Follow-Up Biomarkers in the Evolution of Prostate Cancer, Levels of S100A4 as a Detector in Plasma

The management and screening of prostate cancer (PC) is still the main problem in clinical practice. In this study, we investigated the role of aggressiveness genetic markers for PC stratification. We analyzed 201 plasma samples from PC patients and controls by digital PCR. For selection and validation, 26 formalin-fixed paraffin-embedded tissues, 12 fresh tissues, and 24 plasma samples were characterized by RNA-Seq, immunochemistry, immunofluorescence, Western blot, and extracellular-vesicles analyses. We identified three novel non-invasive biomarkers; all with an increased expression pattern in patients (PCA3: p = 0.002, S100A4: p ≤ 0.0001 and MRC2: p = 0.005). S100A4 presents the most informative AUC (area under the curve) (0.735). Combination of S100A4, MRC2, and PCA3 increases the discriminatory power between patients and controls and between different more and less aggressive stages (AUC = 0.761, p ≤ 0.0001). However, although a sensitivity of 97.47% in PCA3 and a specificity of 90.32% in S100A4 was reached, the detection signal level could be variable in some analyses owing to tumor heterogeneity. This is the first time that the role of S100A4 and MRC2 has been described in PC aggressiveness. Moreover, the combination of S100A4, MRC2, and PCA3 has never been described as a non-invasive biomarker for PC screening and aggressiveness.


Introduction
Prostate cancer (PC) accounts for 43% of all cancer cases in men, which amounts to more than one in five in new diagnoses. However, survival rates (98%) are the highest compared with other tumors [1]. There is controversial data about the different rates among countries suggesting that prostate-specific antigen (PSA) tests could be one of these reasons [2]. PSA shows some disadvantages; a randomized controlled trial in the USA demonstrated PSA-testing decreasing PC deaths [3]. However, PC is detected on repeated biopsy in 10-35% of patients with a negative first biopsy. Widespread PSA testing also leads to the diagnosis of clinically insignificant tumors, resulting in potential overdiagnosis and overtreatment [4]. 2

of 10
There is a great clinical necessity for an accurate detection of PC, not only in screening stages reducing unnecessary biopsies, but also in PC management and monitoring. Therefore, new PC biomarkers have been described in recent years. Emerging biomarkers include those applying serum, urinary, or tissue samples as a test substrate [5], such as PCA3 and TMPRSS2-ERG gene fusions, which are specific biomarkers that are measurable in urine. They are usually found overexpressed in PC cells in comparison to normal or benign prostate cells [6,7]. PCA3 has been suggested in many studies as one of the most promising non-invasive methods with acceptable sensitivity and specificity in PC diagnosis, distinguishing between patients and healthy individuals [8]. It has been indicated as one of the candidates to be established as a non-invasive biomarker in diagnosis and prediction of clinical PC outcomes [9].
Other markers, such as MRC2 (also known as uPARAP, Endo180, or CD280), are involved in extracellular matrix remodeling by mediating collagen degradation, and have many roles in different tumors or diseases. Data are reporting that high expression of MRC2 produces tumor growth and is related to metastasis; thus, it usually associated with worse prognosis in several cancers [10]. Similar happens with S100A4, whose expression has been related as the most significant predictor of patient survival and a possible effective disease progression marker. Accumulating research evidence indicates overexpression patterns correlated with patient outcomes in gastric, colorectal, and renal cells, among others. Moreover, it has also been associated with patient survival in other tumor types, such as ovarian, pancreatic, bladder, and hepatocellular, among others [11]. Recently, the role of S100A4 has been suggested to be a key element in the regulation of bone metastasis [12]. Moreover, Shiqin L. et al. indicated that CASP8 and S100A4 are protein candidate biomarkers for high-risk PC, although this study ultimately only found significant values in CASP8 [13]. There are currently no conclusive data on both markers in PC, with this article being the first case in which a strong effect of MRC2 and S100A4 on PC aggressiveness is reported.
Recent massive strategies provide large amounts of data of novel biomarkers, such as PTGFR, NREP, SCARNA22, DOCK9, FLVCR2, IK2F3, USP13, and CLASP1, as potential biomarkers to predict PC progression [14], or YWHAZ in PC aggressiveness [15]. Moreover, machine learning models also reported new biomarkers related to PC development and progression such as PIAS3 and UBE2V2 [16]. However, none of them are currently used in clinical routine.
One of the main aims of current medicine is to avoid invasive methods; that is why a liquid biopsy is increasing as one of the most universal methodologies for monitoring and stratifying tumors such as PC. Liquid biopsy is one of the most efficient non-invasive modalities which can be performed and monitored in real-time, useful for early tumor diagnosis, therapeutic guidance, and recurrence monitoring [17]. Many molecules can be analyzed in liquid biopsy like circulating tumor cells, microvesicles, circulating cellfree DNA, circulating cell-free RNA, or miRNAs [18]. Friedemann et cols. indicated that RASSF1A and GSTP1 methylation patterns in circulating cell-free DNA from serum offer superior sensitivity of epigenetic biomarker analyses in the early detection of PC metastases [19]. Furthermore, genomic alterations of RB1, TP53, MYC, and DNA repair pathways are detected in liquid biopsy and associated with poor clinical outcomes in PC patients [20]. However, RNA levels are also a good strategy for non-invasive monitoring, enabling the determination of prognostic individualized therapy for aggressive or progressive PC [21]. Furthermore, there are commercial tests like AdnaTest, which allowed the analysis of the androgen receptor mRNA, which helps predict patients' survival [22].
Here, we reinforce the role of PCA3 as a non-invasive biomarker for PC and indicate for the first time the role of two other genes, MRC2 and S100A4, as good aggressiveness biomarkers in PC. Moreover, the combination of these three genes improves their accuracy as liquid biopsy biomarkers for PC aggressiveness. Tumor biomarkers play important roles in tumor growth, invasion, and metastasis, and genomic ones are crucial, being one of the most informative non-invasive strategies. Present data could reinforce the role of genetic markers in liquid biopsy to improve the current aggressiveness classification. The strategy of using genetic non-invasive biomarkers for PC management could improve precision medicine among all these patients.

Results
A summary representation of the present methodological study is illustrated in Figure 1. As can be seen, the present article was carried out in two relevant stages. biomarkers in PC. Moreover, the combination of these three genes improves their accuracy as liquid biopsy biomarkers for PC aggressiveness. Tumor biomarkers play important roles in tumor growth, invasion, and metastasis, and genomic ones are crucial, being one of the most informative non-invasive strategies. Present data could reinforce the role of genetic markers in liquid biopsy to improve the current aggressiveness classification. The strategy of using genetic non-invasive biomarkers for PC management could improve precision medicine among all these patients.

Results
A summary representation of the present methodological study is illustrated in Figure 1. As can be seen, the present article was carried out in two relevant stages. (C) microvesicles selection and dyeing in MRC2 and S100A4 for visualizing in cytometry; (D) immunofluorescence in cell lines and tissues for MRC2 and S100A4; (E) immunohistochemistry in PC tissue (MRC2 and S100A4). Previously, markers were selected by NGS analysis (2) by using filters obtained from different resources.

RNA Analysis and Gene Prioritization
RNA-Seq analysis provided about 500 genes. Those genes with higher expression patterns compared to healthy tissues (n = 6) were selected. At least 250 genes were chosen in each patient, always with an adjusted p-value below 0.05 and log2FoldChange out of range −0.10/0.10. (C) microvesicles selection and dyeing in MRC2 and S100A4 for visualizing in cytometry; (D) immunofluorescence in cell lines and tissues for MRC2 and S100A4; (E) immunohistochemistry in PC tissue (MRC2 and S100A4). Previously, markers were selected by NGS analysis (2) by using filters obtained from different resources.

RNA Analysis and Gene Prioritization
RNA-Seq analysis provided about 500 genes. Those genes with higher expression patterns compared to healthy tissues (n = 6) were selected. At least 250 genes were chosen in each patient, always with an adjusted p-value below 0.05 and log2FoldChange out of range −0.10/0.10.
Afterward, only those present in at least two plasma samples of PC patients were picked, reducing the total to 207 genes. Subsequently, using the GeneAnalytics web tool from GeneCards, genes were categorized by disease, and those included in cancer and PC categories were selected, reducing the list to 27 genes.
Finally, using related published articles, those genes previously described in plasma and cancer were selected for analysis (details in Supplementary Table S2). According to function, pathway, and previous studies from the gene set, we solely focused the analysis on S100A4, MRC2, and PCA3.

Differential Gene Expression in CP Patients Versus Controls
A total of 178 plasma samples (n = 89 patients and n = 89 controls) were analyzed for gene expression by dPCR. All three candidate biomarker genes (PCA3: p = 0.002, S100A4: p ≤ 0.0001, and MRC2: p = 0.005) demonstrated statistically significant differences in expression between PC patients and healthy controls in plasma samples (see details in Supplementary Figure S1).
Among patients, we have also performed comparisons between ISUP grade (abbreviated ISUP hereinafter) (1-2 vs. 3-5), PSA (<20 vs. >20 ng/mL), and the presence of metastasis. According to these groups, we have found that S100A4 is the best biomarker with statistically significant data when comparing ISUP (p = 0.044), PSA (p = 0.017) and metastasis (p = 0.025), whereas PCA3 is only statistically significant in PSA (p = 0.008) and MRC2 in none of the previous comparisons. Moreover, we have also developed the combination of all these genetic markers with clinical parameters, and those combined with S100A4 reached statistical significance (Table 1).  Subsequently, we selected formalin-fixed paraffin-embedded (FFPE) tissue samples of patients and controls to validate that these patterns are followed in PC tissue-affected areas. These samples were marked by a pathologist indicating tumor and non-tumor areas; both areas were selected for the analysis to correctly compare expression patterns. As can be seen in Supplementary Figure S2, S100A4 is the one with the highest expression patterns in patients' samples.

Western blot Analysis
It also demonstrated, as did previous data in genomic analysis, that blood samples of severe patients have a higher intensity of the band in MRC2 compared to moderate and controls (Supplementary Figure S3), as well as more expression patterns as can be seen in Supplementary Figure S4, p = 0.002.

Extracellular Vesicles Analysis
Lastly, we analyzed positive S100A4 and MRC2 extracellular vesicles (EVs) extracted of plasma from patients and healthy donors (n = 23). Due to the difficulty of finding EVs in plasma, we reinforced this point using three strategies: (I) use of ImageStream MKII (Amnis, Luminex) ( Figure 2); (II) analysis of the data with IDEAS ® software v.6.0 (selecting easily those EVs with double events); (III) estimate of the number of objects/mL for an exact calculation of EVs ( Supplementary Figures S5 and S6).
As can be seen, plasma EVs are higher in PC samples than in controls, having double the positive markers in 60.000 acquired events (Figure 2). Similarly, it was proven with similar patterns in renal cancer in our group [23,24] (details in Supplementary Figures S5  and S6). The same pattern is proven with ImageStream and IDEAS ® v.6.0; S100A4 fluorescence is increased in patients versus controls, not significant values. In MRC2, this pattern is repeated with significant values (p = 0.0356). However, we should be cautious due to the small sample size in these analyses.

Immunohistochemistry and Immunofluorescence Analysis
Firstly, we evaluated the ability of S100A4 and MRC2 as biomarkers by analyzing tumor and healthy prostatic regions with immunohistochemical techniques (n = 11). We discovered some differences in the staining of MRC2, and a Western blot was performed in this marker to develop a non-invasive quantification. Then, MRC2 and S100A4 gene expression analysis was undertaken using tumor and healthy paraffin-embedded prostatic tissue (n = 13). In this case, we determined a significant difference for S100A4 and MRC2, detecting an increased expression of both molecules for the tumor group, but conclusive results could only be obtained for MRC2 (details in Supplementary Figures  S7-S9)

Sensitivity and Specificity and Biomarker Correlation with Overall Survival or Progression-Free Survival
Given the high expression of PCA3, MRC2, and S100A4, and according to all overall results, we have performed a ROC curve analysis and calculated the AUC to investigate their individual ability to discriminate between PC patients and controls (see Figure 3). We also combined these genes to examine their potential diagnostic advantages. An AUC with a 95% of confidence interval (CI) was obtained to evaluate diagnostic accuracy of the individual blood-derived circulating mRNAs of these genes. In our study population overall survival (OS) rates were 96.6% at 3 years, 95.1% at 10 years, and 88.6% at 20 years. S100A4 was significant for OS, showing reasonable specificity and sensitivity values to predict disease progression. As can be seen, plasma EVs are higher in PC samples than in controls, having double the positive markers in 60.000 acquired events ( Figure 2). Similarly, it was proven with similar patterns in renal cancer in our group [23,24] (details in Supplementary Figures S5 and S6). The same pattern is proven with ImageStream and IDEAS ® v.6.0; S100A4 fluorescence is increased in patients versus controls, not significant values. In MRC2, this pattern is repeated with significant values (p = 0.0356). However, we should be cautious due to the small sample size in these analyses.

Immunohistochemistry and Immunofluorescence Analysis
Firstly, we evaluated the ability of S100A4 and MRC2 as biomarkers by analyzing tumor and healthy prostatic regions with immunohistochemical techniques (n = 11). We discovered some differences in the staining of MRC2, and a Western blot was performed in this marker to develop a non-invasive quantification. Then, MRC2 and S100A4 gene expression analysis was undertaken using tumor and healthy paraffin-embedded prostatic tissue (n = 13). In this case, we determined a significant difference for S100A4 and MRC2, detecting an increased expression of both molecules for the tumor group, but conclusive results could only be obtained for MRC2 (details in Supplementary Figures S7-S9).

Sensitivity and Specificity and Biomarker Correlation with Overall Survival or Progression-Free Survival
Given the high expression of PCA3, MRC2, and S100A4, and according to all overall results, we have performed a ROC curve analysis and calculated the AUC to investigate their individual ability to discriminate between PC patients and controls (see Figure 3). We also combined these genes to examine their potential diagnostic advantages. An AUC with a 95% of confidence interval (CI) was obtained to evaluate diagnostic accuracy of the individual blood-derived circulating mRNAs of these genes. In our study population overall survival (OS) rates were 96.6% at 3 years, 95.1% at 10 years, and 88.6% at 20 years. S100A4 was significant for OS, showing reasonable specificity and sensitivity values to predict disease progression. Although all markers are statistically significant (see Table 1) S100A4 is the biomarker with the best performance, with an AUC value of 0.735 (95% CI = 0.668-0.803) and a specificity of 90.32%. Analysis of the other genes did not show a high discriminatory capacity (AUC < 0.70). However, the different combinations of present genes analyzed in blood provide slightly additional benefits with AUC levels > 0.74 (details in Table 2).

Discussion
Fluids biomarkers currently in use for diagnosis, prognosis, and relapse-monitoring of localized PC remain centered around PSA, with new strategies for genomic biomarkers still presenting a challenge [24]. At present, the established prognostic factors (ISUP, stage, and PSA) are insufficient to separate PC patients with high risk of cancer progression; moreover, molecular markers for metastatic propensity remain elusive [25]. RNA biomarkers are experiencing an increased interest in predicting progression-free survival and OS [26]. In the present article, we reinforced the role of PCA3, MRC2, and S100A4 in combination, and S100A4 on its own, as RNA biomarkers in PC aggressiveness stratification.
In PC, the role of S100A4-antibody therapy and its clinical applicability in treating immunosuppressive PC patients has previously been suggested [25]. Its role in promot- Figure 3. Characteristics of S100A4 as a biomarker in plasma. (A) ROC (Receiver operating characteristic) curves analysis in S100A4; (B) overall survival in days calculated to S100A4 optimal cut-point.
Although all markers are statistically significant (see Table 1) S100A4 is the biomarker with the best performance, with an AUC value of 0.735 (95% CI = 0.668-0.803) and a specificity of 90.32%. Analysis of the other genes did not show a high discriminatory capacity (AUC < 0.70). However, the different combinations of present genes analyzed in blood provide slightly additional benefits with AUC levels > 0.74 (details in Table 2).

Discussion
Fluids biomarkers currently in use for diagnosis, prognosis, and relapse-monitoring of localized PC remain centered around PSA, with new strategies for genomic biomarkers still presenting a challenge [24]. At present, the established prognostic factors (ISUP, stage, and PSA) are insufficient to separate PC patients with high risk of cancer progression; moreover, molecular markers for metastatic propensity remain elusive [25]. RNA biomarkers are experiencing an increased interest in predicting progression-free survival and OS [26]. In the present article, we reinforced the role of PCA3, MRC2, and S100A4 in combination, and S100A4 on its own, as RNA biomarkers in PC aggressiveness stratification.
In PC, the role of S100A4-antibody therapy and its clinical applicability in treating immunosuppressive PC patients has previously been suggested [25]. Its role in promoting cell proliferation and epithelial-mesenchymal transition characteristics in tumor cells, inducing regulating bone metastasis, has also been proposed [12]. However, there are no reports on its predictive role as an aggressiveness non-invasive biomarker; as we have proven in present article, it has an AUC value of 0.735 (95% CI = 0.668-0.803) and a specificity of 90.32%. Moreover, there are no reports of its increased efficiency at screening the non-invasive biomarkers combination of PCA3, MRC2, and S100A4, with an AUC value of 0.761 (95% CI = 0.688-0.835) and specificity of 95.65%.
Although previous studies reported that type I transmembrane collagen receptor Endo180 (CD280, CLEC13E, KIAA0709, MRC2, TEM9 or uPARAP) was a strong prognostic indicator for PC survival [27], this is the first time that its role in PC aggressiveness is proved in blood. The role of PCA3 is known, however is not currently used in largescale clinical practice [28]. We employed dPCR to measure PCA3, MRC2, and S100A4 expression in different grades of PC patients. Our results demonstrate the value of these genes as PC aggressiveness biomarkers. Moreover, Western blot analysis reveals the same expression patterns described in dPCR. The strength of this project is the consistency of data obtained from FFPE, fresh tissues, and blood samples analysis. Furthermore, concerning exosome studies from blood samples, S100A4 allows us to detect the most aggressive stages with high specificity (90.32, p ≤ 0.0001) as previously reported in blood-derived non-extracellular vesicles.
One of the most critical challenges of current medicine is to offer more individualized and personalized medicine to each patient. Several biomarker-based screening kits, such as 4Kscore ® Test (blood) and Select MDx (urine), focus on reducing unnecessary biopsies as well as predicting the risk of developing PC in combination with clinical data. None of these screening biomarkers are used in routine clinical practice. Although precision medicine has come a long way to ameliorate cancer disease management, screening strategies in PC patients are still limited. This study validates the potential of PCA3, MRC2, and S100A4 biomarkers in the screening of PC aggressive stages as an easy, efficient, and not high-cost method, using sensitive dPCR assays in blood samples.
Limitations of our study include the sample size in some experiments; nevertheless, a high number of different methodological analyses approve of the same hypothesis and follow the same pattern. We highlight that, firstly, the analysis was conducted with a massive strategy (RNA-Seq); its costs and the vast data of its analysis assure the success of the following steps. Subsequently, those analyses developed with smaller sample numbers performed in FFPE samples (such as immunohistochemistry/fluorescence), were conducted just as a proof of concept of the previous analysis.

Patients and Sample Collection
A total of 89 PC plasma samples, selected from an extended collection of samples of PC 312 (C.0005252), were used for molecular analysis. We have recruited clinical data from all patients such as T-stage, serum PSA levels, ISUP, and D'Amico risk (low, intermediate, and high risk) (details in Supplementary Table S1). Only patients with PSA ≥ 4.0 ng/mL who met the criteria for undergoing a prostate biopsy were recruited from 2012 to 2014 and included in this study; those with negative biopsy results were included as controls. All patients had localized diseases and treatment-naive at the time of the collection. All individuals underwent a systematic 20-core ultrasound-guided biopsy to limit the false negative rate. Moreover, we have collected buccal swabs and fresh tissue samples. All samples (blood, fresh tissues, and buccal swabs) were stored at −80 • C until they were processed. FFPE tissue samples were also collected from 13 samples (n = 5 controls vs. 8 patients) according to their clinical aggressiveness (n = 4, ISUP > 3; and n = 4, ISUP ≤ 3). Eighty-nine plasma samples of control individuals (mean age of 66) were also analyzed. For more details on sample analysis and selection see Figure 1. The study was approved by the Research Ethics Committee of Granada Center (CEI-Granada internal code 1503-M2-20) following the Helsinki ethical declaration and all study participants provided written informed consent before being enrolled.

Methodology
As was previously described in the results section, we divided the methodology into the same two stages. Each section has a specific protocol and methodology design; the second phase (validation), especially, is performed in fewer samples just to validate the data obtained in the first part of the biomarkers search.

Biomarker Searching
This section includes: (I) RNA extraction and quantification according to the QIAamp RNA miRNeasy Serum/Plasma protocol (QIAGEN, Hilden, Germany); (II) NGS analysis using TruSeq Stranded mRNA Library Prep kit (Catalog # 20020594) (Illumina, San Diego, CA, USA); (III) bioinformatic analysis and potential biomarkers searching using the reference genome GRCH38 by STAR tool (see all details in Supplementary Material). Moreover, in silico analysis validation using Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene Ontology (GO), PathCards Pathway UD, and Reactome is detailed in Supplementary Table S3.
The main strategy of detection is focused on digital PCR (dPCR) analysis in plasma and tissue samples by a QuantStudio™ 3D Digital PCR System (Waltham, MA, USA). Details of probes in Supplementary Table S4.

Validation of the Biomarkers
This section includes: (I) Western blot; (II) EVs analysis; (III) immunohistochemistry and immunofluorescence in cell lines and samples analysis. All these techniques were carried out following conventional procedures (all details in Supplementary Material). Moreover, just to remind, several methodologies such as Western blot are limited to those markers that produce proteins or validate immunohistochemistry analysis (details in Supplementary Table S3).

Conclusions
The present study indicates for the first time the use of PCA3, MRC2, and S100A4 RNA biomarkers in blood for PC aggressiveness. Moreover, S100A4 on its own could serve as an accurate biomarker in PC management of high-grade patients screening. These biomarkers, in combination with clinical data, could offer promising strategies to anticipate the follow-up of aggressive patients. Current clinical strategies are lacking in the application of non-invasive molecular biomarkers in PC patients' stratification. The present discovery opens a new approach in non-invasive genetic monitoring of PC, suggesting the inclusion of these genes as markers in clinical routine.