Prostate Cancer Dormancy and Recurrence

Abstract

Prostate cancer can progress rapidly after diagnosis, but can also become undetectable after curative intent radiation or surgery, only to recur years or decades later. This capacity to lie dormant and recur long after a patient was thought to be cured, is relatively unique to prostate cancer, with estrogen receptor positive breast cancer being the other common and well-studied example. Most investigators agree that the bone marrow is an important site for dormant tumor cells, given the frequency of bone metastases and that multiple studies have reported disseminated tumor cells in patients with localized disease. However, while more difficult to study, lymph nodes and the prostate bed are likely to be important reservoirs as well. Dormant tumor cells may be truly quiescent and in the G0 phase of the cell cycle, which is commonly called cellular dormancy. However, tumor growth may also be held in check through a balance of proliferation and cell death (tumor mass dormancy). For induction of cellular dormancy, prostate cancer cells respond to signals from their microenvironment, including TGF-β2, BMP-7, GAS6, and Wnt-5a, which result in signals transduced in part through p38 MAPK and pluripotency associated transcription factors including SOX2 and NANOG, which likely affect the epi-genome through histone modification. Clinical use of adjuvant radiation or androgen deprivation has been modestly successful to prevent recurrence. With the rapid pace of discovery in this field, systemic adjuvant therapy is likely to continue to improve in the future.

NATURAL HISTORY OF PROSTATE CANCER DORMANCY AND RECURRENCE

Unlike most cancers, prostate cancer (PCa) can recur many years after the patients’ disease was thought to be cured by prostate radiation or surgery. In the largest case series of PCa recurrence after radical prostatectomy surgery (1997 patients total, 304 with recurrence), in 25% of patients, the initial sign of recurrence occurred ≥ 5 years after surgery¹. Also, unlike most other cancers, PCa has a sensitive and specific biomarker for recurrent disease – the prostate specific antigen (PSA). Therefore, most recurrences of PCa are initially PSA only – or biochemical recurrence (BCR) and do not represent enough cancer to be detectable on physical exam or imaging, or to cause symptoms. Therefore, the time to a recurrence capable of causing symptoms is even longer. In the same large series, though 25% of patients with recurrence had their PSA rise ≥ 5 years after surgery, it took an additional median of eight more years for a clinically meaningful metastatic recurrence². These late recurrences were not merely indolent cancers, but also resulted in patient deaths, though at a slightly lower frequency than more rapid recurrences. Although characteristic of PCa, late recurrences occur in other diseases as well. In a meta-analysis of nearly 63,000 patients with localized hormone receptor positive breast cancer, there was no decrease in the risk of recurrence out to 20 years². Curiously, although hormone receptor negative breast cancers have a high risk of early recurrence and death the recurrence risk subsequently abates and is lower than the recurrence risk for hormone receptor positive cancers at ten years or later³. The observations of continued risk of recurrence in some cancers are consistent with a model where cancer cells spread distantly early in the disease process, some of which later reactivate, start growing and lead to metastatic disease. Researchers subsequently began to try to understand this process and have subsequently gained significant insight into the cellular and molecular mechanisms. Below, we summarize the current state of knowledge of PCa dormancy and recurrence, describe how this knowledge has been applied to treatment thus far, and discuss future areas for study.

RESERVOIRS FOR LATER RECURRENCE

The spread of cancer cells while a prostate tumor is still small could occur through invasion into the surrounding stroma, lymphatic spread, hematogenous spread, or perhaps more likely a combination of these processes. For example, a cancer cell could invade the tumor stroma, migrate to a lymph vessel, then enter the blood stream. It is difficult or impossible to determine the anatomic location where most cancer cells lie dormant prior to recurrence experimentally in patients. However, it seems reasonable to presume that the most important “reservoirs of dormancy” are where recurrences are later found clinically. In an autopsy series of 631 patients with stage IV prostate cancer, 66% of patients had involved lymph nodes and 90% of patients had bone metastases. Lung and liver metastases were next most common, with 46% and 25% respectively⁴. Similarly, in a study of biochemically recurrent PCa patients evaluated with Choline C¹¹ PET/CT scans, disease was found most commonly in lymph nodes, bones and the prostate bed⁵. This population of patients with biochemical recurrence (elevated prostate specific antigen but no detectable disease on physical exam and most imaging tests) can sometimes have a low level of recurrent PCa visualized on Choline C¹¹ PET/CT or other “molecular imaging” tests. Lesions detected on these scans represent the initial sites of relapse and are presumably where previously dormant PCa cells exited quiescence and began to grow – unless they migrated to an additional anatomic site while still in a dormant state.

Therefore, the prostate bed, lymph nodes, and bones appear to be the most important sites for PCa to lie dormant and undetectable after initial treatment of the primary tumor with radiation or surgery. Of these sites, only bone has been the subject of patient samples research into the detection and biology of rare disseminated PCa cells in patients without detectable disease outside their prostate tumors. Presumably, this is because bone marrow aspirates can be obtained with appropriate patient informed consent, but research samples of lymph nodes or prostate bed tissue are very difficult to obtain. Evidence for the presence of these cells, termed disseminated tumor cells (DTCs) or disseminated cancer cells (DCCs) has been found in bone marrow aspirates of patients with localized PCa in multiple reports over the past three decades in frequencies ranging from about 30% to 70% of patients^6–11. Techniques included RT-PCR for the KLK3 (PSA) gene¹², immunocytochemistry¹³, immune-magnetic beads and single cell isolation⁶, and fluorescence activated cell sorting (FACS)¹⁴

CELL BIOLOGY AND GENOMICS

Investigators have found evidence for several non-mutually exclusive cellular processes to explain the clinical observations that PCa and other cancers can remain undetected for many years after curative intent therapy but then relapse with fatal consequences^15,16. Most investigators refer to the processes which keep DTCs from growing into clinically detectable tumors, yet fail to totally eradicate the cancer as “dormancy maintenance.” Conversely, the processes that convert DTCs from rare and asymptomatic to clinically detectable and problematic are referred to as “dormancy escape.” Conceivably, the phenomenon of dormancy could be accomplished through one or more of several patterns of cell growth or lack thereof, or interaction with the microenvironment. DTCs might divide very slowly or not at all – termed “cellular dormancy.” This lack of cell division is thought to be the result of interactions with the microenvironment and may be accompanied by reversible epigenetic reprogramming of the DTCs^11,16. Alternatively, dormancy might result from the balance of cell division and cell death, so that very little change in the total number of DTCs occurs – termed “tumor mass dormancy¹².” In addition, dormancy might result from “angiogenic dormancy” where DTC or micrometastasis growth is held in check by lack of a blood supply, or “immunologic dormancy” where the immune system prevents DTC or micrometastasis outgrowth¹².

In PCa and most other cancers, “cellular dormancy” has been the most studied mode of dormancy maintenance. Curiously, many of the initial observations of cellular dormancy were not in PCa, but in head and neck squamous cell carcinoma. Aguirre-Ghiso and colleagues developed head and neck squamous cell carcinoma sub-lines from a growth permissive site (lung) or an inhibitory / dormant site (bone marrow). After isolation from the mice, these sub-lines retained the expected growth patterns for weeks to months; i.e. the lung sub-lines grew much more rapidly than the bone marrow derived sub-lines. This maintenance of phenotype after removal from the microenvironment suggested that there is an important epigenetic component to dormancy regulation. With these dormant vs. aggressive sub-lines, they were able to interrogate the signaling pathways and epigenetic modifications,^17,18 including the importance of p38 vs. Erk MAP kinase signaling, which we discuss below.

One labor intensive study built on the concept of cellular dormancy and found evidence that at least some DTCs in patient samples cycle slowly or not at all. Chery et al isolated single DTCs from PCa patients with either “no evidence of disease (NED)”; defined as no sign of recurrence after prostatectomy, or “advanced disease (ADV)”; defined as metastases at the time of prostatectomy or PSA recurrence after prostatectomy or radiation, and then analyzed the cells by microarrays¹⁹. After excluding cells thought to be contaminating normal marrow cells using an erythroid gene signature, they analyzed 44 cells from 10 patients. They compared cells from NED vs. ADV patients to discover difference between dormant and active cells. However, they subsequently found that about half of the cells from ADV patients had gene expression that clustered with NED patients, whereas the other half of cells from ADV patients did not. This data supports a hypothesis that most of the DTCs in patients with dormant disease are quiescent, whereas patients with active disease have some cells that are quiescent, but others that are actively cycling. In a different study of microarray analysis of isolated DTCs some of which had their DNA analyzed by comparative genome hybridization, Guzvic et al observed an unexpectedly high number of transcripts for hematopoietic genes and hypothesized that DTCs had a very plastic phenotype – akin to an extreme form of epithelial mesenchymal transition²⁰. The study of DTC dormancy is difficult to adapt to model systems. Therefore, the literature directly examining the cell cycle status of PCa DTCs in in vivo model systems is limited. Owen et al used a label retention based approach and single cell RNA sequencing of bone marrow DTCs isolated from an animal model and found evidence for the importance of another mode of dormancy maintenance; immunologic dormancy²¹. In the current review, we concentrate on recent advances. Other authors have reviewed literature specific to PCa prior to the past five years – which provide complimentary discussions to the present work^22,23.

A factor potentially affecting PCa dormancy particularly difficult to study in model systems is the impact of genomics (DNA sequence changes). Most investigators would agree that the time period for model systems is too short to study genomic changes in dormancy regulation, though innovative investigators might solve this problem in the future. Using genomic hybridization techniques, Holcomb et al investigated large genomic changes (megabase and chromosomal level) in DTCs from metastatic vs. non-metastatic patients. They observed changes characteristic of PCa including 8p loss, 8q amplification and amplification of Xq, which contains the androgen receptor gene. They observed more genomic changes in metastatic than non-metastatic patients, which is consistent with a hypothesis that acquired genomic changes in DTCs could be important for metastatic recurrence. To again borrow from another malignancy, Werner-Klein and colleagues sequenced tumor cells from melanoma tumor cells and DTCs isolated from the sentinel lymph nodes. They found more differences between primary tumors and DTCs than expected. These included differential presence of typical driver mutations such as the BRAF gene, which are usually thought to be clonal²⁴. They concluded that it was very likely that DTCs were spreading to the draining lymph nodes when the primary tumor was very small (0.5 mm thick) and continuing to acquire genomic changes at the lymph node site.

MICROENVIRONMENT AND INTRACELLULAR SIGNALING

In a similar fashion as the work with DTCs from patient samples, the majority of the research into PCa dormancy regulation has focused on the bone marrow microenvironment, though other sites are very likely to be important as well. Presumably, this is likely due to practical considerations; i.e. the difficulty of studying lymph nodes in a model organism the size of a mouse. The bone marrow is a complex organ with multiple disparate cell types including; precursors and mature hematopoietic cells (myeloid, lymphoid, erythroid, and thrombocytic), fat, sinusoids, arterioles, nerves, osteoblasts, and osteoclasts. Functionally, Shiozawa et al found that PCa DTCs appear to share the same environment, or niche, as hematopoietic stem cells by showing that PCa cells displace hematopoietic stem cells from the bone marrow and vice versa²⁵. However, the precise microanatomic location of this niche is poorly defined. The components of the hematopoietic stem cell niche have been hotly debated, with the likely conclusion that many different cells can function as niche components²⁶.

Given that DTCs are integrated into the bone microenvironment, it may not be surprising that they are regulated by protein ligands that are also important in other bone and bone marrow processes, most notably proteins of the tumor growth factor beta (TGFβ) and bone morphogenetic protein (BMP) families. Kobayashi and colleagues reported that BMP-7 signals through BMPR2 to keep PCa DTCs dormant. They focused their analyses on a sub-population purported to have cancer stem cell like characteristics. They found that BMP-7 / BMPR2 signaling regulated the quiescence marker, senescence associated β-galactosidase, in vitro and lengthened metastasis free survival using xenograft animal models. They also showed that BMPR2 inversely correlated with metastasis free survival in patient specimens²⁷. Subsequently, Sharma et al found that secreted protein acidic and rich in cysteine (SPARC) stimulated BMP-7 expression in bone marrow to maintain dormancy²⁸. TGF-β2 signaling was first shown to maintain dormancy in in other malignancies¹⁷. Then, Yumoto and colleagues showed that TGF-β2 also acted to maintain dormancy in PCa and that it required AXL, a tyrosine kinase receptor from another family²⁹. AXL is a member of the TAM family of receptor tyrosine kinases, named for the three members; TYRO3, AXL, and MERTK. The TAM family has multiple protein ligands with varying affinities for the three receptors, with the most studied ligands being GAS6 and Protein S (PROS1). GAS6 signaling through AXL promotes PCa cell survival and causes a G1/G0 arrest during docetaxel treatment^30,31. In agreement with this, Taichman et al, profiled the cell surface expression of AXL and TYRO3 in subcutaneous tumors, bone marrow DTCs, and bone metastatic tumors. They found that AXL was expressed more highly on DTCs, whereas TYRO3 was expressed more highly on macroscopic tumors at either the subcutaneous or metastatic sites. This is consistent with a model where AXL signaling regulates dormancy maintenance and TYRO3 regulates dormancy escape. However, they did not examine MERTK in this study and did not perform functional studies³². Subsequently, we showed that MERTK knockdown inhibited cell cycle progression in vitro and lengthened metastasis free survival after intra-cardiac left ventricle injection in vivo³³. Therefore, AXL seems to promote dormancy maintenance and TYRO3 and MERTK promote dormancy escape. In related studies, norepinephrine was shown to stimulate PCa dormancy escape directly by signaling through the β2-adrenergic receptor and indirectly by decreasing GAS6 expression in bone marrow stroma³⁴. More recently, WNT5A was shown to maintain PCa dormancy. Ren and colleagues found that WNT5A induced Siah E3 Ubiquitin Protein Ligase 2 (SIAH2), which repressed β-catenin signaling to maintain dormancy³⁵. This study showed an unexpected signaling pathway because WNT5A, typically a ligand for non-canonical (β-catenin independent) WNT signaling, acted indirectly to affect canonical (β-catenin dependent) WNT signaling.

However, with regard to downstream intracellular signaling, the bulk of the literature points to MAP kinases rather than β-catenin. Largely in head and neck squamous cell carcinoma model systems, p38 MAPK (MAPK14 gene) was found to be a key node regulating dormancy related intracellular signaling. A key component of this signaling module is the relative abundance of phosphorylated (active) p38 vs. phosphorylated ERK1/2 kinases (MAPK3 and MAPK1 genes). In this model, phosphorylated p38 favors dormancy maintenance, while phosphorylated ERK1/2 favors dormancy escape, proliferation and relapse^17,18,36. In line with these studies, Ingenuity Pathway Analysis predicted p38 to be a key node regulating a quiescent vs proliferative phenotype in bone marrow DTCs isolated from PCa patients. In this study, Chery et al isolated single DTCs from either patients with active PCa or patients with resected localized PCa and no evidence of recurrence for at least seven years. They then amplified the cDNA and performed microarray analysis on the single cell samples. By comparing gene expression in patients with inactive vs. active PCa, they derived a 21-gene signature that converged on p38 by Ingenuity Pathway Analysis¹⁹. The balance of phosphorylated p38 vs phosphorylated ERK1/2 or p38 alone has been shown to be important in model system studies of PCa dormancy as well, downstream of multiple ligands and receptors including; TGF- β2, GAS6 / AXL, MERTK, BMP-7 / BMPR2, SPARC and norepinephrine / β2-adrenergic receptor^{27–29,33,34}. Recently, another intracellular pathway regulating dormancy was discovered; PPARγ signaling through mTOR to maintain dormancy. Curiously, this was stimulated by the plant phytohormone abscisic acid³⁷. Continuing further downstream to the nucleus, less is known about how transcription factors and epigenome modifying enzymes regulate PCa dormancy. In other malignancies, the embryonic stem cell transcription factors SOX2 and NANOG, and the orphan transcription factor NR2F1 appear to be important for dormancy maintenance, at least in part through histone H3 post-translational modifications^17,18. This discovery prompted study of the potential use of all trans retinoic acid and 5-azacitadine as a treatment approach to revert cycling PCa cells to a dormant phenotype in and ongoing clinical trial discussed more below. In keeping with these findings, we showed that MERTK knockdown delayed dormancy escape and induced higher levels of NR2F1, SOX2 and NANOG and induced higher levels of the inactivating histone H3 marks; trimethylated lysine 9 and lysine 27³³.

Much of the work discussed above focuses on “cellular dormancy,” i.e. the lack of cell division of DTCs induced by paracrine signaling in the microenvironment, often accompanied by epigenetic reprogramming. There has been comparatively little work on the role of the immune system (immunologic dormancy) on PCa dormancy maintenance and escape. This may be due partly to the fact than many of these studies were conducted using immune compromised animals. A recent study used single cell RNA sequencing to compare slowly vs. rapidly cycling DTCs in an immune competent mouse PCa model and discovered a novel mechanism of PCa dormancy maintenance involving both the immune system and epigenetic reprogramming²¹. They found that tumor cell intrinsic (as opposed to microenvironmental) type 1 interferon was critical for PCa dormany maintenance and observed the expected effects on the immune system and reversal of dormancy by administration of systemic interferon. They also saw reversal of the effect with a histone deacetylase inhibitor, which highlights a potential interaction between immune regulated and epigenetic regulated dormancy. It seems likely that investigators will continue to unearth interactions between different dormancy regulatory mechanisms in the near future. Please see Figure 1 for an illustration of the key molecular mechanisms regulating PCa dormancy.

Figure 1.

Illustration of key molecular mechanisms regulating prostate cancer dormancy.

ADJUVANT TREATMENT TO PREVENT RECURRENCE

A major practical application of cancer dormancy research is to design adjuvant therapies – to prevent recurrence in patients with treated primary tumors and no current clinically detectable cancer. In PCa there has been some progress in adapting existing therapies to the adjuvant setting, though no molecular targeted therapies have been approved in this setting thus far. As an illustration that the prostate bed and lymph nodes can serve as a reservoir of residual disease, adjuvant radiation therapy after radical prostatectomy can increase event free survival. However, early salvage radiation (soon after initial PSA relapse) appears to be as effective as salvage radiation³⁸. Indefinite androgen deprivation therapy (ADT) with medical castration improves survival if given immediately to patients with involved lymph nodes discovered at the time of radical prostatectomy³⁹. However, this therapy has significant side effects, especially if continued life long. Similarly, adjuvant ADT is beneficial in localized PCa patients treated with definitive radiation rather than surgery. While ADT is often given concurrently with radiation to sensitize the cancer cells to the effects of radiation, if continued long term, ADT also acts as an adjuvant therapy in addition to its effect as a radiosensitizer. Three years of ADT (radiosensitizer and adjuvant) given with prostate radiation results in longer overall survival than only six months (radiosensitizer only) of ADT⁴⁰. In a later trial, an intermediate period of radiation (18 months) was shown to have superior prostate cancer specific mortality to six months of ADT when given with radiation⁴¹.

However, clinical trials of PCa systemic adjuvant therapy other than ADT with medical castration have met with much less success. The first randomized trial of a chemotherapy drug for prostate cancer adjuvant therapy after curative intent radical prostatectomy randomized patients to ADT or ADT plus mitoxantrone – a drug much less commonly used today as when the first patients were enrolled to the trial in 1999⁴². Unfortunately, the trial needed to be stopped early by the data safety monitoring board because of concern for an increased number of cases of acute myeloid leukemia in the mitoxantrone group. In longer-term follow-up, the monitoring board’s concerns appear justified. At a median follow-up of 11.2 years, there was no improvement in overall survival for the mitoxantrone group, but there was an increased risk of death from other cancers. A later trial of a contemporary chemotherapy drug, docetaxel, for PCa adjuvant therapy was also unsuccessful. The trial, named TAX-3501, was terminated early when the pharmaceutical company sponsor withdrew support – citing enrollment below expectations⁴³. Most recently, the Radiation Therapy Oncology Group (RTOG) sponsored a trial to test if addition of docetaxel improved survival for high risk localized PCa when added to definitive radiation with concurrent ADT. Initially there appeared to be an overall survival advantage with addition of docetaxel, prompting listing this approach as an option in the United States National Comprehensive Cancer Network (NCCN) guidelines. However, the survival curves converged with longer follow-up^44,45. Therefore, with the modest success of PCa systemic adjuvant therapy, new agents and approaches are needed.

One such approach is an ongoing randomized crossover pilot trial of the addition of all trans retinoic acid and 5-azacitadine to ADT in patients with PCa biochemical recurrence⁴⁶. In biochemical recurrence, patients initially have no clinical evidence of disease after treatment of localized PCa with surgery or radiation, but then have an increase in prostate specific antigen, but do not have enough disease to be detected by physical exam or imaging. These cases of biochemical recurrence are presumably from reactivation of residual PCa in the prostate bed or DTCs in metastatic sites such as bones or lymph nodes. The purpose of addition of 5-azacitadine and all trans retinoic acid to ADT is to restore the reactivated residual cancer to a dormant state rather than to necessarily eradicate it completely. This approach was largely based on work highlighting the importance of the transcription factor NR2F1 in maintaining cancer dormancy¹⁸. In this study, NR2F1 was induced by all trans retinoic acid and by histone H3 post-translational modifications. In their treatment model, all trans retinoic acid was sufficient to convert active cells to a dormant state temporarily, but addition of 5-azacitadine was required for a lasting effect. We anxiously await results of this clinical trial, which are expected in August, 2022 and expect that this will be just the first of many trials of targeted agents for PCa adjuvant therapy.

CONCLUSIONS

Investigators have made great progress in understanding how PCa can disseminate early in the disease course and relapse months or decades later. Consistent with bone as a common site of PCa metastasis, researchers have found evidence of disseminated tumor cells (DTCs) in bone marrow of many, and in some cases the majority, of patients with clinically localized PCa. Laboratory researchers, predominantly in model systems, have gained a significant amount of insight in to the molecular mechanisms controlling PCa dormancy. TGF-β2, GAS6, BMP-7, and WNT5A are extracellular molecules that maintain dormancy, whereas adrenergic signaling signaling stimulates dormancy escape. Dormant DTCs also have increased histone H3 post-translational modifications and have increased activity of the transcription factors; SOX2, SOX9, NANOG,and NR2F1. A recent work found importance for tumor intrinsic type 1 interferon as maintaining dormancy and lack of type 1 interferon for stimulating dormancy escape. This work brings up the importance of immunologic dormancy in addition to cellular dormancy, which was the subject of most of the other work. Although clinical researchers have had modest gains thus far, much more work is needed to refine the basic science and translate this knowledge to the clinic.

HIGHLIGHTS.

• Prostate cancers can lie dormant for many years after treatment

• Likely sites of dormancy are the prostate bed, lymph nodes and bone marrow

• Multiple investigators have found disseminated prostate cancer cells in the bone marrow of patients with localized disease

• Key molecular regulators include; BMP-7, TGF-β2, GAS6, WNT5A, p38 MAPK, SOX2 and NANOG

• There has been clinical progress in adjuvant therapy using androgen deprivation and radiation, but more laboratory and clinical research is needed