Genomic Proling of Chinese Urothelial Carcinoma Patients

Urothelial carcinoma (UC) is the most common genitourinary malignancy in China. In this study, we studied the genomic features in Chinese UC patients, and investigated the concordance of genetic alterations between serum circulating tumor free DNA (ctDNA) and matched tumor tissue. urothelial of bladder; tract

last three years, especially with the development and approval of the ve immune checkpoint inhibitors (ICIs) and one FGFR inhibitor by the U.S. Food and Drug Administration (FDA).
Previous studies have widely characterized the genomic features of urothelial carcinoma, but the samples were mainly limited to the Caucasians [4]. To date, rare research had studied the molecular features of Chinese UC patients. The genomic alterations between western and Chinese UC patients may differ, especially considering the exposure of aristolochic acid and the different genetic backgrounds [5].
Meanwhile, as most of the genomic research in UC conducted on the tumor tissue samples, patients without su cient and valid achieved tumor samples or unable to biopsy will miss the chance of target therapies. Recently, circulating cell-free tumor DNA (ctDNA) has been widely studied and applied in cancer diagnosis, treatment selection and monitoring, especially for lung cancer. For localized bladder cancer, ctDNA in plasma and urine was proved to be detectable even at the early phase of disease and associated with disease recurrence or progression [6]. However, whether ctDNA in blood or urine would be an effective substitution for tissue for genetic testing in UC is still unknown.
In the present study, we analyzed the germline and somatic gene alterations in 112 Chinese UC patients.
The basic genomic alteration pro les were described and compared with corresponding Western data.
Meanwhile, we compared the concordance of genetic alterations between ctDNA in plasma and matched tumor tissue to comprehend the utility of liquid biopsy samples in UC patients.

Samples source and ethic data
In total, formalin-xed, para n-embedded (FFPE) tumor tissues and matched blood in EDTA (for germline tests) from 112 diagnosed UC patients (Supplemental Table 1) were collected with the signing of informed consent and agreements to publish the results for research purpose. All tumor tissues were pathologically reviewed to contain at least 20% tumor cells. 34 patients who cannot provide su cient or valid tumor tissue samples were collected blood for ctDNA testing instead. Among enrolled patients, 20 of them agreed to provide additional blood samples for comparison of the genomic differences between blood and matched tumor.

Dna Isolation
The FFPE tissues and peripheral blood mononuclear cells were collected to extract DNA using DNeasy Blood & Tissue Kit (Qiagen, Inc.) under the manufacturer's instructions. Cell-free DNA (cfDNA) was extracted from serum using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Inc.) following the protocol of the manufacturer. The puri ed gDNA and cfDNA were quanti ed using the Qubit 3.0 Fluorometer (Life Technologies, Inc.) and StepOnePlus System (Life Technologies, Inc.).

Target Next-generation Sequencing
For the matched germline and tumor samples, 100 ng of DNA was sheared with a Covaris E210 system (Covaris, Inc.) to get 200 bp fragments. We performed next generation sequencing of tumor and gDNA matched germline DNA using Accel-NGS 2S DNA Library Kit (Swift Biosciences, Inc.) for library preparation and xGen Lockdown Probes kit (IDT, Inc.). The custom xGen Lockdown probe was synthesized by IDT, Inc. for the exons and selected intronic regions of 92 genes (Supplemental Table 2). The prepared library was quanti ed using the Qubit 3.0 Fluorometer (Life Technologies, Inc.), and quality and fragment size were measured using Agilent 2100 Bioanalyzer (Agilent Technologies, Inc.). Samples underwent paired-end sequencing on an Illumina Nextseq CN500 platform (Illumina Inc) with a 150-bp read length. Mean coverage of 1260.5×, 3759.3 × and 223.6 × were achieved for tumor gDNA, blood cfDNA and peripheral blood mononuclear cells gDNA, respectively.

Data Processing
Raw sequencing data were aligned to the reference human genome (UCSC hg19) by Burrows-Wheeler Aligner. After removing duplicate and local realignment, we applied Genome Analysis Toolkit (GATK) for single nucleotide variation (SNV), insertion and deletion (inDel) calling. Next, after removing the germline alterations from matched blood samples, the somatic alterations were obtained. Variants were annotated using the ANNOVAR software tool. Copy number was analyzed using CNVkit. Genomic alterations data from The Cancer Genome Atlas database (TCGA) and Memorial-Sloan Kettering Cancer Center (MSKCC) was downloaded from cBioPortal (http://www.cbioportal.org).

Statistical analysis
Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) statistical package and Graphpad (Prizm 8). Differences were considered signi cant if p < 0.05.

Demographic and clinical data
The demographic and clinical data of patients enrolled in our cohort were summarized in Table 1. Of the enrolled patients, 72.32% were bladder cancer (81/112), and the rest were diagnosed as upper tract UC (27.68%, 31/112), including carcinoma of the renal pelvis and ureteral. The median age at diagnosed was 67 (range . The sex ratio between males and females was 2.61:1, which was close to the sex ratio (2.97:1) in Chinese UC patients reported by the national cancer center of China. The amounts of enrolled smoker and non-smoker were the same (56 versus 56).

Germline Alteration
Excluded of variants identi ed as benign or likely benign according to the American College of Medical Genetics and Genomics (ACMG) guideline, 75.89% (85/112) patients in our cohort harbored at least one germline alteration (Supplemental Table 3). After exclusion of known and likely benign variants, 244 germline variants were identi ed. However, only two patients (1.85%) had variants that could be identi ed as pathogenic or likely pathogenic, including a ERCC4-p. Lys481fs and a BRCA2p.Thr3030fs (Supplemental Table 3). The patient harbored deleterious BRCA2 germline variants was also concurrent with prostate cancer. None pathogenic or likely pathogenic germline variant associated with Lynch syndrome, including MLH1, MSH2, MSH6 and PMS2 genes, was identi ed in UTUC and UC patients of our cohort.

Differences of somatic gene alterations in UC patients between our cohort and Western cohort
To determine the potential difference of genomic features between Western and Chinese UC patients, we compared the alterations data of the selected 92 genes between our cohort and Western cohorts (UC data published by TCGA and UTUC data published by MSKCC). The prevalence of alterations in FGFR4, KDM5C, TERT, PDGFRB, FLT3, FLCN, MSH6, FLT1 were higher in UCB patients in our cohort, compared with TCGA ( Fig. 2A). As for UTUC, a higher mutated frequency of multiple genes, including TP53, LRP1B, KMT2D, FAT4, BRCA1, FGFR2 and BRIPI, were found in our cohort (Fig. 2B). On the contrary, the mutation frequency of FGFR3 gene in the MSKCC cohort was three times higher than that in our cohort (48.24% versus 16.13%, p < 0.001).
Genomic Alterations In Fgfr Genes 28.57% (32/112) patients had at least one somatic alteration in FGFR genes (including FGFR1, FGFR2, FGFR3 and FGFR4), but only 11 patients' alternations could be de ned as gain of function (Fig. 3A). Notably, six of these patients had multiple oncogenic alterations in FGFR genes. Same to previous studies' results, the most altered gene was FGFR3 (13.39%). Hotspot variants, including FGFR3p.Ser249Cys, p.Arg248Cys and p.Tyr373Cys were identi ed in six samples, while FGFR3-TACC3 fusion was only identi ed in two patients. Three novel variants (de ned as unreported in any other research or single nucleotide polymorphism database before and without ExAC frequency), including p.His349Asn, p.Val166Met and p.Thr755Lys of FGFR3 were identi ed, though the functions were still unknown (Fig. 3B). Nine patients (8.04%) were identi ed with FGFR2 alterations, of which only three alterations could be de ned oncogenic, including one p.Asn549Lys, one p.Lys659Met and one copy number gain (copy number 32.87). No FGFR2 fusion was identi ed. Four novel variants with unknown function, including p.Gly305Arg, p.Tyr207Phe, p.Met803Leu and p.Gln683Ter of FGFR2 gene were identi ed in this study. Only six patients carried FGFR1 gene alterations, and half of them were ampli cation. Eight patients (all were bladder cancer) carried nonsynonymous single nucleotide variants with unknown function in FGFR4 gene.
Genomic Alterations In Erbb Family 20 of our cases (17.86%) carried somatic alterations in the ERBB genes (EGFR, ERBB2 and ERBB3), and three of them had dual alterations (Fig. 3C). Five patients (4.46%) carried EGFR gene alteration and two of them had EGFR copy number gain. Notably, one patient had an activated EGFR exon 20 insertion (EGFR-p.Val769_Asp770insAspAsnPro) and copy number gain. We observed ERBB2 alterations in 12 patients (10.71%) s, and most of the oncogenic alterations located in furin-like cysteine rich region (amino acid 190-343) and protein tyrosine kinase domain (amino acid 721-975) (Fig. 3D). Two of the ve identi ed ERBB3 alterations were in the furin-like cysteine rich region (amino acid 182-332) and oncogenic.
Concordance of genetic alterations between ctDNA in serum and matched tumor tissue Twenty matched tumor tissue and blood samples were collected to comprehend the concordance and discordance of genomic alterations between different sample types in UC. Only three patients' matched samples were collected with an interval time over one month (7 months for P074, 18 months for P084 and 34 months for P110, respectively). In total, 99 and 91 somatic alterations were identi ed in tumor and serum ctDNA samples from 20 patients who underwent genomic testing for matched tumor tissue and serum ctDNA samples, respectively. Only one patient's ctDNA was negative for valid somatic alterations identi cation, while matched tumor samples were positive for alterations calling. Conversely, in Patient 20, ctDNA analysis revealed oncogenic mutations in PIK3CA, TP53, ARID1A and KDM6A with allele fractions all beyond 3%, for which there was no corresponding valid alteration in the matched tissue sample (Fig. 5A). There was no signi cant difference between tissue and blood in the median number of genomic alterations (4 versus 4). By comparison of blood and matched tumor tissue, the overall concordance for genomic alterations and altered genes identi ed in matched samples was 42.97% (0-100%) and 46.83% (0-100%), respectively ( Table 2). Among the genomic alterations, 48 of them were in concordance between ctDNA and tumor tissue. Fifteen actionable alterations were identi ed in tissue de ned according to the Oncobkb database, and 60% of them were shared in matched ctDNA (Fig. 5B).

Discussion
Compared with UC patients in Western countries, more Chinese UC patients were diagnosed in the advanced stage for multiple reasons [9]. Meanwhile, fewer drugs were developed and approved in China for advanced UC which left no choice but only chemotherapy for most advanced Chinese UC patients. Meanwhile, because of the aristolochic acid exposure history, many Chinese UC patients, especially UTUC, tend to have chronic kidney disease, which made them un t for platinum treatment [5]. Therefore, it's of great importance to understand the genomic features of Chinese UC patients for further therapy development and clinical trials. Remarkably, over 60% of UC patients in this study were identi ed to have at least one potentially actionable somatic alteration, which would afford more choice for further treatment.
To date, FGFR2 and FGFR3 genes were the most actionable genes in UC, especially for UTUC, as erda tinib (a FGFR inhibitor) was proved to have a 40% objective response rate in previously treated locally advanced and unresectable or metastatic UC with FGFR2/3 alterations [10]. We identi ed FGFRs gene alteration in 26.85% of UC patients, which was close to the corresponding ratio in the TCGA database [4]. Though may have an increased sensitivity to FGFR inhibitor, UC patients with FGFRs alterations may have a lower response rate to the ICIs therapy because of the suppressed in ltration of tumor immune cells in the tumor environment [11,12]. Meanwhile, alterations in ERBB pathway, especially in ERBB2, raised increasing interest in UC recently, as 12.4% and 11% UCB patients had ERBB2 overexpressing (as tested by HER2 immunohistochemistry (IHC) staining) and mutations, respectively [13]. Similarly, ERBB2 alteration was observed in 10.71% of cases, which was equal to the data in Western UC patients [4]. An interesting correlation between ERBB2 gene mutations and a higher response possibility to platinum-based neoadjuvant chemotherapy was found in 61 muscle-invasive UCB [14]. Meanwhile, anti-Her2 therapy by Trastuzumab combined with chemotherapy has been proved to have an outstanding response rate and median overall survival of 70% and 14.1 months, respectively, in 44 advanced UC patients with HER2 overexpression in a single arm phase II trial [15]. Other anti-Her2 therapies, such as tyrosine kinase inhibitors, antibodies and antibody-drug conjugate, had been widely developed in UC with abnormal ERBB2 [16].
As previous work reported, DDR alterations were associated with a higher response rate and clinical bene t for both immune checkpoint inhibitors (ICIs), platinum-based neoadjuvant therapy and rst-line chemotherapy in UC patients [17]. The previous study found 23.9% UCB patients had alterations in Homologous recombination repair gene pathway [18]. In our study, we found 47.32% and 14.29% UC patients had alterations and deleterious alterations in DDR pathway, respectively. The ratios were similar to the results analyzed by the panel involved the same genes [19]. Except for ICIs, tumors with deleterious DDR gene mutations, especially for BRCA1/2 gene, were also associated with better sensitivities with poly (ADP ribose) polymerase inhibitors (PARPi) [20]. Though it is still lack of evidence that proved the e ciency of PARPi in UC patients, PARPi combined with ICIs has been demonstrated to have cooperative effects in treating UC patients with HR mutations [21].
The prevalence of germline mutations of cancer susceptibility genes in patients with unselected UC was con icting. Although previous studies established the point that UC was a carcinoma with rarely cancer susceptibility genes alterations, the latest research found that 14% of 586 unselected UC patients carried pathogenic or likely pathogenic germline variants, and 11.26% (66/586) of them had deleterious variants in DNA repair pathway [22]. Until now, the prevalence of germline mutations of cancer susceptibility genes in Chinese UC patients remains unclear. In our cohort, we only identi ed 1.79% of patients with pathogenic or likely pathogenic germline variants, which may represent a low prevalence of cancer susceptibility genes in Chinese UC patients compared to the corresponding Caucasian patients. In this study, we found a novel likely-pathogenic ERCC4 gene germline variants in a UTUC patient. Previously, ERCC4 gene mutations were identi ed in fanconi anemia, skin-photosensitive, nucleotide excision repair (NER)-de cient disorders xeroderma pigmentosum and XFE progeroid syndrome [23]. 2.03% of UC patients had alterations in ERCC4 gene, while no deleterious germline ERCC4 carrier of UC had been identi ed [24]. Meanwhile, Previous research found that 7%-8.3% unselected UTUC patients had lynch syndrome-related features, such as de cient mismatch repair (dMMR), microsatellite instability high (MSI-H) or deleterious alterations of MMR genes, while only 2.1% occurred in UC patients [22,25]. In the present research, no germline variant in mismatch repair pathway was found, and the possible reasons may lie in the differences between Chinese and Western UC patients, limited UTUC patients enrolled and technology limitations (such as lack of detecting in MLH1 promoter hypermethylation and large rearrangements) [26].
Moreover, though previous studies suggested that ctDNA could serve as an effective biomarker in cancer diagnosis, risk strati cation, therapy monitoring and early relapse in UC, the controversy about liquid biopsy remained [27]. Several prior studies published controversial results about the concordance levels between tissue and blood in other genitourinary carcinomas [28]. We found a concordance of 42.97% (0-100%) of the genetic alterations between 20 ctDNA and matched tissue samples, which was not reported before. Though factors, including tumor stage, metastasis status treatments, the interval time of sample collection and analyzed panel of genes would affect the concordance of genomic alterations between ctDNA and tissue, lipid biopsy is still an expanding way to overcome the intratumoral and intertumoral heterogeneity {Meeks, 2020 #632}. Our result support that genetic testing on tissue is still the gold standard for UC patients, meanwhile liquid biopsy would serve as a complementary source to provide additional and meaningful results to comprehend the genomic feature more precisely.

Conclusion
In this study, we identi ed a unique genomic feature of germline and somatic alterations in Chinese UC patients by next-generation sequencing. In total, 62.50% of 112 Chinese UC patients had at least one actionable genomic alteration, which would bene t from matched or related target therapies. Furthermore, our study found a good genomic concordance between ctDNA and matched tumor, suggesting the potential application of liquid biopsy in UC.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementalTable2Genelist.xlsx SupplementalTable1Patientsclinicalcharacteristics.xlsx SupplementalTable3Germlinevariants.xlsx