Prognostic significance of NOD2/CARD15 variants in HLA-identical sibling hematopoietic stem cell transplantation: effect on long-term outcome is confirmed in 2 independent cohorts and may be modulated by the type of gastrointestinal decontamination

Ernst Holler, Gerhard Rogler, Julia Brenmoehl, Joachim Hahn, Hans Herfarth, Hildegard Greinix, Anne M. Dickinson, Gerard Socié, Daniel Wolff, Gottfried Fischer, Graham Jackson, Vanderson Rocha, Beate Steiner, Guenther Eissner, Jeorg Marienhagen, Juergen Schoelmerich, Reinhard Andreesen


To assess the role of NOD2/CARD15 variants on the long-term outcome of allogeneic stem cell transplantation in a genetically homogeneous group, we extended our previous study (cohort I, n = 78) and typed DNA for NOD2/CARD15 single nucleotide polymorphisms (SNPs) from an additional 225 recipients and their HLA-identical sibling donors (cohort II) treated at four other European centers. Results of genotyping were compared with clinical outcome. The strong association of NOD2/CARD15 variants with transplantation-related mortality (TRM) was confirmed in univariate and multivariate analysis; TRM increased from 20% in cohort I/22% in cohort II in recipient/donor pairs without any NOD2/CARD15 variants to 47% in cohort I/32% in cohort II in the presence of one variant in either donor or recipient and further to 57% in cohort I/74% in cohort II in the presence of 2 or more variants (P < .002 in both cohorts). NOD2/CARD15 SNPs were not associated with relapse rate but had a strong impact on overall survival. In an analysis of center effects, the type of gastrointestinal decontamination was the only factor interfering with the prognostic significance of NOD2/CARD15 SNPs. Our data further support an interaction between gastrointestinal defense mechanisms, activation of the innate immune system, and specific transplant-related complications.


Pathophysiologic concepts of graft-versus-host disease (GvHD) have been based on several hypotheses. Although there is no doubt that activation of donor T lymphocytes is the central event in the alloreactivity, the concomitant interaction of the innate immune system also plays an important role and has been conceptually addressed by several investigators. The findings of van Bekkum et al1 on prevention of GvHD by breeding mice under germfree conditions initiated a longstanding discussion on the role of gnotobiotic prophylaxis and the mechanisms of interaction between microbial and specific T-cell activation. With description of cytokines as major effectors2,3 of at least acute GvHD, the endotoxin hypothesis postulated that the translocation of bacterial toxins across the gastrointestinal mucosa gave rise to costimulation of cytokine release and an alloimmune response. However, the findings of Beelen et al,4 demonstrating a strong effect of elimination of anaerobic bacteria on GvHD in patients, and the recent observations that mice lacking Peyer patches5 or receiving prophylaxis with lactobacilli6 have a strongly reduced incidence of GvHD, suggested again a more direct interaction of bacterial ligands and activation of specific immune responses.

Recently, our group reported a significant association between the occurrence of single nucleotide polymorphisms (SNPs) within the NOD2/CARD15 gene and GvHD as well as transplantation-related mortality (TRM).7 The NOD2/CARD15 gene had been identified as the first “susceptibility” or “disease” gene in Crohn disease, a form of inflammatory bowel disease.8,9 It codes for an intracytoplasmic sensor of the bacterial cell-wall compound muramyl-dipeptide (MDP). Ligation of MDP to NOD2/CARD15 induces a signal transduction cascade finally leading to activation of the transcription factor NF-κB and subsequent inflammatory response in epithelial cells of the ileum10 as well as in monocyte/macrophage-derived cells. Because the first series analyzed consisted of a heterogeneous group of allogeneic HLA-identical sibling and unrelated donor hematopoietic stem cell transplantations (HSCTs), it did not allow a more detailed analysis within these genetically different transplantation groups. Therefore, we now have performed an analysis focusing only on HLA-identical sibling transplantations including patients from the first cohort and a larger, second and independent cohort of patients, and addressed the impact of NOD2/CARD15 SNPs on long-term transplantation outcome. The strong impact of NOD2/CARD15 variants on acute GvHD and early TRM was confirmed and translated as a strong impact on long-term overall survival (OS). Furthermore, our data suggest an interference of NOD2/CARD15 genotype with prophylactic use of gastrointestinal decontamination.

Patients, materials, and methods

Patients and donors

A total of 303 consecutive patients from five different centers were included in this analysis, the cohort of 78 patients from Regensburg (center 1; cohort I) had been partially included in our first single-center report on the role of NOD2/CARD15 in allogeneic HSCTs but so far had not been analyzed for long-term outcome. To these 78 patients, 225 patients (cohort II) were added from the centers of Vienna (center 2; n = 38), Newcastle (center 3; n = 90), Rostock (center 4; n = 37), and Paris (center 5; n = 60). All patients received grafts from HLA-identical sibling donors. Patients had been treated according to center-specific protocols of conditioning regimens, immunosuppressive prophylaxis, and gastrointestinal decontamination. Major patient characteristics and outcome variables were reported and analyzed based on uniform case record forms (CRFs). In all centers, patients and their HLA-identical sibling donors gave informed consent approved by local ethics review boards to analyze genetic risk factors associated with GvHD and outcome following allogeneic SCT. Approval was obtained from the University of Regensburg Medical Center's and from the other participating centers' Institutional Review Board for these studies. Detailed characteristics of patients and transplantation-specific strategies are given in Table 1. Strategies of gastrointestinal decontamination were grouped according to efficacy against Gram-positive bacteria. Type 1 decontamination included no decontamination at all or decontamination addressing mainly Gram-negative and anaerobic bacteria (ciprofloxacin and late metronidazole starting from day +1); type 2 decontamination also covered Gram-positive bacteria either via the use of amoxicillin and ciprofloxacin (center 5) or via oral vancomycin and nonabsorbable polymyxins (center 2). The median follow-up time was 30 months (range, 1-220 months) in surviving or censored patients and 6 months (range, 0-36 months) in patients dying from complications.

View this table:
Table 1.

Characteristics of patients according to transplantation centers and cohorts

Typing for NOD2/CARD15 SNPs

Prior to admission, DNA from recipients and their respective donors had been isolated from EDTA blood and stored until central analysis at the University of Regensburg. In some centers, EBV cell lines from donors and recipients had been established and DNA was isolated from cryopreserved EBV cell lines. TaqMan polymerase chain reaction (PCR) for SNPs 8, 12, and 13 of the NOD2/CARD15 gene was performed as previously described.11

Analysis of data

An SPSS 12.0 database was established based on the CRFs, which included the typing results for NOD2/CARD15 and was used for analysis of frequencies, survival data, and risk factors. Cox regression analysis for TRM and OS included NOD2/CARD15 status, cohort and center, age, status of disease at the time of HSCT, sex combination (female donor in male recipients versus others), intensity of conditioning (standard versus reduced intensity), stem cell source, and type of gastrointestinal decontamination. The NOD2/CARD15 status of the patients was grouped according to the number of NOD2/CARD15 variants observed in an individual donor-recipient pair: wild-type was absence of any variant in recipient and donor; one variant was one heterozygous variant in either recipient or donor; 2 or more meant compound heterozygous or homozygous variants in recipients (n = 1) or donors (n = 1) or in both (n = 26).

Significance testing was performed by using the likelihood ratio test (P < .05). For calculation of TRM and incidence of GvHD and relapse, the cumulative incidence method with either relapse or non–relapse-related mortality treated as competing risks was used with the help of NCSS 2004 software (NCSS, Kaysville, UT). All group comparisons were done using the Fisher exact test. P < .05 was considered as statistically significant.


Distribution of NOD2/CARD15 variants and number of NOD variants in the transplantation cohorts

As shown in Table 2, the distribution of NOD2/CARD15 variants within recipients and donors as well as the distribution of the number of NOD2/CARD15 variants were similar in both cohorts of patients.

View this table:
Table 2.

Distribution of major NOD2/CARD15 variants according to cohorts analyzed

Effect of NOD2/CARD15 variants on GvHD III/IV, 1-year TRM, and overall TRM in both cohorts

When we checked the impact of NOD2/CARD15 variants on major outcome variables, similar associations were observed in both cohorts. Although the relative increases in incidences seemed to be more pronounced in cohort I as compared to cohort II, results were uniformly significant for all major parameters (Table 3) including GVHD III/IV, gastrointestinal GvHD, 1-year TRM, and overall TRM. Causes of death were comparable in both cohorts. Death could be directly attributed to GvHD in 24%, to major bacterial or viral infection in 19%, to fungal infections in 10%, to primary or secondary respiratory failure in 35%, and to organ toxicities in 12%. Within patients with NOD2/CARD15 mutations, there was a trend to a higher proportion of patients dying from respiratory failure (data not shown).

View this table:
Table 3.

Number of NOD2/CARD15 variants and outcome in both cohorts of patients

Multivariate Cox regression analysis including major known risk factors such as older age (> 40 years), diagnosis, status of disease at the time of HSCT, sex combination, type of gastrointestinal decontamination, intensity of conditioning, and stem-cell source confirmed the impact of NOD2/CARD15 status, especially for the high-risk group bearing 2 or more variants, as the only significant and independent risk factors for TRM in both cohorts of patients (Table 4).

View this table:
Table 4.

Multivariate risk factor analysis for overall TRM

Long-term effects: association of NOD2/CARD15 variants with TRM demonstrates a strong impact on OS

In contrast to TRM, NOD2/CARD15 status had no major impact on relapse incidence and relapse-related mortality. For all patients, cumulative incidence of relapse-related mortality was 14% (95% CI, 10%-20%) in recipient and donor without NOD variants, 19% (95% CI, 11%-36%) in recipient and donor with one heterozygous variant, and 11% (95% CI, 4%-32%) in transplants with 2 or more variants. In line with this, the incidence of overall chronic GvHD at 1 year after SCT showed only a trend for a higher rate of chronic GvHD in NOD2/CARD15 variants but did not reach statistical significance (data not shown).

Thus, association of NOD2/CARD15 status with TRM demonstrated a strong and highly significant impact on OS for the whole group of patients (Figure 1) and also for individual cohorts (wild-type cohort I 60%, cohort II 53%; 1 variant: cohort I 15%, cohort II 47%; 2 or more variants: cohort I 0%, cohort II 21%; log rank for cohort I, P < .001; for cohort II, P < .02). In multivariate analysis, NOD2/CARD15 status remained the only significant factor associated with survival in both cohorts whereas advanced stage was of significance (HR 2.67; 95% CI, 1.18-5.96) only in cohort I (Table 5).

View this table:
Table 5.

Multivariate risk factor analysis for overall long-term survival

When OS was analyzed in relation to the occurrence of isolated or combined recipient or donor variants, a similar pattern as seen for TRM in the first report was observed. Whereas 221 patients with wild-type NOD2/CARD15 of recipients and donors had an OS rate of 56%, a strong impact of either recipient (n = 27, OS 23%, P = .005) or simultaneous recipient and donor variants (n = 26, OS 17%, P = .001) was seen. If only the donors possessed NOD2/CARD15 SNPs, then OS was intermediate (40%, no significant difference from the wild-type group).

Figure 1.

Overall survival according to the number of NOD2/CARD15 variants P = .02 for one variant versus wild-type, P < .06 for 2 variants versus one variant, P ≤ .001 for 2 variants versus wild-type (Kaplan Meier, log rank).

Center effects may be explained by the type of gastrointestinal decontamination

Because associations of NOD2/CARD15 seemed to be less stringent in cohort II when compared to our first series, we checked for the prognostic role of NOD2/CARD15 in individual centers. Although influenced by different numbers of patients, the overall role of NOD2/CARD15 seemed to be consistently strong in 3 centers but less clear in 2 others. Significant associations for both TRM and OS were observed in centers 1, 3, and 4, whereas minor trends occurred in centers 2 and 5 (Table 6).

View this table:
Table 6.

Center effects for the association of the number of NOD2/CARD15 variants and overall TRM

Because patient and transplant characteristics showed major differences among centers, we next checked whether these factors interfered with the prognostic significance of NOD2/CARD15. After stratification for risk factors differing between participating centers such as age of the patient, status of disease at the time of HSCT, underlying disease, standard versus reduced-intensity conditioning, and marrow versus peripheral-blood stem cells (PBSCs) as stem-cell source, NOD2/CARD15 remained a significant risk factor for GvHD and overall TRM in the subgroups. However, after stratification for the type of intestinal decontamination, the clear effect of NOD2/CARD15 variants on GvHD and overall TRM was only seen in patients receiving decontamination with ciprofloxacin or no decontamination at all (as used in centers 1, 3 and 4) but was strongly reduced in patients receiving decontamination including Gram-positive bacteria (as used in centers 2 and 5; Figure 2). As a control, association of TRM and survival with the described risk factors was tested in the high-risk group bearing 2 or more NOD2/CARD15 variants only. Again, only the type of decontamination reached significance in multivariate analysis (HR for decontamination including Gram-positive bacteria 0.08 (0.01-0.79 95% CI, P < .03).

Figure 2.

Major impact of the type of gastrointestinal decontamination on prognostic significance of NOD2/CARD15 variants. Differences were highly significant (P < .001) for patients receiving Gram-negative decontamination or no decontamination at all, but did not reach significance in patients with combined Gram-negative and Gram-positive decontamination. Numbers in parentheses indicate the number of patients analyzed for each type of gastrointestinal decontamination. (A) GvHD grade III/IV and type of decontamination (n = 289, 14 patients were excluded from analysis due to early death). (B) Overall TRM and type of decontamination (n = 303).


After our first report on an association of NOD2/CARD15 variants with GvHD and TRM following allogeneic SCT, we now were able to extend our observations to a second independent cohort and to analyze for the first time a genetically homogeneous group of HLA-identical sibling transplant recipients and their respective donors from 5 different European centers. Our data not only confirm the previously reported associations in a large second cohort but also add substantial new information on the role of NOD2/CARD15 SNPs with regard to long-term outcome, graft-versus leukemia effects, and the potential role of gene dosage and interfering factors.

Within the large and genetically homogeneous group of HLA-identical sibling transplant recipients, we analyzed the long-term effect of NOD2/CARD15 mutations on transplantation outcome. Many strategies or factors associated with GvHD are inversely and directly linked to the major therapeutic principle of allogeneic SCT and the graft-versus-leukemia (GvL) effect, and patients with severe GvHD have a low risk of relapse and vice versa.12,13 In contrast, analysis of survivors with NOD2/CARD15 variants indicated an almost identical cumulative incidence of relapse in all cohorts. As a result, the NOD2/CARD15 status had a strong and independent influence on OS. This influence was also stronger than the role of the otherwise most established clinical risk factor, the status of disease at time of HSCT. Thus, our data give rise to the hypothesis that the part of the pathophysiology of GvHD, which links the interactions of microbial load with innate immunity and the activation of the specific immune response, is not directly involved in GvL effects. This hypothesis would be in line with experimental results suggesting that modulating the inflammatory pathway induced by LPS prevents GvHD but spares GvL effects.14 Furthermore, absence of cross-reaction between bacterial antigens in the gut and tumor antigens may also be derived from careful analysis of risks for various non–gastrointestinal cancers in patients suffering from Crohn disease because these studies did not reveal any protective effect of intestinal inflammation.15

So far, we do not know exactly how an impaired function of NOD2/CARD15 may mediate GvHD and TRM. The paradigm of a (at least partial) NOD2/CARD15 defect and an increased susceptibility to inflammation has been extensively discussed in the setting of Crohn disease,16 because both gain of function with increased IL-1β–driven inflammation17 and loss of function with deficient NF-κB response, and NF-κB–dependent inflammation18 and bacterial killing19 have been reported in patients and mice with NOD2/CARD15 variants. However, even given that patients with NOD2/CARD15 variants are susceptible to uncontrolled inflammation, the cellular players are still not clear; NOD2/CARD15-driven inflammation might directly contribute to increased alloreactivity by delivering altered danger signals to antigen-presenting cells thus affecting T-cell activation. Alternatively, NOD2/CARD15 variants may have a more indirect effect via a dysregulated macrophage/monocyte or epithelial response to microbial challenges.

The hypothesis of a role of specific bacterial pathogens in gastrointestinal inflammation and subsequent GvHD was raised by the reports from Beelen et al4 on the role of elimination of anaerobic bacteria. The observed association of the type of gastrointestinal decontamination and the prognostic significance of NOD2/CARD15 status in our present study provides some additional evidence for the role of specific pathogens, because they suggest that a Gram-positive bacterial population might be relevant for activation of NOD2/CARD15-mediated or -regulated inflammation. Because MDP, the specific ligand of NOD2/CARD15, is a bacterial cell-wall compound of both Gram-negative and Gram-positive bacteria,20 the hypothesis seems reasonable. Of course, our conclusion is based on a relatively small number of patients and confirmation of the role of decontamination in relation to NOD2/CARD15 variants in a larger prospective series of patients is urgently needed.

In our previous paper, we reported the strongest association of NOD2/CARD15 variants with outcome in patients with combined donor/recipient variants. Combined donor/recipient variants may have the strongest impact on GvHD and TRM because of the additive effects of damage in affected cells, for example, recipient epithelium and donor monocytes but also represent an additive effect of mutation dose in macrophage-derived cells of the donor and the recipient. Because mutation dose of NOD2/CARD15 has been shown to have a strong impact not only on the relative risk of Crohn disease but also on functional defects like MDP-induced IL-8 secretion,21-23 we preferred to perform the present analysis based on the number of variants present in an individual donor/recipient pair and were able to show strong cumulative effects. However, the evaluation of the exact contribution of donor and recipient cells and the respective cellular components will require sophisticated mouse experiments addressing isolated NOD2/CARD15 deficiencies in donors and recipients and T-cell versus macrophage populations.

Clinically, it might be desirable but difficult to overcome the impact of NOD2/CARD15 variants by altered prophylactic strategies. Selecting an alternative donor might be helpful especially if a NOD2/CARD15 variant is present in the recipient. Reducing the epithelial damage by the use of nonmyeloablative strategies, which alter the pattern of GvHD,24 might be another approach, although so far our analysis including standard and reduced-intensity conditioning (RIC) strategies did not show any effect of intensity of conditioning. Adapting the strategy of gastrointestinal decontamination might be a further option, but clearly prospective trials are needed to confirm the association of the type of decontamination and the prognostic role of NOD2/CARD15 status. However, there is no doubt that these findings move forward our understanding of the clinical pathophysiology of transplantation because they demonstrate a link between alterations of innate immunity and specific posttransplantation complications. An association of the gastrointestinal defense against pathogens and GvHD has been clinically postulated for many years and now seems to find confirmation with defined genetic data. Finally, our observations raise the hope that assessment of SNPs within inflammatory or immunoregulatory pathways might indeed lead to individualized and risk-adapted clinical strategies.


The expert technical assistance of Heike Bremm, Manuela Gunckel, Daniela Vogl (Regensburg); Elisabeth Douglas, Christina Basford, and Gerard Cain (University of Newcastle); Moema Nene dos Santos and Antoine Toubert (Paris); Volker Weirich (University of Rostock); and Ingrid Fae (Blood Group Serology Vienna) is gratefully acknowledged.


  • Reprints:
    Ernst Holler, Department of Haematology/Oncology, Medical Centre, University of Regensburg, Franz-Josef Strauß Allee 11, 93042 Regensburg, Germany; e-mail: ernst.holler{at}
  • Supported by grants QLRT-2000-00010 (“EUROBANK”), QLRT-CT-2001-01936 (“TRANSEUROPE”), and MRTN-CT-2004-512253 (“TRANS-NET”) from the European Commission.

  • The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 U.S.C. section 1734.

  • Prepublished online as Blood First Edition Paper, January 19, 2006; DOI 10.1182/blood-2005-09-3741.

  • Submitted September 19, 2005.
  • Accepted January 6, 2006.


View Abstract