Genetic variation within the MHC is inherited en bloc in classical Mendelian fashion as a haplotype of markers on the same chromosomal strand. With current donor-matching criteria, less than 5% of the total MHC resident gene content is evaluated, and this opens up the potential for novel undetected variation as a cause of GVHD after HLA-matched transplantation. Together with HLA alleles, non-HLA loci in the MHC travel together, and their biological effects may synergize with those stemming from donor-recipient HLA mismatches. There has been strong interest in exploring non-HLA loci as a source of variation responsible for GVHD, particularly in the setting of HLA-matched transplantation ^{57, 58} . Studies have approached the search for novel transplantation determinants through candidate gene approaches or fine mapping with the aid of single nucleotide polymorphisms (SNPs) that define blocks of tightly linked markers (tagSNPs). In the case of unrelated donors and patients, identity for HLA alleles does not guarantee identity for other MHC loci, particularly if the patient’s and donor’s haplotypes are different ⁵⁹ .

A growing number of studies have taken a candidate gene approach to explore the clinical significance of the non-classical class I loci including HLA-E, HLA-G, and MHC class I related chain A ( MICA). Each of these genes is polymorphic and each has unique features that make them interesting candidates for transplantation determinants. A total of 25 HLA-E alleles that give rise to 18 proteins have been recognized; however, two alleles, E*01:01 and E*01:03, are the most frequently observed in most populations studied thus far ¹⁷ . Given its participation in both the innate and the adaptive immune pathways, HLA-E has been an attractive candidate gene with the potential to influence the risk of GVHD after transplantation; however, the evidence to date is heterogeneous ^{60– 64} . In HLA-matched sibling transplantation, homozygosity for HLA-E*01:03 was protective for GVHD and associated with improved survival ^{60, 62} . The presence of E*01:03 in transplant donors, however, was associated with higher grades (II–IV) of acute GVHD ⁶¹ . Two additional studies did not find an association of HLA-E with clinical outcome ^{63, 64} .

HLA-G is a non-classical class I gene best studied for its role in tolerance at the maternal-fetal placental interface ^{65, 66} . HLA-G encodes a total of 53 allelic variants giving rise to 18 proteins ¹⁷ . However, the most intriguing characteristic of this non-classical gene is its ability to form soluble as well as membrane-bound protein as a result of alternative splicing. A total of four membrane-bound and three soluble isoforms differ with respect to their size, structure, and ability to bind β2 microglobulin ^{67, 68} . The transcriptional regulation of HLA-G is complex and includes, but is not limited to, genetic variation within the 5’ and the 3’ UTRs ^{69– 71} . The 3’ UTR is particularly interesting, as it is characterized by a 14 basepair (bp) insertion/deletion and by haplotypes of SNPs, of which rs1063320 has been the subject of intense investigation as a basis for HLA-G expression and disease association ^{71– 74} . The 14 bp insertion results in the removal of 92 bases from exon 8 and is correlated with lower expression of HLA-G transcript ^{75, 76} . Its participation in both T and natural killer (NK) cell-mediated immune pathways ^{68, 77– 80} has prompted investigation into a role for HLA-G gene products in cancer, autoimmunity, and transplantation ^{74, 79, 81} .

In HCT, homozygosity for the 14 bp deletion correlated with higher risk of acute GVHD compared to homozygosity for the 14 bp insertion ⁸² . In an independent study, however, the 14 bp insertion was found to be a risk factor for acute GVHD ⁸³ . No correlation of the 14 bp insertion was found for acute GVHD, but an association with lower overall survival and disease-free survival was observed ⁸⁴ . Still, other investigations have not found associations among 3’ UTR haplotypes, the 14 bp insertion/deletion, and clinical outcome after HCT ^{85, 86} .

Although it has classically been considered that HLA-G expression is restricted to the maternal-fetal interface, the thymus, and the cornea, several recent studies have found increased levels of HLA-G in the plasma ^{87– 89} and in GVHD target organs in patients receiving allogeneic HCTs ⁸⁷ . In the first 30 days after HCT, levels of the soluble G5, G6 and G7 proteins were significantly higher compared to pre-transplant levels; higher levels of soluble HLA-G proteins were found in patients without GVHD compared to lower levels in patients who developed grades II–IV acute GVHD ⁸⁹ . Very intriguing data on the recovery of CD14 ⁺ HLA-G ⁺ cells in the plasma of healthy and transplant patients and the ability to antagonize the suppressive function of these cells through HLA-G blockade provide new information on the contribution of HLA-G-expressing monocytes in the immune response ⁸⁷ . Transplant recipients were found to have a higher frequency of HLA-G ⁺ CD8 ⁺ T cells after transplantation. Furthermore, neo-expression of HLA-G in the epidermis of patients with clinical GVHD, and a direct correlation of HLA-G expression with the severity of skin GVHD, suggests that up-regulation of HLA-G is involved in the etiology or clinical manifestation of this disease. In an independent study of patients undergoing HCT for hematologic malignancies, high levels of soluble HLA-G proteins within the first month after HCT could be recovered in patients who did not develop acute GVHD; the level of soluble HLA-G proteins correlated with the frequency of T regulatory cells with the CD4 ⁺ CD25 ⁺ CD152 ⁺ phenotype in transplant recipients ⁸⁸ .

In addition to HLA-E and -G, the MICA gene has been an attractive candidate to explain GVHD risks in related and unrelated donor transplantation. A total of 106 unique alleles giving rise to 82 proteins are recognized ¹⁷ . Although MICA shows sequence homology with HLA-A, -B, and -C, it lacks association with β2-microglobulin and does not bind or present peptides. MICA is expressed on the epithelium of the gastrointestinal tract (hence the interest in GVHD) and is induced by cellular stress ⁹⁰ . The MICA CD94/NKG2D activating receptor is expressed on most ϒδ T cells, αβ T cells, and NK cells ⁹¹ . Engagement of MICA with NKG2D leads to NK-mediated killing of target cells and CD8-positive ϒδ T cell-mediated activation of CTLs ^{92– 96} . It is no surprise then that MICA-NKG2D has been implicated in host resistance and susceptibility to infection ⁹⁷ , tumor surveillance ^{98– 100} , and autoimmunity ^{101– 103} . Particular attention has been paid to the rs1051792 variation that gives rise to methionine or valine at residue 129 of the MICA α domain; importantly, this position is associated with lower (valine) or higher (methionine) binding affinity to NKG2D ¹⁰⁴ and impacts the strength of NKG2D signaling and co-stimulation of CD8+ T cells ^{95, 97} .

Early investigation into the HLA-B–HLA-C region of class I uncovered donor-recipient mismatching for a series of markers, which were identified as MICA and MICB, and their potential relevance in HCT outcomes ¹⁰⁵ . Subsequently, a series of studies have shed light on the importance of donor-recipient MICA mismatching and the Met129Val polymorphism in clinical outcome after related and unrelated donor transplantation ^{96, 106– 110} . The extensive linkage disequilibrium across the MHC favors MICA matching among transplant pairs who are HLA-A, -B, -C, -DRB1, and -DQB1-matched. The clinical significance of MICA mismatching on GVHD risk has been heterogeneous and is likely complicated by a low mismatch rate and population differences. Higher risk of acute GVHD, in particular GVHD of the gastrointestinal tract, has been observed with donor-recipient MICA mismatching in some ^{106, 109} but not in other studies ¹⁰⁷ . Recently, a large retrospective analysis of unrelated donor-matched and -mismatched transplants performed in Germany demonstrated higher risks of mortality, lower disease-free survival, and higher rates of acute GVHD with donor-recipient mismatching for Val129Met ¹⁰⁸ . The presence of 129Met in recipients correlated with better overall survival and lower risk of death due to GVHD; however, homozygous 129Met/Met patients had increased risks ⁹⁶ . A retrospective study of MICA performed by the CIBMTR found no association of MICA mismatching nor Val129Met on clinical outcome ¹¹⁰ . The reasons for these heterogeneous results are unclear but may be the result of multifactorial genetic and environmental factors.

The use of SNPs for fine mapping novel transplantation determinants within the MHC is founded on the concept that tightly linked markers serve as proxies for one another. The use of tagSNPs makes no a priori assumptions for the likely disease-causing genes or their associated pathways; for a region that contains the most immune-related genes anywhere in the genome, tagSNPs provide an efficient and robust way to locate novel transplantation determinants. The elucidation of conserved extended HLA haplotypes provided the much-needed reference sequence(s) for designing SNP arrays for disease mapping ^{111– 113} . The availability of NGS methods has permitted the completion of in-depth interrogation of extended haplotypes that are homozygous for the classical HLA alleles, providing unprecedented annotation of coding and non-coding regions of the MHC ¹¹⁴ . When the classical HLA-A1, -B8, -DR3 and HLA-A2, -B, -DR15 haplotypes are aligned, extreme levels of sequence conservation over 3 Mb in length are evident ¹¹¹ . Further appreciation of the nature and organization of genetic variation on extended HLA haplotypes is demonstrated in the analysis of common haplotypes in ethnically diverse populations. In Japan, for example, unrelated individuals show remarkable sequence conservation for the three most common HLA haplotypes, spanning 3.3 Mb in length ¹¹⁵ .

Demonstration that haplotype-based approaches can facilitate the identification of novel transplantation determinants was shown in a study of HLA-matched unrelated donor-recipient pairs using a long-range phasing method ¹¹⁶ . The physical linkage of HLA-A, HLA-B, and HLA-DR on the same chromosomal strand was performed to identify matched pairs with the same HLA-A, -B, and -DR haplotypes and matched pairs with different haplotypes. HLA-matched pairs with different haplotypes had a significantly increased risk of grade III–IV acute GVHD, lower relapse, and similar overall survival ⁵⁹ . These observations suggest that the HLA haplotype can be used as a surrogate marker for GVHD risk and for the identification of specific risk loci. To this end, SNP arrays have been used to query the MHC in matched and mismatched unrelated donor transplants. In HLA-matched transplantation, two SNPs were validated as determinants of survival and acute GVHD ¹¹⁷ and a similar strategy is informative for single-locus mismatched unrelated donor transplants ¹¹⁸ . Collectively, the data thus far point to the presence of MHC resident variation that may confer risks alongside those stemming from HLA mismatching between the donor and the recipient. This information will enhance our understanding of the MHC as a critical region of the genome in transplantation biology and provide potential novel approaches for donor selection and for targeted immunotherapy.

NK cells do not directly cause GVHD after HLA-matched or HLA-mismatched transplantation. Unlike T cells, which recognize recipient major HLA and minor histocompatibility differences in host leukemia and normal tissues, NK cells recognize target cells that lack class I ligands (missing self; missing ligand), a situation that may arise with viral infection or malignant transformation. Inhibitory KIR receptors interact with their cognate ligands during NK development, which induces tolerance to target cells ¹²¹ . Target cells that are virally infected or are transformed by malignancy may lose their self-class I ligands; these target cells are recognized by licensed NK cells, leading to target cytotoxicity. In HLA-matched transplantation, patients who lack KIR ligands (missing ligand) experience lower risk of relapse and improved overall survival ¹²² , consistent with cytotoxicity of host residual leukemia by unlicensed NK cells. In HLA-mismatched transplantation, wherein the patient lacks KIR ligands that are present in the donor, licensed NK cells can mediate cytotoxicity against the KIR ligand mismatch ¹²³ . The reduction in relapse in both scenarios is not accompanied by increased GVHD. These unique properties of NK cells are the basis for the development of NK cells in adoptive immunotherapy ¹²⁴ .

Inhibitory KIR genes demonstrate a range of allelic variation and the alleles may have higher or lower degrees of inhibition that add to the diversity of individual immune responses. One example illustrates that features of both the receptor, KIR3DL1, and the ligand, HLA-Bw4, can be associated with different risks of relapse and mortality depending on the strength of inhibition (high or low) and the specific residue 80 sequence polymorphism in Bw4-positive cells ¹²⁵ . An added layer of polymorphism in the KIR genetic system is the organization of genes into two major haplotype groups defined by the number and nature of genes with inhibitory and activating potential ¹²⁶ ; whereas “A” haplotypes encode more inhibitory than activating genes, “B” haplotypes tend to encode more activating genes than do “A” haplotypes. Haplotype-based analyses of KIR demonstrate the importance of the number and nature of inhibitory and activating genes on A and B haplotypes on transplant outcomes. The presence of at least one B haplotype is associated with improved survival compared to lack of any B haplotype (i.e. only A haplotypes). These data support a role for activating KIRs in transplantation ¹²⁷ . Among the activating genes that have been studied thus far, the KIR2DS2 receptor and its HLA-C ligands expressing Asn77 and Lys80 (“C2” ligands) are capable of mediating strong anti-leukemic potential ¹²⁸ . The unique features of activating and inhibitory receptors together with their ligands provide avenues for lowering risk of relapse without an increase of GVHD through KIR-based algorithms applied to donor selection ¹²⁹ .

CD8 ⁺ and CD4 ⁺ T cells play a role in antigen-mediated recognition leading to GVHD after allogeneic transplantation. By definition, minor histocompatibility antigens are non-self peptides in which one or several polymorphisms within the homologous proteins between a transplant donor and patient may lead to altered binding of peptide to HLA and/or recognition of the HLA-peptide complex by T cells ^{130, 131} . Due to the diversity of minor histocompatibility antigens that can be presented by HLA, donor T cell-mediated allorecognition of disparities in minor histocompatibility antigens presented by patient target cells may induce GVHD in both HLA-matched and HLA-mismatched transplantation from both related and unrelated donors.

The millions of nucleotide sequence variants within the human genome provide a rich source for potential minor histocompatibility antigens ^{132– 134} . Variation within autosomal genes and genes of the Y chromosome contribute minor antigens of importance to transplantation ¹³⁵ . One of the best-studied minor histocompatibility antigens is the HMHA1 gene-derived “HA-1” nonamer peptide presented by HLA-A*02:01 allotypes; for this minor antigen, both HLA and T cell receptor binding features contribute to the immunogenicity of HA-1 ^{136, 137} .

Since SNPs represent the most common form of genetic variation accounting for the generation of minor histocompatibility antigens, the availability of SNP arrays to query whole genomes has yielded new insight into the loci that contribute to the pool of clinically relevant minor antigens ^{33, 47, 138} . In a Japanese study of HLA 10/10-matched unrelated donor transplants, autosomal SNPs have been identified that significantly increase the risk of GVHD. The identification of the true causative genes awaits further investigation in independent cohorts.

To predict the extent to which patient mismatching for minor histocompatibility antigens can contribute to GVHD risk, a single-center GWAS analysis compared the degree of genome-wide mismatching of donor anti-host recognition between HLA-matched sibling transplants and HLA-matched unrelated donor transplants ³³ . On average, 17.3% of unrelated transplants were mismatched for coding SNPs compared to 9.4% of HLA-identical sibling pairs. HLA-matched unrelated donor transplants overall had low GVHD-related outcome risks. The risk was higher among HLA-DP mismatched unrelated donor transplants compared to HLA-matched sibling transplants. These data suggest that GVHD risk after unrelated donor transplantation is conferred in large part by HLA disparity through direct recognition of the mismatched HLA as well as HLA-peptide complexes. Future analysis of larger cohorts will be required to fully examine whether coding SNP variation is a robust proxy for overall degree of minor antigen mismatching and whether genome-wide patient SNP mismatching can contribute to GVHD after unrelated donor transplantation.