Rosiglitazone and trametinib exhibit potent anti-tumor activity in a mouse model of muscle invasive bladder cancer

Abstract

Muscle invasive bladder cancers (BCs) can be divided into 2 major subgroups-basal/squamous (BASQ) tumors and luminal tumors. Since Pparg has low or undetectable expression in BASQ tumors, we tested the effects of rosiglitazone, Pparg agonist, in a mouse model of BASQ BC. We find that rosiglitazone reduces proliferation while treatment with rosiglitazone plus trametinib, a MEK inhibitor, induces apoptosis and reduces tumor volume by 91% after 1 month. Rosiglitazone and trametinib also induce a shift from BASQ to luminal differentiation in tumors, which our analysis suggests is mediated by retinoid signaling, a pathway known to drive the luminal differentiation program. Our data suggest that rosiglitazone, trametinib, and retinoids, which are all FDA approved, may be clinically active in BASQ tumors in patients.

New treatments are needed for muscle invasive bladder cancers. Here, the authors show that combined Pparg activation and MEK inhibition using FDA approved drugs shrinks tumor volume and induces a Basal/Squamous-to-Luminal shift in the urothelium as well as in invading tumors.

Introduction

Urothelial carcinoma of the bladder (bladder cancer; BC) is currently the 6th most common form of cancer in the United States with about 81,000 new cases and 17,000 deaths per year. It occurs three times more frequently in men than in women^1–3, but women tend to have more aggressive disease⁴. Most cases of BC (70%) present as non-muscle invasive bladder cancer (NMIBC) that includes patients with stage Ta and T1 tumors as well as carcinoma in situ. All NMIBCs are treated with complete transurethral resection, and patients with high-risk disease are also treated with adjuvant intravesical BCG. However, 50–70% of patients experience recurrence and up to 20% of high-grade NMIBCs progress to muscle invasive bladder cancer [MIBC²]. MIBC is associated with a 50% five-year survival rate, and metastatic progression almost always leads to death^5,6.

Neoadjuvant cisplatin-based chemotherapy followed by definitive surgery (radical cystectomy) has been the preferred treatment for MIBC since the early 2000s^7–9, although radiation-based trimodal therapy (TMT) is equally effective in well-selected patients¹⁰. Immune-checkpoint inhibitors (ICIs) such as atezolizumab and pembrolizumab are active in approximately 20% of patients with advanced and metastatic disease^11–13, and novel combinations of ICIs and antibody-drug conjugates are displaying even stronger activity, although the durability of this benefit is still unclear. In contrast to basal/squamous (BASQ) tumors, many luminal papillary (LP) tumors are poorly infiltrated with immune cells, and this subset of tumors has a lower response rate to immune checkpoint inhibitors^2,12,14–17. Overall, there is still a great need to identify new avenues for treatment of BC.

Pparg is a likely regulator of tumor subtype in bladder cancer. PPARG mutations and genomic alterations occur frequently in bladder cancer; 20–25% of luminal tumors display activation of Pparg-dependent transcription due to amplification of the PPARG gene or mutations in its binding partner RXRA¹⁸. On the other hand, PPARG expression is downregulated in BASQ subtype tumors^19–22. Bladder cancers arise from the urothelium, a specialized epithelia composed of basal and luminal cell types. Pparg has been shown to induce urothelial differentiation in human urothelial cells in vitro, and to promote the luminal differentiation program in mouse models and organoids in vivo^23,24. On the other hand, Pparg loss-of-function mutants display squamous differentiation in the urothelium which worsens with injury²⁵. Studies in a mouse model of breast cancer indicate that a combination of Rosiglitazone (Rosi), a synthetic Pparg agonist, and Trametinib (Tram), an MEK/ERK inhibitor, reduced growth and metastases and induced tumor cells to differentiate into post-mitotic adipocytes²⁶. These observations prompted us to investigate whether a similar protocol could be efficacious in the treating MIBC. We tested the combined and individual effects of Rosi and Tram in mice harboring BASQ tumors induced by BBN, a carcinogen that results in formation of BASQ tumors in situ after 4–5 months of exposure. We find that BASQ tumors in BBN-treated mice dosed with Rosi or Tram alone displayed reduced proliferation while combined Rosi+Tram induced apoptosis within 7 days and decreased tumor volume by 91% after one month of administration. Apoptosis was accompanied by down-regulation of Bcl2 and pro-survival genes as well as Ccnd1 and Cdkn2a. Pparg is known to down-regulate Ccnd1 and cell cycle regulators in urothelial cells²⁷, while Bcl2 and Ccnd1 are known targets of MEK/ERK signaling³. In addition to anti-tumor effects, Rosi+Tram treatment induced luminal differentiation in the urothelium of mice harboring BASQ tumors and induced a shift from BASQ to luminal differentiation in BBN963 cells, which are derived from a BBN-induced BASQ tumor. Our studies suggest that these events are mediated at least in part by retinoid signaling, which is activated in Rosi+Tram treated tumors. Retinoids and Pparg have similar effects on urothelial differentiation: Retinoic acid (RA), which is the active metabolite of Vitamin A (retinol), is required for differentiation of luminal urothelial cell types^28,29, and landmark studies from the 1940s indicate that vitamin A deficiency induces keratinizing squamous metaplasia in the urothelium of rats bred on vitamin A deficient diets, in which the endogenous urothelial cell types are replaced by a skin-like sheet^30–32.

RA signaling is mediated by RA-receptors which, like Pparg, belong to the nuclear receptor superfamily^33–35. Retinoic acid receptors (Rars) are ligand activated transcription factors that bind to DNA response elements in promoters and enhancers of target genes as heterodimers with a second nuclear receptor family member, Rxr. In the absence of RA, Rar/Rxr heterodimers are transcriptionally inactive or repressive where they are bound by co-repressors. RA binding to the DNA binding domain of the Rar subunit induces a conformational change, promoting the release of co-repressors and the recruitment of co-activators. Rars and Rxrs are widely expressed; however, transcriptional activity is limited by the local availability of RA which is synthesized from retinol (vitamin A), an inactive precursor by retinol and retinaldehyde-synthesizing enzymes³³. Our studies reveal that RA-synthesizing enzymes which are inactive in BASQ tumors are up-regulated in tumors treated with Rosi+Tram, inducing expression of canonical RA targets including Rarb, which is regulated by an RA-response element in its promoter sequences³⁶.

We also identify changes in AP-1 and Kdm6a in BASQ tumors treated with Rosi+Tram. Kdm6a, a lysine demethylase that is part of the Mll3/4 complex, is undetectable in BBN-induced BASQ tumors is up-regulated in BASQ tumors treated with Rosi+Tram, while AP1, a pathway that promotes BASQ differentiation that is directly suppressed by Rars^37,38, is the most enriched motif among down-regulated genes in Rosi+Tram treated tumors, suggesting a possible role as a regulator of BASQ versus LP differentiation. Recently published work indicates that Kdm6a both facilitates RA-signaling and inhibits AP-1 activity in urothelial cells and in bladder cancer cells³⁹. On the other hand, AP-1, a regulator of keratinocyte differentiation, has recently been shown to bind to super-enhancers that control BASQ differentiation⁴⁰. Together our findings suggest that Rosi+Tram treatment inhibits growth and survival of BASQ tumors and also opposes a squamous differentiation program, up-regulating retinoids and Kdm6a, which promote luminal differentiation, and down-regulating AP-1, driving a BASQ differentiation program.

Although the FDA restricted rosiglitazone prescriptions in 2010 due to concerns of increased acute myocardial infarction and risk and cardiac deaths, further evaluation of longitudinal data has led to removal of these restrictions in 2013 and the Risk Evaluation and Mitigation Strategy (REMS) was eliminated in 2015^41–43. Both Trametinib (Mekinist) and all-Trans-RA (Tretinoin) are FDA approved and used for prevention and treatment of cancer. Tretinoin is used to treat APL and for prevention of cSCC, PML, and Karposi’s sarcoma^44–46, while Trametinib (Mekinist) is used to treat cancers including melanoma, non-small cell lung cancer, and anaplastic thyroid cancer^47–49. Our studies here using the BBN carcinogen-induced mouse model suggests that Rosi, Tram and RA, all of which are FDA approved, may be efficacious for treating MIBC in a clinical setting.

Results

Combined Rosiglitazone and Trametinib induce tumor regression in a mouse model of BBN-induced MIBC

Pparg is undetectable in BASQ tumors at the level of mRNA and protein. We thus wanted to test the effects of activating Pparg signaling in BASQ tumors using the synthetic Pparg agonist rosiglitazone (hereafter referred to as Rosi). We used a mouse model of BBN [N-butyl-n-(4- hydroxybutyl)-nitrosamine]-induced carcinogenesis that produces BASQ tumors after 4–5 months of exposure^18,50–52. Wild type mice were exposed to BBN for 5 months then treated with Rosi or vehicle by daily oral gavage for one month. Ultrasound analyses of tumor growth in control animals treated with vehicle revealed robust growth, where the median change was 730.9% (Table 1; Fig. 1A, B, I). Treatment with Rosi alone was not sufficient to eliminate BBN-induced BASQ tumors.

Table 1.

Tumor volume based on ultrasound from before and after treatments

Treatment	Before Volume (mm^3)	After Volume (mm^3)	Change (mm^3)	Average Change (mm^3)	Change (%)	Average Change (%)	Median Change (%)
5 M BBN 1 M Vehicle	0.265	2.190	1.925	12.651	726.4%	27839.0%	730.9%
	0.729	6.057	5.328		730.9%
	4.278	24.047	19.769		462.1%
	3.988	25.881	21.893		549%
	0.496	1.522	1.026		206.9%1
	0.562	9.868	9.306		1655.9%
	0.890	27.691	26.801		3011.3%
	0.010	24.083	24.073		240730.0%
5 M BBN 1 M RosiTram	4.467	2.190	−2.277	−3.198	−51.0%	−91.0%	−100.0%
	11.943	0.315	−11.628		−97.4%
	2.604	0.000	−2.604		−100.0%
	5.408	0.000	−5.408		−100.0%
	0.516	0.000	−0.516		−100.0%
	0.562	0.000	−0.562		−100.0%
	0.523	0.000	−0.523		−100.0%
	1.461	0.000	−1.461		−100.0%
	10.014	1.241	−8.773		−87.6%
	6.261	0.000	−6.261		−100.0%
	1.518	0.507	−1.011		−66.6%
	1.148	0.329	−0.819		−71.3%
	0.406	0.000	−0.406		−100.0%
	2.523	0.000	−2.523		−100.0%
5 M BBN 1 M Rosi only	2.379	2.903	0.524	3.412	22.0%	115.2%	114.9%
	6.140	20.129	13.989		227.8%
	3.906	0.000	−3.906		−100.0%
	7.947	24.451	16.504		207.7%
	20.929	0.887	−20.042		−95.8%
	3.120	16.525	13.405		429.6%
5 M BBN 1 M Tram only	1.554	12.217	10.663	3.783	686.2%	234.1%	2.32%
	2.226	20.224	17.998		808.5%
	6.152	4.197	−1.955		−31.8%
	4.149	5.564	1.415		34.1%
	6.545	2.424	−4.121		−63.0%
	6.133	4.829	−1.304		−21.3%

Fig. 1. Combined Rosi+Tram treatment reduces tumor burden in mice with BBN-induced MIBC.

Wildtype male mice were given BBN (0.05%) in drinking water ad libidum for 5 months. Mice were then treated for one month by daily via oral gavage with vehicle (Control), Rosi alone, Tram alone, or with a combination of Rosi+Tram. Tumor volume was measured prior to euthanasia, and bladders were collected. Ultrasound images and 3D reconstruction of bladders were obtained from mice before and after 1 month treatment. A–H Representative images of ultrasound imaging of a mouse exposed to BBN for 5 months before A and after B administration of vehicle for 1 month. Ultrasound image of a mouse exposed to BBN for 5 months before C and after D administration of Rosi for 1 month. Ultrasound images and 3D reconstruction of a bladder from a mouse treated for 5 months with BBN before E or after F treatment for 1 month with Tram. Ultrasound images and 3D reconstruction of bladders from mice treated for 5 months with BBN before G and after H 1 month of combined Rosi+Tram treatment. I. Box plot showing tumor volume measurements based on 3D reconstruction before and after 1 month vehicle (n = 7), Rosi (n = 6), Tram (n = 6), or Rosi+Tram (n = 13). Box plots display minima, maxima, and interquartile range (IQR). Significance calculated by one-way ANCOVA with Dunnet’s test for pairwise comparisons ***p = 0.0007. J–N Panoramas showing histological analysis of a bladders from BBN exposed mouse after 1 month of vehicle J, after 1 month of Rosi treatment K, after 1 month of Tram L, after 1 month of Rosi+Tram M, and untreated wild type control N. O–S Magnified regions in boxed regions shown in images J–N. Scale bars: 50 µm. Source data are provided as a Source Data file. Cartoons in Fig. 1 created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

Studies in a mouse model of breast cancer show that a combination of Rosi+Tram prevents both tumor growth and metastasis, inducing tumor cells to undergo adipogenesis and exit the cell cycle²⁶. The MEK/ERK pathway promotes survival and growth of tumors and suppresses Pparg transcriptional activity in tumor cells by phosphorylation at residue 273 (S273) in the Pparg ligand binding domain, which can result in its degradation^53–55. Inhibition of MEK/ERK signaling using Trametinib can maintain Pparg signaling activity and prevent its phosphorylation and degradation⁵⁶. MEK/ERK is activated downstream of RTK pathways, including EGFR, ERBB2, FGFR3, HRAS, K-RAS, BRAF, and MERTK⁵⁷, which are active in MIBC⁵⁸. On the other hand, PPARG expression is low or undetectable in BASQ MIBC, although the mechanism by which this occurs is not clear. To determine whether Tram increases Pparg signaling in MIBC tumors, BBN-treated mice were treated with either vehicle, Rosi alone, Tram alone, or a combination of Rosi+Tram by oral gavage for one month, and tumor growth was assessed by ultrasound imaging. However, in mice treated with Rosi+Tram, tumors decreased in size by an average of 91%; in most cases (9/14), residual tumors were not detectable (Table 1; Fig. 1G, H, I). Histopathological evaluation confirmed the observations from ultrasound of Rosi+Tram treated tumors. Bladders from animals dosed with vehicle for 1 month contained large tumors, many of which were muscle invasive and exhibited signs of squamous differentiation [T2-T3; Table 2, Fig. 1J, O]. Tumors from animals that received either Rosi or Tram alone displayed less severe histopathological changes [Table 2; (Dysplasia-T1)] compared to vehicle control tumors and lacked overt signs of squamous differentiation (Fig. 1K, L, P, Q). Consistent with the ultrasound results, bladders from animals treated with Rosi+Tram contained few tumors, and the histology of the urothelium was quite similar to that in untreated healthy animals (Table 2; Fig. 1I, M, N, R, S). These findings together suggest that Rosi or Tram alone can reduce tumor growth, but the combination of both drugs is more powerful and induces tumor regression after one month of continuous treatment.

Table 2.

Histopathological tumor evaluation based on H&E

Treatment	Histopathological findings	Tumor Stage
5 M BBN 1 M Vehicle	Severe urothelial dysplasia/carcinoma in situ with focal squamous differentiation	Tis
	Invasive urothelial carcinoma with squamous + glandular, surface with squamous dysplasia	T3
	Invasive urothelial carcinoma with extensive squamous and focal glandular differentiation	T3
	Invasive urothelial carcinoma with extensive squamous differentiation, surface CIS	T3
	Invasive urothelial carcinoma with poorly differentiated/sarcomatoid features, surface CIS	T1
	Invasive urothelial carcinoma, surface CIS	T1
	Invasive urothelial carcinoma, surface CIS and papillary carcinoma with extensive squamous differentiation	T1
	Invasive urothelial carcinoma, surface CIS with squamous differentiation	T1
5 M BBN 1 M RosiTram	Invasive urothelial carcinoma, surface CIS	T1
	Normal urothelium
	Urothelial atypia/dysplasia
	Normal urothelium
	Urothelial atypia
	Urothelial atypia
	Urothelial atypia
	Urothelial atypia
	Urothelial atypia/dysplasia
	Urothelial atypia
	Normal urothelium/focal hyperplasia
	Urothelial atypia/dysplasia
	Normal urothelium/focal hyperplasia
5 M BBN 1 M Rosi	Urothelial atypia/dysplasia
	Invasive sarcomatoid urothelial carcinoma, surface CIS	T2
	Severe urothelial dysplasia/carcinoma in situ	Tis
	Urothelial dysplasia
	Urothelial dysplasia
	Invasive urothelial carcinoma, surface CIS, extensive inflammation	T1
5 M BBN 1 M Tram	Urothelial dysplasia
	Urothelial dysplasia	T0
	Invasive urothelial carcinoma, surface dysplasia	T1
	Severe urothelial dysplasia/carcinoma in situ	Tis
	Urothelial dysplasia
	Invasive urothelial carcinoma, surface CIS	T1

As is the case in human BASQ tumors, BBN-induced tumors in mice express squamous markers including K14 and K6a and lack intermediate cells and superficial cells, populations that normally express luminal markers including Krt20, Pparg, Foxa1 and Gata3 (Fig. 2A, B, E–H). Analysis of the urothelium in animals treated either with Rosi or Tram alone revealed that upper levels of the urothelium, which were initially populated by BASQ cells, were now populated with intermediate and superficial cells expressing appropriate luminal markers, including Upk2 and K20. However, the basal-most layers still contained squamous cells expressing Krt6a+ Krt14+ cells, indicating that neither Rosi nor Tram alone were sufficient to respecify the basal population (Fig. 2E, I–N). In animals treated with combined Rosi+Tram, however, few residual tumors remained, squamous differentiation in the urothelium was reversed, and endogenous urothelial populations (K5-Basal cells, K14-Basal cells, intermediate cells and superficial cells) were restored (Fig. 2C–E, O–T).

Fig. 2. Combined Rosi+Tram treatment impairs tumor growth and restores normal urothelial differentiation in mice with BBN-induced BASQ tumors.

A–D Panoramas showing luminal (Upk2, K20) and basal/squamous (K14/K6a) marker expression in mouse exposed to BBN for 5 months then treated for 1 month with vehicle (n = 4) A, B or Rosi+Tram (n = 5) C, D. E Composition of the cell types in the urothelium in animals treated with vehicle, Rosi, Tram, Rosi+Tram, and no BBN controls. F, I, L, O, R. Upk2, K20, and P63 expression in animals exposed to BBN for 5 months then treated with vehicle (n = 4) F, Rosi (n = 5) I, Tram (n = 4) L, Rosi+Tram (n = 5) O, and untreated adult mouse controls (n = 3) R. G, J, M, P, S. Expression of K14, P63, and K6a in the urothelium of mice exposed to BBN for 5 months then treated for one month with vehicle (n = 4) G, Rosi (n = 5) J, Tram (n = 4) M, combined Rosi+Tram (n = 5) P, and untreated controls (n = 3) S. H, K, N, Q, T. Expression of K14, K5, and P63 in the urothelium of mice exposed to BBN for 5 months then treated for one month with vehicle (n = 4) (H), Rosi (n = 5) (K), Tram (n = 4) N, combined Rosi+Tram (n = 5) Q, and untreated controls (n = 3) (T). U–X Pparg/Fabp4 expression in urothelium of BBN-induced tumors after treatment for one month with vehicle U, Rosi V, Tram W and Rosi+Tram X. Y, Z Bar graphs showing changes in the percentage of the basal-most layer of the urothelium expressing Pparg Y and FABP4 Z after 1 month treatment with vehicle (n = 3), Rosi (n = 3), Tram (n = 3), and Rosi+Tram (n = 3). Data given as means ± SD, significance calculated using two-tailed unpaired Welch’s t-test. A′–D′. Expression of K14, Pparg, and FABP4 in basal tumors treated for 7 days with vehicle A′ or Rosi+Tram (B′–D′). E′. Quantification of percentage of lesions expressing FABP4 after 7 days of treatment with vehicle (n = 4) or Rosi+Tram (n = 4). Data given as means ± SD, significance calculated using two-tailed unpaired Welch’s t-test, p = 0.0001. Scale bars: 50 µm. Source data are provided as a Source Data file.

Basal and suprabasal layers of invasive BASQ tumors display different sensitivities to Rosi

The basal layer the urothelium both the urothelium and invasive tumors undergoes a shift from BASQ to luminal differentiation after treatment with both Rosi+Tram, while only the suprabasal layers respond to Rosi alone and undergo luminal differentiation, suggesting that there are fundamental differences between these populations that prevent Pparg expression and transcriptional activation. To address this question, we first examined the distribution of Pparg and its direct transcriptional target Fabp4, after treatment with vehicle, Rosi, Tram, or Rosi+Tram. Interestingly, Pparg activation was detected in nearly 67% of invasive tumors in Rosi+Tram treated mice compared to controls treated with vehicle, where neither Pparg nor Fabp4 were detected (Fig. 2U–E′). In BASQ tumors treated with Rosi alone, Pparg and Fabp4 expression were co-localized in upper layers of invasive tumors, indicating a response to Rosi treatment, however few cells in the basal-most layers expressed Pparg or Fabp4 (Fig. 2V). In contrast, treatment with either Tram alone or Tram+Rosi induced Pparg activation in basal-most layers of invasive tumors (Fig. 2W, X, B′–E′), indicating that Tram sensitized these basal-most cells, which are Krt14+, proliferative and invasive to Pparg activation. Similar changes were observed as early as 7 days after administration of Rosi+Tram, suggesting that the shift from BASQ to luminal differentiation occurs relatively early after treatment (Fig. 2A′–E′). These findings suggest that suprabasal and basal layers of both the urothelium and invasive tumors contain different cell populations with different sensitivities to Rosi.

Bladder cancers display high levels of MEK/ERK signaling down-stream from EGFR, ERBB2, K-RAS and BRAF^58–60. MEK/ERK signaling can negatively regulate Pparg activity by inducing phosphorylation of serine 273 (S273) in the Pparg ligand binding domain that can lead to its degradation⁵⁵. Based on gene set variation scores, the MEK/ERK signaling pathway is highest in BASQ tumors compared to other subtypes (Supplementary Fig. 1A). The MEK/ERK signaling pathway can be activated down-stream of RTKs including EGFR which, based on analysis of TCGA-BLCA data, is also high in BASQ tumors and low in luminal and neuronal subtypes. On the other hand, PPARG expression is high in luminal tumors, and low in BASQ and neuronal subtype BC (Supplementary Fig. 1B). Furthermore, pERK, a down-stream target of the MEK/ERK pathway is also high in BASQ tumors from humans and mice, and low in papillary tumors (Supplementary Fig. 1C–E).

These observations suggest that absence of Pparg signaling in basal layers of Rosi-treated tumors may be linked to MEK/ERK dependent phosphorylation. Phosphorylation of Pparg S273 is associated with changes in a set of genes including Adiponectin, Adipsin, Aplp2, Car3, Cidec, Ddx17, Nr1d1, Nr1d2, Nr3c1, Peg10, Rarres2, and Selenbp1 which are up-regulated, and Gdf3 which is down-regulated in cells in response to S273 dephosphorylation⁵⁶. RNA-seq analysis of Rosi+Tram treated BBN-induced tumors compared to vehicle treated tumors revealed alterations in eight of twelve signature genes after 5 days or 1 month of treatment. These include Aplp2, Car3, Nr1d1, Nr1d2, Nr3cs, Peg10, and Selenbp1, which were increased in response to Rosi+Tram treatment, as well as Gdf3 which was decreased [Supplementary Fig. 1E⁵³].

To further confirm Tram-induced changes in Pparg phosphorylation, we directly assessed the distribution of S273-phosphorylated Pparg in serial sections of tumors using an antibody specific for the serine 273 phosphorylation site (Pparg pS273). This analysis revealed widespread expression of the phosphorylated from of Pparg in control tumors treated with vehicle, but little expression in tumors after Rosi+Tram or Tram alone (Supplementary Fig. 1G, L–O). Together these observations suggest that MEK/ERK dependent phosphorylation of S273 may be important for suppressing Pparg activity in BASQ tumors, which can be alleviated by inhibition of MEK/ERK signaling. The observation that basal and suprabasal populations in BASQ tumors have distinct response to Rosi and Tram suggests that they may contain distinct cell types.

Rosi and Tram treatment drives apoptosis and cell cycle exit in BBN-induced tumors

The dramatic decrease in tumor volume (91%) observed after 1 month of Rosi+Tram treatment suggests that the 2-drug combination may induce cell death in BASQ tumors. To address this question, we analyzed tumors from BBN-exposed mice 1 day, 4 days, and 7 days after Rosi+Tram administration for expression of activated caspase-3, a marker expressed in cells undergoing apoptosis. These studies revealed large numbers of caspase-3-expressing tumor cells in BBN-induced BASQ tumors after 7 days of Rosi+Tram treatment (Fig. 3A–E) as well as a marked decrease in proliferation evidenced by decreased numbers of Ki67-positive cells (Fig. 3F).

Fig. 3. Rosi/Tram treatment induces apoptosis and reduces proliferation in BBN-induced BASQ tumors after 7 days.

A, B Expression of activated Caspase-3 in tumors of mice exposed to BBN for 5 months then treated with vehicle for 7 days, image shown with A and without B Ecad. C, D Expression of activated Caspase-3 in 5 M BBN tumors treated with Rosi+Tram for 7 days, image shown with C and without D Ecad. E Bar graph showing the percentage of cells expressing activated caspase-3 in BBN-induced tumors treated for 7 days with vehicle (n = 3) or Rosi+Tram (n = 3). Data given as means ± SD, significance calculated using two-tailed Mann-Whitney test, ***p = 0.0001. F Bar graph showing the percentage of cells expressing Ki67 in BBN-treated mice after vehicle (n = 3) or Rosi+Tram (n = 3) for 7 days. Data given as means ± SD, significance calculated using two-tailed Mann-Whitney test, ****p < 0.0001. G Apoptosis assayed by AnnexinV-eFluor450/7-AAD double staining in BBN963 cells treated with DMSO, Rosi, Tram, or combined Rosi+Tram for 72 h. Cells undergoing early apoptosis were AnnexinV+/7-AAD: late apoptosis: AnnexinV+/7-AAD+. H Heatmap showing normalized expression of proapoptotic genes, caspases, and pro-survival genes in Rosi+Tram treated BBN963 cells compared to controls based on RNAseq analysis. I–L Decreased proliferation determined by Ki67 staining in BBN963 cells after Rosi (n = 4) J, Tram (n = 4) K, or combined Rosi+Tram (n = 4) L compared to controls (n = 4) I. M–P Down-regulation of Ccnd1 in BBN963 cells after treatment with Rosi N, Tram O, or Rosi+Tram P compared to controls M. Scale bars: 50 µm. Q Heatmap of normalized gene expression showing down-regulation of Ccnd1 and genes that positively regulate cell cycle progression and up-regulation of genes that inhibit progression in BBN963 cells treated with Rosi+Tram for 72 h compared to DMSO treated controls. R Bar graph showing number of Ccnd1-positive BBN963 cells after 72-h treatment with DMSO (n = 4), Rosi (n = 4), Tram (n = 4), and combined Rosi+Tram (n = 4). Data given as means ± SD, significance calculated using two-tailed unpaired Welch’s t-test, ***p = 0.0006, ****p < 0.0001. Source data are provided as a Source Data file. Cartoons in Fig. 3 created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

To further investigate the individual and combined roles of Rosi and Tram on BASQ growth and survival, we performed in vitro and in vivo studies using BBN963 cells, a murine bladder cancer line derived from a BBN-induced BASQ tumor⁶¹. BBN963-derived orthotopic tumors and cultured cells express BASQ markers including Krt6a and Krt14 (Supplementary Fig. 2A–K). Importantly, analysis of orthotopic grafts and BBN963 cells in culture indicate that this cell line responds to Rosi+Tram in a similar way as BBN-induced BASQ tumors: we observed up-regulation of luminal markers such as Pparg and Upk and down-regulation of basal markers in both orthotopic tumor grafts and 2D cultures of BBN963 cells (Supplementary Fig. 2). Analysis of BBN963 cells by Annexin V staining revealed low numbers of cells in early/late apoptosis when treated for 72 h with DMSO or Rosi alone, while Tram treatment alone led to increased numbers of cells in late apoptosis (Fig. 3G). Combined treatment with Rosi+Tram resulted in further increases in the percentage of cells in both early and late apoptosis (Fig. 3G), suggesting that the 2-drug combination acts in a synergistic manner to induce cell death. Consistent with this, analysis of Rosi+Tram treated cultures revealed significant upregulation of pro-apoptotic members of the Bcl-2 family (Bad, Noxa, Bmf, Bik, Bid, Puma, Bak, and Bim), and downregulation of pro-survival members [(Bcl-2 and Bcl-X; Fig. 3H)]. Rosi+Tram also exhibited profound effects on the cell cycle (Fig. 3I–R, Supplementary Fig. 3A–H). RNA-seq analysis of after combined Rosi+Tram treatment in both BBN-induced mouse tumors and in BBN963 cells revealed decreased expression of Ccnd1 and other positive regulators of the cell cycle including Cdk1, Cdk4, and E2f2 along with increased expression of negative regulators of the cell cycle, including Rb, Cdkn1a and Cdkn2b (Fig. 3Q). These observations together suggest that Rosi+Tram treatment efficiently induced apoptosis and reduced proliferation in BASQ tumors in vivo and in vitro within one week of treatment.

RA-signaling Kdm6a and AP1 are differentially regulated in BASQ tumors by Rosi+Tram treatment

In addition to changes in survival and proliferation, Rosi+Tram treatment induced a luminal differentiation program in BASQ orthotopic tumors and in the urothelium of BBN-exposed mice. We first confirmed that our paired RNA-seq/ATAC-seq analysis detected changes in Pparg signaling after treatment with Rosi+Tram versus vehicle. Analysis 5 days after treatment revealed increased expression of Pparg-target genes in Rosi+Tram treated tumors, including Fabp4, Plin4, Cpt1a, and Cpt2 (Fig. 4A; Fig. Supplementary Fig. 4A, B). Cpt1a is directly regulated by Pparg and is a key enzyme in fatty acid oxidation⁶². Immunostaining confirmed decreased expression of Cpt1a in BBN-induced BASQ tumors compared to tumors treated with Rosi+Tram (Supplementary Fig. 4C, D). Pparg is a master regulator of adipogenesis and lipid storage, that controls lipid droplet formation in obesity, atherosclerosis, and fatty liver disease. In tumors treated with Rosi+Tram, we observed accumulation of neutral lipid droplets by Oil-red-O staining, which were not detected in controls (Supplementary Fig. 4E, F); similar alterations were previously observed in mice expressing a constitutively active form of Pparg in the urothelium⁶³.

Fig. 4. RosiTram treatment of BBN-induced tumors increases Kdm6a and retinoid signaling and decreases AP-1.

A Pathways changed in ATAC-seq/RNA-seq analysis after 1 M and 5 days of RosiTram vs vehicle. Bubble color/size corresponds to the p-value (p < 0.05) and number of genes enriched in pathways (GeneRatio), respectively. B Scatter plot showing differentially expressed/accessible genes after 5 days RosiTram vs vehicle. Plotted as Fold Change ATACseq RPM [Sample/Control] and Fold Change RNAseq RPM [Sample/Control], each dot is one gene. C IGV tracks showing gene expression and chromatin accessibility of retinoid signaling pathway genes in tumors treated for 5 days with RosiTram or vehicle. D Heatmap showing normalized gene expression of RA targets and RA synthesizing enzymes after 5 days of RosiTram treatment versus vehicle controls (from RNA-seq). E–L Expression of Aldh1a1 and Aldh1a2 in 5 M BBN tumors treated with vehicle (n = 3) E, I Rosi (n = 3) F, J, Tram (n = 4) G, K, and combined Rosi+Tram (n = 3) H, L. Scale bars: 50 µm M. Volcano plot showing upregulated (red) and down-regulated (blue) differentially expressed genes (DEGs) from RNAseq analysis of BBN963 cells after treatment with RosiTram (n = 5) vs DMSO controls (n = 5). Each dot is one DEG, colored points have a p-value ≤ 0.05. N–S Expression of Krt14, Kdm6a, and Fabp4 in BBN-induced tumors after treatment with vehicle (n = 3) N, Rosi (n = 3) O, Tram (n = 3) Q, or Rosi+Tram (n = 3) R, patient MIBC BASQ tumor (n = 2) P, and patient luminal papillary tumor (n = 3) S. Scale bars: 50 µm. T. HOMER motif analysis after ATAC-seq showing enrichment of the Junb regulatory network in less accessible peaks after Rosi+Tram for 5 days compared to vehicle. Heatmap showing normalized expression of AP-1 pathway genes after 5-day treatment with RosiTram compared to vehicle (bulk RNA-seq). U–X. Expression of Fosb U, V or JunB W, X after 7-day RosiTram or vehicle treatment of BBN-induced tumors. Scale bars: 50 µm. Y, Z Bar graphs quantitating the percentage of FosB Y and JunB Z-positive tumor cells after 7-day treatment with vehicle (n ≥ 3) or Rosi+Tram (n = 3). Data given as means ± SD, significance calculated using two-tailed unpaired Welch’s t-test, ****p < 0.0001. Source data are provided as a Source Data file.

In addition to increased Pparg signaling, Rosi+Tram treated tumors displayed increased expression of luminal markers (Fgfr3, Upk1b, Grhl3, Foxa1, Erbb2). These markers were low or undetectable in tumors treated with vehicle, which instead expressed high levels of BASQ markers (Fig. 4A, B). Interestingly, Rosi+Tram treatment induced expression of canonical targets of the retinoic acid signaling pathway including Rarb, a direct RA-target that harbors an RA-response element in its promoter³⁶, as well as Dhrs3 and Dhrs4, which are also known RA-targets (Fig. 4A–D, Supplementary Data 4). Pparg has been shown to up-regulate RA-signaling by inducing expression of RA-synthesizing enzymes such as retinol and retinaldehyde dehydrogenases that mediate the first and second steps in RA-synthesis, respectively³³. Analysis of BBN-induced tumors treated with Rosi+Tram revealed up-regulation of retinol and retinaldehyde dehydrogenases, including Rdh10 and retinaldehyde dehydrogenase family members Aldh1a1 and Aldh1a2 (Fig. 4B–D). Consistent with results from RNA-seq, IHC analysis of tumors treated with vehicle revealed little expression of Aldh1a1 or Aldh1a2, whereas expression of both genes was increased after treatment with Rosi+Tram (Fig. 4E–L). We obtained similar results with BBN963 cells: RNA-seq analysis after Rosi+Tram treatment revealed up-regulation of luminal markers including Fgfr3 and Upk2 and down-regulation of BASQ markers, consistent with the Rosi+Tram induced shift from BASQ to luminal differentiation. In addition, we observed increase in RA-regulated genes including Rarb (Fig. 4M, Supplementary Data 5).

In addition to upregulating the retinoid signaling pathway, Rosi+Tram treatment up-regulated expression of Kdm6a, a lysine demethylase that, based on recent studies, potentiates RA-signaling³⁹. Analysis of expression of Kdm6a in vehicle treated BBN-induced tumors revealed little or no detectable expression, while Kdm6a expression increased with Rosi or Tram alone (Fig. 4N, O, Q) and expanded to include most cells in residual tumors and the urothelium after treatment with both drugs (Fig. 4R). Analysis of the distribution of KDM6A in BASQ and LP human tumors revealed non-nuclear expression in suprabasal layers of BASQ MIBC, while nuclear expression of Kdm6a was widespread in LP tumors- these observations are consistent with the known role of Kdm6a as a promoter of luminal differentiation (Fig. 4P, S). Similar findings have been observed in mouse and human urothelial cells³⁹. Interestingly, despite the increase in Kdm6a, we did not observe changes in H3K27 trimethylation (Supplementary Fig. 5A–H), consistent with the observations from other groups that Kdm6a-regulated transcription does not always depend on its demethylase activity.

Pathway analysis to identify genes that were both less accessible and down-regulated after Rosi+Tram treatment reveals decreased accessibility and down-regulation of Tram targets (MAPK1/MAPK3 and ERK) and Keratinocyte differentiation (Fig. 4A, Supplementary Data 2 and 3). HOMER motif analysis of less accessible sites after ATAC-seq analysis revealed enrichment of the JunB motif and down-regulation of AP-1 family members including Jun, Batf, Atf3, Fosb and Fosl1 (Fig. 4T). Supporting this, analysis of expression of Junb and Fosb in BBN-induced tumors revealed high expression of both genes in tumors treated with vehicle, while both Fosb and Junb were undetectable after treatment with Rosi+Tram (Fig. 4U–Z). Similar findings were obtained from RNA-seq analysis of Rosi+Tram treated BBN963 cells (Fig. 4M).

Pparg-signaling is known to directly inhibiting expression of Rela, a critical component of the Nf-kb complex^64–67 and inhibits immune infiltration in a luminal model of BC⁶³. Consistent with this, we observed decreased expression of Rela and other components of the Nf-kb signaling pathway after Rosi+Tram treatment (Fig. 4A), and HOMER motif analysis of less accessible sites after ATAC-seq analysis revealed enrichment of Rela motifs within the Nf-kb network, including Nfkb1 and Nfkb2 (Supplementary Fig. 6A), which are positively regulated by Rela. Consistent with the decrease in Nf-kb expression, interleukins, cytokines, MHC Class II, and genes important for T-cell signaling were also down-regulated after Rosi+Tram treatment (Fig. 4B; Supplementary Fig. 4B; Supplementary Data 1). IHC analysis of Rela and Cd45 expression confirmed these observations. We found significant expression of Rela and significant immune infiltration in untreated tumors that were reduced after Rosi or Tram treatment (Supplementary Fig. 6C–L). Rela was undetectable after Rosi+Tram treatment and Cd45 staining was greatly reduced, consistent with the state of low immune infiltration associated with luminal tumors (Supplementary Fig. 6K–L).

Retinoid signaling controls basal vs luminal differentiation in the urothelium and in BC

Retinoids are required for urothelial development and differentiation in adults, regulating formation of luminal cell types (intermediate and superficial cells) in the urothelium^28,29. Vitamin A deficiency, which decreases retinoid signaling, induces keratinized squamous differentiation in the urothelium, which is replaced by a cornified sheet resembling skin^30,31. We observed a similar effect in the urothelium of mice expressing a dominant inhibitory form of Rara that globally suppresses RA-signaling^29,68,69. The observation that retinoid signaling is required for formation of luminal cell types in the urothelium, suppresses squamous differentiation, and is up-regulated in tumor cells undergoing a BASQ to luminal shift in differentiation suggests that retinoids may be important down-stream of Pparg for promoting the luminal differentiation program. To address this question, we used BBN963 cells⁶¹ to assess the effects of RA on tumor differentiation and subtype. To generate tumors, BBN963 cells were introduced into bladders of syngeneic mice by ultrasound guided injection. After 2 weeks of tumor cell grafting, the mice were dosed with RA or vehicle (DMSO) via oral gavage for 7 days and then analyzed to assess the effects on tumor marker expression, proliferation, and growth. This analysis revealed up-regulation of luminal markers (Fig. 5A, C) and decreased proliferation in tumors treated with RA compared to controls (Fig. 5B, D, F). RNA-seq analysis confirmed these observations and revealed up-regulation of targets also that were also increased by Rosi+Tram treatment, including RA-regulated genes (Rarb, Cyp26b1, Dhrs3, Rbp4, Stra6), luminal markers (Fgfr3, Krt19, Krt20, Upks) and Kdm6a (Fig. 5E, G, Supplementary Data 6). Down-regulated genes after RA treatment included BASQ markers (Krt14, Krt16, Krt6a, Cd44) which were also down-regulated after RosiTram treatment (Fig. 4E, G). Interestingly, both RA and Rosi+Tram treatment resulted in decreased expression of AP-1 family members, including Fosb, Fosl1 and Fosl2. AP-1 has long been known to be directly inhibited by RA-receptor signaling^37,38, and recent studies indicate that AP-1 promotes BASQ differentiation in the urothelium and in BC^39,40. These observations suggest that Pparg-induced BASQ-to-Luminal differentiation is likely to be driven at least in part by RA-signaling which directly inhibits AP1 and is likely to be facilitated by Kdm6a.

Fig. 5. Retinoic acid induces BBN 963 cells to undergo a Basal-to-luminal shift in differentiation and lowers proliferation.

A, B Ki67 and Foxa1 staining in an orthotopic tumors from BBN963 cells engrafted into bladders of C57Bl/6 mice after treated with vehicle for 7 days. C, D Ki67 and Foxa1 staining in an orthotopic tumors from BBN963 cells engrafted into bladders of C57Bl/6 mice after treatment with RA for 7 days. Scale bars: 50 µm E. Heatmap showing normalized gene expression changes in BASQ and luminal markers after RA treatment of BBN963 cells compared to controls from RNAseq analysis. F Bar graph showing decreased proliferation (via percentage of Ki67+ cells) in BBN963 orthotopic tumors after treatment with RA (n = 4) versus vehicle (n = 4) for 7 days. Data given as means ± SD, significance calculated using two-tailed unpaired Welch’s t-test. G Volcano plot showing upregulated (red) and down-regulated (blue) differentially expressed genes (DEGs) from RNAseq analysis of BBN963 cells treated either with vehicle (n = 4) or RA (n = 4). Each dot is one DEG, colored points have a p-value ≤ 0.05. Source data are provided as a Source Data file. Cartoons in Fig. 5 created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

Discussion

The integration of neoadjuvant chemotherapy with radical cystectomy in the early 2000’s improved the survival of patients with MIBC, but little substantive progress has been made since that time. Immunotherapy likely has a role in the management of this disease, but the optimal integration of this modality into the care of patients with MIBC is still a work in progress^1,7–9. As such, there is still a pressing need to explore therapy targeting other biological vulnerabilities of MIBC.

Here we sought to test the efficacy of a combination of Rosiglitazone (Rosi), a Pparg synthetic agonist, and Trametinib (Tram), a MEK/ERK inhibitor, in preclinical models of MIBC. Treatment of BBN-induced BASQ tumors with Rosi or Tram alone resulted in decreased proliferation. Treatment with both drugs, however, induced apoptosis in BASQ tumors within 7 days of treatment and reduced overall tumor volume by 91% after one month. These changes were accompanied by decreased expression of the Bcl2 pro-survival pathway as well Ccnd1 and cell cycle regulators. In addition to killing tumors, the combined drug treatment exerted a powerful effect on cell specification, inducing a shift from BASQ to luminal differentiation in orthotopic tumors and restoring luminal differentiation in the urothelium of tumor bearing mice, which had undergone squamous differentiation. These changes were accompanied by up-regulation of the retinoid signaling pathway and Kdm6a and down-regulation of AP-1. RA-signaling, which is a major regulator of luminal differentiation, is induced down-stream of Pparg and can independently drive luminal differentiation in BASQ tumors. Our studies suggest that Kdm6a facilitates RA signaling to drive the luminal differentiation program down-stream of Pparg while inhibition of AP-1 suppresses BASQ differentiation. These studies suggest that the combination of Rosi+Tram and retinoids, all of which are FDA approved, may be effective in a clinical setting for treatment of MIBC.

Basal and suprabasal layers of BASQ tumors are populated by distinct cell types with different sensitivities to Pparg activation

Our studies support a model in which invasive BASQ tumors are composed of distinct basal and suprabasal cell populations, as is the case for many epithelial structures, that leads to differential sensitivities to Rosi/Tram treatment. Pparg signaling, evidenced by expression of Pparg and Fabp4, a direct transcriptional target, is not detected in BASQ BBN-induced tumors treated with vehicle, while treatment with Rosi+Tram induced Pparg/Fabp4 expression in 67% of T0/T1 tumors which was mainly restricted to suprabasal layers (Fig. 2U′–E′). As in the urothelium, we noted an increased number of cells with activated Pparg signaling in basal-most layers of invasive tumors treated with Rosi+Tram or Tram alone compared to Rosi alone (Fig. 2U–E′), suggesting that these K14+ basal-cells which are proliferative and invasive, are refractory to Rosi treatment. That Tram-induced MEK/ERK inhibition led to increased Pparg and Fabp4 expression in basal-most layers of invading tumors, indicates that Pparg signaling is likely to be suppressed down-stream of the MEK/ERK pathway in these cells. Rosi treatment has been shown to induce Pparg resistance in the context of Type 2 Diabetes by a similar mechanism, and several studies suggest that MEK/ERK-dependent phosphorylation at Pparg serine 273 (S273) leads to suppression of Pparg signaling which can lead to degradation^53–55. Supporting this mode of action, analysis of BASQ tumors with an antibody specific for phosphorylated Pparg (Pparg p-S273) shows lower expression of pS273 Pparg in Tram treated BASQ tumors compared to untreated BASQ tumors (Supp Fig. 1H–O).

Based on scRNA-seq analysis, the basal layer of BASQ tumors is an invasive front, a common feature of epithelial tumors⁷⁰, populated by proliferative cells expressing EMT markers⁷¹. The basal layer in BBN-induced tumors in mice also contains tumor-forming progenitors⁷² which are Krt14-positive⁷³. The observation that different layers of BASQ tumors can be populated by cells with different drug sensitivities is an important consideration in designing cancer therapeutics in order to ensure robust response and prevent cancer relapse.

Pparg activation up-regulates RA signaling by inducing expression of RA-synthesizing enzymes

Our studies show that a combination of Pparg activation and MEK inhibition induced by Rosi+Tram induce a BASQ-to-luminal shift in differentiation (Fig. 6). We find that the shift to luminal differentiation is accompanied by up-regulation of RA-signaling, a pathway that is not expressed in BASQ tumors and normally drives luminal differentiation in the embryonic and adult urothelium²⁹. Activation of RA-signaling in Rosi+Tram-treated BASQ tumors is most likely driven by expression of RA-synthesizing enzymes that produce RA locally and are required for transcriptional activation of RA-receptors³³. We observe up-regulation of the RA-signaling pathway, evidenced by expression of canonical RA-targets, including Rarb, which is accompanied by increased expression of RA-synthesizing enzyme that are required for activation of RA-signaling. We observe increased expression of Rdh10, a retinol dehydrogenase that mediates the first step in RA-synthesis^74,75, and two of three retinaldehyde dehydrogenases (Aldh1a1 and Aldh1a2) that mediate the second step in RA synthesis^76,77. Pparg activation has been shown to be a driver of RA signaling in a number of contexts including dendritic cells, where Rosi treatment induced the RA-signaling pathway driven by expression of RA-synthesizing enzymes Rdh10 and Aldh1a1⁷⁸, findings that are quite similar to ours (Fig. 4B–L). Rosi+Tram also induced expression of Kdm6a, a lysine demethylase that is undetectable in untreated BASQ tumors (Fig. 4M–R). Kdm6a, which associates with the Mll3/4 complex, is known to positively regulate RA-signaling, recruiting the Mll3/4 complex to promoters and enhancers where Rar/Rxr heterodimers are bound⁷⁹. Based on recent studies, Kdm6a induces a luminal differentiation in urothelial cells and is important both for maintaining RA-signaling and suppressing the activity of AP-1, a pathway required for differentiation of the epidermis that promotes BASQ differentiation in the urothelium³⁹. These observations together suggest that Rar signaling in Rosi+Tram treated BASQ tumors may be driving the shift from BASQ-to-luminal differentiation, most likely down-stream of Pparg activation; events facilitated by Kdm6a that are likely to be opposed by AP-1.

Fig. 6. Combined and single activities of Rosi+Tram regulate tumor survival and subtype.

Rosi or Tram alone decrease tumors size and growth, but combined Rosi + Tram treatment induces apoptosis of BASQ tumors 7 days after treatment. Tram helps maintain Pparg activity by preventing its phosphorylation, which can lead to degradation. Rosi induces Pparg activation, which leads to up-regulation of RA-synthesizing enzymes that activate the retinoid signaling pathway. Retinoids and Kdm6a cooperate to promote luminal differentiation, suppressing AP-1, a pathway that promotes BASQ differentiation. Source data are provided as a Source Data file. Cartoons in Fig. 6 created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en). Cartoons in Fig. 6 created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

Rosi, Tram, and retinoids as possible treatments for MIBC

Rosi (aka Avandia) has been extensively used in the past as an insulin sensitizer in patients with Type 2 Diabetes. Although the FDA restricted rosiglitazone prescriptions in 2010 due to concerns of increased risk of acute myocardial infarction and cardiac deaths, further evaluation of longitudinal data led to removal of these restrictions in 2013 and the Risk Evaluation and Mitigation Strategy (REMS) was eliminated in 2015⁴². Since then, Rosi has been considered to be safe for treatment of Type 2 Diabetes by the FDA but is rarely used to treat patients^41,43. Trametinib (Mekinist), is used to treat cancers including melanoma, non-small cell lung cancer, and anaplastic thyroid cancer^47–49. RA (Tretinoin) is also FDA approved and is used as a differentiation therapy for the treatment of acute promyelocytic leukemia as well as for the prevention of cutaneous squamous cell carcinoma and cutaneous Kaposi’s sarcoma^44–46. Our studies suggest that these FDA approved drugs may be used for treatment of basal/squamous MIBC in a clinical setting.

Methods

Mice

All work with mice was approved by and performed under the regulations of the Columbia University Institutional Animal Care and Use Committee. Animals were housed in the animal facility of Irving Cancer Research Center, Columbia University. Animals were housed in standard cage of 75 square inches at or below maximum cage density permitted by IACUC protocol. Temperature was maintained between 68 and 79 °F. Humidity was maintained between 30 and 70%. A timed-controlled lighting system was used for a uniform diurnal lighting cycle. Wild-type Swiss Webster mice were obtained from Taconic Biosciences (Tac:SW). K5CreERT2^fl/fl mice (FVB.Cg-Tg(KRT5-cre/ERT2)2Ipc/JeldJ)3233 were obtained from D. Metzger and P. Chambon and Rosa26mTmG reporter mice (Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J)⁵² mice from Jackson Laboratory (stock #007576).

Cell lines

BBN963 cell line (W. Kim, University of North Carolina at Chapel Hill) were cultured inside tissue culture-treated plates with Dulbecco’s modified Eagle medium high glucose (DMEM) with 10% fetal bovine serum (Gibco) and 1% penicillin-streptomycin (100 U/ml; Gibco) at 37 °C and 5% CO₂. Cells were split twice per week, and cell viability was measured using trypan blue staining in the Countess II automated cell counter (Thermo Fisher Scientific).

Orthotopic model

BBN963 cells were detached from tissue culture plates using 0.25% trypsin- EDTA at 37 °C. The cells were injected orthotopically under ultrasound guidance into the bladder lamina propria of C57BL/6 J male mice at 5 × 10⁶ cells suspended in 500 uL sterile PBS using 30 G needle with syringe. Mice between 8 and 10 weeks of age were used in these experiments.

BBN treatment

BBN (0.05%; Sigma cat#B8061-1G) was administered in the water supply ad libidum for up to 24 weeks to induce bladder cancer. Mice between 8 and 10 weeks of age were placed on BBN. Mice were euthanized between 20 and 24 weeks. All bladders were removed and embedded for sectioning and staining.

Rosiglitazone and Trametinib treatment

In vivo: Tumor growth was confirmed via ultrasound (Visualsonics VEVO 3100), and wild type male mice were divided into four treatment groups as follows: Group 1: Control treated with vehicle (0.5% hydroxypropyl methylcellulose and 0.2% Tween 80 in distilled water and 0.7% DMSO). Group 2: Rosiglitazone (Adipogen CAS# 122320-73-4; 20 mg/kg) dissolved in vehicle. Group 3: Trametinib (LC laboratories T-8123; 0.3 mg/kg) dissolved in vehicle. Group 4: Rosiglitazone (20 mg/kg) and Trametinib (0.3 mg/kg) dissolved in vehicle. Mice were treated daily via oral gavage for 4 days, 5 days, 7 days, or 1 month. Bladders were removed for embedding, RNAseq, or ATACseq.

In vitro: BBN963 cells were seeded at density of 20,000 cells/cm² and treated for 24 or 72 h. The cells were divided into four treatment groups as follows: Group 1: DMSO control. Group 2: Rosiglitazone (2 µM). Group 3: Trametinib (100 µM). Group 4: Rosiglitazone (2 µM) and Trametinib (100 µM). Triplicates of each treatment group were performed.

All-trans retinoic acid treatment

In vitro: BBN963 cells were seeded at density of 20,000 cells/cm². DMSO control or ATRA (1 µM) (Sigma Cas# 302-79-4) dissolved in DMSO were added for 24 or 72 h. Triplicates of each treatment group were performed.

Ultrasound

Ultrasound images and scanning were acquired using VEVO 3100 Ultrasound Imaging System (FUJIFILM VisualSonics, Toronto, Canada) located within the mouse barrier in the Herbert Irving Cancer Center Small Animal Imaging facility. Tumor volume was calculated via 3D reconstruction program (Vevo LAB).

Apoptosis assay

The BBN963 cells were harvesting using 0.25% trypsin-EDTA at 37 °C. AnnexinV/7-AAD detection was performed according to manufacturer’s instructions (eBioscience 88800672). Flow cytometry was performed on Sony MA900 Multi-Application Cell Sorter.

RNA-Seq

For in vivo experiments, single cell suspensions from healthy control, vehicle control, and RosiTram treated bladders were filtered through a 70 μm filter (Fisherbrand cat#22363548) and then sorted on a BD Aria II Cell Sorter using a 130 μm nozzle aperture and 13 psi pressure to collect DRAQ5 + , DAPI-negative live urothelial cells. Gating strategy was performed on BD FACSDiva Software v 8.0. Cells were then centrifuged at 500 x g for 30 min at 4 °C. The supernatant was discarded, and the pellet was processed for total RNA extraction. Samples with a RIN (regulation identification number) >8 were used for RNA-seq. The libraries were prepared using the SMART- Seq® v4 Ultra® Low Input RNA Kit for Sequencing (TaKaRa) followed by Nextera XT (Illumina), both according to manufacturer’s instructions. They were sequenced to a targeted depth of 40 M 2x100bp reads on a NovaSeq 6000 (Illumina). For in vitro experiments, total RNA was extracted and samples with a RIN > 8 were used for RNA-seq. The libraries were prepared using Illumina Stranded mRNA prep (Illumina 20040532) according to manufacturer’s instruction. They were sequenced to a targeted depth of 400 M 2x75bp reads on a NextSeq 550 (Illumina).

RNA-seq data analysis

Differential expression analysis was performed by reading kallisto counts files into R using the R packages tximport (v.1.10.1) and biomaRt (v.2.34.2) and running DESeq2 (v.1.18) to generate log fold change values and p-values between the two experimental groups. The heatmap and PCA plots were visualized after transforming the counts using VST (variance stabilizing transformation). Gene set analysis by ConsensusPathDB⁸⁰ were used to identify significantly changed pathways.

ATAC-seq

Chromatin accessibility assays utilizing the bacterial Tn5 transposase were performed as described⁸¹ with minor modifications. Cells (3.0 × 10⁵) were lysed and incubated with transposition reaction mix containing PBS, 1% Digitonin, Tween-20, and Transposase (llumina). Samples were incubated for 30 min at 37 °C in a thermomixer at 1000 rpm. Prior to amplification, samples were concentrated with the DNA Clean and Concentrator Kit-5 (Zymo). Samples were eluted in 20 uL of elution buffer and PCR-amplified using the NEBNext 2X Master Mix (NEB) for 10 cycles and sequenced on a NextSeq 500 (Illumina).

ATAC-seq data analysis

ATAC-seq reads were mapped to the mouse genome assembly mm10 using HISAT2 (v2.1.0, parameter: -X 2000). Potential PCR duplicates were removed by the function “MarkDuplicates” (parameter: REMOVE_DUPLICATES=true) of Picard (v2.23.1). The correlation analysis for genomic distribution of ATAC-seq signals was performed by the functions “multiBigwigSummary” (parameter: --binSize 1000) and “plotCorrelation” (--corMethod pearson -- skipZeros) of deepTools (v3.3.2). Peaks of ATAC-seq data were called using macs2 (v2.1.2, default parameters) and were annotated by the R package “ChIPseeker”. The distribution of ATAC-seq signals near promoter regions were visualized by the functions “computeMatrix” and “plotProfile” of deepTools (v3.3.2). The reads number for each peak was measured by featureCounts (v1.6.1). The differential accessibility of promoters was calculated by the R packages DESeq2 (v1.28.0) and visualized by ggplot2 (v3.2.1).

Statistics and reproducibility

All quantitation was performed on at least three independent biological samples, using the ImageJ software. Data presented in bar graphs are mean values ± S.D. Statistical analysis was performed using the R version 4.0.4 and GraphPad Prism 10.2.3. In two group comparisons, statistical significance was determined using unpaired two-tailed Welch's T-tests, as well as ANCOVA with Dunnet's test which was used to assess the effects of RosiTram treatment on tumor growth in BBN-induced mice. The number of samples used in the experiments is included in figure legends. All immunostainings and H&E experiments were performed with at least 3 biological replicates and 3 technical replicates for each condition.

Immunostaining

Bladders were embedded in paraffin and serial sections were generated. For immunohistochemistry, paraffin sections were deparaffinized using HistoClear and rehydrated through a series of Ethanol and 1× PBS washes. Antigen retrieval was performed by boiling slides for 15 min in pH 9 buffer or 30 min in pH 6 buffer. Primary antibodies in 1% horse serum were incubated overnight at 4 °C. The next day, slides were washed with PBST three times for 10 min each and secondary antibodies were applied for 90 min at room temperature. DAPI (4′,6- diamidino-2-phenylindole) was either applied as part of the secondary antibodies cocktail or for 10 min, for nuclear staining, and then the slides were sealed with coverslips. Conditions of antibodies used is detailed in Table below.

Fluorescent microscopy

Zeiss Axiovert 200 M microscope with Zeiss Apotome were used to collect immunofluorescent images. Bright-field images were collected using a Nikon Eclipse TE200 microscope. Data were analyzed using the Fiji package of ImageJ (v.1.0) and Photoshop screen overlay method (v. 21.1.0).

Bioinformatic analysis and visualization of PPARG and EGFR mutations in bladder cancer

This analysis was performed with R packages including “ggplot2”95 to generate jitter plots, “ComplexHeatmap” 96 and “circlize”97 to generate heatmap and “GSVA” 98 to calculate GSVA enrichment scores. Gene sets for BIOCARTA_MAPK_PATHWAY and BIOCARTA-ERK- PATHWAY were downloaded from molecular signature database in GSEA website.

Antibodies	Source	Identifier	Clone	Dilution
Chicken Polyclonal Anti- Keratin 14	Biolegend	Cat#906001	N/A	1:400
Chicken Polyclonal Anti- Keratin 5	Biolegend	Cat#905901	N/A	1:400
Rabbit Polyclonal Anti- Krt6A	LSBio	Cat#LS-B12036-100	N/A	1:2000

Rabbit Polyclonal Anti- Keratin 10	Covance	Cat#PRB-159P	N/A	1:500
Mouse Monoclonal Anti- Cytokeratin 20	AgilentDako	Cat# M701929-2	Ks20.8	1:200
Rabbit Polyclonal Anti-PPARG	Cell Signaling Technology	Cat# 2435	N/A	1:200
Goat Polyclonal Anti- FABP4	R&D Systems	Cat# AF1443	N/A	1:1000
Mouse Monoclonal Anti- FOXA1	Seven Hills Bioreagents	Cat# WMAB-2F83	2F83	1:1000
Rabbit Polyclonal Anti-p63	GeneTex	Cat#GTX102425	N/A	1:300
Goat Polyclonal Anti-p63	R&D Systems	Cat#AF1916	N/A	1:200
Rat Monoclonal Anti-CD45	BD Bioscience	Cat#550539	30F11	1:100
Rabbit Polyclonal Anti- Ki67	Abcam	Cat#ab15580	N/A	1:200
Rabbit Monoclonal Anti- Cyclin D1	Abcam	Cat#ab16663	SP4	1:200
Rabbit Polyclonal Anti- Active Caspase-3	Promega	Cat#G7481	N/A	1:300
Goat Polyclonal Anti-E- Cadherin	R&D Systems	Cat#AF748	N/A	1:400
Rabbit Polyclonal NFKB p65 (Rela)	Abcam	Cat#AB19870	N/A	1:300
Alexa Fluor 488 Donkey Anti-Rabbit IgG	Jackson Immunoresearch	Cat#711-545-152	N/A	1:700
Alexa Fluor 488 Donkey Anti-Mouse IgG	Jackson Immunoresearch	Cat#711-545-150	N/A	1:700

Alexa Fluor 488 Donkey Anti-Chicken IgG	Jackson Immunoresearch	Cat#703-545-155	N/A	1:700
Alexa Fluor 488 Donkey Anti-Goat IgG	Jackson Immunoresearch	Cat#705-545-003	N/A	1:700
Cy3 Donkey Anti-Rabbit IgG	Jackson Immunoresearch	Cat#711-165-152	N/A	1:700
Alexa Fluor 594 Donkey Anti-Mouse IgG	Jackson Immunoresearch	Cat#715-585-151	N/A	1:700
Alexa Fluor 594 Donkey Anti-Chicken IgG	Jackson Immunoresearch	Cat#703-585-155	N/A	1:700
Alexa Fluor 594 Donkey Anti-Goat IgG	Jackson Immunoresearch	Cat#705-585-147	N/A	1:700
Alexa Fluor 647 Donkey Anti-MouseIgG	Jackson Immunoresearch	Cat#715-605-150	N/A	1:400
Alexa Fluor 647 Donkey Anti-Rabbit IgG	Jackson Immunoresearch	Cat#711-605-152	N/A	1:400
Alexa Fluor 647 Donkey Anti-Chicken IgG	Jackson Immunoresearch	Cat#703-605-155	N/A	1:400
Alexa Fluor 647 Donkey Anti-Goat IgG	Jackson Immunoresearch	Cat#705-605-003	N/A	1:400

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, Peptides, and Recombinant proteins
N-butyl-N-(4-hydroxybutl) nitrosamine	Sigma	Cat#B8061-1G
Hank’s Balanced Salt Solution (HBSS)	ThermoFisher Scientific	Cat#14170-112
Bovine Serum Albumin	Sigma	Cat#A2058

Bacillus licheniformis protease	Sigma	Cat#P5459
CaCl2	Sigma	Cat#21115
DNAse I recombinant	Sigma	Cat#4716728001
Antigen unmasking solution	Vector Labs	Cat#H3300
Citrate buffer	Thermo Fisher Scientific	Cat#AP-9003-500
Horse Serum, heat inactivated	Gibco	Cat#26050070
Tween80	Sigma	Cat#P1754
DMEM/F12	Thermo Fisher Scientific	Cat#H7904
Critical commercial assays
SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing	TaKaRa	Cat#634889
Illumina stranded mRNA Prep	Illumina	Cat# 20040532
Annexin V Apoptosis Detection Kit	eBioscience	Cat#88-8007-72
Software and Algorithms
R version 4.0.4	R Core Team, 201682	http://www.r-project.org/
DESeq2	Love et al., 201483	https://github.com/mikelove/ DEseq2
Gene Set Enrichment Analysis, v4.1.0	Subramanian et al., 200584	https://www.gsea- msigdb.org/gsea/index.jsp
Other
VEVO 3100Ultasound Imaging System	FUJIFILM VisualSonics	N/A

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

S.A.P. performed RNA-seq analysis of Rosi+Tram and RA-treated BBN963 cells and tumors, immunostaining as well as statistical analysis and helped write the manuscript. T.T. conceived the idea behind this study and performed the first round of experiments. T.T. also prepared RNAseq samples from the 1 M RosiTram samples. H.A. and M.B. performed pathological grading of tumors; X.C., K.G., A.D.V., C.L. and B.A. helped with ATAC-seq data analysis and assisted in preparing sequencing figures; A.M. and G.W. performed experiments related to the RaraDN mutant; E.B. and J.L. performed immunostaining; W.Y. helped with analysis TCGA data analysis and B.L helped with statistical analysis. J.M., B.G, C.D, D.J.M. provided advice on studies and manuscript; C.L.M. helped design experiment, interpret results, and wrote the paper.

Peer review

Peer review information

Nature Communications thanks Joshua Meeks, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The RNAseq and ATACseq data generated in this study have been deposited in the GEO database under accession code GSE234327. Source data are provided with this paper.

Competing interests

C.D. is creator of intellectual property owned by UT/MDACC related to the use of genetic alterations as a predictive biomarker for response to Nadofaragene firadenovec and has stock options and personal compensation from CG Oncology for consulting and advisory services. C.L.M. has performed studies in collaboration with FLARE Therapeutics focused on PPARG inverse agonists in NMIBC. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-50678-2.