Catalog of prognostic tissue-based biomarkers in patients treated with neoadjuvant systemic therapy for urothelial carcinoma of the bladder: a systematic review

PURPOSE: The present systematic review aimed to identify prognostic values of tissue-based biomarkers in patients treated with neoadjuvant systemic therapy (NAST), including chemotherapy (NAC) and checkpoint inhibitors (NAI) for urothelial carcinoma of the bladder (UCB). MATERIAL AND METHODS: The PubMed, Web of Science, and Scopus databases were searched in August 2020 according to the PRISMA statement. Studies were deemed eligible if they compared oncologic or pathologic outcomes in patients treated with NAST for UCB with and without detected pretreatment tissue-based biomarkers. RESULTS: Overall, 44 studies met our eligibility criteria. Twenty-three studies used immunohistochemistry (IHC), 19 − gene expression analysis, three - quantitative polymerase chain reaction (QT PCR), and two − next-generation sequencing (NGS). According to the currently available literature, predictive IHC-assessed biomarkers, such as receptor tyrosine kinases and DNA repair pathway alterations, do not seem to convincingly improve our prediction of pathologic response and oncologic outcomes after NAC. Luminal and basal tumor subtypes based on gene expression analysis showed better NAC response, while claudin-low and luminal-inﬁltrated tumor subtypes did not. In terms of NAI, PD-L1 seems to maintain value as a predictive biomarker, while the utility of both tumor mutational burden and molecular subtypes remains controversial. Speciﬁc genomic alterations in DNA repair genes have been shown to provide signiﬁcant predictive value in patient treated with NAC. QT PCR quantiﬁcation of speciﬁc genes selected through microarray analysis seems to classify cases regarding their NAC response. CONCLUSION: We believe that the present systematic review may offer a robust framework that will enable the testing and validation of predictive biomarkers in future prospective clinical trials. NGS has expanded the discovery of molecular markers that are reﬂective of the mechanisms of the NAST response. (cid:1) 2021 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)


Introduction
Urothelial carcinoma of the bladder (UCB) is one of the most frequently diagnosed and harmful cancers worldwide [1]. Neoadjuvant cisplatin based combination chemotherapy (NAC) prior to radical cystectomy is the preferred first treatment in cisplatin eligible patients with muscle-invasive UCB [2,3]. However, multiple reasons impeded the widespread uptake of NAC such as the fear of unnecessary chemotoxicity, its perceived relatively modest survival benefit, and/or the fear of a delay to radical treatment [4,5]. Moreover, UCB is a highly heterogeneous disease with varied response rates when therapies are given in unselected patient populations. Identification of the patients who are unlikely to respond to NAC could allow better selection of patients to immediate radical cystectomy or allocation of different systemic therapies such as checkpoint inhibitors (CPI).
Modern medical decisions can be tailored to the individual patient based on predicted response or risk of disease. Understanding the molecular basis of disease has ushered in a new age of precision medicine. Molecular markers are promising tools that may give insight into which UCB patients will or will not benefit from neoadjuvant systemic therapy (NAST) and which have the potential to overcome the limitations of conventionally used prognostic risk factors. In addition, a biomarker-based strategy to identify patients who should undergo NAC is more cost-effective compared to the current unselected use of NAC or radical cystectomy alone [6]. Numerous publications provided data on potential molecular markers associated with NAC response in UCB patients; however, none is yet validated or widely used in the clinical practice [7−9].
In this systematic review we aimed to summarize the available evidence as well as to determine whether pretreatment tissue-based biomarkers may help predict oncologic and pathologic outcomes in patients treated with NAST for UCB. This review is a benchmark for future developments.

Literature search
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement [10]. This study's protocol was registered a priori on the International Prospective Register of Systematic Reviews (PROSPERO; Registration ID CRD42020208417).
The PubMed, Web of Science, and Scopus databases were searched in August 2020 to identify studies reporting on the prognostic value of tissue-based biomarkers in patients treated with NAST for UCB. A comprehensive systematic literature search was independently performed by two authors. The keywords used in our search strategy included: (NAC OR neoadjuvant) AND (bladder OR urothelial) AND (cancer OR tumor OR malignancy OR carcinoma) AND (biomarker). In addition, we manually searched for potentially relevant trials from the references of selected studies. The primary outcome of interest was both oncologic and pathologic outcomes in patients treated with NAST for UCB.
After removing duplicates, two independent reviewers screened the titles and abstracts. Any citation which either reviewer thought should be included or unclear for inclusion was identified for full text screening. Subsequently, reviewers reviewed full texts of eligible articles for final inclusion and data extraction. In cases of disagreement, the authors consulted with the co-authors, and final decisions were reached by consensus.

Inclusion and exclusion criteria
We included all non-randomized observational studies that reported on the prognostic value of tissue-based biomarkers in UCB.
The PICO in this study was the following: patients treated for UCB with detected pretreatment tissue-based biomarkers. Intervention included NAST for UCB. Control group included those patients without pretreatment tissuebased biomarkers. The outcome included any measure of association between oncologic and pathologic outcomes and the candidate biomarker, the diagnostic performance of the biomarker.
We excluded reviews, letters, editorials, animal studies, study protocols, case reports, meeting abstracts, replies from authors, brief correspondence, and articles not published in English. Furthermore, we excluded the studies that did not provide data regarding the oncologic or pathologic outcomes. References of all papers included were scanned for additional studies of interest.

Data extraction
Data extracted from each study were independently extracted by two independent reviewers. Extracted data included the following: first author's name, publication year, study design, demographics characteristics including age range, sample size, pathological T stage, follow-up duration, NAC regime, definition of response, type of biomarkers, methods of biomarkers detection, % of patients with high expression, and Main results. Subsequently, the hazard ratios (HR) and 95% confidence intervals (CI) of tissue-based biomarkers associated with each outcome were retrieved.

Evidence synthesis
The literature search identified 624 unique references. Among them, 233 records were removed due to duplication, and 261 articles were excluded due to unrelated outcomes during the screening process ( Figure 1). Of the 130 full-text articles assessed for eligibility, 86 were excluded based on the selection criteria.
Overall, 44 studies were finally included in the present systematic review. Characteristics of the studies are shown in Table 1. Fifteen of the included studies had a prospective study design, and twenty-nine were retrospective.

Immunohistochemistry (IHC)
Twenty-three studies provided data on the pretreatment biomarkers detected at IHC.

DNA repair pathway alterations
A study assessing DNA repair pathway alterations found that a strong expression of ERCC1 was associated with pathological response in patients treated with neoadjuvant gemcitabine and cisplatin (GC) (p=0.01) [15]. Choueiri et al. [16] reported a pathological response (<pT1) rate of 43% in ERCC1-positive and 60% in ERCC1-negative UCB patients treated with dose dense MVAC.

Receptor tyrosine kinases
Yang et al. [15] reported that receptor tyrosine kinases (ERBB2, FGFR3, and PIK3CA) were more commonly altered in the responders (p<0.01) compared to the non-responders; FGFR3 mutations were significantly enriched in patients with a response to GC based regimen (p=0.01). In contrast, Kiss et al. [17] failed to report on the association between ERBB2 alterations and both pathological response (<ypT2N0) or OS.

Biomarkers for combination of NAC and radiotherapy
Three studies reported IHC biomarkers in patients treated with combination of NAC and radiotherapy [18] − [20]. Positive p53 and p21 were independently associated with decreased disease free survival (DFS) in a retrospective study of 82 patients (p<0.005 and p<0.009, respectively); additionally, p53 overexpression was associated  with poor OS (p<0.03) [18]. Alteration of the combination of p53 and p21 was a strong and unfavorable prognostic factor for both DFS (p<0.003) and OS (p<0.02). Hemdan et al. [19] demonstrated that patients with negative emmprin (extracellular matrix metalloproteinase inducer) expression had significantly greater OS in 125 UCB patients treated with radiotherapy and NAC (71% vs. 38%, p <0.001); cancer specific survival (CSS) in patients with negative and positive emmprin expression was 76% and 56%, respectively (p=0.027). Turker et al. [20] reported that patients exhibiting Bcl-2 negative expression had a significantly increased OS (p=0.009). In summary, pretreatment p53, p21, emmprin, and Bcl-2 have been suggested to exhibit predictive value in UCB patients treated with NAC and radiotherapy. However, further studies are needed to improve our understanding of the radiotherapy impact on inflammation status, which could affect biomarker expression.
According to the currently available literature, IHC biomarkers, including receptor tyrosine kinases and DNA repair pathway alterations, do not seem to clearly improve our prediction of pathological response or oncologic outcomes in UCB patients treated with NAC.

Gene expression and genomic DNA analyses
Nineteen studies provided data on the pretreatment biomarkers detected using gene expression analysis.
Over the last decade, molecular subtyping has led to distinct or partially overlapping molecular classifications of UCB. The arising molecular subtypes based on these classifications have been shown to be clinically useful in predicting the likelihood of therapy response. Whole transcriptome analysis suggests that luminal and basal tumors, compared to claudin-low or luminal-infiltrated tumors, might have the best response to platinum-based NAC (p<0.05) [21,22]. Supporting this data, Choi et al. [23] reported response rate of 0% in p53-like, 40% -basal-like, and 67% -luminal-like subtypes (p=0.018). Efstathiou et al. [24] detected worse DSS and OS among patients with claudin-low tumors at transcriptome-wide gene expression profile analysis (p=0.01 and p=0.068, respectively). Taking together, luminal and basal tumor subtypes showed better NAC response, while claudin-low and luminal-infiltrated tumor subtypes did not.
Surprisingly, during comprehensive genomic profiling, molecular subtypes were not significantly associated with response (ypT0N0) in both studies assessing NAC and Pembrolizumab (all p>0.2) [25]. Notably, immune signatures explored in this study had a significant association with the pathologic response in the PURE-01 cohort (all p<0.02), but not in the NAC cohort (p>0.7) [25]. Among other studies on predictive biomarkers for neoadjuvant immunotherapy, Necchi et al. [26] reported an association of tumor mutational burden (TMB) and PD-L1 combined positive score with both the pT0 and the pT1 response to Pembrolizumab (all p<0.03). In contrast, Bandini et al. [27] found that TMB was not associated with response (pT0N0) to Pembrolizumab on multivariable analysis (OR 1.04, 0.98-1.10, p=0.09). These results were supported by Powles et al. [28] in a study of 95 patients treated with neoadjuvant Atezolizumab. Summing up, in terms of neoadjuvant immune-checkpoint inhibitors (CPI), PD-L1 seems to maintain value as a predictive biomarker, while the utility of TMB and molecular subtypes is still controversial.
Among other predictive biomarkers detected with gene expression analysis, Plimack et al. analyzed molecular alterations in baseline tumor samples and did not find a correlation between p53 deleterious mutations and response to NAC [29]. Defects in DNA repair genes (ATM, RB1, and FANCC) were shown to predict pathological response in both MVAC (p<0.001) and dose dense GC (p = 0.033) cohorts and at the same time with better OS after MVAC (p = 0.007) [30]. Another DNA repair pathway alteration (ERCC2) was also significantly mutated in cisplatin responders compared to non-responders (p<0.01) [31]. In contrast, genetic alterations in genes associated with cell cycle checkpoints and regulators (E2F3, JUN, FBXW7) suggested potential resistance [32].
Summing up, according to the currently available literature, alterations in DNA repair genes seem useful to predict pathological response and even oncologic outcomes in UCB patients treated with NAC. However, these data should be supported by future large-scale trials.

Polymerase chain reaction (PCR)
Three studies provided data on the pretreatment biomarkers detected at quantitative PCR [33]− [35].
In order to investigate the predictive role of the breast cancer susceptibility gene 1 (BRCA1) mRNA expression in UCB, tumor samples of 57 patients treated with GC or CMV (cisplatin, methotrexate, vinblastine) for UCB were retrospectively analyzed using quantitative PCR [33]. 66% of patients with low/intermediate BRCA1 levels attained a pathological response (pT0-1) compared to 22% of those with high BRCA1 levels. Furthermore, median survival was longer in patients with low BRCA1 expression (168 and 34 months, respectively, p=0.002). Thus, BRCA1 expression could be a useful tool for selecting UCB patients who are likely to benefit from cisplatin-based NAC. The authors suggested that taxane-based therapy for patients with high BRCA1 expression could be explored in further studies.
Among studies on other tissue-based biomarkers detected with PCR, Kato et al. [34] identified 12 candidates genes tested in tissue microarrays derived from baseline biopsies of 37 patients treated with NAC. Among these genes, IPO-7 and SLC22A18 were upregulated in non-responders. Vinall et al. [35] found that higher let-7c expression had higher odds of responding (OR 2.493, 95% CI 1.121-5.546, p=0.023), and let-7c levels allowed predicting response (pT0) with an accuracy of 72%. Nevertheless, larger scale studies are certainly warranted to confirm and validate these results.
In general, quantitative PCR results for the expression of genes selected through microarray analysis might correctly classify cases with regard to their NAC response.

Next-generation sequencing (NGS)
Two studies provided data on the pretreatment biomarkers detected at NGS [36,37].
In a study of Groenendijk et al. [36], ERBB2 was strongly associated with NAC response, defined as ypT0N0 (p=0.006), whereas ERCC2 mutations were not. Miron et al. [37] found that mutations in ATM, RB1, or FANCC were significantly associated with improved OS (p=0.0043) and DSS (p=0.0015) in 58 patients treated with NAC (GC or MVAC). The authors hypothesized that, based on understanding the function of ATM, RB1, and FANCC and their involvement in DNA damage repair, mutations in these genes sensitize tumors to cisplatin because of a baseline deficiency in DNA repair.

Discussion
This review on the impact of using pretreatment tissuebased biomarkers to select patients who are most likely to benefit from NAST generated several important findings.
First of all, there is no clear benefit of using predictive biomarkers, including receptor tyrosine kinases and DNA repair pathway alterations, detected at IHC to predict pathologic response or oncologic outcomes in UCB patients treated with NAC. The controversial results can be explained by the small sample size as well as the retrospective nature of most included studies, leading to heterogeneity between NAST cohorts, differences between NAC settings, and definitions such as that of pathologic response as well as non-standardized sample collections and arbitrary cut-offs during assay analysis. Moreover, we believe that for the initial development of a putative marker model as well as markers with combinations, it is essential to reflect the molecular understanding of the tumor and its microenvironment.
We found out that specific genomic alterations in DNA repair genes (e.g., ATM, RB1, FANCC, and ERCC2) provide predictive value for predicting pathologic response and oncologic outcomes after NAC. Quantitative PCR results for the expression of genes selected through microarray analysis (e.g., BRCA1) could correctly classify cases with regard to their NAC response. However, it should be stressed that the utility of genetic profiling has historically been limited to small gene panels and costly molecular diagnostics. Hence, biomarkers detected at IHC can still be a simple and less expensive alternative. To facilitate inclusion into routine urological practice, precise identification of tissue-based biomarkers with accurate detection technology seems to be of necessity. The continuous improvement in high throughput technologies, the development of novel analytical tools based on artificial intelligence need for biomarker-driven preclinical and clinical trials. Nowadays, NGS is becoming a complementary diagnostic tool, guiding the decision-making progress with the goal of facilitating precision medicine. We believe that with the incorporation of NGS, physicians will have the ability to obtain a more comprehensive understanding of the molecular alterations driving an individual urothelial cancer [38].
In terms of predicting the likelihood of responding to neoadjuvant CPI, TURBT PD-L1 seems to have value as an accurate but not ideal biomarker [39]. Indeed, a higher pathologic response rate was shown in patients with PD-L1 positive tumors compared to those with PD-L1 negative tumors; while the utility of TMB or molecular subtypes in patients treated with neoadjuvant CPI is still unclear, at best. Moreover, it was recently shown that indicate molecular subtypes may not be useful due to tumor heterogeneity and various models of changes in molecular profiles before or during progression [40,41]. Understanding the stability of molecular subtypes over time and the subtype heterogeneity within tumors and patients remains challenging. Future areas certainly include conceptual molecular pathways (e.g., FGFR3 pathway) that would allow for targeted therapy approaches. New clinical trials that use molecularly guided therapy selection will determine the clinical efficacy of the integration of genomics and other molecular predictive biomarkers to guide daily therapeutic decision-making.
Our systematic review is not free from limitations. First, the inconsistencies in evaluation of the tissue-based biomarkers among the enrolled trials could lead to potential confounding and bias. The second limitation is the retrospective and heterogeneous nature of most included studies which also suffered from single-center designs. Third, the small cohort size of most of the included studies may have limited their power to detect a statistically and/or clinically significant associations. Therefore, well-designed comparative trials with larger cohorts are required to validate some of the most promising findings inherent to the present systematic review.

Conclusions
Pretreatment tissue-based biomarkers still hold promise in selecting the ideal UCB patient who is most likely to benefit from NAST. However, due to the lack of prospective, well-designed, large scale data, no molecular biomarkers could be recommended for the routine use. The present systematic review offers a robust framework to enable the testing and validation of predictive biomarkers in future prospective clinical trials.

Ethical standards
Not applicable.

Declaration of Competing Interest
None