Determination of ERG(+), EZH2, NKX3.1, and SPINK‐1 subtypes to evaluate their association with clonal origin and disease progression in multifocal prostate cancer

Abstract Background The prognostic relevance of prostate cancer (PCa) molecular subtypes remains controversial, given the presence of multiple foci with the possibility of different subtypes in the same patient. Aim To determine the clonal origin of heterogeneity in PCa and its association with disease progression, SPOP, ERG(+), EZH2, NKX3.1, and SPINK‐1 subtypes were analyzed. Methods A total of 103 samples from 20 PCa patients were analyzed; foci of adjacent non‐tumor prostate tissue, HGPIN, GL3, GL4, GL5, and LN were examined to determine the presence of the TMPRSS2‐ERG fusion and ERG, EZH2, NKX3.1, and SPINK‐1 expression levels, using RT‐PCR. Mutations in exons 6 and 7 of the SPOP gene were determined by sequencing. The presence of subtypes and molecular patterns were identified by combining all subtypes analyzed. To establish the clonal origin of multifocal PCa, molecular concordance between different foci of the same patient was determined. Association of these subtypes with histopathological groups and time to biochemical recurrence (BCR) was assessed. Results No mutation was found in SPOP in any sample. The ERG(+) subtype was the most frequent. The molecular pattern containing all four PCa subtypes was only detected in 3 samples (4%), all LN, but it was the most frequent (40%) in patients. Molecular discordance was the predominant status (55%) when all analyzed molecular characteristics were considered. It was possible to find all subtypes, starting as a preneoplastic lesion, and all but one LN molecular subtype were ERG(+) and NKX3.1 subtypes. Only the expression of the NKX3.1 gene was significantly different among the histopathological groups. No association was found between BCR time in patients and molecular subtypes or molecular concordance or between clinicopathological characteristics and molecular subtypes of ERG, EZH2, and SPINK‐1. Conclusion The predominance of molecular discordance in prostatic foci per patient, which reflects the multifocal origin of PCa foci, highlights the importance of analyzing multiple samples to establish the prognostic and therapeutic relevance of molecular subtypes in a patient. All the subtypes analyzed here are of early onset, starting from preneoplastic lesions. NKX3.1 gene expression is the only molecular characteristic that shows a progression pattern by sample.

subtypes in a patient. All the subtypes analyzed here are of early onset, starting from preneoplastic lesions. NKX3.1 gene expression is the only molecular characteristic that shows a progression pattern by sample. Nevertheless, these criteria do not establish a prognosis with sufficient certainty. 2 Numerous studies have proposed different classifications of PCa molecular subtypes that can contribute to determining prognosis; among them, ETS(+) subtypes characterized by chromosomal fusions that cause the overexpression of ETS transcription factors. The most prevalent ETS fusion in PCa is TMPRSS2-ERG, which leads to ERG overexpression. [3][4][5] ERG overexpression, in turn, generates an increase in the expression of EZH2, which acts as a histone methyltransferase enzyme, and a decrease in the expression of the NKX3.1 tumor suppressor gene, which is an androgen-regulated prostate-specific homeobox gene and a transcription factor. NKX3.1 also binds to TMPRSS2 upstream sequences, and in TMPRSS2-ERG cases, it negatively regulates ERG expression and is a negative feedback control since ERG directly represses NKX3.1. 6,7 EZH2 and NKX3.1 had also been described as PCa subtypes. Another subtype reported in PCa is characterized by SPINK-1 overexpression, which specifically happens in a subset of ETS(À) and is associated with another subtype, the SPOP subtype, characterized by a mutation in this gene. 3 SPINK-1 protein is a trypsin inhibitor, which can function as an autocrine growth factor, 8,9 while SPOP is an E3 ubiquitin ligase substrate-binding subunit of the proteasome complex that mediates the ubiquitination of target proteins, leading more frequently to their proteasomal degradation. 10 The prognostic relevance of PCa molecular subtypes remains controversial. [11][12][13][14][15][16][17][18][19][20] The presence of multiple foci with the possibility of different subtypes in the same patient with PCa could contribute to the difficulty of predicting the clinical behavior of a specific PCa subtype. [21][22][23] It is not well known whether these foci have a monoclonal origin that expands through the prostate, giving rise to new foci that share some molecular characteristics by origin but begin to acquire changes in subclonal events; or whether each focus appears independently, having a multiclonal origin, so they are expected to be very heterogeneous at the molecular level. 22,24,25 To elucidate the origin of heterogeneity in PCa and its association with disease progression, ERG(+), SPOP, EZH2, SPINK-1, and NKX3.1 subtypes were analyzed in 103 samples from prostatic foci and regional metastasis to lymph nodes (LN) from 20 PCa patients with poor prognosis. Molecular patterns are also established by sample and patient, combining the presence of all subtypes analyzed.

| Samples
Formalin-fixed paraffin-embedded (FFPE) tissue samples obtained from radical prostatectomy (RP) of 20 PCa patients with poor prognosis (regional metastases and/or biochemical recurrence) and 5-year follow-up were included. Cases were chosen from the project "Exploration of potential predictor biomarkers in patients diagnosed with prostate cancer," approved by the Ethics Committee of the Instituto Nacional de Cancerología (Bogotá, Colombia) to develop this study.
Since tissues were obtained as part of routine care and not specifically for this study, and the data used in this study were previously anonymized, the Ethics Committee of the Instituto Nacional de Cancerología (Bogotá, Colombia) waived the need for informed consent for this study.
Four to five FFPE samples per patient were selected to identify foci with different degrees of differentiation; a total of 83 samples were analyzed. These samples corresponded to HGPIN (n = 17), PCa foci with different Gleason scores (n = 53; GL3 n = 23; GL4 n = 23; and GL5 n = 7), and LN (n = 13). A sample of adjacent non-tumor prostate tissue was used from each patient as a reference (n = 20).
Two frozen PCa tissue samples were also used as a positive control of the TMPRSS2-ERG fusion, as well as the PC3 cell line was assessed as a negative control for gene fusions. All methods complied with relevant guidelines and regulations at the national and international levels.

| Cell culture
The PC3 cell line was derived from androgen-independent PCa. This line was kindly provided by Dr. Fabio Aristizabal, from the Universidad Nacional de Colombia. The culture conditions were as follows: 5% CO 2 at 37 C, in DMEM medium (Dulbecco/Vogt modified Eagle's minimal essential medium), supplemented with 10% fetal bovine serum and a cocktail of penicillin, streptomycin, and 1% amphotericin B.
2.3 | DNA/RNA extraction and quantification from FFPE tissue samples DNA/RNA was isolated from FFPE tissue by automated extraction in the QIAcube using a modified protocol of the commercial kit AllPrep DNA/RNA FFPE (QIAGEN, Hilden, Germany). The modification consisted of incubating the sample before starting the automated protocol with proteinase K at 56 C for 3 h, instead of 15 min. The RNA was quantified by using the Qubit ® 2.0 fluorometer RNA HS assay (Invitrogen, Waltham, MA). The RNA was stored at À70 C until its use. Good quality DNA was obtained with this same kit, which was also quantified using the Qubit ® 2.0 fluorometer DNA HS assay and stored at À20 C. T A B L E 1 Primers used in the study for RT-PCR, qRT-PCR, and PCR assays with their respective standardized optimal conditions.

| Determination of PCa molecular subtypes
PCa molecular subtypes were assigned as follows: ERG(+) subtype, when TMPRSS2-ERG fusion is present and/or ERG mRNA is highly expressed; SPOP subtype, in the presence of mutations in the SPOP gene; EZH2 subtype, when EZH2 mRNA is highly expressed; NKX3.1 subtype, when NKX3.1 mRNA is lowly expressed; and SPINK-1 subtype, when SPINK-1 mRNA is highly expressed.
These subtypes were assigned by sample and patient; to consider that a patient is positive for a subtype, at least one sample must have this subtype. A molecular pattern is the combination of the presence of all subtypes analyzed in samples and patients.  Figure 1A-E graphically describe the five concordance patterns found: two concordant patterns (patterns A and B), two partially concordant patterns (patterns C and D), and one discordant pattern (pattern E).

| Data analysis
Clinicopathological characteristics were analyzed as continuous vari- The assumption of normality in the quantitative variables was vali-   Table 2.    Table 3.  Table 3). LN samples were excluded from these analyses.  Figure 1F shows concordance patterns with respect to each molecular alteration evaluated by patient. The concordant pattern was the predominant state for all evaluated alterations (40 to 55%), except for SPINK-1 expression, where the partially concordant state was the predominant pattern in 45%. The partially concordant pattern ranged between 20% and 45%, and total molecular concordance, which includes concordant and partially concordant cases, was the most common pattern for all molecular characteristics analyzed, ranging between 70% and 90% (90% for ERG, 85% for fusion status, NKX3.1 and SPINK-1 expression, and 70% for EZH2 expression), which is compatible with monoclonal origin. Discordance was the least common pattern (10%-30%), which would support a multiclonal origin ( Figure 1F). T A B L E 3 Determination of molecular characteristics, subtypes, and molecular patterns by focus and patient, presented by concordance pattern of all molecular characteristics in prostatic foci by patient.

| Determination of concordance in prostatic foci for all molecular alterations evaluated in each patient
After analyzing concordance for each molecular characteristic separately, concordance was evaluated simultaneously for all molecular characteristics analyzed in this study in prostatic foci (n = 70) ( Table 3). A total of 3 out of 20 patients were concordant (15%), 6 patients were partially concordant (30%), and 11 patients were discordant (55%), which shows that heterogeneity and the possibility of multiclonal origin increases from 15 to 30% by molecular characteristic analyzed separately to 55% when the five molecular characteristics are analyzed simultaneously.
Intratumoral concordance was also assessed in foci with the same GL score in 6 patients who had samples with more than one focus with the same GL. Although the foci of the same GL in patients 1, 2, and 9 showed concordance, there were differences in patients 7, 10, and 8 in the presence of fusion and expression of ERG, NKX3.1, and SPINK-1, which indicates intratumoral heterogeneity (Table 3).

| Molecular alterations in metastasis concerning prostatic foci
Metastatic material from the lymph nodes of 13 patients was analyzed ( Table 3). None of the LN analyzed showed the same pattern  3.6 | Histopathological and clinical significance of the presence of ERG, EZH2, SPINK-1, and NKX3.1 subtypes Figure 3A shows the frequency of molecular subtypes by histopathological group. All analyzed subtypes were detected starting from the preneoplastic lesion (HGPIN) in a range of 47%-53%, but the SPINK-1 subtype had a low percentage (18%   Figure 4).
An analysis was carried out by comparing clinicopathological characteristics, molecular subtypes, and concordance patterns in all 20 patients by BCR time, categorized as ≤3 and >3 months ( Table 2).
A statistically significant association was evidenced between BCR time with tumoral percentage (p = 0.040), pT (p = 0.049), and pN (p = 0.029), but no association was found between BCR time and molecular subtypes or molecular concordance. A comparison of the clinicopathological characteristics by molecular subtype with ERG, EZH2, NKX3.1, and SPINK-1 subtypes did not show any statistically significant association (Tables S1-S4).

| DISCUSSION
Although PCa is a multifocal disease, it is not clear to what extent molecular heterogeneity in these foci is a determining factor in the natural history of the disease or its role as a prognostic molecular biomarker for this disease. 32,33 The novelty of this work lies in the infrequency of molecular analyses of many foci in the same PCa with different lesion degrees (HGPIN, GL3, GL4, GL5, and LN). This approach allowed contributing to a better understanding of the molecular heterogeneity of a multifocal disease. In this study, five molecular subtypes associated with PCa were analyzed: ERG(+), EZH2, NKX3.1, SPINK-1, and SPOP mutation, to establish their association with different lesion degrees and regional metastasis in LN in 20 PCa patients with poor prognosis, as well as the clonal origin of the foci in these patients. The ERG(+) subtype includes the TMPRSS2-ERG fusion and ERG overexpression. Since no mutations were identified in the SPOP gene, which may be due to sample size, this was not considered for analysis.
Initially, when determining the presence of molecular subtypes, the ERG(+) subtype was found in 60% of the samples and in 85% of T A B L E 4 Matching lymph node and localized prostatic lesions to show which neoplastic foci contributed to lymph node metastasis.

Most similar prostate lesion
Observed change in LN with respect to the most similar prostate lesion Total number of references (% of cases) Total molecular concordance, which includes concordant and partially concordant cases, was high (70%-90%) for each alteration and supports a monoclonal origin with subclonal events; between 15% and 30% of the cases show discordance, which could be associated with a multiclonal origin. When analyzing all alterations simultaneously per patient, the degree of discordance (55%) was higher than when alterations were evaluated separately (10%-30%), as expected.
Other studies have reported a decrease in molecular concordance as the number of analyzed molecular alterations increases; in particular, the whole genome or transcriptome shows that the foci vary widely in the same patient. [40][41][42] However, based on the results of our study, it cannot be discarded that some multifocal PCa may have a monoclonal and others a multiclonal origin; that is, there is no single origin for all PCa.
It is noteworthy that there were intratumoral differences in foci with the same GL in 3 out of 6 patients who had samples with more than one focus with the same GL. Nevertheless, concordance in the other 3 patients supports the probability of a monoclonal origin. Wei et al. 22 showed that the correlation between different PCa foci in an individual is low. These findings highlight the importance of evaluating the role of multiclonality, which presents different alterations and subtypes in the same patient, complicating its application in translational oncology since alterations other than the index tumor may be determining metastases. 23,43 LN were not considered for concordance analysis, but it should be noted that none of the 13 LN studied was the same as the other foci analyzed, which underlines that malignant cells need other subclonal events to form metastatic foci in LN. This is in accordance with what was said above about subclonal alterations, which must provide additional advantages for such sophisticated processes as metastasis. 40 Regarding the association between subtypes and histopathological groups, it was possible to observe all of them in HGPIN lesions, which leads to the conclusion that all the evaluated alterations are early events in the natural history of PCa. It is worth noting that the constant in all but one LN molecular subtype was the ERG(+) and Only the expression of the NKX3.1 gene was significantly different among the histopathological groups. Its expression decreased in more advanced lesions. Thus, the lowest levels were found in the LN group, indicating regional metastasis; this finding is consistent with other reports where NKX3.1 has been found downregulated in many PCa tissues. It has been reported lost or completely eliminated even in the most dysplastic and metastatic PCa tumors. 11,[44][45][46] No association was found between BCR time and clinicopathological characteristics, molecular subtypes, or molecular concordance, which may be due to our sample size and highlights the importance of future studies with larger sample sizes that are more representative.

| CONCLUSIONS
The results of this study support the notion that, given the molecular heterogeneity of PCa, the molecular evaluation of a single focus does not provide a complete molecular picture of the disease in each patient. Although it is possible to assign molecular subtypes in this way, this type of analysis hinders the association of molecular subtypes with prognosis and the choice of appropriate treatment.