Association between BMI-1 expression, acute graft-versus-host disease, and outcome following allogeneic stem cell transplantation from HLA-identical siblings in chronic myeloid leukemia

The management of chronic myeloid leukemia (CML) has changed radically since the introduction of tyrosine-kinase inhibitors. The decision whether to offer a patient allogeneic stem cell transplantation (allo-SCT) must be carefully weighed. Historically, various attempts have been made to determine prognosis and outcome for individual patients with CML at the time of allo-SCT. Thus, the European Group for Blood and Marrow Transplantation (EBMT)–Gratwohl score based on histocompatibility, stage of disease at time of allo-SCT, age and sex of donor and recipient, and time from diagnosis to allo-SCT, proved to be the most useful scoring system to assess the risk of mortality when counseling and decision-making.^1,2  However, it is established that CML exhibits marked heterogeneity in prognosis,³  which is reflected even in response to modern therapies.⁴  We have recently shown in patients with CML who did not receive allo-SCT that the combination of CD7, proteinase-3 (PR-3), elastase-2 (ELA-2), and the polycomb group (PcG) gene BMI-1 expression levels at diagnosis can reflect the intrinsic molecular heterogeneity of the disease, especially disease aggressiveness and duration of chronic phase (CP).^5,6  As immune responses to both PR-3 and ELA-2 peptides have been implicated in the eradication of CML following allo-SCT,^7,8  we investigated whether these genes might also serve as biomarkers to predict outcome of patients with CP-CML receiving allo-SCT.

A total of 84 consecutive patients with CP-CML with complete follow-up who underwent transplantation between 1981 and 2002 and who had cryopreserved leukaphereses collected before allo-SCT as standard policy at the Hammersmith Hospital were studied. Informed consent for the use of these cells for research was approved by the Hammersmith and Queen Charlotte's and Chelsea Research Ethics Committee Institutional Review Board in accordance with the Declaration of Helsinki. Diagnosis and staging of CML^5,9  and acute and chronic graft-versus-host disease (GVHD) were graded according to standard criteria.^10,11 

Peripheral blood mononuclear cells (PBMCs) from cryopreserved material were isolated by density gradient centrifugation (Lymphoprep; Nycomed, Oslo, Norway). Total RNA was extracted using the Qiagen RNeasy kit (Qiagen, Crawley, United Kingdom), treated with DNase I (Invitrogen, Paisley, United Kingdom) to eliminate genomic DNA, and reverse-transcribed into cDNA according to standard methods. Expression of CD7, PR-3, ELA-2, BMI-1, and GAPDH was assessed by quantitative real-time reverse transcriptase–polymerase chain reaction (Q-RT-PCR) using the Applied Biosystems 7300/7500 Real Time PCR System (Applied Biosystems, Foster City, CA). All Q-RT-PCR reactions were performed in 25-μL volume.⁵ GAPDH expression was used as the endogenous cDNA quality control. The Applied Biosystems Assays-on-Demand TaqMan probe-and-primer reagents for CD7, PR-3, ELA-2, BMI-1, and GAPDH were used according to the manufacturer's instructions. The median gene expression level was used to segregate the patients into a “low” group (gene expression less than median) and a “high” group (gene expression greater than median).

The following post–allo-SCT outcome parameters were analyzed: engraftment, acute GVHD, chronic GVHD, nonrelapse mortality (NRM), leukemia-free survival (LFS), and overall survival (OS). Groups were compared using the Mann-Whitney test for continuous data and the Fisher exact test for categoric data. Probabilities of OS and LFS were calculated using the Kaplan-Meier method, and groups were compared using the log-rank test. The cumulative incidence procedure was used to calculate probabilities of NRM and GVHD. Genes identified from univariate analyses with P values of less than .20 (P < .20) were entered into Cox regression analyses that contained established prognostic variables (Gratwohl score, year of transplantation, GVHD prophylaxis regimen, and TBI dose). All quoted P values are from 2-sided tests with values less than .05 considered significant.

Patient, disease, and transplantation characteristics are summarized in Table 1. All patients received allo-SCT from an HLA-identical sibling and were mainly treated in the pre-imatinib era. The median follow-up time for the patients alive after allo-SCT was 9.8 years (range, 1.7-23.9 years). Patients underwent transplantation at a median of 9 months (range, 3-94 months) after diagnosis. The median EBMT-Gratwohl calculated score was 3. A total of 20 (24%) patients died of NRM (secondary malignancies, n = 3; severe GVHD, n = 4; opportunistic infections, n = 9; other causes, n = 4), while 6 died of disease progression. CD7, PR-3, ELA-2, and BMI-1 expression did not show any significant association with neutrophil or platelet engraftment, or with graft rejection. CD7, PR-3, and ELA-2 expression was not associated with OS. These data corroborate those from Yong et al,⁸  where PR-3 and ELA-2 expression prior to allo-SCT was prognostic only in advanced-phase CML but not in CP-CML. However, and in sharp contrast to our previous findings in the non-allo-SCT setting,⁶  patients displaying a “high” BMI-1 expression level prior to allo-SCT had significantly better OS than those with low expression (P = .005; Figure 1A). Furthermore, when BMI-1 expression was included in a multivariate survival model and adjusted for the other prognostic variables (EBMT-Gratwohl score, allo-SCT era, GVHD prophylaxis, total body irradiation [TBI] dose), a high expression level was found to be an independent marker associated with better survival (relative risk [RR] = 2.72; 95% confidence interval [CI], 1.1-6.9; P = .034), suggesting that measuring such expression for survival prognostication may be more powerful than the previously established criteria. The predictive value of BMI-1 expression in terms of OS was also significant when the analysis was restricted to patients (n = 68) who did not receive any form of T-cell depletion for GVHD prophylaxis (RR = 4.19; 95% CI, 1.31-13.4; P = .015).

Given the impact of BMI-1 expression level on OS, without a significant association with relapse, and since neither BMI-1 nor the other tested genes showed any significant association with LFS (Figure 1B), we assessed their impact on NRM. In multivariate analysis, BMI-1 and ELA-2 showed significant associations with NRM (RR = 3.02; 95% CI, 1.05-8.71; P = .041; RR = 2.7; 95% CI, 1.02-7.16; P = .045, respectively). Expression of CD7, PR-3, and ELA-2 was not associated with acute GVHD occurrence or severity. However, there was a striking and significant association between acute GVHD and BMI-1 expression, not only in overall incidence (“low” BMI-1: grades 0-1 (n = 21), grade 2 (n = 10), grades 3-4 (n = 9); “high” BMI-1: grades 0-1 (n = 32), grade 2 (n = 9), grades 3-4 (n = 1); P = .005), but also in cumulative incidence at 100 days (24% vs 48%; P = .016; Figure 1C). In multivariate analysis, a “low” BMI-1 expression level was associated with an increased risk of grades 2 to 4 acute GVHD (RR = 2.85; 95% CI, 1.3-6.4; P = .011).

Our previous observations suggested that BMI-1, an essential gene for the self-renewal of normal as well as cancer stem cells,^12-14  plays an important role in CML pathophysiology and prognosis in the nontransplantation setting.⁶  However, in contrast to patients treated with hydroxyurea and interferon-α (in the pre-imatinib era), where high expression of BMI-1 was associated with a worse prognosis, the opposite was observed in patients treated with allo-SCT. Although these novel and unexpected observations require confirmation by larger studies, several hypotheses can be offered to help understanding them. As previously established in CML and other malignancies,^6,15-18 BMI-1, acting in cooperation with other oncogenes, can induce neoplastic transformation,^19,20  and its overexpression contributes to disease aggressiveness. However, in the context of allo-SCT for CP-CML, a gain of a neoplastic proliferative advantage within the leukemia stem cell pool through increased BMI-1 expression may be neutralized by an immune response in donor cells against BMI-1, which was shown to be a genuine tumor-associated antigen.²¹  Moreover, our data on the association between BMI-1 and acute GVHD add to the growing evidence that PcG genes are also involved in the regulation of immune functions. A higher expression of BMI-1 was shown to facilitate T helper 2 (Th2) cell differentiation in a Ring finger–dependent manner by regulating GATA3 protein stability,²²  reinforcing the view that the balance between Th1 and Th2 immune responses is critical for controlling the severity of acute GVHD.^23,24  An alternative hypothesis is that the initiating role of inflammatory cytokines associated with GVHD may be influenced by BMI-1 expression.²⁵ 

With the caveat that supportive care and thus GVHD incidence might have changed during the relatively long follow-up period of this study, our data suggest that BMI-1 may be a potential “biomarker” to identify those patients likely to develop GVHD and its corollary NRM. Thus, in addition to the well-established EBMT-Gratwohl score,¹  measurement of BMI-1 expression level at diagnosis can improve risk assessment of candidate patients with CML for allo-SCT. At present, long-term results of tyrosine-kinase inhibitors are still unknown, unlike allo-SCT, for which evidence is solid and long-standing. Allo-SCT is a viable option for 25% of patients with CML who exhibit primary or secondary resistance to tyrosine-kinase inhibitors, and remains the first-line therapy in many countries due to prevailing economical reasons.²⁶  However, the risk of high NRM associated with acute GVHD must be balanced against the benefit of allo-SCT, hence the current challenge to identify subsets of patients best suited to receive this type of therapy. In this regard, and in addition to other immune parameters,⁸ BMI-1 is likely to represent a marker of CML outcome after allo-SCT, and may offer a valuable tool toward tailored therapeutic interventions, including allo-SCT.

We thank the staff of the Stem Cell Laboratory at the Hammersmith Hospital (London, United Kingdom) for help with identification of stored cells, and Daniela Passos, Tamara Law, Maria Makri, and Chee Han Lim for excellent technical assistance.

This work was supported by a grant from the Hammersmith Hospital Trust Research Committee.

Contribution: M.M. conceived and designed the study, collected patient data, processed samples, performed experiments, analyzed data, performed bibliographic search, and wrote the report; R.M.S. performed the statistical analysis and helped write the report; A.Y. conceived and designed the study, performed bibliographic search, and commented on the manuscript; J.F.A. provided clinical care and recorded clinical data; J.M.G. provided clinical care, recorded clinical data, and commented on the manuscript; and J.V.M. conceived and designed the study, supervised its execution, helped write and revise the report and had full access to all the data in the study and had final responsibility for the decision to submit it for publication.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Junia V. Melo, Department of Hematology, Institute of Medical & Veterinary Science, Frome Road, Adelaide, South Australia 5000, Australia; e-mail: junia.melo@imvs.sa.gov.au.