p53 mitigates the effects of oncogenic HRAS in urothelial cells via the repression of MCOLN1.

Inhibition of TRPML1, which is encoded by MCOLN1, is known to deter cell proliferation in various malignancies. Here, we report that the tumor suppressor, p53, represses MCOLN1 in the urothelium such that either the constitutive loss or ectopic knockdown of TP53-in both healthy and bladder cancer cells-increased MCOLN1 expression. Conversely, nutlin-mediated activation of p53 led to the repression of MCOLN1. Elevated MCOLN1 expression in p53-deficient cancer cells, though not sufficient for bolstering proliferation, augmented the effects of oncogenic HRAS on proliferation, cytokine production, and invasion. Our data suggest that owing to derepression of MCOLN1, urothelial cells lacking p53 are poised for tumorigenesis driven by oncogenic HRAS. Given our prior findings that HRAS mutations predict addiction to TRPML1, this study points to the utility of TRPML1 inhibitors for mitigating the growth of a subset of urothelial tumors that lack p53.

Introduction

By coordinating the synthesis and breakdown of macromolecules with signaling events that regulate metabolism, endolysosomes play vital roles in cellular homeostasis (Xu and Ren, 2015). While insufficient endolysosomal function is associated with inborn errors of metabolism, lysosomal storage, and degenerative diseases, increased biogenesis of these organelles is a recurring theme in cancers (Davidson and Vander Heiden, 2017). Several studies have described roles for the endolysosomal cation channel, TRPML1, in different malignancies (Hu et al., 2019; Jung et al., 2019; Kasitinon et al., 2019; Xu et al., 2019; Yin et al., 2019). A common theme in these studies is that TRPML1 is necessary for malignancy-promoting signaling cascades involving RAS-MAPK and mTORC1. For instance, HRAS-driven cancers exhibit increased TRPML1 abundance and activity, which underlies endolysosomal exocytosis necessary for recycling cholesterol-rich membranes and oncogenic HRAS to the plasma membrane (Jung and Venkatachalam, 2019; Jung et al., 2019). As a result, TRPML1 inhibition or MCOLN1 knockdown disrupts ERK activation and cell proliferation in HRAS-driven cancer cells. TRPML1-dependent vesicle exocytosis is also necessary for the release of lysosomal ATP, which in the case of triple-negative breast cancers, stimulates cell migration and metastases (Xu et al., 2019). In other cancers, TRPML1 promotes untrammeled proliferation and tumor growth via mTORC1 (Hu et al., 2019; Xu et al., 2019).

These studies make a compelling argument for understanding the pathways that govern MCOLN1 expression in cancers. Of major significance here are TFEB, TFE3, and MITF—transcription factors that induce the expression of MCOLN1 and other endolysosomal genes harboring “coordinated lysosomal expression and regulation” (CLEAR) motifs (Martina et al., 2014; Palmieri et al., 2011; Ploper et al., 2015; Sardiello et al., 2009; Settembre et al., 2011, 2012). Signaling events that influence the nucleocytoplasmic distribution of these transcription factors function as de facto regulators of endolysosomal biogenesis (Martina et al., 2012; Medina et al., 2015; Palmieri et al., 2017; Roczniak-Ferguson et al., 2012; Settembre et al., 2011). Several cancers are characterized by either the constitutive localization of TFEB/TFE3/MITF in cell nuclei or by amplifications that upregulate their expression (Argani, 2015; Blessing et al., 2017; Calcagnì et al., 2016; Kauffman et al., 2014; Kundu et al., 2018; Li et al., 2019; Perera et al., 2015, 2019; Ploper et al., 2015). The attendant potentiation of endolysosomal biogenesis supports cancer cell survival, proliferation, and metastases. Because inhibition of TRPML1 arrests proliferation or induces cell death in cancers with elevated expression of the CLEAR network genes (Hu et al., 2019; Jung and Venkatachalam, 2019; Jung et al., 2019), tumors with hyperactivated TFEB/TFE3/MITF signaling are addicted to TRPML1.

Despite all this progress, many questions persist. For instance, we still do not fully understand how the CLEAR network and MCOLN1 expression are pathologically augmented in cancer. In healthy cells, feedback mechanisms keep endolysosomal biogenesis in check. For instance, mTORC1 suppresses endolysosomal biogenesis by driving phosphorylation and cytosolic retention of TFEB and TFE3 (Martina et al., 2012, 2014; Roczniak-Ferguson et al., 2012; Wang et al., 2015), which is why high mTORC1 activity is correlated with repression of the CLEAR network. In cancers, however, these axes are often dysregulated—a notion exemplified by KRAS-driven pancreatic cancer cells, which exhibit hyperactivation of both mTORC1- and TFEB-driven gene expression owing to failure of the mutually inhibitory mechanisms linking these entities (Perera et al., 2015). In addition, genomic landscapes of cancer cells encourage qualitative shifts in the regulatory pathways that control endolysosomal biogenesis. Signaling that normally curtails endolysosomal gene expression morphs into drivers of vesicular biogenesis in cancer. For example, while the tumor suppressor, p53, activates TFEB and TFE3 in normal fibroblasts exposed to DNA damage, loss of p53 in cancers is also associated with the paradoxical activation of the TFEB/TFE3 endolysosome axis (Brady et al., 2018; Tasdemir et al., 2008a, 2008b; Zhang et al., 2017). The lack of clear insights into the regulation of TFEB/TFE3-driven endolysosomal biogenesis hinders our ability to exploit TRPML1 addiction as a therapeutic strategy. To fill the gaps in knowledge, we surveyed MCOLN1 expression in different cancers using the Cancer Genome Atlas (TCGA) (Cancer Genome Atlas Research Network, 2014). This analysis prompted a focus on bladder carcinoma (BLCA) (Robertson et al., 2017), in which primary tumors exhibited significant elevations in MCOLN1 expression. Further investigation revealed a role for p53 in repressing TFEB-driven MCOLN1 expression. Therefore, loss of p53 augmented TRPML1 abundance, which in turn fostered cell proliferation, inflammation, and invasion stemming from oncogenic HRAS. Our study uncovers an axis by which MCOLN1 expression is regulated and suggests that TRPML1 inhibitors could mitigate tumorigenesis in p53-deficient bladder cancers.

Results

Expression of MCOLN1 is inversely correlated with p53 targets in bladder cancer

By comparing mRNA levels in tumors relative to matched normal tissues, we found that MCOLN1 expression was elevated in the TCGA BLCA data set (log2FC = 0.5, FDR = 0.001; Figure 1A and Table S1), which prompted us to select this disease as a suitable model for the identification of cancer-related pathways that depend on MCOLN1 induction. We also reasoned that ontologies of genes whose transcription correlates with MCOLN1 would reveal the pathways that depend on MCOLN1 expression. In BLCA, MCOLN1 exhibited significant positive and negative correlation with 4737 and 3611 genes, respectively (Figure 1B and Table S2). Targeted gene set enrichment analysis (GSEA) (Subramanian et al., 2005) using the correlation coefficients revealed the expected enrichment of CLEAR targets in genes that are positively correlated with MCOLN1 (Figure 1C) (Palmieri et al., 2011). Likewise, upon probing the correlation coefficients against MSigDB, we found that genes that positively correlated with MCOLN1 exhibited enrichment for modules related to lysosomes and lytic vacuoles, endocytosis and phagocytosis, and vesicular exocytosis and secretion (Figures 1D and S1). In addition, MCOLN1 expression was positively correlated with genes that are upregulated during ultraviolet (UV)-induced DNA repair and negatively correlated with p53 target genes and genes that are repressed during UV-induced DNA repair (Figures 1D and S1).

Figure 1

Pathways correlated with MCOLN1 expression in BLCA

(A) Volcano plot showing DEGs in BLCA primary tumors relative to matched normal tissues. Genes that were significantly (FDR<0.05) upregulated and downregulated in tumors are depicted by blue and red dots, respectively. MCOLN1 (light blue circle) belonged to the set of significantly upregulated genes. Gray dots represent genes whose expression was not significantly altered in tumors.

(B) Plot showing correlation of 20,177 genes with MCOLN1. Values on the Y axis represent Spearman's correlation. Genes that showed significant (FDR <0.05) positive and negative correlation with MCOLN1 are represented by magenta and green dots, respectively. Gray dots represent genes that were not significantly correlated with MCOLN1.

(C) GSEA shows enrichment for CLEAR targets in genes that showed significant positive correlation with MCOLN1.

(D) Functional annotation of genes that were significantly correlated with MCOLN1 in BLCA. All the genes that were significantly correlated with MCOLN1 were subjected to GSEA. Sets with GSEA FDR<0.05 are shown as nodes. Node color represents NES from GSEA. Node size represents number of genes that make up that set. Thickness of the lines connecting the nodes indicates extent of overlap. Unconnected nodes have no overlap. Abbreviations: FC, fold change; NES, normalized enrichment score.

MCOLN1 expression was elevated in primary BLCA tumors with TP53 mutations

Next, we applied the principles of information theory (Shannon, 1948) to determine whether mutations in any of the 722 genes belonging to the Cancer Gene Census (https://cancer.sanger.ac.uk/census; Table S3) correlated with MCOLN1 expression in BLCA. We sought to identify those genes that were mutated in tumors with either high or low MCOLN1 expression, such that partitioning the set of tumors on the basis of MCOLN1 expression would decrease the stochasticity associated with the appearance of mutations (i.e., an increase in “Shannon information,” see STAR methods). We found that 145 genes exhibited significant information gain upon partitioning the tumors on the basis of MCOLN1 expression (Figure 2A and Table S4). Of these genes, only four (TP53, KDM6A, ARID1A, RB1) were mutated in ≥15 tumors (Figure 2A). To validate further the relationship between MCOLN1 expression and mutations in these four genes, we performed targeted GSEA by ranking BLCA tumors on the basis of MCOLN1 expression. This analysis revealed that mutations in TP53 (TP53) and RB1 (RB1) were significantly enriched in tumors with highest MCOLN1 expression, whereas wild-type alleles of both genes (TP53 and RB1) were enriched in tumors with lowest MCOLN1 expression (Figure 2B). In agreement, pairwise comparison of normalized MCOLN1 FPKM revealed significantly higher values in TP53 and RB1 in comparison with TP53 and RB1 tumors, respectively (Figure 2C). In contrast, neither were the mutations in ARID1A (ARID1A) or KDM6A (KDM6A) enriched in tumors with either the highest or lowest MCOLN1 expression nor was MCOLN1 expression significantly altered in ARID1A or KDM6A tumors relative to tumors with wild-type alleles of these two genes (ARID1A or KDM6A, respectively) (Figures 2B and 2C). Together, these data point to elevated MCOLN1 expression in tumors harboring TP53 and RB1 mutations. Because tumors with RB1 mutations belonged to the larger set with TP53 mutations (Figure S2), we conclude that tumors harboring TP53 mutations constitute the broadest set of samples with elevated MCOLN1 expression.

Figure 2

MCOLN1 expression is elevated in TP53 BLCA tumors

(A) Waterfall plot showing mutations in 30 genes with significant information gain. MCOLN1 expression is depicted as median normalized FPKM values at the top. Bars mark tumors harboring mutations in respective genes, and bar colors represent the type of mutation. On the right, Shannon information associated with partitioning tumors into three groups based on MCOLN1 expression are show; and “freq.” depicts number of tumors in the BLCA data set that exhibit mutations in the respective genes. For all the genes shown, Q < 0.0001.

(B) GSEA traces show enrichment for indicated mutations in tumors ranked on the basis of MCOLN1 expression (represented as median normalized FPKM values).

(C) Box plots showing MCOLN1 expression (represented as median normalized FPKM values) in BLCA sorted on the basis of the presence or absence of mutations in the indicated genes. Each dot represents the values from a different tumor and lines in the box plots represent median values. ∗∗, p < 0.01, t test; ∗∗∗∗, p < 0.0001, Mann-Whitney tests. Abbreviation: NES, normalized enrichment score; q, FDR value; n.s., not significant.

Approximately 47% of BLCA tumors in the TCGA data set were TP53. Using the IARC database (https://p53.iarc.fr/Default.aspx) (Bouaoun et al., 2016), we found that among the TP53 BLCA tumors, ∼86% harbored TP53 mutations (TP53) that predicted to result in loss of transactivation function owing to either nonsense mutations leading to premature termination or missense mutations in the DNA-binding domain. Among the remainder, ∼7% harbored TP53 mutations (TP53) that were either unlikely to affect transactivation or have remained unclassified. Another ∼3% carried TP53 mutations in splice acceptor or donor sites (TP53), and ∼3% harbored multiple TP53 mutations. Finally, one tumor harbored an inframe mutation. In sum, majority of TP53 mutations in BLCA were predicted to result in loss of p53 transactivation. Analyses of the RNA sequencing data revealed that in comparison to the values inTP53 tumors, MCOLN1 expression was significantly higher in TP53 tumors, but not in TP53 or TP53 tumors (Figure S3A). Furthermore, targeted GSEA revealed that TP53 mutations were significantly enriched in tumors with high MCOLN1 expression (Figure S3B). Therefore, MCOLN1 expression was elevated in tumors that harbored mutations in the DNA-binding domain of p53.

MCOLN1 expression is elevated in p53-deficient bladder cancer cell lines

Next, we sought to examine the relationship between p53 and MCOLN1 expression in both bladder cancer and healthy urothelial cells. We made use of five different bladder cancer cell lines—HT1197, RT4, SW780, 5637, and T24 (Bubeník et al., 1973; Fogh, 1978; Rasheed et al., 1977; Rigby and Franks, 1970). As per the Cancer Cell Line Encyclopedia (CCLE), HT1197, RT4, and SW780 cells are wild type for TP53, whereas 5637 cells harbor a missense TP53 mutation that is predicted to encode a transactivation-deficient (R280T) variant, and T24 cells are homozygous for a nonsense mutation in the TP53-coding sequence that is predicted to yield a null allele (Table S5). In agreement with CCLE, we detected p53 protein in extracts from HT1197, RT4, SW780, and 5637 cells, but not in extracts from T24 cells (Figures S4A and S4B). Interestingly, p53 abundance was higher in 5637 cells relative to RT4 and SW780 (Figure S4A). Application of the p53-stabilizing agent, nutlin (Vassilev et al., 2004), led to a significant increase in p53 protein in RT4 cells but a relatively smaller increase in 5637 cells (Figure S4A). These data suggest that basally higher p53 in 5637 cells involves compensatory upregulation of the protein via the Mdm2–p53 axis targeted by nutlin (Vassilev et al., 2004). Using RT-PCR, we found that in comparison to the values in HT1197, RT4, SW780, or 5637, MCOLN1 expression was significantly higher in T24 cells (Figure 3A). Therefore, loss of p53 in bladder cancer cells was associated with higher MCOLN1 expression.

Figure 3

p53 has a necessary and sufficient role in repressing MCOLN1 in both transformed and healthy urothelial cells

(A) Bar graph showing relative MCOLN1 expression in the indicated bladder cancer cell lines treated with control siRNA or siRNA against TP53 (75 nM) for either 2 or 5 days. For each cell type values were normalized to HT1197 control mean. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗∗, p < 0.0001, t-tests with Bonferroni correction in case of samples used in multiple pairwise comparisons.

(B) Bar graph showing relative MCOLN1 expression in the indicated bladder cancer cell lines with or without treatment with 10 μM Nutlin for 48 h. For each cell line, values were normalized to cell numbers in the untreated cohorts. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗∗∗∗, p < 0.0001,; n.s., not significant, t-tests.

(C) Same as in (A) but in healthy urothelial cells.

(D) Same as in (B) but in healthy urothelial cells.

(E) Same as in (A) but with or without addition of TP53 siRNA (TP53 KD) and expression of HRAS as indicated below the graph.

(F) Representative Western blot showing p53 and tubulin abundance in extracts generated from the cell lines indicated above the blot that were treated as indicated below the blot.

(G) Quantification of the data shown in (F). Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗, p < 0.05; ∗∗, p < 0.01; n.s., not significant, t-tests with Bonferroni correction in case of samples used in multiple pairwise comparisons.

To analyze further the regulation of MCOLN1 expression by p53 in the bladder cancer cells, we first examined the consequences of knocking down of TP53 expression. Application of TP53 siRNA (Xu et al., 2009), which elicited the expected decrease in p53 protein abundance in HT1197, RT4, SW780, or 5637 cells (Figures S4B and S4C), significantly elevated MCOLN1 expression in those cells (Figure 3A). Conversely, activation of p53 by application of nutlin (Vassilev et al., 2004) decreased MCOLN1 expression in HT1197, SW780, and RT4 cells (Figure 3B), all of which carry wild-type alleles of TP53. Nutlin did not repress MCOLN1 in the p53-deficient, T24 cells (Figure 3B), which indicates that nutlin lowered MCOLN1 mRNA via p53. Consistent with TFEB being a transcriptional activator of MCOLN1 (Sardiello et al., 2009), knockdown of TFEB in T24 cells restored MCOLN1 expression to control levels (Figure S4D). Furthermore, simultaneous knockdown of TFEB dampened the extent of MCOLN1 induction stemming from TP53 knockdown in RT4 cells (Figure S4E). We also asked whether the role for p53 in MCOLN1 repression is an exclusive feature of transformed cells or whether p53 restrict MCOLN1 expression even in healthy, noncancerous urothelial cells. Examination of healthy urothelial cells revealed that knockdown of TP53, which led to the expected decrease in p53 abundance (Figures S4B and S4C), augmented MCOLN1 expression in those cells (Figure 3C). Furthermore, application of nutlin lowered MCOLN1 expression in the healthy urothelial cells (Figure 3D). Taken together, our data indicate that p53 is necessary and sufficient to regulate MCOLN1 expression in both healthy urothelial and transformed bladder cancer cells.

The human genome encodes two other MCOLN1 paralogs called MCOLN2 and MCOLN3 (Karacsonyi et al., 2007; Di Palma et al., 2002). Interestingly, both MCOLN2 and MCOLN3 exhibited increased expression in T24 cells compared with HT1197, RT4, and SW780 cells (Figure S4F). Arguing against a role for p53 in the regulation of the MCOLN1 paralogs, however, was our finding that knockdown of TP53 did not alter MCOLN2 expression in HT1197, RT4, and SW780 cells (Figure S4F). MCOLN3 expression, though insensitive to TP53 knockdown in HT1197 and RT4 cells, was slightly elevated upon application of TP53 siRNA in SW780 cells (Figure S4F). On the whole, these data indicate that while MCOLN1 expression in bladder cancer cells was clearly dependent on p53, expression of MCOLN2 and MCOLN3 were relatively insensitive to TP53 knockdown.

Forced expression of oncogenic HRAS lowered p53 abundance in bladder cancer cells

We previously demonstrated the existence of an evolutionarily conserved axis by which MCOLN1 expression was elevated in cells expressing oncogenic HRAS (Jung and Venkatachalam, 2019; Jung et al., 2019). In agreement with our previous findings, forced expression of HRAS by lentiviral transduction in HT1197, RT4, SW780, and 5637 cells increased MCOLN1 expression to extents comparable with those elicited by TP53 knockdown (Figures 3E and S4G). Knockdown of TP53 in cells that ectopically expressed HRAS did not increase further the levels of MCOLN1 mRNA (Figure 3E). The lack of additivity raised the possibility that HRASG12V and TP53 knockdown increased MCOLN1 expression via the same pathway—a notion consistent with the findings that activating mutations in HRAS lead to p53 destabilization (Ries et al., 2000). In agreement, p53 abundance was significantly decreased in bladder cancer cells that ectopically expressed HRAS (Figures 3F and 3G). Because the decrease in p53 induced by HRASG12V was comparable with that elicited by TP53 knockdown, our data are consistent with the notion that HRASG12V promotes MCOLN1 expression by lowering p53 abundance.

Roles for TRPML1 in proliferation and prevention of cell cycle arrest in p53-deficient bladder cancer cells

Next, we identified the genes whose expression was significantly different in TP53 tumors relative to those that were TP53. At FDR<0.05, we identified 5659 differentially expressed genes (DEGs). MCOLN1 belonged to the cohort of significantly upregulated genes in TP53 tumors, whereas expressions of both TP53 and RB1 were significantly diminished (Figure 4A and Table S6). GSEA identified several functional modules in the DEGs including those related to positive regulation of cell cycle progression and proliferation and immunity and inflammation (Figure 4B and Table S7). Additional upregulated modules comprised gene sets for cell adhesion, RNA processing, and response to zinc (Figure 4B and Table S7). Downregulated modules involved those related to detoxification/drug metabolism via cytochrome p450, hormone signaling, lipase activity, ribosome and protein translation, and solute pumps (Figure 4B and Table S7).

Figure 4

Transcriptional alterations in BLCA tumors harboring TP53 mutations

(A) Volcano plot showing DEGs in TP53 compared with TP53 tumors. Genes that were significantly (FDR <0.05) upregulated and downregulated in TP53 tumors are depicted by magenta and blue dots, respectively. MCOLN1 belonged to the set of genes upregulated genes, whereas TP53 and RB1 belonged to the set of downregulated genes. Gray dots represent genes whose expression was not significantly altered.

(B) Functional annotation of DEGs in TP53 relative to TP53 BLCA tumors. All DEGs were subjected to GSEA. From the results of this analysis, only the gene sets with GSEA FDR <0.05 are shown as nodes. Node color represents NES from GSEA. Node size represents number of genes that make up that set. Thickness of the lines connecting the nodes indicates extent of overlap. Unconnected nodes have no overlap. Abbreviations: FC, fold change.

Because modules related to cell cycle progression were upregulated in TP53 tumors, we asked whether TRPML1 might be necessary for the proliferation of bladder cancer cells lacking p53. In agreement with our previous findings (Jung et al., 2019), the MTT cell proliferation assay revealed that pharmacological inhibition of TRPML1 by application of ML-SI1 (Samie et al., 2013) for 2 days significantly attenuated T24 cell proliferation (Figure 5A). HT1197, RT4, and SW780 cells were much less sensitive to TRPML1 inhibition (Figure 5A). Extending the duration of ML-SI1 treatment to 5 days led to even greater attenuation of T24 cell numbers compared with the effect of the drug on HT1197, RT4, or SW780 cells (Figure 5A). To rule out the possibility that the antiproliferative effects of ML-SI1 reflect a specific influence on the MTT assay, we assessed colony formation. ML-SI1 application led to significantly greater reduction in the number of colonies in T24 relative to HT1197 (Figure 5B). In agreement with our findings using ML-SI1, MCOLN1 knockdown curtailed the proliferation of T24 cells to a significantly greater extent than it did in HT1197, RT4, SW780, or 5637 cells (Figure 5C). Ruling out the possibility that the efficacy of MCOLN1 siRNA varies between the cell types, extent of MCOLN1 knockdown in HT1197, RT4, and 5637 cells (∼50%, Figure S5A) was in the same range as that elicited by this siRNA in T24 cells (Jung et al., 2019).

Figure 5

TRPML1 is required for the proliferation of T24, but not HT1197, RT4, SW780 or 5637 bladder cancer cells

(A) Bar graph showing relative number of the indicated cells assessed using the MTT cell proliferation assay. Numbers at the bottom represent duration of 10 μM ML-SI1 application in days. Circles represent independent biological repeats and the values shown represent mean ± SEM.

(B) Left, representative images of the indicated cell lines treated with DMSO or 10 μM ML-SI1 for 5-days. Cells grown on the dished were stained with crystal. Right, bar graph showing intensity of crystal violet staining in ML-SI1 treated cells normalized to the values in untreated cells. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗∗∗∗, p < 0.0001, t test.

(C) Same as (A) but with 200 nM MCOLN1 siRNA. ∗∗∗∗, p < 0.0001, t test with Bonferroni tests in case of samples used in multiple pairwise comparisons.

(D) Bar graph showing fraction of T24 cells in the G0/G1 phase of the cell cycle in response to the indicated treatments for the indicated durations. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗∗∗, p < 0.001, t-tests.

(E) Same as (D) but showing BrdU-labeled cells. ∗∗, p < 0.01; ∗∗∗, p < 0.001, t-tests. Drug concentrations in (D) and (E), 200 nM MCOLN1 siRNA and 10 μM ML-SI1.

(F) Same as in (A) but in the indicated cell lines for the indicated drugs and for the indicated durations. ∗∗, p < 0.01; ∗∗∗∗, p < 0.0001, t-tests. Concentrations, 10 μM cisplatin and 10 μM ML-SI1. Abbreviations: n.s., not significant.

Analysis of the fractions of T24 cells in various stages of cell cycle revealed that MCOLN1 knockdown or TRPML1 inhibition led to a significant increase in the fraction of cells in the G0/G1 phase of the cell cycle (Figure 5D). Accumulation of cells in G0/G1 was evident within 2 days of treatment and persisted at the 5-day mark (Figure 5D). The larger fraction of cells in G0/G1 was accompanied with a decline in the fraction of BrdU-labeled cells (Figure 5E) and cells in the G2/M phases (Figure S5B). Therefore, the decline in T24 cancer cell numbers upon the knockdown or inhibition of TRPML1 was a consequence of accumulation in the G0/G1 phase of the cell cycle with an attendant decline in the fractions in S and G2/M phases.

Because ML-SI1 slowed progression through the cell cycle, we did not anticipate the drug to synergize with a cytotoxic agent. Indeed, cisplatin decreased the number of T24 and RT4 cells to similar extents (Figure 5F). Treatment with a combination of ML-SI1 and cisplatin led to a further decline in T24, but not RT4 cell numbers (Figure 5F). Normalization to the effects of ML-SI1 showed that sensitivity toward cisplatin treatment was not synergistically enhanced by TRPML1 inhibition (Figure S5C). Neither did cisplatin change the expression of MCOLN1 (Figure S5D). Thus, ML-SI1 and cisplatin exert their effects on cancer cell number via distinct pathways leading to an additive effect upon simultaneous application of both drugs.

TRPML1-dependent cytokine production regulates cell proliferation, cell invasion, and immune microenvironment in bladder cancers

Given the induction of MCOLN1 and inflammation in TP53 BLCA tumors (Figure 4), we asked whether TRPML1 could be participating in cytokine gene expression. We focused our attention on genes encoding IL6 and TNFα, which are involved in various aspects of tumorigenesis (Caetano et al., 2016; Mantovani et al., 2017; Wu and Zhou, 2010). Suggesting a relationship between MCOLN1 and cytokine gene expression, top half of the tumors in the TCGA BLCA dataset, sorted on the basis of MCOLN1 expression, also expressed significantly higher levels of TNF and IL6 than did the remainder of the tumors (Figure S6A). In agreement, T24 cells expressed significantly higher levels of IL6 and TNF than did HT1197, RT4, and SW780 cells (Figure 6A). Knockdown of RELA, which encodes the p65 subunit of NF-κB (Nolan et al., 1991), partially attenuated IL6 and TNF transcription in all cell types except SW780 (Figure 6A). Notably, application of ML-SI1 or MCOLN1 knockdown mitigated the expression of the two cytokines in T24 cells to the levels that were typical of HT1197, RT4, and SW780 cells (Figure 6A). These data indicate that cytokine gene expression in p53 deficient bladder cancer cells requires functional NF-κB and TRPML1.

Figure 6

TRPML1 is required for cytokine gene expression in T24, but not HT1197, RT4, SW780, or 5637 bladder cancer cells

(A) Bar graph showing relative cytokine expression in the indicated cell lines treated as described. All values were normalized to the HT1197 mean. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗∗, p < 0.01, ∗∗∗, p < 0.001, ∗∗∗∗, p < 0.0001, t-tests with Bonferroni corrections in case of samples used in multiple pairwise comparisons. Concentrations, 50 nM RELA siRNA, 200 nM MCOLN1 siRNA, and 10 μM ML-SI1.

(B) Bar graph showing relative cell numbers in response to 40 μg/mL etanercept with or without ML-SI1 as indicated at the bottom. All values were normalized to the cell numbers obtained in the absence of etanercept. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗∗∗, p < 0.001, t test.

(C) Left, representative images of crystal violet stained cells from the indicated type in Matrigel. Images in the bottom row are magnifications of the regions depicted by boxes shown in the top row. T24 cells were treated with control siRNA or 200 nM MCOLN1 siRNA. Right, bar graph showing the number of invading cells relative to the RT4 mean. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗∗, p < 0.01, ANOVA.

(D) Relative fractions of the indicated immune cell types in BLCA tumors sorted on the basis of MCOLN1 expression (FPKM values). Fractions of immune cells were determined using CIBERSORT. ∗, FDR <0.05, ∗∗, FDR <0.01, Wilcoxon ranked test followed by two-stage step-up method of Benjamini, Krieger, and Yakutieli. Abbreviations: n.s., not significant.

To assess the influence of TNFα on the bladder cancer cell proliferation, we applied a TNFα chelating drug, etanercept (Suffredini et al., 1995), to T24, HT1197, RT4, and SW780 cells. Although all four lines were sensitive to etanercept, T24 cell numbers declined by significantly greater extents than did the others (Figures S6B and 6B). Moreover, simultaneous application of etanercept and ML-SI1 induced a further decline of cell number in T24 but not RT4, HT1197, or SW780 cells (Figure 6B). These data argue in favor of a role for TNF, whose expression depended on TRPML1, in promoting the proliferation of bladder cancer cells in an autocrine manner. When grown on Matrigel, significantly larger number of T24 cells (TNFα high) invaded the matrix than did RT4 cells (TNFα low) (Figure 6C). Morphology of cells within the matrix also differed between the two lines. RT4 cells formed tightly packed clusters of 10–20 cells, whereas T24 cells invaded as individuals with typical mesenchymal morphology (Figure 6C). Knockdown of MCOLN1 in T24 cells decreased the number of invading cells, and the cells that permeated the matrix had acquired morphologies resembling those of RT4 cells, that is elevated cell–cell association (Figure 6C). Given the established roles for TNFα in metastases (Rossi et al., 2018; Wu and Zhou, 2010; Zhu et al., 2014), our data are consistent with TRPML1 regulating the invasiveness of bladder cancer cells via the regulation of TNF expression.

IL6 modulates the immune microenvironment of tumors by forcing tumor-associated macrophages (TAMs) to adopt the antiinflammatory and protumorigenic M2 state (Caetano et al., 2016; Chen et al., 2018; Fu et al., 2017; Mantovani et al., 2017). Using CIBERSORT (Newman et al., 2015), we found that tumors in the top half of MCOLN1 expression harbored a significantly higher density of M2 macrophages and lower density of activated dendritic cells than did tumors with lower MCOLN1 expression (Figures S6C and 6D). Enrichment for M2 TAMs also correlated with higher IL6 expression (Figure S6C). Therefore, BLCA tumors with higher MCOLN1 and IL6 expression exhibited a protumorigenic immune signature consistent with higher densities of M2 TAMs.

TRPML1 plays a permissive, but not sufficient, role in the regulation of proliferation and inflammation in bladder cancer cells

So far, we have shown roles for TRPML1 in promoting proliferation, invasion, and cytokine expression. Furthermore, knockdown of wild-type TP53 in HT1197, RT4, SW780, and 5637 cells was sufficient to increase MCOLN1 expression. These data prompted us to ask whether increased MCOLN1 expression in TP53 knockdown cells would be sufficient to augment proliferation and cytokine gene expression. However, treatment of HT1197, RT4, and SW780 cells with TP53 siRNA increased neither the rates of cell proliferation nor the sensitivity to ML-SI1 (Figure 7A). Similarly, TP53 knockdown alone did not elevate IL6 or TNF mRNA levels in HT1197, RT4, and SW780 bladder cancer cells (Figure 7B). These data demonstrate that increased MCOLN1 expression upon loss of p53 plays a strictly permissive role in bladder cancer cell proliferation and inflammation.

Figure 7

TRPML1 is necessary but not sufficient for cell proliferation and inflammation in bladder cancer cells

(A, C, and D) Bar graphs showing relative cell numbers in the indicated lines exposed to the conditions described below the graphs. In (A) and (C), values were normalized to HT1197 treated with DMSO alone. In (D), values were normalized to HT1197 treated with control siRNA. Circles represent independent biological repeats and the values shown represent mean ± SEM; black ∗∗∗∗, p < 0.0001, t-tests; red ∗∗, p < 0.01, ANOVA; red, ∗∗∗, p < 0.001, ANOVA.

(B and E) Bar graphs showing relative cytokine expression in the indicated cell lines treated as described. All values were normalized to the mean in DMSO-treated HT1197 cells. Circles represent independent biological repeats and the values shown represent mean ± SEM; ∗∗∗∗, p < 0.0001, ANOVA.

(F) Schematic showing that TRPML1 is necessary for HRASG12V–MEK–ERK signaling. Therefore, increased MCOLN1 expression in the absence of p53 permits MAPK-driven cell proliferation and inflammation. Concentrations, 75 nM TP53 siRNA, 200 nM MCOLN1 siRNA, and 10 μM ML-SI1. Abbreviation: n.s., not significant.

The requirement for TRPML1 in maintaining the localization and signaling competency of HRASG12V (Jung et al., 2019) raised the possibility that bladder cancer cells lacking p53 would be highly suitable for the subsequent appearance of HRAS mutations. In support of this notion, lentiviral transduction of HRAS into HT1197, RT4, and 5637 led to slight increases in rates of proliferation and the acquisition of de novo sensitivity toward TRPML1 inhibition or MCOLN1 knockdown (Figures 7C and 7D, respectively). These data indicate that oncogenic HRAS instilled the requirement for TRPML1 function in bladder cancer cells. Ectopic HRAS also induced a >2-fold increase in IL6 and TNF transcription (Figures 7E and S7A). Indicating that cytokine expression in HRAS-expressing cells was driven by the MEK-ERK pathway, inhibition of MEK1/2 using the highly selective drug, U0126 (MEKi) (Duncia et al., 1998), attenuated expression of both IL6 and TNF in T24 cells (Figure S7B). Cytokine expression driven by HRASG12V was also abolished by ML-SI1 (Figure 7E). Taken together these data indicate that increased MCOLN1 expression after the loss of p53 has a necessary role of HRASG12V-driven cell proliferation and inflammation but is not sufficient for either in the absence of HRASG12V.

Discussion

p53 has a necessary and sufficient role in repressing MCOLN1 in the urothelium

In this study, we provide multiple lines of evidence that point to a necessary and sufficient role for p53 in the regulation of MCOLN1 expression in both malignant and healthy urothelial cells. First, in TCGA data sets, we found that MCOLN1 was upregulated in primary BLCA tumors harboring TP53 mutations that are predicted to ablate the transactivation function of p53. In tumors that were either wild type for TP53 or harbored mutations in the regions of p53 that do not bind DNA, MCOLN1 expression remained unchanged. Another way to frame these findings is that increased MCOLN1 expression in BLCA speaks to the preponderance of transactivation-deficient p53 mutations in this disease. Second, ectopic knockdown of TP53 in either healthy urothelial cells or bladder cancer lines was sufficient for augmenting MCOLN1 expression. The MCOLN1 paralogs, MCOLN2 and MCOLN3, were not as responsive to TP53 knockdown, which suggests an element of selectivity in the relationship between p53 and the mucolipin genes. Third, we found that forced stabilization of the p53 by application of nutlin led to the repression of MCOLN1 in both bladder cancer and healthy urothelial cells.

In agreement with prior reports that oncogenic HRAS-driven MEK-ERK signaling mediates p53 degradation via Mdm2 (Ries et al., 2000), we found that ectopic expression of HRAS in bladder cancer cells lowered p53 abundance. These data suggest that p53 could be an intermediary in the induction of MCOLN1 by activated HRAS (Jung et al., 2019). Given that TRPML1 ensures the transduction of signals from HRASG12V to MEK–ERK (Figure 7F) (Jung et al., 2019), the relationship between activated HRAS, p53, and TRPML1 we describe is inherently protumorigenic. Although additional studies will be needed for understanding of the mechanisms by which p53 represses MCOLN1, the finding that TFEB knockdown prevented the induction of MCOLN1 in the constitutive absence or acute knockdown of TP53 hints at the involvement of TFEB (Figure 7F). However, TFEB, TFE3, or MITF do not belong to the list of p53 transcriptional targets (Allen et al., 2014). Therefore, it is likely that p53 represses the TFEB-MCOLN1 axis indirectly—for instance, by influencing nucleocytoplasmic transport of TFEB. Reports of constitutively nuclear TFEB upon the loss of p53 in certain cancers (Zhang et al., 2017) agree with this idea. It is also noteworthy that in normal mouse fibroblasts exposed to DNA-damaging agents, p53 activates, rather than represses, TFEB/TFE3 (Brady et al., 2018). Therefore, the qualitative relationship between p53 and endolysosomal biogenesis could also depend on the cell and tissue type under observation.

TRPML1 supports oncogene-induced proliferation in p53-deficient bladder cancer cells

Given that TP53 mutations are amongst the earliest events in bladder cancer (Gerstung et al., 2020), our data raise the possibility that elevated expression of MCOLN1 could be occurring early in the trajectory to malignancy. These transcription changes, in turn, would increase the probability that TRPML1-dependent oncogenic pathways are subsequently activated in those cells. In agreement, our bioinformatic analyses revealed that MCOLN1 and gene sets related to cell proliferation are concurrently induced in TP53 BLCA tumors. Furthermore, in the p53-deficient and HRASG12V-positive T24 bladder cancer cells, inhibition of TRPML1 led to decreased rate of proliferation. The antiproliferative effects of TRPML1-inhibition results from an increase in the fraction of cells in the G0/G1 phase of the cell cycle and corresponding decreases in the fractions of cells in S and G2/M phases. These data agree with G1 arrest being an outcome of MEK-ERK inhibition in RAS-transformed cancer cells (Chambard et al., 2007; Vasjari et al., 2019; Yamamoto et al., 2006). Because ML-SI1 slowed the progression of cancer cells through the cell cycle rather than inducing cell death, simultaneous application of the drug with the cytotoxic agent, cisplatin, led to a purely additive decline of T24 cell number.

Although TP53 knockdown in RT4, HT1197, 5637, and SW780 cells increased MCOLN1 expression, these alterations were not sufficient to accelerate cell proliferation. These data agree with the general understanding that loss of p53 is not in and of itself sufficient to initiate neoplasia but rather predisposes cells toward accumulating oncogenic mutations (Schwitalla et al., 2013). Thus, it is possible that the induction of oncogenic signaling would be more favorable in cells lacking p53 owing to attendant the elevation in MCOLN1 expression. In agreement, introduction of HRAS into HT1197, RT4, SW780, and 5637 cells was sufficient to confer upon those cells a requirement for functional TRPML1. Therefore, inhibition of TRPML1 or MCOLN1 knockdown had a greater cytostatic effect on those cell lines after the ectopic expression of HRAS. These data suggest that MCOLN1 induction, although not tumorigenic per se, sets the stage for oncogenic mutations that require higher TRPML1 abundance. Because TRPML1 recycles and maintains plasma membrane cholesterol (Jung et al., 2019), any signaling axis that is dependent on surface cholesterol would in principle be sensitive to TRPML1 inhibition. If so, increased MCOLN1 expression after the loss of p53 would favor proliferation driven by several oncogenic pathways.

TRPML1 supports oncogene-induced inflammation in p53-deficient bladder cancer cells

Consistent with reports that loss of p53 triggers inflammation (Schwitalla et al., 2013), gene sets related to inflammation were upregulated in TP53 BLCA tumors. Likewise, IL6 and TNF mRNA were significantly higher in T24 than in RT4 or HT1197 cells. Because endolysosomal proteins such as TRPML1 are needed for NF-κB-dependent cytokine production (El-Houjeiri et al., 2019; Sun et al., 2015; Visvikis et al., 2014; Wong et al., 2017), MCOLN1 knockdown or TRPML1 inhibition ablated the induction of cytokines. As was the case with rates of cell proliferation, increased MCOLN1 expression after TP53 knockdown was not sufficient to augment cytokine gene expression. Rather, this requirement for TRPML1 also depended on the presence of activated HRAS, which drove IL6 and TNF expression via MEK as an intermediary.

What might be the consequences of cytokine gene expression to the cancer cells? Prior studies have shown that TNFα can promote the proliferation of malignant cells (Gakis, 2014; Ham et al., 2016; Michaud, 2007; Wang et al., 2014; Zhu et al., 2014). In agreement, we found that application of the TNFα inhibitor, etanercept, diminished the proliferation of T24 cells to a significantly extent than did RT4, HT1197, and SW780 cells. Furthermore, the effects of etanercept on T24 cell number were enhanced by ML-SI1—a phenotype that was not observed in RT4, HT1197, and SW780 cells. TNFα has also been reported to promote cancer cell invasion (Rossi et al., 2018; Wu and Zhou, 2010; Zhu et al., 2014). In concordance with the elevation in TNF expression, T24 cells were significantly more invasive than were RT4 or HT1197 cells. MCOLN1 knockdown mitigated the invasiveness of T24 cells, which agrees with TRPML1 being needed for TNF expression. IL6 is secreted by primary cancer cells and interacting stromal entities such as fibroblasts and compels TAMs into the anti-inflammatory, M2 state (Caetano et al., 2016; Chen et al., 2018; Cho et al., 2018; Fu et al., 2017; Mantovani et al., 2017). In agreement with our finding that TRPML1 is needed for augmented IL6 expression, BLCA tumors with higher MCOLN1 expression exhibited significantly greater density of M2 macrophages. Given that M2 TAMs discourage the infiltration of antitumorigenic T lymphocytes (Mantovani et al., 2017), our data raise the possibility that higher MCOLN1 expression is predictive of an immune-cold tumor microenvironment, and thus, poor patient prognosis (Gardner and Ruffell, 2016; Gu-Trantien et al., 2013). Future studies could evaluate whether the simultaneous application of TRPML1 inhibitor and checkpoint blockers enhance the efficacy of immunotherapy.

Limitations of the study

Despite detecting an inverse relationship between p53 and MCOLN1 expression in several transformed and normal urothelial cells, we found that the 5637 bladder cancer cells did not exhibit constitutive overexpression of MCOLN1 despite harboring a missense mutation in the DNA-binding domain of p53 (p53R280T). Given that p53 abundance was higher in 5637 cells than in those with wild-type alleles of TP53, it is possible that compensatory upregulation of p53 abundance in 5637 cells results in partial restoration of function. Indeed, siRNA-mediated knockdown of TP53 in 5637 cells augmented MCOLN1 expression. These data indicate that in the case of cancer cells with missense LOF mutations in p53, compensatory pathways may preclude the upregulation of MCOLN1 expression and potentially limit the efficacy of TRPML1 inhibition. We also note that upon exposure to DNA damaging agents, p53 has been shown to activate, rather than repress, the TFEB/TFE3 transcriptional axis in normal mouse fibroblasts activates (Brady et al., 2018). Therefore, the qualitative relationship between p53 and TFEB/TFE3 targets such as TRPML1 could be a function of the cell and tissue type under observation. Finally, the relationship between MCOLN1 expression and the immune-cold tumor microenvironment, as detected in the TCGA BLCA datasets, would need to be experimentally evaluated in tumors before targeting TRPML1 to augment immunotherapy.

STAR★methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Kartik Venkatachalam (kartik.venkatachalam@uth.tmc.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All bioinformatic data used in this manuscript were obtained from TCGA as indicated in the Method Details and are included in the supplemental tables. No new datasets were generated in this study. The code generated during this study are available at GitHub (https://github.com/kvenkatachalam-lab/Jung-et-al.-2021-iScience-paper).

Experimental model and subject details

Cell lines and primary cultures

The following cancer cell lines were used in the study:

1. HT1197 cells (RRID: CVCL_1291).

Origin: Human bladder carcinoma cell line obtained from a 44-year-old male.

Culture media and conditions: HT1197 cells were obtained from ATCC and were cultured in DMEM high glucose media containing 2mM L-glutamine and supplemented with 10% (v/v) inactivated fetal bovine serum (FBS), penicillin-streptomycin, non-essential amino acids, and pyruvate. Cells were maintained in a 37°C incubator under 5% CO2. We did not authenticate this cell line in our laboratory.

2. RT4 cells (RRID: CVCL_0036).

Origin: Human bladder carcinoma cell line obtained from a 63-year-old male.

Culture media and conditions: RT4 cells were obtained from ATCC and were cultured in DMEM high glucose media containing 2mM L-glutamine and supplemented with 10% (v/v) inactivated fetal bovine serum (FBS), penicillin-streptomycin, non-essential amino acids, and pyruvate. Cells were maintained in a 37°C incubator under 5% CO2. We did not authenticate this cell line in our laboratory.

3. SW780 cells (RRID: CVCL_1728).

Origin: Human bladder carcinoma cell line obtained from an 80-year-old female.

Culture media and conditions: SW780 cells were obtained from ATCC and were cultured in DMEM high glucose media containing 2mM L-glutamine and supplemented with 10% (v/v) inactivated fetal bovine serum (FBS), penicillin-streptomycin, non-essential amino acids, and pyruvate. Cells were maintained in a 37°C incubator under 5% CO2. We did not authenticate this cell line in our laboratory.

4. 5637 cells (RRID: CVCL_0126).

Origin: Human bladder carcinoma cell line obtained from a 68-year-old male.

Culture media and conditions: 5637 cells were obtained from ATCC and were cultured in DMEM high glucose media containing 2mM L-glutamine and supplemented with 10% (v/v) inactivated fetal bovine serum (FBS), penicillin-streptomycin, non-essential amino acids, and pyruvate. Cells were maintained in a 37°C incubator under 5% CO2. We did not authenticate this cell line in our laboratory.

5. T24 cells (RRID: CVCL_0554).

Origin: Human bladder carcinoma cell line obtained from an 82-year-old female.

Culture media and conditions: T24 cells were obtained from ATCC and were cultured in DMEM high glucose media containing 2mM L-glutamine and supplemented with 10% (v/v) inactivated fetal bovine serum (FBS), penicillin-streptomycin, non-essential amino acids, and pyruvate. Cells were maintained in a 37°C incubator under 5% CO2. We did not authenticate this cell line in our laboratory.

The following primary cells were used in the study:

1. Primary urothelial cells (ATCC PCS-420-010)

Origin: Human primary urothelial cells. Sex of the cells, which are batch specific, was not known.

Culture media and conditions: Primary urothelial cells were cultured in bladder epithelial cell basal medium (ATCC-PCS-420-032) supplemented with bladder epithelial cell growth kit (ATCC-PCS-420-042). We did not authenticate this cell line in our laboratory.

Method details

Bioinformatic analyses

Analyses of differential gene expression

Raw counts (CPM) from the RNA-seq analyses of BLCA tumors and matched controls were obtained from GDC (https://portal.gdc.cancer.gov). Differential expression was determined using EdgeR (Robinson et al., 2010) with FDR < 0.05 as the cutoff.

Analyses of correlation with MCOLN1

Spearman's correlation coefficients of coexpression with MCOLN1 in BLCA were obtained from cBioPortal (https://www.cbioportal.org). FDR < 0.05 was the cutoff for determining genes that were significantly correlated with MCOLN1.

GSEA

GSEA was performed using the standalone GSEA_4.0.3 application downloaded from www.gsea-msigdb.org. Pre-ranked GSEA was performed using either custom gene sets or gene sets obtained from MSigDB. For the analyses, no collapse was used, and the default minimum and maximum exclusion criteria of 15 and 500, respectively.

Analyses of networks

We used CytoScape (Shannon et al., 2003) to generate networks using the results of GSEA analyses. Gene sets were nodes and extent of overlap between the sets represented edge values. Node size represented number of genes, whereas node color represented normalized enrichment score (NES) from GSEA analyses.

Determination of Shannon information

List of somatic mutations (SNPs and small INDELs) that were determined using MuTect2 variant aggregation and masking were obtained from the UCSC Xena functional genomics browser (https://xenabrowser.net/datapages/). Using a custom R script, we filtered all synonymous mutations, and genes that did not belong to the consensus list of 722 CGCs (https://cancer.sanger.ac.uk/census; Table S3). Tumors were then ranked on the basis of MCOLN1 FPKM values. The ranked set was then partitioned into groups comprised of tumors that belonged to either the top third or the bottom third of the ranked list. Conditional probability that mutations in a gene would occur in tumors that belonged to a group was calculated using the formula:

Entropy associated with gene mutations occurring in a group was calculated using the formula:

Shannon information was calculated using the formula:

For the permutation test, we first generated a null dataset comprised of information values for each gene obtained after randomly reordering the assignment of MCOLN1 FPKM values. Default number of permutations was 1000. Nominal P-values were calculated by comparing computed information for each gene with the respective values from the null dataset using a one-sample t-test. FDR Q values were calculated from the nominal P-values by applying the Benjamini and Hochberg correction.

CIBERSORT analyses

Distribution of immune cell types in the BLCA tumors was determined using the online portal (https://cibersort.stanford.edu/index.php). Fractions of each immune cell type were then compared in groups of tumors sorted on the basis of MCOLN1 FPKM values. Pairwise comparisons were made using the Mann-Whitney test, and FDR was analyzed using the two-stage step-up method of Benjamini, Krieger and Yakutieli.

Cell culture

T24, HT1197, RT4, SW780, and 5637 cell lines (all from ATCC) were cultured in DMEM high glucose media containing 2mM L-glutamine and supplemented with 10% (v/v) inactivated fetal bovine serum (FBS), penicillin-streptomycin, non-essential amino acids, pyruvate (all from Fischer Scientific). Primary urothelial cells (ATCC PCS-420-010) were cultured in bladder epithelial cell basal medium (ATCC-PCS-420-032) supplemented with bladder epithelial cell growth kit (ATCC-PCS-420-042). All of the cell lines were maintained at 37°C and 5% CO2. For drug treatment, we added DMSO (vehicle) or the drugs directly to the cell media.

Gene knockdown by RNA interference

For siRNA transfections, cells were transfected with siRNA oligonucleotides using X-treme GENE 9 DNA Transfection reagents (Roche) following manufacturer’s instructions. Subsequent analyses were performed after 2- or 5-days of siRNA treatment. The following sequences were used:

MCOLN1 — 5’-CCCACATCCAGGAGTGTAA-3’ (200 nM)

TP53 — 5’-AGACTCCAGTGGTAATCTA-3’ (75 nM)

RELA — 5’-GGAGTACCCTGAGGCTAT-3’ (50 nM)

TFEB — 5’-CCGCCTGGAGATGACCAACAA-3’ (50 nM)

Control siRNA (SIC001, Sigma-Aldrich) was used as a negative control. Concentration of control siRNA was equal to that of the specific siRNA.

Western blotting

When generating cellular extracts for Western blotting, we harvested cells in a lysis buffer containing protease inhibitors and DTT. We loaded the cell extracts onto 4–20% gradient Tri-glycine gels (Bio-Rad) for SDS–PAGE. After transfer to nitro- cellulose membranes, the blots were blocked using the Odyssey blocking buffer (LI-COR Biosciences) followed by incubation with primary antibodies and secondary antibodies described below. All blots were imaged using the Odyssey imaging platform (LI-COR Biosciences). For quantification, we determined band intensities using Fiji/ImageJ (NIH) (Schindelin et al., 2012). When probing the blots with antibodies from the same species, we stripped the blots using the NewBlotTM nitrocellulose stripping buffer as per the manufacturer’s instructions (LI- COR Biosciences). All antibody incubations were performed in the Odyssey blocking buffer (LI-COR Biosciences). Primary antibodies used were mouse anti-p53 (1:2000, DO-1, Santa Cruz Biotechnology), and mouse anti-β-tubulin (1:2000, E7, DSHB). Secondary antibodies used were goat anti-mouse IRDye 800CW (LI-COR Biosciences).

Lentiviral transduction

To ectopically express HRASG12V in bladder cancer cells, we transduced cells with a HRAS-GFP lentiviral construct. Briefly, we treated cells with cell culture media containing lentivirus and 4 μg/ml of hexadimethrine bromide (Sigma) for 12 hours. After 12 hours, we removed this media and added full growth media. We incubated the cells for 48 hours before adding media with puromycin (1:5000 dilution) for selection of cells expressing HRAS-GFP.

Analyses of cell proliferation

WST-1 assay

Cells were seeded in 96-well plates at the density of 104 cells/well and incubated for 12 hours at 37°C (CO2). Subsequently, fresh media containing DMSO or drugs or control or specific siRNA was applied to the cells. Cell numbers were assessed 2- or 5-days later using the WST-1 (4-(3-(4- iodophenyl)-2-(4-nitrophenyl)-2H-5-tetrazolio)-1,3-benzene disulfonate, Roche Applied Sciences) assay as per the manufacturer’s protocol. Absorbance was measured using a microplate ELISA reader (Cytation 5, BioTek) at 450 nm.

Crystal violet staining

Cells were seeded in 6-well plates (2x105 cells/well), and exposed to 10 μM MLSI for the indicated durations. Cells were then fixed for 5 minutes with 4% PFA, and stained for 10 minutes with 0.5% crystal violet (Sigma) diluted in PBS. After a series of washes in plain water, plates were dried and images were captured. Subsequently, crystal violet signal from the plates was solubilized in 100 μl methanol, and absorbances of the solutions were read at 540 nm.

Gene expression analysis

When performing RT-qPCR, total RNA was extracted with RNeasy Mini Kit (Qiagen) following manufacturer’s instructions. 1μg of total RNA was reverse-transcribed with High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time qPCR was performed using SYBR Green JumpStart Taq ReadyMix (Sigma) as per the manufacturer’s protocol. Primers used have been described previously (Chiang et al., 2017; Jung et al., 2019; Zhou et al., 2018), and were as follows:

GAPDH:

Forward: 5’-GAAGGTGAAGGTCGGAGTC-3’

Reverse: 5’-GAAGATGGTGATGGGATTTC-3’

MCOLN1:

Forward: 5’-CTGGTGGTCACGGTGCAG-3’

Reverse: 5’-CTGCTCCCGCGTGTAGG -3’

MCOLN2:

Forward: 5’-CGGCAGCCTTATCGTTTTCC-3’

Reverse: 5’-GCCATTGCATTTCTGACGGT-3’

MCOLN3:

Forward: 5’-TCTCCTCCCGTCTGACTCTG-3’

Reverse: 5’-CAGGATCTGCCATCTCTGGG-3’

RELA:

Forward: 5’-TGAACCGAAACTCTGGCAGCTG-3’

Reverse: 5’-CATCAGCTTGCGAAAAGGAGCC-3’

TFEB:

Forward: 5’-CCAGAAGCGAGAGCTCACAGAT-3’

Reverse: 5’-TGTGATTGTCTTTCTTCTGCCG-3’

IL6:

Forward: 5’-AGACAGCCACTCACCTCTTCAG-3’

Reverse: 5’-TTCTGCCAGTGCCTCTTTGCTG-3’

TNFA:

Forward: 5’-CTCTTCTGCCTGCTGCACTTTG-3’

Reverse: 5’-ATGGGCTACAGGCTTGTCACTC-3’

Analyses of cell cycle

Number of cells in G1, S or G2/M phases of the cell cycle was determined by measuring DNA content using propidium iodide (PI) and bromodeoxyuridine (BrdU) labeling followed by flow cytometry analysis. Cells were pulsed with 10 μM BrdU for 45 minutes, harvested and fixed with 70% ice-cold ethanol overnight. Cells were washed in BrdU wash solution (0.5% Tween 20, 0.5% BSA in PBS) and resuspended in 2 N HCl for 20 minutes at room temperature. 0.1 M sodium borate was added to neutralize, and the cells were washed twice with BrdU wash solution. Cells were incubated with anti-BrdU antibody (BD 347580 mouse IgG) at room temperature for 30 minutes, washed thrice in BrdU wash solution and incubated with Alexa Fluor 488 goat anti-mouse IgG (Invitrogen) for 30 minutes at room temperature. After three washes with BrdU wash solution, the cells were treated with 100 ug/ml RNase and 25 μg/ml propidium iodide for 20 minutes at 37°C. Analyses were performed using a BD LSRFortessa instrument.

Cell invasion assay

3×103 cells in 100 μL serum-free DMEM were placed into the upper chamber of a 24-well Transwell insert (8 μm pore size; Trevigen) coated with 10% of Matrigel (v/v, Corning). 500 μL complete media was added to the lower chamber. After 24 hours, cells remaining on the upper membrane were removed with cotton wool. Cells that had migrated or invaded through the membrane were stained with the crystal violet (0.5%, Sigma) after fixation with 4% PFA as described above. Stained cells were then photographed and counted using ImageJ.

Quantification and statistical analysis

We used either a parametric or a nonparametric test of statistical significance on the basis of whether the data were normally distributed. Multiple comparisons were made by ANOVA. R, Excel (Microsoft) and Prism (GraphPad) were used for statistical analyses. Statistical significance was defined as a P < 0.05. P-values were shown on the figures as asterisks: ∗, P < 0.05; ∗∗, P < 0.01; ∗∗∗, P < 0.001; ∗∗∗∗, P < 0.0001. Independently performed biological replicates (n values) are indicated as circles in the bar graphs.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-p53 antibody (DO-1)	Santa Cruz Biotechnology	Cat#: sc-126; RRID: AB_628082
Mouse monoclonal anti-β-tubulin antibody (E7)	DSHB	Cat#: E7; RRID: AB_528499
Goat anti-mouse IRDye 800CW secondary antibodies	LI-COR Biosciences	Cat#: 925-32210; RRID: AB_2687825
Mouse anti-BrdU (clone B44)	BD Biosciences	Cat#: 347580
Alexa Fluor 488 goat anti-mouse secondary antibodies	Invitrogen	Cat#: A11001
Bacterial and virus strains
LV-HRASG12V-GFP	This paper	N/A
Chemicals, peptides, and recombinant proteins
Nutlin-3a	Tocris	Cat#: 6075
ML-SI1 (GW-405833)	Sigma	Cat#: G1421
Cisplatin	Sigma	Cat#: PHR1624
Etanercept	Sigma	Cat#: Y0001969
U0126 (MEKi)	Sigma	Cat#: 19-147
Propidium iodide	Sigma	Cat#: 81845
Bromodeoxyuridine (BrdU)	Acros	Cat#228590010
RNAse	Sigma	Cat#: R6513
Hexadimethrine bromide	Sigma	Cat#: H9268
Crystal violet	Sigma	Cat#: C6158
Critical commercial assays
X-treme GENE 9 DNA Transfection reagents	Roche	Cat#: 6365779001
WST-1 cell proliferation assay	Roche	Cat#: 5015944001
Cultrex® Cell Migration Assay	Trevigen	Cat#: 3465-024-K
RNeasy Mini Kit	Qiagen	Cat#: 74004
High-Capacity cDNA Reverse Transcription Kit	Applied Biosystems	Cat#: 43-688-14
SYBR Green JumpStart Taq ReadyMix	Sigma	Cat#: S4438
Experimental models: Cell lines
Human: HT1197 cells	ATCC	RRID: CVCL_1291
Human: RT4 cells	ATCC	RRID: CVCL_0036
Human: SW780 cells	ATCC	RRID: CVCL_1728
Human: 5637 cells	ATCC	RRID: CVCL_0126
Human: T24 cells	ATCC	RRID: CVCL_0554
Human: Primary urothelial cells	ATCC	Cat#: PCS-420-010
Oligonucleotides
MCOLN1 siRNA — 5’-CCCACATCCAGGAGTGTAA-3’	Jung et al. (2019)	N/A
TP53 siRNA — 5’-AGACTCCAGTGGTAATCTA-3’	Xu et al. (2009)	N/A
RELA siRNA — 5’-GGAGTACCCTGAGGCTAT-3’	This paper	N/A
TFEB siRNA — 5’-CCGCCTGGAGATGACCAACAA-3’	Jung et al. (2019)	N/A
MISSION® siRNA Universal Negative Control (Control siRNA)	Sigma	Cat#: SIC001
GAPDH sequencing:Forward: 5’-GAAGGTGAAGGTCGGAGTC-3’Reverse: 5’-GAAGATGGTGATGGGATTTC-3’	Jung et al. (2019)	N/A
IL6 sequencing:Forward: 5’-AGACAGCCACTCACCTCTTCAG-3’Reverse: 5’-TTCTGCCAGTGCCTCTTTGCTG-3’	Zhou et al. (2018)	N/A
MCOLN1 sequencing:Forward: 5’-CTGGTGGTCACGGTGCAG-3’Reverse: 5’-CTGCTCCCGCGTGTAGG -3’	Jung et al. (2019)	N/A
MCOLN2 sequencing:Forward: 5’-CGGCAGCCTTATCGTTTTCC-3’Reverse: 5’-GCCATTGCATTTCTGACGGT-3’	Jung et al. (2019)	N/A
MCOLN3 sequencing:Forward: 5’-TCTCCTCCCGTCTGACTCTG-3’Reverse: 5’-CAGGATCTGCCATCTCTGGG-3’	Jung et al. (2019)	N/A
RELA sequencing:Forward: 5’-TGAACCGAAACTCTGGCAGCTG-3’Reverse: 5’-CATCAGCTTGCGAAAAGGAGCC-3’	This paper	N/A
TFEB sequencing:Forward: 5’-CCAGAAGCGAGAGCTCACAGAT-3’Reverse: 5’-TGTGATTGTCTTTCTTCTGCCG-3’	Jung et al. (2019)	N/A
TNFA sequencing:Forward: 5’-CTCTTCTGCCTGCTGCACTTTG-3’Reverse: 5’-ATGGGCTACAGGCTTGTCACTC-3’	Chiang et al. (2017)	N/A
Recombinant DNA
LV-HRASG12V-GFP	This paper	N/A
Software and algorithms
Fiji (ImageJ)	Schindelin et al. (2012)	https://imagej.net/Fiji
Prism8	Graphpad	https://www.graphpad.com/scientific-software/prism/
RStudio	RStudio	https://www.rstudio.com
EdgeR	Robinson et al. (2010)	https://bioconductor.org/packages/release/bioc/html/edgeR.html
GSEA	Subramanian et al. (2005)	https://www.gsea-msigdb.org/gsea/index.jsp
CytoScape	Shannon et al. (2003)	https://cytoscape.org
Cibersort	Newman et al. (2015)	https://cibersort.stanford.edu
Determination of Shannon information	This study	https://github.com/kvenkatachalam-lab/ung-et-al.-2021-iScience-paper