Both donor and recipient NOD2/CARD15 mutations associate with transplant-related mortality and GvHD following allogeneic stem cell transplantation

Acute graft-versus-host disease (GvHD) is still a major and the most severe complication following allogeneic stem cell transplantation (SCT). Although activation of donor T lymphocytes by minor or major HLA differences plays a pivotal role in induction of GvHD, there is increasing evidence of involvement of cytokine-mediated inflammation.^1,2  Experimental models indicate the primacy of gastrointestinal damage: conditioning-related damage of the intestinal epithelium results in bacterial translocation followed by increased cytokine release by macrophages/monocytes and T-cell activation.³  Clinically, gastrointestinal GvHD is the most severe manifestation of acute GvHD, occurs only in a subgroup of patients, and frequently precedes multiorgan complications.

The NOD2/CARD15 protein has been described as an intracellular sensor to muramyl dipeptide (a compound of the bacterial cell wall) mediating consecutive nuclear factor κ B (NF-κB) activation⁴  and belongs to a large family of proteins involved in intracellular pathogen recognition.⁵ NOD2/CARD15 expression is—as known so far—restricted to intestinal epithelial cells (IECs) and cells of monocyte/macrophage lineage.^5,6  A variety of single nucleotide polymorphisms (SNPs) in the NOD2/CARD15 gene have been reported. Within these SNPs, high-risk alleles in SNPs 8, 12, and 13 have been identified which confer an increased risk for Crohn disease.^7,8  Patients heterozygous for these mutated forms have a 2- to 4-fold increased risk to develop Crohn disease, and the risk increases to 20- to 40-fold in homozygous patients.⁹  Based on the homologies in the pathophysiology of GvHD and inflammatory bowel disease, we speculated that the genetic background shown for NOD2/CARD15 polymorphisms might also have an impact on the pathophysiology of GvHD.

Included in the study were 169 consecutive patients admitted at our transplant unit for allogeneic SCT between June 1998 and July 2003. Conditioning and prophylactic immunosuppression were performed according to standard protocols; reduced-intensity conditioning (RIC) consisted of fludarabine, BCNU (1,3-bis(2-chloroethyl)-1-nitrosourea), and melphalan (Table 1).¹⁰  There were 78 patients who received grafts from HLA-identical siblings (matched related donors [MRDs]), 87 who received grafts from unrelated donors (URDs), and 4 patients who received grafts from one-HLA-antigen–different relatives (RDs). Within the unrelated donor group, 64 donor/recipient pairs were matched for HLA-A, HLA-B, DRB1, and DQB1 as determined by low-resolution typing for class I and high-resolution typing for class II, whereas 1- or 2-allele mismatches for DRB1 and DQB1 had been accepted in the remaining 23 pairs. GvHD was graded according to Glucksberg criteria¹¹  and major outcome variables such as transplant-related mortality (TRM; death in remission) and causes of death were recorded in a database updated monthly. Median observation time for surviving patients or patients dying from relapse was 450 days (range, 33-1767 days); median observation time for patients dying from TRM was 172 days (range, 14-1065 days).

DNA from 169 patients and from 168 of their donors was collected, frozen, and retrospectively analyzed to assess the role of NOD2/CARD15 mutations. DNA from donors and recipients had been prepared from peripheral blood cells at the time of admission for SCT. At admission and prior to stem cell collection, patients and donors had given their informed written consent to prospective collection of DNA samples for assessment of genetic risk factors associated with GvHD and TRM by SNP analysis for cytokines and inflammatory proteins. The collection of samples had been approved by the local ethics committee (institutional review board approval no. 02/220).

Single nucleotide polymorphisms 8, 12, and 13 were determined by a Taqman protocol.¹²  Control samples that were confirmed by sequencing were included in each Taqman run. The primers and probes were synthesized by MWG Biotech (Ebersberg, Germany). Probes were synthesized with reporter dye 6-FAM or TET covalently linked at the 5′ end and a quencher dye TAMRA linked to the 3′ end of the probe. The optimized reaction mix of a final volume of 10 μL consisted of Universal Master Mix (PE Applied Biosystems, Foster City, CA), 400 nM of each forward and reverse primer, 250 nM of each fluorogenic probe, and the DNA template (20 ng/well). All reactions were performed in a 384-well plate (Abgene, Epsom, United Kingdom) and amplified using the thermocycler Primus-HT (MWG Biotech). The cycle conditions were 50°C for 2 minutes, then 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds (melting step) and 60°C for 1 minute (anneal/extend step). After the polymerase chain reaction (PCR) run, the released fluoresence was measured by the ABI PRISM 7700 Sequence Detection System (PE Applied Biosystems).

Increases in the amount of reporter dye fluorescence during the 40 cycles of amplification were monitored using Sequence Detector software (SDS version 1.6; PE Applied Biosystems). Changes in reporter dye fluorescence of 6-FAM versus changes in reporter dye fluorescence of TET were plotted (homozygous FAM versus homozygous TET versus heterozygous FAM/TET).

At the time of availability of the NOD2/CARD15 assay for our patient population, we had collected DNA from 169 consecutive patients and their respective donors with an actual TRM of 33% observed in these patients. Based on the distribution of unmutated and mutated NOD2/CARD15 SNPs in the control population reported by Hampe et al,¹²  we assumed a 2:1 ratio of patients with wild-type or mutated SNPs. We further assumed that occurrence of NOD2/CARD15 mutations should at least double the incidence of TRM from 25% to 50% or result in a hazard ratio of 2.0 between patients with wild-type SNPs and those with mutated SNPs. Given a doubling of TRM, we calculated that the available sample size of 169 would be sufficient to detect a difference concerning TRM with a significance of P equal to .05 and a power of 84%. Given a hazard ratio of 2.0, the difference would be detected with a power of 72%. Calculations were performed using the respective models of the PASS software (version 2000; NCSS, Kaysville, UT).

Three major outcome variables (TRM at 365 days is equal to 1-year TRM; overall grade of severe acute GvHD, grade III/IV; and stage of gastrointestinal GvHD) were analyzed in relation to NOD2/CARD15 mutations. Cumulative incidence reflecting the competing risks of death from other toxicities or relapse for GvHD, and death from relapse for TRM were calculated for each of these parameters. Time to the clinical event (TRM, GvHD) was measured from the date of SCT. For overall GvHD and gastrointestinal GvHD, events occurring until day 100 were included in the model. For 1-year TRM, events occurring until day 365 after SCT were included in the model. Stepwise multivariate Cox regression models were adjusted testing the independent prognostic relevance of NOD2/CARD15 mutations. The limit for reverse selection procedures was 0.2. For the multivariate comparison of major risk factors for TRM and GvHD, only age of the patient, disease stage at the time of transplantation, type of donor (HLA-identical sibling vs URD vs RD), and NOD2/CARD15 mutation status were considered. For analysis of NOD2/CARD15 SNPs, the less frequent high-risk alleles associated with Crohn disease and an impaired NF-κB production were defined as mutated, whereas the more frequently observed alleles were referred to as wild-type alleles. For analysis of mutations in relation to clinical events, groups were formed according to occurrence of (1) any mutations in recipients or donors and (2) any mutations in recipients only, donors only, or in both.

Cumulative incidence allowing adjustment for competing risks was calculated using NCSS software (version 2004), further statistical analysis was performed by the help of SPSS software (version 11.05, Chicago, IL).

NOD2/CARD15 mutations occurred with a frequency of 21.8% in patients and 13.7% in donors typed in this study. A homozygous mutation was observed in only one patient and in his HLA-identical donor, whereas all other patients and donors with mutations were heterozygous. Calculated haplotype frequencies for mutated SNP 8 were 0.056 for recipients (R) and 0.045 for donors (D); for mutated SNP 12 were 0.021 (R) and 0.009 (D); and for mutated SNP 13 were 0.027 (R) and 0.034 (D). Thus, the haplotype frequencies were close to the range of controls reported in studies on Crohn disease¹²  (SNP 8: 0.0478; SNP 12: 0.0072; and SNP 13: 0.0406). This resulted in an overall percentage of 70.5% of transplantations where both donors and recipients had unmutated SNPs (wild-type group). In 22 pairs (13%), only recipients had mutated SNPs (R mut); in 16 pairs (9.5%), mutations occurred in donors only (D mut); and in 12 pairs (7%), both donors and recipients revealed mutations (R & D mut). In our consecutive cohort of patients, patient- and donor-specific characteristics as well as transplant-specific procedures were equally distributed between patient/donor pairs with and without mutations (Table 1).

Patients were followed for a median of 16 months (range, 0.5 to 59 months). We first compared the outcome of patients from wild-type recipient/donor pairs (n = 119) with the outcome of patients in pairs with any type of mutations (n = 50). Overall, severe GvHD grade III/IV, severe gastrointestinal GvHD, and TRM at one year were significantly increased in pairs with NOD2/CARD15 mutations as compared with the wild-type group (Table 2). Deaths from relapse occurred in both groups: 18 of 119 patients with wild-type SNPs and 7 of 50 patients with mutated pairs (data not shown).

When mutations were split according to their occurrence in recipients or donors or both (Table 3; Figure 1), there was an increased risk of overall severe GvHD in recipients with donor mutations and a dramatic rise for both overall and gastrointestinal GvHD in the small subgroup of pairs with mutations of both recipients and donors. As a consequence, cumulative incidence of 1-year TRM increased from 20% in wild-type recipient/donor pairs to 49% in pairs with recipient (P = .03), to 59% in pairs with donor (P = .004), and to 83% if donor and recipient revealed mutations (P = .001). Again, deaths from relapses were almost equally distributed within these subgroups (data not shown).

The strong association of TRM with NOD2/CARD15 mutations was partially explained by an increased risk not only for death from GvHD but also from respiratory failure due to diffuse primary or secondary pulmonary damage. Whereas in the wild-type group only 11 of 30 (37%) deaths in remission were caused by GvHD and respiratory failure, the proportion increased to 17 of 26 (65%) in pairs with NOD2/CARD15 mutations.

We next compared the relevance of donor and recipient mutations in the different immunogenetic subgroups included in this analysis. Although the associations of TRM with NOD2/CARD15 mutations were more prominent for HLA-identical sibling transplantations, there were parallel trends for URD transplantations (Table 4; Figure 2). Within the URD group, cumulative incidence of 1-year TRM rose from 27% in wild-type pairs to 55% in mutated pairs, if recipients and donors were matched for HLA-A, HLA-B, DRB1, and DQB1. TRM increased from 29% to 50% in pairs with one or 2 DRB1- or DQB1-allele mismatches (data not shown). Studies in larger cohorts of unrelated donor transplantations are clearly needed to assess the definitive role of NOD2/CARD15 mutations in unrelated donor transplantation and in relation to the class I and class II allele or even antigen mismatches. The small group of antigen mismatched–related transplantations showed a high mortality and did not allow any assessment of the role of NOD2/CARD15 mutations in this particular subgroup.

In multivariate Cox regression analyses, age, donor type, and NOD2/CARD15 mutations in donors alone or in both recipients and donors were independent risk factors for TRM (Table 5). The hazard ratios for TRM following allogeneic SCT concerning NOD2/CARD15 mutations were 2.4 (P = .02; 95% confidence interval [CI] 1.1-5.0) for recipient mutations, 2.5 (P = .02; 95% CI 1.2-5.4) for donor mutations, and 6.0 (P = .000; 95% CI 2.6-14.1) in the case of simultaneous mutations in donors and recipients. When we included overall GvHD grade II-IV as a risk factor for TRM in our model, GvHD (P = .006) as well as NOD2/CARD15 mutations (P = .000) remained significant risk factors for TRM. Even after inclusion of severe GvHD grade III/IV, the contribution of NOD2/CARD15 remained significant (data not shown), clearly confirming the strong and independent impact of NOD2/CARD15 mutations on the outcome of allogeneic SCT.

Our study is the first to report a strong association of mutations in SNPs of the NOD2/CARD15 gene with severe GvHD, gastrointestinal GvHD, and TRM following allogeneic SCT. So far, only one casuistic report on de novo occurrence of Crohn disease after allogeneic SCT suggested that donor mutations might transfer the risk of Crohn disease to the patients and result in an increased rate of complications.¹³  Our findings, however, indicate a major direct impact of NOD2/CARD15 mutations in both recipient and donor cells on outcome following allogeneic SCT from related as well as unrelated donors. Although the exact contribution of each single SNP mutation and, more importantly, the various rare combinations of donor and recipient mutations can only be defined after analysis of large cohorts of patients, our data allow identification of patient/donor pairs with an increased risk of GvHD and subsequent death from TRM. Thus our findings should have major implications for pathophysiologic concepts of GvHD as well as the clinical management of patients.

As well known for toll-like receptors (TLRs) on the cell membrane, NOD2/CARD15 proteins are reported to play a major role in the intracytoplasmatic response to bacterial cell wall products (muramyl dipeptide) and subsequent NF-κB activation in monocytes/macrophages as well as in specialized epithelial cells of the ileum.^4,14,15 A contribution of an altered epithelial reactivity in the ileum is not unexpected in GvHD, and an impaired antimicrobial response with diminished defensin production¹⁵  and increased translocation of bacteria as shown in experimental models¹⁶  could well contribute to excess morbidity in pairs with mutations of the recipient or in both recipient and donors. The additional and strong donor associations are surprising for several reasons. Recipient monocytes are rapidly replaced by donor cells in allogeneic SCT, and our data indicate a broader role of donor monocytes and their reactivity to bacterial toxins in GvHD. So far, the interaction of bacterial toxins and monocytes/macrophages in GvHD has been explained by up-regulation of inflammatory cytokine responses in IFNγ–sensitized macrophages.^17-19  The association of NOD2/CARD15 mutations characterized by a diminished NF-κB response, however, with GvHD and TRM suggest a predominant role of an impaired monocyte/macrophage response resulting in dysregulation of inflammation. The exact mechanisms of interaction of a diminished monocyte/macrophage response and altered behavior of epithelial cells resulting in a possible defect in the defense against bacteria, and T-cell activation in allogeneic SCT remains to be eludicated. Our findings, however, strongly remind us of van Bekkum and Knoan's results,²⁰  which suggested that translocation of intestinal microflora directly interferes with T-cell activation and donor T-cell expansion. The predominant role of the gastrointestinal tract in T-cell activation is further supported by a recent study showing that Peyer patches²¹  are essential for induction of T-cell responses in murine GvHD.

Besides Crohn disease and other inflammatory bowel diseases, the role of an impaired monocyte/macrophage response due to NOD2/CARD15 mutations has so far only been discussed in the context of spondylarthritis^22,23  and other immune-mediated diseases like psoriatic arthritis and lupus erythematodes.^24,25  Our data, however, suggest not only an increased mortality from GvHD in patients bearing NOD2/CARD15 mutations but also from respiratory failure. Thus, it might well turn out, that, in addition to gastrointestinal inflammation, monocyte/macrophage dysfunction in organs like the lung is affected by NOD2/CARD15 function and polymorphisms thus contributing to pathophysiology of pulmonary complications. In allogeneic SCT, a gut-liver-lung axis¹  of impaired elimination of intestinal lipopolysaccharide (LPS) by the liver has been postulated to be involved in development of pulmonary damage. Direct involvement of an impaired monocyte/macrophage or even epithelial response in the lung, however, might be an alternative explanation and should be addressed in careful analyses in future studies.

Clinically, our data suggest that avoiding a stem cell graft from a donor bearing mutations, at least if the patient has mutated SNPs, might reduce the risk of developing severe complications. This is also strongly supported by the results of multivariate analysis. As NOD2/CARD15 mutations are not linked to HLA and the associations described in our study turned out to be true for both related and unrelated donors, our results might have a future influence on typing and donor selection strategies. Of course, assessment of NOD2/CARD15 mutations in further cohorts of donor/recipients in both related and especially unrelated SCT is required to verify this hypothesis. Our data strongly argue for a detailed analysis comparing the relevance of allele matching with assessment of NOD2/CARD15 mutations in unrelated donor transplantation.

In summary, our data describe NOD2/CARD15 mutations as a new and highly significant genetic factor allowing risk assessment in a subgroup of patients receiving allogeneic SCT. As discussed for Crohn disease, consideration of GvHD as a result of a monocyte/macrophage defect resulting in impaired tolerance against bacteria rather than of uncontrolled excess stimulation should widen pathophysiologic and hopefully also prophylactic and therapeutic concepts of this major complication of allogeneic SCT.

The authors wish to acknowledge the skillful help of Heike Bremm, Division of Hematology/Oncology, University of Regensburg, in collecting samples and DNA from recipients and donors. They acknowledge provision of positive controls for NOD2/CARD15 typing by Stefan Schreiber, Department of Internal Medicine, University of Kiel.