Loss-of-Function Mutations of BCOR Are an Independent Marker of Adverse Outcomes in Intensively Treated Patients with Acute Myeloid Leukemia

Simple Summary Acute myeloid leukemia (AML) is a genetically heterogeneous disease. Clinical phenotypes of frequent mutations and their impact on patient outcome are well established. However, the role of rare mutations often remains elusive. We retrospectively analyzed 1529 newly diagnosed and intensively treated AML patients for mutations of BCOR and BCORL1. We report a distinct co-mutational pattern that suggests a role in disease progression rather than initiation, especially affecting mechanisms of DNA-methylation. Further, we found loss-of-function mutations of BCOR to be independent markers of poor outcomes in multivariable analysis. Therefore, loss-of-function mutations of BCOR need to be considered for AML management, as they may influence risk stratification and subsequent treatment allocation. Abstract Acute myeloid leukemia (AML) is characterized by recurrent genetic events. The BCL6 corepressor (BCOR) and its homolog, the BCL6 corepressor-like 1 (BCORL1), have been reported to be rare but recurrent mutations in AML. Previously, smaller studies have reported conflicting results regarding impacts on outcomes. Here, we retrospectively analyzed a large cohort of 1529 patients with newly diagnosed and intensively treated AML. BCOR and BCORL1 mutations were found in 71 (4.6%) and 53 patients (3.5%), respectively. Frequently co-mutated genes were DNTM3A, TET2 and RUNX1. Mutated BCORL1 and loss-of-function mutations of BCOR were significantly more common in the ELN2017 intermediate-risk group. Patients harboring loss-of-function mutations of BCOR had a significantly reduced median event-free survival (HR = 1.464 (95%-Confidence Interval (CI): 1.005–2.134), p = 0.047), relapse-free survival (HR = 1.904 (95%-CI: 1.163–3.117), p = 0.01), and trend for reduced overall survival (HR = 1.495 (95%-CI: 0.990–2.258), p = 0.056) in multivariable analysis. Our study establishes a novel role for loss-of-function mutations of BCOR regarding risk stratification in AML, which may influence treatment allocation.


Introduction
Acute myeloid leukemia (AML) is a genetically heterogeneous disease [1]. In the last decade, next-generation sequencing (NGS) has been introduced in hematological practice, and with the use of myeloid gene panels, rare and recurrent mutations have been unveiled, with a majority of patients harboring more than one mutation even within defined AML entities [2]. The discovery of common recurrent mutations, such as NPM1 and FLT3-ITD, among others, has led to a better understanding of the molecular landscape of AML [1,3], with distinct consequences for prognostication and treatment allocation [4]. However, the role and potential impact of rare mutations on prognosis is not well understood, and warrants further studies to improve risk assessment, implying a more precise approach to AML treatment, as relapse and mortality rates are still unsatisfactory.
The BCL6 corepressor (BCOR) gene, and its homolog, the BCL6 corepressor-like 1 (BCORL1) gene, are located on chromosomes Xp11.4 and Xq26.1, respectively [5,6]. BCOR was originally described as a repressor of BCL6 [5], but also interacts with PCGF1, KDM2B, MLLT3, and IRF8, while BCORL1 interacts with PCGF1, KDM2B, CTBP1, and HDAC [7]. Both BCOR and BCORL1 are core proteins of the polycomb repressive complex PRC1.1. PRCs are essential for the maintenance of cell identity and cell differentiation, and their perturbance is an important factor in carcinogenesis [8]. PRC1.1 plays a role in epigenetic modification by adding an ubiquitin to histone H2A at lysine 119 [9]. This process (among others) leads to a silencing of Hox gene clusters [10], and mediates transcriptional repression [11]. BCOR is highly expressed in embryonic stem cells, where a role in maintaining the primed pluripotent state has been suggested [7,12]. As a mediator of stemness and differentiation, BCOR is involved in the development of B-and T-cells, as well as erythrocytes [13].

Data Set
We retrospectively analyzed a multi-center cohort of 1529 AML patients. Eligibility criteria were newly diagnosed AML according to WHO definitions [29], age ≥ 18 years, and available biomaterial at diagnosis. All patients were treated with intensive regimens in the following clinical trials: AML96 [30], AML2003 [31], AML60+ [32], and SO-RAML [33] or were enrolled in the German Study Alliance Leukemia (SAL)'s AML registry (NCT03188874). Detailed information on treatment regimens is given in the respective references. All mentioned studies were carried out under the auspices of the SAL, and approved by the Institutional Review Board of the Dresden University of Technology (Dresden, Saxony, Germany). All participants gave their written, informed consent, in accordance with to the Declaration of Helsinki.

Definitions
AML was defined as de novo when neither previous malignancy nor previous treatment with chemo-and/or radiotherapy was reported. When myeloid neoplasms were documented prior to AML diagnosis, AML was defined as secondary (sAML). Prior exposure to chemo-and/or radiotherapy before the initial diagnosis defined therapy-associated AML (tMN). Early death was defined as death by any cause within 30 days of the initial diagnosis (ED 30 ). Remission and survival criteria were defined according to ELN2017 recommendations [4].

Molecular Analysis
Next generation sequencing (NGS) using a TruSight Myeloid Sequencing Panel (Illumina, San Diego, CA, USA) was performed on pre-treatment bone marrow or peripheral blood, targeting 54 genes (Table S1) associated with myeloid neoplasms, including full coding exons of BCOR and BCORL1, as previously described in detail [34,35]. A DNeasy blood and tissue kit (Qiagen, Hilden, Germany) was used to extract DNA, and subsequent quantification was performed using a NanoDrop spectrophotometer. Pooled samples were sequenced paired-end (150 bp PE) using a NextSeq NGS instrument (Illumina). The SE-QUENCE PILOT software package (JSI medical systems GmbH, Ettenheim, Germany) was used for sequence data alignment, variant calling, and filtering. A 5% variant allele frequency (VAF) cut-off was used. Genome-mapping algorithms were referenced to human genome build HG19. We compared VAFs of BCOR/BCORL1 mutations with VAFs of co-mutated drivers for dichotomization of dominant, subclonal, and secondary mutations. For putative subclonal mutations a minimum VAF difference of 10% was applied.

Statistical Analysis
We compared categorical variables between groups using the chi-squared test, while continuous variables were compared using the Kruskal-Wallis test. The Kaplan-Meier method was used to estimate survival probabilities. The logrank test was used to compare survival time distributions between groups. Cox regression was used to estimate univariate and adjusted hazard ratios. For the binary endpoint of complete remission, logistic regression models were fitted to estimate univariate and adjusted odds ratios. A significance level of 0.05 was used to determine statistical significance. Calculations were performed in R 4.0.3.

Discussion
We analyzed a large cohort of newly diagnosed and intensively treated AML patients according to their mutational status of BCOR and BCORL1. The respective proportions of mBCOR and mBCORL1 in the cohort were comparable to those reported in recent studies [6,21,26,27]. We found both mBCOR with LOF and mBCORL1 to be more prevalent in patients in the ELN2017 intermediate-risk group [4]. The proportion of sAML amongst patients harboring mBCOR was significantly higher than in their wild-type counterparts, confirming previous reports [37]. Mutations of BCORL1 were significantly more prevalent in females than in males, as has been previously suggested [27], and patients harboring mBCOR with LOF had significantly lower WBC, and a trend for lower peripheral blood blast counts. Mutations of BCOR frequently co-occurred with mutations of DNMT3A, RUNX1, TET2, NRAS, and BCORL1. Since both BCOR and DNMT3A function as epigenetic modifiers [7,38], a synergistic role in leukemogenesis has been reported recently [39,40]. In murine models, the co-occurrence of mBCOR and mutations of TET2, another epigenetic modifier [41] frequently associated with mBCOR in our cohort, has been reported to induce MDS [25,42]. This may further reinforce the notion that BCOR's interaction with other DNA methylators plays a crucial role in leukemogenesis. RUNX1 and BCOR both play essential roles in the proliferation and differentiation of myeloid cells [10,43]. Therefore, the interplay of mBCOR and other essential regulators of normal myeloid development appears to be a factor that may promote leukemogenesis. Recent studies suggest, however, that perturbations of BCOR function alone do not suffice to induce malignant transformation [10,21,44], and thus co-mutations appear to be needed in order to drive leukemogenesis or, alternatively, sequential, secondary acquisition of LOF in BCOR following an oncogenic event, e.g., MDS driver mutation triggers disease progression towards AML, as suggested by the higher frequency of sAML among mBCOR patients. Accordingly, in our cohort of patients with mBCOR AML, the majority had at least two co-mutations, while only one patient harbored no other co-mutations targeted by our panel (Figure 1B). Similarly to mBCOR, co-mutations of mBCORL1 were RUNX1, DNMT3A, and TET2, as well as BCOR and FLT3-ITD, which were only rarely co-mutations of mBCOR. Again, the majority of patients harboring mBCORL1 had at least two other co-mutations, and only one patient had no other co-mutations revealed by our panel ( Figure 2B). As for mBCOR, this suggests a potential interplay of impaired BCORL1 function with other dysfunctional mechanisms of DNA methylation, cell differentiation, and signal transduction in leukemogenesis.
Regarding outcomes, we found no statistically significant differences between patients with mBCOR and mBCORL1 regarding CR rate and ED 30, compared to wild-type patients. Interestingly, while we observed lower median EFS, RFS, and OS for both mBCOR and mBCORL1 in general, only LOF mutations of mBCOR were associated with significantly reduced EFS and RFS and a trend of reduced OS in multivariable testing adjusted for age, AML type, and ELN2017 risk. In MDS, Abuhadra et al. [23] recently reported significantly reduced OS for patients with frameshift mutations of BCOR, while general mutation status did not affect OS. Previous studies of mBCOR in AML have reported poorer outcomes to often be associated with distinct co-mutations; however, a significant association of LOF mutations of BCOR as independent markers of poor outcomes has not yet been reported in AML. Terada et al. [27] reported reduced OS in AML patients who were younger than 65 years, wild-type FLT3, and had intermediate-risk cytogenetics. Grossmann et al. [26] reported a higher prevalence of mBCOR and a trend of reduced OS in normal-karyotype AML with no mutations in NPM1, FLT3-ITD, CEBPA, or MLL-PTD. However, in a validation cohort no association between mBCOR and OS was observed. Nevertheless, in both cohorts reduced EFS was detected [26]. Recently, Eisfeld et al. [45] reported reduced survival for patients in the ELN2017 favorable-risk group who harbored mutations in BCOR or SETBP1. Our findings underline the complexity of the mutational landscape of AML, where even mutational variants of rare mutations have to be considered in order to determine their clinical and prognostic effects. Future work needs to focus on the implementation of LOF mutations of BCOR in risk stratification tools for AML management.

Conclusions
In conclusion, both mBCOR and mBCORL1 are rare but recurrent mutations in AML. While previous studies suggested poor outcomes for mBCOR in AML, especially in the context of co-occurring mutations, we found loss-of-function mutations of mBCOR to be independent markers of poor outcomes in AML, while mBCORL1 was not significantly associated with outcomes in multivariable testing.