Background
Urothelial bladder cancer (UBC) is the 10
th most prevalent cancer worldwide and the 4
th among men [
1]. Long-term surveillance of patients at risk of recurrence or progression is costly for healthcare providers and burdensome for patients, necessitating repeated cystoscopies and imaging. However, existing classifications, which depend solely on clinicopathological variables [
2,
3], are insufficient for the precise prediction of treatment responses or disease trajectory. Consequently, there is an urgent and unmet need to identify accurate local and systemic predictive and/or prognostic biomarkers. Proximity Extension Assays (PEAs) are sensitive tools for biomarker discovery, but are yet to be explored in-depth in liquid biopsies from UBC patients.
UBC is associated with carcinogen exposure, often resulting in tumours characterised by high mutational load [
4,
5]; such features may lead to immunogenicity and an immunotherapy-responsive disease by increasing the likelihood of HLA-presentable tumour neoantigens that are able to initiate an immune response [
6‐
8]. High numbers of macrophages [
9‐
11], neutrophils [
12], and myeloid-derived suppressor cells [
13,
14] in UBC are further associated with poor survival and limited treatment responses.
This study aimed to investigate whether a sensitive protein detection method such as PEA technology could aid in prognostic biomarker discovery in liquid biopsies by exploring the systemic and local proteomic contexture of urothelial cancer in relation to immune- and tumour-related processes. To achieve this, we utilised an immuno-oncology-focused multiplexed PEA panel to analyse plasma and urine samples from a Swedish UBC patient cohort. Findings were validated using PEA data from an independent UK UBC cohort [
15], and publicly-available RNA-sequencing data were interrogated to identify gene expression changes and the cellular sources of the prognostic biomarkers.
Discussion
The majority of UBC patients require intensive, long-term surveillance which increases morbidity and healthcare-related costs. With the current standard-of-care treatments for NMIBC, 5-year recurrence-free survival ranges from 23 to 43%, while the 5-year progression-free survival ranges from 54 to 93% [
27]. However, based solely on clinical and pathological characteristics, patient tumour recurrence and survival outcomes are difficult to predict. Although NMIBC and MIBC are molecularly complex and heterogeneous [
28], recent developments in transcriptomic classification offer promise for better treatment stratification and personalised therapy [
29,
30]. These approaches rely upon laboratory-based deep sequencing of tumour tissue, necessitating invasive biopsies, specialist public or commercial genomics laboratories, and a turnaround time of 5–10 days. Accurate protein-based prognostic biomarkers assessed in liquid biopsies could overcome some of these limitations and easily permit repeated assessments during the patient’s disease course.
In this study, we used an assay which combines high technical sensitivity and specificity with a high-throughput multiplex format, permitting the assessment of 92 protein markers of important immune- and tumour-related processes; the clinical samples were collected at the first diagnosis of UBC before therapeutic interventions. Nevertheless, stage-specific patient classification or clustering based on protein levels was not achieved – a result of the heterogeneity in proteomic patient profiles which did not follow stage- or grade-related patterns.
Released extracellular vesicles or shed immune-related receptors/ligands such as MICA/B, which was one of the 12 identified immune-related receptor/ligand proteins positively correlating between plasma and urine, may have a systemic impact on the immune response with an increased risk of developing immune evasion later in the disease trajectory in several cancer types [
31].
Interestingly, we found that receptor-bound proteins were elevated in our liquid based analysis. The origin of elevated soluble CD40 in the urine of NMIBC compared to MIBC patients is likely to be antigen-presenting cells and its decreased levels at MIBC might be a biomarker of immune suppression in the local microenvironment, there are discrepancies as to whether increased expression of CD40 has beneficial or adverse effects on cancer prognosis and therapy responses [
32,
33]. On the other hand, soluble CD27 is reduced in the serum of cancer patients compared to healthy individuals, while in the serum of cancer patients post immunotherapy, soluble CD27 correlates with improved survival probability [
34].
MMP7 and CCL23, involved in tissue degradation and T cell recruitment signalling, were found to be significantly elevated in the plasma of MIBC patients compared to NMIBC patients. Significantly higher MMP7 levels were also observed at MIBC compared to NMIBC in the UK cohort [
15], in gene expression data of an independent patient cohort [
20], and in other studies [
35]. In our study, analysis of tumour tissue gene expression indicated that
MMP7 was expressed at higher levels in tumours of MIBC patients compared to NMIBC.
Despite stage-specific differences, none of the aforementioned proteins reached a high UBC stage classification accuracy (AUC > 0.85). High levels of MMP7 have been reported to be associated with shorter overall survival in UBC [
15,
36,
37]. However, neither MMP7 nor CCL23, CD27 or CD40 were significantly associated with overall survival when included as continuous variables in multivariable Cox-regression analyses (data not shown) in our dataset.
Instead, analysis of the data by application of an RSF model ranked increased plasma levels of MMP12 as the main driver for overall survival in the cohort, and as an independent prognostic biomarker associated with decreased overall survival of UBC patients following multivariable analysis comprising clinically relevant confounders, such as stage and grade. Due to the high inter-patient heterogeneity and collinearity between markers, a non-parametric and non-linear method, such as random forests, offers an alternative to classical approaches that are affected by many noise variables and high-order interactions between investigated variables [
26]. Despite the differences in liquid biopsies (plasma and serum), sampling procedures, and the use of different multiplex panels, the association between high systemic MMP12 levels and poor survival were recapitulated in the UK cohort.
To evaluate if MMP7 and MMP12 can be identified as biomarkers when using clinical biopsies, we investigated their gene expression in 306 UBC patients [
20]. Here, we found that MIBC had higher
MMP7 mRNA levels overall. Interestingly, compared to MMP12 protein levels,
MMP12 mRNA expression was significantly higher at the MIBC stage. However, when studying the impact of
MMP12 on survival in tissue biopsies analysed by transcriptomics, we could not corroborate the findings of the liquid biopsy data. This could be due to the bias introduced in the selection of the malignant cells retrieved for pathological scoring as well as the difference between transcriptomic versus proteomic analyses or the high number of NMIBC patients compared to late stage patients in the transcriptomics cohort. A biopsy for clinical diagnosis also focuses on sampling the most malignant tumor area, rather than the complete tumour microenvironment.
To explore the source of MMP7 and MMP12 in UBC, we analysed UBC patient tumour samples that underwent single-cell sequencing and identified epithelial cells as the source of
MMP7 and macrophages as the source of
MMP12 mRNA transcripts. Previous studies have shown that increased numbers of tumour-infiltrating macrophages correlate with advanced tumour stage and Bacillus Calmette–Guérin (BCG) treatment failure in UBC [
38,
39]. The importance of sex in UBC in the context of immunomodulatory therapies is still not fully explored [
40], additionally, macrophages are usually expressed at a higher number and in a more active state in females compared to males [
41], thus we included that variable in the single-cell RNA analysis for
MMP12. However, we did not see any sex dependent difference in that aspect.
The family of MMPs is largely orthologous between humans and mice, sharing structure and function [
42]. Utilising single-cell RNA-sequencing data of an autochthonous murine model of UBC,
Mmp12 mRNA expression was attributed to tumour-infiltrating macrophages. As observed in the human single-cell transcriptomic dataset, expression was also elevated in MIBC compared to NMIBC in male and female mice. However,
Mmp7 mRNA transcripts in the murine data set were only detected in a minor subset of urothelial cells and its expression was inadequate to warrant further inspection.
MMPs degrade the tissue matrix by extracellular proteolysis, thus facilitating tumour cell migration, invasion and metastasis. Moreover, MMPs are involved in several physiological and tumour-supporting cellular processes, including loss of cell adhesion, tumour angiogenesis, cell proliferation, epithelial-to-mesenchymal transition and apoptosis [
43]. They have been extensively analysed in UBC, which has revealed potential roles for certain MMPs as diagnostic markers and prognostic factors at different stages of the disease course [
44]. MMP12 is linked to reduced overall survival rates in numerous cancers [
45‐
47]; in UBC, certain genetic polymorphisms of the gene have been shown to increase invasiveness [
48], while higher mRNA expression of
MMP12 in tumour tissue has been reported to correlate with higher tumour grade [
49].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.