Patient baseline data and prognosis
The clinicopathological data of 58 collected patients were shown in Table 1. The mean age was 61.91 years (standard deviation: 7.67 years), and the median follow-up time was 32.86 months (interquartile range: 21.00–44.49 months). There were no significant differences in baseline clinicopathological data between BCG and EPI subgroups except for T stage. For BCG subgroup, recurrence occurred in 9 patients, with 6 of them recurrent during course of intravesical BCG therapy. Two years RFS rate for BCG group was 82.2%; while for EPI subgroup, 7 patients recurrent and five occurred during EPI treatment course. Two years RFS rate for EPI was 58.3%. Univariate analysis revealed that significant better RFS was seen in
BCG treated subgroup (P=0.020) and those with incipient tumor (P=0.001 for all 58 patients and P=0.014 for 46 BCG treated patients) (Additional file 1 table A1).
Genomic landscape of patients with NMIBC and relationship with prognosis
NGS was performed on the total 58 patients, the most frequently altered genes were FGFR3 (48%), KDM6A (47%), KMT2D (43%), KMT2C (34%), and STAG2 (31%). Oncoprints for these genes are presented in Additional file 2 Figures A1–S7). Significant mutations of FGFR3 were observed in 73.1% of intermediate-risk patients (19/26) and 15.6% of high-risk patients (5/32). For STAG2, these numbers were 46.2% (12/26) and 18.8% (6/32), respectively. Mutations of FGFR3, STAG2, and PRKDC were significantly associated with tumor risk stage (P<0.05, Additional file 1 Table A2). Moreover, we found that mutations of 17 genes (such as KDM6A and ARID1A) were significant correlated with a higher EORTC score, while mutations of other 17 genes (such as TP53) were significantly correlated with a higher CUETO score (Additional file 1 Table A3).
Cox univariate analysis was performed to identify correlations of genomic mutations with recurrence. In all 58 cases, 18 genes were significantly correlated with recurrence (Table 2 and Additional file 1 Table A4); all these mutations were positively associated with poor prognosis. However, among these genes, only mutations of NEB, MLH1, GATA3, FGFR1 and RAF1 occurred in >5% of patients. Similar results were observed in the BCG subgroup (46 cases), with 10 genes significantly associated with recurrence (P<0.05, Table 2 and Additional file 1 Table A5). The multivariate Cox analysis included all genes significantly associated with recurrence and the results demonstrated that NEB, MLH1, FGF12 and FGFR1 were independently predicted poor prognosis for all 58 patients (P=0.001, 0.001, 0.001 and 0.007 respectively), while NEB, FGFR1 and SDHC played a role as an independent prognostic predictor for recurrence in patients treated with BCG (P=0.001, 0.004 and 0.017 respectively) (Additional file 1 Table A6).
Genomic pathways and recurrence
Canonical genomic pathways were analyzed to define whether mutations were associated with prognosis (Figure 1). Mutations of the receptor tyrosine kinases/RAS/phosphatidylinositol 3-kinase (RTK/RAS/PI3K) pathway were found in 87.9% of patients, and were associated with a higher risk stage (P=0.013, Additional file 1 Table A7). Epigenetic-related gene mutations were also frequently detected in patients with NMIBC (49/58, 84.4%), with markedly lower rates of mutations in the TP53/cell cycle pathway, Switch/sucrose nonfermentable pathway, DNA damage pathway, and alternative splicing pathway (50.0%, 31.0%, 32.8%, and 34.5%, respectively). Meanwhile, correlations between the number of pathway mutations and prediction models were also analyzed: larger numbers of mutated genes in the epigenetic pathway, histone modification pathway, and Switch/sucrose nonfermentable pathway were significant correlated with a poorer EORTC score, whereas there was no significant correlation found with the CUETO score (Additional file 1 Table A7).
Correlations of pathway mutations with recurrence were analyzed by Cox regression in qualitative and quantitative dimensions (i.e., whether mutations are present and the number of mutations in each pathway) (Table 3). In qualitative analysis, mutations in epigenetic-related genes, RTK-PI3K pathway, SWI/SNF pathway and alternative splicing pathway were protective factors (hazard ratio [HR]<1), while TP53/ cell cycle pathway mutations were risk factor, no matter in all 58 patients’ cohort or in BCG subgroup. Besides, mutations in epigenetic-related genes were significantly associated with recurrence in qualitative analysis (P=0.020 and 0.023 for total patients and for BCG subgroup separately). Nevertheless, when quantitative analysis performed, we found that no pathways mutations were significantly correlated with recurrence (P>0.05, Table 3).
DNA damage response and repair (DDR)genes and TMB
DDR genes have been associated with increased mutation load and the effect of immunotherapy against urothelial carcinoma9. We also analyzed the correlation between DDR, TMB, and prognosis (the list of DDR genes is provided in Additional file 1 Table A8 and the alteration status is shown in Figure S8). In our study, the percentages of DDR mutations in 58 cases with NMIBC and BCG subgroups were 86.2% and 84.8%, respectively (the number of mutations was 165 and 141, respectively). TP53 was the most frequently mutated gene among the DDR genes (19/165 in the total cases and 17/141 in the BCG subgroups). Of note, missense variant was the most common alteration in TP53 (16/19 and 14/17 in the total and BCG group, respectively) (Additional file 3 Figure A8). In our previous analysis, TP53 mutation was correlated with a poorer CUETO score; however, there was no correlation found with tumor risk group stratification, EORTC score, or RFS (Additional file 1 Tables A2–5). When considering mutations in all DDR genes, there was no correlation demonstrated with the predicted models and RFS (P>0.05) (Table 3 and Additional file 1 Table A7).
In the analysis of TMB, significant correlation was found with the EORTC model (P-values for correlation with risk group stratification, CUETO, and EORTC scores were 0.485, 0.706, and 0.037, respectively). There was no significant correlation between TMB and RFS (HR: 0.972, P=0.316 for 58 NMIBC cases and HR: 0.976, P=0.497 for 46 BCG subgroup cases). However, significant correlations were demonstrated between alterations in DDR and TMB, and between DDR mutated numbers and TMB (Figures 2A, B). In addition, TP53 mutations were also significantly associated with TMB (Figure 2C).