medfusion是什么牌子SFPQ-TFE3 reciprocally regulates mTORC1 and induces lineage plasticity in a mouse model of renal tumorigenesis

We generated a transgenic mouse which conditionally expresses the human SFPQ-TFE3 fusion, downstream of a LoxP-Stop-LoxP (LSL) cassette, under control of a strong chicken beta-actin (CAG) promoter in the Rosa 26 locus (SFPQ-TFE3^LSL; hereafter referred to as ST mice). Briefly, the LSL-SFPQ-TFE3 allele, containing the human SFPQ CDS (exon 1-9) /human TFE3 CDS (exon 5-10) (type 2 fusion)¹, was cloned into intron 1 of the Rosa26 locus in reverse orientation, separated from the native CAG promoter by a stop sequence flanked by LoxP sites (LSL), thus enabling conditional excision of the stop sequence upon Cre-mediated recombination and expression of the fusion transgene (Supplementary Fig. 1A). To validate conditional expression of the allele, renal tubular epithelial cells from adult C57BL/6 controls or ST transgenic mice were harvested, dissociated, and cultured in the presence of a control adenovirus or adenovirus expressing Cre-recombinase. Primary cells from ST, but not control mice, showed strong induction of SFPQ-TFE3 expression and the canonical MiT/TFE E-box target GPNMB⁴ on immunoblotting (Supplementary Fig. 1B). Genotypes were additionally confirmed by PCR (Supplementary Fig. 1C, D).

To conditionally express SFPQ-TFE3 in tubular epithelial cells during renal development, we crossed ST transgenic mice with Ksp-Cadherin (Cadherin-16 [Cdh16]) -Cre mice, in which the Cre recombinase is specifically expressed from the Cdh16 promoter in distal tubular epithelial cells and collecting ducts of the kidney starting from embryonic day E12.5³², to generate SFPQ-TFE3^LSL; Ksp-Cre mice (hereafter referred to as STK mice). While the STK mice were born in expected mendelian ratios and did not manifest any gross anatomic differences compared to littermate controls at birth, no STK mice survived beyond postnatal day 20 (P20) (Fig. 1A). Histologically, STK kidneys had a disorganized appearance at P1 with mild and scattered tubular dilation and abundant expression of nuclear TFE3 in tubular epithelial cells by IHC (Fig. 1B). By P15, STK kidneys were grossly enlarged and markedly cystic, with disruption of the cortical architecture (Fig. 1C, D), and significantly elevated kidney to body weight ratios, as well as blood urea nitrogen (BUN) and serum creatinine (Fig. E–G). Indirect immunofluorescence for LTL, which labels the brush border of differentiated proximal tubules, showed diminished staining in the STK kidney tubules, compared to littermate controls, with rare, dilated tubules expressing distal tubular markers such as DBA (Fig. 1H). Histological analysis of kidneys from 3-week-old STK mice demonstrated tubular enlargement by solid nests of large epithelioid cells with abundant clear-to-eosinophilic cytoplasm and monomorphic nuclei (Fig. 1I). The mutant kidneys also showed luminal occlusion with frequent psammomatous calcification (Fig. 1I), a feature of human tRCC⁶, and trichrome staining highlighted inter-tubular fibrosis (Supplementary Fig. 2A). By P15, there was diffuse tubular nuclear TFE3 expression in STK kidneys, and upregulation of an array of melanocytic/lysosomal markers over-expressed in human MiT/TFE-related neoplasms (PMEL, MelanA and GPNMB)^3,4,6 (Fig. 1J–L). Consistent with a pro-tumorigenic role of SFPQ-TFE3, proliferation as measured by BrDU positivity, Ki67 staining, and phosphorylated H3 were all significantly elevated in STK kidneys compared to littermate controls by 2 weeks of age (Supplementary Fig. 2B–D). Cumulatively, these results indicate that neonatal expression of renal SFPQ-TFE3 disrupts renal development with subsequent renal failure culminating in early postnatal death. Though there was evidence of increased tubular cell proliferation, the early renal failure precluded aging mice to assess tumor development in the STK model.

To examine effects of SFPQ-TFE3 expression in renal tubular cells following completion of kidney development, we leveraged a conditional, tamoxifen-inducible Pax8 Cre-ERT model³³. Pax8 is a critical renal lineage transcription factor, thus Cre is diffusely expressed in most renal epithelial cells of the proximal and distal renal tubules, loops of Henle, collecting ducts, and the parietal epithelial cells of Bowman’s capsule following tamoxifen exposure in this model. SFPQ-TFE3^LSL; Pax8-CreERT (STP) mice were born at expected Mendelian ratios and survived to adulthood, when they were injected with tamoxifen at 8–12 weeks of age to induce Cre expression. When mice were sacrificed at 3.5 months following tamoxifen, STP kidneys were variably enlarged with bilateral solid tumors (Fig. 2A), showing strong, diffuse, nuclear TFE3 induction (Fig. 2B). Tumor cells were diffusely infiltrative around normal renal tubules, leading to disappearance of the normal cortico-medullary junction, and tumor replacement of a large part of the kidney parenchyma (Fig. 2B). Kidney/ body weight ratios and BUN levels were also elevated in some STP mice by 3.5 months after tamoxifen, though some variability in tumor burden was evident in this inducible model, with tamoxifen-injected, female STP mice frequently demonstrating higher kidney/ body weight ratios and BUN levels than males at 4.5 months, consistent with previous studies in humans and mice^26,34 (Fig. 2C, D). To validate these observed sex-based differences in tumor burden, we digitally annotated histologic tumor volume based on GPNMB positivity in larger cohorts of STP mice at 3.5 months, demonstrating statistically significant sex-based differences at this age (Supplementary Fig. 3A, B). Histologically, the tumors were comprised of nests of monomorphic epithelioid cells with clear to eosinophilic cytoplasm and large round nuclei (Fig. 2E). Though they were invasive, the tumor cells were similar in morphology to intratubular proliferations seen in STK kidneys (Fig. 1I) and broadly resembled those seen in human PEComas (Supplementary Fig. 3C). By IHC and immunoblotting, STP tumor cells showed strong expression of nuclear TFE3, the canonical TFE3 target GPNMB, as well as expression of melanocytic and lysosomal markers (PMEL, MelanA and Cathepsin K) commonly upregulated in human MiT/TFE-related neoplasms (Fig. 2F, G). In contrast, STP tumor cells entirely lacked expression of proximal (LTL) and distal (DBA) renal tubular markers (Supplementary Fig. 3D), as well as cytokeratins, such as type II (CK8) or type I keratins (including CK10, 13, 14, 15, 16, 17 and 18) (Fig. 2H, left panels). Notably, STP tumor cells showed minimal vimentin expression, and were negative for smooth muscle marker (α-SMA) by IHC (Fig. 2H, right panels). Cumulatively, these findings suggested that post-natal expression of SFPQ-TFE3 within murine renal tubular cells results in highly penetrant, infiltrative epithelioid tumors, with brisk expression of melanocytic/lysosomal markers. Importantly, these markers are characteristic of human MiT/TFE-related tumors, such as the tRCC cases inadvertently included in The Cancer Genome Atlas (TCGA) papillary RCC (KIRP) cohort³⁵ (Fig. 2I) or tuberous sclerosis (TSC)-related PEComas (angiomyolipomas)^3,36 (Fig. 2J).

mTOR signaling has previously been shown to be activated in human tRCC samples and pre-clinical models by transcriptomic and proteomic profiling^{18,24,25,27,28}, including in a recent mouse model of ASPSCR1-TFE3 tRCC²⁶, though this signaling pathway was not examined in the previously published PRCC-TFE3 transgenic mice⁴. We performed gene set enrichment analyses (GSEA)³⁷, using Hallmark gene sets³⁸ on publicly available gene expression data from the PRCC-TFE3 mice⁴, and found significant enrichment of genes associated with PI3K/ AKT/ MTOR signaling in the transgenic mice at 7 months compared to controls (Supplementary Fig. 4A). By immunoblotting, there were increased levels of mTORC1 substrate p-4E-BP1 in PRCC-TFE3; KSP-Cre kidney tumor lysates compared to controls, though total 4E-BP1 levels were increased as well (Supplementary Fig. 4B) as has been seen previously in other mouse models with mTORC1 activation in the kidney²⁰. We then characterized mTORC1 signaling in the SFPQ-TFE3 transgenic models. Phosphorylation of canonical mTORC1 substrates and downstream signaling intermediates (p70 S6K, 4E-BP1 and S6) was elevated in P15 STK transgenic kidney tumor lysates compared to controls, by immunoblotting (Fig. 3A) and IHC (Fig. 3B). Levels of phosphorylated canonical mTORC1 substrates and downstream signaling intermediates were also elevated in kidney tumor lysates from 3.5-month STP mice compared to treated littermate controls (with substantial increase in total substrate levels seen in these tumors) by immunoblotting (Fig. 3C), and similar evidence of mTORC1 activation seen by IHC (Supplementary Fig. 4C). Notably, in the STP mice, expression of phosphorylated mTORC1 substrates were specifically increased in the TFE3 (+) tumor cells by IHC, with low expression in surrounding normal kidney (Fig. 3D). In addition, levels of nuclear TFEB, an established non-canonical mTORC1 substrate³⁹, were also lower in the tubular cells from STK transgenic mice compared to littermate control kidneys by IHC, and a similar pattern was seen in tumor cells from STP transgenic mice, compared to internal control surrounding normal renal tubules (Fig. 3E). These findings are consistent with TFEB hyperphosphorylation and cytoplasmic retention due to increased mTOR signaling.

To validate these findings, we also examined mTORC1 activation in multiple tRCC human cell line models⁴, comparing them to ccRCC cell lines by immunoblotting. Phosphorylation of canonical (p70 S6K) and non-canonical (TFEB) mTORC1 substrates and downstream signaling intermediates (S6) was generally elevated in the tRCC cell lines with SFPQ-TFE3 (UOK145), PRCC-TFE3 (UOK120, UOK124, UOK146) and NONO-TFE3 (UOK109) fusion expression, compared to ccRCC controls (UOK111, UOK140 and UOK150) (Fig. 3F). Because comparison of genetically disparate UOK cell lines does not allow examination of the isolated effects of TFE3 fusions on mTORC1 activation in an isogenic setting, we engineered HEK293 cells with doxycycline-inducible expression of WT-TFE3, S321A-TFE3 (an mTORC1-site phosphodeficient TFE3 point mutant that is constitutively nuclear localized⁴⁰), SFPQ-TFE3, PRCC-TFE3 and NONO-TFE3, using the Flp-In-T-Rex^TM system in HEK293 cells. Doxycycline-mediated induction of the variably sized fusion proteins and canonical E-box targets GPNMB⁴, RRAGD⁴¹ and FLCN ⁴¹ was confirmed by immunoblotting (Supplementary Fig. 4D). Phosphorylation of mTORC1 substrates and downstream signaling intermediates was strongly elevated upon induction of SFPQ-TFE3 and PRCC-TFE3 in this system, with subtler induction due to WT-TFE3, S321A-TFE3, and minimal induction with NONO-TFE3 (Supplementary Fig. 4D). To validate these results in a renal tubular cell line, we leveraged a previously described HK2 cell system (human proximal renal tubular epithelial cell line) with doxycycline-inducible expression of WT-TFE3 and TFE3 fusion proteins using the rtTA3 (Tet-on) construct⁴². These immunoblotting experiments confirmed increased phosphorylation of mTORC1 substrates and downstream signaling intermediates with expression of TFE3 fusion proteins in human renal tubular epithelial cells (Supplementary Fig. 4E). A dose response time course for doxycycline confirmed increased TFEB phosphorylation in HK2 cells with inducible SFPQ-TFE3 expression (Supplementary Fig. 4F). Taken together, these data suggest that MiT/TFE fusion gene expression induces increased mTORC1 activity in human cells.

The mechanism of mTORC1 activation downstream of TFEB expression has previously been ascribed to increased MiT/TFE-mediated transcription of RRAGD, an essential component of the Rag GTPase nutrient-sensing complex that regulates amino acid-induced mTORC1 activation at the lysosome^{19,20,29,30,43}. However, to our knowledge, increased RRAGD transcription has been documented in only a single cell line model of tRCC²⁹. To explore a potential role for RRAGD, or its functionally redundant paralog RRAGC ⁴⁴, in mediating mTORC1 activation in tRCC models, we first examined RRAGC and RRAGD gene expression in human tRCC specimens³⁵ (Supplementary Fig. 5A), where both genes were significantly upregulated in tRCC samples compared to papillary RCC without gene fusion expression. We then examined RRAGC and RRAGD gene expression in the aforementioned panel of 8 UOK tRCC cell lines, and confirmed increased expression of both genes, most notably RRAGD, compared to clear cell RCC control cell lines (Supplementary Fig. 5B, C), with similar findings at the protein level by immunoblot (Supplementary Fig. 5D). RagD protein expression was similarly upregulated in HEK293 and HK2 cells upon doxycycline-induced expression of TFE3 fusion proteins (Supplementary Fig. 4D, E). Finally, we examined Rragc/d gene expression in STK and STP transgenic mice kidneys, where it was also increased compared to control kidneys (Supplementary Fig. 5E, F).

We next examined whether suppression of RRAGC or RRAGD activation via amino acid deprivation or transient RRAGC or RRAGD silencing was sufficient to suppress mTORC1 signaling in cell line models expressing TFE3 fusion proteins. In HK-2 cells with inducible SFPQ-TFE3 expression, amino acid starvation decreased phosphorylation of direct mTORC1 substrates 4E-BP1, p70S6K, and TFEB to levels near control cells lacking fusion induction, though relative increased phosphorylation of substrates in cells expressing SFPQ-TFE3 compared to those without induction remained evident (Supplementary Fig. 5G). Either transient RRAGC and/or RRAGD knockdown in SFPQ-TFE3-expressing cells resulted in variably decreased phosphorylation of 4EB-P1, S6, and/or TFEB compared to control, though similar results were not seen for p70S6K phosphorylation (Supplementary Fig. 5H). Similarly, we transiently overexpressed increasing concentrations of HA-tagged, inactive RRAGC^GTP (Q120L) or RRAGD^GTP (Q121L)⁴³ in HK2/ SFPQ-TFE3 cells (Supplementary Fig. 6A). In cells without doxycycline induction of SFPQ-TFE3, over-expression of either inactive RRAGC or RRAGD decreased TFEB phosphorylation as previously described^44,45, while the effects were more modest in cells expressing SFPQ-TFE3, with no change in p-TFEB, and a slight reduction of p-4E-BP1 observed at the highest concentrations of inactive RRAGD transfection. Collectively, these data indicate that while RRAGC and RRAGD expression is increased with TFE3 fusion expression, transient knockdown or expression of inactive mutants of RRAGC and/or RRAGD is insufficient to completely suppress phosphorylation of mTORC1 substrates in this context.

Because mTORC1 recruitment to the lysosome is mediated by RRAGC/D in wild type cells and is associated with kinase activation, we next examined lysosomal levels of the mTORC1 subunit Raptor in HK-2/ SFPQ-TFE3 cells. Relative to cytoplasmic levels, lysosomal levels of Raptor were increased in HK-2/ SFPQ-TFE3 cells with increasing duration of doxycycline (Supplementary Fig. 6B, top panels; Supplementary Fig. 6C, D). As a control, mTORC1 inhibition with Torin1 increased relative lysosomal Raptor as expected^19,46. In contrast, RRAGC or RRAGD siRNA only variably and non-significantly decreased lysosomal Raptor levels (Supplementary Fig. 6C, D), consistent with variable effects of RRAGC and/or RRAGD knock-down on mTORC1 substrate phosphorylation. Taken together, our results suggest that RRAGC and/or RRAGD may contribute to mTORC1 signaling in tRCC but are likely not the only mechanism leading to increased mTORC1 activity.

In addition to RRAGC/D, the vaculoar H⁺-ATPse (v-ATPase) is an MiT/TFE transcriptional target⁴⁰ and an important component of the lysosomal machinery that activates mTORC1⁴⁷. Reciprocally, mTORC1 activation itself drives v-ATPase expression in cells and mice⁴⁸, forming a positive feedback loop. Accordingly, we found that expressions of multiple v-ATPase subunits were significantly elevated in TFE3-fusion RCC cases in the TCGA KIRP cohort, compared to ccRCC (Fig. 4A). In HK-2 cells with inducible SFPQ-TFE3 expression, V-ATPase subunit expression was similarly elevated by qRT-PCR, whole cell- and lysosomal fractionation- immunoblotting, and down-regulated by mTOR inhibition via Torin1 or RHEB siRNA (Fig. 4B-D, Supplementary Fig. 6B-lower panels). Expression of multiple v-ATPase subunits was also variably elevated in the UOK cells with TFE3-fusions, compared to ccRCC (Fig. 4E). ATP6V0C is an evolutionarily conserved v-ATPase subunit that interacts with Ragulator and controls mTORC1^47,49. We found that siRNA-mediated depletion of ATP6V0C was sufficient to decrease p-4EBP1 in HK-2/SFPQ-TFE3 cells (Fig. 4F). Finally, treatment of HK-2/SFPQ-TFE3 cells with BafilomycinA1 (BafA1)-a chemical v-ATPase inhibitor that binds to the V0C subunit⁵⁰-downregulated phosphorylation of multiple mTORC1 substrates in a dose- and time-dependent manner (Fig. 4G, Supplementary Fig. 6E), further substantiating the v-ATPase complex as a key contributor to mTORC1 activation in HK-2/SFPQ-TFE3 cells.

Although the tumors in our STP model were epithelioid and definitively derived from a PAX8-positive cell of origin due to use of the Pax8-CreERT model³³, the lack of keratin and renal tubular marker expression raised the possibility that these tumors more closely resembled malignant PEComas rather than tRCC. To further characterize the lineage of these tumors, we carried out RNA-seq on 15-day STK and 3.5-month tamoxifen-treated, STP transgenic kidneys, as well as 7-month PTK kidneys, and examined differentially expressed genes compared to their respective normal control kidneys (Supplementary Data 1-3). Notably, the core set of differentially expressed genes overlapping between human tRCC and the ASPSCR1-TFE3 mouse model²⁶ was strikingly enriched in the STP, STK, and PTK transgenic kidney tumors (Supplementary Fig. 7A), highlighting the transcriptional overlap between our mouse models and tRCC. However, gene set enrichment analyses using standard KEGG sets were notable for convergent negative enrichment for the peroxisome and numerous peroxisome-associated metabolic pathways⁵¹ associated with renal tubular epithelial cells in the STK, STP, and PTK models (Supplementary Data 4–6). Consistent with this finding, there was negative enrichment of genes associated with renal epithelial cell subsets from the cell type signature gene sets (C8) in STK and STP, and to a lesser extent in PTK transgenic kidneys, consistent with downregulation of keratins and renal tubular markers in these models (Supplementary Fig. 7B).

To further probe loss of renal tubular cell identity in the STP model, we examined expression of core renal lineage transcription factors in tubular cells with or without SFPQ-TFE3 expression within the same kidney. We used GPNMB expression to identify renal tubules with mosaic SFPQ-TFE3 fusion transgene expression in mice treated for only two weeks with tamoxifen. Strikingly, nuclear PAX8 expression was conspicuously absent in GPNMB+ tubular epithelial cells, in contrast to adjacent GPNMB- neighboring tubular cells (Fig. 5A), indicating that suppression of PAX8 occurs rapidly after SFPQ-TFE3 fusion expression. Concordant results were seen for keratin (CK8) immunostaining at this timepoint, with CK8 loss occurring specifically in cells with fusion gene expression (Supplementary Fig. 8A). After 3.5 months of tamoxifen treatment, the full-blown tumors in the STP model lacked detectable PAX8 (Fig. 5B), consistent with loss of keratin in this model (Fig. 2H). A more heterogeneous pattern of PAX8 and PAX2 suppression was seen in the STK kidneys at P15 (Fig. 5C) as well as the PTK kidney tumors at 7 months age (Fig. 5D) however these differences were highly significant on digital quantification for PAX8 IHC expression (Fig. 5E, F), indicating that both SFPQ-TFE3 and PRCC-TFE3 expression drives varying degrees of renal lineage factor loss in the murine kidney.

The PAX8 transcription factor hub drives a core regulon of target genes in the kidney to promote proximal tubular epithelial cell fate, including GATA3, LHX1 and WT1, among others⁵², and we confirmed these targets in HK2 proximal tubular epithelial cells following PAX8 knockdown using immunoblotting and qRT-PCR (Supplementary Fig. 8B, C). Loss of PAX8 phenocopied expression patterns in renal TSC-related PEComas (angiomyolipomas), which show striking downregulation of PAX8 or PAX2 and their core target genes (GATA3, WT1, LHX1)^36,52 compared to surrounding kidney parenchyma. (Supplementary Fig. 8D). Indeed, expression of GATA3 and pan (type I) keratin was also markedly decreased in STP tumors by 3.5 months after tamoxifen treatment, consistent with loss of PAX8 (Supplementary Fig. 8E) and STK, STP and PTK transgenic kidneys showed significant enrichment of both up- and downregulated genes from renal TSC-related PEComas compared to controls (Fig. 5G)³⁶. Finally, to explore the temporal dynamics of downregulation of PAX8, PAX2 and downstream target genes following SFPQ-TFE3 fusion expression, we leveraged cultured primary renal tubular epithelial cells from ST mice, treated with empty or Cre-recombinase-expressing adenovirus in vitro. By immunoblotting, expressions of PAX8, PAX2 and their transcriptional targets GATA3 and WT1, as well as pan-keratin (Type 1) were decreased within two days of SFPQ-TFE3 induction (Fig. 5H, Supplementary Fig. 8F). Taken together, both SFPQ-TFE3 and PRCC-TFE3 fusion expression are accompanied by downregulation of renal lineage transcription factors and their downstream targets in vivo and in vitro in the murine kidney. This finding is most dramatic in the STP model, where resulting tumors are best classified as renal epithelioid malignant PEComas, based on total loss of PAX8, PAX2, cytokeratin and renal tubular marker expression, with accompanying upregulation of melanocytic and lysosomal markers. Since SFPQ-TFE3 expression is limited to PAX8-expressing cells in the STP model due to use of the Pax8-CreERT model, this model serves as a lineage tracing experiment, thereby substantiating renal tubular epithelial cells as the cell of origin for a TFE3 fusion-driven murine PEComa model.

To validate our murine results in human systems and to further test whether findings were conserved across different TFE3 fusion partners, we next examined expression of PAX8 and PAX2 in the HK2 proximal renal epithelial cell line with doxycycline-inducible expression of common TFE3 fusions. Similar to findings in our mouse models, PAX8 and PAX2 protein expression were most strikingly and specifically downregulated with induction of SFPQ-TFE3 and this was accompanied by robust upregulation of the melanocytic marker PMEL (Fig. 6A). Notably, WT-TFE3 over-expression did not affect PAX8 or PAX2 expression, while PRCC-TFE3 induction was accompanied by a mild suppression of PAX2, without discernable effects on PAX8 expression. PAX8 nuclear localization was also significantly decreased in HK2/SFPQ-TFE3 cells by nuclear-fraction immunoblotting (Fig. 6B) and immunofluorescence (Fig. 6C). These findings were corroborated at the gene expression level, where PMEL and GPNMB expression were dramatically increased upon SFPQ-TFE3 induction (Fig. 6D), and accompanied by decreased mRNA expression of PAX8, PAX2 and the downstream transcriptional targets of these renal lineage transcription factors, including LHX1, HNF1B and GATA3 (Fig. 6E). The downregulation of PAX8 with doxycycline induction of SFPQ-TFE3 was also rescued in HK2/SFPQ-TFE3 cells stably expressing TFE3-targeting CRISPR sgRNAs (Fig. 6F), providing evidence that PAX8 silencing is due to SFPQ-TFE3 expression and not a confounding effect of doxycycline treatment in this system.

Though tRCC is distinguished from PEComa by its retention of detectable keratin and PAX8 expression in clinical practice, the downregulated expression of PAX8 expression in the PTK tRCC model and the frequent underexpression of cytokeratin and EMA expression seen in human tRCC¹⁸ led us to test whether there is partial PAX8 and PAX2 loss in human tRCC samples that had not been previously appreciated. We first examined the series of patient-derived tRCC cell lines, where UOK145 cells expressing SFPQ-TFE3 showed a striking reduction in PAX8 protein expression compared to ccRCC cell lines, but nearly all tRCC cell lines showed some decrease in PAX2 expression compared to the average expression seen in ccRCC lines (Fig. 6G, H). Leveraging the TCGA papillary RCC (KIRP) cohort^35,53, mRNA expression levels of PAX2 and PAX8 were significantly decreased in TFE3 fusion-RCC cases, compared to ccRCC (Fig. 6I), while expression levels of related family members PAX5 and PAX6 remained unaltered (Supplementary Fig. 8G), though sample numbers were too small to examine the effects of specific fusions on PAX8 and PAX2 expression. Cumulatively, our results in human samples indicate that SFPQ-TFE3 expression is associated with particularly potent suppression of renal lineage transcription factor expression compared to other common TFE3 fusion genes, with resulting lineage plasticity towards a PEComa phenotype, while other fusion genes have a milder effect. These findings are entirely consistent with the fact that SFPQ is the most common TFE3 fusion partner in human PEComas and further support the fidelity of current MiT/TFE mouse tumor models, where the PRCC-TFE3 model shows partial retention of PAX8 and epithelial features consistent with tRCC⁴, while our SFPQ-TFE3 model most closely approximates a malignant epithelioid PEComa phenotype.

To investigate potential mechanisms by which SFPQ-TFE3 expression down-regulates PAX8 mRNA expression, we performed chromatin immunoprecipitation sequencing (ChIP-seq) for SFPQ-TFE3 in 2 independent clones of HK2/SFPQ-TFE3 cells. HOMER motif analysis revealed an expected enrichment in CA[C/T] [G/A]TG sequences consistent with the E-box/TFE3-binding motif (Supplementary Data 7–10). Additionally, several de novo motifs (e.g. TGA[G/C] TCA) were also enriched, indicative of potential binding by TFs like Jun, AP1, Fosl2, etc, or RUNX1 as recently described⁵⁴. Notably, we observed strong binding peaks of HA-SFPQ-TFE3 within intragenic and upstream transcription start site regions of the PAX8 gene (Supplementary Fig. 8H, Supplementary Data 11–12), with identical peaks observed for both clones (e.g. 22831 in PSF6_8_12 and 14370 in PSF6_9_9). ChIP-qPCR using primers specific for the indicated peaks confirmed binding of SFPQ-TFE3 to these upstream and intragenic regions of PAX8 following doxycycline treatment in HK2/SFPQ-TFE3 cells (Supplementary Fig. 8I). These SFPQ-TFE3 binding peaks within PAX8 corresponded to histone modification marks associated with enhancer activity [e.g. H3K27ac binding tracks] in nephron organoids from ENCODE data (Supplementary Fig. 8H), suggesting SFPQ-TFE3 binding at regulatory regions within PAX8 that could impact its expression.

Next, we examined whether mTOR signaling activity is required for renal tubular cell lineage plasticity downstream of SFPQ-TFE3 fusion expression. Leveraging our in vitro models for pharmacologic and genetic mTOR modulation, we first asked whether suppression of mTOR signaling concurrent with induction of fusion TFE3 expression in HK2/SFPQ-TFE3 cells affected PAX2/PAX8 expression or MiT/TFE transcriptional target gene expression. Treatment of HK2/SFPQ-TFE3 cells with the mTOR kinase inhibitor Torin1 rescued PAX2 and PAX8 protein expression in a dose-dependent fashion and simultaneously downregulated expression of the PEComa marker PMEL, as well as multiple MiT/TFE-regulated proteins (LC3A/B, RRAGD, RAB7A, and GPNMB), by immunoblotting (Fig. 7A, B, Supplementary Fig. 9A, B). These effects were reproduced by pre-treatment of HK2/SFPQ-TFE3 cells with RHEB siRNA to knock down a key component of the mTOR complex prior to doxycycline induction (Fig. 7C, D, Supplementary Fig. 9B), providing genetic confirmation. SFPQ-TFE3-induced downregulation of keratin expression was also completely rescued by Torin1 and RHEB siRNA (Supplementary Fig. 9B). Strikingly, mTOR inhibition concurrent with doxycycline induction via either Torin1 or RHEB siRNA was associated with significantly lower SFPQ-TFE3 fusion protein expression, and this was particularly evident for RHEB siRNA (Fig. 7A–D, Supplementary Fig. 9A, B) and mildly dose-dependent for Torin1 treatment (Supplementary Fig. 9A). Accordingly, mTORC1 inhibition with Torin1 rescued PAX8 nuclear localization in doxycycline-treated HK2/SFPQ-TFE3 cells, by nuclear-fraction immunoblotting (Fig. 7E) and immunofluorescence (Fig. 7F), with both experiments demonstrating a concurrent decrease in nuclear SFPQ-TFE3 levels. Doxycycline-mediated induction of SFPQ-TFE3 in HK2 cells resulted in a change in morphology from a cobblestone, epithelial pattern to a more mesenchymal appearance, which was also completely reversed on mTOR inhibition (Supplementary Fig. 9C). The dose-dependent decrease in TFE3 fusion protein expression upon Torin1 and RHEB siRNA treatment was generalizable to other cell line systems and TFE3 fusion partners examined by immunoblotting, including HEK293 cells with doxycycline-inducible expression of SFPQ-TFE3 (Supplementary Fig. 9D) or PRCC-TFE3 (Supplementary Fig. 9E) as well as UOK120 and UOK124 cells (Supplementary Fig. 9F) constitutively expressing PRCC-TFE3. At the gene expression level, Torin1 treatment of HK2/SFPQ-TFE3 cells downregulated MiT/TFE transcriptional targets (GPNMB and PMEL) (Fig. 7G), rescued PAX2 and PAX8 gene expression and also upregulated expression of PAX2/8-regulated transcripts (LHX1, GATA3, HNF1B), by qRT-PCR (Fig. 7H). These findings are consistent with the fact that PAX2/8 regulation by SFPQ-TFE3 is mediated at the gene expression level, downstream of elevated mTOR activity.

Importantly, genetic or pharmacological inhibition of the v-ATPase with ATP6V0C siRNA (Fig. 7I), or BafilomycinA1 (Supplementary Fig. 9G, H) markedly decreased SFPQ-TFE3 fusion protein expression and rescued PAX8 expression (Fig. 7I), consistent with potent mTORC1 suppression exerted by v-ATPase inhibition (Fig. 4F, G, Supplementary Fig. 6E). In contrast, transient knockdown (Supplementary Fig. 5H), or expression of inactive mutants of RRAGC and/or RRAGD (Supplementary Fig. 6A), had no effect on SFPQ-TFE3 or PAX8 expression, indicative of a more modest impact of RRAGC/D silencing on mTORC1 signaling in this context. Taken together, these findings suggest that inhibition of v-ATPase expression and/or activity, but not RRAGC/D expression or activity, is sufficient to disrupt the reciprocal SFPQ-TFE3-mTORC1 feedback loop and restore PAX8 expression in the context of inducible SFPQ-TFE3 expression.

We next examined the mechanisms by which mTORC1 signaling regulates SFPQ-TFE3 fusion expression. Treatment of HK2/SFPQ-TFE3 cells with the translation inhibitor cycloheximide revealed similar SFPQ-TFE3 degradation kinetics in vehicle- or torin1-treated cells, suggesting that the decrease in SFPQ-TFE3 expression with mTORC1 inhibition was not due to decreased protein stability or increased degradation (Supplementary Fig. 10A, B). However, we further investigated this hypothesis in additional experiments. Since mTORC1 inhibition may increase autophagic degradation of SFPQ-TFE3 protein, we also employed a fluorescent spectrophotometric approach to quantitatively estimate autophagic flux or intracellular Cathepsin B activity in vitro. Both CTSB activity (Supplementary Fig. 10C) and Cyto-ID fluorescence (Supplementary Fig. 10D), were expectedly elevated upon induction of SFPQ-TFE3, consistent with increased lipidated LC3-II observed on immunoblotting (Supplementary Fig. 9B). mTOR inhibition with Torin1 or RHEB siRNA actually decreased LC3-II (likely due to downregulation of SFPQ-TFE3-associated LC3 gene expression/LC3-I protein expression), and neither CTSB activity nor Cyto-ID fluorescence were altered by mTORC1 inhibition. Accordingly, blocking autophagic flux with concurrent treatment with BafilomycinA1 did not reverse the Torin1-induced decrease in SFPQ-TFE3 protein levels (Supplementary Fig. 10E). Co-immunoprecipitation-ubiquitination analyses of HK2/SFPQ-TFE3 cell lysates showed a slight decrease in Ub chains associating with immunoprecipitated SFPQ-TFE3 following Torin1 or RHEB siRNA treatment, proportionate to the decrease in total SFPQ-TFE3 protein expression with mTORC1 inhibition (Supplementary Fig. 10F, right panels), and concurrent proteasomal inhibition with MG132 or Bortezomib also did not reverse the Torin1-induced decrease in SFPQ-TFE3 protein levels (Supplementary Fig. 10G). Given the lack of evidence for altered fusion protein stability or degradation with mTOR inhibition, we then examined impacts on fusion gene transcription. Significantly, mTORC1 inhibition with Torin1 or RHEB siRNA reduced SFPQ-TFE3 mRNA levels in doxycycline-induced, HK2/SFPQ-TFE3 cells (Supplementary Fig. 10H) and also decreased PRCC-TFE3 mRNA expression in UOK124 cells (Supplementary Fig. 10I), by qRT-PCR. Cumulatively, these results suggested that mTORC1 signaling most likely regulates SFPQ-TFE3 fusion expression via impacts on fusion gene expression levels.

Finally, we examined whether mTOR inhibition via Torin1 reduced kidney tumor burden in STP mice. Following induction of SFPQ-TFE3 expression with tamoxifen, we treated male and female cohorts of mice with vehicle or Torin1 for 4 weeks, following which kidney FFPE sections were immunostained for p-S6 to confirm mTORC1 inhibition (Supplementary Fig. 11A) and GPNMB to quantify tumor area in digital whole slide images of kidney (Supplementary Fig. 11B). While Torin1 treatment did not significantly impact renal tumor burden in female STP mice, there was a small and borderline significant reduction in tumor area in male STP kidneys, which show a lower tumor burden overall to their female counterparts (Supplementary Fig. 11C). Taken together, our findings indicate that SFPQ-TFE3 fusion expression is sufficient to induce renal tumorigenesis in mouse models, driving V-ATPase expression and concomitant downstream mTORC1 signaling which, in turn, positively and reciprocally increases SFPQ-TFE3 gene expression levels. This positive feedback cycle reinforces a lineage switch in renal tubular cells, driving cells towards a PEComa cell phenotype characterized by reduced transcription of renal lineage transcription factors. Inhibition of mTOR signaling interrupts this positive feedback, and by reducing SFPQ-TFE3 fusion gene expression levels, indirectly rescues PAX8 expression and restores renal lineage identify in vitro but requiring additional work to substantiate in vivo impact (Fig. 8).