1. Introduction
Androgen deprivation therapy (ADT) combined with radiotherapy (RT) is a standard treatment for patients with localized high-risk prostate cancer (PCa) [
1]. However, many patients will eventually regress after therapy and develop metastatic disease [
2]. Therefore, there is an urgent need for a biomarker, which can reliably identify patients at high risk for recurrence and metastases [
2]. For those patients, intensification of treatment beyond standard combination of ADT with radiation may be required [
2].
All the options of intensification of therapy, including longer durations of ADT, combination with intense androgen blockade, chemotherapy or novel agents to target androgen activity through different pathways [
3], are associated with additional toxicity. Therefore, it is critically important to ascertain the subgroup of patients in need of that next step. While clinical factors such as prostate-specific antigen (PSA) values, T-category, and Gleason scores have traditionally been used to risk-stratify prostate cancer patients, their accuracy is low (25%–40%) to predict important end points such as progression or metastases after radiation and ADT [
4,
5]. Due to the lack of a better surrogate, PSA measurement continues to be the main method to monitor treatment response and recurrence after treatment for prostate cancer [
6]. However, this is fraught with difficulty as there is no reliable method to differentiate PSA produced by tumor vs. normal prostate tissue [
5].
Many studies have evaluated the use of circulating tumor cells (CTCs) as a biomarker to predict disease progression and survival in patients with metastatic, advanced, or even early-stage PCa, as well as an endpoint marker in clinical trials [
7,
8,
9,
10,
11,
12]. As CTCs are responsible for distant metastasis, their analysis could potentially provide information about treatment response [
7]. High CTC numbers are associated with aggressive disease, increased metastasis, and decreased time to relapse in men with castration-resistant and metastatic prostate cancer [
10,
12,
13,
14]. However, the value of CTCs detection in men with localized high-risk prostate cancer is unknown. Even though CTC collection from blood is a less-invasive method and can be used as a real-time liquid biopsy during regular follow-up [
7], there are many challenges for the use of CTCs as a prognostic and/or predictive biomarker. For example, the number of CTCs found in patient samples depends on the isolation method used, because of CTCs immunophenotype heterogeneity, CTCs derived from the same tumor can present different expression of epithelial markers, such as EpCAM [
15]. This difference might limit CTC detection by EpCAM-dependent technologies, like CellSearch [
15]. In addition, many apoptotic CTC cells are also isolated and analyzed which may not necessarily be representative of potential metastatic cells; and, finally, CTCs can be absent in some non-metastatic PCa patients [
15,
16]. For a better clinical use of CTCs in PCa as a prognostic and/or predictive biomarker, a combination of enrichment (isolation), detection (identification), and characterization strategies (such as molecular profile), are necessary to improve our ability to identify high-risk lethal prostate cancer in patients with clinically localized high-risk prostate cancers [
17,
18].
In previous studies, we have demonstrated the potential of single-cell analysis of CTCs, combining a filtration-based CTC isolation technology with prostate cancer cell-specific antibodies, followed by the use of 3D telomere profiling to identify PCa patient subgroups [
18,
19]. Telomere shortening is one of the earliest molecular genomic events in prostate cancer tumorigenesis and can generate genomic instability [
20]. The detection of shorter telomeres is associated with increased occurrence of lethal prostate cancer and decreased survival [
21]. Additionally, androgen receptor (AR) inactivation by knockdown, androgen deprivation, or treatment with bicalutamide in LNCaP cells (prostate cancer cell line) can induce telomere breaks and telomere fusion [
22]. However, telomere dysfunction was not observed following bicalutamide treatment in the AR-negative PC-3 prostate cell line [
23]. Clinical studies assessing the effects of ADT on telomeres using prostate cancer patient samples are limited. In 2017, Cheung et al. reported no evidence that ADT deprivation accelerates telomerase shortening in men who have been diagnosed with prostate cancer [
24]. However, leucocyte DNA was used for this analysis [
24].
In a previous work, we began to follow the early dynamics of CTCs using 3D telomere analysis during ADT (a pilot study composed of 20 patients, where consenting treatment-naïve patients with cT3, Gleason 8-10, or prostate-specific antigen > 20 ng/mL and non-metastatic prostate cancer were included) [
19]. We analyzed CTCs from high-risk prostate cancer patient´s samples at different time points: before ADT and RT (+0 month, untreated), after 2 months of ADT (+2 months) but prior RT, and 2 months after the final fraction of RT (+6 months). ADT begun 2 months before the start of RT and continued after RT was completed. At each time point, we enumerated CTCs from the blood, collected PSA values and investigated the nuclear 3D telomere architecture in CTCs derived from patients with non-metastatic high-risk prostate cancer before, post-ADT, and post-RT [
19]. Contrary to CTC enumeration and PSA serum levels, we showed that nuclear 3D telomere architectural analysis is highly sensitive in detecting cellular events that affect the genome stability in CTCs, and we described three distinct telomere signatures in CTCs. Our previous data also indicated that only one-third (6/20, 30%) of patients with non-metastatic high-risk prostate cancer may be able to fully benefit from a synergistic ADT/radiotherapy treatment. However, a 2-month post-RT time point cutoff was too early to conclude disease outcome in our previous study and for a complete assessment of the effects of ADT and RT on 3D telomere architecture of CTCs, those patients need to be followed up for longer period.
Therefore, in the current study, we assessed if the 3D CTC telomere profiles can predict response to treatment when compared with PSA (the standard evaluation). We are comparing different PSA end points as early surrogates for tumor response, such as six-months PSA end levels after ADT, six-month PSA end levels after RT, and twelve-months PSA end levels after completed ADT (+36 months). The cutoff values were chosen on the basis of previous reports in which PSA end levels above or equal to 0.1 ng/mL after radiotherapy and long-course androgen deprivation therapy are associated with an increased risk of recurrence [
6,
25,
26,
27].
3. Discussion
In most cancers, early detection allows for improved outcomes, but for localized high-risk prostate cancer patients, early detection can also result in overdiagnosis and overtreatment [
29]. In addition, localized high-risk prostate cancer patients face a more serious problem, a reliable prognostic tool capable of predicting whether the cancer will eventually develop into a lethal metastatic disease [
18]. Although PSA levels are used for disease monitoring, the predictive value of PSA testing and screening is low (around 35%) and have been associated with a high rate of overdiagnosis/overtreatment in clinical trials [
32,
33,
34]. In addition, other clinical parameters such as clinical stage and Gleason score tumor grade have limitations to detect and predict disease outcome [
35]. This scenario demonstrates the importance for novel and less-invasive biomarkers that can reliably identify patients at high risk for recurrence and metastases.
Telomere shortening is one of the earliest events in prostate cancer tumorigenesis and continue during tumor progression [
18]. Since detection of shorter telomeres is associated with increased occurrence of lethal prostate cancer and decreased survival times, the 3D telomere assessment could potentially improve prostate cancer screening by adding, to the current approaches, prognostic information to better stratify patients requiring active surveillance or more definitive treatment, such as surgical castration [
18,
31].
To our knowledge, the current study is the first to investigate the dynamics of the nuclear 3D telomere architecture in CTCs derived from patients with non-metastatic high-risk prostate cancer before, post-ADT, and post-RT (until 36 months after initial treatment). CTCs isolated before treatment was divided into five distinct telomere signatures. Remarkably, CTC analysis showed distinct dynamic changes in their 3D telomere signatures, which were unique to each group during ADT+RT treatment. Recent studies have provided insights to the clinical value of CTCs collected from blood in prostate cancer [
17,
36,
37,
38]. In the present study, we have used the ScreenCell filter device, which allows a size-based separation of CTCs from whole blood of patients with non-metastatic high-risk prostate cancer [
28]. Captured CTCs underwent 3D telomere analysis to determine their nuclear 3D telomere profile before treatment and to investigate dynamic changes to their 3D telomere profiles during and after treatment with ADT and RT. The CTCs can be detected in blood before the occurrence of clinically relevant metastases providing insight into the genetics of the primary prostate tumor [
39,
40,
41,
42,
43,
44,
45]. In this study, we show that nuclear 3D telomere architectural analysis is highly sensitive in detecting telomere changes that affect the genome stability in CTCs and may prove to be a valuable tool in monitoring treatment response in patients with non-metastatic high-risk prostate cancer.
As stated above, in a previous study, we showed the usefulness of PSA serum levels and CTC blood counts after treatment [
17]. In studies using the CellSearch® system to enrich and isolate CTC in localized high-risk prostate cancer, no correlation between CTC count and other clinical-pathological parameters was found [
46,
47,
48]. This highlights the importance of molecular analysis of CTC instead of only CTC count to provide additional information about the tumor. For example, AR-V7 expression in CTCs was found to be a predictor of response for treatment with abiraterone/enzalutamide and disease outcome [
46].
Here, we assessed the effects of ADT and RT on 3D telomere architecture of captured CTCs until 24 months. We found that the 100 high-risk patients could be stratified into five distinct telomere signatures based on telomere numbers and telomere length (intensity). Furthermore, each of the five CTC groups responded to the combined treatment with different changes to telomere profiles, thus, providing unique insight about the complexity by which high-risk prostate cancer cells can adapt to those treatments. We also showed in previous studies the potential of 3D telomere architecture in predicting disease outcome and patient survival [
16,
17,
47,
48,
49]. Based on our 3D telomere analysis, Group 2 may qualify as a non-responder, since the telomere profiles were stable throughout treatment. However, for Group 1, the treatment resulted in a dramatic decrease in the number of telomeres with an intensity less than 10,000 a.u. Intriguingly, this 3D telomere profile remained unaltered despite radiation-induced cell death. For Group 4, all patients presented a peak after ADT started, which decreased in later time points. Zhou et al. have already demonstrated that AR inactivation by androgen deprivation, in LNCaP cells (prostate cancer cell line) can induce telomere breaks and telomere fusion [
20]. The peak that we observed in +2m can be a consequence of the ADT treatment, which induce genomic instability and consequently death. Similar to Groups 1 and 4, the treatment for Groups 3 and 5 also had the number of short telomeres decreased. However, at later time points, the population with short telomeres started to increase reflecting a positive selective pressure of ADT plus RT in favor of resistant prostate cancer clones.
The ability of cancer cells to survive specific treatments, such as ADT and RT, involves changes in 3D telomere architecture and reflects the effects of complex cellular processes in which genomic stability, instead of causing death, ensure tumor cell survival. The effect of ADT plus RT on specific 3D telomere profiles may reflect the evolution of heterogeneous prostate tumor sub clones, as showed at later time points for Groups 3 and 5. The cellular mechanisms responsible for the dynamic telomere alterations in these patients are currently unknown. It is important to recognize that the observed heterogeneity in telomere phenotype was limited to five unique 3D telomere signatures in 100 localized high-risk patient samples. The effect of these profiles on patient survival awaits future analysis. However, we used PSA after 6 months of ADT, after 6 months of finished RT, and 36-months after initial treatment as an early surrogate for tumor response. In all time points, increase of nuclear volume, total number of signals, and total number of aggregates were significantly different between the two patient population (< 0.1 ng/mL and ≥ 0.1 ng/mL PSA end). The cutoff value was chosen on the basis of previous reports in which increasing levels of PSA above 0.1 ng/mL after ADT and radiotherapy was associated with an increased risk of recurrence [
6,
22,
23,
24,
25]. Additionally, the total intensity (associated with telomere length) decreased at 6 months of continued ADT and increased after 36 months after initial treatment. We attributed this to a process where the decrease of telomere length leads to a decrease of total intensity; however, as the formation of telomere aggregates (clusters of telomeres) continues, the resulting high intensity values of these clusters ensued an increase in the total intensity measurements in 36 months. It is important to highlight that we compared our data with the only approved biomarker guiding for treatment decisions in PCa [
7]. However, PSA values often do not represent the current tumor status, potentially misleading therapeutic decisions [
7].
Nevertheless, we found no association between the PSA end values and the 3D profile with CTCs dynamics over time (p = 0.38). Our data also predicted that only 32% (Groups 1 and 4) of the patients with non-metastatic high-risk prostate cancer could benefit from the combination of ADT/radiotherapy treatment. Additionally, 3D telomere analysis of circulating tumor cells offers a non-invasive method to follow up prostate patients during their treatment cycle.
The centroid cluster analysis identified three clusters, using all TeloviewTM parameters (number of telomeres, total intensity/length, telomere aggregates, and nuclear volume), which separated patients with different levels of genomic instability. Cluster 3 seems to correspond to somewhat more aggressive phenotypes than Cluster 2 and 1. We observed that in PSA 6 months after ADT, 33.33% of patients had PSA values above 0.1 ng/ml. This percentage decrease in the second time point (PSA 6 after RT), however, in the third time point (PSA after 36 months), cluster 3 is the only group that return to the same scenario found in the first time point (PSA 6 months after ADT), with 33.33% of the patients above 0.1 ng/ml (PSA). In the other clusters (1 and 2), the percentage of patients above 0.1 ng/ml after 36 months of treatment is lower than previous time points. In spite of a 3-years follow-up be too short to detect prostate cancer related mortality, our results demonstrate that CTCs retain important genetic information that could be used as a real time liquid biopsy to guide therapeutic decision and avoid overtreatment.
The current study has two important limitation. First, the 3-years follow-up was too short to detect prostate cancer related mortality, which affects the correlation of our data with a “real” clinical end-point. Second, our results were correlated with PSA end levels. Although PSA measurement after radiotherapy and androgen deprivation for localized prostate cancer has been proposed as an early prognostic biomarker [
50,
51]. PSA positive predictive value is only 25–40% [
52]. We found that our telomere parameters are significantly different between PSA groups, which highlights the potential of our biomarker to be equal or superior to PSA. However, the effect of our findings on patient survival still awaits future analysis.