Bap1 is essential for kidney function and cooperates with Vhl in renal tumorigenesis

Significance

Despite the discovery of the von Hippel–Lindau (VHL) gene in 1993, and that inactivating germ-line mutations of VHL cause multiple kidney lesions, including clear-cell renal cell carcinoma (ccRCC), Vhl inactivation in the mouse does not lead to ccRCC and a mouse model has been lacking. We discovered that the BRCA1-associated protein-1 (BAP1) two-hit tumor suppressor gene is mutated in ccRCC, and one BAP1 allele is frequently somatically codeleted with VHL in tumors. In the mouse, Vhl and Bap1 are on different chromosomes. We show that SIX homeobox 2 (Six2)-Cre;Vhl^F/F;Bap1^F/+ mice develop premalignant lesions and malignant ccRCC resembling VHL syndrome. More broadly, differences in tumor predisposition across species may result from differences in the location of two-hit tumor suppressor genes across the genome.

Abstract

Why different species are predisposed to different tumor spectra is not well understood. In particular, whether the physical location of tumor suppressor genes relative to one another influences tumor predisposition is unknown. Renal cancer presents a unique opportunity to explore this question. Renal cell carcinoma (RCC) of clear-cell type (ccRCC), the most common type, begins with an intragenic mutation in the von Hippel–Lindau (VHL) gene and loss of 3p (where VHL is located). Chromosome 3p harbors several additional tumor suppressor genes, including BRCA1-associated protein-1 (BAP1). In the mouse, Vhl is on a different chromosome than Bap1. Thus, whereas loss of 3p in humans simultaneously deletes one copy of BAP1, loss of heterozygosity in the corresponding Vhl region in the mouse would not affect Bap1. To test the role of BAP1 in ccRCC development, we generated mice deficient for either Vhl or Vhl together with one allele of Bap1 in nephron progenitor cells. Six2-Cre;Vhl^F/F;Bap1^F/+ mice developed ccRCC, but Six2-Cre;Vhl^F/F mice did not. Kidneys from Six2-Cre;Vhl^F/F;Bap1^F/+ mice resembled kidneys from humans with VHL syndrome, containing multiple lesions spanning from benign cysts to cystic and solid RCC. Although the tumors were small, they showed nuclear atypia and exhibited features of human ccRCC. These results provide an explanation for why VHL heterozygous humans, but not mice, develop ccRCC. They also explain why a mouse model of ccRCC has been lacking. More broadly, our data suggest that differences in tumor predisposition across species may be explained, at least in part, by differences in the location of two-hit tumor suppressor genes across the genome.

It is well established that different animal species are predisposed to a different spectrum of tumors. For instance, although it varies among breeds, dogs and mice (and humans) are not equally predisposed to particular tumor types. These differences may be explained by both genetic and environmental factors. In humans, at least in adults, cancers are thought to result from five to 10 somatic mutations in driver genes (1). Although some genes, such as tumor protein p53 (TP53), are broadly mutated across different tumor types, other cancer genes are associated with particular tumors (2). This association is most evident in familial cancer syndromes, where germ-line mutations in cancer genes predispose to specific tumor types (3). Cancer genes are broadly classified into oncogenes and tumor suppressor genes, which are most commonly involved in familial syndromes (2). Oncogenes are subject to activating mutations, typically involving one allele, and they induce tumor development. Tumor suppressor genes are inactivated by mutation; typically, one allele is incapacitated through an intragenic mutation, and the second is lost as part of a deletion [which results in loss of heterozygosity (LOH)]. Given that deletions are typically large (usually several megabases), the clustering of multiple tumor suppressor genes in a chromosomal region could affect tumor predisposition. Thus, differences in the location of tumor suppressor genes with overlapping tissue specificity of tumor suppressor action could account for differences in the spectrum of tumors across species.

Kidney cancer provides a unique opportunity to study the impact of tumor suppressor gene location on tumorigenesis (4). Kidney cancer ranks in the top 10 most common cancers in the United States, with roughly 60,000 new cases diagnosed every year (5). Over 70% of kidney cancers are clear-cell renal cell carcinoma (ccRCC) (6), and ccRCC is characterized by mutation of the von Hippel–Lindau (VHL) gene. VHL mutation is associated with familial as well as sporadic ccRCC (7, 8). Inactivating germ-line VHL mutations cause a form of VHL syndrome characterized by, among other features, a high incidence of both premalignant and malignant renal lesions, including ccRCC (9). The VHL gene is also mutated (or silenced) in >80% of sporadic ccRCCs (8, 10). VHL functions as a classic two-hit tumor suppressor gene, in which both alleles are inactivated during tumorigenesis. Typically, one allele is disrupted through an intragenic point mutation (or indel) and another as part of a large deletion, often involving the whole 3p chromosome arm (where VHL is located). Intragenic mutations in VHL and loss of 3p are early events during tumorigenesis (referred to as “truncal” events) and are thought to initiate ccRCC development (11, 12).

Recently, three other tumor suppressor genes were found to be frequently mutated in ccRCC: Polybromo-1 (PBRM1), BRCA1-associated protein-1 (BAP1), and Set domain-containing 2 (SETD2). Interestingly, all three of these tumor suppressor genes are located on chromosome 3p, within a 50-Mb region that also includes VHL, which is lost in the vast majority of ccRCCs (4). PBRM1, which encodes BAF180 (herein referred to as PBRM1), a component of a switch/sucrose nonfermentable (SWI/SNF) nucleosome remodeling complex, is mutated in ∼50% of ccRCCs (13). BAP1, which encodes a nuclear deubiquitinase, and SETD2, which encodes an H3K36 methyltransferase, are each mutated in ∼15% of ccRCCs (14–16). Notably, tumors tend to have mutations in either PBRM1 or BAP1, and simultaneous inactivation of both PBRM1 and BAP1 in the same tumor cells occurs at a lesser frequency than expected by chance alone (4, 15, 17). How SETD2 fits into this classification is less clear, but SETD2 loss seems to cooperate with PBRM1 loss in tumor development (4). Importantly, mutations in BAP1 and PBRM1 define two different subtypes of ccRCC with different gene expression and biology (15, 18). Furthermore, tumors deficient for either BAP1 or PBRM1 have markedly different outcomes; retrospective studies of patients with localized ccRCC have shown a significantly shorter kidney cancer-specific survival for patients with tumors deficient for BAP1 compared with tumors deficient for PBRM1 (17, 18).

Despite the important role of VHL in ccRCC development, a mouse model of ccRCC reproducing the molecular genetics is lacking. Whereas VHL heterozygous patients are predisposed to ccRCC, Vhl^+/− mice do not develop ccRCC (19). Notably, in the mouse, the Vhl gene is on a different chromosome than Pbrm1 and Bap1 (or Setd2). These discoveries provide a potential explanation for why VHL heterozygous humans, but not Vhl^+/− mice, develop ccRCC (LOH in the Vhl region in the mouse still leaves two copies of Pbrm1, Bap1, and Setd2 intact).

Results

Lineage Tracing Experiments Using a SIX Homeobox 2-Cre Driver Show Broad Reporter Expression in Nephron Cells.

The development of a mouse model of ccRCC is further confounded by uncertainty about the cell type of origin. Although ccRCCs are generally thought to arise from epithelial cells in the proximal renal tubule, this belief is largely based on correlative protein expression markers, and some studies have reported that both proximal and distal renal tubule cells may be involved (20–22). To investigate the role of Bap1 and Vhl in the kidney, we used a SIX Homeobox 2 (Six2)-Cre driver line, which expresses Cre recombinase in the cap mesenchyme of the developing kidney (23). Six2 is expressed in multipotent nephron progenitor cells (NPCs) that give rise to the different cell types of the main body of the nephron (23). To trace cells derived from those cells expressing Six2, we intercrossed Six2-Cre–expressing mice with mice harboring a CAG-loxP-stop-loxP-tdTomato reporter in the ubiquitous ROSA26 locus (Fig. S1A, Upper). Expression of Cre recombinase leads to recombination of loxP sites and indelibly labels these cells and their offspring with tdTomato. Lineage tracing experiments showed that glomerular and renal tubular cells were broadly derived from Six2-expressing progenitors (Fig. S1B).

Bap1 Is Expressed in Cells Derived from Six2-Expressing Progenitor Cells.

To determine whether Six2-Cre would be an appropriate driver to study Bap1 function in the kidney, we characterized Bap1 expression in the mouse kidney and compared Bap1-expressing cells with those cells derived from Six2-expressing progenitors. Specifically, we assessed the overlap between cells expressing Bap1 and the offspring of Six2-expressing NPCs. To assess Bap1 expression, we generated mice harboring a β-gal gene trap in the Bap1 locus (Fig. S1A, Middle). We then crossed them with the Six2-Cre;ROSA26-tdTomato mice. This approach allowed us to examine Bap1 expression (by assessing β-gal expression) and relate it to the offspring of Six2-Cre–expressing cells by evaluating its overlap with tdTomato. As determined by β-gal expression, Bap1 was expressed in many kidney cell types (Fig. S1B), and there was broad overlap between β-gal and tdTomato in the renal cortex (Fig. S1B). Thus, Bap1 was broadly expressed in the Six2 lineage. Among the Six2-lineage cells expressing Bap1, there were cells lining proximal tubules, as determined by costains using Lotus tetragonolobus lectin (LTL) (Fig. S1C). Overall, these data indicate that a Six2-Cre driver would induce Bap1 loss in many nephron cells that normally express Bap1, suggesting that it would be a fitting driver to study Bap1 tumor suppressor function in the kidney.

Bap1 Is Required for Normal Kidney Development and Function.

To evaluate the role of Bap1 in the kidney, we generated Six2-Cre;Bap1^F/F mice (F represents floxed allele, which becomes inactivated upon Cre expression; Fig. S1A, Lower). In these mice, Bap1 loss was induced primarily in the kidney (Fig. S1D). PCR analysis showed the Bap1 recombination product in kidney but not in other organs, such as brain, heart, and liver, although some level of recombination was observed in the stomach, consistent with previous reports (24). Quantitative RT-PCR (qRT-PCR) analyses of WT Bap1 mRNA levels in the kidney showed a 60% decrease in expression in Six2-Cre;Bap1^F/F mice compared with controls (Fig. S1E). The remaining Bap1 transcripts may arise from other cell types not derived from Six2-expressing cells, such as stroma or endothelial cells.

Six2-Cre;Bap1^F/F mice were born at expected Mendelian ratios, but newborns appeared sickly and were smaller compared with controls. Six2-Cre;Bap1^F/F mice uniformly died by day 30 (Fig. 1A). As determined by gross examination, the mutant kidneys were large and pale, and they harbored numerous cysts (Fig. 1B). H&E-stained sections showed that cysts could be detected in the Bap1-mutant kidneys as early as postnatal day 0 (P0) (Fig. 1C). Cortical atrophy was already observed in Bap1-mutant kidneys at P0, and the cysts became more pronounced over time (Fig. 1C). Residual glomeruli exhibited glomerular enlargement and glomerulosclerosis (Table S1). As mice became moribund, tubules appeared atrophied. The Bap1-mutant kidneys in moribund mice exhibited a number of other abnormal features, including cystically dilated tubules lined by epithelial cells with cytoplasmic clearing, rare epithelial proliferation, and mild nuclear atypia (Table 1).

Fig. 1.

Bap1 is required for normal kidney development and postnatal kidney function. (A) Survival curve showing reduced viability of Six2-Cre;Bap1^F/F mice compared with controls (n = 10 for each group; mice that were still alive at the end of the study period were censored. The estimates of percentage of survival were adjusted to account for the changing number of mice at risk due to the censoring.). (B) Representative image of kidneys from P15 Six2-Cre;Bap1^F/F mutant mice and control mice. (C) H&E staining of kidney sections from Six2-Cre;Bap1^F/F and Bap1^F/F (control) mice at P0 and P15. Arrows show examples of cysts in the Bap1-mutant kidneys. BUN (D) and Cr (E) measurements in P15 Six2-Cre;Bap1^F/F and control mice (n = 3 for each group). **P < 0.01; ***P < 0.001.

Table 1.

Pre- and neoplastic lesion characterization

Age	Genotype	Cytoplasmic clearing	Preneoplastic and neoplastic lesions		Immunohistochemistry
			Atypical cyst	Neoplastic micronodules	pS6	CAIX
2–3 wk (n = 5)	Six2-Cre;Bap1F/F	+	+	—	—	—
7–15 mo (n = 3)	Six2-Cre;Bap1F/+	—	—	—	++*	—
2–3 wk (n = 3)	Six2Cre;VhlF/F	—	—	—	—	—
9–11 mo (n = 2)	Six2Cre;VhlF/+	—	—	—	—	—
3–15 mo (n = 13)	Six2-Cre;VhlF/F	—	—	—	—	++†
2–3 wk (n = 5)	Six2-Cre;VhlF/F;Bap1F/+	—	—	—	—	—
2–6 mo (n = 7)	Six2-Cre;VhlF/F;Bap1F/+	+	+++	+++	+++‡	+++‡

+, mild/infrequently observed; ++, moderate/somewhat frequently observed; +++, strong/frequently observed.

^*In normal tubules.

^†In normal tubules and cysts.

^‡In neoplastic nodules and cysts.

The abnormal morphological features and cystic phenotype of the kidneys in Six2-Cre;Bap1^F/F mice suggested that kidney dysfunction was likely to be the cause of death. To evaluate kidney function, we measured blood urea nitrogen (BUN) and serum creatinine (Cr) levels. Both BUN and serum Cr were markedly elevated in Six2-Cre;Bap1^F/F mutants, providing an explanation for the early lethality (Fig. 1 D and E). Overall, these data demonstrate that Bap1 is essential for kidney development and function. In addition, cytoplasmic clearing, epithelial proliferation, and nuclear atypia suggest that Bap1 loss may predispose to tumor development.

Vhl Deletion in Kidney Progenitor Cells Causes Delayed Cystic Degeneration and Renal Failure.

We next evaluated the role of Vhl in NPCs. We crossed mice with a floxed Vhl allele (19) with Six2-Cre mice to generate Six2-Cre;Vhl^F/F mice. Six2-Cre;Vhl^F/F mice were born at expected Mendelian ratios and, in contrast to Six2-Cre;Bap1^F/F mice, did not appear sickly at birth. At 3 wk of age, Six2-Cre;Vhl^F/F mice had grossly normal kidneys that were indistinguishable from the controls (Table 1, Fig. S2, and Table S1). However, starting at ∼4 mo of age, Six2-Cre;Vhl^F/F mice began to die (Fig. S2A). Kidney sections from Six2-Cre;Vhl^F/F mutants at the moribund stage (from 90 to ≥250 d) revealed cystically dilated tubules with global glomerulosclerosis, tubular atrophy, interstitial inflammation, and fibrosis (Fig. S2C). BUN and Cr levels were significantly increased in moribund Six2-Cre;Vhl^F/F mice (Fig. S2 D and E), providing an explanation for the premature demise. These data show that loss of Vhl throughout the nephron results in cystic degeneration and renal failure over time and that Vhl is required for postnatal kidney function. Importantly, despite the presence of variably sized cysts and cystically dilated tubules in these mice, we did not observe premalignant features, such as epithelial proliferation, atypia, or cytoplasmic clearing (Table 1). The finding that mice with loss of Vhl fail to develop any signs of malignancy suggests that Vhl disruption, by itself, is not sufficient for renal tumorigenesis.

Vhl and Bap1 Cooperate in Renal Tumorigenesis, Leading to Renal Cell Carcinoma Development.

We hypothesized that tumor development may result from the combined inactivation of the Vhl and Bap1 genes, which are located on the same chromosome arm in humans. We intercrossed mice with floxed alleles of Vhl and Bap1 with a Six2-Cre driver line. Not surprisingly, Six2-Cre;Vhl^F/F;Bap1^F/F mice died before a month of age. However, we recovered Six2-Cre;Vhl^F/F;Bap1^F/+ mice.

At three weeks of age, the kidneys from Six2-Cre;Vhl^F/F;Bap1^F/+ mice looked morphologically and histologically normal (Fig. 2A, Table 1, and Table S1). However, MRI revealed the development of extensive abnormalities over time. T2-weighted images showed multiple rounded/ovoid, homogeneously hyperintense lesions consistent with fluid-filled cavities (Fig. 2B). Mutant kidneys displayed various abnormal features (Fig. 2C). Histologically, there was cortical atrophy with glomerulosclerosis, prominent tubular atrophy, interstitial fibrosis, and chronic inflammation (Fig. 2A). Glomerulomegaly and cystically dilated tubules were observed, suggestive of compensatory hyperfiltration nephropathy (Fig. 2A). These findings were more pronounced than those findings observed in Six2-Cre;Vhl^F/F mice, but renal dysfunction and survival in Six2-Cre;Vhl^F/F;Bap1^F/+ mice was similar (Fig. 2 D and E and Fig. S3).

Fig. 2.

Six2-Cre;Vhl^F/F;Bap1^F/+ mice develop premalignant and malignant lesions. (A) H&E staining of kidney from Six2-Cre;Vhl^F/F;Bap1^F/+ mice at 3 wk or moribund stage. (B) Coronal T2-weighted images of a Six2-Cre;Vhl^F/F;Bap1^F/+ mouse (368 d) demonstrating multiple rounded hyperintense lesions consistent with fluid-filled cysts (arrows) in both kidneys. Thin septations can be appreciated on these images (green arrowhead). (C) H&E staining of kidneys from Six2-Cre;Vhl^F/F;Bap1^F/+ mice showing abnormal features, including simple cysts lined by a single layer of epithelium (1 and 2); atypical cysts with multilayering and micropapillary epithelial proliferations (3); neoplastic nodules [1 (arrowheads) and 4–12]; enlarged nuclei (5–8); nuclear crowding, hyperchromasia, nuclear membrane irregularities, clumped chromatin, and nucleolar prominence (6–8); cytoplasm that was eosinophilic (6–8) or clear (9 and 10); and mitotic activity (5, 7, and 8; arrows), intratumoral thin-walled vasculature (11, arrows), and a focus suspicious for lymphovascular invasion (12, arrowheads). (Phenotype was observed in six of seven mice examined.) BUN (D) and Cr (E) levels were determined for controls, Six2-Cre;Vhl^F/F mice (also reported in Fig. S2 D and E), and Six2-Cre;Vhl^F/F;Bap1^F/+ mice at 3 wk (n = 3 for each group). **P < 0.01; ***P < 0.001. Note that age-matched control mice were used as controls for the “moribund” mutant mice, which became moribund at later ages.

Interestingly, Six2-Cre;Vhl^F/F;Bap1^F/+ kidneys developed renal cell carcinoma (RCC). Akin to kidneys from patients with VHL syndrome, we observed a spectrum of premalignant and malignant tumors. These lesions ranged from simple cysts/dilated tubules to atypical cysts and neoplastic nodules (Fig. 2 A and C and Fig. S4). Cysts ranged in size from 100 to 2,700 μm. Simple cysts were lined by a single layer of cuboidal epithelium (Fig. 2C, 1 and 2). Atypical cysts showed multilayering and micropapillary tufting (Fig. 2C, 3). The neoplastic nodules measured 250–1,800 μm in diameter [Fig. 2 A and C, 1 (arrowheads) and 4–12]. These expansile neoplastic nodules included both cystic and solid masses (Fig. 2C, 1 and 4–6). Architecturally, the neoplastic cells grew in either a micropapillary (Fig. 2C, 4) or solid cystic growth (Fig. 2C, 5 and 6) pattern. Cytologically, the neoplastic cells were characterized by enlarged nuclei (Fig. 2C, 6–8), nuclear crowding, hyperchromasia, nuclear membrane irregularities, clumped chromatin, and nucleolar prominence (Fig. 2C, 6–8 and Fig. S4 B–D). Their cytoplasm was eosinophilic (Fig. 2C, 5–8) or clear (Fig. 2C, 9 and 10), similar to cytoplasm seen in human ccRCC. In addition, the neoplastic cells displayed increased mitotic activity [Fig. 2C, 5, 7, and 8 (arrows) and Fig. S4C]. Arborizing, thin-walled vasculature, which is characteristic of human ccRCC, was also focally observed in neoplastic nodules [Fig. 2C, 11 (arrows) and Fig. S4D]. In addition, we found areas suspicious for lymphovascular invasion (Fig. 2C, 12 and Fig. S4A). Overall, these findings indicate that simultaneous inactivation of Vhl and one Bap1 allele is sufficient to trigger ccRCC development.

Fig. 3.

Premalignant and malignant lesions in Six2-Cre;Vhl^F/F;Bap1^F/+ mice stain for CAIX and show mTORC1 activation. Immunohistochemical analysis in kidneys of Six2-Cre;Vhl^F/F;Bap1^F/+ mice show increased nuclear Ki-67 expression in atypical cysts (A) and neoplastic nodules (B), membranous and cytoplasmic expression of CAIX in atypical cysts (C) and neoplastic nodules (D), and increased expression of phospho-S6 ribosomal protein (phospho-S6 Ser240/244) in some of the atypical cysts and neoplastic nodules (E and F). Six2-Cre;Bap^F/+ mice showed increased expression of phospho-S6 in some of the tubules (G) and a lack of CAIX expression (H). Six2-Cre;Vhl^F/F mice showed a lack of phospho-S6 expression in the cysts and the tubules (I) and patchy membranous CAIX expression in the tubules (J).

We also observed an increase in the number of cells staining for the proliferation marker Ki-67 (antigen identified by monoclonal antibody Ki67) in atypical cysts and neoplastic nodules (Fig. 3 A and B). Carbonic anhydrase IX (CAIX), an hypoxia-inducible factor (HIF)-target gene and a marker of ccRCC, had the classic membranous (and cytoplasmic) staining pattern in cysts and neoplastic nodules (Fig. 3 C and D). In addition, we observed increased staining in phosphorylated S6 ribosomal protein [pS6 (Ser240/244)] in the neoplastic cells (Fig. 3 E and F). Phospho-S6 is a marker of mammalian target of rapamycin complex 1 (mTORC1) activation, and we previously showed it to be up-regulated in BAP1-deficient human ccRCC (15). Parenthetically, staining with pS6, although weaker, was seen focally in morphologically unremarkable cortical tubules in Six2-Cre;Bap1^F/+ mice but not in Six2-Cre;Vhl^F/F mice (Fig. 3 G and I). In contrast, CAIX showed membranous staining in a few morphologically unremarkable cortical tubules and cysts in Six2-Cre;Vhl^F/F mice but not in Six2-Cre;Bap1^F/+ mice (Fig. 3 H and J). Overall, these data show that the lesions observed in Six2-Cre;Vhl^F/F;Bap1^F/+ kidneys have the morphological and immunohistochemical characteristics of the corresponding lesions in humans.

Finally, loss of VHL results in the accumulation of HIF-α subunits, leading to activation of the HIF-2 (and possibly HIF-1) transcription factors (25). To determine whether Vhl loss may similarly induce Hif in the mouse, we performed qRT-PCR assays for a panel of Hif-2 and Hif-1 target genes, including VEGF, PAI-1, Glut1, IGFBP3, and PGK1 (26). As shown in Fig. S5, we found that Six2-Cre;Vhl^F/F;Bap1^F/+ kidneys exhibited a uniform increase in all HIF-target genes examined. Further experiments will be required to ascertain the relative role of Hif-1 and Hif-2 in ccRCC tumorigenesis in the mouse.

Discussion

Whereas inactivation of a single VHL allele in the germ line results in ccRCC in humans and the development of numerous premalignant and malignant kidney lesions over a lifetime, Vhl^+/− mice fail to develop renal cancer (27). This observation has remained an enigma since this report by Gnarra et al. (27) was published in 1997. Vhl^−/− mice die during gestation (27), but simultaneous inactivation of both Vhl alleles in kidney epithelial cells similarly failed to induce renal tumorigenesis (28–31). These experiments are confounded by lack of knowledge about the cell type of origin of ccRCC and restricted Cre expression (28–31); however, as we show, even when Vhl is inactivated in multipotent NPCs, Vhl loss is insufficient for renal tumorigenesis.

Recent tumor sampling studies show that intragenic mutations in VHL and loss of chromosome 3p are truncal events in ccRCC development (11, 12). Tumor development begins with a VHL mutation and the loss of the second allele, and along with it, the loss of one allele of other genes on chromosome 3p, including BAP1 (4). We modeled this process in the mouse by generating Six2-Cre;Vhl^F/F;Bap1^F/+ mice, where both alleles of Vhl and one allele of Bap1 were inactivated in NPCs. These mice developed a spectrum of lesions ranging from simple cysts to atypical cysts and neoplastic nodules. Kidneys from Six2-Cre;Vhl^F/F;Bap1^F/+ mice resembled kidneys of patients with VHL syndrome. In patients who have VHL syndrome, lesions evolve from cysts with a single layer of cells, which are considered benign, to cysts lined by two or three layers of cells with focal papillary tufting (atypical cysts), as well as cystic and solid tumor nodules.

Differentiating benign from malignant neoplasms can be a biological and clinicopathological dilemma. This problem is particularly true for small renal tumors, for which traditional morphological features, such as local infiltration and stromal desmoplasia, cannot be used to differentiate a benign tumor from RCC. In humans, it is not infrequent to find minute areas of atypical epithelial proliferation within cystically dilated renal tubules that surround large ccRCC tumors. These lesions are more frequent in kidneys of patients with VHL syndrome and in end-stage kidneys. These small proliferative lesions may be considered clear-cell adenomas; however, the term “clear-cell adenoma” is not currently used because even when small, clear-cell tumors may have acquired the capacity to metastasize (although rarely), and all clear-cell tumors, regardless of size, are presently regarded as carcinomas (32). In contrast, the WHO 2004 classification for papillary renal cell tumors makes a provision for papillary adenomas for tumors with a maximum diameter of 5 mm and without clear-cell morphology or high-grade nuclear features. In this context, adenomas and carcinomas likely represent a continuum in the biological spectrum of the disease.

The neoplastic lesions of Six2-Cre;Vhl^F/F;Bap1^F/+ kidneys are most similar to early-stage human ccRCC. The grade of the neoplastic cells in Six2-Cre;Vhl^F/F;Bap1^F/+ mice resembles Fuhrman nuclear grade 3: a nuclear size more than three- to fourfold of normal renal tubular cell nuclei, an irregular nuclear membrane, and prominent nucleoli. The following features strongly suggest that these murine renal nodules are RCC: The lesions are large for the size of a mouse kidney; focal lymphovascular invasion was observed; there are foci of intratumoral, branching, fibrovascular septa; and there is increased membranous CAIX expression. Conventional features of malignancy, such as anaplasia, pleomorphism, hyperchromasia, nuclear membrane irregularities, loss of polarity, frequent and atypical mitoses, and high Ki-67 staining, are also present. Although the tumors are not morphologically identical to human ccRCC, cytoplasmic clearing was observed in some nodules and both VHL and BAP1 are rarely mutated in RCC of non–clear-cell types. In addition, we observed evidence of mTORC1 activation in lesions in Six2-Cre;Vhl^F/F;Bap1^F/+ kidneys, which we have previously reported in BAP1-deficient human ccRCC (15, 18). Although we suspect that the second Bap1 allele is lost, the lesions were small and no IHC test is available to evaluate mouse Bap1. Overall, these data show that combined inactivation of Vhl and Bap1 is sufficient to induce RCC development. However, these tumors did not get very large or metastasize, suggesting that other events are likely required for further progression.

Interestingly, BAP1 is not the only tumor suppressor gene on chromosome 3p in humans. The short arm of chromosome 3 contains two other kidney tumor suppressor genes, PBRM1 and SETD2. Our previous work shows that 60–70% of ccRCC has mutations in either BAP1 or PBRM1 (15). These numbers may be larger in metastatic tumors. BAP1 and PBRM1 mutations define two subtypes of ccRCC with different gene expression, biology, and outcomes (15, 18, 33, 34). Thus, following VHL inactivation and loss of 3p (11, 12), subsequent inactivation of either the remaining BAP1 or PBRM1 allele may lead to tumors of different grade and aggressiveness (33). To test this hypothesis, experiments are ongoing to evaluate the effects of simultaneous inactivation of Vhl and Pbrm1 in cells of the Six2 lineage. For reasons that are not well understood, simultaneous loss of BAP1 and PBRM1 is negatively selected for in tumors. Tumors tend to lose one or the other tumor suppressor protein, but the loss of both occurs at lower than expected frequencies (15, 17, 18, 33). Experiments are also ongoing to explore this finding.

Overall, our data suggest that VHL is a weak tumor suppressor gene. Vhl inactivation in NPCs in Six2-Cre;Vhl^F/F mice failed to cause malignant or even premalignant lesions. The only lesions we observed were benign cysts lined by a single layer of cells. Although the development of more advanced lesions may be precluded by the early demise of the mice, these mice survived to more than 1 y of age in some instances. Overall, these data are consistent with findings that VHL loss does not induce tumorigenic properties in cells in culture (35). In contrast, in Six2-Cre;Bap1^F/F mice, which died by 1 mo of age, the kidneys already showed premalignant features, including cystically dilated tubules lined by epithelial cells with cytoplasmic clearing, rare epithelial proliferation, and mild nuclear atypia. These data suggest that Bap1 is a stronger tumor suppressor gene than Vhl in the kidney. Furthermore, the development of premalignant lesions in Six2-Cre;Bap1^F/F kidneys is consistent with the finding that germ-line mutation of BAP1 in humans predisposes to familial renal cancer (36, 37). Interestingly, ccRCCs arising in patients with a BAP1 germ-line mutation may not always have mutations in VHL (36). Taken together, these data suggest that BAP1 may be a more potent tumor suppressor gene than VHL in the kidney. However, for reasons that are not understood, VHL mutations occur at a much higher frequency than BAP1 mutations in both familial and sporadic ccRCC.

Our data show that both Vhl and Bap1 are required for kidney function and that inactivation of either in NPCs causes renal failure. However, whereas loss of Bap1 caused renal failure and the death of mice by 1 mo of age, some mice survived more than 1 y after Vhl inactivation. Thus, in keeping with the greater role of Bap1 in tumor suppression, Bap1 seems to play a more significant role in kidney development as well.

Herein, we report the first animal model, to our knowledge, reproducing the hallmark features of ccRCC and the molecular genetics of the human disease. Our results suggest that the differential predisposition to renal cancer between VHL heterozygous humans and Vhl^+/− mice is due to differences in the collinear arrangement of two-hit tumor suppressor genes in the genome. These data serve as a paradigm for how the arrangement of tumor suppressor genes in the genome may influence tumor predisposition across species. This principle may also be applied to the discovery of novel tumor suppressor genes. Germ-line and somatic deletions may be examined for syntenic differences that could explain tumorigenesis in one species but not another.

Materials and Methods

Phenotype Evaluation and Histopathology.

Kidneys were removed and fixed in 10% (vol/vol) neutral-buffered formalin for ∼24 h, and then routinely processed, embedded in paraffin, sectioned at 5 μm, and stained with H&E. Stained sections were evaluated by a board-certified pathologist (P.K.).

BUN and Cr Measurements.

Blood was collected, kept at room temperature (RT) for 30 min before centrifugation, and serum was isolated by spinning samples at 3,600 × g for 10 min. Serum samples were submitted to the mouse metabolic phenotyping core for measurement of BUN and Cr levels by VITROS MicroSlideTM technology (University of Texas Southwestern).

Immunostaining.

For immunofluorescence, isolated kidneys were fixed in 4% (wt/vol) PFA, embedded in optimal cutting temperature (OCT) compound, and sectioned on a cryostat. Frozen sections were washed with PBS, blocked with 5% (wt/vol) BSA/PBS for 1 h at RT, incubated with a primary antibody to anti–β-gal (ab9361; Abcam) at 4 °C overnight, and detected by secondary antibodies with Cy5 (Jackson ImmunoResearch Laboratories). Sections counterstained with LTL were incubated with biotinylated Lotus tetragonolobus agglutinin (FL-1321; Vector Laboratories) at RT for 30 min before washing. For immunohistochemistry, paraffin-embedded tissue was sectioned at 4 μm and staining was performed as described (15). Immunohistochemistry for phospho-S6 ribosomal protein (Ser240/244) and CAIX was conducted using an AutostainerLink 48 (Dako, Agilent Technologies). Briefly, formalin-fixed, paraffin-embedded tissue was cut in 3- to 4-μm sections and air-dried overnight. The sections were deparaffinized, rehydrated, and subjected to heat-induced epitope retrieval using low pH target retrieval solution (for phospho-S6 IHC) and high pH target retrieval solution (for CAIX) (Envision FLEX Target Retrieval Solution, High and Low pH; Dako). Sections were incubated with primary antibodies (phospho-S6 ribosomal protein − phospho-S6 (Ser240/244), 1:100 dilution; Cell Signaling Technology or CAIX, 1:200 dilution; Thermo Fisher Scientific). Before applying the CAIX primary antibody, slides were incubated for 1 h with Mouse Ig Blocking Reagent (Vector Laboratories). For signal detection, the Envision FLEX System (Dako, Agilent Technologies) was used according to the manufacturer’s instructions [with the exception that with pS6, a goat/HRP-labeled polymer was used (1:1,000 dilution; Fisher Scientific)]. Slides were developed using 3,3′-diaminobenzidine chromogen and counterstained with hematoxylin. Appropriate positive and negative controls were used for each run of immunostains.