Abstract

VHL is the most frequently mutated gene in ccRCC

Inactivation of the VHL tumor suppressor gene is the causal event in the pathogenesis of clear cell renal cell carcinoma (ccRCC). Approximately 75% of the RCCs are of the clear cell type. Among them, 70–80% of ccRCC tumors harbor biallelic inactivation of VHL through mutation, deletion, or hypermethylation of its promoter that results in loss of expression [1]. In hereditary kidney cancer patients, a germline VHL mutation predisposes individuals to earlier onset bilateral kidney cancer. Because one allele is already defective, only the remaining copy of VHL must be altered for biallelic inactivation to occur. pVHL, the protein product of the VHL tumor suppressor gene, encodes the substrate recognition unit of an E3 ubiquitin ligase complex, which also includes Cul2 and Rbx1. This complex targets the α subunits of the heterodimeric transcription factor hypoxia-inducible factor (HIF) for poly-ubiquitylation and proteasomal destruction. pVHL recognizes HIFα only when HIFα is hydroxylated on either of two prolyl residues by members of the EglN family (also called PHDs or HPHs) (Fig. 1). When pVHL is inactivated, HIFα proteins are synthesized, accumulate and form a complex with HIF1β protein. The complex then activates transcriptional response to hypoxia in the nucleus. The constitutively active HIF activity subsequently drives tumorigenesis and growth of ccRCC tumors [2]. Interestingly, not all HIF-induced genes are tumor-promoting [3,4]. pVHL also has HIF-independent functions, but their relevance to tumor suppression remains unclear[5–7].

Open in a separate window

Figure 1

Schematic representation of HIF regulation by pVHL. pVHL is the substrate recognition subunit of an E3 ubiquitin ligase complex that contains Elongin B, C, Cul2 and Rbx1. Hypoxia-inducible factor α subunits (HIFα) are constantly translated. In the presence of oxygen, prolyl hydroxylases promote proline hydroxylation on HIFα proteins, which leads to recognition by pVHL and proteasome-mediated degradation. Tumor derived VHL mutations would either destroy the domain that mediates its interaction with the hydroxyl-prolines on HIFα, or disrupt the domain that is responsible for Elongin C binding, leading to HIFα stabilization and constitutive activation of HIF. This in turn leads to ccRCC initiation and development.

Restoration of pVHL expression and function in VHL-defective kidney cancer cells suppresses their ability to form tumors in nude mice[8], while stabilization of HIF2α in VHL-proficient cancer cells drives tumor growth [9,10]. Conversely, reduction of high-level endogenous HIF2α expression in VHL-defective kidney cancer cells severely blocks their ability to form tumors in a xenograft model [11,12]. Consistent with the notion that an abnormally activated HIF pathway is critical to tumor growth and maintenance, sunitinib, a small molecule antagonist against VEGF and PDGF receptors (which are activated by two HIF-responsive genes), generates a 30% to 40% response rate and a five-month increase in overall survival (OS) of kidney cancer patients [13]. Other anti-angiogenesis drugs such as sorafenib, pazopanib, and axitinib hit the same targets and have similar effects. However, only a third of the patients respond strongly to these drugs, and almost all the tumors will develop drug resistance over time. Given that so many drugs are directed at the same molecular targets and that a large portion of ccRCC patients are not benefiting from them, other drug target(s) are urgently needed.

PBRM1 is another key tumor suppressor in kidney cancer

While VHL is frequently mutated in ccRCC and is has been shown to be essential to development in this disease, it is a well-known fact that most types of cancer harbor multiple driver mutations, sometimes dozens of them, which collectively establish the hallmarks of cancer that are too numerous for one mutation to achieve [14,15]. Large-scale sequencing efforts have been undertaken to uncover genes mutated at a significant level in addition to VHL mutations in ccRCC. Varela et al. reported that 41% of ccRCC tumors harbor inactivating mutations in a SWI/SNF chromatin remodeling complex gene known as PBRM1 [16]. Interestingly, PBRM1 is also mutated in Bladder carcinoma, Chromophobe and Papillary Renal Cell Carcinoma (Table 1), although at lower frequencies. The high rate of PBRM1 mutations in ccRCC have been confirmed by multiple studies, along with mutations in several other tumor suppressor genes, including BAP1, SETD2, JARID1C and PTEN, as well as the histone H3 lysine 27 demethylase gene UTX (KMD6A)[17–21]. However, the mutations of these other genes occur at much lower rates than that of PBRM1.

PBRM1 is likely a tumor suppressor as the reported mutations are mostly inactivating ones; the main PBRM1 mutations that have been observed are missense, frameshift truncation, frameshift insertion, and nonsense mutations, all of which are inactivating in nature. Knockdown of PBRM1 has been shown to significantly increase cell proliferation, cell migration, and colony formation in soft agar, indicating that PBRM1 loss leads to a transformed phenotype. Transcriptionally, PBRM1 suppression led to altered pathways governing chromosome instability and cell proliferation [16]. The importance of PBRM1 mutations to ccRCC development is further supported by the following discovery: PBRM1 mutations, like those of VHL, occur early in ccRCC tumorigenesis, with mutations being present in all cancer cells within a tumor in many cases. Notably, however, it was discovered through multi-region exome sequencing that ccRCCs display profound intra-tumor heterogeneity [22–24]. The majority of the driver mutations, including those of BAP1, SETD2 and JARID1C, are subclonal, meaning that they occur later during tumor development and are present only in a subpopulation of the cancer cells. Based on the mutations identified from different regions of a tumor, phylogenetic trees of mutations were constructed and the driver mutations were mapped onto them. VHL mutations were ubiquitous and mapped to the trunk of the mutational phylogenetic trees in the all cases, suggesting that it is the fundamental event of ccRCC development. PBRM1 mutations were mapped onto the trunk in ~50% of cases when mutated, suggesting that in many cases PBRM1 mutations happen very early as well. The mutations of all the other genes mentioned above were mapped onto the branches, suggesting that they occur later during tumor development [23]. Collectively, these lines of evidence strongly suggest that, like VHL, PBRM1 is a key tumor suppressor in ccRCC.

To date, the contribution of PBRM1 mutations to the clinical outcome of ccRCC patients has been controversial. Some groups reported that PBRM1 mutations occurred at a similar rate in different tumor stages, and these mutations did not seem to correlate with adverse patient survival[25,26]. However, other groups have reported that PBRM1 mutations are positively linked to tumor invasiveness[27] and, based on IHC findings, loss of PBRM1 protein was associated with advanced tumor stage, high Fuhrman grade and poor overall survival[28,29]. Our own studies of PBRM1 protein loss in ccRCC tissue microarrays also indicate that defects in PBRM1 are strongly linked to higher tumor stages (Jiang, et al., unpublished). Thus, it is possible that PBRM1 protein deficiency itself is strongly tumor promoting, but the dampening effects of other concurrent mutations may mask its effect.

With the high percentage of ccRCC tumors carrying PBRM1 mutations and deficiency, understanding its molecular functions, its contribution to clinical disease progression and outcome, its interaction of other mutated genes in ccRCC, and how to suppress or kill PBRM1-deficient ccRCC cancer cells therapeutically are important areas of renal cancer research. It is worth mentioning that expression of ARID1A, another targeting subunit of the SWI/SNF complex that competes with PBRM1 for binding to core subunits, is absent or downregulated in more than half of ccRCC tumors [30,31]. Since its mutation rate in ccRCC is low[16,20], the widespread loss of protein expression likely arises through other mechanisms such as promoter methylation and/or silencing by microRNA.

PBRM1’s molecular and biological functions

PBRM1 was originally discovered as BAF180, which is a component of a SWI/SNF-B complex that resembles the yeast Rsc chromatin remodeling complex [32]. This complex also contains BRD7 and BAF200 as the specificity subunits [33]. This complex is called PBAF to distinguish it from the BAF complex that has ARID1A or ARID1B as specificity subunits. Both PBAF and BAF complexes share the core subunits such as BAF155/BAF170, BAF57, BAF47, ACTL6, BAF60, and the enzymatic subunits BRG1. Both complexes possess ATP-dependent mono-nucleosome disruption activity that can slide or eject nucleosomes on DNA [32]. PBRM1 deficiency in mice leads to embryonic lethality[34], thought to result from a failure in cardiac chamber maturation. It was later discovered that PBRM1-deficient mice more likely have a failure in coronary vessel development [35]. Deletion of PBRM1 in mouse T cells enhanced Th2 differentiation and significantly elevated IL-10 expression [36]. PBRM1 is mutated in some human breast cancers, and its re-expression inhibits cell proliferation and induces p21 expression [37]. Furthermore, PBRM1 and BRD7 are found to be critical for p53 to induce replicative senescence [38]. Interestingly, genes that are regulated by PBRM1 are highly dependent on the cell state, as its repression targets are very different in resting and stimulated Th2 cells [36]. PBRM1 was also found to be required for retinoic acid-dependent gene activation [34]. Recently, PBRM1 was found to be required for centromeric cohesion and prevention of genomic instability [39], and it is also reported to help repress transcription and repair at DNA damage sites [40]. Some tumor-derived PBRM1 mutants failed to perform these functions, suggesting that they might contribute to PBRM1’s tumor suppressor functions [39,40].

The bromodomain 2 of PBRM1 has high affinity for acetylated K14 on histone H3 (H3K14Ac)

The PBRM1 protein contains six bromodomains, two bromo-adjacent homology (BAH) domains, and one high-mobility-group (HMG) motif (Fig. 2). The bromodomains are well-characterized motifs that bind to acetylated lysine [41]. The BAH domains are proposed to be involved in protein-protein interactions, and the HMG motif has been shown to bind to DNA [42]. It was hypothesized that PBRM1’s ability to regulate gene expression is dependent upon its bromodomain-mediated interaction with lysine-acetylated histones. In one investigation, individual bromodomains of PBRM1 displayed high affinity for different acetylation sites on histone H3 tails, with the dissociation constants (K_d) in the sub-micromolar range [43]. Another report suggested that the bromodomain 2 of PBRM1 preferentially recognizes the acetylated lysine 14 on histone H3 (H3K14Ac) (Fig. 3), with a K_d of about 500μM [44]. Because of the many discrepancies among these reports, it will be important to characterize the specificity of full-length PBRM1 toward acetylated-lysines on histone H3. It is highly probable that PBRM1, through its multiple bromodomains, would interact with other lysine-acetylated tumor suppressors or oncogenes to exert its tumor suppressor functions.

Open in a separate window

Figure 2

The tumor-derived somatic mutations on the newly discovered cancer genes in ccRCC. PBRM1 has six bromodomains (known to bind to acetyl lysines), two BAH domains (postulated for protein-protein interaction) and one HMG domain (known to bind to DNA), and the mutations are inactivating ones [16]. BAP1 has an UCH domain that possesses deubiquitinase activity, and an H domain that mediates interaction with HCF-1. There are numerous mutations affecting the UCH domain, including many missense mutations[19]. The SET domain in SETD2 is required for its histone demethylase activity, and the mutations seem to be evenly distributed[27,85]. The Jumonji domains in JARID1C are important for its histone demethylase activity. Many identified mutations are either inside or near the Jumonji domains[27]. The Jumonji domain is also critical to the histone demethylase activity of UTX[79].

Open in a separate window

Figure 3

The newly identified tumor suppressors in ccRCC are reader, writer, and erasers of adjacent key histone marks on the tails of histone H2A and H3. PBRM1 recognizes H3K14Ac, BAP1 deubiquitinates K119ub on H2A, SETD2 methylates H3K36Me3, JARID1C demethylates H3K4Me3, UTX demethylates H3K27Me3.

BAP1 is a key tumor suppressor in ccRCC and other cancers

Inactivating mutations of the BRCA1 associated protein-1 gene (BAP1) gene have been reported in 10–15% of ccRCC[18,19] (Fig. 2 and Table 1). In other urological malignancies such as Bladder carcinoma and Papillary Renal Cell Carcinoma, BAP1 mutations are also found at lower frequencies (Table 1). BAP1 was initially discovered as a BRCA1-interacting protein in a yeast two-hybrid screen[45], and BAP1 protein inhibited the E3 ubiquitin ligase activity of BRCA1/BARD1 complex [46]. However, this interaction was not easy to detect in some systems, suggesting that they might function together under specific conditions. Notably, the BAP1 gene is located on the short (p) arm of chromosome 3, at band location 3p21, very near (<140,000 base pairs away) from the PBRM1 gene. SETD2 and VHL are also located in chromosome 3p, and allelic losses of all four of these tumor suppressor genes are frequently observed in ccRCC tumors due to chromosomal deletions of large segments of 3p. As with VHL, PBRM1 and SETD2, biallelic inactivation of BAP1 via mutation and allelic loss of the remaining wild type allele fit a two-hit tumor suppressor profile.

Germline mutations of BAP1 predispose patients to ccRCC and other type of cancers

Somatic mutations of BAP1 have been reported in nearly 85% of metastasizing uveal melanomas and in 20–25% of malignant pleural mesotheliomas [47–49]. In 2011, Testa et al. reported germline mutations of BAP1 in two families with multiple mesothelioma and various other malignant tumors[49]. One of the two families also exhibited two uveal melanomas and two skin cancers, whereas several BAP1 mutation carriers in the second family were afflicted with carcinomas of the kidney, breast or ovary. In a concurrent report, Wiesner et al. described germline BAP1 mutations in two families with high incidence of benign atypical melanocytic tumors as well as several uveal melanomas and cutaneous melanomas[50]. Collectively, the observations in these high-risk cancer families have led investigators to propose the existence a BAP1 cancer susceptibility syndrome characterized by a high incidence of malignant mesothelioma, uveal and other melanocytic neoplasms, and most likely in other cancers[51]. The latter possibility is now certain, based on the subsequent identification of multiple BAP1 families with one or more ccRCCs[51–53]. Other studies suggest that the clinical phenotype of the BAP1 syndrome may be expanded to other cancers such as lung adenocarcinoma, meningioma, breast carcinoma and paraganglioma[54–57].

In vivo studies have revealed that systemic or hematopoietic-restricted homozygous deletion of Bap1 in adult mice recapitulates features of human myelodysplastic syndrome[58]. Bap1 deletion was found to result in decreased levels of host cell factor-1 (HCF-1) and O-linked N-acetylglucosamine transferase, indicating a critical role for BAP1 in stabilizing these epigenetic regulators. To address experimentally the question of why mesothelioma is the predominate malignancy in some BAP1 syndrome families but not in others, Xu et al. exposed heterozygous Bap1 knockout mice to asbestos[59]. They found that germline mutation of Bap1 accelerates the rate of development and progression of asbestos-induced mesothelioma, although no spontaneous mesotheliomas were seen in unexposed Bap1 knockout mice followed for up to 21 months of age. Drawing parallels to the human disease counterpart, these in vivo genetic findings indicate that BAP1 mutation carriers are predisposed to the tumorigenic effects of asbestos and suggest that high penetrance of mesothelioma requires such environmental exposure.

The molecular functions of BAP1 that might be important to its tumor suppressor functions

BAP1 is a histone deubiquitinase, with the ubiquitin C-terminal hydrolase (UCH) domain located at its N-terminal domain (Fig. 2). It also harbors a BARD1 interaction domain (BA), and its interaction with HCF-1 needs the NHNY sequence (H). The C-terminus of BAP1 also contains a UCH37-like domain (ULD), a YY1-bindng domain (Y), and two nuclear localization signals (NLS) (Fig. 2). BAP1 was found to deubiquitinate the mono-ubiquitinated K119 residue on histone 2A (Fig. 3), and possibly other proteins as well[58,60]. However, its ability to repress cell growth did not seem to be associated with its ability to deubiquitinate H2A. As BAP1 can both promote and suppress cell proliferation based on the cell types, its ability to repress cellular growth in vitro does not necessarily correlate with its major tumor suppressor function in vivo[61]. Indeed, even though BAP1’s deubiquitinase activity is not required to suppress cell growth in vitro, its N-terminal UCH region is often targeted by tumor-derived mutations, suggesting that this activity is important to it to be a tumor suppressor.

In addition to its deubiquitinase activity, BAP1 also has a HBM motif that mediates the interaction with HCF-1, which could serve as a chromatin scaffold for a few chromatin-remodeling complexes[62–64]. Interestingly, BAP1 with a mutation in the HBM motif that failed to interact with HCF-1 also failed to repress cell growth in RCC cells[19]. It remains to be determined whether its deubiquitinase activity and/or its ability to interact with HCF-1 is required for its tumor suppressor function.

In addition to a potentially critical role for BAP1 inactivation in epigenetic deregulation in ccRCC, BAP1 has been found to promote DNA double-strand break repair by homologous recombination, thereby adding to its tumor suppressor function[24]. In other work, germline mutations in BAP1 were found to impair its function in DNA double-strand break repair[65]. BAP1 was shown to mediate rapid poly(ADP-ribose)-dependent recruitment of the polycomb deubiquitylase complex PR-DUB to sites of DNA damage and was identified as a phosphorylation target for the ATM, further suggesting a DNA repair response mechanism that also contributes to the tumor-suppressive function of BAP1.

BAP1 mutations are not ubiquitous in ccRCC but do associate with tumor aggressiveness and poor outcome

The mutations rate of BAP1 was 10–15% compared to PBRM1’s 30–40%. The mutations of BAP1 are mostly inactivating mutations (Fig. 2). Multiple studies have revealed that in a substantial portion of ccRCC tumors with BAP1 mutations, the mutations are subclonal and the protein losses are focal[21–24,66]. Such tumor heterogeneity suggests that BAP1 alterations occur late in disease progression. Interestingly, in ccRCC tumors, the mutations of BAP1 and PBRM1 are largely mutually exclusive[19,25]. Such a reciprocal pattern usually occurs when two genetic mutations affect the same pathway, such that there is no selection pressure to mutate the second gene once the first is inactivated. However, this does not seem to be the case for BAP1 and PBRM1, as both can occur in the same tumor.

Using whole-genome and exome sequencing, Peña-Llopis et al. identified inactivating mutations of BAP1 in 15% of ccRCCs, and BAP1 loss was associated with high tumor grade[19]. In a retrospective analysis of 145 patients with primary ccRCC, Kapur et al. reported that median overall survival was significantly shorter for patients with BAP1-mutant tumors than for patients with PBRM1-mutant tumors[25]. A second independent cohort from The Cancer Genome Atlas (TCGA) was used for validation, with similar differences between these two patient groups. Similarly, a study of 609 patients with ccRCC from two distinct cohorts revealed that mutations of 3p21 epigenetic regulators BAP1 and SETD2 were associated with worse cancer-specific survival (CSS), suggesting that they play key roles in disease progression[26]. In an independent European study of ccRCC, BAP1-mutated tumors were associated with metastatic disease at presentation, advanced clinical stage, and a trend towards shorter recurrence free survival when compared with tumors exclusively mutated for PBRM1[67]. These results provide further support indicating a potential role for BAP1 and PBRM1 mutations in the risk stratification of ccRCCs. Based on their roles as ccRCC driver genes and the function of their gene products, investigators have proposed ways in which mutations in these genes may be exploited therapeutically[68].

SETD2 gene

SETD2 protein is a histone H3 Lysine 36 trimethylating (H3K36Me3) enzyme (Fig. 3), and it is mutated in 10–15% of ccRCC tumors[17,20,21,26] (Fig. 2 and Table 1). In other urological malignancies SETD2 mutations are found as well (Table 1). As mentioned before it is also located on chromosome 3p and it is a two-hit tumor suppressor. Mutational analysis indicated that SETD2 mutations tended to be subclonal[22–24], and sampling analysis found that different SETD2 mutations could exist in different regions of the same tumor[22], suggesting high selection pressure for SETD2 mutations. It was suggested that SETD2 mutations were associated with worse cancer specific survival[26]. Furthermore, SETD2 mutations were found to occur much more frequently when PBRM1 was mutated, suggesting that these mutations were cooperating with each other[69].

It is not clear how SETD2 mutations lead to ccRCC. Without SETD2, the H3K36Me3 mark on histone is significantly down-regulated. This mark is usually associated with active transcription, although it is also reported to be associated with alternative splicing, transcriptional repression[70]. Indeed, it was reported that SETD2 mutations the ccRCC tumors associated with changes in chromatin accessibility and DNA methylation[71], and another group reported that SETD2 mutations were linked to variation in chromatin accessibility and widespread RNA processing defects[72]. Multiple reported also revealed that SETD2 played an important role in DNA damage repair through promoting homologous recombination, mismatch repair, and activation of p53-mediated checkpoint[73–75]. It remains to be seen which function of SETD2 plays the most significant role in ccRCC tumor suppression, and whether it is worthwhile to target the dysfunctional pathway associated with SETD2 inactivation in ccRCC.

JARID1C/KDM5C

JARID1C/KDM5A is a histone demethylase that takes off methyl group from lysine 4 on histone H3 (H3K4Me3) (Fig. 3). Its gene was inactivated with mutations in 3–7% of ccRCC tumors[16,17,20,21] (Fig. 2 and Table 1). Interestingly, its mutations were mostly subclonal, and if multiple regions of ccRCC tumors were sampled, 25% of them would harbor JARID1C mutations[24]. JARID1C/KDM5A mutations or amplification are also identified in other urological malignancies (Table 1). Its prognosis value is unknown, but JARID1C was found to be a HIF target[3]. Without JARID1C, the overall level of H3K4Me3 in VHL-defective kidney cancer cells would go up, and these cancer cells would form much larger tumors in a xenograft model, suggesting that HIF-induced JARID1C expression and function constituted a negative feedback loop in regard to tumor growth[3]. H3K4Me3 is a histone mark that is tightly linked to actively transcribing gene[76] as it provides the targeting site for TFIID[77]. Recently it was found that H3K4Me3 breadth is linked to cell identity, and the broadest H3K4Me3 domains exhibited enhanced transcriptional consistency rather than transcriptional levels[78].

UTX/KDM6A

UTX/KDM6A is a histone H3 lysine 27 demethylase (Fig. 3). It was mutated at a low frequency in ccRCC tumors [17,79] (Fig. 2 and Table 1). Strikingly, its mutation rate in Bladder cancer is quite high (26.1%), and its mutations are also discovered in Prostate adenocarcinoma and Papillary Renal Cell Carcinoma (Table 1). Very little is known how UTX mutations could impact ccRCC. However, it is worth noting that H3K27Me3 is the most prominent histone mark that is associated with repressed transcription[76], and the enzymes that modify this site such as EZH2 and UTX were frequently mutated and/or altered in human cancers[79,80]. How UTX inactivation contribute to ccRCC tumor biology remains to be determined.

Conclusion

The field of kidney cancer research had focused on VHL-HIF axis for two decades. It produced tremendous amount of insights and led to the development of anti-angiogenesis therapies against ccRCC, which are partially and transiently effective. The complete sequencing of the kidney cancer genomes in the last few years revealed that many chromatin regulators such as histone code reader (PBRM1), writer (SETD2), and erasers (BAP1, JARID1C and UTX) were also mutated in ccRCC to various extents, suggesting that these factors could contribute to ccRCC initiation, development, and evolvement into more aggressive states. It is also worth noting that the same genes are also significantly altered in other subtypes of kidney cancer and other urological malignancies (Table 1), indicating the pan-cancer significance of these genes, and suggesting that the lessons learned from ccRCC might be applicable to other cancers as well. Interestingly, even though the chromatin landscape is vast, these proteins all targeted toward a crowded area of histone H2 and H3 tails. Some of these sites were only a few amino acids apart (Fig. 3), suggesting that these proteins might be functionally linked. Currently their studies in the cancer biology are still in early stage, so it is hard to predict or speculate their therapeutic relevance. As they are tumor suppressor genes, direct inhibition of their activities is unlikely to generate positive outcome. However, as the story of VHL/HIF indicates, the loss of an important tumor suppressor will lead to the activation of major oncoprotein(s), whose inhibition could be therapeutically beneficial. Undoubtedly the intense studies of these factors in ccRCC will provide significant insights into how their losses contribute, either alone or in cooperation with HIF and/or each other, to cancer biology of ccRCC tumors. These studies will likely lead to the identification of new target(s) and development of new therapeutics against ccRCC tumors that are sorely needed.

Acknowledgments

Footnotes

References