1. Introduction

Globally, bladder cancer (BC) stands as the tenth most prevalent form of cancer, impacting individuals across all gender identities. In the year 2020 alone, it was calculated that BC was the diagnosis given in ~573,000 new cancer cases while also being responsible for roughly 12,000 deaths [1]. The tumor microenvironment (TME) in BC is characterized by its dynamic nature, marked by significant alterations and complex intercellular dialogues during phases of tumor development, metastatic expansion, and responses to treatments. Consequently, BC is characterized as a complex biological network, inclusive of an array of nonmalignant cells and their intricate interactions embedded within the tumor matrix [2]. Despite this complexity, predominant research efforts have been focused on exploring the genetic and epigenetic modifications occurring within the urothelial cell clusters. There arises an unequivocal need to delve into the influence exerted by the stromal segment of the TME on the evolution and progression of BC. This amplified focus on the cellular and molecular activities within the urothelial cells has inadvertently led to a gap in understanding the comprehensive landscape of BC. A nuanced exploration into the roles and impacts of the stromal compartment, especially in the context of its interactions with cancerous cells and contributions to the TME’s dynamism, could unveil novel insights into BC pathophysiology and potential therapeutic interventions.

Mesenchymal stem cells (MSCs) are noted for their distinct ability to home to various tumor types, significantly influencing tumor progression [3, 4]. In colon cancer, MSCs enhance stemness and epithelial–mesenchymal transition through interleukin-8 (IL-8)/CXC chemokine receptor 2 (CXCR2)/mitogen-activated protein kinase (MAPK) and fibroblast growth factor 10 (FGF10)–protein kinase A (PKA)–protein kinase B (Akt)–β-catenin signaling pathways [5]. A substantial body of evidence, supported by both in vitro and in vivo studies, confirms the role of MSCs in engendering resistance to chemotherapy and radiotherapy [6, 7]. Mitochondrial transfer from adipose MSCs to cancer cells occurs via tunneling nanotubes and contributes to multidrug resistance [8, 9]. Carcinoma-associated MSCs (CA-MSCs) further enhance ovarian cancer metastasis by donating mitochondria to “mito-poor” tumor cells, thereby restoring their proliferative capacity and chemotherapy resistance [10]. Additionally, the activation of CXCR1/Akt signaling induced by MSC-derived IL-8 promotes anoikis resistance and pulmonary metastasis in osteosarcoma cells [11]. Furthermore, MSC-derived RAB22A-induced extracellular vesicles (EVs) containing activated stimulator of interferon genes (STING) can effectively induce antitumor immunity by promoting interferon beta (IFNβ) expression in monocytes and enhancing immune responses [12]. The presence of MSCs has been confirmed not only in human bladder tissue but also in urine, highlighting their pervasive nature [13–15]. Conditioned media from MSCs are rich in chemokine ligand 2 (CCL2)/monocyte chemoattractant protein-1 (MCP-1), matrix metalloproteinases (MMPs), and ILs, which promote tumor aggressiveness and chemoresistance in BC cells [16]. In light of these findings, delving into the implications of stromal MSC-associated factors within the sphere of BC is deemed essential.

Targeting the TME through innovative treatments is recognized as a promising strategy in combating cancer. This approach is advantageous because it acts on genetically stable cells, mitigating the risk of developing drug resistance [17]. Prior research has illuminated the potential of inhibiting stromal components within the TME to arrest tumor advancement [18–20]. A holistic and system-level approach, encompassing diverse therapeutic targets within the TME, is essential for optimizing antitumor results [17]. However, the field lacks extensive studies on identifying and developing therapeutic targets specifically for MSCs.

In the present study, a unique five-gene MSC signature was developed. The constructed model exhibited predictive capabilities for both prognosis and therapeutic responses. Meanwhile, the effective drugs that targeted MSC-related genes were screened by molecular docking. Evidence indicates that this MSC-based model could pave the way for new therapeutic interventions targeting MSCs within the BC landscape.

2. Materials and Methods

3. Results

3.1. The Association Between Higher MSC Infiltrations and Stromal Scores and Poorer OS in BC Patients

The flowchart of the study is illustrated in Figure 1. MSC infiltration was assessed using xCellAnalysis, while the stromal score was determined using ESTIMATE. The predictive utility of these measures for OS was evaluated utilizing log-rank tests. Kaplan–Meier curves in TCGA-BLCA cohort (Figure 2A,B) and GSE31684 cohort (Figure 2C,D) showed a negative correlation between OS and higher MSC infiltration and stromal scores. Given these observations, an in-depth analysis of genes associated with MSC and stromal characteristics in BC was imperative.

3.2. The Coexpression Network Analysis of MSC and Stromal Scores

WGCNA analysis was conducted to dissect gene expression frameworks in the TCGA-BLCA and GSE31684 datasets. Specific soft threshold powers (four for TCGA-BLCA and eight for GSE31684) were applied to construct scale-free topology networks (Figures 3A,B). Hierarchical clustering trees revealed 11 unique coexpression patterns for TCGA-BLCA (Figure 3C), wherein the blue module exhibited a noteworthy positive association with MSC proportion (correlation = 0.37, p = 2e−14) and the stromal score (correlation = 0.90, p = 4e−147) (Figure 3E). Concurrently, the analysis of GSE31684 yielded 10 coexpression patterns (Figure 3D), with the brown module drawing attention for its robust positive association with MSC proportion (correlation = 0.45, p = 6e−6) and stromal score (correlation = 0.92, p = 1e−39) (Figure 3F). Encapsulated within these modules were 1081 genes for the blue and 560 for the brown, necessitating an in-depth analysis. This investigation of scatter plots pertaining to the blue module underscored stark correlations between MM and GS for MSC (correlation = 0.56, p = 2.9e−90) and stromal scores (correlation = 0.95, p < 1e−200) (Figure 3G). A correspondingly potent association was discerned in the brown module concerning MSC (correlation = 0.69, p = 6.6e−92) and stromal scores (correlation = 0.96, p < 1e-200) (Figure 3H). The strategic application of MM > 0.8 and GS > 0.4 as threshold parameters led to the recognition of 376 and 380 pivotal genes within the blue and brown modules of TCGA-BLCA and GSE31684, respectively. These genes are significantly intertwined with both MSC and stromal scores.

3.3. Enrichment Analyses Conducted on Hub Genes

A Venn diagram reveals the identification of 133 shared genes at the intersection of two distinct hub gene sets (Figure 4A). GO and KEGG analyses were performed on the 133 identified genes, revealing significant enrichment in BPs related to ECM organization, external encapsulating structure organization, and extracellular structure organization. Notably, collagen-containing ECM and ECM structural constituents were the major enriched terms in CCs and MFs, respectively (Figure 4B). The KEGG pathway analysis revealed significant enrichment in protein digestion and absorption, ECM–receptor interaction, and focal adhesion (Figure 4C).

3.5. Strong Correlation Was Observed Between MSC Signature Genes and MSC Markers

To validate the robustness of the MSC model as a predictor of MSC infiltrations, we conducted Spearman’s correlation analyses between the MSC risk score and both the stromal score and MSC abundances predicted by xCell. The results consistently demonstrated a strong correlation between the MSC risk score, MSC infiltrations, and the stromal score across the TCGA-BLCA (Figure 5A) and GSE31684 (Figure 5B) cohorts. We examined the relationship between the MSC risk score, the expression levels of the five pivotal genes, and various MSC markers, revealing a strong positive correlation consistent across both cohorts, TCGA-BLCA (Figure 5C,E), and GSE31684 (Figure 5D,F).

3.6. Comparison of Chemotherapy and Immunotherapy Sensitivity among Different MSC-Risk Groups

The applicability of the MSC risk score as a predictor of immunotherapy efficacy in BC patients was evaluated using the TIDE method, which showed a significantly higher TIDE score in the high-risk subgroup compared to the low-risk cohort in both the TCGA-BLCA and GSE31684 datasets (p  < 0.001; Figure 6A,D). A detailed analysis revealed that patients in the low-MSC-risk group exhibited a significantly higher responsiveness to immunotherapy compared to those in the high-MSC-risk group (p  < 0.001; Figure 6B,E). The efficacy assessment was further supported by AUC values, showing 0.819 (95% CI: 0.775–0.859) for the TCGA-BLCA dataset (Figure 6C) and 0.787 (95% CI: 0.684–0.874) for the GSE31684 dataset (Figure 6F), highlighting the robust performance of our MSC model in predicting immunotherapy response. The IC50 values of various drugs were estimated using data from the GDSC database. Wilcoxon analyses revealed that BC patients classified in the high-risk group, with elevated MSC scores, exhibited heightened sensitivity to gemcitabine, vincristine, paclitaxel, gefitinib, and sorafenib. Conversely, the low-MSC score subgroup demonstrated increased sensitivity to cisplatin, a trend consistently observed in both the TCGA-BLCA (Figure 6G) and GSE31684 (Figure 6H) cohorts. These findings highlight the predictive capability of the MSC model regarding drug responsiveness in BC.

3.7. GSEA Performed on the Five-Gene MSC Signature

To enhance understanding of the functional implications linked to the MSC signature, GSEA was performed on the TCGA-BLCA and GSE31684 datasets, comparing high- and low-MSC-risk groups. The results revealed significant enrichment in key KEGG signaling pathways, including ECM–receptor interaction, focal adhesion, and cytokine–cytokine receptor interactions (Figure 7A). Additionally, genes within the high-MSC-risk group showed notable enrichment in hallmark gene sets associated with external encapsulating structure organization and collagen fibril organization (Figure 7B). The ssGSEA results demonstrated a significant positive correlation between the MSC risk score and the enrichment scores of ECM–receptor interaction, focal adhesion, and cytokine–cytokine receptor interactions in both TCGA-BLCA (Figure 7C) and GSE31684 (Figure 7D), highlighting the intricate functional dynamics associated with elevated MSC risk classifications.

3.8. Validation of Key Genes Using the HPA Databases

To investigate the protein expression patterns of the MSC signature genes, we analyzed IHC images from the HPA database. The results revealed strong staining of matrix remodeling-associated 7 (MXRA7), ADP-ribosylation factor-like 4C (ARL4C), and cerebral endothelial cell adhesion molecule (CERCAM) in the BC stroma (Figure 8), indicating their potential as specific markers for MSCs.

3.9. Possible Targets and Molecular Docking of MSC-Related Genes in BC

In this study, we conducted further investigation into the drugs that effectively targeted five MSC-related genes in patients with BC. Our findings revealed that the drugs corresponding to four genes exhibited the potential to specifically impact BC (ZNF165, Haijinsha, Kuxingren, and Kunbu et al.; MXRA7, Chuanlianzi, Dazao, and Gehua et al.; CEMIP, Diercao, Maoyancao, and Muxiang et al.; and ARL4C, Baixianpi, Baihe, and Beiliujinu et al.), as evidenced by the data presented in Table S1. Subsequently, we selected the most promising drugs and their corresponding genes (ZNF165, Haijinsha; MXRA7, Chuanlianzi; CEMIP, Diercao; and ARL4C, Baixianpi) for detailed analysis. The TCMSP database was utilized to analyze the primary chemical constituents of the target drugs Haijinsha, Chuanlianzi, Diercao, and Baixianpi (Table 1). Subsequently, the drug exhibiting the highest OB (%) value was selected for molecular docking, aiming to validate the potential interaction between the drug and the target. The results of molecular docking demonstrated that kaempferol exhibited favorable docking with ZNF165 (Vina score = −7.7) (Figure 9A), quercetin exhibited favorable docking with MXRA7 (Vina score = −6.7) (Figure 9B), mairin exhibited favorable docking with CEMIP (Vina score = −7.6) (Figure 9C), and limonin diosphenol exhibited favorable docking with ARL4C (Vina score = −8.3) (Figure 9D). The stability observed in molecular docking experiments suggests that these compounds hold promise as effective therapeutic agents for targeting MSCs in the TME.

Table 1. Composition and relevant information of drugs corresponding to hub genes.

Gene	Medicine	Molecule ID	Molecule name	OB (%)	DL	HL
ZNF165	Haijinsha	MOL001506	Kaempferol	41.88	0.24	14.74
	—	MOL001689	Acacetin	34.97	0.24	17.25
	—	MOL001790	Linarin	39.84	0.71	16.07
	—	MOL002879	Diop	43.59	0.39	3.6
	—	MOL002881	Diosmetin	31.14	0.27	16.34
	—	MOL002882	[(2R)-2,3-Dihydroxypropyl] (Z)-octadec-9-enoate	34.13	0.3	5.19
	—	MOL002883	Ethyl oleate (NF)	32.4	0.19	4.85
	—	MOL000296	Hederagenin	36.91	0.75	5.35
	—	MOL000358	Beta-Sitosterol	36.91	0.75	5.36
	—	MOL000422	Supraene	33.55	0.42	2.72
MXRA7	Chuanlianzi	MOL001494	Quercetin	46.43	0.28	14.4
	—	MOL001495	Ethyl linolenate	46.1	0.2	6.2
	—	MOL002045	Stigmasterol	43.41	0.76	6.29
	—	MOL002047	Melianone	40.73	0.81	6.92
	—	MOL002048	Nimbolidin D	30.38	0.53	10.82
	—	MOL002053	Nimbolin A	32.11	0.34	7.33
	—	MOL002056	(E)-3-[(2S,3R)-2-(4-hydroxy-3-methoxy-phenyl)-7-methoxy-3-methylol-2,3-dihydrobenzofuran-5-yl]acrolein	54.74	0.4	8.03
	—	MOL002058	40957-99-1	57.2	0.62	2.04
	—	MOL000098	Mandenol	42	0.19	5.39
CEMIP	Diercao	MOL000211	Mairin	55.38	0.78	8.87
	—	MOL000358	Beta-Sitosterol	36.91	0.75	5.36
	—	MOL000359	Sitosterol	36.91	0.75	5.37
	—	MOL000422	Kaempferol	41.88	0.24	14.74
	—	MOL006772	Poriferasterol monoglucoside_qt	43.83	0.76	5.32
	—	MOL007879	Tetramethoxyluteolin	43.68	0.37	15.45
	—	MOL007880	3,5,7,3,5-Pentahydroxy flavonol	63.64	0.24	1.84
	—	MOL000098	Quercetin	46.43	0.28	14.4
ARL4C	Baixianpi	MOL006236	Limonin diosphenol	91.69	0.55	−3.24
	—	MOL006257	Dictamnusine_qt	69.27	0.19	−11.1
	—	MOL006235	Dasycarpamin	59.14	0.21	2.19
	—	MOL006272	9alpha-Hydroxyfraxinellone-9-o-beta-d-glucoside	55.52	0.62	−6.45
	—	MOL006244	(1R,4R,6S,8aR)-1-(3-Furyl)-4,6-dihydroxy-5,8a-dimethyl-4,6,7,8-tetrahydro-1H-isochromen-3-one	49.09	0.19	−11.68
	—	MOL006233	3′-O-methyl taxifolin	48.36	0.3	14.86
	—	MOL000098	Quercetin	46.43	0.28	14.4
	—	MOL006270	Tirucallane	44.02	0.75	5.97
	—	MOL013352	Obacunone	43.29	0.77	13.04
	—	MOL006268	Preskimmianine	42.14	0.21	1.64
	—	MOL002663	Skimmianin	40.14	0.2	−2.43
	—	MOL005043	Campest-5-en-3beta-ol	37.58	0.71	4.43
	—	MOL000358	Beta-Sitosterol	36.91	0.75	5.36
	—	MOL000006	Luteolin	36.16	0.25	15.94
	—	MOL006262	7alpha-Acetylobacunol	36.08	0.71	−11.88
	—	MOL006271	Trichirubine,b	35.51	0.27	19.24
	—	MOL006261	Isomaculosidine	31.99	0.2	12.96
	—	MOL000173	Wogonin	30.68	0.23	17.75

• Abbreviations: DL, drug likeness; HL, drug half-life; OB, oral bioavailability.

4. Discussion

To our knowledge, this study is the first to establish an MSC-related signature to predict prognosis and therapeutic responses in BC. We utilized WGCNA to estimate MSC and stromal infiltrations in BC. We identified hub modules that exhibited the strongest correlation with stromal and MSC infiltrations. Subsequently, through the application of univariate and LASSO Cox regression analyses, we identified ZNF165, MXRA7, CEMIP, ARL4C, and CERCAM as prognostic MSC markers. Furthermore, we constructed a five-gene MSC signature that demonstrated the ability to predict prognosis and therapeutic responses in BC. Our findings suggest that the MSC model may represent a novel therapeutic approach targeting MSCs in BC.

The TME is intricately composed, with a myriad of CCs beyond immune cells, encompassing a variety of stromal cells like MSCs, fibroblasts, endothelial cells, and pericytes, each contributing to the complex milieu [34]. Despite the existence of prognostic models associated with cancer-associated fibroblasts [35], a dedicated model centered on MSCs is currently lacking. The propensity of MSCs to be recruited to tumor sites has been a consistent observation across diverse tumor types [36–44]. This phenomenon of MSC recruitment is not just incidental but plays a pivotal role in determining the trajectory of tumor progression. Existing literature accentuates the dual role MSCs can potentially play; however, a substantial body of evidence inclines toward their protumorigenic influence, where they are implicated in augmenting tumor growth, promoting angiogenesis, facilitating metastasis, and instigating resistance to therapeutic drugs [26]. In alignment with this, the findings from the current research elucidate a direct correlation between elevated MSC infiltration and diminished OS, underscoring the imperative to understand this dynamic for enhanced therapeutic strategies.

MSCs exert a pivotal influence on the chemosensitivity of tumor entities, driven by their multifaceted secretory profile. Insight from a murine BC model revealed that bone marrow–derived MSC exosomes carry miR-23b [45]. This discrete microRNA is instrumental in counteracting the myristoylated alanine-rich C-kinase substrate, an essential protein kinase C substrate. This strategic interference catalyzes a dormancy stage in metastatic niche-residing cancer stem cells, fostering a robust defense mechanism against the chemotherapy agent docetaxel [45]. In a related observation, the therapeutic potential of platinum compounds witnessed a stark nullification with the intravenous introduction of bone marrow MSCs (BM-MSCs), as documented across three experimental subcutaneous tumor paradigms [46]. Upon encountering platinum-based chemotherapy, BM-MSCs respond by discharging polyunsaturated fatty acids. These specific secretions are recognized for their role in shielding various cancer forms, namely, colon, lung, and breast carcinomas, against the cytotoxic influence of platinum-derived drugs [46]. The current investigation accentuates the relevance of a five-gene MSC signature in forecasting the responsiveness to an array of chemotherapeutic agents, encompassing cisplatin, gemcitabine, vincristine, and paclitaxel. This genetic constellation holds substantial promise in refining therapeutic strategies, anchored by its predictive precision.

The landscape of BC treatment has undergone substantial transformation with the advent of immune checkpoint inhibitors (ICIs). These agents have proven effective, particularly for a subset of patients, instilling a durable immunologic memory [47]. However, the response rate of UC to ICIs hovers around 15%–21% in the context of unselected metastatic cases [48]. Distinct from targeted therapies, the realm of ICI treatment still grapples with the absence of predictive molecular biomarkers, rendering the identification of likely responders a significant clinical challenge. Given MSCs’ established role in modulating both innate and adaptive immune elements, conditioning the TME for facilitating neoplastic proliferation and progression [49], the impetus has shifted toward leveraging MSC-based insights for biomarker discovery. This strategy holds promise in refining patient screening processes for ICI therapy, contributing to more personalized therapeutic protocols. Reinforcing this perspective, findings from the present analysis indicate that an MSC-related prognostic construct holds efficacy in anticipating responses to ICI, marking a potential paradigm shift in oncological personalization.

The therapeutic landscape in cancer research has recently pivoted toward the TME, recognizing its crucial influence on tumor behavior and treatment outcomes [17]. This investigation pinpointed ZNF165, MXRA7, CEMIP, ARL4C, and CERCAM as promising prognostic indicators within the context of MSCs in BC, underscoring their potential as innovative therapeutic intervention points. ZNF165 emerges as a constituent of the Kruppel-like cadre of zinc-finger transcription modulators, notorious for its upregulation across a spectrum of malignancies encompassing urinary bladder transitional cell carcinoma [50], gastrointestinal cancer [51], and hepatocellular carcinoma [52], triple-negative breast cancer [53]. Interaction paradigms between ZNF165 and SMAD3 underscore its regulatory capacity on transforming growth factor beta (TGF-β)-responsive gene transcription [54], with distinct pathways elucidated in the proliferation and motility augmentation within hepatocellular carcinoma contexts, notably via tryptophan/kynurenine/AhR/CYP1A1 axis engagement [52]. CEMIP, documented across various tumor types, correlates with enhanced cellular motility, invasiveness, and fortified resistance against pharmacological interventions [55–57]. Specific interactions with Bip within the endoplasmic domain, precipitating Ca²⁺ liberation and subsequent PKCα signal transduction, have been implicated in breast cancer metastasis [55]. A distinct scaffolding function was observed in CEMIP, mediating GRAF1 and MIB1 coordination, thereby instigating metastatic events in colorectal cancer through CDC42/MAPK pathway activation [58]. Similarly, ARL4C, abundant within neoplastic tissues, perpetrates its tumorigenic influence across various cancer forms including, but not limited to, colorectal, pulmonary, and hepatic carcinomas [59–61]. Mechanistic insights link its activity to Rac1 stimulation, Rho attenuation, and PIK3CD amplification [59, 61], with further implications in tumor–stroma interplay impacting pancreatic cancer resilience and expansion [62]. In a parallel vein, CERCAM’s dysregulated expression profiles in diverse cancers hint at its contributory role in oncogenesis and malignant transformation [63, 64]. Empirical data accentuate CERCAM’s role in fostering cellular survival, DNA synthetic processes, and invasive potential specifically within BC matrices [65]. In summation, the molecules ZNF165, MXRA7, CEMIP, ARL4C, and CERCAM collectively represent prospective therapeutic nexus points in impeding tumor advancement. However, a comprehensive understanding of the exact mechanistic pathways endorsing BC proliferation via these molecules necessitates further exploration.

Acknowledging the limitations of this study is essential. Our research primarily utilized a retrospective bioinformatics approach, drawing from two publicly available gene expression datasets. This highlights the need for robust validation of the prognostic and therapeutic significance of the MSC model through multicenter and prospective studies. Additionally, the specific biological roles of the MSC signature biomarkers in BC should be confirmed through detailed molecular and animal experiments. Despite these limitations, our findings provide a valuable foundation for future investigations into the role of MSCs in BC.

5. Conclusions

The current five-gene prognostic MSC signature demonstrates reliability in predicting outcomes and effectively assessing responses to clinical chemotherapy and immunotherapy in BC patients. The genes ZNF165, MXRA7, CEMIP, ARL4C, and CERCAM show promise as therapeutic targets for anti-MSC interventions.

Conflicts of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Author Contributions

Conceptualization, W.Z.P., Y.E.G., W.H.Z., and Z.X.Y.; methodology, Y.E.G., J.L.H., Z.X.Y., L.P., Z.L.Y., and J.S.S.; software, Y.E.G., Z.X.Y., L.P., Z.L.Y., and J.S.S.; writing–original draft preparation, Y.E.G., J.L.H., Z.X.Y., Z.L.Y., W.H.Z., and J.S.S.; writing–review and editing, Y.L., T.J.Q., Z.Y.F., W.H.Z., and W.Z.P.; visualization, Y.E.G., Z.X.Y., Z.L.Y., J.S.S., and L.P.; supervision, Y.L., T.J.Q., and W.Z.P.; project administration, Y.L., T.J.Q., and W.Z.P.; funding acquisition, W.Z.P. All authors have read and agreed to the published version of the manuscript. Y.E.G., J.L.H., and Z.X.Y. contributed equally to this work.

Funding

This work is supported by the National Natural Science Foundation of China (NSFC: 81874088) and the Cuiying Scientific Training Program for Undergraduates of Lanzhou University Second Hospital (Grant Number: CYXZ2021-13).

Supporting Information

Additional supporting information can be found online in the Supporting Information section.