Loss of FLCN-FNIP1/2 induces a non-canonical interferon response in human renal tubular epithelial cells

Germline mutations in the Folliculin (FLCN) tumor suppressor gene cause Birt–Hogg–Dubé (BHD) syndrome, a rare autosomal dominant disorder predisposing carriers to kidney tumors. FLCN is a conserved, essential gene linked to diverse cellular processes but the mechanism by which FLCN prevents kidney cancer remains unknown. Here, we show that disrupting FLCN in human renal tubular epithelial cells (RPTEC/TERT1) activates TFE3, upregulating expression of its E-box targets, including RRAGD and GPNMB, without modifying mTORC1 activity. Surprisingly, the absence of FLCN or its binding partners FNIP1/FNIP2 induces interferon response genes independently of interferon. Mechanistically, FLCN loss promotes STAT2 recruitment to chromatin and slows cellular proliferation. Our integrated analysis identifies STAT1/2 signaling as a novel target of FLCN in renal cells and BHD tumors. STAT1/2 activation appears to counterbalance TFE3-directed hyper-proliferation and may influence immune responses. These findings shed light on unique roles of FLCN in human renal tumorigenesis and pinpoint candidate prognostic biomarkers.


Introduction
Renal cell carcinoma (RCC) is the most common form of kidney cancer representing up to 5% of newly identified cancer cases (Ferlay et al., 2019;Lopez-Beltran et al., 2006;Siegel et al., 2018). Generally, RCCs are diagnosed in adults, with the exception of translocation RCC, which is driven by a hyper-activated fusion protein of the transcriptional activators TFE3 or TFEB and comprises 20-75% of RCCs in childhood (Ambalavanan and Geller, 2019;Caliò et al., 2019;Lee et al., 2018). Birt-Hogg-Dubé syndrome (BHD) is a dominantly inherited kidney cancer syndrome caused by mono-allelic germline loss-of-function mutations of the essential and conserved Folliculin (FLCN) gene (Nahorski et al., 2011;Nickerson et al., 2002). The lifetime risk for BHD patients to develop RCC is~10 times higher than for the unaffected population (Houweling et al., 2011;Pavlovich et al., 2005;Toro et al., 2008;Zbar et al., 2002). BHD patients are predisposed to bilateral and multifocal renal tumors (Nickerson et al., 2002;Schmidt et al., 2005;Zbar et al., 2002) and dependent on surveillance by renal imaging for early detection and curative treatment prior to metastasis (Johannesma et al., 2019). Loss of heterozygosity, by gene silencing or an inactivating somatic mutation of the wild-type FLCN allele, is a prerequisite for kidney cancer development in BHD patients (Vocke et al., 2005).
Here, we present the molecular and cellular consequences of knocking out FLCN or its binding partners FNIP1/FNIP2 in a human renal proximal tubular epithelial cell model, representing the cells of origin of RCC (Holthö fer et al., 1983). We performed RNA sequencing (RNAseq) and proteomics, followed by pathway analyses and mining of regulatory promotor motifs of differentially expressed genes, revealing that FLCN loss induces two separate transcriptional signatures. The first is characterized by E-box controlled genes and confirms TFE3 as a main target of the FLCN-FNIP1/2 axis (El-Houjeiri et al., 2019;Endoh et al., 2020;Hong et al., 2010b;Petit et al., 2013). Secondly, we discovered that loss of FLCN-FNIP1/2 induces a set of genes under control of interferon-stimulated response elements (ISREs). The ISRE gene activation program is directed by upregulate STAT1 and STAT2 and may explain why loss of the FLCN tumor suppressor, paradoxically, reduces cellular proliferation. We propose that TFE3 and STAT1/2 are the two main, independent transcriptional effectors of FLCN-FNIP1/2 loss in human renal epithelial cells. Preliminary data indicate that these gene networks may also be activated in BHD tumors. Taken together, our findings may help the development of prognostic biomarkers or targeted therapies.

Knocking out FLCN activates TFE3 in renal proximal tubular cells
To study the effects of FLCN loss in a context relevant for oncogenesis, we used an immortalized, diploid renal proximal tubular epithelial cell line (RPTEC/TERT1, ATCC CRL-4031; Wieser et al., 2008, hereafter called RPTEC) as a model system, which retains the capacity to form 3D tubular structures ( Figure 1A). First, we constructed an inducible Cas9-expressing RPTEC cell line and verified doxycycline-induced Cas9 protein expression by immunoblots ( Figure 1B). Because a TP53dependent DNA damage response prohibits effective gene editing in some cell types (Haapaniemi et al., 2018;Ihry et al., 2018), we simultaneously knocked-out TP53 and FLCN to improve targeting efficiency. Guide RNAs (gRNAs) targeting 5' coding exons of TP53 and FLCN ( Figure 1C) were co-transfected and targeted exons of single-cell derived clones were sequenced. Indel analysis showed that each clone carried a unique genetic disruption of FLCN and TP53 (Figure 1-figure supplement 1A). By karyotypic analysis, we found that four out of six knock out (KO) cell lines had become aneuploid ( Figure 1D). We conclude, however, that aneuploidy occurred spontaneously during cell line formation because it appeared independently of FLCN, TP53, or inducible Cas9 status ( Figure 1D). At a later stage, we indeed developed diploid FLCN knock out RPTEC cell lines using more advanced gRNA delivery assays (see below). However, considering that kidney tumors are often aneuploid (Kardas et al., 2005;Morlote et al., 2019), we first used the cell   Figure 1D to identify FLCN-specific effects that occur independently of karyotype. We started by comparing two groups of cell lines: RPTEC ('WT'), RPTEC tet-on Cas9 ('Cas9'), and RPTEC tet-on Cas9 TP53 -/-('TP53 KO ') were assigned to the FLCN POS group, while three individually isolated RPTEC tet-on Cas9 TP53 -/-FLCN -/clones ('FLCN KO C1','FLCN KO C2', and 'FLCN KO C3') were assigned to the FLCN NEG group. C1 and C2 were created by gRNAs targeting FLCN exon 5 ('gFLCN_5') and C3 by gRNAs targeting FLCN exon 7 ('gFLCN_7'). Loss of FLCN and TP53 protein expression was confirmed by immunoblots, which also showed that expression of the renal proximal tubular marker aquaporin-1 (AQP1) was unchanged (Bedford et al., 2003; Figure 1E and Previous studies reported that FLCN prevents nuclear localization of TFE3 under nutrient-rich conditions. TFE3, and its close family member TFEB, are transcription factors directing an autophagy and stress tolerance gene program under growth restrictive conditions (Hong et al., 2010a;Wada et al., 2016). To investigate the status of this pathway in our FLCN knock-out RPTEC cell models, we visualized TFE3 localization in the presence or absence of FLCN. Immunofluorescence co-staining of TFE3 and the lysosomal marker LAMP2 in either fed, starved, or amino acid (AA)depleted conditions are shown in Figure 1F and Figure 1-figure supplement 1C. RPTEC cells showed cytoplasmic TFE3 under normal growth conditions but TFE3 completely translocated to the nucleus after withdrawal of serum and amino acids ( Figure 1F, upper panels). In contrast, TFE3 localized in the nucleus of the majority of cells from each of the three RPTEC FLCN NEG cell lines, independent of starvation or refeeding, revealing that FLCN NEG cells fail to convey proper nutrient sensing and TFE3 responses ( Figure 1F, lower panels and Figure 1-figure supplement 1C). We conclude that, in the absence of FLCN, TFE3 is constitutively active in renal epithelial cells, confirming TFE3 as a bona fide target of the FLCN tumor suppressor (El-Houjeiri et al., 2019;Wada et al., 2016).

Overlapping transcriptomic and proteomic alterations induced by FLCN loss
Subsequently, we determined changes in gene transcription and protein expression patterns in our FLCN POS and FLCN NEG RPTEC cell lines by mRNA sequencing (RNAseq) and proteomic workflows shown in Figure 2A. For specificity, we only compared profiles of the FLCN POS and FLCN NEG cell lines between the groups defined by rectangles in Figure 2 To correct for possible clonal effects on global transcription, three different TP53 knock-out clones were included in the RNAseq analysis. We used edgeR (Robinson et al., 2010) to identify FLCN-related effects and visualized differential expressed genes in volcano plots ( Figure 2B). Green circles show genes upregulated in FLCN NEG cells and pink circles indicate genes expressed at a higher level in control (FLCN POS ) cells. The threshold line represents a false discovery rate (FDR) of <0.05. The edgeR results of RPTEC FLCN POS versus FLCN NEG comparison are in Supplementary file 1 and a volcano plot with annotated gene names is shown in Figure 3-figure supplement 1A. Interestingly, the majority of significantly altered genes was induced, rather than repressed, by FLCN inactivation: in FLCN NEG cells 426 genes were upregulated at least tenfold while only 62 genes were downregulated to the same extent (FDR < 0.05,  FLCN NEG cells using normalized spectral counts (beta-binomial test, ), identified the differential expression pattern presented in Figure 2C (threshold line indicates p-values of <0.05). The U(-), and U(+), columns next to the volcano plots show proteins that were detected in only one of the conditions. Volcano plots annotated with corresponding protein names in are shown in Figure 3-figure supplement 1A and the complete comparative analysis can be found in Supplementary file 1.
In concordance with our RNAseq data, FLCN loss predominantly induced protein expression: 209 proteins were expressed at least fivefold higher in FLCN NEG cells, versus 41 proteins that were similarly downregulated, out of a total of 914 differentially expressed proteins (p<0.05 and average count >1.5; Figure 2-figure supplement 2E, lower right). Conversely, as a control, deletion of the transcriptional activator TP53 alone (Farmer et al., 1992;Fields and Jang, 1990;Vogelstein and Kinzler, 1992)

FLCN loss activates a specific set of TFE3 target genes in kidney epithelial cells
To determine which TFE3/TFEB (TFE) driven genes are activated by FLCN loss, we collected 248 previously reported TFE targets (Hong et al., 2010a;Martina et al., 2014;Palmieri et al., 2011;Santaguida et al., 2015) and performed k-means Pearson correlation clustering of our FLCN POS vs. FLCN NEG RPTEC RNAseq data to categorize the effects. Interestingly, we found a specific subset of 115 known TFE targets to be upregulated in all three FLCN NEG cell lines ( Figure 3A, cluster 3, boxed yellow, Supplementary file 2). Consistent upregulation of TFE target genes FNIP2, GPNMB, SQSTM1, RRAGC, GABARAP, ARHGAP12, AMDHD, WIPI1 and the more recently identified tumor growth-promoting TFE target RRAGD (not included in Figure 3A; Di Malta et al., 2017) was validated in all three FLCN NEG RPTEC clones using quantitative RT-PCR ( Figure 3B). Western blots of GPNMB, RRAGD, SQSTM1, and FNIP2, proteins associated with lysosome function, also showed these were strongly upregulated by FLCN inactivation, irrespective of TP53 status ( Figure 3C; TP53 wt FLCN KO refers to a TP53 positive FLCN knock out clone, see Figure 3-figure supplement 1B). Induction of GPNMB, RRAGD, SQSTM1, FNIP2 and other E-box targets was significantly ameliorated in all three FLCN NEG RPTEC clones by simultaneously knocking down TFE3 and its close family member TFEB ( Figure 3D; siTFE3 only is shown in Figure 3-figure supplement 1C), an effect further confirmed in a TP53 wt FLCN KO cell line shown in Figure 8-figure supplement 1C. We conclude that FLCN loss activates a specific category of TFE target genes in renal epithelial cells, independent of TP53 or karyotype, many of which function in autophagy and lysosome regulation.
Since the nutrient-sensing capabilities of mTOR had been associated with both FLCN and TFE3 activity in previous studies (Baba et al., 2006;Bastola et al., 2013;El-Houjeiri et al., 2019;Hasumi et al., 2009;Hong et al., 2010b;Hudon et al., 2010;Petit et al., 2013;Takagi et al., 2008), we next assessed FLCN-dependent changes in canonical mTOR signaling. However, phosphorylation levels of two direct mTORC1 targets, 4E-BP1 (as judged by its electrophoretic mobility shift) and S6 kinase (S6K_T389) were not changed by FLCN loss in RPTECs, while also AKT/ PKB_S437 phosphorylation remained intact ( Figure 4A). The dynamic subcellular localization of   Figure 3 continued on next page mTOR in response to starvation did not differ between FLCN POS and FLCN NEG as detected by immunofluorescent co-staining of mTOR and lysosomal marker LAMP2 ( Figure 4B, FLCN NEG C3 is representative of the three FLCN NEG clones which all show normal mTOR dynamics). So, in contrast to several previous studies, but in line with a recent report (El-Houjeiri et al., 2019), we found no evidence for directly altered mTOR signaling or nutrient sensing in the absence of FLCN in renal tubular cells.

An interferon (IFN) gene signature is induced in the absence of FLCN
To further identify the main biological processes influenced by FLCN expression, we performed Molecular Signatures Database (MSigDB) gene set enrichment analyses (GSEA) (Subramanian et al., 2005) on both RNA and protein data sets (pre-ranked list of p-values, classic ES). Figure 5A displays hallmark gene sets ranked by normalized enrichment score (NES) and significance (FDR is represented by the size of the dot). Gene sets significantly enriched in either RNA or protein (FDR < 0.05) are shown, with biological processes enriched in FLCN NEG indicated in green, and processes enriched in FLCN POS in pink. An overview of less significant (FDR > 0.05) hallmark gene sets is shown in Figure 6-figure supplement 1A. We found a higher representation of cell cycle related processes in FLCN POS cells, in both RNA and protein data ( Figure (Maere et al., 2005) gene ontology analyses of differential expression patterns in FLCN NEG RPTECs also revealed many overlapping immune and IFN response signature genes, including ISG15, IFIT1, IF16, MX1, OAS2, and STAT2 in both RNA and protein data ( Figure 5C,D; orange circles show overlap). These findings indicate that the IFN response signature program is a key target of FLCN in renal epithelial cells.

Two distinct transcriptional programs are strongly induced by FLCN loss
To specify how transcription is changed in FLCN NEG RPTEC, we used iRegulon (Janky et al., 2014), which prioritizes candidate regulatory transcription factors based on enriched promotor motifs upstream of the transcription start sites (TSS) ( Figure 6A). The iRegulon analysis of distinct promotor motifs directing genes upregulated in FLCN NEG (n = 711, FDR < 0.05 and logFC > 2) are shown in Figure 6B. A similar overview is shown in Figure 6C, based on significantly upregulated proteins in FLCN NEG RPTEC (n = 498, p<0.05 and FC > 2).
Importantly, the majority of regulatory elements enriched in FLCN NEG RPTEC can be assigned to either the basic helix-loop-helix E-box motif group (e.g. regulated by TFE3/TFEB) or to the Interferon-Stimulated Response Element (ISRE) motif group ( Figure 6D). In FLCN POS RPTECs, iRegulon replicates. To determine quantitative gene expression levels, data were normalized to the geometric mean of two housekeeping genes. See Figure 3source data 1 for raw qRT-PCR values and fold change calculations. (C) Western blots of RPTEC/TERT1 tet-on Cas9 cell lines. All FLCN NEG clones show strong induction of protein expression of TFE targets GPNMB, RRAGD, SQSTM1, and FNIP2. GAPDH and Actin were used as loading controls. Western blots were performed three times. (D) Knock down of TFE3/TFEB (10 nM siRNA, 72 hr) ameliorates the TFE expression gene signature induced by FLCN loss in three FLCN NEG clones. Expression levels were determined by qRT-PCR, normalized to siNT-treated clones and are representative of three independent experiments. To determine quantitative gene expression data levels were normalized to the geometric mean of two housekeeping genes. Also see Figure     Upregulation of E-box or ISRE motif-dependent genes induced by FLCN loss in our RNAseq and proteomics analyses are shown in Figure 6E-H. Significant induction of the ISRE targets MX1, ISG15, IRF9, IFIT1, STAT1, and STAT2 next to TFE3 and E-box targets GPNMB, RRAGD, FNIP2, CTSD, and SQSTM1 were validated by qPCR in Figure 6-figure supplement 1C.
We then analyzed protein expression levels compared to normal human kidney lysates using mass spectrometry of two independent BHD kidney tumors ( Figure 6-figure supplement 1D and E).
Although full interpretation of these results awaits the analysis of a larger tumor sample size, we observed elevated expression of both the ISRE as E-box associated genes in FLCN NEG BHD tumors ( Figure 6I). Based on these results, we conclude that FLCN loss upregulates two major gene classes: TFE3/TFEB regulated E-box targets and IFN-associated ISRE targets. Our integrated analysis reveals these genes and their protein products as candidate positive biomarkers for FLCN loss in BHDrelated kidney cancer.

FLCN-FNIP1/2 loss upregulates STAT2 in a cell-type-specific manner
FLCN acts in a protein complex with FNIP1 and FNIP2 (Baba et al., 2006;Hasumi et al., 2015;Hasumi et al., 2008). When FNIP1 and FNIP2 are inactivated simultaneously, mice develop kidney cancer (Hasumi et al., 2015). To investigate whether deletion of both FNIP1 and FNIP2 in RPTEC had a molecular effect similar to that of FLCN loss, we created a FNIP1/FNIP2 NEG RPTEC cell line ( Having validated a new gene program targeted by the FLCN-FNIP1/2 complex in RPTECs, we next determined whether FLCN-dependent control of ISRE target genes occurred in cells of different tissue origin, too. Using a similar approach as in RPTEC, we created a FLCN NEG retinal pigment epithelial cell line (RPE1/TERT tet-on Cas9 TP53 KO Benedict et al., 2020;Figure 7-figure supplement 1C). Quantitative RT-PCR analyses revealed that, strikingly, ISRE and E-box associated genes were not induced in RPE1 cells by FLCN loss ( Figure 7B). RRAGD was the only exception yet was induced~10 times less strongly in RPE1 as compared to RPTEC cells. These results indicate that the two main gene induction programs directed by FLCN-FNIP1/2 are renal specific.
To identify the dominant signature of regulatory elements responding to FLCN loss in RPTEC, we further looked into the shared hits of RNA and protein data sets. Upon selecting the most significantly overlapping effects (FDR < 0.01, n = 181) ( Figure 7C), iRegulon analysis of differential RNAs overlapping with differential proteins (r > 0.8, n = 49) upon FLCN loss predicted STAT1 and STAT2binding motifs as the most enriched upstream regulatory elements ( Figure 7D).
To prove that these upregulated gene programs were truly FLCN dependent, we complemented FLCN NEG RPTEC C2 by re-introducing FLCN (Figure 7-figure supplement 1D). FLCN re-expression in FLCN NEG RPTEC C2 restored regulation of TFE3 localization and nutrient sensing (Figure 7figure supplement 1E). Moreover, re-introducing FLCN completely reverted the ISRE expression phenotype ( Figure 7E). To confirm the roles of STAT1/2, we knocked them down using siRNAs in FLCN NEG RPTEC C2. Quantitative RT-PCR showed that upregulation of the ISRE-associated gene program was entirely STAT1/2-dependent ( Figure 7E). In addition, immunoblotting subcellular fractions showed higher protein levels of both STAT1 and STAT2 in FLCN NEG cells ( Figure 7F). processes that were significant in either RNA and/or protein data are depicted in this Figure. An extended version with all identified gene sets is shown in Figure 6-figure supplement 1A. (B) FLCN NEG RPTECs grow significantly slower (p=8.31E-11) when compared to FLCN POS RPTEC. Cell lines were seeded in equal densities and total cell number was counted for 7 consecutive days. Results shown are representative for two independent experiments. (C) Gene Ontology (biological processes, BinGO) analysis of mRNAs higher expressed in FLCN NEG RPTECs reveals highly overlapping (orange circles) clusters of immune-and interferon-response-related genes between both data sets. Shade of green nodes represents fold change. (D) Gene Ontology (biological processes, BinGO) analysis of proteins higher expressed in FLCN NEG RPTECs reveals highly overlapping (orange circles) clusters of immune and interferon response related genes between both data sets. Shades of green nodes represent different levels of fold change.
Black nodes indicate uniquely detected proteins in FLCN NEG RPTEC.
ns ns Figure 6. Identification of regulatory elements activated by FLCN loss in RPTEC and BHD tumors. (A) Identification of transcriptional regulatory elements associated with loss of FLCN expression. Regulons were identified by iRegulon (Janky et al., 2014), using an input a list of differential expressed genes ( Figure 3). (B) Upstream regulons enriched in FLCN NEG RPTEC based on significantly upregulated genes derived from our transcriptomic data set (n = 711, FDR < 0.05 and logFC > 2). Transcription factors with normalized enrichment scores (NES) higher than 3.5 are shown, Figure 6 continued on next page Importantly, nuclear STAT2 was exclusively bound to chromatin in the absence of FLCN ( Figure 7G) identifying STAT2 as a key target of FLCN. Canonical IFN signaling follows IFNa or IFNg stimulation of IFN receptors, resulting in auto-phosphorylation of Janus Kinases 1/2 (JAK1/JAK2) and Tyrosine Kinase 2 (TYK2). As a consequence, STAT1 and STAT2 are phosphorylated enabling formation of (homo)dimers or the ISGF3 complex (composed of STAT1, STAT2, and IRF9) which translocate to the nucleus to initiate transcription of ISRE genes (Majoros et al., 2017). To understand how STAT2 is activated upon FLCN loss, we measured IFN levels in supernatant of FLCN NEG RPTEC cell lines, using a flow-cytometry-based cytometric bead array (CBA) or an enzyme-linked immunosorbent assay (ELISA). However, we did not detect any secreted IFNg or IFNa, indicating that autocrine stimulation of IFN signaling did not cause the upregulated ISRE gene program or chromatin binding of STAT2 in FLCN NEG RPTEC (Figure 7-figure supplement 1F). Also, we did not detect IFNA, IFNB and IFNG expression, nor differential expression of the IFN receptors (IFNAR1, IFNAR2, IFNGR1 and IFNGR2) in FLCN NEG RPTEC (Figure 7-figure supplement 1G). Finally, in a recent phospho-proteomic analysis of our FLCN NEG cell lines (not described in this paper), loss of FLCN did not lead to phosphorylation of upstream kinases JAK2 and TYK, or STAT2 itself. Lack of induced STAT1 phosphorylation as detected by immunoblot further confirmed that FLCN NEG cells do not activate the IFN receptor (Figure 7-figure supplement 1H). Together, these results show that FLCN loss in RPTEC leads to a non-canonical, IFN-independent activation of unphosphorylated STAT1 and STAT2.

FLCN loss counteracts TFE3-induced hyperproliferation
As the experiments described above confirm our hypothesis that both TFE3 and STAT2 are FLCN targets, we started to investigate their potential contribution to renal cell transformation in vitro. We aimed to compare the effect of TFE3 activation resulting from FLCN loss to that of an active TFE3 gene fusion that is constitutively nuclear. Xp11 translocation RCC is a rare subtype of kidney cancer associated with various TFE3, TFEB, or MITF gene fusions, resulting in oncogenic, nuclear forms of these transcription factors (Caliò et al., 2019). This type of RCC behaves remarkably aggressively, with poor progression-free survival rates (Lee et al., 2018). One common fusion partner for TFE3 is the DNA-binding splicing factor SFPQ (PSF), which we expressed in RPTEC. For comparison, we created a new, diploid FLCN knock-out cell line by a recently improved gRNA and Cas9 delivery protocol (Figure 8-figure supplement 1A and B). In both SFPQ-TFE3 and FLCN KO RPTEC cell lines, we confirmed high expression of TFE targets GPNMB, RRAGD, FNIP2, and WIPI1 ( Figure 8A). TFE3 or Figure 6 continued together with detected number of targets, motifs, and elements. ISREs are highlighted in orange. Upper part shows motifs enriched 500 bp upstream from transcription start site (TSS), lower part shows motifs enriched 20 kb around TSS. (C) Upstream regulons enriched in FLCN NEG RPTEC based on significantly upregulated proteins derived from our proteomic data set (n = 498, p<0.05 and FC > 2). Transcription factors with normalized enrichment score (NES) higher than 3.5 are shown, together with number of targets, motifs, and elements detected. ISREs are highlighted in orange and E-boxes in purple. STAT1 appears twice due to the fact that iRegulon ranks this transcription factor to be the most likely upstream regulator for two sets of targets genes, containing slightly different ISRE-motifs 20 kb upstream from the TSS. (D) Two major enriched motif elements detected in iRegulon analysis of genes upregulated in FLCN NEG RPTEC. Regulons can be assigned to E-box (in purple) or ISRE (in orange) motif group. (E) Bar graphs of RNA expression levels of genes associated with an E-box motif, derived from RPTEC transcriptomic data set.    Importantly, SFPQ-TFE3 expression did not induce IFN response genes in RPTECs, showing that the IFN gene induction signature is a specific effect of FLCN loss ( Figure 8A). Concordantly, STAT2 recruitment to chromatin was enhanced in the absence of FLCN but not detected after expression of SFPQ-TFE3 ( Figure 8B).

S T A T 1 S T A T 1 S T A T 2 S T A T
In agreement with the known growth inhibitory effects of the IFN stimulated gene program, but unexpected considering that FLCN is a tumor suppressor gene, FLCN loss reduced RPTEC colony formation and slowed cellular proliferation, regardless of TP53 status ( Figure 8C upper panels, Figure 5B). Indeed, re-introducing FLCN expression in FLCN KO RPTEC rescued cellular proliferation ( Figure 8D). Reversely, the absence of FLCN dominantly repressed the hyper-proliferative effects of an active, oncogenic TFE3 fusion protein in RPTEC, similar to the growth reduction observed after treating SFPQ-TFE3 RPTECs with 100 IU/ml IFNg ( Figure 8C lower panels, Figure 8E). The growth inhibitory effect of knocking out FLCN in SFPQ-TFE3 RPTECs also correlated with strong induction of ISRE genes (Figure 8-figure supplement 1D). In conclusion, these results show that FLCN loss induces a STAT2-mediated IFN signature that results in growth inhibition, which counteracts the hyperproliferative effects of constitutive activation of TFE3 (Figure 8-figure supplement 1E). This indicates that, next to its growth stimulatory effects on cell proliferation via TFE3 activation, loss of FLCN also has a growth suppressive effect executed by STAT2 activation in renal tubular cells. STAT2 activation may also contribute to a pro-oncogenic state by contributing to an inflammatory response (Figure 8-figure supplement 1E). Determining the role of STAT2 activity in tumorigenesis will require further analysis of bio-markers and immune-infiltrates in a large set of BHD tumors.

Discussion
FLCN, together with its binding partners FNIP1 and FNIP2, forms a regulatory protein complex found in species ranging from yeast to humans (Nookala et al., 2012;Pacitto et al., 2015;Zhang et al., 2012). Pleiotropic effects resulting from its mutation in different model systems have hampered a clear understanding of the biological function of the FLCN-FNIP1/2 axis. At the organismal level, FLCN also plays different roles in different tissues. This is already apparent from the clinical

(B) qRT-PCR levels of genes with ISRE or E-box motif in FLCN POS and FLCN NEG retinal pigment epithelial cells (RPE/TERT1 tet on Cas9
TP53 KO ) reveal that the identified FLCN dependent gene signature is absent in an epithelial cell type of another tissue origin. Results shown are representative for two independent experiments. To determine quantitative gene expression data levels were normalized to the geometric mean of two housekeeping genes. *OAS2 level in FLCN NEG RPE was too low to detect using qRT-PCR. See Figure 7-source data 1 for raw qRT-PCR values and fold change calculations. (C) Spearman correlation analysis reveals overlapping FLCN-dependent RNA and protein data. FLCN differential mRNAs and proteins (FDR < 0.01, n = 181, red line) showed a higher correlation than the overlap of all identified mRNAs and proteins in our datasets (blue line). Statistical methods are described in Materials and methods section. (D) iRegulon analysis of differentially expressed genes (FDR < 0.01) with highest correlation with differentially expressed proteins (r > 0.8, n = 49) reveal STAT1, STAT2, IRF1, and IRF3 as most obvious upstream transcriptional regulators. Only regulons displaying normalized enrichment scores (NES) > 4.5 are shown. STAT2 appears twice due to the fact that iRegulon ranks this transcription factor to be the most likely upstream regulator for two sets of targets genes, containing slightly different ISREmotifs upstream from the transcription start site (TSS). (E) Reintroducing FLCN (overexpression, OE) or siRNA-mediated knock down of STAT1/STAT2 (10 nM, 72 hr) revert the IFN expression gene signature induced by FLCN loss in RPTEC FLCN NEG C2. FLCN OE also lowers the enhanced expression of E-box-associated target genes but knock down of STAT1/2 has no effect on E-box-associated genes. Expression levels were determined by qRT-PCR and are representative of two independent experiments. To determine quantitative gene expression data levels were normalized to the geometric mean of two housekeeping genes. See   manifestations in BHD syndrome, varying from fibrofolliculomas of the skin, to pulmonary cysts with increased risk for pneumothorax, and renal cysts associated with increased cancer risk (Birt et al., 1977;Houweling et al., 2011;Nickerson et al., 2002;Schmidt et al., 2005;Zbar et al., 2002). While manifestations in the skin and lung in BHD probably reflect an effect of FLCN haplo-insufficiency, kidney tumorigenesis in BHD carriers starts by complete functional inactivation of the remaining wild type FLCN allele (van Steensel et al., 2007;Vocke et al., 2005). Here, we modeled the molecular and cellular effects of FLCN inactivation in the cell type most relevant for kidney tumorigenesis. The RPTEC/TERT1 cell line is widely accepted as an appropriate in vitro model system for human kidney function (Aschauer et al., 2015). Nevertheless, it is important to emphasize that many of our conclusions are based on independent clones derived from a single human cell line. Here, we observed that loss of FLCN had three main effects in RPTEC/TERT1: (1) severely reduced cellular proliferation, regardless of TP53 activity or alterations in karyotype; (2) nuclear accumulation and activation of TFE3, concomitant with TFE3/TFEB-dependent upregulation of a specific set of E-box genes linked to autophagy and lysosomal control; (3) IFN-independent upregulation and chromatin binding of STAT2, activating a gene program of typical IFN response genes, possibly in cooperation with STAT1. Upregulated genes besides STAT1 and STAT2 include MX1, IFIT1, ISG15, IRF9, and OAS2. We found the combination of these two master gene programs to be a renal-specific response to FLCN inactivation, also observed after knocking out FNIP1 and FNIP2. Re-expressing FLCN reverted these effects, proving that they are specific to FLCN loss.

RPTEC/TERT1 S T A T 1 S T A T 1 S T A T 1 S T A T 2 S T A T 2 S T A T
How these gene expression programs contribute to oncogenesis needs to be resolved, but TFE3 activation was clearly sufficient to promote uncontrolled, enhanced proliferation and loss of contact inhibition in RPTEC cells (e.g. Figure 8). Knocking out FLCN in the context of active TFE3 slowed cellular proliferation to the level observed after IFNg treatment. The upregulated IFN signature could thus explain why loss of the FLCN tumor suppressor, paradoxically, represses cellular proliferation. In relation to cancer, slow growth induced by FLCN loss and STAT2 upregulation may form a barrier to TFE3-driven renal tumorigenesis in BHD patients. On the other hand, it is possible that the activation of the IFN program is associated with a pro-oncogenic inflammatory response. The TFE3 and IFN signature programs are connected in this respect: TFE3/TFEB upregulate cytokines in macrophages that elicit an innate immune response linked to pathogen resistance (El-Houjeiri et al., 2019). In our RPTEC FLCN NEG cell models however, we observed a clear downregulation in secreted cytokine IL-8 and no consistent changes in IL-6 expression (data not shown), suggesting that, at least in renal cells, the IFN program induced by FLCN inactivation is not connected to inflammation per se.
Preliminary analysis of BHD tumor material showed the presence of immune cells at tumor margins (unpublished findings) and upregulation of ISRE genes ( Figure 6I). Collectively our data lead to the hypothesis that renal tumorigenesis in BHD patients could follow two different paths: either proinflammatory effects of specific E-box and/or ISRE genes are further aggravated by secondary mutations during tumor evolution, or growth-suppression by IFN signature genes, which is dominant over the effects of TFE3 activation, is gradually lost by additionally acquired mutations thus leading to  Figure 8-source data 1 for raw qRT-PCR values and fold change calculations. (B) Western blots of subcellular fractions show enhanced binding of TFE3 to DNA in FLCN NEG RPTEC and SFPQ-TFE3 RPTEC. STAT2 DNA-binding was enhanced upon FLCN loss but reduced by SFPQ-TFE3 over-expression in RPTEC. STAT2 was blotted on separate blots of the same lysates. Histone H3 levels were used as loading control and as marker for chromatin fraction. Western blot was performed two times, using independent fractionations. (C) Colony formation assays show that loss of FLCN in wild-type or SFPQ-TFE3 RPTEC reduces colony outgrowth. SFPQ-TFE3 RPTEC show more colonies than wild type RPTEC after 10 days. Insets show bright field images (Â20 magnification). Cells were seeded in three technical replicates and experiment was performed twice. (D) Loss of FLCN in RPTEC results in slower growth, which is reverted when FLCN expression is restored by over-expression. Cell lines were seeded in equal densities and total cell numbers were counted three times within 11 days. Results shown are representative for two independent experiments. (E) Treatment with IFNg (100IU/ml) or combining FLCN KO in SFPQ-TFE3 RPTEC results in growth inhibition. Cell lines were seeded in equal densities and total cell number was counted twice within 7 days. Results shown are representative for two independent experiments. The growth curve of FLCN KO RPTEC (D) is added for comparison. The online version of this article includes the following source data and figure supplement(s) for figure 8: Source data 1. Raw qRT-PCR values and fold change calculations belonging to Figure 8A and  TFE3-driven tumor progression. This suggests that certain IFN response genes which strongly reduce proliferation form a deeper tumor suppressive layer protecting against uncontrolled proliferation in the absence of sufficient FLCN expression. This hypothesis fits with the observation that renal tumors specifically driven by TFE3 activation behave more aggressively as compared to slowly growing BHD tumors. Speculatively, BHD tumors fall into two mutually exclusive classes, with either high or low IFN signatures and/or STAT2 expression depending on their growth properties.
From our results, it is clear that FLCN loss promotes STAT2 binding to chromatin but we did not resolve a clear mechanism by which FLCN loss promotes these changes. Because TFE3 activation is not sufficient to induce STAT2 or upregulate ISRE genes, and reciprocally STAT1/2 siRNA does not downregulate the E-box genes, we conclude that the two gene activation programs are separate effects of FLCN loss. We considered that FLCN might affect STAT2 protein stability, in a similar manner as observed for a number of viral proteins that bind directly to STAT2 (Grant et al., 2016;Morrison et al., 2019) but found no evidence for direct complex formation between STAT2 and FLCN. Also, the protein half-life of STAT2 was similar in the absence of FLCN and no shift was observed in STAT2 nucleo-cytoplasmic distribution (data not shown).
With respect to candidate TFE3 target genes that drive renal epithelial cell transformation, it is clear that in the absence of FLCN, particularly GPNMB, RRAGD, ASAH1 and FNIP2, the latter in an apparent feedback mechanism, are strongly upregulated. In renal cells lacking FNIP1 and FNIP2, GPNMB and RRAGD are also very strongly induced. This illustrates the potential value of GPNMB and RRAGD as positive biomarkers for FLCN inactivation such as in BHD tumors.
GPNMB is a transmembrane protein frequently upregulated in a wide variety of tumors, including lung and renal cancer, yet it is unclear whether GPNMB overexpression in itself is tumorigenic (Taya and Hammes, 2018). Importantly, GPNMB can be targeted therapeutically using antibodydrug conjugates which are in clinical trials for cancer therapy, such as glembatumumab vedotin (Rose et al., 2017), providing an entry point for the evaluation of glembatumumab in the treatment of BHD tumors. Furthermore, RRAGD is a candidate oncogenic target gene of TFE3, previously associated with loss of FLCN (Di Malta et al., 2017;Tsun et al., 2013). Recently, Napolitano et al., 2020 described TFEB to be the main driver of kidney abnormalities in a BHD mouse model. Their results show that TFEB is phosphorylated by mTORC1 in substrate-specific mechanism that is mediated by Rag GTPases. FLCN is a key regulator of Rag GTPases but despite the fact that we found clear upregulation of both RagC and RagD, we found no evidence for mTORC1 hyperactivation in our human in vitro model system. There might be crucial differences in renal tumorigenesis between mice and humans, where mTORC1 activation may occur at a later stage in oncogenic transformation of FLCN NEG renal epithelial cells.
In addition to the upregulated TFE3 and ISRE programs, we observed that several genes were downregulated by FLCN loss. GSEA analyses of downregulated genes showed overlapping biological processes but iRegulon analyses failed to reveal a clear common upstream transcriptional regulator of these genes in mRNA and protein data. Taken together however, the repository of FLCN target genes and proteins presented here provides a clear basis for further investigations into specific roles in kidney cancer and other BHD-related symptoms. This could facilitate the discovery of biomarkers for early-stage tumorigenesis and therapeutic strategies to prevent or treat RCC metastases in BHD patients.

Cell culture
Renal proximal tubular epithelial cells (RPTEC/TERT1, ATCC CRL-4031) were maintained in DMEM/ F12 (Gibco, Life Technologies, Thermo Fisher Scientific Inc, Waltham, Massachusetts, US) according to the manufacturer's protocol with addition of 2% fetal bovine serum (FBS, Gibco). To maintain the selective pressure for immortalization 0.1 mg/ml G418 Sulfate (Calbiochem, Merck, Darmstadt, Germany) was added. Cell lines were cultured in a humidified atmosphere at 37˚C and 5% CO 2 . Retinal pigment epithelial cells (RPE-1/hTERT, ATCC CRL-4000) were maintained in DMEM (Gibco, Life Technologies) with addition of 8% FBS and 1 mM Sodium Pyruvate (Gibco, Life Technologies). The generation of used RPE1-hTERT tet-on Cas9 TP53 KO cells was described earlier (Benedict et al., 2020). Both cell lines were obtained recently from ATCC and experiments were performed exclusively with low-passage cell lines which were regularly tested to exclude Mycoplasma infections; follow-up authentication was performed on the basis of functional assays, gene expression patterns and cellular morphology.

Karyotype analysis
After standard cytogenetic harvesting and GTG banding, at least 35 metaphase cells were analyzed and described according to ISCN 2016(McGowan-Jordan et al., 2016.

Virus production and infection
To create an inducible Cas9 RPTEC cell line, lentiviral production and transduction took place according to the Lenti-X Tet-On 3G Inducible Expression System (Clontech, Takara Bio, Japan) technical manual. In short, Cas9 cDNA was cloned into the pLVX-Tre3G plasmid where after Tre3G-Cas9 and Tet3G lentiviral particles were produced in HEK293T cells. For transduction RPTEC cells were seeded (250 k/well) in a 6-wells plate one day prior to infection. The next day growth media was removed and 1 ml media containing viruses was added. Cells were incubated overnight and after 24 hr media was replaced with 2 ml fresh media. The next day cells were transferred to 10 cm plates and Puromycin (3 mg/ml, Sigma-Aldrich, St. Louis, Missouri, USA) was added to select for successfully transduced cells. G418 was already present in growth media to maintain selective pressure for immortalization and therefore not added for Tet3G selection. For FLCN rescue experiments FLCN cDNA was cloned into pLenti CMVie-IRES-BlastR (gift from Ghassan Mouneimne, Addgene plasmid #119863 [Puleo et al., 2019]). The SFPQ-TFE3 fusion sequence was derived from gene expression analysis of a patient-derived pediatric RCC (Calandrini et al., 2020) and subsequently cloned as geneblock into pLKO-Ubc lentiviral backbone (Fumagalli et al., 2017) using Gibson assembly. For both lentiviral particles were produced in HEK293T and transduced into RPTEC. Blasticidin (15 mg/ ml, Invitrogen, Life Technologies) was added for selection of successfully transduced cells and protein overexpression was confirmed by western blotting.
After 24 h cells were transferred to 10 cm plates and Nutlin-3 (10 mM, Selleck Chemicals, Houston, Texas, USA) was added for selection of TP53 knock-out and thus successfully transfected cells. After selection cells were grown in limiting dilution in 96-wells plates to generate single cell clones. Subsequently, knockout status was assessed by western blot and Sanger sequencing. FLCN KO RPTEC cell line described in Figure 8 was created using Synthego's Synthetic cr:tracrRNA Kit and corresponding manual. Cas9/gRNA (FLCN_exon 4 GAGAGCCACGAUGGCAUUCA + modified EZ scaffold) RNP complexes were transfected transiently using Neon Electroporation System (Thermo-Fisher). Subsequently cells were grown in limiting dilution in 96-well plates to generate single-cell clones and knockout status was assessed by western blot and Sanger sequencing. Sequenced samples were analyzed by manual alignment or using the Synthego ICE analysis (ice.synthego.com) tool which gives a quantitative spectrum of indels that are formed around the cut site.
DNA isolation, PCR, and sequencing DNA was extracted from dry cell pellet (~1.5E6 cells) according to technical manual of DNA isolation kit (QIAamp DNA Blood Mini Kit, Qiagen, Venlo, Netherlands). Subsequently equal amounts of DNA were amplified by PCR. Tubes were placed in a thermal cycler (Veriti, Thermo-Fischer Scientific) for amplification with specific PCR primer mixes (10 mM). PCR program used for amplification was 1 cycle of 94˚C for 3 min, 5 cycles of 94˚C for 30 s, 65˚C for 30 s, 72˚C for 120 s, 30 cycles of 94˚C for 30 s, 60˚C for 30 s, 72˚C for 2 min, 72˚C for 10 min and ending in a rapid thermal ramp to 10˚C. Here after PCR purification (ExoSAP-IT PCR product cleanup (USB products, Affymetrix, Santa Clara, California, USA) and Sephadex G-50 superfine gel (GE Healthcare, Chicago, Illinois, USA)) took place and samples were further analyzed by sequencing. Sequencing was either performed in-house or at Eurofins Genomics. For PCR and sequencing, following primers were used: In-house sequencing was done according the BigDye Terminator v1.1 Sequencing Kit (Life Technologies) protocol. For each sample~75 ng DNA and specific sequencing primers (3.3 mM) were used. Tubes were placed in a thermal cycler for amplification with specific PCR primer mixes. PCR program used for amplification was: 1 cycle of 95˚C for 1 min, 30 cycles of 95˚C for 10 s, 55˚C for 5 s, 60˚C for 2 min and ending in a rapid thermal ramp to 10˚C.

Differential expression analysis of RNAseq data
We used the R package edgeR (Robinson et al., 2010) to compare RNA-sequencing profiles between FLCN POS and FLCN NEG replicates, as well as between TP53 POS and TP53 NEG . This involved reading in the gene-level counts, computing library-size normalizing factors using the trimmed-mean of M-values (TMM) method and then fitting a model to estimate the group effect. Obtained p-values were corrected for multiple testing using the Benjamini-Hochberg false discovery rate (FDR) step-up procedure (Benjamini and Hochberg, 1995).

Sample preparation for LC-MS/MS
Equal protein lysates, of each cell line in duplicate, were separated on precast 4-12% gradient gels using NuPAGE SDS-PAGE (Invitrogen, Carlsbad, California, USA). Gels were fixed in 50% ethanol/ 3% phosphoric acid solution and stained with Coomassie R-250. Gel lanes were cut into five bands (see cutting scheme Figure 2-figure supplement 2B) and each band was cut into~1 mm 3 cubes. Gel cubes were washed with 50 mM ammonium bicarbonate/50% acetonitrile and were transferred to a microcentrifuge tube, vortexed in 50 mM ammonium bicarbonate for 10 min and pelleted. The supernatant was removed, and the gel cubes were again vortexed in 50 mM ammonium bicarbonate/50% acetonitrile for 10 min. After pelleting and removal of the supernatant, this wash step was repeated. Subsequently, gel cubes were reduced in 50 mM ammonium bicarbonate supplemented with 10 mM DTT at 56˚C for 1 hr, where after supernatant was removed. Gel cubes were alkylated in 50 mM ammonium bicarbonate supplemented with 50 mM iodoacetamide for 45 min at RT in the dark. Next, gel cubes were washed with 50 mM ammonium bicarbonate/50% acetonitrile, dried in a vacuum centrifuge at 50˚C and covered with trypsin solution (6.25 ng/ml in 50 mM ammonium bicarbonate). Following rehydration with trypsin solution and removal of excess trypsin, gel cubes were covered with 50 mM ammonium bicarbonate and incubated overnight at 25˚C. Peptides were extracted from the gel cubes with 100 ml of 1% formic acid (once) and 100 ml of 5% formic acid/50% acetonitrile (twice). A total of 300 ml extracts were stored at À20˚C until use. Prior to LC-MS, the extracts were concentrated in a vacuum centrifuge at 50˚C, volumes were adjusted to 50 ml by adding 0.05% formic acid, filtered through a 0.45 mm spin filter, and transferred to a LC auto sampler vial.

LC-MS/MS
Peptides were separated by an Ultimate 3000 nanoLC-MS/MS system (Dionex LC-Packings, Amsterdam, Netherlands) equipped with a 45 cm Â 75 mm ID fused silica column custom packed with 1.9 mm 120 Å ReproSil Pur C18 aqua (Dr Maisch GMBH, Ammerbuch-Entringen, Germany). After injection, peptides were trapped at 6 ml/min on a 10 mm Â100 mm ID trap column packed with 5 mm 120 Å ReproSil Pur C18 aqua in 0.05% formic acid. Peptides were separated at 300 nl/min in a 10-40% gradient (buffer A: 0.5% acetic acid (Fisher Scientific), buffer B: 80% ACN, 0.5% acetic acid) in 60 min (100-min inject-to-inject). Eluting peptides were ionized at a potential of +2 kVa into a Q Exactive mass spectrometer (Thermo Fisher, Bremen, Germany). Intact masses were measured at resolution 70,000 (at m/z 200) in the orbitrap using an AGC target value of 3E6 charges. The top 10 peptide signals (charge-states 2+ and higher) were submitted to MS/MS in the HCD (higher-energy collision) cell (1.6 amu isolation width, 25% normalized collision energy). MS/MS spectra were acquired at resolution 17,500 (at m/z 200) in the orbitrap using an AGC target value of 1E6 charges, a maxIT of 60 ms, and an underfill ratio of 0.1%. Dynamic exclusion was applied with a repeat count of 1 and an exclusion time of 30 s.

Protein identification and label-free quantitation
MS/MS spectra were searched against the reference proteome FASTA file (42161 entries; swisspro-t_2017_03_human_canonical_and_isoform). Enzyme specificity was set to trypsin, and up to two missed cleavages were allowed. Cysteine carboxamidomethylation (Cys, +57.021464 Da) was treated as fixed modification and methionine oxidation (Met, +15.994915 Da) and N-terminal acetylation (N-terminal, +42.010565 Da) as variable modifications. Peptide precursor ions were searched with a maximum mass deviation of 4.5 ppm and fragment ions with a maximum mass deviation of 20 ppm. Peptide and protein identifications were filtered at an FDR of 1% using the decoy database strategy. The minimal peptide length was seven amino acids. Proteins that could not be differentiated based on MS/MS spectra alone were grouped into protein groups (default MaxQuant settings). Searches were performed with the label-free quantification option selected. Proteins were quantified by spectral counting, that is, the number of identified MS/MS spectra for a given protein (Liu et al., 2004) combining the five fractions per sample. Raw counts were normalized on the sum of spectral counts for all identified proteins in a particular sample, relative to the average sample sum determined with all samples. To find statistically significant differences in normalized counts between sample groups, we applied the beta-binomial test , which takes into account within-sample and between-sample variation using an alpha level of 0.05. For proteomic analyses of tissue material, human kidney lysates (obtained commercially NB820-59231 (HK1) and sc-363764 (HK2)) and fresh BHD tumor tissues (BHDT1 and BHDT2, see details below) lysed in NP40 lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP40) were used and analysed as described above.

Patient material
BHD T1 and BHD T2 tumor material was obtained with informed consent for research use according to Amsterdam UMC Medical Ethics Committee guidelines. Both tissues are leftover material from surgery isolated by a pathologist and stored at the Amsterdam UMC Birt-Hogg-Dubé biobank filed under number 2019.359.

Gene ontology and gene set enrichment analysis
Protein-protein relations were retrieved from the STRING database v11 (Szklarczyk et al., 2019) and imported as a network in Cytoscape v3.5.1 (Shannon et al., 2003). Enriched gene ontology terms (biological processes) were obtained with the BiNGO v3.0.3 app for Cytoscape (Maere et al., 2005), using hypergeometric testing with Benjamini-Hochberg multiple-test correction. Ontology analysis was performed with both the major connected component of the network and subclusters obtained with the ClusterONE v1.0 app for Cytoscape (Nepusz et al., 2012). Gene set enrichment was performed with GSEA software (Subramanian et al., 2005) using gene sets from Molecular Signatures Database v6.1 (Liberzon et al., 2011). Pre-ranked GSEA was performed with ranking by the statistics of the differential FLCN NEG -FLCN POS tests. As a ranking metric, we used the product of the sign of the fold change and the negative value of the log10-transformed p-value. Visualization of results was performed in R using the ggplot2 package (Wickham, 2016).

Amino acid starvations and immunofluorescence
For immunofluorescence experiments RPTECs were grown on coverslips (15 mm diameter) on day 1. The medium was refreshed at the end of day 2 or replaced by serum free medium. Amino acid starvation was done with cells that had been serum starved overnight in custom-made DMEM without amino acids but containing 2 mM L-glutamine for 2 hr. Amino acid addback was done by adding an equal volume of starvation medium containing twice the amount of essential amino acids for 15 min. Next, cells were washed twice with ice-cold PBS and fixed in 4% para-formaldehyde for 20 min at room temperature. After washing with PBS, cells were permeabilized with 0.25% Saponin (Sigma Aldrich) in PBS and blocked with blocking solution (10% FCS, 0.25% Saponin in PBS). Coverslips were incubated overnight at room temperature with primary antibodies in blocking solution. The next day, coverslips were washed in 0.25% Saponin in PBS and incubated with secondary antibodies from Fisher Scientific (Alexa 488-goat anti-mouse; 10696113 and Alexa 568-goat anti-rabbit; A11036) for 2 hr at room temperature. Hereafter cells were mounted with Immu-Mount (Thermo Scientific Shandon). Specimens were visualized under Zeiss LSM510 microscope and imaged using Zeiss vision software. The following antibodies were used: mTOR (7C10, CST 2983; 1:300), Lamp2 (H4B4, Ab24631; Abcam; 1:400) and TFE3 (CST 14779; 1:300). For western blot analyses of starvation experiments, RPTECs were seeded in 6-wells plates and treated in like manner. Instead of fixation cells were scraped into 1x Laemmli sample buffer (Sigma Aldrich) and immunoblotted as described below. Starvations, as well as immunofluorescent stainings and immunoblotting were performed three times.

Immunoblotting
For immunoblotting dry cell pellets (~1.5E6 cells) were lysed in NP40 lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP40) supplemented with protease and phosphatase inhibitors (Roche). Subsequently samples were boiled at 70˚C for 5 min in 1x NuPAGE LDS sample buffer (Novex NP0007, Thermo-Fischer) with 10% 1M DTT (Sigma) and equal amounts were separated by 4-15% or 8-16% SDS-PAGE (BioRad) and blotted onto polyvinylidene fluoride (PVDF) transfer membranes (Merck). After transfer, membrane was blocked for 1 hr at room temperature with 5% milk (ELK, Campina, Amersfoort, Netherlands) in TBST. The primary antibody incubation was ON at 4˚C in 2.5% milk in TBST. Next day, membrane was washed and incubated with appropriate secondary antibody (Dako) for 3 hr at 4˚C in 2.5% milk in TBST. For detection of phosphorylated proteins, blocking and incubation steps were performed with Bovine Serum Albumin (BSA Fraction V, Roche) instead of milk. After incubation the membrane was thoroughly washed where after bands were visualized by chemoluminescence (ECL prime, Amersham, VWR, Radnor, Pennsylvania, USA) in combination with ChemiDoc Imaging Systems (BioRad, Hercules, California, USA). For subcellular fractionation experiments the Subcellular Protein Fractionation Kit for Cultured Cells (78840, Thermo Scientific) was used according to manufacturer's protocol.

Growth curve modeling
We modeled the growth curve data with a random intercept model (Mirman, 2016;Molenberghs and Verbeke, 2000): with Y ijk the number of cells of knock-out j at time k, I TP53;j a binary variable indicating if the i th observation is coming from the Cas9-TP53 KO (I TP53;j ¼ 1) or from the Cas9-FLCN KO (I TP53;j ¼ 0), b 0~N 0; s 2 b À Á , and " ijk~N 0; s 2 " À Á . Scheme 1 shows the modeled growth curves for both knock-outs. Based on the estimated values of the random intercept model (Table 1), we can conclude that there is a statistically significant difference (a = 0.05) in the growth curves of Cas9-FLCN KO and Cas9-TP53 KO (b 4 6 ¼ 0; p-value<<0.05), and that there is a statistically significant increase in the number of cells over time (b 1 6 ¼ 0; p-value<<0.05; and b 2 6 ¼ 0; p-value<<0.05).
Scheme 1: Growth rates of FLCNKO and TP53KO cells with fitted curves Scheme 1. Growth curves of FLCNKO and TP53KO with the fitted curves.
Cytometric bead array and enzyme-linked immunosorbent assay (ELISA) To detect IFNg in supernatants, the Human IFN-g Flex Set (560111, BD Biosciences, Franklin Lakes, New Jersey, USA) was used according to the manual for the BD CBA Human Soluble Protein Master Buffer Kit. Measurements were done on BD LSRFortessa Flow Cytometer (BD Biosciences) and FCAP Array Software. Supernatants were measured twice in duplicate. For detection of IFNa in supernatants, the VeriKine-HS Human IFN-a All Subtype ELISA kit (#41115, PBL assay science, Piscataway, New Jersey, USA) was used according to manufacturer's protocol. For calculation of IFNa concentrations, blank optical densities were subtracted from standard and sample optical densities. Supernatants were measured twice in duplicate.

Clonogenic assays
To assess clonogenicity a Crystal Violet (CV) staining was performed. Cell lines were plated in three technical replicates in a six-well plate at low concentration (3000 cells/well). Plates were incubated in a humidified atmosphere at 37˚C and 5% CO 2 for approximately 10 days until plates displayed colonies with substantially good size. At that moment, cells were fixed with ice-cold methanol (Methanol !99.8%, Sigma-Aldrich) and plates were incubated for 5 min at RT after which cells were washed with PBS and stained with CV (0.05% solution, Merck) for 30 min at RT. Then CV was washed away with tap water, plates were air dried and scanned for further analysis. Detail photographs (Â20) were obtained with AxioScope microscope camera and corresponding software (AxioVision SE64 Rel. 4.9.1, Carl Zeiss, Jena, Germany). At least, two independent experiments with three technical replicates were used to determine colony formation capacities. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.  Figure 3A). The following datasets were generated: