Folliculin-interacting proteins Fnip1 and Fnip2 play critical roles in kidney tumor suppression in cooperation with Flcn

Significance The role of FLCN as a tumor suppressor in kidney cancer has been well documented, whereas the functional roles of folliculin (FLCN)-interacting proteins 1 and 2 (FNIP1 and FNIP2) in kidney are unknown. In this study, we demonstrate that double inactivation of Fnip1 and Fnip2 leads to enlarged polycystic kidneys or kidney cancer, which mimics the phenotypes seen in Flcn-deficient kidneys and underscores the significance of Fnip1 and Fnip2 in kidney tumor suppression. Moreover, we found that Fnip1/Fnip2 mRNA ratios differ among organs, which may reflect tissue-specific roles for each Fnip. Our findings define Fnip1 and Fnip2 as critical components of the Flcn complex that are essential for its tumor suppressive function and will aid in the development of novel therapeutics for kidney cancer. Folliculin (FLCN)-interacting proteins 1 and 2 (FNIP1, FNIP2) are homologous binding partners of FLCN, a tumor suppressor for kidney cancer. Recent studies have revealed potential functions for Flcn in kidney; however, kidney-specific functions for Fnip1 and Fnip2 are unknown. Here we demonstrate that Fnip1 and Fnip2 play critical roles in kidney tumor suppression in cooperation with Flcn. We observed no detectable phenotype in Fnip2 knockout mice, whereas Fnip1 deficiency produced phenotypes similar to those seen in Flcn-deficient mice in multiple organs, but not in kidneys. We found that absolute Fnip2 mRNA copy number was low relative to Fnip1 in organs that showed phenotypes under Fnip1 deficiency but was comparable to Fnip1 mRNA copy number in mouse kidney. Strikingly, kidney-targeted Fnip1/Fnip2 double inactivation produced enlarged polycystic kidneys, as was previously reported in Flcn-deficient kidneys. Kidney-specific Flcn inactivation did not further augment kidney size or cystic histology of Fnip1/Fnip2 double-deficient kidneys, suggesting pathways dysregulated in Flcn-deficient kidneys and Fnip1/Fnip2 double-deficient kidneys are convergent. Heterozygous Fnip1/homozygous Fnip2 double-knockout mice developed kidney cancer at 24 mo of age, analogous to the heterozygous Flcn knockout mouse model, further supporting the concept that Fnip1 and Fnip2 are essential for the tumor-suppressive function of Flcn and that kidney tumorigenesis in human Birt–Hogg–Dubé syndrome may be triggered by loss of interactions among Flcn, Fnip1, and Fnip2. Our findings uncover important roles for Fnip1 and Fnip2 in kidney tumor suppression and may provide molecular targets for the development of novel therapeutics for kidney cancer.

Folliculin (FLCN)-interacting proteins 1 and 2 (FNIP1, FNIP2) are homologous binding partners of FLCN, a tumor suppressor for kidney cancer. Recent studies have revealed potential functions for Flcn in kidney; however, kidney-specific functions for Fnip1 and Fnip2 are unknown. Here we demonstrate that Fnip1 and Fnip2 play critical roles in kidney tumor suppression in cooperation with Flcn. We observed no detectable phenotype in Fnip2 knockout mice, whereas Fnip1 deficiency produced phenotypes similar to those seen in Flcn-deficient mice in multiple organs, but not in kidneys. We found that absolute Fnip2 mRNA copy number was low relative to Fnip1 in organs that showed phenotypes under Fnip1 deficiency but was comparable to Fnip1 mRNA copy number in mouse kidney. Strikingly, kidney-targeted Fnip1/Fnip2 double inactivation produced enlarged polycystic kidneys, as was previously reported in Flcn-deficient kidneys. Kidney-specific Flcn inactivation did not further augment kidney size or cystic histology of Fnip1/Fnip2 double-deficient kidneys, suggesting pathways dysregulated in Flcn-deficient kidneys and Fnip1/Fnip2 double-deficient kidneys are convergent. Heterozygous Fnip1/homozygous Fnip2 double-knockout mice developed kidney cancer at 24 mo of age, analogous to the heterozygous Flcn knockout mouse model, further supporting the concept that Fnip1 and Fnip2 are essential for the tumor-suppressive function of Flcn and that kidney tumorigenesis in human Birt-Hogg-Dubé syndrome may be triggered by loss of interactions among Flcn, Fnip1, and Fnip2. Our findings uncover important roles for Fnip1 and Fnip2 in kidney tumor suppression and may provide molecular targets for the development of novel therapeutics for kidney cancer.
The first FLCN interacting protein FNIP1 was identified through protein-protein interaction studies of the FLCN protein (11). FNIP1 binds to the C terminus of FLCN and to AMPactivated protein kinase (AMPK) (11), a critical molecule for energy sensing, further underscoring a central role for the FLCN/ FNIP1 pathway in cellular metabolism. A second folliculininteracting protein FNIP2 was discovered through bioinformatics searches for sequences similar to FNIP1 (12,13). Similar to FNIP1, FNIP2 was found to bind to the C terminus of FLCN and to AMPK (12), suggesting a potential functional redundancy with FNIP1. Recent studies with Fnip1 knockout mouse models have demonstrated that Fnip1 is required for B-cell development

Significance
The role of FLCN as a tumor suppressor in kidney cancer has been well documented, whereas the functional roles of folliculin (FLCN)-interacting proteins 1 and 2 (FNIP1 and FNIP2) in kidney are unknown. In this study, we demonstrate that double inactivation of Fnip1 and Fnip2 leads to enlarged polycystic kidneys or kidney cancer, which mimics the phenotypes seen in Flcn-deficient kidneys and underscores the significance of Fnip1 and Fnip2 in kidney tumor suppression. Moreover, we found that Fnip1/Fnip2 mRNA ratios differ among organs, which may reflect tissue-specific roles for each Fnip. Our findings define Fnip1 and Fnip2 as critical components of the Flcn complex that are essential for its tumor suppressive function and will aid in the development of novel therapeutics for kidney cancer. (14,15). Interestingly, a Flcn knockout mouse model using the tamoxifen-inducible ER (mutated form of the ligand-binding domain of the estrogen receptor)-Cre system also displayed defects in B-cell development (14), suggesting Fnip1 knockout mice might develop phenotypes similar to those that develop as a consequence of Flcn deficiency. Furthermore, FLCN has been shown to have a variety of functions that might potentially link AMPK, mTOR, and Ppargc1a with other important pathways. Crystallographic studies have shown that the C terminus of FLCN may be distantly related to Differentially Expressed in Normal Cells and Neoplasia (DENN) domain proteins and may possess guanine nucleotide exchange factor activity toward RAB35 (16). FLCN modulates TFE3 localization (17), which may play an important role in the exit of stem cells from pluripotency (18), and interacts with other signaling pathways including the von Hippel-Lindau-hypoxia inducible factor-vascular endothelial growth factor axis (19)(20)(21), the TGF-beta pathway (22,23), Rho A signaling (24,25), cell cycle regulation (26,27), Rag-mediated amino acid sensing (28,29), and autophagy (30,31). These findings underscore FLCN as an important molecule, inactivation of which affects multiple pathways.
To clarify the function of FLCN-interacting proteins Fnip1 and Fnip2, we inactivated Fnip1 and/or Fnip2 in mouse kidneys, muscle, and heart and investigated the effect on the mTOR pathway and mitochondrial metabolism. The absolute mRNA copy number of Fnip1 and Fnip2 was measured using droplet digital PCR (ddPCR) technology. To evaluate functional synergy of Fnip1 and Fnip2 with Flcn, we also inactivated Flcn, Fnip1, and Fnip2 simultaneously in mouse kidneys. Finally, we searched for latent tumor development in Fnip1 and Fnip2 knockout mice.

Results
Neither Kidney-Targeted Fnip1 nor Fnip2 Knockout Mice Develop a Kidney Phenotype. To investigate Fnip1 function in mouse kidney, we crossbred mice carrying loxP-flanked Fnip1 alleles (floxed, f) (14) with cadherin 16 (CDH16)-Cre transgenic mice, which express Cre recombinase driven by the CDH16 promoter, thereby deleting Fnip1 gene sequences specifically in kidney. We observed no significant phenotype in the Fnip1-deficient kidneys except occasional tiny cysts (Fig. 1A). Therefore, we decided to analyze Fnip2 function in kidney by generating a Fnip2 conditional mouse carrying loxP-flanked Fnip2 alleles (floxed, f) ( Fig. 1 B-D) and crossbreeding with CDH16-Cre transgenic mice. However, kidney-targeted Fnip2 inactivation also did not cause any phenotype in mouse kidney (Fig. 1E). Indeed, we could not find any phenotype in whole-body Fnip2 knockout mice that affected life span.
The Relative Expression Levels of Fnip1 and Fnip2 Differ from Organ to Organ. Previously, we reported that Fnip1 knockout mice showed B-cell developmental defects, which were also observed in Flcn knockout mice using the tamoxifen-inducible ER-Cre system (14), suggesting Fnip1 knockout mice might show phenotypes similar to those resulting from Flcn deficiency. In addition to the B-cell phenotype, we found similar Flcn-deficient and Fnip1-deficient phenotypes in skeletal muscle and heart. Muscletargeted Fnip1 knockout mice showed red-colored muscle with increased mitochondrial biogenesis (myoglobin and cox4 readouts; Fig. 2 A and B), as well as cardiac hypertrophy with elevated mTORC1 activity (Fig. 2 C-F), which we had previously observed in muscle-targeted Flcn knockout mice (6,32). Sequence similarity between FNIP1 and FNIP2 and the shared interaction of FNIP1 and FNIP2 with FLCN and AMPK (12) implied that FNIP1 and FNIP2 might be functionally redundant. In support of this, we observed that expression of either FNIP1 or FNIP2 in Fnip1/Fnip2 null mouse embryonic fibroblasts (MEFs) suppressed Ppargc1a mRNA and ATP production ( Fig. 2 G-I). Because of the potential functional redundancies between Fnip1 and Fnip2, we postulated that Fnip1 and Fnip2 expression might differ from organ to organ and that this variable expression may determine the specific roles for Fnip1 and Fnip2 in those tissues. The recent technology of ddPCR enabled us to measure absolute mRNA copy number. Using this technology, we compared the absolute copy number of Fnip1 and Fnip2 mRNA in wild-type mouse tissues. Interestingly, we observed dominant expression of Fnip1 in heart, skeletal muscle, and bone marrow, the tissues in which we observed phenotypes in Fnip1 knockout mice, whereas there was no significant difference between Fnip1 and Fnip2 mRNA copy number in kidney (Fig. 2J), in agreement with the absence of kidney phenotypes in Fnip1 and Fnip2 knockout mice. These data support the idea that in kidney, the Fnip2 expression level, which is commensurate with that of Fnip1, might maintain Fnip function in Fnip1deficient kidneys, and therefore, double inactivation of Fnip1 and Fnip2 would be necessary to develop a kidney-specific phenotype.

Fnip1/Fnip2 Double-Deficient Kidneys Are Identical to Flcn-Deficient
Kidneys. Previously we observed increased mTOR activity (5) and Ppargc1a-driven mitochondrial biogenesis in Flcn-deficient kidneys (6). In fact, Fnip1/Fnip2 double-deficient kidneys showed increased protein expression of Ppargc1a and signaling molecules in the mTOR pathway, including the downstream target of mTORC1, phospho-Ulk1 at Ser757, that suppresses autophagy, which was confirmed by the accumulation of sequestosome-1 (SQSTM1)/p62 ( Fig. 4 A and B). Increased respiratory capacity (n = 4 each; P < 0.001) (Fig. 4C) and increased mitochondrial surface area (13 cells each; P < 0.001) (Fig. 4D) were also observed in the Fnip1/Fnip2 double-deficient kidneys. These data further support the concept that double inactivation of Fnip1/ Fnip2 in kidney cells mirrors the same phenotype as that associated with kidney-targeted Flcn inactivation. We next asked whether the enlarged polycystic kidney phenotype in kidneytargeted Fnip1/Fnip2 double-knockout mice developed through the same pathway that had produced the identical phenotype in kidney-specific Flcn knockout mice. To answer this question, we crossbred kidney-specific Fnip1/Fnip2 double-knockout mice with kidney-specific Flcn knockout mice to see whether Flcn inactivation might have a synergistic effect on Fnip1/Fnip2 doubly inactivated kidneys. In fact, Flcn inactivation did not enhance the size (n = 8; P = 0.554) (Fig. 4E) or alter the histology (Fig. 4F) of Fnip1/Fnip2 double-deficient kidneys, suggesting that the enlarged polycystic kidney phenotypes of kidney-specific Fnip1/Fnip2 double-knockout mice and kidneyspecific Flcn knockout mice developed through the same pathway, possibly triggered by loss of an important functional interaction among Flcn, Fnip1, and Fnip2.
Fnip1/Fnip2 Double-Knockout Mice Develop Kidney Cancer. Our previous finding that heterozygous Flcn knockout mice developed kidney tumors by 24 mo of age after the loss of the remaining Flcn allele, thereby mimicking human BHD renal tumorigenesis, strongly supports a tumor-suppressive role for FLCN in kidney and underscores FLCN as a classical tumor suppressor (8,9). Because Fnip1/Fnip2 double deficiency in kidney resulted in the identical enlarged polycystic kidney phenotype seen under Flcn deficiency, and knockdown of FNIP1/FNIP2 was permissive for proliferative cell growth in a FLCN-restored human BHD-associated kidney cancer cell line (Fig. 5A), we postulated that double inactivation of Fnip1 and Fnip2 might lead to tumor development in the kidney. Double homozygous inactivation of Fnip1 and Fnip2 in whole body resulted in embryonic lethality (Table 1), further supporting the essential nature of the Fnips and the similarity between Flcn and Fnip1/Fnip2 double-deficient phenotypes. Homozygous Fnip1 knockout mice did not survive long enough to observe latent tumorigenesis (median survival = 292 d) because of severe immunodeficiency resulting from B-cell developmental defects (14). Therefore, we decided to achieve double inactivation of Fnip1 and Fnip2 by heterozygous knockdown of Fnip1 together with homozygous knockdown of Fnip2. Strikingly, these mice developed kidney cancer without developing any particular extrarenal phenotype, whereas neither heterozygous Fnip1 knockout mice nor homozygous Fnip2 knockout mice displayed a kidney tumor phenotype (n = 42; median kidney tumor-free survival = 796 d; P < 0.0001) (Fig. 5B). Kidney cancer in human BHD syndrome presents as multiple tumors with a variety of histologies including hybrid oncocytic tumors, chromophobe renal cell carcinoma, oncocytoma, papillary renal cell carcinoma, and clear cell renal cell carcinoma (4). Indeed, we observed multiple kidney tumor lesions with a variety of histologies in heterozygous Fnip1/homozygous Fnip2 double-knockout mice, most of which were hybrid oncocytic tumors (Fig. 5C). We observed increased protein expression of mTORC1/2 pathway members and Ppargc1a in these tumors by Western blot analysis (Fig. 5D) 500m  The size of the Fnip1-deficient kidney was not significantly different from that of the control kidney. n = 11 each at 3 wk of age. Mean ± SD. Student t test (Middle). Representative H&E staining of 3-wk old control and Fnip1-deficient kidneys did not show differences except for infrequent tiny cysts (Right). (B) Fnip2 gene-targeting vector was constructed by recombineering methodology using homologous recombination. A neomycin resistance (Neo r) cassette flanked by Frt (bar) and loxP (triangle) sequences was inserted into intron 11 for positive selection, and the thymidine kinase gene was included for negative selection. A second loxP sequence was inserted into intron 13. Correctly targeted embryonic stem cells were identified by Southern blot analysis and injected into blastocysts to produce chimeras. Backcrossing to C57BL/6 mice produced heterozygous F1 offspring with germline transmission of the Fnip2 floxed (f)-Neo allele. The Neo cassette flanked by Frt sites was excised in vivo by crossing with mice expressing the Flp recombinase transgene under the β-actin promoter. To produce the Fnip2 deleted (d) allele, Fnip2 f/+ mice were crossed with mice expressing the Cre recombinase transgene under the ubiquitous β-actin promoter. Deletion of exon 12 and 13 resulted in a frameshift and premature termination codon in exon 14, which was predicted to cause mRNA degradation by the nonsense-mediated decay mRNA surveillance system. (C) The targeted embryonic stem cells were screened by Southern blotting of BamH1and EcoRV-digested DNA, using two different external probes located outside the targeting sequence, as shown in B. (D) PCR-based genotyping was performed using DNA extracted from mouse tails for routine monitoring of inheritance in offspring. Locations of PCR primers are indicated by arrows. (E) Kidney-specific inactivation of Fnip2 was achieved by crossing with CDH16-Cre transgenic mice. Inactivation of Fnip2 mRNA was confirmed by real-time PCR. n = 6 each at 3 wk of age. Mean ± SD. Student t test (Left). Size of Fnip2-deficient kidney was not significantly different from that of control kidney. n = 6 each at 3 wk of age. Mean ± SD. Student t test (Middle). Representative H&E staining of 3-wk old control and Fnip2-deficient kidneys did not show any difference in histology (Right). and immunostaining (Fig. 5E), further supporting the idea that these tumors might develop through the same pathway as tumors that develop in the heterozygous Flcn knockout mouse model.

Discussion
Here we report kidney tumor-suppressive roles for Fnip1 and Fnip2 in cooperation with Flcn. Fnip1/Fnip2 double inactivation targeted to mouse kidney resulted in an enlarged multicystic kidney phenotype shortly after birth, which was identical to the phenotype observed in Flcn-deficient kidneys (5). The ratio of absolute Fnip1 to Fnip2 mRNA copy number was high in the organs that demonstrated a phenotype after Fnip1 inactivation, whereas in kidney where no Fnip1-deficient or Fnip2-deficient phenotype was observed, the absolute Fnip2 mRNA copy number was comparable to that of Fnip1. Therefore, these data suggest that the ratio of absolute Fnip1 to Fnip2 mRNA copy number may determine the function of each Fnip in a particular organ. Moreover, whole-body heterozygous Fnip1/homozygous Fnip2 double-knockout mice developed kidney tumors at 24 mo of age, implying that loss of interaction between Flcn, Fnip1, and Fnip2 may trigger kidney cancer development.
According to protein sequence, FNIP1 and FNIP2 show 49% identity and 74% similarity (12), suggesting possible functional redundancy. Double homozygous inactivation of Fnip1 and Fnip2 specifically in mouse kidney resulted in an enlarged polycystic kidney phenotype. However, expression of one allele of either Fnip1 or Fnip2 in kidney-targeted Fnip1/Fnip2 knockout mice was sufficient to rescue this phenotype, suggesting Fnip1 and Fnip2 may have functional redundancy. Because worms and flies have only one Fnip protein (12), it is possible that a second Fnip with overlapping functions evolved from the primary Fnip to ensure redundancy and conserve its critical role in regulating kidney cell proliferation through its interaction with Flcn.
In contrast, the differences between FNIP1 and FNIP2 amino acid sequences imply potentially distinct functions for each FNIP. Therefore, the nonoverlapped functions of the FNIPs, as well as additional overlapping functions, will need to be elucidated in future experiments.
Crystallographic studies revealed that the C terminus of FLCN shows distant homology to DENN domain proteins, a family of GDP-GTP exchange factors that activate Rab GTPases involved in membrane trafficking in eukaryotes (16). Notably, a subsequent bioinformatics study reported that FNIP1 and FNIP2 also have novel DENN modules (33), raising the possibility that complex assembly of FLCN-FNIP1 or FLCN-FNIP2 might form a noncanonical DENN module critical for GDP-GTP exchange that would suffer functional arrest if any of the components were absent. Disruption of a noncanonical DENN module, whose conformation may be controlled by Flcn/Fnip1 or Flcn/ Fnip2 interactions and may be essential for regulation of proper kidney cell proliferation rates, might result in development of the polycystic kidney phenotype in our kidney-targeted in vivo models. (J) ddPCR showed Fnip1 expression was significantly higher than Fnip2 expression in bone marrow, heart, and quadriceps of C57BL/6 mice but was not significantly higher than Fnip2 expression in kidney of C57BL/6 mice. n = 6 each at 6 wk of age. Mean ± SD. P < 0.001 for bone marrow, heart, quadriceps. N.S., not significant. Student t test.
After FLCN was identified as a two-hit tumor suppressor gene for BHD-associated chromophobe, hybrid oncocytic, and clear cell kidney cancers, a search for FLCN mutations in a broad spectrum of sporadic kidney tumors was conducted, but genetic analyses of these samples only rarely identified mutations in the FLCN gene (34,35). Somatic FNIP1 or FNIP2 mutations have been detected in several cancers by whole-exome sequencing as part of The Cancer Genome Atlas (TCGA) project. Sequencing efforts by TCGA project detected infrequent FNIP1/FNIP2 mutations in urologic cancers, including clear cell renal carcinomas (7/424; 1.7%) and bladder cancer (7/130; 5.4%), and in all but one of the urologic cancer samples, the tumors had a single mutation of either FNIP1 or FNIP2 (36). In comparison, uterine corpus endometrioid cancer displayed the highest percentage of tumors with FNIP1 and FNIP2 alterations (17/240; 7.1%), of which 5 (2.1%) had mutations in both genes (37). Infrequent somatic mutations of FLCN, FNIP1, and FNIP2 genes in sporadic kidney tumors indicate that genetic alteration of these genes is not the direct cause of sporadic kidney tumorigenesis, but rather, that the status of the FLCN/FNIP1 or FLCN/FNIP2 interactions might be critical for sporadic kidney tumor suppression. Proteinprotein interaction studies of sporadic kidney cancers would be important to elucidate the functional status of the FLCN/FNIP1/ FNIP2 complex in sporadic kidney cancer, especially in sporadic chromophobe renal cell carcinoma and oncocytoma, which are the most frequent histologic subtypes observed in human BHD syndrome (Fig. S1).
The findings of this study, which characterize the phenotype of Fnip1, Fnip2, and Fnip1/Fnip2 knockout mouse models in multiple organs, further our understanding of the Flcn tumor suppressor pathway and underscore important roles for Flcn/Fnip1/Fnip2 interactions in inhibiting kidney cancer development. These data may lead to the development of novel diagnostics and therapeutics for kidney cancer that target the FLCN-FNIP pathway.

Materials and Methods
Animals. Mice carrying Flcn alleles and Fnip1 alleles flanked by loxP sites (floxed, f) and mice carrying a Fnip1 deleted (d) allele were generated as previously described (5,14). Mice carrying Fnip2 alleles flanked by loxP sites (floxed, f) and mice carrying a Fnip2 deleted (d) allele were generated using the same strategy (5,14). Briefly, the Fnip2 targeting vector was generated by inserting a neomycin resistance cassette flanked by Frt and loxP sequences into intron 11 of Fnip2 and inserting a second loxP sequence into intron 13. Deletion of exon 12 and exon 13 in the Fnip2 d/+ mice resulted in a reading frameshift and premature termination codon in exon 14. CDH16-Cre transgenic mice, which express Cre recombinase under the cadherin 16 (CDH16) promoter specifically in adult renal tubules and developing genitourinary tract (38), were crossed with mice carrying floxed (f) alleles of Flcn, Fnip1, and Fnip2 to inactivate those genes. Because we did not observe any phenotypic difference between Proportion surviving procedures followed the National Cancer Institute Animal Care and Use Committee guidelines.
MRI Imaging. T 2 -weighted images were obtained using a fast spin echo sequence (rapid acquisition of relaxation enhancement) with an echo time of 13 ms and a repetition time of 2,500 ms by a 7 T MRI scanner controlled with ParaVision 5.0 (Bruker BioSpin MRI GmbH).