UC patients, comprising 131 UTUC (79 male cases & 52 female cases, and diagnosis age ranged from 36 to 86 years old) and 118 UCB (92 male cases & 26 female cases, and diagnosis age ranged from 19 to 87 years old) patient samples, were enrolled between January 1^st, 2018 and October 9^th, 2021, and UTUC or UCB disease was diagnosed by a qualified physician. Written informed consents of all involved patients have been collected, and this study was conducted according to the International Ethical Guidelines for Biomedical Research Involving Human Subjects, the Declaration of Helsinki, and approved by the ethics committee of Peking University First Hospital. In the present study, there was no limit regarding the gender, age, weight of involved patients when collecting UTUC or UCB patient samples; however, pregnant or breastfeeding female samples were excluded. Besides, a total of 3 qualified surgeons participated into the surgical operations and all patient samples were collected using simple randomization. Formalin-fixed, paraffin-embedded (FFPE) tumor tissues and germline DNA from matched blood samples were subsequently analyzed by using a targeted NGS platform. The representative FFPE tumor tissues were estimated by a qualified pathologist to determine whether the tumor content was sufficient (at least 20%). Of these samples, 1 UTUC & 2 UCB samples were removed because of insufficient tumor content and 1 UCB samples were lost owing to technical problems during DNA extraction, thus resulting in a total of 131 UTUC and 118 UCB samples available for subsequent analysis. Of note, all involved workers were blinded as of sample collection, processing, outcome assessment, and bioinformatic analysis.

The FFPE tumor tissue samples and white blood cell samples, including tumor genomic DNA (gDNA) and germline gDNA, respectively, were initially extracted by using QIAamp Genomic DNA Kit (Qiagen, CA, USA) according to the instructions of manufacturer. Next, the quality and quantity of purified DNA were detected and analyzed via using Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA) and Qubit 3.0 Fluorometer (Thermo Fisher Scientific Life Technologies, MA, USA). Then, 100 ng gDNA was sheared by a Covaris E210 system (Covaris, MA, USA), and both the FFPE tumor gDNA and germline gDNA library construction were prepared by using Accel-NGS 2S DNA Library Kit (Swift Biosciences, MI, USA). The target-enriched libraries were constructed by using xGen Lockdown Probes Kit (Integrated DNA Technology, IA, USA), and the probes were synthesized by Integrated DNA Technology, Inc for a panel of 618 genes ( Supplementary Table 1). In the present study, bioassays were replicated three times. Eventually, paired-end sequencing with 150 bp length each read was conducted for the target-enriched libraries on the Illumina Novaseq 6000 platform (Illumina, CA, USA). The coverage of at least 1000× and 200× was achieved for tissue and germline gDNA, respectively.

The FASTQ files of paired-end sequencing reads were obtained by the software FASTP (FASTP, RRID : SCR_016962, https://github.com/OpenGene/fastp), and the quality control was also conducted via CASAVA (CASAVA, RRID : SCR_001802, http://support.illumina.com/sequencing/sequencing_software/casava.html). The raw sequencing data were then aligned to the UCSC Human Genome Reference hg19 by Burrows-Wheeler alignment tool and a binary sequence alignment map (BAM) file was generated. The duplicate removal and local realignment were performed by using Picard (Picard, RRID : SCR_006525, http://broadinstitute.github.io/picard/) to improve the mapping quality. Variant calling (single nucleotide variation (SNV), small insertions/deletions (indels), etc.) was based on the Genome Analysis Toolkit (GATK, RRID : SCR_001876, https://software.broadinstitute.org/gatk/download/), and they were annotated by using the software ANNOVAR (ANNOVAR, RRID : SCR_012821, http://www.openbioinformatics.org/annovar/). In addition, TMB was defined as the number of identified variants per megabase (Mb). The somatic alterations were interpretated according to the standards and guidelines of Association for Molecular Pathology, American Society of Clinical Oncology, and College of American Pathologists. The germline variants were determined by a molecular geneticist in accordance with the standards and guidelines of American College of Medical Genetics and Genomics, and the Association for Molecular Pathology, it is necessary to state that variants with unknown significance were also evaluated but not reported in the present study. The oncogenic alterations were identified in ClinVar (ClinVar, RRID : SCR_006169, http://www.clinvar.com/). Actionable GAs analysis was conducted in OncoKB (OncoKB, RRID : SCR_014782, https://www.oncokb.org/).

Thirty-four DDR genes ( Supplementary Table 2 ), included in six canonical pathways of Checkpoint, Fanconi Anemia (FA), Homologous Recombination (HR), Mismatch Repair (MMR), Nucleotide Excision Repair (NER), and others, were obtained as previously described (11), which was utilized for the analysis of somatic alteration profiling. Besides, an additional EPCAM was usually used to characterize potential UC patients with LS as a MMR-associated gene (17) during the investigation of germline variants in our cohort. Deleterious alterations were determined according to the interpretation of OncoKB database via cBioPortal (cBioPortal, RRID : SCR_014555, http://www.cbioportal.org/).

The landscape of GAs of UC patients was exhibited in the Oncoprint plots by using the package maftools. According to the previously described oncogenic signaling pathways in 33 cancer types (18), somatic GAs in our study were categorized into following canonical pathways: TP53, RTK/RAS, DDR, NOTCH, PI3K, Cell Cycle, WNT, HIPPO, TGF-, MYC, and NRF2. Besides, the comparison of GA types of selected genes between UTUC and UCB was visualized in the lollipop plots also by using the package maftools. In addition, 30 COSMIC mutational signatures were downloaded from the COSMIC website (COSMIC, RRID : SCR_002260, http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/), and mutational signature analysis was conducted by using nonnegative matrix factorization from the 96-channel mutational profiles (19).

Eventually, clinical and genomic data of UTUC (MSK, Nat. Commun. 2020) (20), namely MSKCC-UTUC cohort in the present study, were downloaded to investigate the difference in GAs between Chinese and Western UTUC patients, and the clinical comparison analysis was shown in Supplementary Table 3 . In addition, clinical and genomic data of UCB from TCGA, PanCancer Atlas (21) and MSKCC, Eur Urol 2014 (22), namely TCGA-UCB and MSKCC-UCB cohort, respectively, were retrieved to investigate the difference in GAs between Chinese and Western UCB patients, and the clinical comparison analysis between Chinese and TCGA-UCB or MSKCC-UCB cohorts was exhibited in Supplementary Table 4, 5 . In the present study, muscle-invasive bladder cancer (MIBC) patients were screened out for subsequent analysis. It should be mentioned that a shared panel of 209 genes ( Supplementary Table 6 ) was used when comparing GAs between Chinese UTUC and MSKCC-UTUC cohorts, and also the same shared panel of 209 genes was used when comparing GAs between Chinese UCB and MSKCC-UCB cohorts, besides, a panel of 618 genes ( Supplementary Table 1 ) was used to filter GAs of UCB patients from the TCGA-UCB cohort for subsequent analysis. The IMvigor210 cohort (23), including UC patients treated with immunotherapy by anti-PD-L1 therapy, was downloaded to explore the relationship between TMB level and immunotherapy response. The defined criteria of therapy response: CR: complete response, PR: partial response, SD: stable disease, PD: progressive disease.

Statistical analysis was conducted in R studio (v.3.4.3, https://rstudio.com/) or by using the software Statistical Package for the Social Sciences (SPSS, RRID : SCR_002865, http://www-01.ibm.com/software/uk/analytics/spss/) and GraphPad Prism v8.0 (GraphPad Prism, RRID : SCR_002798, http://www.graphpad.com/), and chi-square and Fisher’s exact tests were used to determine the statistical significance of differences among demographic categorical variables, and the Kruskal-Wallis’ test was used to compare the continuous variables. The p-value < 0.05 was considered statistically significant.

After investigation, all involved Chinese UC samples ( Table 1 ), including 131 UTUC and 118 UCB tumors (UCB samples had no prior history of UTUC), were found with valid somatic alterations, and the median and mean TMB values were 12.00 [0.66, 277.32] and 17.64, respectively. Besides, the median and mean TMB values of UTUC vs. UCB were 9.88 vs. 14.00 and 17.31 vs. 18.00, respectively. Except a significantly higher proportion of female patients diagnosed with UTUC tumors instead of UCB, no statistically significant difference in other clinical features was observed between UTUC and UCB in our cohort ( Table 1 ).

In total, genomic landscape of Chinese UC samples demonstrated that the most frequently altered genes ( Figure 1A ) included TP53 (44.58%), ARID1A (18.00%), and KDM6A (15.26%); whereas, it was noticed that the alteration frequencies of FGFR3 were only 7.63% and 5.93%, respectively in UTUC and UCB. When grouping GAs by functional significance, it was revealed that the top prevalently enriched pathways were TP53 (50.60%), DDR (49.00%), RTK/RAS (45.78%), NOTCH (34.94%), PI3K (27.71%), and Cell Cycle (21.29%). While 93.98% (234/249) of Chinese UC patients were identified as having oncogenic alterations, which were the most abundantly enriched in TP53 signaling pathway (48.59%, Figure 1B ).

Furthermore, a total of 141 (56.63%) Chinese UC patients with 220 actionable alterations were identified in our cohort. Irrespective of UTUC or UCB, it was found that more patients had level 3 actionable GAs (level 3 vs. level 4 vs. level 1, UTUC: 44.44% vs. 32.22% vs. 23.33%; UCB: 39.33% vs. 37.08% vs. 23.60%, Supplementary Figure 1 ). UC patients had the most actionable GAs in ARID1A (N = 34, 13.65%; UTUC vs. UCB: 12.21% vs. 15.25%). Besides, remaining actionable GAs were mostly enriched in the RTK/RAS, PI3K, and DDR related signaling pathways ( Figure 1C ). According to the statistical analysis, there was no difference in the prevalence of actionable GAs between Chinese UTUC and UCB, except that more AKT1 and ERCC2 actionable GAs were presented in Chinese UCB (p < 0.05, Figure 1C ). In comparison with the TCGA-UCB cohort, more AKT1 (4.24% vs. 0.24%), EGFR (1.69% vs. 0.00%), and PALB2 (2.54% vs. 0.24%) actionable GAs occurred in the Chinese UCB cohort; whereas, ERBB2 (9.29% vs. 1.69%), FGFR3 (12.96% vs. 3.39%), KDM6A (20.29% vs. 0.00%), and PIK3CA (18.34% vs. 5.08%) actionable GAs were prevalent in the TCGA-UCB cohort (p < 0.05, Supplementary Figure 2 ). Regarding the comparison of actionable GAs between Chinese UCB and MSKCC-UCB cohort, it was discovered that MSKCC-UCB cohort had more prevalent actionable GAs in ARID1A (27.52% vs. 15.25%), FGFR3 (18.35% vs. 3.39%), KDM6A (37.61% vs. 0.00%), and PIK3CA (22.02% vs. 5.08%), while ERCC2 (7.63% vs. 0.92%) actionable GAs were more prevalent in the Chinese UCB cohort (p < 0.05, Supplementary Figure 3 ). In addition, FGFR3 (44.34% vs. 8.47%), KDM6A (23.48% vs. 0.00%), and TSC1 (10.43% vs. 2.54%) actionable GAs were more frequent in the MSKCC-UTUC cohort (p < 0.05, Supplementary Figure 4 )

In addition, a total of 32 P/LP germline variants were identified within 11.24% (28/249) of individuals with UC in our cohort, including 14.50% and 7.63% of UTUC and UCB patients ( Table 2 ). Interestingly, there were four UTUC patients harboring two P/LP germline variants. The top prevalently altered genes in UC were BRCA1 (N = 3, 1.20%, UCB vs UTUC: 2:1), CHEK2 (N = 3, 1.20%, UCB vs UTUC: 1:2), MSH2 (N = 2, 0.80%, UCB vs UTUC: 0:2), ERCC5 (N = 2, 0.80%, UCB vs UTUC: 1:1), BRCA2 (N = 2, 0.80%, UCB vs UTUC: 1:1), BAX (N = 2, 0.80%, UCB vs UTUC: 0:2), and PALB2 (N = 2, 0.80%, UCB vs UTUC: 0:2). In addition, 25 out of 32 P/LP germline variants were in genes associated with DDR in UC, of which 5 were in MMR-associated genes, including MSH2 (N = 2), MSH6 (N = 1), PMS1 (N = 1), EPCAM (N = 1). In UCB, it was observed that the Chinese UCB cohort had the germline variants of MMR-associated PMS1 (N = 1), and other DDR-associated BRCA1 (N = 2), BRCA2 (N = 1), BRIP1 (N = 1), CHEK2 (N = 1), ERCC5 (N = 1), RAD54L (N = 1), & RECQL4 (N = 1). Whereas, the Chinese UTUC cohort harbored the germline variants of MMR-associated MSH2 (N = 2), MSH6 (N = 1), EPCAM (N = 1), and other DDR-associated BRCA1 (N = 1), BRCA2 (N = 1), CHEK2 (N = 2), ERCC3 (N = 1), ERCC4 (N = 1), ERCC5 (N = 1), FANCL (N = 1), MUTYH (N = 1), PALB2 (N = 2), RAD51C (N = 1), and non-DDR-associated ASXL1 (N = 1), BAX (N = 2), CDKN2A (N = 1), SDHA (N = 1) & VHL (N = 2) ( Figure 2 ).

GA enrichment analysis revealed that KMT2D (25.19% vs. 3.39%), KMT2C (12.98% vs. 1.69%), GNAQ (7.63% vs. 0.85%), KMT2A (5.34% vs. 0.00%), RUNX2 (5. 34% vs. 0.00%), and SPTA1 (5.34% vs. 0.00%) were significantly more prevalently altered in UTUC; whereas, ERCC2 (11.02% vs. 2.29%), RB1 (7.63% vs. 1.53%), and PPM1D (6.78% vs. 0.76%) were more prevalently altered in UCB instead ( Figure 3A ). Grouped by signaling pathway, it was found that there was no significant difference in the prevalence of altered pathway between UTUC and UCB tumors, with the exception of Cell Cycle signaling pathway which was more prevalent in UCB (27.97% vs. 15.27%, Figure 3B ).

The difference in the prevalence of KMT2D between UTUC and UCB was consistent in both the diseases with early (I/II) and advanced (III/IV) clinical stages. Meanwhile, some novel difference emerged, such as more ERCC2 (18.75% vs. 2.13%), CDKN1A (18.75% vs. 2.13%), BRCA2 (15.63 vs. 2.13%) and Cell Cycle signaling pathway alterations (37.50% vs. 8.51%) were prevalent in early stage UCB ( Figures 3C, D ). For tumors in advanced clinical stage (III/IV), KMT2C (13.10% vs. 1.19%) and PPM1D (1.19% vs. 9.52%) showed significantly differed prevalence in UTUC and UCB ( Figure 3E ). More prevalent alterations in the NRF2 pathway (p = 0.06) but less alterations in NOTCH pathways (p < 0.05) were identified in patients with advanced UTUC, comparing with those with advanced UCB ( Figure 3F ). Additionally, identification of FGFR3 alteration site further revealed that p.Arg248Cys (0.00% vs. 10.64%, p < 0.05, Supplementary Figure 5 ) was less prevalent in early-stage UCB when compared with early-stage UTUC.

In UTUC, the most prevalently altered genes were TP53 (38%), KMT2D (23%), and FGFR3 (17%) in clinical stage I/II tumors, while TP53 (51%), KMT2D (26%), and ARID1A (21%) in clinical stage III/IV tumors ( Figure 4A ). Of note, the frequency of FGFR3 (17.02% vs. 2.38%) and EP300 (14.89% vs. 4.76%) was significantly higher in UTUC patients with low clinical stage, along with a higher prevalence of GAs in RTK/RAS pathway (38.30% vs. 21.43%, Figure 4B ).

For UCB, TP53 (46%), KDM6A (33%), and CDKN1A (15%) were the most prevalent in clinical stage I/II tumors; whereas, TP53 (41%), ARID1A (23%), and EP300 (14%) were the most prevalent in clinical stage III/IV tumors ( Figure 4C ). In addition, it was found that KDM6A (34.38% vs. 14.28%) was more prevalent in UCB tumors with lower clinical stage, but HRAS (9.52% vs. 0.00%) was more prevalent in UCB tumors with advanced clinical stage. While there was no difference in the frequency of altered signaling pathways between lower and higher clinical stage UCB tumors ( Figure 4D ).

To fully understand the genomic profile of Chinese UTUC patients, we further compared the genomic difference between UTUC patients with different disease sites. The oncoprint plot ( Figure 5A ) initially exhibited the genomic landscape of Chinese UTUC samples including 74 renal pelvis & 57 ureter tumors ( Table 3 ). GA enrichment analysis based on the primary tumor location (renal pelvis vs. ureter) revealed that ELF3 GAs (14.86% vs. 3.51%) were more enriched in renal pelvis tumors; conversely, the frequency of TP53 (61.40% vs. 35.14%), PMS2 (8.77% vs. 0.00%), FAT4 (8.77% vs. 0.00%) was significantly higher in patients with ureter tumors ( Figure 5B ). Correspondingly, significantly more GAs in TP53 (70.18% vs. 40.54%) and Cell Cycle (22.81% vs. 9.46%) signaling pathways were observed within patients with ureter tumors, compared with those with renal pelvis tumors ( Figure 5C ).

In our cohort, a total of 122 (49.00%) UC patients, including 66 (50.38%) UTUC and 56 (47.46%) UCB cases, were found to harbor at least one GA in 34 DDR-related genes ( Figure 6A ). Of note, the top prevalently altered genes ( Figure 6B ), respectively in six canonical pathways of Checkpoint, FA, HR, MMR, NER, and others, were ATM (N = 24, 9.64%), BRCA2 (N = 20, 8.03%), BRCA1 (N = 16, 6.43%), MSH6 (N = 10, 4.02%), ERCC2 (N = 16, 6.43%), and RECQL4 (N = 16, 6.43%). Statistical analysis further disclosed that the frequency of RECQL4 (9.16% vs. 3.39%), was significantly higher in UTUC; whereas, there were significantly more ERCC2 (11.02% vs. 2.29%) and PMS1 (3.39% vs. 0.00%) GAs in UCB ( Figures 6A, B ). GAs distributed in the pathways of HR & others (19.85%) and FA (19.49%) were the most prevalent in UTUC and UCB, respectively ( Figure 6C ). Besides, no significant difference was observed in the distribution of GAs among these six DDR related pathways between UTUC and UCB, with one exception in the pathway of others which was more prevalent in UTUC (19.85% vs. 7.63%, p < 0.01). Additionally, a total of 70 (28.11%) UC patients, including 30 (22.90%) UTUC and 40 (33.90%) UCB patients, had at least one deleterious/likely deleterious GA in DDR related pathways ( Figure 6D ). Whereas, statistical analysis showed that there was no significant difference in the number of deleterious/likely deleterious GAs among these six canonical pathways between UTUC and UCB.

Known to have a high level of heterogeneity, UTUC and UCB have distinct mutational signatures and etiologies. Of note, it was revealed in our cohort that sig 3 (correlated with failure of DNA double-strand break-repair by HR) was the most prevalent both in UTUC (18.26%) and UCB (34.55%) ( Figure 7A ). Other dominant mutational signatures in UTUC included sig 1 (correlated with spontaneous deamination of 5-methylcytosine), sig 2 (correlated with APOBEC cytidine deaminase, C>T), sig 4 (correlated with exposure to tobacco mutagens), sig 10 (defective DNA mismatch repair), sig 16 (etiology: unknown), and sig 22 (correlated with exposure to AA). On the other hand, sig 1, 4, 13 (correlated with APOBEC-mediated mutagenesis), and 22 were predominantly included in UCB instead. Notably, sig 5 (etiology: unknown) and sig 23 (etiology: unknown) were exclusive to UTUC and UCB samples, respectively, albeit with a low prevalence.

Of note, Chinese UC patients with sig 22, particularly UTUC patients, had markedly elevated TMB level ( Figure 7B ). Besides, significantly differed TMB levels were also identified in Chinese UTUC patients presenting with sig 2, 12 and 25 or UCB patients with sig 6, 7 and 13, respectively ( Figure 7B ). Impressively, the analysis of IMvigor210 cohort further revealed that TMB-high group had the significantly better clinical outcomes, and importantly, TMB-high was markedly correlated with complete/partial responses of immunotherapy ( Supplementary Figure 6 ). Moreover, KDM6A, ATM, POLD1, and BCOR were more frequently altered in UTUC patients with sig 22 than in those without it ( Figure 7C ). Whereas in UCB, NCOR1, BRCA2, ARID1B, and PBRM1 were more frequently altered in patients with sig 22 than the counterparts without it ( Figure 7D ). In addition, NCOR1, FBXW7, ATR, RNF43, and AKT2 were more prevalent in UCB patients with sig 13 ( Figure 7E ).

Finally, GA comparison analysis was conducted to reveal the differentiation of molecular characterization of UTUC and UCB between Chinese and Western cohorts. TP53 was significantly more prevalent in the Chinese UTUC cohort compared to the MSKCC-UTCC cohort; on the contrary FGFR3, KDM6A, and KMT2C were more frequently altered in the MSKCC-UTUC cohort ( Figure 8A , Supplementary Table 7 ). Accordingly, TP53 signaling pathway was significantly more common in our UTUC cohort; whereas, RTK/RAS and PI3K signaling pathways were more prevalent in the MSKCC-UTUC cohort ( Figure 8B ). In addition, PPM1D and ERCC5 GAs were prevalent in Chinese UCB cohort, compared with TCGA-UCB cohort ( Figure 8C , Supplementary Table 8 ), but ERCC2 GAs were frequently altered in our cohort, compared with MSKCC-UCB cohort ( Figure 8D , Supplementary Table 9 ). In contrast, ARID1A, PIK3CA, RB1, ERBB2, and FAT1 GAs were significantly more prevalent both in the TCGA- and MSKCC-UCB cohort. Pathway enrichment further revealed that WNT signaling pathway was more prevalent in the Chinese UCB cohort when compared to the TCGA-UCB or MSKCC-UCB cohort; whereas, GAs in TP53, RTK/RAS, HIPPO and PI3K signaling pathways were more prevalent in these two Western cohorts ( Figures 8E, F ).