Comprehensive Analysis of CXCR4, JUNB, and PD-L1 Expression in Circulating Tumor Cells (CTCs) from Prostate Cancer Patients

CXCR4, JUNB and PD-L1 are implicated in cancer progression and metastasis. The current study investigated these biomarkers in CTCs isolated from metastatic prostate cancer (mPCa) patients at the RNA and protein levels. CTCs were isolated from 48 mPCa patients using the Ficoll density gradient and ISET system (17 out of 48). The (CK/PD-L1/CD45) and (CK/CXCR4/JUNB) phenotypes were identified using two triple immunofluorescence stainings followed by VyCAP platform analysis. Molecular analysis was conducted with an EpCAM-dependent method for 25/48 patients. CK-8, CK-18, CK-19, JUNB, CXCR4, PD-L1, and B2M (reference gene) were analyzed with RT-qPCR. The (CK+/PD-L1+/CD45-) and the (CK+/CXCR4+/JUNB+) were the most frequent phenotypes (61.1% and 62.5%, respectively). Furthermore, the (CK+/CXCR4+/JUNB-) phenotype was correlated with poorer progression-free survival [(PFS), HR: 2.5, p = 0.049], while the (CK+/PD-L1+/CD45-) phenotype was linked to decreased overall survival [(OS), HR: 262.7, p = 0.007]. Molecular analysis revealed that 76.0% of the samples were positive for CK-8,18, and 19, while 28.0% were positive for JUNB, 44.0% for CXCR4, and 48.0% for PD-L1. Conclusively, CXCR4, JUNB, and PD-L1 were highly expressed in CTCs from mPCa patients. The CXCR4 protein expression was associated with poorer PFS, while PD-L1 was correlated with decreased OS, providing new biomarkers with potential clinical relevance.


Introduction
Prostate cancer is the second most common cancer and the fifth leading cause of cancer mortality in men worldwide [1].Newly diagnosed patients are usually treated with the standard treatment approach of androgen deprivation therapy, although in many cases the development of resistance leads to metastasis [2].Prostate cancer patients with distant metastases, particularly to the bone, have limited effective treatment options and poorer disease outcomes [2,3].Therefore, early diagnosis combined with the discovery of new treatment approaches is crucial for effective treatment of even the most aggressive forms of prostate cancer [4].Liquid biopsy is an effective and minimally invasive tool for discovering new therapeutic targets and monitoring their efficacy in real time.Tumor circulome in liquid biopsies, unlike tissue biopsies, can elucidate the clonal variation of tumors [5].Liquid biopsy components which include circulating tumor cells (CTCs), circulating tumor DNA (ctDNA), RNA, extracellular vehicles (EVs), and proteins, provide information about potential metastatic lesions, and reveal the cancer profile [6,7].
CTCs play a key role in the spread of metastases as they enter the bloodstream and migrate to form distant lesions [8].CTCs that migrate into the bloodstream exhibit both epithelial and mesenchymal properties [9].The mesenchymal nature of CTCs can be an obstacle to their isolation [9,10], as many methods are epithelial-dependent, such as CellSearch TM , the gold standard for detection and isolation of CTCs, particularly in breast [11], colorectal [12], and prostate cancer [13].CTCs can survive either as individual cells or as cell clusters that can remain inactive for many years until they develop recurrences or metastasis, indicating the intolerance of CTCs to the lethal effects of existing therapeutic approaches [14,15].
Programmed cell death protein 1 (PD-1) is normally located on the surface of activated T cells, and the corresponding ligand (PD-L1) is physiologically expressed on macrophages, dendritic cells, natural killer cells, B lymphocyte cells, and vascular endothelial cells [16].In patients, PD-L1 is expressed in cancer cells and inhibits the cytotoxic effect of CD8+ T cells against them [3,16].Several studies have shown increased PD-L1 protein expression in CTCs of different cancer types, including prostate cancer, which is potentially related to patient outcomes [15][16][17][18][19]. Furthermore, the evolution of molecular assays for the detection of PD-L1 mRNA expression has shown that the detection of PD-L1 expression in CTCs is practicable and can provide real-time clinical information [20].In head and neck squamous cell carcinoma (HNSCC), the detection of CTCs overexpressing PD-L1 using an RT-qPCR method at the end of definitive non-surgical treatment including chemoradiation correlates with a lower probability of achieving complete response (CR) and a higher risk of relapse and death [21].
CXCR4 is normally involved in homeostatic processes in the bloodstream, such as leukocyte transport and hematopoiesis [22].On the other hand, it may also be involved in carcinogenesis and metastasis in various cancers, including prostate cancer [22][23][24].This is confirmed by a meta-analysis showing that increased CXCR4 expression in prostate cancer samples correlates to metastases [25].
JUNB is a transcription factor that can act as an oncogene and induce abnormal proliferation of quiescent cells [26].In the prostate cancer cell line PC-3, JUNB has been indicated to promote migration and antagonize the opposite function of JUND [27].In addition, a full transcriptome analysis showed that JUNB is present in severely affected patient groups and metastatic lesions, which are thought to play a role in prostate cancer progression [2].On the other hand, JUNB demonstrates a new upstream signaling role triggering the activation of p16/pRb, which is crucial for the initiation and maintenance of senescence.Additionally, activated JUNB effectively hinders the malignant progression of prostate cancer as measured by invasion and metastasis [28].
Both CXCR4 and JUNB are highly expressed in CTCs isolated from metastatic breast [29] and non-small cell lung cancer (NSCLC) patients [30], as well as in disseminated tumor cells (DTCs) isolated from early-stage breast cancer migrated in bone marrow [31].The presence of JUNB in CTCs from breast cancer patients is correlated to poor prognosis [29].In addition, CXCR4 and JUNB have been associated with poorer overall survival (OS) of early-stage breast cancer patients [31] and lower progression-free (PFS) and overall survival of NSCLC patients [30].In the current study, we investigated for the first time the expression, at the RNA and protein level, of CXCR4, JUNB, and PD-L1 in CTCs from (mPCa) patients.Furthermore, we investigated the correlation between the results on protein and RNA levels and the clinical relevance of the different CTC phenotypes and mRNA patterns with the clinical outcome.

Characteristics of the Patients
Patients' characteristics for every cohort, depending on the analyzed method used, are presented in Tables S1-S3.A total of 48 patients with mPCa and 10 healthy donors were included in the study.The 48 patients provided blood collection for CTC identification through Ficoll density gradient isolation (Table S1).A total of 25 out of 48 patients were additionally evaluated for mRNA analysis (Table S2), and 17 out of 48 patients were additionally evaluated for CTC identification through ISET system isolation (Table S3).The mean age of the 48 patients was 75 years (range: 42-87 years), of the 25 patients was 73 years (range: 44-87 years), and of the 17 patients was 74 years (range: 65-83 years).A total of 38 patients were treatment naïve (baseline sample), and 10 patients were enrolled before the initiation of second-line treatment.The protocol was approved by the ethics and scientific committees of all participating institutions [Venizeleion, General Hospital, (136/26/26-10-2022); Metropolitan General Hospital, (172/18-09/2020); Larissa General University Hospital, (32710/3-8-20); ST Andrews General Hospital of Patras, (521/13-10-2020)].The patients and the healthy donors provided their written informed consent for blood collection and for the use of clinical outcome information for research endeavors.

Blood Sampling and Cytospin Preparation in Patients
Ten mL of peripheral blood was collected from all patients and from 10 healthy donors, using K2EDTA tubes.All blood samples were collected by mid-vein puncture, and the first 5 mL was discarded to avoid contamination with skin epithelial cells.Peripheral blood mononuclear cells (PBMCs) from mPCa patients were isolated by Ficoll-Hypaque (d = 1.077 g/mol) density centrifugation at 1800 rpm for 30 min without dissertation.The Ficoll density method enables isolation of CTCs contained in the mononuclear fraction of peripheral blood mononuclear cells, which is size-and EpCAM-independent.After washing twice with PBS and centrifuging at 1500 rpm for 10 min, aliquots of 500,000 patients' PBMCs/500 µL were cyto-centrifuged at 2000 rpm for 2 min on Superfrost glass slides (Thermo Fisher Scientific, Waltham, MA, USA).Cytospins were dried and stored at −80 • C.

ISET Isolation
In a subgroup of 17 mPCa patients, who were able to provide additional 10 mL of blood, ISET (Isolation by Size of Epithelial Tumor cells) technology was applied.This device employs size-based, label-independent isolation to effectively capture CTCs with diverse phenotypic profiles, as CTCs usually have a larger size compared to most leukocytes [32].Following the manufacturer's guidelines, CTC isolation was performed within 2 days from blood sampling, using 10 mL of peripheral blood collected in K2EDTA tubes.The blood was diluted in a 1:10 ratio with ISET buffer (Rarecells, Paris, France) for 10 min at room temperature, facilitating erythrolysis and maintaining CTC integrity.The diluted samples were then filtered through the ISET membrane at a depression of 10 kPa.Fixed cells, including CTCs unable to pass through the pores (8 µm), were held on the membrane, creating 10 spots.The threshold of method's detection, specified by the manufacturer, is 1 CTC per 10 mL of blood.For each patient, a spot containing 10 6 cells was utilized for CTC identification and the assessment of PD-L1 expression.This was accomplished by employing triple immunofluorescence staining, subsequently followed by VyCAP microscopy analysis.

Triple Immunofluorescence Analysis
One slide from each patient was used for triple immunofluorescence staining to identify CTCs and assess PD-L1 expression, using antibodies against cytokeratins (CK), PD-L1, and CD45 (CK/PD-L1/CD45).The presence of CK, using an A45-B/B3 antibody for the detection of CK-8, CK-18, and CK-19, was used to characterize a nucleated cell as CTC.CD45 expression was used as a negative hematopoietic biomarker expressed only in PBMCs.Many CTCs with a more mesenchymal nature may have decreased levels of CK, so, additionally, the cytomorphologic criteria described by Meng et al. (high nuclear/cytoplasmic ratio, larger cells than white blood cells, irregular nuclear shape, and size) were applied to identify a cell as a CTC [32].Furthermore, morphological characteristics of CTCs described by Adam et al. and Park et al. were observed in subpopulations of CTCs detected in the present study [33,34].According to the CTC morphology observed on the (CK/PD-L1/CD45) staining, a second slide from each patient was analyzed for the assessment of CXCR4 and JUNB expression in CTCs, using antibodies against CK, CXCR4, and JUNB (CK/CXCR4/JUNB), as we have previously reported [19,30].
Positive and negative controls were prepared for each experiment using H1299 or PC-3 cells spiked into PBMCs from healthy volunteers (1000 H1299 cells or PC-3 cells/100,000 PBMCs) to imitate CTCs' microenvironment in patients' cytospins slides.Negative controls were generated by excluding the incubation of corresponding primary antibodies and incubating the cells with the corresponding secondary antibody.Each experiment included three negative controls (one negative for every primary antibody used) and one positive control for all antibodies, to assess methods' sensitivity and specificity.
Slides from both triple immunofluorescence stainings were analyzed using the Vy-CAP system (VyCAP B.V., Enschede, The Netherlands).The VyCAP system operates as an imaging platform, where cytospins slides can be scanned automatically using four different channels.In the current study, we used DAPI, CK, PD-L1, CD45 for the first immunofluorescent staining and DAPI, CK, JUNB, CXCR4 for the second immunofluorescent staining.The corresponding frames were analyzed to identify and characterize patients' CTCs.Additionally, the corresponding frames of patients' slides were also double-checked using the ACCEPT software (automatic software for CTCs detection, University of Twente, Enschede, The Netherlands).Images for control samples and CTCs were captured based on the determined exposure time and examined on negative and positive controls for each antigen.The identification of CTCs was conducted blind to clinical data.
Results were presented as percentage of patients expressing specific phenotype among the total CTC-positive patients.Specifically, percentages of patients for a specific phenotype were calculated as follows: Percentage of patients = (number of patients with the phenotype/total CK-positive patients) * 100%, Regarding the percentage of CTCs to address the variability in the number of isolated CTCs per patient and ensure that each patient's contribution was weighted equally, we evaluated the percentages of CTCs corresponding to specific phenotypes for each patient individually.This involved calculating the percentage of CTCs exhibiting a specific phenotype within the total CTCs of each patient.Here is how this was performed: Frequency of CTCs per patient = (patient CTCs with the phenotype/total patient CTCs) * 100%, ( Then the average of all patients' frequencies was calculated, for every phenotype.

mRNA Analysis
CTC Isolation, RNA Extraction, and cDNA Synthesis In total, 25 mPCa patients and 10 healthy individuals were included, and 10 mL of peripheral blood was collected from all of them for mRNA analysis.Prior to analysis, the PB was immediately mixed by gentle inversion 10 times.Following the addition of 30 mL of red cell lysis buffer (containing NH 4 Cl, 155 mmol/L; KHCO 3 , 10 mmol/L; and EDTA, 0.1 mmol/L, pH = 7.3), samples were incubated for 20 min at room temperature, with occasional gentle inversion for mixing.After centrifugation of 530× g at room temperature for 20 min, the supernatant was removed, and, subsequently, red cell lysis buffer (30 mL) was added.After centrifugation (530× g, room temperature, 10 min), the supernatant was removed, and 10 mL of red cell lysis buffer was added.Magnetic beads, coated with the monoclonal antibody BerEP4 against the human epithelial antigen EpCAM, were further used for CTC enrichment (Dynabeads®Epithelial Enrich, Life Technologies, Waltham, MA, USA) [35].

External Quality Controls
For the development of molecular assays for CXCR4 and JUNB, synthetic standards corresponding to two different concentrations were used: one low (10 copies) and one medium (10 3 copies).By using these two different positive controls, we selected the optimal concentration of each component that ensured the best performance of the two assays.The analytical linearity and sensitivity of the two RT-qPCR assays for JUNB and CXCR4 were evaluated by analyzing two different types of external standards: (a) synthetic DNA standards ranging from 10 5 copies/µL to 10 copies/µL (Supplementary Figure S1A,B) and (b) cancer cell lines ranging from 10 4 cells to 10 cells (Supplementary Figure S1C,D).H1299 was served as a positive control for mRNA expression of JUNB, while MDA-MB-231 was utilized as a positive control for mRNA expression of CXCR4.

RT-qPCR
The detection of mRNA of PD-L1, CK-19 46, CK-8, CK-18, and the reference gene of beta-2-microglobulin (B2M) [37], which was conducted through RT-qPCR assays to confirm the quality of the analyzed samples, has been reported previously.Two novel RT-qPCR assays were developed for the detection of CXCR4 and JUNB mRNA transcripts and validated for analytical specificity and sensitivity.The development of two single RT-q PCR assays was relied on an in silico design of highly specific probes and primers.The in silico design was performed using Primer Premier 5.0 software (Premier Biosoft, San Francisco, CA, USA).Homology searches in the nucleotide database (NCBI, Nucleotide BLAST) were used as an initial evaluation of specificity of the primers and probes.We also designed two synthetic DNA oligos that we used as standards for the analytical validation of RT-qPCR assays for the detection of CXCR4 and JUNB mRNA transcripts.

Normalization for mRNA Expression Data
Normalization to quantify the expression of CXCR4, JUNB, and PD-L1 in CTCs was performed using the 2 −∆∆Cq method for the expression of B2M as a reference gene [39].More specifically, the expression of the CXCR4, JUNB, and PD-L1 genes was estimated as a relative ratio to the expression of B2M both in the CTC fraction of the patients and in the corresponding CTC fraction of the healthy donor group used as a calibrator.

Statistical Analysis
The statistical software SPSS version 27 (IBM, Armonk, NY, USA) was used for statistical tests at the significance level p < 0.05.Progression-free survival (PFS) was estimated as the time within patient enrollment in the study and disease progression.Overall survival (OS) was estimated as the time within patient enrollment and last follow-up or death from any cause.In order to determine the impact of each CTC phenotype on PFS, univariate Cox regression analysis was performed.Kaplan-Meier analysis was used to investigate the presence of CTC numbers or specific CTC phenotypes in correlation with patients' clinical outcomes.To compare Kaplan-Meier curves and Cox regression analysis for PFS and OS, the log-rank test was used.χ 2 was performed to investigate whether there was a statistically significant difference between the observed phenotypes with PD-L1 in Ficoll density gradient isolation versus ISET isolation.The Wilcoxon rank test was performed to examine quantitative differences between the median values of all tested groups.The Mann-Whitney test was performed to determine the differences between the median fold changes of CXCR4, JUNB, and PD-L1 levels between groups.
No statistically significant differences were observed between treatment-naive patients and patients who had yet to undergo initiation of second-line treatment regarding PD-L1 expression.

CXCR4 and JUNB Protein Expression in CTCs from mPCa Patients
Representative images of a CTC isolated from an mPCa patient with simultaneous expression of CXCR4 and JUNB is shown in Figure 3.In addition, positive controls with spiked H1299 or PC-3 cells in 100,000 PBMCs from healthy donors are presented in Supplementary Figure S3.
No statistically significant differences were observed between treatment-naive patients and patients who had yet to undergo initiation of second-line treatment regarding PD-L1 expression.
No statistically significant differences were observed between treatment-naive patients and patients who had yet to undergo the initiation of second-line treatment regarding CXCR4 or JUNB expression.S5).No statistically significant differences were observed between treatment-naive patients and patients who had yet to undergo the initiation of second-line treatment regarding CXCR4 or JUNB expression.

Limit of Detection (LOD) and Limit of Quantification (LOQ) of RT-qPCR for CXCR4 and JUNB
The LOD and LOQ were determined using absolute quantification of low-input calibrators.In all cases, the LOD was shown to correspond to 3 copies/µL.The limit of quantification (LOQ), defined as three times the LOD, corresponds to 9 copies/µL.This is consistent with the MIQE guidelines for RT-qPCR assay development [38], according to which the minimum number of copies in a sample that can be accurately measured is equivalent to 3 copies/µL [36].In addition, the assays developed demonstrated linearity across the entire quantification range (10 5 -10 copies) and correlation coefficients exceeding 0.99 in all cases, which indicated a precise log-linear relationship.

Relative Fold Change of CXCR4, JUNB and PD-L1 Levels
Before proceeding with the combined gene expression analysis, the quality of all cDNAs of the two different groups of samples was checked by RT-qPCR for B2M (reference gene) (Figure S4).B2M expression is used as an internal control for sample quality to avoid false negative results but also as a reference gene for relative quantification.As shown in Supplementary Figure S4, the median B2M Cq value in the HD group was 22.52, while it was 22.45 in the cancer group (Mann-Whitney test, Ζ = −0.530,p = 0.596).The LOD and LOQ were determined using absolute quantification of low-input calibrators.In all cases, the LOD was shown to correspond to 3 copies/µL.The limit of quantification (LOQ), defined as three times the LOD, corresponds to 9 copies/µL.This is consistent with the MIQE guidelines for RT-qPCR assay development [38], according to which the minimum number of copies in a sample that can be accurately measured is equivalent to 3 copies/µL [36].In addition, the assays developed demonstrated linearity across the entire quantification range (10 5 -10 copies) and correlation coefficients exceeding 0.99 in all cases, which indicated a precise log-linear relationship.

Relative Fold Change of CXCR4, JUNB and PD-L1 Levels
Before proceeding with the combined gene expression analysis, the quality of all cDNAs of the two different groups of samples was checked by RT-qPCR for B2M (reference gene) (Figure S4).B2M expression is used as an internal control for sample quality to avoid false negative results but also as a reference gene for relative quantification.As shown in Supplementary Figure S4, the median B2M Cq value in the HD group was 22.52, while it was 22.45 in the cancer group (Mann-Whitney test, Z = −0.530,p = 0.596).The Kruskal-Wallis test was the same between the two different sample groups (Kolmogorov-Smirnov, Z = 0.535, p = 0.938) (Figure S4).

Molecular CTC Profile of mPCa Patients
In 25 of these patients, we had material available for subsequent molecular analysis after CTC isolation.The first 5 mL of blood was discarded for CTC isolation from PB to prevent contamination of skin epithelial cells.The peripheral blood was then collected in a 10 mL K2EDTA tube (BD Vacutainer, Plymouth, UK) at the blood collection centers and transported to the ACTC laboratory.CTC isolation was performed within 2 days, as CTCs are detectable as early as 48 h after blood collection when the PB is stored in K2EDTA tubes.
We used the Wilcoxon signed-rank test for related samples to determine quantitative variations between the median values of the relative fold changes for each gene.The median differences between the analysis of mRNA expression of JUNB and CXCR4 (p = 0.979) and the analysis of mRNA expression of JUNB and PD-L1 (p = 0.06) were not found to be significant.However, when analyzing the relative fold-change data between PD-L1 and CXCR4 (p = 0.020), significant differences were found (Figure 6).

Clinical Significance
Survival analysis was performed for all observed molecular and protein p The univariate Cox regression analysis for the Ficoll density gradient isolation, patients out of 48 with available data of PFS duration, showed a significantly high of relapse in the group of patients harboring CTCs of the (CK+/CXCR4+/JUNB-) p type (quantitative variable) (HR: 2.5, p = 0.049).No statistically significant corr was found in the survival analysis for PD-L1 in the Ficoll density gradient isolatio Regarding the ISET isolation, 13 patients had available data on PFS duratio 15 patients out of 17 had available data on OS duration.Kaplan-Meier analysis ind a significantly shorter PFS (HR: 1137.5, p = 0.046) for 2 patients with ≥2 CTCs com to 11 patients who had ≤1 CTCs detected (3 months with a range of 1.6-4.4 vs. 7.5 m with a range of 5.8-9.1), and a decreased OS (HR: 9.0, p = 0.031) for 3 patients w CTCs compared to 12 patients who had ≤1 CTCs detected (3.7 months with a ra

Combined Analysis between Molecular and Protein Analysis
Our results from the molecular CTC analysis were compared with the immunofluorescence analysis.The agreement (negative and positive samples) between mRNA and protein expression was (a) for JUNB, 17/25 samples (68.0%),(b) for CXCR4 expression, 12/25 (48.0%),(c) for PD-L1 expression, 13/25 (52.0%).Our results showed a higher sensitivity for the combination of molecular and IF analysis.Specifically, the positivity rate in these 25 patients increased for JUNB to 10/25 (40.0.0%), the positivity rate for CXCR4positive patients increased to 15/25 (60.0%), and the positivity rate for PD-L1-positive patients increased to 17/25 (68.0%) (Supplementary Figure S5a).

Clinical Significance
Survival analysis was performed for all observed molecular and protein profiles.The univariate Cox regression analysis for the Ficoll density gradient isolation, for 23 patients out of 48 with available data of PFS duration, showed a significantly higher risk of relapse in the group of patients harboring CTCs of the (CK+/CXCR4+/JUNB-) phenotype (quantitative variable) (HR: 2.5, p = 0.049).No statistically significant correlation was found in the survival analysis for PD-L1 in the Ficoll density gradient isolation.
Regarding the ISET isolation, 13 patients had available data on PFS duration, and 15 patients out of 17 had available data on OS duration.Kaplan-Meier analysis indicated a significantly shorter PFS (HR: 1137.5, p = 0.046) for 2 patients with ≥2 CTCs compared to 11 patients who had ≤1 CTCs detected (3 months with a range of 1.6-4.4 vs. 7.5 months with a range of 5.8-9.1), and a decreased OS (HR: 9.0, p = 0.031) for 3 patients with ≥2 CTCs compared to 12 patients who had ≤1 CTCs detected (3.7 months with a range of 0.6-6.7 vs. 12.8 months with a range of 10.5-10) (Figure 7a,b).Furthermore, the detection of ≥1 CTC in six patients was also associated with lower OS compared to nine patients who had no CTCs (HR: 115.6, p = 0.032) (3.5 months with a range of 1-14 vs. 5 months with a range of 0-10) (Figure 7c).In addition, a significantly lower OS through Kaplan-Meier analysis in ISET isolation was observed for 4 patients harboring CTCs positive for PD-L1 (CK+/PD-L1+/CD45-) contrary to 11 patients without this phenotype (HR: 262.7, p = 0.007) (3 months with a range of 1-14 vs. 4 months with a range of 0-10) (Figure 7d).No other correlations were found between CTC numbers and phenotypes with other patients' characteristics.

Discussion
The diversity of CTCs regarding their morphological features, protein expression pa erns, and molecular profile has been previously examined by many studies [33,34,41,42].To date, there are many different methods for CTC isolation based on distinct biological or morphological characteristics [40,43,44], each of which isolates subpopulations and not all the existing CTCs.This fact makes the detection of CTCs a major challenge [9,10].Furthermore, previous studies by our group have indicated a high expression of CXCR4, JUNB, and PD-L1 in CTCs and/or DTCs of breast cancer and NSCLC patients, related to their survival [29][30][31].In the present study, we combined the results of different CTC isolation methods, (a) an EpCAM-independent method (Ficoll) [19,30] and (b) an EpCAM-dependent method [35,36], in association with two different detection methods, (a) immunofluorescence analysis [19,30] and (b) molecular analysis [21], for a more holistic approach regarding the expression status of the above biomarkers in the CTCs of prostate cancer patients.
In 17 patients with enough volume of blood for further analysis, we also performed CTC isolation through the size-based ISET technology, which raised the CTCs' detection rate from 37.5% (18 out of 48) to 47.1% (8 out of 17), in line with our previous study [40].A higher recovery rate through ISET filtration compared to Ficoll was also obvious for In addition, a significantly lower OS through Kaplan-Meier analysis in ISET isolation was observed for 4 patients harboring CTCs positive for PD-L1 (CK+/PD-L1+/CD45-) contrary to 11 patients without this phenotype (HR: 262.7, p = 0.007) (3 months with a range of 1-14 vs. 4 months with a range of 0-10) (Figure 7d).No other correlations were found between CTC numbers and phenotypes with other patients' characteristics.

Discussion
The diversity of CTCs regarding their morphological features, protein expression patterns, and molecular profile has been previously examined by many studies [33,34,41,42].To date, there are many different methods for CTC isolation based on distinct biological or morphological characteristics [40,43,44], each of which isolates subpopulations and not all the existing CTCs.This fact makes the detection of CTCs a major challenge [9,10].Furthermore, previous studies by our group have indicated a high expression of CXCR4, JUNB, and PD-L1 in CTCs and/or DTCs of breast cancer and NSCLC patients, related to their survival [29][30][31].In the present study, we combined the results of different CTC isolation methods, (a) an EpCAM-independent method (Ficoll) [19,30] and (b) an EpCAM-dependent method [35,36], in association with two different detection methods, (a) immunofluorescence analysis [19,30] and (b) molecular analysis [21], for a more holistic approach regarding the expression status of the above biomarkers in the CTCs of prostate cancer patients.
In 17 patients with enough volume of blood for further analysis, we also performed CTC isolation through the size-based ISET technology, which raised the CTCs' detection rate from 37.5% (18 out of 48) to 47.1% (8 out of 17), in line with our previous study [40].A higher recovery rate through ISET filtration compared to Ficoll was also obvious for the absolute number of CTCs [7 CTCs (range 0-97) vs. 1 CTC (range 0-6)].However, no statistically significant differences were found between the Ficoll and ISET methods for the number of total CTCs and the (CK+/PD-L1+/CD45-) and (CK+/PDL1-/CD45-) phenotypes, through the Wilcoxon rank test (p = 0.591, p = 0.714, and p = 0.206, respectively) and χ 2 test (p = 0.057, p = 0.644, and p = 1, respectively).This may be owed to the smaller number of patients enrolled in ISET isolation compared to the number of patients enrolled in Ficoll density isolation (17 versus 48).Further studies with larger patient cohorts will be needed to confirm these results.In samples processed through the Ficoll density gradient, simultaneous identification of PD-L1-positive and PD-L1-negative CTCs was not observed.These results could be attributed to the limited number of CTCs isolated with the Ficoll density gradient [40].Conversely, the ISET system evades this issue by isolating larger numbers of tumor cells.The ISET platform facilitates the CTCs' enrichment, as the membrane mostly holds cells with diameters equal to or exceeding 8 µm, while other blood cells and constituents are discarded.Employment of the ISET isolation enabled the observation of PD-L1 heterogeneity in CTCs of prostate cancer patients.Particularly, 37.5 of the CTC-positive patients harbored both phenotypes (PD-L1-positive and PD-L1-negative) in their blood, providing more informative results regarding tumor cells metastasis.
CellSearch TM stands as the only FDA-approved system for identifying CTCs retrieved from individuals diagnosed with mPCa [45][46][47][48].However, the detection of CTCs with a mainly mesenchymal phenotype, due to epithelial-mesenchymal transition (EMT), characterized by the absence of EpCAM, may be inadequate with CellSearch TM [10,33,42].An earlier study involving 34 human breast cancer cell lines revealed the inability of CellSearch TM to identify normal-like breast cancer cell lines [49].Moreover, in another study, a greater number of CTCs was identified in castration-resistant patients, using an antibody-targeting cell-surface vimentin, in contrast to CellSearch TM [10].It has been reported that different detection systems could complement the EpCAM-based system of CellSearch TM by improving the overall rate of CTC detection [33,50,51].For instance, a recent report involving 30 patients with breast and prostate cancer evaluated CTC isolation using a microfiltration system (CellSieve) and the CellSearch TM system.The study revealed that the CellSieve yielded a higher number of isolated CTCs with different morphological features, while the CellSearch TM system detected a subpopulation of CellSieve CTCs [33].Furthermore, the presence of CTCs detected using CellSearch TM as an EpCAM-dependent method and two EpCAM-independent detection systems is predictive of decreased OS in patients with metastatic breast cancer [52].Our study is in line with the above notions, as the positivity rate combining mRNA analysis and immunofluorescent stainings for (CK+JUNB+), (CK+CXCR4+), and (CK+PD-L1+) cells was increased (Figure S5a).
Our results indicated that the (CK+/PD-L1+/CD45-) phenotype is predominant in patients with mPCa using both Ficoll density gradient method (61.1%) and ISET filtration (75.0%).Interestingly, in ISET isolation the (CK+/PD-L1+/CD45-) phenotype was correlated with decreased OS with a Kaplan-Meier analysis (HR: 262.7, p = 0.007).These results are in line with a previous study on triple-negative breast cancer (TNBC) patient CTCs showing that the expression of PD-L1 is linked to lower OS in Kaplan-Meier analysis (HR: 8.7, p < 0.001) [19].Furthermore, the investigation of PD-L1 expression in CTCs using the CellSearch TM platform has shown that PD-L1 is frequently expressed in metastatic breast cancer CTCs (68.8%) [18].Similar results have been shown for advanced NSCLC patients, indicating that PD-L1-positive CTCs are correlated to poor prognosis [53].In prostate cancer, Zhang et al. have shown that at least one PD-L1-positive CTC was detected at baseline in 40% of men with metastatic hormone-sensitive prostate cancer (mHSPC), 60% of men with metastatic castration-resistant prostate cancer (mCRPC) starting abiraterone acetate/prednisone or enzalutamide (pre-ARSI), and 70% of men with mCRPC post-ARSI [17].In addition, a study investigating the expression of PD-L1 on circulating epithelial tumor cells (CETCs) of different types of cancer, including prostate cancer, showed that PD-L1 is expressed in 100% of CETCs from prostate cancer patients [16].The detection of PD-L1 in prostate cancer could lead to the customization of immunotherapies [54], and its combination with other drugs could bring encouraging results and benefits [3].Therefore, our study provides an interesting protocol for the evaluation of PD-L1 in CTCs from PC patients with potential clinical significance.Furthermore, our data showed that both CXCR4 and JUNB were expressed at the protein level in the CTCs of most CK-positive mPCa patients.Similar observations regarding high expression of CXCR4 and JUNB were noted in our prior investigations involving metastatic [29] and early breast cancer patients [31], as well as in NSCLC patients [30].The most frequent phenotype was (CK+/CXCR4+/JUNB+) (62.5%), as we have previously shown (90% in breast cancer and 50% in NSCLC) [30,31], and (CK+/CXCR4-/JUNB+) showed the lower percentage of positivity (18.8%) in contrast to the breast cancer [31] and NSCLC study [30], where (CK+/CXCR4+/JUNB-) was the rarest phenotype (5% and 6%, respectively).The number of CTCs expressing the (CK+/CXCR4+/JUNB-) phenotype was associated with poorer PFS in the Cox regression analysis (HR: 2.5, p = 0.049).Consistent with this observation, in extensive-stage small cell lung cancer (SCLC) patients, the presence of ≥4 CXCR4-positive CTCs was correlated to decreased OS (HR: 5.01, p = 0.041) [30].Recently, CXCR4-positive CTCs were shown to persist in the blood of prostate cancer patients treated with radiotherapy for up to 3 months, suggesting that CXCR4-positive CTCs may represent aggressive CTC subclones that contribute to treatment resistance [23].CXCR4 antagonists could be used in a stratified manner in the treatment of CRPC [55,56].In PC-3 tumor xenograft models, AMD3100 has been shown to inhibit tumor growth and reduce microvessel formation by inhibiting CXCR4/Akt signal transduction [55], and CTCE-9908 has been shown to reduce tumor spread and angiogenesis in an orthotopic prostate cancer model [56].Therefore, anti-CXCR4 treatment in prostate cancer could be a potential alternative, and our protocol may provide a companion diagnostic test.
Gene expression analysis for PD-L1, CXCR4, and JUNB was performed on a small cohort of 25 patients.In an earlier study by Zavridou et al., PD-L1 mRNA expression in EpCAM-positive CTCs and plasma-derived exosomes isolated from mCRPC patients was investigated in a comparative study [35].In addition, the expression of AR splice variants of mCRPC in CTCs and exosomes was also investigated [36].Many studies have shown that sequential biopsies provide important prognostic and predictive data based on PD-L1 status changes [57][58][59].We have already developed and clinically tested a molecular assay to detect PD-L1 transcripts in CTCs from HNSCC patients, which offers prognostic information [21].According to our results, almost half of the patients with mPCa were found to overexpress PD-L1.High PD-L1 expression of over 50% was also found in NSCLC CTCs during radiotherapy [60].
In addition, we have developed two novel RT-qPCR assays for the detection of mRNA expression of CXCR4 and JUNB, which are characterized by their analytical sensitivity and specificity.Molecular assays have the advantage of detecting gene expression at very low levels, as we have previously shown [36].A high prevalence of CXCR4 transcripts compared to JUNB transcripts was observed.This is the first time that a molecular assay has been developed and clinically evaluated for the detection of CXCR4 and JUNB transcripts in mPCa CTCs.
We compared the median differences between the analysis of mRNA expression of all the genes tested.We found no significance between the relative fold-change data of JUNB and CXCR4 and JUNB and PD-L1.Previous work in metastatic breast cancer has shown that CXCR4 and JUNB are linked in a common signaling pathway by bioinformatic analysis [29].Furthermore, JUNB can dock in the region of the PD-L1 promoter and thus enable transcription of the PD-L1 gene [61].In addition, when analyzing the relative foldchange data between PD-L1 and CXCR4, significant differences were found.The CXCR4-CXCL12 axis regulates the transmission of diverse downstream signaling pathways crucial for tumor cell survival, proliferation, and migration [62], and PD-1/PD-L1 controls immune tolerance [63].Recent data have shown that a combination approach of drugs targeting the function of CXCR4 and PD-L1 enhances the efficacy of anti-PD-L1 therapy in TNBC by preventing two important signaling pathways of tumor proliferation [64].
The very low frequency of CTCs among blood cells makes the Isolation and detection methods particularly challenging [40].We also performed a combined investigation in molecular and protein levels for the three analyzed common gene targets.Results indicate an increased detection rate of cytokeratin-positive patients through RT-qPCR compared to IF for the cohort of 25 patients (76% vs. 60%), probably attributed to the fact of an increased sensitivity of the molecular assay for CK-detection or due to the Ficoll density gradient's low-recovery yield of CTCs [65].However, previous studies in metastatic breast cancer have shown that positive detection of CTCs using either RT-qPCR or IF is correlated with a notably decreased median overall survival [52].According to our results, the agreement regarding CTC-positivity between these three different assays ranged from 48.0% to 68.0%.This observation aligns with previous studies highlighting the heterogeneity of CTCs, which contributes to the variability in results obtained from different isolation and detection methods.While these methods may not yield identical findings, they often provide complementary insights into CTC biology and characteristics [37].Nevertheless, the combined and overlapping results between the two different detection methods resulted in a higher percentage of positivity for the examined biomarkers-CXCR4, JUNB, and PDL1.
In summary, we investigated the expression of CXCR4, JUNB, and PD-L1 at the protein and molecular level in CTCs from mPCa patients.The protein-level study included EpCAMindependent isolation methods and two triple IF methods, while the molecular-level study included an EpCAM-dependent method and three individual RT-qPCR assays.The comprehensive analysis of CTCs using two different isolation and detection systems showed a high probability of positive CTC detection.Furthermore, the (CK+/CXCR4+/JUNB-) phenotype was associated with disease progression, and PD-L1 protein expression was also related to poorer OS for mPCa patients.This is a pilot study, as the patient number is limited; further studies with larger patient groups, encompassing different time points, could better clarify the above results.However, the study can provide insightful indications regarding the importance of these biomarkers for monitoring patient survival and possible therapeutic targets.Clinical trials for immunotherapies against the PD-1/PD-L1 axis have been tested in mPCa patients [3,66], while a combination of CXCR4-targeted therapy together with radiotherapy was more effective, inhibiting metastatic growth [67].As long as a combination of PD-L1-and CXCR4-targeted therapies have been tested for TNBC patients [64], a relevant treatment combination could be tested in prostate cancer patients to examine their therapeutic impact.
One limitation of this study is the relatively small sample size, as well as the variability in the number of patients analyzed using different methods, such as the Ficoll density gradient, ISET platform, and RT-qPCR.The reason for these differences was the limited available blood volume from some patients.Future studies should aim to enroll a larger number of patients for each isolation technique and to analyze CXCR4, JUNB, and PD-L1 expression in a larger cohort of patients, with long-term follow-up to gain deeper insights and a more comprehensive understanding of their clinical relevance.

Conclusions
The present study examined CXCR4, JUNB, and PD-L1 at molecular and protein levels in circulating tumor cells (CTCs) derived from patients with mPCa.These biomarkers are known to play significant roles in cancer progression and metastasis.Our results showed complimentary overexpression of these biomarkers in CTCs of prostate cancer at both mRNA and protein levels.Meanwhile, the single-cell analysis revealed that the presence of the (CK+/CXCR4+/JUNB-) and (CK+/PD-L1+/CD45) phenotypes were correlated with poorer PFS and OS, respectively, providing interesting biomarkers for this type of cancer.S1: Characteristics of patients enrolled in Ficoll density isolation (48 patients); Table S2: Characteristics of patients enrolled in mRNA analysis (25 patients); Table S3: Characteristics of patients enrolled in ISET isolation (17 patients); Table S4: CTCs per patient after Ficoll density gradient isolation and triple immunofluorescent staining CK/PD-L1/CD45; Table S5: Wilcoxon signed-rank test results for the identified phenotypes from triple immunofluorescence stainings; Table S6: CTCs per patients after ISET isolation and triple immunofluorescent staining CK/PD-L1/CD45; Table S7: CTCs per patient after Ficoll density gradient isolation and triple immunofluorescent staining CK/CXCR4/JUNB.