A total of 17 patients diagnosed with urothelial carcinomas were retrospectively analyzed in our study. Resection was performed for each tumor lesion at the PLA General Hospital. None of them underwent chemotherapy or radiation therapy. Of the 17 patients, 11 (p1-p11) had metachronous recurrences with an interval time of more than three months. After-surgery tissues of primary and paired recurrent tumors were collected for NGS and immunofluorescence-based microenvironment analysis. Clinical data were obtained, including the patient’s age, sex, AA- intake history, body mass index (BMI), concomitant diseases, tumor size, pathological stage, tumor sites, and recurrence-free survival (RFS) from the electronic medical health records. This study was approved by the Ethics Committee of the PLA General Hospital. Written informed consent was obtained from each patient.

AA exposure assessment was defined according to previous research assessed by self-reported data (17, 18): (1) the presence of a definite history of taking AA-containing herbal drug, Guan Mu Tong (Aristolochia manshuriensis), Guang Fangchi (Aristolochia fangchi), Qing Mu Xiang (Radix Aristolochiae), Ma Dou Ling (Fructus Aristolochiae), Tian Xian Teng (Caulis Aristolochiae), Xun Gu Feng (herba Aristolochiae mollissimae), and Zhu Sha Lian (Kaempfer Dutchmanspipe Root); (2) the duration of exposure of the above drugs was more than three months.

Genomic profiling was performed using a panel covering 520 cancer-related genes (OncoScreen plus, Burning Rock Biotech, Guangzhou, China), allowing for the evaluation of complex biomarkers such as tumor mutational burden (TMB). The Catalogue of Somatic Mutations in Cancer (COSMIC) mutational signatures were downloaded from the COSMIC website (https://cancer.sanger.ac.uk/cosmic).

Tumor tissues were used to perform immunofluorescence-based tissue microenvironment analysis. Multiplex staining for seven key immune markers, PD1, PD-L1, CD3, CD8, CD56, CD 68, and CD163, was performed with PANO 7-plex IHC kit (Panovue, Beijing, China) according to the manufacturer’s protocol. All stained tissues were independently scored by two pathologists who were blinded to the clinical parameters. Our study reported the density and positive rate of tumor-infiltrating lymphocytes (TILs) in both parenchymal (iTILs) and stromal (sTILs) compartments.

Statistical analysis was performed using R version 4.1.0. The Fisher exact test or a non‐parametric test was used to compare categorical data. The Wilcoxon test between groups analyzed the difference in the infiltration of immune markers and TMB. P-values <0.05 were considered statistically significant.

Targeted NGS was conducted across multiple tumor specimens in 17 patients, 9 of whom have documented AA exposure. At diagnosis, the median age was 67.2, and the median primary tumor size was 3.1 cm. Most patients were women (n = 10). No Obese (BMI > 30 kg/m²) was identified in all the patients. Most patients were overweight (BMI 25.1–30 kg/m²) and normal weight (BMI 18.5–25 kg/m²). Nine had at least one concomitant disease, including aristolochic acid nephropathy, chronic kidney failure, coronary artery disease, diabetes, hypertension, and renal failure. Fourteen patients had UC originating from the upper tract, and the remainder had LTUC of the bladder (n = 3). The majority of tumor stage was T2 or worse, and none of them had shown metastasis. 82.4% of tumors were high-grade. Patient demographics and clinical features were provided in Table 1 .

Next, we sought to determine whether UC showed specific mutational signatures with AA exposure. Using the COSMIC, SBS V2 signatures of mutational processes in human cancer, we identified 24 mutational signatures ( Figure 1A ). Hierarchical clustering based on the COSMIC signatures confirmed two significant cohorts: AA Sig (n=11) and non-AA Sig (n=6) ( Figure 1B ). There was a significantly higher number of signature 22 mutations in the AA Sig cohort than in the non-AA Sig cohort. Analysis showed that the AA Sig cohort was associated with AA-containing herbal drug intake, recurrence ( Table 1 ), and higher TMB (35 vs. 6 mut/Mb, Figure 1C ). The AA mutational signature was observed in two patients’ tumors without previous indication of AA exposure ( Table 1 ).

We identified 864 mutations in the two cohorts, with a median of 38 (range, 19 to 137) somatic mutations per tumor in AA UC patients versus 18 (range, 11 to 69) in non-AA. 276 and 102 genes were mutated in the AA Sig cohort and non-AA Sig cohort in total, making it difficult to distinguish mutated drivers from mutated passenger genes. The most frequently mutated genes in AA Sig and non-AA Sig included KMT2D (72.7% and 66.6%), TP53 (63.6% and 50%), and FLT1 (72.7% and 33.3%, Figure 1D ). The mutation frequency of FLT1, SPTA1 and PRKDC was higher in AA Sig patients than non-AA Sig (72.7% vs. 33.3%, 72.7% vs. 16.7%, and 72.7% vs. 0%, respectively; p < 0.05). FGF3 was higher in non-AA Sig patients (36.3% vs. 66.7%).

The vast majority (74.1% and 41.3% in AA Sig and non-AA Sig cohorts, respectively, Figure 2A ) of these mutations were single-nucleotide variation (SNV) missense mutations. The SNVs in the AA Sig showed a marked mutagenic signature, with T>A missense accounting for 72.6% of SNVs. In contrast, C>T missense accounted for 61.1% in the non-AA cohort ( Figure 2B ). In AA Sig, we also observed a preference for a C in the base preceding the mutated T residue and a preference for G at the following base in T>A mutation ( Figure 2C ). We also found copy number variations (CNVs) in both cohorts, and the frequency of CNVs in non-AA Sig was higher ( Figure 2D ).

To delineate differences between recurrent AA Sig and non-AA Sig, we compared the mutational signatures of 30 tumor specimens from 11 recurrent UC patients, in which 4 of them developed multiple recurrences (p3, 6, 8, and p10) ( Figure 3A ). 9 out of the 11 recurrent patients identified AA signature and the recurrent tumors in the AA Sig cohort consistently showed AA signature, which indicated that AA exposure might contribute to recurrences ( Figure 3B ). In both cohorts, the TMB in recurrent tumors was similar to primary tumors ( Figure 3C ). The most frequently mutated genes in primary and recurrent tumors included KMT2D (72.7% and 84.2%), TP53 (54.5% and 84.2%), FAT1 (72.7% and 42.1%), and FLT1 (72.7% and 42.1%, Figure 3D ). These results demonstrated the mutation signature between primary and recurrent tumors was similar. The AA mutational signature was consistently identified in paired tumors of AA UC cohort but not non-AA cohort.

The clonal evolution theory of UC, especially the UC with AA signature, is not well studied. The single or multiple recurrent tumors arising in individual patients are clonally related recurrences if they inherit an identical or similar set of somatic genetic alterations; otherwise, they may represent distinct primary tumors. To further determine the clonal relatedness of the recurrent tumors, we probed the genetic aberrations identified by NGS of all primary and recurrent tumors, including multiple recurrences in patients 3, 6, 8, and 10. Of the 11 patients (p1-p11), nine were from the AA Sig cohort (p1-p9) and two were from the non-AA Sig cohort (p10, p11). Our results revealed three types of clonal evolution patterns in our patients ( Figure 4 ). Clonal architecture of UCs indicated that the paired recurrent tumors in 7 patients (63.6%) shared minor genetic alterations with the primary tumors, revealing their different clonal origins ( Figure 4A ). Less than 2% of somatic mutations were present in primary and recurrent tumors (Fig 4D). These recurrent cases were considered second primary tumors since the primary and recurrent tumors arose from independent clones. Of note, they were all classified as AA Sig cohort. However, in the non-AA Sig cohort, shared mutations were found in primary tumors and recurrent tumors ( Figure 4B ). The shared mutations accounted for 59.09% of all the mutations in p11, and 56.25% and 50% of all the mutations in p12 who developed recurrences twice ( Figure 4E ), demonstrating shared clonal origin of primary and recurrent tumors.

In addition, the clonal origins of some patients showed heterogeneity ( Figure 4C ). When compared with the primary tumor of p6, 84.21% and 86.84% of shared mutations were present in the second (p6-2) and fourth (p6-4) tumor, but only 2.63% were found in the third tumor (p6-3), which was considered second primary tumor ( Figure 4F ). In another patient, p3, the second tumor inherited an identical set of somatic genetic alterations from the primary tumor ( Figure 4F ). However, the third tumor only shared 4.35% somatic genetic alterations with the primary tumor. It was considered the second primary tumor, with which more than 85% of mutations of the later recurrences were shared ( Figure 4G ). These data suggest that these AA UCs likely have multiple independent origins.

To characterize the TME of UCs, we compared the density and positive cell rate of seven key immune markers, PD1, PD-L1, CD3, CD8, CD56, CD68, and CD163, with multiplex staining. In primary tumors, there was no significant difference between AA Sig and non-AA Sig patients ( Figure S1A ). We further compared the expression of each marker in sTILs and iTILs, and AA UC patients had significantly higher CD56+ iTILs ( Figure S1B ).

In particular, we analyzed the expression of these immune markers between primary and recurrent tumors based on their clonal relatedness. The primary tumors of each patient were classified as group “First primary.” The recurrent tumors were classified as group “First primary recurrent,” in which shared mutations were detected in the recurrent tumors in p6-2, p6-4, and p3-2. Little or no common mutations of primary tumors were observed in the recurrent tumors in p1-2, p2-2, p4-2, p5-2, p7-2, p8-2, p9-2, p3-3, and p6-3 (group “Second primary”), and their clonally related recurrent tumors p3-4/5/6 were classified as group “Second primary Recurrent.” The TME of second primary tumors showed different TME between first primary and recurrent tumors in tumor nests and stroma ( Figures 5A, B ). With paired samples Wilcoxon test, second primary tumors were associated with significantly higher CD68+CD163- sTILs (marker of M1 macrophages, Figure 5C ), CD8+ cytotoxic sTILs ( Figure 5D ), PD-L1+ sTILs ( Figure 5E ) and iTILs ( Figure 5F ). These data indicated the immunoreactive TME ( Figure 5G ) in second primary tumors.

To investigate whether the TME plays a role in the development of recurrent tumors, we classified all the tumor specimens into two groups based on their clonal relatedness compared with the tumor before recurrence, similar clonal (clonal-yes: p3-1, p3-3, p3-4, p6-1) and different clonal origin (clonal-no: p1-1, p2-1, p4-1, p5-1, p7-1, p8-1, p9-1, p3-2, p6-2, p6-3). Microenvironment analysis showed that the clonal-yes group had fewer iTILs than the clonal-no group ( Figure 6A ). We found that CD3+ cytotoxic iTILs, CD68+CD163- sTILs, CD56+ iTILs were highly infiltrated in recurrent tumors with different clonal origins ( Figures 6B–D ).

These results suggested the primary AA UC tumors with immunoreactive TME constituted of CD3+ T cells, M1 macrophages, and natural killer (NK) may develop recurrence with new clonal origin.