The basis for the differential impact of BCOR and BCORL1 in aplastic anemia/paroxysmal nocturnal hemoglobinuria and myeloid neoplasia

Abstract BCOR and BCORL1 (BCOR/L1) are X-chromosomalTSGs affected by inactivating or hypomorphic somatic mutations (MT) in myeloid neoplasia (MN). However, BCOR/L1MT are also present in the context of immune-mediated bone marrow failure, including aplastic anemia (AA) and paroxysmal nocturnal hemoglobinuria (PNH) as clonal hematopoiesis. While BCORMT appears to be a poor prognostic factor in MN, in AA/PNH, it represents a self-limited clonal expansion with carriers having a lower risk of secondary MN progression1. This discordant impact of BCOR/L1MT constitutes a clinical conundrum, which can be deconstructed into 2 main pathogenetic issues: i) differential leukemogenic properties in AA/PNH and MN with their disconnection from the secondary MN evolving from prior AA/PNH; ii) immunologically privileged factors leading to the preferential expansion of otherwise benign BCOR/L1MT clones in AA/PNH. There are possible theories addressing these issues including; a) decreased pathogenicity of BCOR/L1 hits in AA/PNH due to the higher proportion of missense mutations and/or their predominance in AA/PNH females; b) occurrence without a context of somatic founder mutations. Either way, preferential expansion of BCOR/L1MT clones may be predicated by extrinsic factors operative only in the context of AA/PNH milieu. They may involve abundance of hematopoietic growth factors, immune attack selecting for phenocopy of PNH due to lack of glycosylphosphatidylinositol-anchored proteins (GPI-APs), or deficient fitness/quantity of remaining stem cells. The latter factors may facilitate somatic gene rescue via the acquisition of BCOR/L1MTmutant clones. To gain more insight into this, we started by analyzing the mutational context of BCOR/L1MT. In 10,332 patients (institutional, 3328; public series, 7004), somatic 372 BCORMT (3.6%) and 66 BCORL1MT (0.6%)were found, including 42 cases with bi-clonal mosaicism. We found similar mRNA expression in males and females, indicating almost complete lyonization. BCOR/L1MT rates were comparable between MN (n=9,414; 5%), AA/PNH (n=918; 4%), and CHIP/CCUS (n=212, 6%). In AML, single-cell NGS data suggested that BCOR is subclonal, confirming its role of a secondary leukemogenic driver2. In AA/PNH, PNH clones were significantly more frequent (96% vs 72.1%, P=.013) in BCOR/L1MTpatients. Next, we assessed the differences between AA/PNH and MN. Patients with BCOR/L1MT in MN were significantly older than in AA/PNH (P<0.001), occurring at the same proportion in old vs. younger AA patients. BCOR/L1MT carriers in AA/PNH were characterized by a higher prevalence of missense and in-frame indels (P=.036) and a significantly smaller clonal burden compared to AML (P<.001). In AA/PNH, no BCOR/L1MT patients coincided with somatic HLA mutations and locus deletions, and the distribution between female and male BCOR/L1MT patients was not significantly higher to that of MN (53% vs 36%, P=.122). In AA/PNH, BCOR/L1MT predominantly occurred alone (n=38; 53%) or as clonal mosaicism with PIGAMT (21%). In contrast, in MN, BCOR/L1MT frequently co-occurred in the same clone with RUNX1MT, DNMT3AMT, and ASXL1MT. We also studied potential factors promoting BCOR expansion using RNA-seq. Following normalization based on cell composition using CIBERSORTX, RNA-seq of 644 immune effector and GPI-linked proteins demonstrated that, BCORMT AA exhibited decreased expression of ICOSLG, CD274, and HPSE but no difference in checkpoint molecules, likePD1, TIM2, and VISTA. The mRNA levels of HLA and β2microglobulin did not differ between BCOR/L1MT and WTcases. However, K562 BCORKO did show HLA class I, downmodulation to isogenic wildtype. Interestingly, BCOR/L1 expression in AA/PNH was significantly reduced compared to controls (P<.001). We also investigated whether BCOR/L1MT results in a phenocopy of PNH by downregulation of GPI-APs. Indeed, certain GPI-APs were downregulated in BCOR/L1MT AA including e.g., CD14, HPSE, ENTPD2, and PRDM12, suggesting that some GPI-APs may partially phenocopy PNH involved in immune advantage in the context of immune attack. In conclusion, while clear differences in the genetic context of BCOR/L1 were found compared to MN, we could not definitely establish a reason for the expansion of BCOR/L1MT clones in the AA/PNH environment. However, downmodulation of certain GPI-APs may resemble the PNH phenotype, and indeed, these two genetically distinct clones may share functional similarities in the context of AA/PNH. 1 2