We investigated 31 urothelial specimens, including 10 CGs, 12 UCs, and nine NUs, using a proteomic approach. A schematic outline of this study is shown in Fig. 1. The median ages were 40, 70, and 63.5 years, respectively, and the male-to-female ratios were 1.0, 0.4, and 0.9, respectively. Half of the specimens that contained CG were accompanied by IM. All UCs were confined to the mucosa (stage Ta), and eight (67%) were of WHO high grade.

From LC-MS/MS analysis, we identified 9,890 proteins across all samples and 1,139 DEPs among the three entities (ANOVA FDR < 0.05). With unsupervised hierarchical clustering, a substantial number of DEPs were discovered to overlap with CG/NU in a way distinct from UC (Fig. 2). CG was not differentiated by the presence of IM (Fig. 2, asterisks). Interestingly, we found a subset of DEP clusters (n = 53, 5%) that were differentially expressed in NU but similarly expressed between CG and UC (Fig. 2, arrows); thus, these proteins were named as “UC-like signature” (Fig. 3A). Since the signature were similar within each group, the patterns of these signatures were thought to represent the general profile of CG.

We hypothesized that this signature might represent the UC-like characteristics of the CG. Using PPI analysis, we explored cellular processes represented by the “UC-like signature”. This resulted in sets of DEP networks within the “UC-like signature”, which were enriched for reactive oxygen species (ROS) and energy metabolism [22–24], growth and DNA repair [25,26], transport [27,28], motility and epithelial-mesenchymal transition (EMT) [29,30], and cell survival [31,32] (Fig. 3B). The unit associated with ROS and energy metabolism, which was connected to the one with growth and DNA repair functions, included the top-ranked DEPs (e.g., SOD2, PRKCD, CYCS, and MRPL23) (Fig. 3B). Using the official pre-ranked GSEA of the “UC-like signature”, we also identified significant enrichment of ultraviolet (UV) response (normalized enrichment score = 1.772, FDR = 0.03) for SOD2, PRKCD, and MRPL23 (Fig. 3C). UV radiation stimulates carcinogenic sequences by promoting ROS generation and DNA damage [33]. These data suggest that CG is predominantly a benign lesion, but oxidative stress might be a gateway theme representing oncogenic potential in CG.

To further characterize the ”UC-like signature”, we shortlisted the top 10 most significant DEPs (Fig. 3A, asterisk), followed by GOBP analysis (Fig. 4A). Subsequently, we constructed a network of connections between the GOBP terms. Consistent with the aforementioned single-protein networks, functional elements related to ROS metabolism, development, and transport were highlighted by GOBPs (Fig. 4B), which suggests that these functions could collectively substantiate the cancerous risks in CG.

Of note, GOBPs were represented almost exclusively by four molecules, i.e., SOD2, PRKCD, CYCS, and HCLS1 (Fig. 4A), and their abundance in UC/CG compared to NU accordantly pointed toward oncologic functions. For example, previous studies have shown an association of decreased level of SOD2 and increased level of PRKCD, as observed in CG/UC versus NU, with ROS production [22,24]. Overt oxidative stress by activating oncogenic signaling pathways, DNA mutations, EMT, and stromal remodeling promotes bladder cancer development and progression [22]. In line with this, altered expression of CYCS and HCLS1 might deregulate oxidative respiration and the RAS signaling pathway, thereby enhancing the malignant behavior of bladder cancer [23,34]. We further examined whether these four statistically and functionally significant molecules were differentially expressed with IM and found no significant differences related to IM (Fig. 4C).

The malignant risk of CG and its association with bladder cancer have been disputed [1,2]. Considering the high prevalence of CG, there is a dire need to clarify its oncologic implications. Using LC-MS/MS, we revealed the proteomic landscape of CG in relation to malignant (UC) and normal (NU) urothelial tissues. CG generally presented an NU-like profile, in contrast to UC, indicating the overall benign nature of CG. By retrospectively observing a handful of patients with CG, consistent with this result, previous studies failed to confirm clear association between UC and CG [7,8]. Conversely, we confirmed that some of the clustered DEPs showed an opposite pattern across the diagnosis; these proteins were shared by CG and UC but not by NU. Using functional and network analyses, we found that these DEPs coded for ROS and energy metabolism, growth and DNA repair, transport, motility and EMT, and cell survival, further supporting the relation of this “UC-like signature” with the malignant risk in CG. To our knowledge, this is the first study to determine the oncogenic characteristics of CG using a comprehensive proteomic analysis. Further large-scale studies are required to determine the risks of UC development in patients with CG.

Using an interconnected functional network analysis of the shortlisted proteins, including SOD2, PRKCD, CYCS, and HCLS1, ROS metabolism, structural development, and transport functions were concordantly found to be significantly enriched in the “UC-like signature.” ROS are tightly regulated in normal cellular environments, and increased ROS act as versatile transducers for the generation and progression of bladder cancer [22]. For example, both UC and CG showed downregulation of SOD2, a scavenging molecule, and upregulation of PRKCD, an oxidative enzyme associated with ROS accumulation. ROS imbalance induces oxidative DNA damage, mutations, propagation of oncogenic signals including RAS, mitogen-activated protein kinases, phosphoinositide 3-kinase, and nuclear factor κB pathways, EMT, and stromal modification in bladder cancer [22,24]. Consistent with this, a previous study suggested that alterations in ROS metabolism participate in inflammation-associated cancer sequences [35,36]. In addition, upregulation of PRKCD can further support a malignant phenotype in bladder cancer by promoting migration and invasion [37]. CYCS is instrumental in regulating oxidative phosphorylation and energy metabolism, and high maintenance of CYCS, as identified in UC/CG, could induce metabolic modification in bladder cancer, or Warburg effect [23]. In addition, UC and CG showed lower HCLS1 levels than NU. HCLS1, an actin-binding molecule supporting cellular transport and trafficking, has been shown to result in adverse outcomes in bladder cancer [34,38]. We believe that the “UC-like signature” might reflect the oncogenic pathway related to altered ROS and energy metabolism and structural modification in CG, and the expression of these signature proteins deserves further study to determine the risk of aggression in CG.

This study lacked specimens with adenocarcinoma in the analysis. The malignant association of CG with IM has been conjectured, typically in terms of bladder adenocarcinoma, but definite evidence has not been confirmed [1]. Instead, altered ROS and energy metabolism have been implicated in the transformation of dysplasia to adenocarcinoma in the esophagus, consistent with the functions enriched in the “UC-like signature” [35]. In addition, the overall proteomic profile and biomarkers relevant to such functions showed no difference regardless of the presence of IM in CG. However, previous studies have suggested that CG with IM may be more advanced than CG without IM regarding tumorigenic potential, as exemplified by more robust telomere shortening or β-catenin activation [4,6]. Therefore, we reasonably speculate that the “UC-like signature” might reflect the malignancy risk encoded in CG irrespective of IM at a global level.

In conclusion, using comprehensive proteomic profiling, we identified a predominantly non-neoplastic landscape of CG that is closer to NU than to UC. Furthermore, we confirmed a small subset of common DEPs in UC and CG, suggesting altered functions of ROS metabolism that might imply potential cancerous risks in CG.