Variant IRF4/MUM1 associates with CD38 status and treatment-free survival in chronic lymphocytic leukaemia

A low penetrance risk allele (defined by rs872071) for chronic lymphocytic leukaemia (CLL) has been identified at the interferon regulatory factor 4 (IRF4)/multiple myeloma oncogene (MUM1) locus at 6p25.3.1 Given a putative role for IRF4 in the aetiology of CLL and a role in regulating B-cell development, we hypothesised that genetic variation in this gene also has a role in determining prognosis in CLL. Peripheral blood samples were obtained from 840 individuals of European origin resident in the UK and diagnosed with CLL between 1998 and 2006 who attended one of 6 clinical centres in the United Kingdom. All patients had typical CLL mature B-lymphocytes expressing CD5, CD19, CD23 and clonally restricted surface immunoglobulin, and diagnosis was made in accordance with World Heath Organisation classification guidelines.2 Indications for treatment were consistent with published guidelines,3 and choice of chemotherapy was at the discretion of local treating clinicians. Mean follow-up time was 2726 days (7.5 years) and 275 patients had died at the time of the last follow-up. The mean age at diagnosis was 65 years (range 24–100 years). Collection of peripheral blood samples and clinicopathological information was undertaken with informed consent and ethical review board approval in accordance with the tenets of the Declaration of Helsinki, and also following review by the respective Institutional Research and Development Committees. IGHV gene sequences were compared with germline using the pre-2007 version of IgBlast (National Centre for Biotechnology Information, Bethesda, Maryland, USA; http://ncbi.nih.gov/igblast/) and those with homology to germline of 98% or higher were regarded as unmutated.4 CD38 and ZAP-70 expressions were determined at each individual clinical centre using diagnostic samples by flow cytometry immunophenotyping5, 6 and a cut-off of 20% was used for positivity. Unmutated IGHV, CD38-positive disease, ZAP-70-positive disease and Binet stage B or C disease at presentation were all significantly associated with shorter treatment-free survival (TFS) and overall survival (OS) in univariate analysis (Table 1), confirming that established prognostic markers for CLL predict prognosis in patients recruited to this study. PCR amplification and digestion with BsmFI was used to determine rs872071 genotype (details available on request). Five percent of samples were genotyped by direct sequencing, and concordance with restriction fragment length polymorphism assay was 100%. Buccal DNA samples were available for 44 cases, and genotype concordance between peripheral blood and buccal samples was 98%. The distribution of genotypes for the IRF4 variant was not significantly different from that expected under Hardy–Weinberg equilibrium (P=0.25). Thirty-eight percent of patients were homozygous (G/G) for the allele associated with increased risk of CLL, 49% were heterozygous (G/A) and 14% were homozygous for the low-risk allele (A/A) (Table 2). There was no significant difference in IRF4 genotype distribution when stratified by gender, age at diagnosis, Binet stage of disease, IGHV mutational status, ZAP-70 expression, lymphocyte doubling time or clinical centre (Table 2). However, there was a significant association between IRF4 genotype and CD38 status (P=0.015; Table 2). Specifically, carriers of the IRF4 allele associated with increased risk of CLL (G/G homozygotes and G/A heterozygotes) had a significantly higher frequency of CD38-positive disease (G/G 38%; G/A 38%) compared with patients homozygous for the low-risk allele (A/A 22%, P=0.004; Table 2), but there was no difference between the two high-risk carrier groups (GG vs GA; P=0.9). To investigate the reproducibility of this association cases were stratified by clinical centre. Carriers of the IRF4 CLL risk allele had a higher frequency of CD38-positive disease than noncarriers in all six cohorts that constitute this multicentre case series (Figure 1a). This association was also significant when a 30% cut-off was used to denote CD38+ve disease (P=0.02). IRF4 is a B-cell specific transcription factor with preference for a characteristic consensus DNA binding sequence (5′-AANNGAAAC-3′). The human CD38 gene contains a putative IRF4 binding site immediately upstream of the 5′ untranslated region. Indeed, a DNA oligonucleotide containing an IRF4 consensus binding site identical to that seen in the CD38 gene (5′ AAGTGAAAC 3′) was used to crystallise the DNA binding domain of IRF4.7 These data suggest a plausible mechanism by which IRF4 could affect CD38 expression. IRF4 genotype was significantly associated with TFS (P=0.015; Figure 1b). There was no significant difference between the two high-risk genotype groups (GG vs GA; P=0.5), but carriers of the IRF4 CLL risk allele had a significantly shorter TFS compared with noncarriers (HR 1.6, 95% confidence interval (CI) 1.2–2.3, P=0.006), suggesting that the CLL risk allele defined by rs872071 acts dominantly with respect to TFS. In patients with mutated IGHV, carriers had a significantly shorter TFS compared with noncarriers (HR 1.9, 95% CI 1.1–3.3, P=0.01; Figure 1d). Similarly, carriers also had a shorter TFS when the analysis was restricted to patients with germline IGHV sequence, although this was not statistically significant (HR 1.5, 95% CI 0.9–2.7, P=0.14; Figure 1d). IRF4 genotype was not significantly associated with overall survival when analysed as a trichotomous variable (P=0.095; Figure 1c), but was significant when analysed as a dichotomous variable (carriers vs noncarriers; P=0.047). Multivariate analysis using Cox regression analysis including age, IGHV mutational status, CD38 status, ZAP-70 status and gender in the model confirmed that carrier status for the IRF4 CLL risk allele was an independent predictor of TFS (Hazard ratio (HR) 1.60, 95% CI 1.01–2.51, P=0.044; Table 3). IRF4 status was also an independent marker of TFS in patients with stage A disease at diagnosis (HR 1.94, 95% CI 1.03–3.62, P=0.04; Table 3). In univariate analysis IRF4 status is a relatively weak marker of TFS (HR 1.6, P=0.006) compared with IGHV mutational status (HR 3.6, P<0.001), CD38 expression (HR 2.9, P<0.001) or ZAP-70 expression (HR 2.6, P<0.001), and for this reason may be of limited value as an independent marker of TFS. This is compounded by the observation that the allelic variant associated with poor prognosis seems to be acting dominantly with respect to both TFS and CD38 expression, and a minority of patients (approximately 15%) are noncarriers and identified as having a relatively good prognosis. Nevertheless, IRF4 status remained a significant predictor of TFS in multivariate analysis. In contrast to other established prognostic markers for CLL, including advanced stage of disease, germline IGHV and positivity for CD38 or ZAP-70 which each identify a minority population (20–40% of cases) with unfavourable outcome, IRF4 status identifies a minority population with relatively favourable outcome. As such, IRF4 polymorphic status could have utility in predicting prognosis in those patients primarily negative for established poor outcome markers, who represent the majority of cases at diagnosis. Such a marker may well be usefully included in multiparameter scoring systems for disease prognostication. IRF4 is expressed in germinal centre and post-germinal centre B-cells committed to plasma cell differentiation, and has a complex role in regulating B-cell maturation and homoeostasis. The functional significance of IRF4 allelic variants are undetermined, although we can speculate based on the location of rs872071 in the 3′ UTR that this polymorphism differentiates between alleles with altered expression profiles, possibly by affecting message stability or translation efficiency. Consistent with this model, initial data suggests an association between IRF4 genotype and altered gene expression in immortalised B-lymphocytes.1 Interestingly, the 3′ UTR of IRF4 includes several established and putative miRNA binding sites, at least two of which are bound by miRNAs giving rise to a reduction in IRF4 protein expression in germinal centre B-cells.8 However, to date there is no direct evidence for miRNA binding to the region incorporating rs872071. IRF4 expression measured by immunohistochemistry or quantitative PCR has been associated with prognosis in several B-cell malignancies, although the data relating to prognosis in CLL are inconsistent. Although we have identified an association between IRF4 genotype and CD38 expression further work will be required to delineate the functional IRF4 variant(s) and to further investigate the nature of that functionality, including any potential effect on both IRF4 and CD38 expression. An understanding of factors regulating CD38 expression would facilitate the development of this molecule as a target for therapy in CLL. This study identifies a common variant in IRF4, previously associated with increased risk of developing CLL,1 as a predictor of a more aggressive disease course and reduced time to first treatment. The strong association with CD38 expression and the identification of a putative IRF4 binding site in the CD38 gene immediately upstream of the transcriptional start site suggests a plausible mechanism by which IRF4 status could affect prognosis. It will be of interest to test the prognostic value of IRF4 polymorphic status in a large prospective population-based study, and to examine whether IRF4 status affects the risk of developing monoclonal B-cell lymphocytosis and progression to CLL. The authors declare no conflict of interest. This work was supported by Leukaemia Research, London, United Kingdom (Grant no. 06002 to JMA). The authors thank Simon Kometa (Newcastle University) for statistical support, Elaine Willmore and Barbara Durkacz (Newcastle University) for supporting this study and Daniel Catovsky (Institute of Cancer Research, Sutton) for critical review of this paper.