Prostate cancer (PCa) is one of the most commonly diagnosed cancer types among men in the industrialized countries [1
]. Early diagnosis has increased since the introduction of serum prostate-specific antigen (PSA) screening. Nevertheless, determination of the risk of recurrence is still incomplete and further research is required to better understand the metastatic PCa cascade. Biochemical recurrence (BR) is still assumed to be the earliest indicator of patient relapse. It has been estimated that approximately 35% of patients manifest BR within 10 years after surgery and overt metastases approximately 8 years after BR [2
]. Further research is needed to identify markers relevant for individualized prognostication and tailoring of therapy.
(Neo)vascularization is considered as one of the putative factors influencing tumor development and in consequence impacting patients’ outcome in different solid tumors including PCa [3
]. Rich vascularity might guarantee appropriate nutrition of tumor cells and putatively facilitate their dissemination. On the other hand, a low number of vascular vessels may result in poor oxygenation, i.e., hypoxia. In PCa, hypoxia occurs even at early stages of disease [9
] and is associated with poor prognosis due to the selection of the most resistant clone(s) of tumor cells [9
]. It was shown to induce and/or maintain metastases-promoting processes such as epithelial-to-mesenchymal transition (EMT), stemness, cell survival, and proliferation [9
To date, the relationship between vascular (VV) and/or lymphatic vessels (LV), and patient outcome in PCa has not been fully clarified. Previous studies showed variable outcomes (i.e., correlations to clinico-pathological parameters and patient survival or lack of) depending on the study design and applied methodology (including origin and type of PCa specimens, type of targeted proteins to detect endothelial cells, area of analysis, type of treatment, etc.) [11
]. Erbesdobler et al. performed a study on the highest (to the best of our knowledge) number of prostatectomy specimens (n
= 3261), showing the correlation between higher VV and disease aggressiveness as well as progression [12
]. As far as we know, data on low vascularity potentially inducing hypoxia are not present in the literature (or do not reach statistical significance) and different therapeutic regimen are scarcely described [13
], and androgen deprivation therapy (ADT) might potentially impact angiogenesis [14
Therefore, in the current study the association between vasculature and the aggressive PCa phenotype and disease progression was investigated in unselected PCa patients, hormone-naïve patients, and those treated with neoadjuvant ADT. CD34 and podoplanin, two proteins commonly used for detection of VV and LV, respectively, were assessed in order to examine the number of vessels, as well as to examine their heterogeneity and elucidate putative clinical relevance. Of note, in contrast to the majority of studies, we also considered minimal vessel numbers and their possible role in tumor progression or the relationship to clinical parameters.
Tumor (neo)vasculature might play a crucial role in tumor development and progression. Here, we show for the first time that low vasculature correlates to worse clinical outcome in hormone-naïve PCa patients after radical prostatectomy and is associated with aggressive phenotype of tumor cells.
The number of vascular and lymphatic vessels is considered as a significant microenvironmental factor impacting tumor fate in different solid tumors [4
]. However, inconsistent study designs and applied methodologies (i.e., different origin, type and area of the analyzed PCa specimens, type of targeted proteins to detect endothelial cells, etc.) result in different outcomes showing both correlations and their lack between number of vessels and clinico-pathological parameters in PCa [11
]. In the current study, two commonly used proteins, CD34 and podoplanin, were examined to define the numbers of VV and LV, respectively. Our evaluation method was similar to commonly used so called Weidner method [20
]. However, the applied study approach allowed for an examination of the clinical significance of both minimal and maximal numbers of vessels detected within the defined fragments of tumors (area equal 0.28 mm2
) without an a priori assumption that only high vasculature might support tumor progression. Of note, tissue microarrays (TMAs) used in this study represented a collection of potentially different tumor samples of individual patients prepared in order to study PCa heterogeneity. In addition, to exclude bias caused potentially by ADT [14
], clinical relevance of vasculature was compared in the unselected cohort, as well as hormone-naïve PCa patients and those treated with ADT preoperatively.
VVs were detected in almost all samples, whereas LVs were found in only 27% of tumor samples. Intratumoral heterogeneity of VV and LV numbers was observed in approximately one-third of the patients, which substantiates the high heterogeneity described in PCa for many factors. VVlow
and LV were found in 32% and 43% of patients, respectively. Only minVVlow
indicating the lowest numbers of vascular vessels (i.e., the lower quartile of minimal values assigned for a patient) correlated to worse clinical outcome in hormone-naïve PCa patients in the timeframe of 3–10 years after surgery. Of note, in the current study minVVlow
was an independent prognostic factor in the multivariate analysis, substantiating its prognostic potential in patients without ADT. We did not observe any correlation to clinical outcome in the cohort of unselected PCa patients or for maxVV and LV, nor for minVV in the PCa patients undergoing ADT before the prostatectomy. However, this cohort was relatively small (n
= 67) which might have biased the outcome. Those results seem to be counterintuitive as the other groups showed that the higher number of VV [12
] and also the presence of LV may increase the risk of BR in the patients undergoing radical prostatectomy after neoadjuvant treatment with androgen blockage [24
], correlating with high Gleason score and other clinico-pathological parameters indicating advanced disease [12
]. In particular, Erbersdobler et al. showed the clinical significance of high number of vessels in, to the best of our knowledge, the largest study on 3261 prostatectomy specimens prepared as TMA [12
]. However, they did not confirm this observation in the multivariate analysis [12
] and Mucci et al. did not observe a correlation between high number of VV and cancer-specific mortality [26
]. Of note, a high number of VV evaluated using CD105, another protein detecting endothelial cells, was also identified as a significant and independent predictor of biochemical recurrence in prostate cancer patients who underwent radical prostatectomy with ADT [13
]. Our results and those of the literature could suggest that vascularization develops differentially under different androgen conditions, regulating tumor development and patient outcome in a different way. In addition, the majority of the studies focused on the so-called “hot spot” with the highest number of VV. This approach might be also biased [28
] and result in exclusion of fragments of tumor with low number of vessels, potentially still crucial for progression but driven by different biological mechanisms.
Indeed, in our study tumors characterized by VVlow
exerted some features of more a aggressive phenotype of tumor cells. They were characterized more frequently by the absence of the epithelial cell marker EpCAM [29
], which might suggest that those tumors undergo EMT known to facilitate migration, and even induce stemness [30
]. They also more frequently expressed higher levels of Loxl-2, a protein related to invasion of tumor cells on extracellular matrix, and in comparison to highly-vascularized tumors this was also the case for proliferation marker Ki-67 and the apoptosis marker. All those proteins are known to correlate to poorer clinical outcome both in PCa and/or other tumor entities [31
]. Of note, they might be expressed under hypoxic conditions expected to occur in less vascularized samples of tumors [32
]. Hypoxia-related markers (such as hypoxia-inducible factor 1-alpha, HIF-1α) were not investigated in this study. However, it might be still speculated that hypoxia occurred in the examined tumors with low numbers of VV, potentially promoting a proliferation/apoptosis imbalance as well as induction of EMT [34
] and acquisition of more aggressive phenotype by tumor cells [30
]. In concordance, Loxl-2 was shown to be upregulated by HIF-1α in a hypoxic tumor microenvironment in hepatocellular cancer [35
], and to induce EMT in colorectal and breast carcinomas [36
4. Material and Methods
4.1. Patient Cohort
Tumor samples from 400 PCa patients were collected following radical retropubic prostatectomy at the Department of Urology at the University Clinic Münster (Germany) after patients gave informed consent, and were prepared in duplex as 2 TMAs as described previously [15
]. Briefly, two tumor samples were selected for each patient from different areas of the tumor (if the tumor was monofocal) or different tumor foci (if the tumor was multifocal). Patients were characterized by different clinico-pathological parameters (Table 2
) and different molecular phenotypes of tumor cells (Supplementary Table S4
). Measurement of serum PSA concentration was performed by two methods: (1) up to 2009 using Tandem-E (Hybritech, San Diego, CA, USA) and (2) from 2010 using Access 2 (Hybritech-calibrated, Beckman Coulter, Brea, CA, USA). BR was defined as two consecutive concentrations of PSA above 0.1 ng/mL. The time point of BR was defined as the first PSA concentration above 0.1 ng/mL. The last follow-up was performed in June 2019 (mean time to BR was 60 months, range 0–201 months).
Sixty-nine of the patients received neoadjuvant ADT (i.e., before surgery), the remaining 331 patients had no hormonal manipulations prior to BR and were termed hormone-naïve. Those two subcohorts were considered separately for statistical analysis. The study was approved by the local Ethics Committee (Ethik Kommission der Aerztekammer Westfalen-Lippe und der Medizinischen Fakultaet der Westfaelischen Wilhelms-Universitaet Muenster, Germany, nr 2007–467–f–S).
4.2. Immunohistochemistry for CD34 and Podoplanin
Immunohistochemical staining was performed on TMA sections (4–5 µm thick) using commercially available ready-to-use mouse monoclonal anti-CD34 antibody (clone QBEnd10, Agilent Dako, Santa Clara, CA, USA) for VV and anti-podoplanin antibody (clone REF 760–4395, Roche, Switzerland) for LV visualized by EnVision FLEX+ system (Dako) and UltraView DAB Benchmark XT (Roche) system, respectively.
4.3. The Evaluation of VV and LV Density
The immunohistochemical staining was evaluated at the 200× magnification using light microscope (Olympus BX 43, Olympus, Japan). The total number of CD34- or podoplanin-positive VV or LV, respectively, with visible light of lumen was documented in each informative tumor sample (0.6 mm diameter, area 0.28 mm2).
In order to examine the possible variability of numbers of vessels, vessels were counted separately in two tumor samples of each patient. The tumor sample with the lower number of vessels was assigned minVV or minLV, the sample with the higher number of vessels was considered as maxVV or maxLV (Figure S1
). If only one tumor sample was informative, the vessel count of that tumor sample was assigned to the patient. The assigned discrete values for minVV, maxVV, minLV, and maxLV were dichotomized based on different mathematical cut-offs (i.e., mean, median, quartiles) and compared to clinical data and patient survival (Figure S1
4.4. Immunohistochemistry for Tumor Cell Markers
Immunohistochemical staining for Ki-67, apoptosis marker (ApopTag), cytokeratins (CK5/6, CK14, CK8/18, CK19), vimentin, E- and N-cadherin, ALDH1, EGFR, EpCAM, and Loxl-2 protein was performed, evaluated, and categorized as negative or positive (and for Loxl-2 as negative, weakly positive, or positive) as described ([16
], brief description in Supplementary Table S5
Immunohistochemical staining for Bcl-2 (dilution 1:250; clone124, DAKO, Denmark) was performed using Autostainer (Dako), and alkaline phosphatase detection kit (Universal LSAB™ Kit/HRP, Rabbit/Mouse/Goat, Dako, Denmark). Staining was categorized based on the intensity as negative (i.e., no or weak expression), or positive (i.e., moderate to strong expression).
4.5. Statistical Analysis
Statistical analysis was performed using SPSS software (IBM) version 25 licensed to the University of Gdańsk. The comparison of number of vessels and clinical data or proteins detected in tumor cells was performed using Chi-squared or Fisher’s exact tests. The association between number of vessels and time to BR was calculated using Mantel Cox test and presented using Kaplan–Meier plots. The uni- and multivariate analysis was performed using the Cox-Hazard-Potential regression model (95% CI). The obtained results were considered statistically significant at p < 0.05. Cases with missing data were excluded from the statistical analysis.