Interferon-γ and interleukin-6 gene polymorphisms associate with graft-versus-host disease in HLA-matched sibling bone marrow transplantation

Graft-versus-host disease (GVHD) is the most common serious complication of allogeneic bone marrow transplantation (BMT) and severe (grade III-IV) acute GVHD (aGVHD) causes increased mortality.1 However immunosuppressive prophylaxis for GVHD increases infection and decreases the graft-versus-malignancy effect.2 Established risk factors for aGVHD include histoincompatibility, age, sex mismatch, viral status, and prophylaxis,1 whereas chronic GVHD (cGVHD) is largely predicted by prior aGVHD.3 Currently there are no widely established approaches to individualized prediction of GVHD. Proinflammatory cytokines including interferon-γ (IFNγ), interleukin-6 (IL-6), and tumor necrosis factor-α (TNFα) are important mediators and regulators of GVHD.4,5 The anti-inflammatory and immunomodulatory cytokine IL-10 is associated with transplantation tolerance6 and decreased GVHD.7 8 

In human clinical BMT, a large number of T-cell clones produce IFNγ,9 whereas levels of IFNγ increase both before and during aGVHD.10,11 IFNγ is also implicated as a mediator of aGVHD in a human skin explant model of GVHD, especially in combination with TNFα.12 Murine models suggest that IFNγ may also play a down-regulatory role in aGVHD.13-15Increased serum IFNγ16 and in situ IFNγ transcription have also been demonstrated in cutaneous cGVHD.17 The IFNγ gene (IFNG) maps to chromosome 12p24 and has a (CA)_n repeat element within the first intron.18 This microsatellite has 2 common alleles, 2 and 3, which exhibit significant differences in IFNγ production in vitro; allele 2 associates with greater IFNγ production than other alleles from mitogen-stimulated peripheral blood mononuclear cells.19,20 The IFNγIntron1 microsatellite has shown association with a variety of autoimmune and alloimmune disease states, including lung transplant fibrosis21 and renal transplant rejection.22 

Serum IL-6 levels increase during aGVHD,23correlating with severity24,25 and prognosis,26 and also are elevated in cGVHD.16 The IL6 gene maps to chromosome 7p21,27 and in the promoter region at position −174 a (G/C) single nucleotide polymorphism (SNP) is found close to a glucocorticoid-response element. The IL-6⁻¹⁷⁴C allele associates with lower in vitro and in vivo production, and IL-6⁻¹⁷⁴CC homozygosity is underrepresented in juvenile chronic arthritis.28 Recently the IL-6⁻¹⁷⁴(G/C) SNP genotype of renal transplant donors has been shown to associate significantly with the incidence and severity of acute rejection.29 Just 3′ to the IL6 gene is an AT-rich minisatellite30 with variable repeat numbers; this polymorphism has been shown to associate with disease states including osteoporosis31 and systemic lupus erythematosus.32 

We have previously demonstrated that homozygosity for the putative high TNFα–producer genotype TNFd3 (TNFd3/d3) associates with grade III to IV aGVHD in cyclosporin A (CyA)-treated HLA-matched sibling BMT.33 Possession of one or more IL-10⁻¹⁰⁶⁴ microsatellite alleles with a high repeat number (i[12-16]) by the recipient also associated with grades III to IV aGVHD in this cohort.33 The association of IL-10⁻¹⁰⁶⁴i[12-16] alleles with severe aGVHD has been confirmed in HLA-matched sibling recipients given methotrexate (MTX) in addition to CyA prophylaxis.34 Similar associations of TNFd and IL-10⁻¹⁰⁶⁴i[12-16] alleles with aGVHD severity have been observed in HLA-mismatched cord blood transplants.35 The combination of genetic risk factors in alloimmune processes has been previously suggested as a means of predicting outcome in renal transplantation.22 36 The current study was designed to test the association of GVHD with candidate IFNγ and IL-6 genotypes in an extended HLA-matched sibling BMT cohort. It was hypothesized that presence of functional polymorphism in the genes of these candidate cytokines would predispose toward differences in aGVHD and cGVHD. Also we wished to assess the relative risk of these cytokine genotypes with respect to established risk factors for GVHD.

Archival DNA samples from 80 nonpediatric (mean age, 29.18 years; SD, 8.88) sibling BMT recipient and donor pairs were studied. All patients underwent transplantation between 1984 and 1997 for hemopoetic malignancy, with acute leukemia (36 acute myeloid leukemia [AML] and 19 acute lymphocytic leukemia [ALL] patients) and chronic myeloid leukemia (CML, 19 patients) being the most common underlying diagnoses. Fifty-two of the recipients and 44 of the donors were men, with 22 male recipients receiving marrow from a female donor. Thirty-three recipients and 35 donors were positive for cytomegalovirus (CMV) by serology prior to BMT.

Conditioning schedules were those appropriate to the disease, remission status, and prior treatment; after 1990, chemotherapy schedules contained melphalan (3 mg/kg) in place of cyclophosphamide (60 mg/kg × 2). Conditioning comprised cyclophosphamide followed by fractionated total body irradiation (TBI; total 1200 cGy in 6 fractions at 25 cGy/min) in 42 patients, melphalan and TBI in 21 patients, and combined cyclophosphamide, melphalan, and TBI for 1 patient. Busulphan (16 mg/kg) was administered in place of TBI for patients who had previously undergone radiation therapy, together with cyclophosphamide in 9 patients, and busulphan with melphalan in 6 patients. One patient with hypoplastic myelodysplasia received cyclophosphamide alone.

HLA matching had been performed serologically for HLA-A and -B antigens and by high-resolution molecular typing for HLA-DRB1. All grafts were non–T cell depleted and GVHD prophylaxis consisted of 3 mg/kg CyA in all patients. Patients at high risk for GVHD, as determined by skin explant assay,37 advanced age, or multiparous female donor, also received either “short course” MTX38(n = 17) or corticosteroids (n = 5). All genotypes were determined with the investigator blinded to clinical outcomes. All patients had given informed consent for participation in the study as approved by the local ethics committee.

IFNγIntron1 microsatellite genotypes were determined by polymerase chain reaction (PCR) using 1 μM of each primer,19 0.5 U Taq polymerase (Bioline, London, United Kingdom), 200 μM dNTP mixture, 2.5 mM MgCl₂ in 1 × KCl buffer (Bioline) in addition to test DNA, to a final volume of 25 μL. PCR was for 30 cycles at 94°C for 30 seconds, 60°C for 60 seconds, and 72°C for 60 seconds, followed by a final extension of 72°C for 7 minutes. PCR products were resolved by polyacrylamide gel electrophoresis (8%; 19:1 acrylamide to bisacrylamide) and visualized by silver staining.

The IL-6⁻¹⁷⁴(G/C) SNP genotypes were determined in an allele-specific PCR. Primers were GAGCTTCTCTTTCGTTCC and either CCCTAGTTGTGTCTTGCC or CCCTAGTTGTGTCTTGCG. Reactions contained 1 μM of each primer, 0.5 U Taq polymerase (Bioline), 200 μM dNTP mixture, and 1.5 mM MgCl₂ in 1 × KCl buffer (Bioline) in addition to test DNA, to a final volume of 25 μL. Amplification was performed on PerkinElmer thermal-cycler (Norwalk, CT) with 30 cycles of 94°C for 30 seconds, 54°C for 60 seconds, and 72°C for 60 seconds, followed by a final extension of 72°C for 7 minutes. PCR products were resolved by agarose gel electrophoresis (2%) and visualized by ethidium bromide.

The IL-6 3′ (AT)-rich minisatellite genotypes were determined in a PCR reaction containing 1 μM of each primer,32 0.5 U Taq polymerase (Bioline), 200 μM dNTP mixture, and 2 mM MgCl₂ in 1 × NH₄ buffer (Bioline) in addition to test DNA, to a final volume of 60 μL. Following a “hot start,” the PCR cycle was as for the IFNγ method above. PCR products were resolved by polyacrylamide gel electrophoresis (4%; 19:1 acrylamide to bisacrylamide) and visualized by silver staining.

Genotypes for the TNFd microsatellite polymorphism were determined as previously described.39 IL-10 haplotypes were determined in an IL-10⁻¹⁰⁸² SNP allele-specific PCR that coamplifies and reports the linked IL-10⁻¹⁰⁶⁴microsatellite repeat.34 

Univariable analysis of data were by ANOVA using Kruskal-Wallis test, including F test analysis of group variances. Survival estimation by Kaplan-Meier analysis, together with bivariable and multivariable forward stepwise logistic regression analysis was performed using SPSS software (version 9; Chicago, IL). P values less than .05 were regarded as statistically significant, and those between .05 and .1 as indicative of a trend.

Four additional patients, who had been genotyped, died prior to day 30 without significant aGVHD and were not analyzed for this outcome. Patients who survived more than 30 days from BMT, or who died prior to 30 days with significant GVHD (grades II-IV) were considered assessable for aGVHD (grading by Glucksberg criteria40). Fifteen of the 80 assessable patients did not develop aGVHD; 27 patients developed grade I, 22 grade II, 11 grade III, and 5 grade IV aGVHD. A further 8 patients died prior to day 100 without cGVHD; overall, the day 100 transplantation-related mortality rate was 22.6% (range, 18.0%-27.2%). Patients who survived more than 100 days from BMT were considered assessable for cGVHD (grading by Atkinson et al41). Thirty of 72 assessable patients developed cGVHD. The presence of aGVHD correlated with cGVHD risk. Two of 15 patients surviving at least 100 days without aGVHD developed de novo cGVHD, whereas 28 of 57 with grades I to IV aGVHD went on to develop cGVHD. At day 180, the GVHD mortality rate was 10.7% (range, 8.7%-12.7%) comprising 7 aGVHD-related and 2 cGVHD-related deaths (with 1 subsequent cGVHD-related death at 35 months), and non-GVHD mortality was 15.5% (range, 12.7%-18.3%).

Genotyping at the IL-6 (AT)-rich minisatellite showed the following allele frequencies: allele A and B both, f = 0.021; allele C, f = 0.148; allele D, f = 0.281; allele E, f = 0.310; allele F, f = 0.472. These frequencies differ from previously reported frequencies in white populations,32 although they are closest to those of a United Kingdom population.31 In this BMT cohort all other polymorphisms examined displayed allele frequencies similar to those previously published.19,28,34,42 43 

Recipient IFNγIntron1-3/3 homozygous genotype was significantly associated with more severe aGVHD. Eight (38.1%) of 21 recipients possessing IFNγIntron1-3/3 genotype developed grades III to IV aGVHD, whereas 8 (13.6%) of 59 recipients with other alleles developed grades III to IV aGVHD (Tables1 and 5). Donor genotype did not associate with aGVHD (Table 2); neither were recipient nor donor IFNγIntron1 polymorphisms associated with cGVHD.

Recipients possessing an IL-6⁻¹⁷⁴G allele showed a strong trend toward development of higher grades of aGVHD than those with other genotypes, but not a significant increase in severe aGVHD (Tables 1 and 5). Recipients possessing an IL-6⁻¹⁷⁴G allele had a nonsignificantly increased incidence of cGVHD, whereas IL-6⁻¹⁷⁴GG homozygotes had a substantially greater incidence of cGVHD than recipients with other genotypes. Fifteen (65.2%) of 23 IL-6⁻¹⁷⁴GG homozygotes developed cGVHD, compared to 15 (30.6%) of 49 recipients with a IL-6⁻¹⁷⁴C allele (Tables 3 and 7). No significant relationship between donor IL-6⁻¹⁷⁴ genotype and aGVHD was demonstrated (Table 2), but donor IL-6⁻¹⁷⁴GG homozygous genotype exhibited a strong trend toward increased frequency of cGVHD. Ten (66.7%) of 15 patients with IL-6⁻¹⁷⁴GG homozygous donors developed cGVHD compared to 13 (37.1%) of 35 patients whose donors had an IL-6⁻¹⁷⁴C allele (Tables4 and 7). There was no association of the IL-6 3′ minisatellite of recipient or donor with aGVHD or cGVHD (data not shown).

TNFd3/d3 homozygous genotype (shared by HLA-matched recipient and donor due to the TNF gene locus position within the major histocompatibility complex class III region) associated with severe aGVHD. Nine (33.3%) of 27 patients with TNFd3/d3 genotype developed grades III to IV aGvHD, whereas only 7 (13.2%) of 53 patients not homozygous for the TNFd3 allele did so (Tables 1 and 6).

Possession of one or more IL-10⁻¹⁰⁶⁴i[12-16] alleles by the recipient associated significantly with overall aGVHD severity (Tables 1 and 5). Recipients possessing IL-10⁻¹⁰⁶⁴i[12-16] alleles were at significantly higher risk of severe aGVHD. Thirteen (29.5%) of 44 recipients possessing an IL-10⁻¹⁰⁶⁴i[12-16] allele developed severe aGVHD, compared to 3 (8.3%) of 36 who possessed only IL-10⁻¹⁰⁶⁴i[7-11] alleles (Tables 1 and 6).

By contrast, no associations were found between GVHD and a number of other polymorphisms, which have been implicated as genetic risk factors in other diseases with immunoregulatory abnormalities: TNFβ1 (NcoI-AspHI haplotype),44 CTLA4 (3′ A_nT_n microsatellite),45 TGFβ1 (⁻⁵⁰⁹ promoter region polymorphism).46 In each case polymorphic alleles showed similar frequency (ie, were similarly distributed) in both aGVHD and cGVHD groups, both for the patient and donor genotypes (data not shown).

Logistic regression confirmed the independent association of recipient IFNγIntron1-3/3 homozygous genotype with severe aGVHD, together with recipient possession of one or more IL-10⁻¹⁰⁶⁴i[12-16] alleles and TNFd3d3 homozygous genotype (Table 8). In addition to these genotypic factors, age was also significantly associated with severe aGVHD; forward stepwise modeling implicated recipient IFNγ, IL-10, and TNF genotypes in addition to age as aGVHD risk factors. IL-6⁻¹⁷⁴GG genotype was confirmed as a risk factor for cGVHD, together with age, gender mismatch (donor female and recipient male), and disease (CML), which have previously been reported as risk factors for cGVHD47 (Table9). Forward stepwise modeling implicated age, IL-6 genotype, and gender mismatch as cGvHD risk factors.

Earlier studies have shown that BMT recipients' possession of certain cytokine gene polymorphism alleles is associated with aGVHD and other inflammatory complications of BMT.33 34 The results of this extended study are consistent with this concept and further demonstrate other candidate gene polymorphisms, which show cumulative effects, as well as demonstrating correlation between cytokine genotype and cGVHD.

Recipients homozygous for IFNγIntron1 allele 3 (linked to lower in vitro IFNγ production by stimulated peripheral blood mononuclear cells) were more likely to develop severe aGVHD. Chronic GVHD showed no trend toward association with recipient or donor IFNγIntron1 genotype in this study, despite cGVHD largely following prior aGVHD. The finding of a recipient genotype linked in other studies to low in vitro IFNγ production associating with more severe aGVHD may have 2 possible explanations. First, it has been suggested that IFNγ may act in a negative feedback regulatory role, as seen in some mouse model studies of GVHD; twice weekly injections of IFNγ in a subacute murine model of GVHD prevented aGVHD, decreasing intestinal lesions and increasing survival.13 This was associated with a reduction in IFNγ–producing T cells, suggesting that exogenous IFNγ may be involved in down-regulation of GVHD via a negative feedback loop.48 Mice receiving transplanted cells from IFNγ knockout donors develop accelerated lethal aGVHD, also supporting a negative feedback role for IFNγ.14 15 

As an alternative explanation for our findings, the in vivo stimulus provided by BMT conditioning may well have a different relationship to the Intron1 polymorphism to that shown by the in vitro response to mitogens.19,20 Macrophage activation as a result of recipient tissue damage by TBI and cytotoxic chemotherapy, in concert with IFNγ production by T and natural killer cells, is pivotal in early conditioning-induced cytokine release.49-51 This network of cytokine interactions initiating aGVHD is more complex than direct in vitro T-cell activation by mitogens.52 The IFNγIntron1 polymorphism results indeed suggest that associations of cytokine gene polymorphic alleles with in vivo clinical outcomes do not necessarily correlate well with in vitro production data.

Recipients of BMT who were homozygous for the low producer IL-6⁻¹⁷⁴C allele had milder aGVHD overall, consistent with the postulated genetic protection from IL-6 release. Chronic GVHD showed significant association with IL-6 genotype, largely to that of recipient but with a trend in respect to donor genotype, despite smaller numbers of available donor samples. The gene dosage effect of possession of the IL-6⁻¹⁷⁴G allele, with homozygotes developing more GVHD, both in terms of aGVHD severity and cGVHD incidence, is also supportive of a true biologic relationship between genetic background and GVHD, especially with reference to in vitro and in vivo production data.28 Intriguingly, there has been little previous examination of the potential role of IL-6 in cGVHD, and the association of functionally associated IL-6 genotypes with cGVHD may point toward future areas of study.

The IL-6 3′ minisatellite polymorphism did not associate with GVHD despite a degree of linkage of the IL-6⁻¹⁷⁴C allele to the F allele at the 3′ (AT)-rich minisatellite. This may suggest that the promoter region polymorphism is more functionally relevant in BMT. In examining the relationship between IL-6 genotype and GVHD in other cohorts it will be important to bear in mind the differing allele frequencies of the IL-6 polymorphisms in various populations. The IL-6⁻¹⁷⁴C allele is uncommon in Afro-Caribbean populations,28 a group some have suggested have a higher incidence of GVHD.53 54 

The relatively small numbers of patients in our study do not allow strong conclusions regarding the relative impact of cytokine genotype-associated GVHD risk with respect to established risk factors such as age, gender mismatch, or CMV status, but would suggest that they are potentially of a similar magnitude (odds ratio between 3 and 5). The absence of an overall relationship between additional immunosuppresssion with MTX/steroids and GVHD probably reflects the targeted use of prophylaxis in our BMT unit.55 The confirmation of significant association of acute and chronic GVHD with these cytokine genotypes suggests that further studies should be undertaken in a prospective fashion. Ideally the association of these cytokine gene polymorphisms should also be tested in other settings, such as mismatched or unrelated donor BMT, and with different prophylaxis regimens, to examine the relative importance of cytokine gene polymorphisms together with other parameters including high-resolution HLA typing, minor histocompatibility antigens, and established clinical risk factors for GVHD. It will be important to examine potential relationships of cytokine genotype with relapse, because an immunomodulatory factor that reduces alloreactivity against host tissues may also do so against residual host malignancy. Cytokines mediate and regulate other BMT complications such as septic shock, multiorgan dysfunction, and interstitial pneumonitis; studies are ongoing in relation to some of these complications.

Relationships demonstrated thus far support the attempted construction, in a prospective fashion, of a GVHD risk index integrating clinical and immunogenetic factors (related to both histocompatibility and nonhistocompatibility including cytokine genotyping). Any risk index will ideally attempt to accurately assign GVHD risk to all patients, but optimizing their prophylaxis may depend on considerations of other factors. Hence, clinical use of cytokine genotype data via an aGVHD risk index may not only be dependent on transplant type (sibling/unrelated, matched/mismatched, marrow/stem cells/cord) but also may be influenced by the indication for BMT and other clinical risk elements.

In conclusion, our findings confirm the principle that the recipient response is critical in BMT outcome, particularly in relation to aGVHD, and that this response involves a substantial genetic component. In addition to reconfirming the association of TNF andIL-10 gene polymorphism alleles, 2 new candidate polymorphisms in IFNG and IL6 genes are shown to associate with aGVHD severity, and in the case of IL6 with cGVHD incidence. Multivariable analysis suggests that these genetic polymorphisms may confirm a similar magnitude of increased GVHD risk to certain other established risk factors, but further studies are clearly needed, ideally involving other BMT groups such as those with mismatched and unrelated donors. Combination of recipient aGVHD severity-associated genotypes increases discrimination as to aGVHD severity. These findings would support the construction of BMT- and prophylaxis-specific genotypic aGVHD risk indices, which could be tested prospectively in combination with established GVHD risk factors. PCR-based cytokine genotyping of recipients and donors could easily be performed alongside HLA typing, when donor options are being assessed and decisions regarding prophylaxis made. A recipient GVHD risk index including cytokine genotype could be used as a guide to more individually tailored GVHD prophylaxis, particularly in combination with other risk factors.