directly from pretransplant allele typing, and HLA-DM genotyping is amenable with widely available DNA-based methods. Upfront delineation of the shared and nonshared haplotypes of the patient and candidate related donors can be performed to assess three-locus HLA-DR-DQ-DM molecule haplotypes. The presence of Val-G2-DM11 either as a shared or nonshared haplotype may help to identify patients who could benefit from additional peritransplant therapy to lower the risks of relapse. Confirmation of these results is warranted in independent transplants. The current study was designed to interrogate the clinical significance of class II variation while accounting for non-HLA variables (such as patient or donor age) that affect the same outcomes.^43,50-54 The relative contributions of HLA and non-HLA characteristics to the selection of candidate donors remain important questions, and analysis will be feasible when larger populations are available to support comparisons of different combinations of factors.

In conclusion, HLA class II appears to contribute to transplant success through the additive effects of molecules with coordinated function in antigen presentation. The elucidation of functional haplotypes has major implications in transplantation and understanding the role of gene structure and function.

DATA SHARING STATEMENT

Requests for data may be made to the corresponding author. Data sharing for research purposes may have partial restrictions consistent with the written informed consent of study participants from whom the data were obtained.

AUTHOR CONTRIBUTIONS

Conception and design: Effie W. Petersdorf

Provision of study materials or patients: Stephen R. Spellman

Collection and assembly of data: Effie W. Petersdorf, Caroline McKallor, Mari Malkki, Meilun He

Data analysis and interpretation: Effie W. Petersdorf, Caroline McKallor, Mari Malkki, Stephen R. Spellman, Theodore Gooley, Philip Stevenson

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

HLA Haplotypes and Relapse After Hematopoietic Cell Transplantation

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).