BCOR involvement in cancer

BCOR is an epigenetic regulator

It is known that tumorigenesis depends on the sequential acquisition of proto-oncogene-activating mutations and/or oncosuppressor loss of function. In this oversimplified perspective, epigenetic alterations are less easily framed, even if the resulting inappropriate expression of oncogenes and oncosuppressors can play an equally crucial role in cancer development and progression. Therefore, epigenetic regulators are potential proto-oncogenes or oncosuppressors, depending on the function of their target genes. A gene implicated in transcriptional regulation, BCOR is increasingly reported as mutated in different human cancers, with a key role in neoplastic transformation or in tumor progression.

BCOR is located on chromosome X, in the Xp11.4 locus, and derives its name from its function as an interacting corepressor of BCL-6 that enhances BCL-6-mediated transcriptional repression [1]. The gene has 16 alternative exons coding several protein isoforms, with the principal isoform, encoded by 14 exons, that gives rise to a protein of 1755 amino acids. BCOR shows virtually ubiquitous expression, but only a few isoforms retain known protein interactions, depending on the domains preserved by alternative splicing.

The function of BCOR is primarily mediated by two domains: the BCL-6-binding domain that interacts with the transcriptional repressor BCL-6 and PUFD, a common domain recurrent in different proteins implicated in histone regulation that binds the RAWUL domain of PCGF proteins (Figure 1).

BCOR takes part into one specific type of polycomb repressive complex that mediates transcriptional repression through epigenetic modifications of histones. Polycomb repressive complexes (PRCs) are molecular complexes involved in histone modification processes that are usually classified into two types: PRC1, which adds an ubiquitin moiety to histone H2A at Lys119 (H2AK119ub1), and PRC2, which catalyzes the addition of one to three methyl groups to histone H3 at Lys27, leading to H3K27me1, H3K27me2 or H3K27me3 [2]. PRCs silent a wide range of genes including HOX group genes, the first identified PRC targets [2,3]. The core protein subunits of PRC2 complex are Suz12 (with a Zinc finger domain), Eed (with a WD repeat domain that recognizes trimethylated peptides) and Ezh2 (the SET-domain containing catalytic subunit) [4]. These core components associate with different proteins that work to regulate PRC2 enzymatic activity or its recruitment to specific genomic loci.

Conversely, all PRC1 complexes contain a core consisting of a RING1 protein (either RING1A or RING1B), which is an E3-ubiquitin ligase that adds an ubiquitin group to histone H2A at Lysine 119, and a PCGF protein (PCGF 1–6). Different PRC1 complexes are defined by the different PCGF protein that associates to the complex and directs interaction with distinct auxiliary proteins, giving rise to PRC1 complexes with differential target specificity and chromatin recruitment patterns [5]. Moreover, based on the cofactor that binds to the RING1 subunit, PRC1 complexes are often divided into ‘canonical’, ‘noncanonical’ or ‘variant’. Canonical PRC1 complexes (cPRC1) are assembled around PCGF2/4 and contain one chromobox protein (CBX2, 4, 6–8) that is able to recognize and bind the H3K27me3 histone mark implemented by the PRC2 complex [6]. Conversely, noncanonical PRC1 complexes (ncPRC1) bind to the RYBP or its paralog YAF2 and associate with PCGF1, PCGF3/5 or PCGF6 to give rise to PRC1.1, PRC1.3/1.5 or PRC1.6, respectively [4,6]. RYBP and YAF2 play a fundamental role in stimulating E3 ligase activity of PRC1 and enhancing H2AK119ub deposition [7]. Importantly, RYBP and YAF2 lack the capacity to bind to H3K27me3 and can be recruited to chromatin in cells lacking functional PRC2.

BCOR participates in the constitution of one of the six currently described noncanonical variants of PRC1, the PRC1.1 (Table 1) [8–17]. PRC1.1’s core components include RING1A/B and PCGF1 that heterodimerize via their N-terminal RING domains (Figure 2). The RING1/PCGF1 heterodimer forms a scaffold for PRC1 assembly. PRC1.1 contains RYBP or its homolog YAF2, altogether with BCOR and other proteins (Figure 2). BCOR associates with the complex by binding to PCGF1 through its PUFD domain located at the C-terminus. Since RYBP and YAF2 are not able to bind H3K27me3, recruitment to chromatin depends on the activity of KDM2B, a H3K36 me2 histone demethylase (Figure 2) that harbors a zinc finger CxxC-binding domain involved in DNA binding that recognizes and drives targeting of PRC1.1 to nonmethylated CpG islands as observed in murine embryonic stem (ES) cells [5,18].

In mammals, PRCs are recruited to target sites by multiple mechanisms, including the binding of hypomethylated CpG islands, the specific interaction with transcription factors and long ncRNAs, and the recognition of chromatin marks deposited by other histone-modifying complexes [6]. This last mechanism is at least partly responsible for the coordinated activity of PRC1 and PRC2, since it is known that the PRC2-dependent H3K27me3 mark increases affinity and recruits CBX-containing cPRC1 [19], while KDM2B-mediated recruitment to nonmethylated CpG islands of ncPRC1 drives H2AK119ub that conversely promotes binding of PRC2 [5]. However, it is also known that in some contexts PRC1.1 can act independently from PRC2; for example, in leukemic stem cells PRC1.1 was found to bind a unique set of genes independently of PRC2 and of the H3K27me3 histone mark [8].

Not surprisingly, the genes encoding for polycomb proteins (PcG) were initially identified, along with trithorax (TrxG), as important elements for homeostasis of body development in Drosophila melanogaster. Indeed, even in mammals and humans, PcG gene germinal mutations often cause severe morphological defects and complex syndromes [20–22]. Specifically, a number of BCOR germinal loss of function mutations (>40 were previously identified) [23] induce oculo-facio-cardio-dental syndrome (OFCD) [24], inherited in an X-linked dominant mode and lethal in males [25]. The typical phenotype of OFCD syndrome is characterized by ocular defects, such as congenital cataracts and microphthalmia, craniofacial dysmorphisms, such as bifid nose tip and palatoschisis, cardiac abnormalities, such as atrial and ventricular septal defects, as well as dental abnormalities, such as radiculomegaly and oligodontia [26]. An attenuated variant of OFCD syndrome, known as Lenz’s microphthalmia, has also been described, holding a classical X-linked recessive inheritance and therefore asymptomatic in females [24].

Nevertheless, the normal evolution of body structure during embryo development is not the only biological process that depends on proper activity of the PcG group proteins. These are also involved in the maintenance of cell identity and are necessary for correct cell differentiation [27]. Studies performed on murine ES showed that BCOR is required for the formation of primitive erythrocytes, plays a role in B- and T-cell development and is necessary for the timely expression of genes regulating ES cell pluripotency and ectodermal and mesodermal development [28]. Moreover, in ES cells, BCOR is highly expressed and required for the maintenance of the pluripotent state, since its depletion induces a robust differentiative phenotype, further supporting the idea that BCOR is a core regulator of the primed pluripotent state [29].

Due to its central role in pluripotency maintenance, differentiation induction and cell fate determination, it is not surprising that mutations in BCOR play a central role in cancer development. This review provides an in-depth description of the different tumor types for which the pathogenetic role of BCOR seems to be particularly relevant.

Sarcomas

Clear cell sarcoma of the kidney

Studies of clear cell sarcoma of the kidney (CCSK) have revealed a lot about the oncogenic role of BCOR somatic alteration on the diagnosis and biological understanding of this disease. CCSK is the second most common renal malignancy in childhood, representing 4–5% of all renal tumors [30] and accounting for approximately 20 new cases per year in the USA [31]. Although CCSK is recognized as an independent entity from Wilms’ tumor, there is still much to learn about this sarcoma.

Microscopically, the classical pattern of CCSK presentation (91% of all cases) [31] is characterized by an architecture consisting of nests or cords of round cells with clear cytoplasm, separated by regular fibrovascular septa [32]. The extracellular matrix, rich in mucopolysaccharides, contributes to confer the ‘clear’ appearance of the tumor. However, classical architecture always accompanies at least one other pattern among the many described, albeit to a lesser extent [31,33].

From a molecular point of view, CCSK appears as a genetically stable tumor, and over the years, the genetic alterations found were few and not recurring. A decisive step forward was made between 2015 and 2016, when five independent studies described and confirmed the recurrence of different internal tandem duplications (ITD) in the last exon of the BCOR gene, involving the PUFD domain (Figure 3A) [34–38]. These events were found in less than 75% of analyzed cases (Table 2), and many different types of ITD were identified in CCSK, all residing in exon 15 of the PUFD domain, and all preserving the protein frame. In particular, the different ITDs involved the region encompassing amino acids 1700–1755, with a minimal identified duplication of 22 amino acids up to a maximum of 38, and a minimal overlapping duplicated region covering protein positions 1725–1737 [38]. The involved region is almost always duplicated, while only in very rare cases was a partial triplication found [38]. These achievements, confirmed by subsequent works [39,40], laid the basis for the investigation of CCSK pathobiology. In fact, the features of duplication, that is strictly in-frame and associated with overexpression of the protein encoded by BCOR, suggested important speculations about the origins of CCSK.

There have been several attempts to attribute biological significance in terms of gain or loss of function to the ITD–BCOR; however, until now, no predicted structural model of the BCOR protein carrying the ITD has been analyzed and released, even though protein–protein interaction studies suggested that ITD–BCOR might represent an hypomorphic allele with altered PCGF1-binding affinity [13]. Currently, we are still far from having conclusive results on the question, and even though there is evidence supporting a loss-of-function effect of ITD-BCOR, (overexpression of Cyclin D1 [43,44], of PRC2 targets and of several classes of HOX genes [36]), the lack of any frameshift or nonsense mutation in this setting is instead in favor of a gain-of-function event.

Primitive mesenchymal myxoid tumor of infancy

Primitive mesenchymal myxoid tumor of infancy (PMMTI) is a newly introduced myofibroblastic pediatric tumor [124]. It is an intermediate grade mesenchymal neoplasm with a high local recurrence rate, but with low metastatic potential [47]. Histologically, it is composed of cells that are fused in a myxoid matrix. Morphologically, differential diagnosis is warranted with respect to other stromal tumors, such as congenital infant fibrosarcoma, fibromyxoid sarcoma, myofibrosarcoma and dermatofibrosarcoma protuberans [47], from which it clearly diverges from the immunophenotypical pattern and specific molecular features. Actually, few studies that investigated PMMTI at a molecular level have found the presence of ITD–BCOR (Figure 3A) associated with the overexpression of BCOR in almost all the cases analyzed [40,52,125], similarly to CCSK. Recently, Kao et al. demonstrated that these two tumors also show a very similar transcriptional signature that, together with epidemiological, morphological and genetic overlap, suggests the possibility that PMMTI is the equivalent of CCSK arising in soft tissue [40].

Small round blue cell sarcomas

The typical pediatric tumors of Ewing sarcoma family can be molecularly distinguished by the presence of translocations involving the EWSR1 gene and one of ETS/FLI family genes. These tumors are morphologically distinct per their microscopic appearance consisting of a bed of round monomorphic cells with regular nuclei and dispersed chromatin, with or without the presence of rosettes [45]. However, progressive diffusion of molecular profiling and high-throughput sequencing technologies has allowed the identification of an increasing number of molecular alterations in undifferentiated sarcomas that are morphologically similar or indistinguishable from Ewing’s sarcoma, but lacking the aforementioned molecular features. The need to set up a new classification scheme for these tumors led to the use of the term ‘Ewing-like small round-blue-cell sarcoma’ (SRBCS), although the specific molecular entities belonging to this subgroup often present atypical, undifferentiated or with nonspecific histology (undifferentiated round cell sarcomas [URCS]). Among the soft tissue URCS, Kao et al. have shown that nearly half of them carry ITD–BCOR [40], whereas Pierron et al. identified a group characterized by the presence of the BCOR–CCNB3 fusion gene (Figure 3B) and overexpression of the resulting chimeric protein [46]. These tumors showed a transcriptional signature different from pediatric tumors of Ewing sarcoma family (they were negative for BCOR immunostaining [54]) and a broader morphological spectrum [46,56], which is why the classification of Ewing-like sarcomas appears reductive. Other works have frequently found cases of BCOR–CCNB3 positive sarcomas, in variable percentages depending on the group of tumors analyzed, but in any case definitely consistent (Table 2) [40,48–51,53,55–59,126–129]. CCNB3 encodes for cyclin B3, a protein with promeiotic function in which its expression is normally limited to the testis. CCNB3 activity in the chimeric protein (evaluated through cellular models of engineered fibroblasts) suggested a possible oncogenic driver role due to the observed ectopic expression of CCNB3 in this type of tumors, although further evaluations are needed to assess the role of BCOR in the chimeric protein, as well as any potential new activity acquired from the fusion of the two proteins [46].

Regarding ‘Ewing-like’ SRBCS, Specht et al. [55] described, in addition to the BCOR–CCNB3 fusion gene, the presence of distinct rearrangements of BCOR in eight cases: two cases carrying BCOR–MAML3 fusion (Figure 3C), two carrying ZC3H7B–BCOR (Figure 3D) and four with internal rearrangements of BCOR. In the same study, the increased expression of HOX genes was reported in one of the two cases with BCOR–MAML3 fusion, suggestive of a loss of function of BCOR and PRC activity. MAML3 encodes a member of the mastermind-like (MAML) family of transcriptional coactivators that constitutes a component of the Notch signaling pathway, with roles in biological processes such as cell proliferation, differentiation and survival. The BCOR–MAML3 fusion transcript contains exons 2-5 of MAML3, thus retaining the transactivation domain, but losing the Notch-binding site of MAML3. ZC3H7B encodes a nuclear protein containing a domain able to interact with the rotavirus NSP3 36 protein [130], as well as several other domains involved in multiple protein interactions and nucleic acid/protein interactions. Little can be said at present about the ways in which these two genes exert their oncogenic roles in SRBCS, since their activity in the context of chimeric proteins containing BCOR requires further study. However, BCOR is involved in its entirety in the BCOR–MAML3 fusion and in its terminal 3′ portion only in the ZC3H7B–BCOR chimeric transcript. The recurrence of these two fusion genes in SRBCS was also confirmed in the most recent study by Watson et al. [126], which highlighted the existence of SRBCS characterized by the presence of ITD–BCOR (Table 2). An extremely interesting result of this last study is that all SRBCS-carrying BCOR alterations generate a common cluster at the transcriptional level, thus reinforcing a view in which BCOR alterations, either ITD or rearrangements, exert a common pathogenic pathway leading to abnormal activity of the PRC1.1 complex. In fact, as in CCSK, a significant enrichment of gene sets regulating morphogenesis, differentiation of neurons and tyrosine kinase receptors was observed, and HOX genes are overexpressed. In the same study, the authors noticed an overexpression of SMARCA2, a component of the chromatin modifier SWI/SNF complex, and its activity was opposite to that of the PRC1 and PRC2 complexes [67], with observed strict association and functional linkage [68]. This insight may suggest that PRCs loss-of-function caused by BCOR alterations can lead to increased SWI/SNF complex activity. This hypothesis is reinforced via in vitro work in which the codeletion of EZH2 reverts the oncogenic phenotype caused by the loss of function of the SMARCB1 gene [66], similarly to SMARCA2, which encodes for a subunit of the SWI/SNF complex. Recently, Kao et al. [51] also identified a single case of URCS harboring the KMT2D–BCOR fusion gene.

Rhabdomyosarcoma

BCOR’s involvement appears to also be relevant in rhabdomyosarcoma (RMS), another typical blue round cell malignant childhood tumor. Histologically, the two major RMS subtypes are alveolar RMS and embryonal RMS, both of which have distinct molecular and clinical hallmarks [131]. Alveolar RMS results in poor prognosis and is genetically defined by the presence of a fusion involving PAX3 or PAX7 genes, two transcription factors. The embryonal RMS subtype typically affects younger children and is generally comprised of PAX-negative tumors. A rare subtype of RMS is anaplastic, where a single case was recently reported to carry a BCOR alteration, together with mutation in ARID1A, a member of the SWI/SNF family, and of SETD2, a histone methyl-transferase [61]. Two studies [131,132] confirmed that BCOR mutations in RMS recur in about 7–8% of total cases, with an apparent major involvement in PAX-negative RMS, for which the percentage nears 10% of cases (Table 2).

Endometrial stromal sarcoma

Endometrial stromal sarcoma (ESS) is a gynecological sarcoma composed of cells that resemble those of the endometrial stroma in the proliferative phase. The WHO classification recognizes four distinct entities [133,134]: benign stromal endometrial nodules; low-grade ESS genetically characterized by various translocations among which the most frequent involves JAZF1 and/or PHF1; high-grade ESS generally carrying YWHAE–NUTM2 fusion gene [135]; and undifferentiated endometrial sarcoma with a complex karyotype.

In particular, the involvement of the YWHAE–NUTM2 fusion gene is noteworthy since this was the first genetic alteration associated with CCSK [136], as well as the second in order of frequency (with a percentage between 0 and 12% of cases in various studies) [34–38,63,137,138]. YWHAE genes (tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein) encode different variants of the 14-3-3 protein family, which includes seven highly conserved and ubiquitously expressed proteins that seem to play a role in modulation of cytoskeletal organization, metabolism, differentiation and proliferation [62]. NUTM2 genes of unknown function belong instead to the FAM22 family, which was more recently renamed NUTM2 because of sequence homology to the NUT gene locus that is involved in the oncogenesis of malignant tumor of midline [64].

Several patients with ZC3H7B–BCOR and/or BCOR–ZC3H7B [60,65,134,139] fusion genes were also described and initially included among low-grade ESS; however, it was recently proposed to identify ZC3H7B–BCOR ESS as an independent subgroup of high-grade ESS (Table 2) due to new prognostic evidence and myxoid leiomyosarcoma-like histologic appearance [65]. Recently, on subsets of 31 various grade ESS, Mariño-Enriquez et al. identified three cases carrying ITD–BCOR, one defined as high-grade ESS and two as undifferentiated ESS, all three cases holding typical immunophenotypic features of ITD–BCOR-positive tumors, such as the overexpression of cyclin D1 and BCOR [70].

This description limited to the mesenchymal tumor histotypes carrying BCOR-related abnormalities clearly underlined that the evident morphological and molecular overlaps between CCSK and PMMTI, as well as the common recurrence of ITD–BCOR, are not the only similarities found among these tumor histotypes. In fact, the YWHAE–NUTM2 fusion gene, typical of some CCSKs, seems to be the driver element of high-grade ESS subgroup. Conversely, low-grade ESS is often characterized by the involvement of BCOR as a fusion gene with ZC3H7B. The absence of substantial histological differences between ITD–BCOR-positive CCSK and those carrying YWHAE–NUTM2, together with the simultaneous presence of BCOR molecular anomalies and YWHAE–NUTM2 fusion in the ESS, as well as the evidence that various BCOR aberrations (such as ITD, ZC3H7B–BCOR and BCOR–CCNB3) lead to the overexpression of BCOR and of cyclin D1 [34–36,40,41,43,44,52,69,70,128,140] are all elements that reinforce the hypothesis that these alterations activate a common pathogenetic pathway, or at least a partially shared one.

It is also interesting that the link between CCSK and SRBCS due to the common involvement of BCOR was also strengthened by the findings of YWHAE–NUTM2B/E and ITD–BCOR in a group of URCS of soft tissue, the last previously identified also in SRBCS [126], as well as by the description of three cases of CCSK carrying the BCOR–CCNB3 fusion gene [77].

Another example of biological intersection between these tumors is the finding of ZC3H7B–BCOR fusion typical of low-grade ESS in some cases of SRBCS [55] and in one case of ossifying fibromyxoid tumor [78], a rare soft tissue tumor.

CNS neoplasms

BCOR is hypothesized to have an active role in neuronal development. In fact, BCOR expression patterns in several tissues during the various phases of embryonic development, conducted by Wamstad and Bardwell on mouse models, revealed intense BCOR expression in the neural tube and retina during development [79].

Therefore, it is not surprising that several CNS tumors carry BCOR alterations (Table 2). Mutations with supposed loss of function (e.g., nonsense, frameshift, splice sites and deletions) have been identified in up to 13% of retinoblastomas [80–82]; between 1 and 14% of various glial tumors, particularly in high-grade tumors [72–74] and also in pediatric patients [73,140], with a 25% peak found in recurrent high-grade astroblastoma [74]. Moreover, BCOR genetic alterations also appear in 2–8% of medulloblastomas [42,75,76,141,142], and in particular, such aberrations seem to be more represented in SHH-driven cases, especially in infant patients [42].

CNS-HGNET-BCOR

Primary CNS neuroectodermal tumors (CNS-PNET) are highly malignant neoplasms that predominantly affect children but may also occur in adolescents and adults. Histologically, CNS-PNETs are characterized by poorly differentiated embryonic cells, with a propensity both for glial and neuronal differentiation [143]. The WHO classification of 2016 [71] has removed CNS-PNET as an independent group of tumors, grouping them in a class of ‘embryonic tumors’ with medulloblastoma, as an example. In the attempt to establish new classification criteria to better define the molecular nature of CNS-PNET, Sturm et al. [140] proposed to evaluate a large series of CNS tumors via genomic methylation patterns. From this, new molecular entities emerged, clustering in well-defined methylation subgroups, but often lacking histological homogeneity. One of these new histologically nonhomogeneous entities was shown to carry in the vast majority of cases ITD–BCOR alteration. This group of tumors, defined CNS-HGNET-BCOR (high-grade primitive neuroectodermal tumors of the CNS with alterations of BCOR), showed a remarkable overexpression of BCOR similar to what is found in other tumors-carrying ITD-BCOR (Table 2). Curiously, as reported by Santiago et al. [52], the triad ‘CCSK/PMMTI/CNS-HGNET-BCOR’, after the group ‘rhabdoid tumor of the kidney/extra-renal rhabdoid tumor of soft tissues/atypical teratoid-rhabdoid tumor of the CNS’, would constitute the second trio of tumors joined by molecular anomalies and kidney, soft tissues, and brain districts. It is however necessary to highlight that Yoshida et al. recently reported important immunophenotypical differences between CNS-HGNET-BCOR and CSSK/PMMTI [144].

In addition to the recurrence of ITD–BCOR, CNS tumors and mesenchymal tumors with BCOR abnormalities also share common characteristics regarding transcriptional regulation. In fact, WNT and in particular SHH pathways, wherein up-regulation had already been reported as a typical feature of CCSK [145], were also found up-regulated in a CNS-HGNET-BCOR patient [146], as well as in many cases of retinoblastoma [147] and even represented the driver element of two subgroups of medulloblastoma [148]. The recurrence of the deregulation of these signaling pathways in at least two entities carrying ITD–BCOR supports the hypothesis that such deregulation may actually affect all ITD–BCOR-positive tumors, and it can be envisaged as a potential therapeutic target [146].

It is also interesting to note that, like mesenchymal tumors, also among CNS tumors with BCOR involvement, there are numerous cases arising in the pediatric setting. This evidence can indicate that, coherently with the functions of BCOR at the embryonic level, mutations in this gene lead to a premature disruption of the differentiation pathway in progenitor stem cells. However, further studies are necessary to support this hypothesis.

Hemolymphopoietic system neoplasms

There is evidence regarding the importance of BCOR for physiological hematopoiesis. Mutations of this gene in knockout organisms result in hematological abnormalities, consisting in functional deficiency of primitive erythrocytes and lymphocyte depletion, and confirming the relevance of BCOR on the activity of BCL-6, a known mitogenic agent for lymphoid cells [28]. Conversely, the loss-of-function of BCOR in murine bone marrow cells produces significantly enhanced proliferation and myeloid differentiation rates with upregulated expression of HOX genes [84]. Even if the introduction of next-generation sequencing (NGS) techniques for the molecular screening of large cohorts of patients affected by hemolymphopoietic system neoplasms highlighted the clear involvement of BCOR alterations in these diseases, at present it is still challenging to clarify the biological role of BCOR in myeloid and lymphoid precursors and to translate this molecular knowledge for clinical applications. Actually, there are still few reports aimed at clarifying the role of BCOR in lymphoid and myeloid oncogenesis [85,86]. Tanaka et al. generated transgenic mice harboring exon 4 deletion leading to loss of the BCL-6-binding domain, which developed lethal acute T-cell lymphoblastic leukemia in a Notch1-dependent manner and demonstrated myc upregulation [85]. Another report by Lefebure et al. [86] showed that BCOR is a tumor suppressor in the Eμ-myc lymphoma murine model and its loss-of-function mutations act as myc-cooperating events in this setting.

Myeloid neoplasms

The involvement of BCOR alterations in clonal disorders of myeloid lineage (Table 2) has been assessed primarily for myelodysplasias [83,87,91–93] and acute myeloid leukemia (AML) [94,102–104], with cases ranging from less than 1–10% in unselected cohorts. By analyzing these and other cohort studies, it was possible to identify clinical contexts where the presence of BCOR alterations is most frequent. Particularly in AML, it is clear that the frequency of BCOR mutations is significantly higher in older patients than in pediatric ones [90,94,96,101], with pediatric AMLs with normal karyotype as the exception [88] and higher in secondary AML with respect to primary AML [102]. NGS techniques allow for the identification of new molecular subgroups potentially related to specific clinical characteristics, in addition to those already identifiable by cytogenetic studies. In the context of these specific entities, association with BCOR comutations (Table 2) is particularly relevant in the following instances: intermediate cytogenetic prognosis and FLT3–ITD-negative AML in adult patients under 65 [89], nonselected myeloid neoplasms with mutations in cohesin-coding genes [95], AML with changes associated with myelodysplasia [92], AML with trisomy 11 or 13 [99,100], AML with balanced or unbalanced chromosomal rearrangements [97], AML with normal karyotype [97,98,107] (in particular, pediatric AML) [88] or if co-occurring mutations in NPM1, CEBPA, FLT3-ITD, MLL-PTD or RUNX1 are present [105,107,117].

The high percentage of BCOR involvement in patients with chronic myelo-monocytic leukemia [92,93,106] or with acute leukemia with mixed phenotype is also noteworthy [149]. Lastly, two cases of a variant of acute promyelocytic leukemia have been described for which BCOR represented the partner of RARα (the transcription factor typically involved in APL) in the chimeric gene (Figure 3E) [150,151].

BCOR alterations found in myeloid clonal diseases are apparently clustered around exon 4 and for the most part are mutations with presumed loss-of-function. This consideration derived both from the analysis of the type of alterations found and the evaluated expression of HOX genes, the target of PRC activity when these genes are strongly overexpressed. The prognostic significance of the presence of BCOR mutations in myeloid clonal diseases was also evaluated. Damm et al. identified the presence of mutated BCOR as an independent negative prognostic factor in MDS [92]. Regarding adult AML, Terada et al. identified BCOR mutations as unfavorable prognostic factors in 5-year overall survival (8.7% BCOR mutated vs 34.0% BCOR WT) [89].

Finally, BCOR alterations have been identified and/or described in several studies conducted on acquired aplastic anemia patients, in percentages similar to those described for nonselected MDS and AML [110–112,152,153]. This suggested the possibility that these mutations may contribute to clonal selection during the onset of myeloid malignancies on the aplastic anemia background.

Lymphoid neoplasms

In the lymphoid line, there is evidence of a significant presence of BCOR genetic alterations, seemingly all causing loss-of-function of the corresponding protein. BCOR alterations have been identified in up to 9% of prolymphocytic T leukemias [114,115], chronic lymphocytic leukemias [108,109,113,116] and, respectively, in 24% and up to 32% of the evaluated cases of diffuse splenic lymphoma of the red pulp [122] and of extranodal NK/T-cells lymphoma, nasal type [118,123]. The latter one is an aggressive lymphoma, strongly associated with EBV infection, which arises primarily, but not exclusively, in nasal and paranasal areas. Considering the high recurrence of mutation and inferences deriving from murine models, it is clear that a wider evaluation of BCOR involvement in these tumors is warranted.

Carcinomas

The functional studies conducted on BCOR showed that the activity of this gene primarily affects cells committed to the mesenchymal and neuroectodermal lineages. In fact, knockout organisms have shown alterations, especially in tissues involved in OFCD syndrome, as well as important hematological abnormalities [28]. Analysis of BCOR expression in various tissues during murine embryogenesis also confirmed strong expression in the eye, limb buds and branchial arches, neural tube, adnexal and nervous system tissues, and craniofacial structures [79]. These studies are partly strengthened by the analysis of tumor histotypes associated with BCOR mutations. The most represented tumors are stromal tumors, CNS neoplasms and hemolymphopoietic system tumors, sometimes concentrated in the head and neck or upper chest districts in organs such as the thymus [119] and salivary glands [120]. In the context of thymic tumors, it is particularly interesting that BCOR mutations recur in 50% of cases of aggressive B3-thymoma [121]. Considering the high percentage of recurrence, an active role of BCOR in the pathogenesis of this thymoma subtype, as well as the potential use of this gene as a classifying marker, is hypothesized. As for salivary glands tumors, the occurrence of BCOR mutations seems to characterize more aggressive diseases, regardless of the histological type. This is particularly true of adenoid cystic carcinoma, which is a rare type of cancer that most frequently occurs in salivary glands, but which can also be found in breast, lacrimal gland, lung, Bartholin’s gland, trachea and paranasal sinuses [154]. While definitive publications on the recurrence and role of BCOR mutations in various epithelial malignancies are still lacking, it is evident from comprehensive sequencing efforts (https://portal.gdc.cancer.gov; www.cbioportal.org) that BCOR abnormalities are recurrent in uterine corpus endometrial carcinoma, colon and stomach adenocarcinoma, and lung tumors regardless of histology. In the context of The Cancer Genome Atlas project, in the EBV + gastric carcinomas subgroup, the percentage of mutations involving BCOR reached 23%. All mutations were nonsense mutations, therefore suggesting a complete loss-of-gene function. The high involvement of BCOR mutations in EBV + gastric carcinomas is emphasized by the consideration that BCOR also affects another EBV-related tumor: extranodal NK/T-cell lymphoma, nasal type.

A special case is that of the uterus because in this organ, BCOR alterations are recurrently involved both in epithelial tumor histotypes (in this case, endometrioid endometrial adenocarcinoma) [155] and in uterine carcinosarcoma (involving BCOR in 23% of cases) [156], in addition to the previously described ESS.

Conclusion

Somatic alterations of BCOR are found in many tumor subtypes, with fusion genes discovered in rare subclasses of sarcomas, internal tandem duplications (ITD) in almost all pediatric clear cell sarcomas of the kidney, in a subgroup of undifferentiated pediatric sarcomas (undifferentiated round cell sarcoma of infancy, primitive myxoid mesenchymal tumor of infancy), in endometrial stromal sarcoma and in the subset of high-grade neuroepithelial tumor of the central nervous system with BCOR gene alteration (CNS HGNET-BCOR). Additionally, loss-of-function mutations were also found in myelodysplasias and acute myeloid leukemias and in selected subtypes of lymphoid neoplasms. A deeper knowledge of the functional role of different BCOR alterations in tumors of the mesenchymal, hematopoietic and neuroectodermal lineages will allow the identification of actionable pathways in these malignancies.

Future perspective

Deeper molecular knowledge of oncology enables better identification of diagnostic or prognostic markers and/or therapeutic targets. The recent introduction of NGS techniques has exponentially amplified the potential to obtain such data and to compare and find similarities among the molecular landscapes of histologically distant tumor types. This review has shown that BCOR represents one of the possible genes and a viable factor to focus our efforts. In fact, the recognition of recurrent alterations of this gene in tumors such as CCSK, PMMT1, SRBCS and CNS-HGNET-BCOR already plays a key role in the diagnosis of these malignancies and there are many examples of new entities wherein BCOR plays a relevant clinical role. Further work will be necessary to dissect the activity of different BCOR mutations in specific cell contexts, and mechanistic and functional studies are warranted to understand the overall effects of these alterations on PRC1.1 function and target recruitment.

We believe that a better knowledge of PRC activity and the epigenetic regulators in which BCOR is an important component can lead to the advancement of new therapies.