A comprehensive search was conducted in the database of Shandong Provincial Hospital Affiliated to Shandong First Medical University, Jinan, China, to identify all patients diagnosed with bladder urothelial carcinoma (BUC) and/or upper tract urothelial carcinoma (UTUC) who underwent surgical resection between 2018 and 2022. Patients who received neoadjuvant chemoradiotherapy were excluded from the study. Clinical data corresponding to the patients were obtained through a retrospective review of their medical records. Following pathological review, triplicate 1 mm diameter cores were sampled from epithelium-rich areas of formalin-fixed paraffin-embedded (FFPE) primary tumor tissue to construct a tissue microarray (TMA) using the TMA Master II instrument (3DHISTECH, Budapest, Hungary). Hematoxylin and eosin-stained (H&E) sections was examined from each TMA block to confirm the adequacy of tumor samples.

Both microarray and whole section slides underwent the same staining technique. 5 μm thick sections of FFPE tissue were deparaffinized, rehydrated, and subjected to antigen retrieval by boiling in a sodium citrate buffer (pH 6.0) in a pressure cooker. Following antigen retrieval, the slides were blocked with 3% H₂O₂ and 5% goat serum, and the incubated with the primary antibody (MLH1: clone ES05, 1:100, PMS2: clone EP51, 1:50, MSH2: clone RED2, 1:100, and MSH6: clone EP49, 1:50, PD-L1: clone 22C3, 1:250, DAKO, Santa Clara, CA, USA; CD8: clone SP16, 1:200, CD4: PD-1: clone OTI4F10, 1:100, CD3, clone LN10, 1:100, ZSGB-BIO, Beijing, China). After washing with PBS, slides were incubated with biotinylated secondary antibody (ZSGB-BIO, Beijing, China), followed by ABC reagent (ZSGB-BIO, Beijing, China) and stable diaminobenzidine (ZSGB-BIO, Beijing, China). All sections were lightly counterstained with hematoxylin and mounted with aqueous mounting medium (ZSGB-BIO, Beijing, China). Negative controls were prepared using mouse or rabbit immunoglobulin-G at the same concentration as the primary antibodies.

The absence of detectable nuclear staining in neoplastic cells was considered indicative of deficient expression of MLH1, MSH2, MSH6, and PMS2. The adjacent non-neoplastic epithelium, stromal cells, or lymphocytes exhibiting intact nuclear staining were used as internal positive controls. CD3+ T cells, CD8+ T cells, and CD4+ T cells were quantified using ImageJ software (https://imagej.nih.gov/ij/index.html). Ten to twenty fields of view were counted at 40 ×, depending on the density of the CD8+ T cells within the tumor. The tumor proportion score (TPS) is determined as the percentage of PD-L1 positive stained tumor cells (TCs) with at least partial membrane staining relative to the total number of TCs, excluding tumor-associated interstitial cells, necrotic, normal, or non-neoplastic cells from the evaluation. The combined positive score (CPS) was defined as the number of PD-L1 or PD-1 positive cells (TCs, lymphocytes, and macrophages) divided by the total number of TCs multiplied by 100.

Genomic DNA was extracted from FFPE dMMR tumor tissues and corresponding normal tissues using the QIAamp DNA FFPE Tissue Kit (Qiagen, Shanghai, China), following the manufacturer’s instructions. The sheared genomic DNA fragments ranging from 200 to 400 bp were purified using the Agencourt AMPure XP Kit (Beckman Coulter, CA, USA). These fragments were then subjected to hybridization with capture probes baits, followed by selection using magnetic beads and amplification. Target capture was performed using a commercial OncoScreen Plus 520 panel (Burning Rock Biotech, Guangzhou, China), which covers genes related UC, tumor mutational burden (TMB), microsatellite instability (MSI), DNA damage response (DDR) pathway genes, as well as other factors relevant to positive and negative prediction and hyperprogression in immune and targeted therapies. The OncoScreen Plus 520 panel was developed through literature research, database analysis, clinical validation, and the selection of validated genes and biomarkers, and has been applied in gene testing studies for various types of cancer in both clinical and experimental settings (12, 13). Indexed libraries were subsequently sequenced on Illumina NextSeq 500 system (Illumina, Inc., USA) with paired-end reads and average sequencing depth of 1000×. The analysis of sequence data, along with variant verification, MSI status assessment, and TMB determination, followed previously described protocols (12, 13). The MSI status was detected with MSIsensor (V0.2), which was defined as the percentage of unstable microsatellite sites divided by total number of microsatellite sites surveyed. MSIsensor scores ≥ 10 and < 10 are defined as MSI-High (MSI-H) and Microsatellite Stable (MSS), respectively (14). TMB was defined as the number of all nonsynonymous mutations and indels per megabase of the genome examined. TMB > 10 Muts/Mb was defined as TMB-High (TMB-H). Tumor neoantigen burden (TNB) was measured as the number of mutations that could generate neoantigens per megabase (Neos/Mb). TNB > 4.5 Neos/Mb was defined as TNB-High (TNB-H) (15).

To test the MSI status based on Polymerase Chain Reaction (PCR) and capillary electrophoresis resolution (MSI-PCR), DNA was extracted from FFPE tumor and normal tissue for each case. Sections were deparaffinized, rehydrated, and digested with proteinase K. From the lysate, the DNA was purified using the Wizard DNA Clean-Up System (Promega, Mannheim, Germany). For PCR, 25 ng of DNA were used per reaction. MSI testing was performed using the MSI Analysis System, Version 1.2 (Promega, Beijing, China), analyzing five nearly monomorphic mononucleotide markers (BAT-25, BAT-26, NR-21, NR-24, and MONO-27) for MSI determination and two polymorphic pentanucleotide markers (Penta C and Penta D) for sample identification. Negative, positive, and water (blank) controls were used. MSI-high (MSI-H) was defined as the presence of novel alleles in the tumor sample, but absence in the corresponding normal sample (indicating MSI) at two or more of the five mononucleotide repeat markers tested. MSI-low (MSI-L) was defined as any cases with peak shifts within one of the five regions tested. Microsatellite stable (MSS) was defined as no microsatellite instability detected at all five mononucleotide repeat markers tested.

Statistical significance of differences between clinical and demographic variables was assessed using Pearson’s Chi-square test or Fisher’s exact test. The comparison of age between dMMR and pMMR groups was conducted using an independent samples t-test. The correlation between the two variables was evaluated using the Pearson correlation coefficient or Spearman’s rank correlation test. Survival analysis was performed using the Kaplan–Meier method with the log-rank test and the Cox proportional hazards model. Hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were estimated from the Cox analyses. A significance level of P < 0.05 was considered statistically significant. All analyses and graphs were conducted using SPSS Statistics 21 software (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 9.5 software (San Diego, CA, USA).

The cohort’s clinicopathological parameters are summarized in Table 1 . We analyzed a total of 374 patients with BUC and 232 patients with UTUC in this study. Among them, we identified 15 cases of UC with simultaneous loss of MSH2 and MSH6 using TMA and IHC. These cases included 1 involved BUC, 9 involved UTUC, and 5 involved synchronous or metachronous occurrences of BUC and UTUC. The incidence of dMMR BUC within the BUC subgroup was only 1.6%. In the cohort, no cases with MLH1 and/or PMS2 loss were identified. dMMR BUC had a slightly lower average onset age (62.3 ± 16.7 vs. 65.4 ± 11.4) compared to proficient mismatch repair (pMMR) BUC. However, these differences were not statistically significant (P > 0.05). Furthermore, dMMR BUC showed an increased likelihood of tumor metastasis (M1 stage) compared to pMMR BUC (P < 0.05). There were no significant differences observed between dMMR and pMMR BUC in terms of patient gender, tumor grade, T stage, or N stage (P > 0.05). Among the UTUC cases, dMMR UTUC accounted for 6% of the total cases. Similar to dMMR BUC, dMMR UTUC had a slightly lower average age of onset (64.0 ± 13.5 vs. 65.7 ± 9.9) compared to pMMR UTUC. However, these differences were not statistically significant (P > 0.05). Unlike pMMR UTUC, dMMR UTUC more frequently presented as low-grade tumors (50% vs. 84.9%) and predominantly fell within the Ta-T1 stage (78.6% vs. 29.8%, Tables 1 and 2 ). Additionally, we observed that dMMR BUC and dMMR UTUC often co-occurred synchronously or metachronously (33.3%). Furthermore, patients with dMMR BUC and UTUC exhibited a significantly higher rate of personal and family histories of cancer than pMMR UC (P < 0.001) ( Table 1 ).

We evaluated the histopathologic characteristics of dMMR UC by reviewing all H&E sections. Inverted growth was observed in all 15 (100%) cases, with 10 cases exhibiting purely inverted characteristics ( Figure 1A ) and 5 cases displaying a mixed morphology characterized by focal inverted and papillary patterns ( Figure 1B ). Among the total cases, 8 (53.3%) cases were classified as high grade. 14 (93.3%) cases demonstrated pushing borders ( Figure 1C ), while only 1 (1.7%) case exhibited destructive infiltrative edges ( Figure 1D ). Immunohistochemical staining in a representative dMMR case revealed loss of MSH2 ( Figure 1E ) and MSH6 ( Figure 1F ).

According to patient informed consent, NGS was conducted on tumor tissues and corresponding normal tissues from 10 cases to investigate the somatic mutational landscape of dMMR UC. The most frequently mutated gene in dMMR UC was KMT2D (60%), with 3 (50%) cases exhibiting nonsense mutations, 2 (33.3%) cases showing frameshift mutations, and 1 (16.7%) case displaying a splice sites mutation ( Figure 2 ). Notably, Patient 8 and 15 harbored the same frameshift mutation site (KMT2D c.3704del, p.G1235fs), while Patient 9 and 10 carried the same nonsense mutation site (KMT2D, c.5707C>T, p.R1903*). The second most frequently mutated gene in dMMR UC was KMT2C (50%), while 3 (60%) cases showing missense mutation, 1 (20%) case exhibiting a nonsense mutation, and 1 (20%) case displaying a frameshift mutation. Other notable variants included FGFR3 (40%), CTCF (40%), KDM5A (40%), MET (40%), MSH3 (40%), NOTCH3 (40%), PIK3CA (40%), POLD1 (40%), and SMARCA4 (40%) ( Figure 2 ). Moreover, Patient 1, 9, 10 and 15 carried the same missense mutation site of FGFR3 (c.742C>T, p.R248C) (40%), with two of them diagnosed as LS, and the other two without MSH2 germline mutations. Less frequent mutations in dMMR UC included ARD1A (30%), ATM (30%), CREBBP (30%), FBXW7 (30%), ROS1 (30%), and others ( Figure 2 ).

In the clinical assessment, 10 (66.7%) out of 15 UC with dMMR had personal histories of Lynch syndrome (LS)-associated neoplasms, such as CRC and EC. Additionally, 9 (60%) cases had family histories of these neoplasms. Six (40%) cases satisfied the Amsterdam II criteria, while 10 (66.7%) cases fulfilled the revised Bethesda guidelines, both indicating a suspicion of LS. Figure 3 displays the representative pedigrees. To identify potential germline variants associated with LS, germline MLH1, PMS2, MSH2, MSH6, and EPCAM genes were tested using NGS in 13 cases of dMMR UC based on patient informed consent. The analysis revealed pathogenic mutations in the MSH2 gene in 7 (53.8%) cases, including 5 cases with nonsense mutations and 2 cases with splice variants ( Table 2 ).

Microsatellite Instability (MSI) status was assessed in dMMR UC using two different methods: NGS (MSI-NGS) and the MSI Promega Analysis System (MSI-PCR). The results are summarized in Supplemental Table S1 . MSI-H was observed in 8 (53.3%) cases using the MSI-NGS method, while MSI-L and MSS were found in 0 and 7 (46.7%) cases, respectively. With the MSI-PCR method, MSI-H was detected in 11 (91.7%) cases, MSI-L in 1 (8.3%) case, and none of the 12 cases showed MSS ( Supplemental Figure S1 ). Among the analyzed cases, instability was observed in 5 mononucleotide markers: BAT-25, MONO-27, NR-21, BAT-26, and NR-24, with frequencies of 75%, 66.7%, 66.7%, 58.3%, and 41.7%, respectively. A total of 12 dMMR UC samples underwent MSI testing using both methods, with an overall concordance rate of 41.7% (5 out of 12 cases). Among the cases identified as MSS by the NGS method, 6 out of 7 cases (85.7%) were shown to be MSI-H using the MSI-PCR method, indicating a sensitivity of 45.5% for the NGS method. The concordance rate for MSI-H status assessed using NGS and dMMR evaluated by IHC was 41.7%. Notably, the concordance rate for MSI-H status between the MSI-PCR testing and dMMR evaluation by IHC was 91.7%, indicating a high level of consistency.

TMB analysis was conducted using NGS. Among the 15 cases examined, 6 (40%) displayed high TMB (TMB-H) with a range of 29.91 to 68.79 mutations per megabase (mut/Mb), and a mean of 44.53 mut/Mb ( Supplemental Table S1 ). TMB exhibited a strong positive correlation with the number of gene mutations (R² = 0.9830, P < 0.01, Supplemental Figure S2A ). Out of the 11 cases analyzed, there was a concordance rate of 4 (36.4%) between the identification of MSI-H by PCR and TMB-H. Furthermore, TMB displayed a positive correlation with the proportion of unstable microsatellite loci detected by NGS (R² = 0.6208, P < 0.01, Supplemental Figure S2B ).

TNB predictions were conducted for patients with dMMR UC. Among the examined cases, 6 (40%) exhibited a high TNB (> 4.5 Neos/Mb), ranging from 20 to 90 Neos/Mb, with a mean of 55.17 mut/Mb. As expected, TNB demonstrated a significant positive correlation with the number of gene mutations (R² = 0.5593, P < 0.05, Supplemental Figure S2C ), the proportion of unstable microsatellite loci detected by NGS (R² = 0.6208, P < 0.01, Supplemental Figure S2D ), and TMB (R² = 0.6460, P < 0.01, Supplemental Figure S2E ).

Morphologically, PD-1 immunostaining was observed on the membrane of immune cells (ICs) with CPS scores ranging from 1 to 60 ( Supplemental Table S1 ). Positive PD-1 expression was found in 9 (60%) cases. Notably, a significant number of intratumoral ICs exhibited intense positive staining of PD-1 ( Figure 4A ). Positive PD-L1 immunostaining was observed on the cell membrane of tumor cells ( Figure 4B ). Based on the TPS and CPS scores for PD-L1 expression, three (20%) cases had TPS ≥ 1% and CPS ≥ 1, while twelve (80%) cases had negative PD-L1 status. The TPS scores of cases with positive PD-L1 expression ranged from 1% to 5%, while the CPS scores ranged from 5 to 15. A strong positive correlation was observed between the TPS and CPS scores of PD-L1 staining (R² = 0.925, P < 0.01, Figure 4C ). However, no correlation was found between PD-1 and PD-L1 expression (P > 0.05).

Peritumoral CD3+ T cell infiltration was observed in 15 cases with dMMR UC, with a density ranging from 724.01 to 4070.54/mm² ( Supplemental Table S2 ). In 80% of the cases, the density of peritumoral CD3+ T cells was > 2000/mm². The density of peritumoral CD8+ T cells ranged from 2347.27 to 84.62/mm², with 6 (40%) cases exceeding 800/mm² ( Figure 4D ). The percentage of peritumoral CD8+ T cells among CD3+ T cells ranged from 10.58% to 65.05%. In 5 (33.3%) cases, the percentage of peritumoral CD8+ T cell was ≥30%. A significant positive correlation was observed between peritumoral CD8+ and CD3+ T cell density (R² = 0.4590, P < 0.01). The density of intratumoral CD8+ T cells ranged from 12.09 to 873.60/mm² ( Supplemental Table S2 ). In 7 (46.7%) cases, the intratumoral CD8+ T cell density exceeded 100/mm². Morphologically, these cases show abundant CD8+ T cells distributed in patches or clusters within the tumor cell nests ( Figure 4E ). The density of peritumoral and intratumoral CD8+ T cells displayed a positive correlation (R² = 0.6091, P < 0.01, Supplemental Figure S2F ). Moreover, positive correlations were also found between intratumoral CD8+ T cells density and PD-L1 score (TPS, R² = 0.531, P < 0.01), PD-L1 score (CPS, R² = 0.4614, P < 0.01), and PD-1 score (CPS, R² = 0.402, P < 0.05) ( Figure 4F ). No correlation was found between the intratumoral CD8+ T cell percentage and TNB (R² = 0.1246, P > 0.05). The evaluation of CD4+ T cell content showed that in 6 (40%) cases, the peritumoral CD4/CD3 ratio was > 50%.

NGS analysis revealed variants in 20 genes related to the DDR pathway ( Figure 5 ). POLD1 and SMARCA4, positively correlated biomarkers for predicting the efficacy of immune checkpoint inhibitors (ICIs), exhibited variations in 4 (40%) cases each. Additionally, ATM, MSH2, and PRKDC mutations were detected in 3 (30%) cases each. On the other hand, the negatively correlated biomarker MET showed variants in 4 cases (40%), with only Patient 1 having a missense mutation (c.3979C>T, p.R1327C) within the tyrosine kinase domain. Patient 1, 9, and 15 exhibited missense mutations in ALK, EGFR, and STK11, respectively. Resistance biomarker analysis for immunotherapy identified JAK1 mutations in Patient 8 and 10, JAK2 and VEGFB/PTEN mutations were found in Patient 9 and 15, respectively. Additionally, DNMT3A missense mutations were observed in Patient 9 and 10, associated with hyperprogressive disease.

In this study, the median follow-up duration for all patients was 31.5 months (range: 7-69 months). While patients with dMMR BUC (41.0 vs. 31.3 months) and dMMR UTUC (32.6 vs. 28.7 months) showed higher average overall survival (OS) compared to their pMMR counterparts. Kaplan-Meier analyses revealed no significant differences in OS between pMMR and dMMR BUC (P > 0.05) or pMMR and dMMR UTUC (P > 0.05) ( Figure 6 ). To investigate the association between clinicopathological characteristics and immunotherapy biomarkers, including age, tumor location, tumor grade, TNM stage, MSI (PCR), TMB values, intrastromal and intratumoral CD8+ T cells density, PD-L1 score (CPS), and OS, univariate Cox regression analysis was performed. The results showed that only the density of intratumoral CD8+ T cells exhibited a significant association with OS in our cohort (P < 0.05, HR = 0.170, 95% CI: 0.034-0.844).

Patient 3, diagnosed with LS, initially underwent radical nephroureterectomy for right UTUC. Two and a half years later, the patient underwent left ureteral resection for upper and lower ureteral UC, along with nephrostomy. The patient subsequently received treatment with Tirelizumab for 5 cycles. Unfortunately, the patient died of systemic sepsis caused by the nephrostomy.

Two cases of metastatic dMMR UC (Patient 14 and 15) achieved complete remission following combination treatment with PD-1 inhibitors and chemotherapy. As our previous reports¹¹, Patient 15 had concurrent ureteral and bladder UC, initially treated with nephroureterectomy and transurethral resection of bladder tumor. Despite disease progression with local bladder recurrence and lung/bone metastasis after cisplatin-based first-line chemotherapy and docetaxel second-line monotherapy, the patient achieved sustained response after five cycles of PD-1 inhibitor sintilimab (200 mg on day 1) combined with docetaxel (120 mg on day 2, every 21 days) over 31 months, showing disappearance of nodules in the bladder, lung, and left proximal femur. Patient 14 underwent partial cystectomy and right ureter reimplantation for BUC. Four years later, imaging revealed retroperitoneal metastasis and right ureteral UC ( Figure 7 ). After 8 cycles of PD-1 inhibitor pembrolizumab (200 mg on day 1) combined with gemcitabine (1.4 g on day 1, every 8 days) therapy, both the retroperitoneal metastatic lesions and ureteral tumor significantly decreased in size. In terms of treatment side effects, Patient 15 experienced sintilimab-associated skin allergies, hypophysitis, and hypothyroidism, which improved with treatment¹¹. Patient 14 experienced vomiting and severe neutropenia, which normalized after receiving granulocyte colony-stimulating factor. Thus, the side effects of immunotherapy in dMMR UC were manageable.