Patient demographic and clinical data
One hundred eighteen patients with UC were enrolled, including 45 patients with UTUC and 73 patients with UCB. Patient characteristics are summarized in Table 1. The UCB patient cohort had a higher ratio of males to females and PD-L1 expression than the UTUC patient cohort (5.08:1 vs. 1.81:1, p = 0.025 and 31.51 vs. 17.78, p = 0.003, respectively). A total of 8.89% (4/45) of patients with UTUC and only 2.74% (2/73) of patients with UCB (p = 0.661) had dMMR. There was no significant differences in other characteristics in both cohorts.
Table 1
Characteristics of patients with urothelial carcinoma
Characteristics | UTUC (n, %) | UCB (n, %) | P value |
N | 45 | 73 | |
Sex | | | 0.025 |
Male | 29 (64.44) | 61 (83.56) | |
female | 16 (35.56) | 12 (16.44) | |
Age | | | 0.342 |
> 65 year | 22(48.89) | 43 (58.90) | |
≤ 65 | 23 (51.11) | 30 (41.10) | |
Smoking | | | 0.231 |
Yes | 13 (28.89) | 30 (41.10) | |
No | 30 (66.67) | 39 (53.42) | |
Missing | 2 (4.44) | 4 (5.48) | |
ECOG | | | 0.621 |
0 | 1 (2.22) | 1 (1.37) | |
1 | 20 (44.44) | 38 (52.50) | |
2 | 4 (8.89) | 4 (5.48) | |
Missing | 20 (44.44) | 30 (41.10) | |
Stage | | | 0.225 |
I | 5 (11.11) | 6 (8.22) | |
II | 11(24.44) | 14 (19.18) | |
III | 14 (31.11) | 21 (28.77) | |
IV | 7(15.56) | 18 (24.66) | |
Missing | 8 (17.78) | 14 (19.18) | |
Metastasis | | | 0.531 |
Yes | 11(24.44) | 23 (31.51) | |
No | 31 (68.89) | 48 (65.75) | |
Msissing | 3 (6.67) | 2 (2.74) | |
Treat-naive | | | 0.671 |
Yes | 31 (68.89) | 48 (65.75) | |
No | 11(24.44) | 22 (30.14) | |
Msissing | 3 (6.67) | 3 (4.11) | |
Recrudesce | | | 0.832 |
Yes | 11(24.44) | 24 (32.88) | |
No | 31 (68.89) | 46 (63.01) | |
Msissing | 3 (6.67) | 3 (4.11) | |
PD-L1 expression | | | 0.033* |
≤ 1% | 22 (48.89) | 21 (28.77) | |
> 1% | 8 (17.78) | 23 (31.51) | |
Msissing | 15 (33.33) | 29 (39.73) | |
MMR | | | 0.661 |
dMMR | 4 (8.89) | 2 (2.74) | |
pMMR | 16 (35.56) | 17 (23.29) | |
Missing | 25 (55.56) | 54 (73.97) | |
Landscape of mutations in UTUC and UCB
To explore the somatic mutation patterns in both UTUC and UCB, we conducted Acornmed 808 panel to identify somatic mutations. One thousand six hundred fifty nonsynonymous protein coding mutations were found in 45 patients with UTUC, ranging from 4 to 333 mutations. Patients with UCB harbored 2017 nonsynonymous protein coding mutations in total, ranging from 11 to 255 mutations. The top 20 most frequently mutated genes were identified in UTUC and UCB. KMT2D, TP53, and BRD4 mutations were the most frequent in both UTUC (Figure 1A) and UCB (Figure 1B). Eight genes were significantly different in diverse tumor location (Figure 1C). Notably, INRRL1 mutations only occurred in UCB. However, compared to UCB, mutations of KMT2C, LRP1B, NCOR1, ZFHX4, KDR, NF1, and NOTCH genes were more likely to be seen in UTUC. In addition, we discovered that 20 genes were significantly different between our UTUC cohorts and TCGA UTUC cohorts, such as KMT2D, TP53, and FGFR3. Moreover, 11 frequently mutated genes were significantly different between our UCB cohorts and TCGA UCB cohorts, such as KMT2C, FAT1, and GANQ.
Prevalence of mutated genes is related to clinical stage
We attempted to determine whether the prevalence of mutated genes was correlated with clinical stage in UTUC and UCB. The significantly different genes in UTUC and UCB, as well as the top 5 mutated genes, were analyzed for various stages, which were divided into two groups by stage (clinical stage I and II vs. stage III and IV), referring to the previous article (7). There was no difference in the genes that were identified as significantly different between UTUC and UCB in various stages (Figure S1A). Surprisingly, there were significant differences for the prevalence of the top 5 genes in various stages and cancer types (Figure S1B-E). For stage Ⅰ and Ⅱ, SPEN mutation was more frequent in UTUC than that in UCB (38% vs. 0%, p=0.004). Nevertheless, FGFR3 mutation was also more frequent in UCB (19% vs. 53%, p=0.004), as previous studies showed. For stage Ⅲ and Ⅳ, ARID1A mutation was more frequent in UTUC than that in UCB (32% vs. 11%, p=0.039). Interestingly, there was no significant difference for gene mutations at various stages in UTUC, although the frequency of SPEN mutation seemed to be higher in stage Ⅰ and Ⅱ than that in stage Ⅲ and Ⅳ (38% vs. 11%, p=0.055). For UCB, FGFR3 mutation was significantly more frequent in stage Ⅰ and Ⅱ than that in stage Ⅲ and Ⅳ (53% vs. 5%, p<0.001). Conversely, SPEN and PTPRT mutations were significantly more frequent in stage Ⅲ and Ⅳ (0% vs. 21%, p=0.042 and 0% vs. 24%, p=0.022, respectively).
Mutational signature for UTUC and UCB
To investigate the molecular characteristics of UTUC and UCB, mutational signature and somatic substitution analysis were undertaken. C to T (C>T) substitutions were the dominant mutation type in both UTUC and UCB (36.5% and 44.0%, respectively). For UTUC, the transversions, which are interchanges of purine for pyrimidine bases involving the exchange of one-ring and two-ring structures, were dominant DNA substitution mutations (Figure 2A). Whereas transitions that are interchanges of two-ring purines (AG) or of one-ring pyrimidines (CT), involving bases of similar shape, were dominant DNA substitution mutations in UCB (Figure 2B). In addition, only signature 1 for spontaneous deamination of 5−methylcytosine existed in both cohorts (41.86% and 26.61%, respectively). Although there were APOBEC Cytidine Deaminase signatures in both groups, they were projected to different COSMIC signatures (signature 2 and 13, respectively), with a difference rate (32.41% and 42.99%, respectively). Signature 22 for Exposure to aristolochic acid (AA) (20.91%) and signature 10 for defects in polymerase POLE (4.81%) were only observed in the UTUC cohort (Figure 2C, E). Conversely, Signature 6 for defective DNA mismatch repair (30.41%) only existed in the UCB cohorts (Figure 2D, F). In addition, 5−methylcytosine mutational signature was the predominant mutational signature seen in UTUC cohorts, but APOBEC was the mutation signature for UCB cohorts.
KEGG enrichment for UTUC and UCB
To better understand the differences in biological functions between UTUC and UCB, KEGG enrichment analysis was performed. In the UTUC cohort, mutated genes were mainly enriched in the cancer pathway, PI3K-Akt signaling pathway, ErbB signaling pathway, HTLV-I infection and Rap1 signaling pathway (Figure 3A). Regarding the UCB cohort, the primary pathways included the cancer pathway, PI3K-Akt signaling pathway, HTLV-I infection, FoxO signaling pathway, and Thyroid hormone signaling pathway (Figure 3B). Rap1 signaling pathway was not discovered in UCB cohorts. To study the proportion of mutational pathways in the UTUC and UCB cohorts further, we analyzed 10 oncogenic signaling pathways reported in previous literature (19). There were no significant differences for the frequency of oncogenic signaling pathways between UTUC and UCB (Figure 3C). The common pathways were the RTK/RAS pathway (68.89% vs. 75.34%), Notch pathway (62.22% vs. 68.90%), p53 pathway (60.00% vs. 58.90%), PI3K pathway (51.11% vs. 52.05%), and Hippo pathway (35.56% vs 38.36%).
Co-occurrence and mutual exclusion among genetic events
To identify correlations among different genes in two UC subtypes, co-occurrence and mutual exclusion were performed for the top 10 gene in each cohort. We discovered that only FGFR3 mutations were mutually exclusive with TP53 mutations in both UTUC and UCB. In the UTUC cohorts, CREBBP mutations were significantly associated (co-occurred) with FAT1, FGFR3, and KMT2C mutations (Figure S2A). Similarly, KMT2C mutations co-occurred with SPEN and KMT2D mutations. In addition, NCOR1 mutations co-occurred with LRP1B mutations. Regarding the UCB cohorts, TP53 mutations were associated with RB1 mutation and were mutually exclusive with KMT2A and ERCC2 mutations. KMT2C mutations co-occurred with FAT1 and KMT2A mutations. Analogously, there were co-occurrences for ERCC2 with FAT1, and GNAQ with BRD4 mutations (Figure S2B).
Inference of mutation clonality for UTUC and UCB
To investigate differences of clonal evolution in UTUC and UCB, we thus inferred the clonal status. Combined with the analysis of all somatic cell mutations, including nonsynonymous SNVs and splice site (generally referred to as nonsilent) and synonymous SNVs (referred to as silent) within exonic, splicing, intergenic and intronic regions, we found that there were 2,040 clonal mutations and 22,330 subclonal mutations in UTUC. Regarding UCB, there were 195 clonal mutations and 2,482 subclonal mutations. Subsequently, clonal status was evaluated by nonsilent mutation, which may play a role in carcinogenesis. Five hundred twenty-seven clonal mutations and 287 subclonal mutations were identified in 82.22% (37/45) of UTUC patients. There were 34 clonal mutations and 106 subclonal mutations in 56.16% (41/73) of UCB patients. To our surprise, UTUC had higher clonal (p<0.001) and subclonal mutation numbers (p=0.015) than those in UCB (Figure 4A). There was a significant difference for the prevalence of genes of clonal mutations between UTUC and UCB. For example, clonal and subclonal mutations of KMT2D, KMT2C, PIK3CA and FAT1 existed only in UTUC (Figure S3).
Clonal driver mutations of Genomic Events
Cancer drivers only occur in a minority of somatic mutations, which confer clonal fitness and are positively selected over the course of tumor evolution (20). Clonal mutations represent the early events in tumor evolution. Conversely, subclonal mutations represent relatively late events based on occurring after the emergence of the most recent common ancestor (21). Thus, clonal driver mutation might contribute to identifying potential therapeutic strategies (20). In this study, we explored the clonal driver gene by the nonsilent with clonal mutation. Interestingly, driver genes were significantly different between UTUC and UCB. TP53, PIK3CA, and FGFR3 mutations are the driver genes for UTUC (Figure 4B), whereas for UCB, the driver gene only was BRCA1 (Figure 4C).
Assessment of DDR gene mutation suggesting potential benefit form immunotherapy
Thirty-four DDR genes related to the response of PD-1/PD-L1 blockade and sensitivity to cisplatin-based regimens were analyzed (13). Twenty-six deleterious DDR gene alterations were observed in 60% (27/45) of UTUC patients, with a median of 2 DDR alterations per patient (range, 1 to 10). There were 27 DDR gene alterations in 71.23% (52/73) of UCB patients, with a median of 2 DDR alterations per patient (range, 1 to 14). There was no difference for the frequency of DDR mutations between UTUC and UCB (60% vs. 71.23%, p=0.572). In UTUC, BRCA2 (n=9), ERCC2, and ATM (n=7 each) were the most frequently mutated genes, while in UCB, the most frequently mutated genes were ERCC2 (n=13), BRCA2, and MDC1 (n=12 each) (Figure 5A). The most commonly interfered pathways or mechanisms related were FA (40% and 47.95%), Checkpoint (37.78% and 43.84%) and MMR (31.11% and 31.51%) pathway in both UTUC and UCB. ERCC4 mutations were more frequent in UTUC than those in UCB (p=0.019). There were not significant differences in both frequency and number of DDR mutations (p=0.231) and other mutated genes (p=0.208) in both cohorts.
Pattern of somatic CNV in UTUC and UCB
It has been demonstrated that CNV can be used to categorize tumors into distinct sensitivity to ICI therapy (22). In this study, ROS1, EGFR,BRCA1, and NTRK1 mutations were the most commonly found in UTUC (Figure S4A) and UCB (Figure S4B), with ROS1 mutations appearing to be more frequent in UTUC than UCB (p=0.055). Conversely, patients with UTUC had less LOC284294 mutations than UCB patients (Figure 5B). The median CNV counts were 49.62 (range from 0 to 706.54) for UTUC and 48.00 (range from 0 to 537.89) for UCB. There were no significant differences for CNV numbers in either cohort (p=0.511, Figure 5C).
TMB comparison between UTUC and UCB
To determine whether the TMB value in UTUC and UCB can be used to screen the potential beneficiaries for immunotherapy, we compared differences in TMB between the two groups. The median TMB was 13.31 (range from 0.89 to 117.46) for UTUC and 15.76 (range from 3.99 to 62.04) for UCB. There were no significant differences in TMB in either cohort (p=0.489, Figure 5D).
Relationship among immune markers for DDR mutation, CNV, TMB, and PD-L1
We further explored the correlation between immune markers, including TMB, PD-L1 expression, and CNV and DDR gene mutation. In this study, we discovered that DDR gene alterations were associated with higher CNV counts in the UTUC cohort (p<0.001, Figure S5A). However, this correlation was not found in the UCB cohort (p=0.435, Figure S5A). Surprisingly, DDR gene alterations were associated with higher TMB in both UTUC and UCB tumors (p<0.001, p<0.001, respectively, Figure S5B). Interestingly, UTUC patients with lower CNV counts had a higher TMB than those with higher CNV counts (p=0.009, Figure S5C). By contrast, higher TMB was more common in UCB patients with high CNV (p=0.006, Figure S5C). Patients undergoing PD-L1 detection were selected to analyze the relationship. We discovered that PD-L1 was not associated with DDR gene mutation, CNV counts or TMB in both UTUC and UCB cohorts (p>0.05, Figure S3D-F).
Assessment of clinical actionability suggesting potential benefit from target therapy
To explore whether there are similar principles in the treatment of UTUC based on UCB and clinical utility for prospective molecular profiling to guide treatment, clinical actionability was evaluated using OncoKB (http://oncok b.org/). We found that 37 (82.22%) UTUC patients harbored at least one actionable alteration, covering 53.33% (24/45) of patients with the targeted drug (Figure 6A). Sixty (82.19%) patients with UCB harbored at least one actionable alteration, including 54.79% (40/73) of patients with the targeted drug (Figure 6B). The rates of receiving targeted drugs were similar in both groups, whereas KIT, NRAS, and CDKN2A only existed in UTUC (Figure 6C), and EGFR, MDM2, CDK4, and BRCA2 in UCB (Figure 6D).