Several restrictive signals have been described in the bone marrow (BM) and lung that maintain DTCs originating from breast, prostate, head and neck squamous carcinoma and multiple myeloma cells in their quiescent state ( Figure 1 ). Therefore, the BM is seen as a sanctuary site for DTCs. Growth arrest-specific 6 (GAS6) (51), Wnt5α (52), leukemia inhibitory factor (LIF) (53) and TGF-β family members such as bone morphogenic protein 7 (BMP7) (54) and transforming growth factor beta 2 (TGFβ2) (55, 56) were shown to exert quiescence of DTCs at the BM niche, whereas, bone morphogenic protein 4 (BMP4) was shown to promote tumor dormancy in the lung (57). Hence, theoretically we can postulate that DTCs can be maintained quiescent for prolonged periods of time by introducing the restrictive mediators that constitute the dormant niche (42).

SPARC, also known as osteonectin, was shown to regulate tumor dormancy of prostate cancer cells by promoting the expression of BMP7 in BM stromal cells. SPARC was shown to be epigenetically silenced in aggressive cells by promoter methylation whereas treatment of prostate cancer cells with the DNA demethylating agent 5-azacytidine (5-AZA) or with the COX2 inhibitor NS398 could restore SPARC expression in malignant prostate cancer cells. This in turn promoted BMP7 expression in BM stromal cells leading to dormancy of prostate cancer cells (58). Hence, reinstating SPARC expression in bone DTCs by treatment with either 5-AZA or with COX2 inhibitor NS398 may offer a therapeutic window to treat recurrent prostate cancer disease ( Figure 1 ).

Another component that may be reinforced is thrombospondin-1 (TSP1), which is found in the suppressive BM and the lung niches. TSP1 secreted by stable microvasculature or by recruited BM-derived myeloid cells was shown to induce tumor dormancy of breast cancer cells at the perivascular niche (PVN) in the BM (59) and prevent metastatic outgrowth of breast and prostate cancer cells in the lungs (60). Given that TSP1 is a large protein, it is not feasible to consider it as a potential treatment. However, TSP1 can be induced by prosaposin. Indeed, administration of the TSP1 mimetic peptide prosaposin (DWLPK) was shown previously to induce TSP1 in BM-derived myeloid cells recruited to the lungs. The latter in turn assembled a metastasis-suppressive microenvironment (60) ( Figure 1 ).

Hence, the approach of inducing natural factors of the suppressive niche such as TSP1 and or SPARC is very appealing. However, this approach will require continuous treatment to ensure indefinite quiescence of the residing dormant DTCs and thus may lead to potential toxicities. Furthermore, TSP1 has been also reported to exert opposing effects which can be also attributed to multiple receptors that mediate TSP1 signaling (61). Likewise, SPARC was reported to have context and tumor dependent impacts on tumor progression (62). Therefore, we must consider TSP1/SPARC’s multiple effects along with the type of tumor and stage of the disease if we want to consider it as a future preventive treatment for cancer recurrence. Another important aspect that should be taken into account when striving to reinforce the dormant niche with TSP1 or other suppressive constituents such as TGFβ family members, is the role they also play in the immune evasion of DTCs (61, 63).

Therefore, it may be more practical to induce common cell-intrinsic mechanisms converging from the different microenvironmental cues comprising the dormant niches.

Pioneering work by Aguirre-Ghiso and colleagues identified p38-MAPK (mitogen-activated protein kinase) induction vs. reduction in ERK (extracellular signal-regulated kinase)-MAPK signaling (p38^high–ERK^low signaling axis) as hallmark of tumor dormancy in several types of tumors (35, 64–66). Interestingly, Debio-0719, an inhibitor of lysophosphatidic acid receptor 1 (LPA1), was shown to induce tumor dormancy of triple-negative breast cancer (TNBC) cells at distant organs by inducing the p38^high–ERK^low signaling axis (67). Though this inhibitor is still at the preclinical stage, some other inhibitors of LPA1 are currently being tested in clinical trials for fibrosis. These inhibitors include LPA1 inhibitor BMS986202 (previously AM152) and BMS-986020 [reviewed in (68)]. Therefore, potential use of these inhibitors in TNBCs patients as means to maintain residual disease at halt should be considered for future investigations ( Figure 1 ).

Orphan nuclear receptor NR2F1 was shown to be a critical node for dormancy induction in head and neck squamous cell carcinoma (HNSCC) and in DTCs of prostate cancer patients (69). NR2F1 was found to control tumor cell dormancy via SOX9 and RARβ-driven quiescence programs (70). Furthermore, combining 5-AZA with trans-retinoic acid (ATRA), reinstated in part the NR2F1-induced dormancy program in HNSCC (70) and induced TGFβ2. TGFβ2 is a BM-derived factor shown previously to impose dormancy in HNSCC and in prostate cancer cells (55, 56). Hence, this combination can induce both dormancy programs and may also contribute to the formation of dormant niche ( Figure 1 ).

The microenvironment of the metastatic niche (34, 36, 40, 71) and its remodeling (35, 38) plays a fundamental role in dictating the fate of residing dormant DTCs by inducing cell-intrinsic mechanisms culminating in the escape from tumor dormancy (39) ( Figure 1 ).

Several reports implicated the role of chronic inflammation (72–74) and/or fibrosis (75, 76) as instigators of DTCs awakening. Fibrosis occurs due to a dysregulated wound healing response. Formation of a fibrotic-like milieu in the lung enriched with type I collagen (Col-I) and fibronectin was part of the tumor ‘permissive’ microenvironment to support dormant mammary DTCs outgrowth (75). Utilizing a 3D model system to study tumor dormancy (77, 78) it was demonstrated that fibronectin and Col-I induced beta 1 integrin (Intβ1) downstream signaling in dormant mammary cells via activation of focal adhesion kinase (FAK) by Src. This activation resulted in downstream activation of ERK, which in turn induced phosphorylation of myosin light chain (MLC) by myosin light chain kinase (MLCK), culminating in F-actin stress fiber organization and transition from quiescence to proliferation. Inhibition of MLCK activation (75, 77) and or Intβ1 expression (75) prevented dormant DTCs outgrowth in vitro and in vivo. Similarly, sustained lung inflammation caused by tobacco smoke exposure or nasal instillation of lipopolysaccharide (LPS) induced the outgrowth of dormant DTCs in the lungs by formation of neutrophil extracellular traps (NETs), which in turn lead to cleavage of laminin-111 by NET-derived elastase and MMP-9 and induction of the Intβ1/Src/FAK/MLCK axis (79). In addition, prostate cancer patient-derived xenograft lines were shown to transition from their dormant state once they engaged with the BM stoma by constitutively activating MLCK (80). Hence, MLCK may serve as potential target to prevent awakening of the dormant DTCs. However, given that MLCK is widely expressed in many normal tissues and in smooth muscle cells, presents a clinical challenge that may require the exploration of other avenues to inhibit MLCK activation in dormant DTCs. One such indirect approach to consider is inhibiting Intβ1 activation.

Notably, several studies highlighted the potential role of Intβ1 activation in regulating the dormant to proliferative switch (81, 82). Previous work reported how the cross talk between Intβ1 and the urokinase receptor can dictate the fate of dormant breast and head and neck cancer cells (83, 84). Intβ1 partners with α subunits to form 12 potential integrin receptors, which bind to extracellular matrix (ECM) molecules such as collagens, laminin, and fibronectin (85).

Indeed, several pre-clinical studies successfully inhibited Intβ1 activity, including α5β1 [reviewed in (86)] which binds the ECM protein fibronectin. In the clinical settings, the anti-α5β1 integrin antibody, volociximab, in combination with carboplatin and paclitaxel demonstrated some encouraging preliminary results in a Phase Ib clinical trial in advanced non-small-cell lung carcinoma (87). Interestingly, JSM6427 a small molecule inhibitor of α5β1 integrin was evaluated in a Phase I clinical trial for the treatment of age-related macular degeneration. It warrants further investigation whether JSM6427 could also be effective in preventing the awakening of dormant DTCs given its mode of action (http://clinicaltrials.gov/ct2/show/NCT00536016) ( Figure 1 ). JSM6427 was shown to inhibit attachment of human retinal pigment epithelium cell (RPE) to fibronectin. This in turn promoted quiescence and cortical organization of the cytoskeleton of RPE. Similarly, inhibition of Intβ1 binding to fibronectin prevented the awakening of dormant mammary cancer cells resulting in their cortical F-actin organization reminiscent of the cytoskeletal organization of dormant DTCs (75, 77).

Therefore, if repurposing JSM6427 to treat cancer patients is being considered, the clinical regimen by which this drug will be administered must also be considered. In light of initial studies demonstrating Intβ1 plays a perquisite role in the reactivation of dormant DTCs [reviewed in (88)] while having no significant impact on actively proliferating metastases (75). Hence, these findings should be considered when designing future clinical trials with α5β1 integrin inhibitors. Changing the therapeutic paradigm of cancer therapy to a preventive treatment targeting early dormant DTCs rather than the current strategies aimed at treating patients with already advanced disease should be considered. Moreover, given that fibronectin and Col-I are part of the fibrotic milieu, preventing engagement of residual cells with such a supportive milieu after local surgery seems crucial.

Indeed, surgical trauma induces local and systemic inflammatory responses that can also contribute to the accelerated growth of residual and micrometastatic disease (89–91). Hence, intervention with inhibitors for α5β1 integrin should be pre-operative and immediately after surgery (post-operative).

Another receptor shown to interact with Col-I at the permissive site is DDR1. Col-I was shown to boost the association of DDR1 with TM4SF1, which, in turn, induced non-canonical signalling through the JAK2/STAT3 axis in dormant breast cancer cells leading to their outgrowth at multiple organ sites (92). Given that development of DDR1 inhibitors are at their infancy, a selective oral JAK2 inhibitor such as fedratinib (Inrebic^®), recently approved by the FDA for the treatment of myeloproliferative neoplasm-associated myelofibrosis (93) may be considered as a potential drug to be tested in a preclinical setting ( Figure 1 ).

In addition to preventing the engagement of dormant DTCs with their permissive niche and inhibiting cell-intrinsic mechanisms induced by signals arising at this niche, inhibiting the formation of such a permissive niche could represent a viable approach. For instance, inhibiting the cross-linking of Col-I by either lysyl oxidase (LOX) or lysyl oxidase like 2 (LOXL2), could prevent formation of the fibrotic milieu. Indeed, it was shown that LOX and/or LOXL2 inhibition significantly decreased pulmonary metastatic burden (94, 95). Furthermore, LOXL2 was shown recently to exert a cell autonomous role in the emergence of dormant DTCs. A study by Weidenfeld and colleagues demonstrated that LOXL2 expression induced by hypoxia in dormant breast DTCs promoted their epithelial to mesenchymal transition (EMT). This in turn endowed the cells with stem-like properties leading to their escape from tumor dormancy both in vitro and in vivo, while inhibiting LOXL2 expression prevented their outgrowth (96, 97). Overall, these studies suggest that LOX/LOXL2 may serve as a therapeutic target to prevent the emergence from tumor dormancy to overt metastases ( Figure 1 ).

Interestingly, a recent report demonstrated how a systemic inflammatory response induced after surgery promotes the emergence of dormant immunogenic DTCs at distant anatomic sites while, preoperative anti-inflammatory treatment with meloxicam, a nonsteroidal anti-inflammatory drug (NSAID), prevented the outgrowth of DTCs in the lungs (74). Notably, these finding are in line with a retrospective analysis carried out on breast cancer patients who received anti-inflammatory analgesics prior to surgery. These patients exhibited reduced incidence of early metastatic relapse (98, 99).

Overall, the studies presented here emphasize the important role of inflammation and/or fibrosis in dormant DTCs outgrowth and also reinforce the notion that intervention in the perioperative stage and immediately after re-section may be critical to prevent local and/or distant recurrences.

The mechanisms responsible for the long-term survival of DTCs are just beginning to emerge.

Previously, Src and ERK1/2 activation were shown to be essential for the survival and outgrowth of dormant breast DTCs. By utilizing a 3D model system and in vivo model system to study tumor dormancy (75, 77, 78) it was shown that only combined inhibition of ERK1/2 and Src in dormant breast cells culminated in their eradication (102). These findings suggest that combining a Src inhibitor such as saracatinib (AZD0530) with the FDA-approved MEK1/2 inhibitor trametinib may eradicate dormant breast tumor cells before they awaken ( Figure 2 ).

The activation of the transcription factor ATF6α was shown to regulate the survival of quiescent squamous carcinoma cells. ATF6α activation induced survival through the up-regulation of Rheb and activation of Akt-independent mTOR signaling (103). Of note, two mTOR inhibitors have been approved by the FDA to treat cancer and several are under clinical investigation as a combination or monotherapy. However, these clinical trials to date are designed to test the drug efficacy in advanced cancer or recurring cancer patients but not as a monotherapy to target dormant DTCs (104) ( Figure 2 ).

Another intrinsic mechanism shown to regulate DTCs survival is autophagy (71, 105, 106) ( Figure 2 ). Inhibition of autophagy in dormant breast and osteosarcoma cells by hydroxychloroquine promoted apoptosis of the dormant breast DTCs that have colonized the lungs and sensitized dormant osteosarcoma cells to cytotoxic anticancer agents (106, 107). Interestingly, autophagy was recently shown to restrict the outgrowth of micrometastases of several murine models of breast cancer while inhibition of autophagy lead to their outgrowth (108). Hence, inhibition of autophagy might differentially impact quiescent solitary dormant DTCs vs. micrometastases.

Notably, hydroxychloroquine is being widely used in cancer clinical trials in order to re-sensitize cancer cells to conventional therapies. Potential use of this drug as a preventive treatment to eradicate early DTCs may warrant further investigation (109). Currently, a Phase II trial of hydroxychloroquine with everolimus for prevention of recurrent breast cancer is ongoing, as well as an ongoing Phase Ib/II trial of gedatolisib (inhibitor of PI3K/mTOR pathway), hydroxychloroquine or combination of both [reviewed in (110)].

Aberrant expression of vascular cell adhesion molecule-1 (VCAM-1) on dormant breast DTCs was also found to ensure their survival once they colonize the lungs ( Figure 2 ). Specifically, VCAM-1 promoted PI3K/Akt activation and cancer cell survival by its engagement with the counter-receptor α4 integrins expressed on leukocytes. Furthermore, antibodies against α4 integrins blocked pro-survival signals induced by VCAM-1 (111). Therefore, disrupting the interaction between VCAM-1 and α4 integrins may potentially serve as a therapeutic target to promote dormant DTC eradication (112). Such drugs are already in clinical trials for the treatment of relapsing multiple sclerosis (MS) and inflammatory bowel disease (IBD). Currently, an orally active α4 integrin antagonist, AJM300, is under clinical trials in IBD patient [reviewed in (112)].

Notably, senescence of tumor cells that have escaped cytotoxic therapy has been proposed by several groups as another form of cell-intrinsic mechanisms that confer tumor dormancy and survival of circulating tumor cells (CTCs) and/or DTCs residing at the perivascular niche [reviewed in (101)]. This emerging concept warrants further exploration as it may open up an interesting opportunity to target these senescent-like cells with senolytics (small molecule drugs with selective killing of senescent cells) (113, 114) ( Figure 2 ). Indeed, several recent papers have demonstrated how senolytic drugs caused cell death of senescence induced cancer cells (115–118).

Overall, these promising studies highlight the importance in further investigating the cell- intrinsic mechanisms governing survival of dormant DTCs.

Signals arising at their foreign niche can induce cell-intrinsic mechanisms governing the survival of dormant DTCs and their escape from cytotoxic therapies. C-X-C motif chemokine ligand 12 (CXCL12) has been reported to constitute aspects of the dormant niche and act as a survival factor. CXCL12 secreted by osteoblasts was shown to induce the survival of disseminated breast tumor cells expressing the CXCR4 receptor by upregulating the Akt pathway via c-Src activation (119). Furthermore, the CXCL12/CXCR4 axis was shown also to mediate the localizing and tethering of prostate cancer cells and of breast cancer cells to the BM (120–122) and regulate their growth. Therefore, disrupting the CXCL12/CXCR4 axis may impinge on the survival of dormant breast tumor cells or induce the outgrowth of prostate tumor cells (123). Plerixafor (AMD3100), a CXCR4 antagonist approved by the FDA, was shown in a subcutaneous xenograft mouse model of human prostate carcinoma to dissociate the prostate cancer cells from their sanctuary site (the BM) and thus sensitize them to chemotherapy treatment (124). This approach was demonstrated to be effective also for other cancers, such as leukemia (125–127). Thus, dormant DTCs may be sensitized to chemotherapy by promoting their proliferation making them vulnerable to chemotherapy treatment ( Figure 2 ).

F-box/WD repeat-containing protein 7 (FBXW7) was shown to promote quiescence by ubiquitylation and proteasomal degradation of cell cycle promoters. Ablation of FBXW7 in dormant breast cancer cells caused DTCs to exit their quiescent state, which sensitized the cells to paclitaxel treatments in mouse xenograft and allograft models (128).

Recently, the TGFβ2/GAS6-Axl axis (56) was shown to be necessary for the induction of dormancy of prostate cancer cells in the BM. Treatment of dormant multiple myeloma cells with inhibitors to Axl such as BMS-777607 or cabozantinib released the cells from the endosteal niche in the BM, causing their reactivation, thus suggesting these cells could be susceptible to chemotherapy treatment (129). Indeed, in previous studies in solid tumors, combining chemotherapy treatment with inhibitor of Axl reduced tumor burden (130).

Another protein mediating the engagement of cancer cells to the BM niche is osteopontin. Osteopontin is an ECM protein secreted by osteoblasts in the BM niche and was shown to anchor acute lymphoblastic leukemia (ALL) cells to the BM culminating in their dormancy. Whereas, inhibition of osteopontin promoted ALL escape from tumor dormancy and sensitized them to cell-cycle–dependent Ara-C chemotherapy (131).

Overall, these studies suggest that dissociating dormant DTCs from their ‘safe haven’ niche can promote the cells to exit their quiescent state and potentially sensitize them to chemotherapy treatments.

Importantly, there are striking similarities between dormant DTCs and quiescent normal hematopoietic stem cells (HSCs), which reside in the BM. These similarities are exhibited by intrinsic mechanisms for survival, quiescence and their mobilization to the BM [reviewed in (43, 132)]. Therefore, when considering a strategy to dissociate dormant DTCs from their sanctuary site in order to sensitize DTCs to cytotoxic chemotherapy treatment it is critical to be sure that this will not facilitate depletion of HSC or cause reawakening of chemoresistant DTCs.

A recent report by Ghajar and his colleagues (100) demonstrated that disrupting the interaction between chemoresistant DTCs with the perivascular niche (PVN) by inhibiting Intβ1 and/or integrin αvβ3 sensitized DTCs to chemotherapy without inducing DTC proliferation ( Figure 2 ). However, 22% of mice still succumbed to bone metastases upon combining integrin β1 inhibition with adjuvant therapy in this study.

Hence, there exists a risk of DTCs mobilization from their dormant niche. Therefore, another strategy that we may contemplate to eradicate dormant DTCs is by preventing their immune evasion.

A recent report by Malladie and colleagues demonstrated that cancer cells selected from lung and breast cancer cell lines for their competence to establish latent metastasis (dubbed LCC), acquired quiescence with stem-like characteristics by expressing the Wnt inhibitor DKK1. The authors demonstrated that autocrine DKK1 helps disseminated LCC cells enter quiescence. Furthermore, these quiescent cells downregulated natural killer (NK) cell ligands leading to evasion of immune surveillance. Specifically, quiescent LCCs downregulated cell surface UL16-binding protein (ULBP) activators of NK cell-mediated cytotoxicity, as well as receptors for cell death signals. Therefore, quiescent (dormant) LCCs escaped cytotoxic killing by NK cells whereas proliferating LCCs were eradicated by activated NK cells (133). Of note, this study was conducted in immune-compromised mice and hence the role of NK cells in fully immune-competent animals remains to be determined.

Overall, these initial findings may open up in the future a novel therapeutic approach to selectively eradicate dormant DTCs by inhibiting their immune evasion. This will require re-expression of NK ligands in dormant DTCs. Interestingly, the inhibitor of histone deacetylases (HDACs) trichostatin A (TSA) was shown to induce ULBP expression in epithelial cancers (134). Whether TSA can inhibit the immune escape of dormant DTCs remains to be explored ( Figure 2 ).

Notably, several studies demonstrated how autophagy in cancer cells can impair the susceptibility of the cancer cells to NK-mediated killing (135). Given that autophagy was shown previously to be launched in dormant DTCs (71, 106), it will be worth exploring whether inhibiting autophagy of dormant DTCs will not only impinge on their direct survival but may also enable cytotoxic killing by NK cells ( Figure 2 ).

Another mechanism by which DTCs were shown to evade the immune system is by downregulation of the expression of major histocompatibility complex class I protein (MHC-I), thus evading CD8+ T cell recognition (136). Furthermore, the MHC-I negative phenotype of DTCs in the BM was shown to be associated with poor survival in curatively resected breast cancer patients without distant metastases (137). Similarly, single quiescent DTCs colonizing livers from patients and mice with pancreatic ductal adenocarcinoma (PDAC) were MHC-I-negative and exhibited unresolved endoplasmic reticulum (ER) stress. Notably, once these quiescent DTCs were resolved from their ER stress the cells emerged from their dormant state but also regained their MHC-I expression. Therefore, outgrowth of these cells occurred only when these quiescent DTCs were resolved from their ER stress and the T cell response was disrupted (138).

Hence, upregulating MHC-I may be an attractive approach to reinstate DTCs vulnerability to immune surveillance ( Figure 2 ). Notably, epigenetic control mechanisms regulating MHC-I expression have been frequently detected [reviewed in (139)]. Furthermore, interferon γ (IFNγ) was shown by several studies to act as an epigenetic modifier upregulating the expression of antigen-presenting machinery genes such as MHC-I (140). Future studies should be pursued in order to study whether treatment of dormant quiescent PDAC/breast DTCs with IFNγ will induce MHC-I antigen expression and eradication by CD8+ T cells. This kind of approach needs to take into account: i) whether the dormant DTCs with MHC-I downregulation can be sensitized to IFN-γ treatment. Given that approximately 30% of human tumor cells exhibit reduced IFN-γ sensitivity as a result of an impaired expression in the different components of the IFN-γ signaling (139) and ii) the clinical stage and context by which this treatment will be applied. Considering that IFN-γ has pleotropic effects on different stages in tumor progression (141). In addition, IFN-γ also is toxic given systemically, so inducing expression locally is likely to be more successful than systemic administration.

Overexpression of immune checkpoint proteins on dormant tumor cells was also shown to facilitate their immune escape. Dormant tumor cells in the DA1-3b/C3H mouse model of AML evade cytotoxic T-lymphocyte (CTL)-mediated killing because they overexpress PD-L1 (B7-H1) and CD80 (B7-1) (142). PD-L1 binds to receptors on CTLs (PD-1) and thus promotes CTL death and exhaustion. Importantly, this immune evasion was overridden by activating NK cells with CXCL10 (143). Overall, these findings suggest that reinstating either MHC-I and/or NK cell ligands, as well as inhibiting immune checkpoints proteins on dormant DTCs, may re-sensitize them to cytotoxic killing by T cells and/or NK cells ( Figure 2 ).

Myeloid–derived suppressor cells (MDSCs) may also indirectly regulate the survival of dormant DTCs. MDSCs represent a population of special cells of the immune system, which consist of immature macrophages, immature granulocytes, and immature dendritic cells. MDSCs suppress activation of T and NK cells through the production of reactive oxygen species (ROS) and arginase 1 (Arg-1), along with recruitment of other immune suppressive cells such as regulatory T cells [reviewed in (144)]. Accumulating evidence suggests that enrichment and activation of MDSCs correlates with cancer recurrence and poor clinical outcome.

Hence, modulating MDSC immunosuppressive activity may in turn prevent immune evasion of dormant DTCs. Inhibitors of phosphodiesterase-5, sildenafil and tadalafil, were shown to inhibit the immunosuppressive activity of MDSCs in preclinical and clinical studies by the downregulation of inducible nitric oxide synthase (iNOS) and Arg-1 activities (144). Promising results with tadalafil have been reported for head and neck squamous cell carcinoma and melanoma patients (145–147). Entinostat, a class I histone deacetylase inhibitor, was also shown to inhibit the immune suppressive activity of MDSCs [reviewed in (144)]. Importantly, entinostat has been evaluated in Phase I and II trials in patients with advanced malignancies, with a favorable risk–benefit profile [reviewed in (148)].

Overall, several drugs that have already been tested in the clinical setting may be considered for inhibiting the immune evasion of dormant DTCs thus potentially leading to their demise ( Figure 2 ).

A potential multifaceted target that has been demonstrated to be a molecular hub in mediating tumor escape from immune surveillance (149) and regulate the outgrowth of dormant DTCs is STAT3. Expansion and immunosuppression of MDSCs is mediated by activation of STAT3 [reviewed in (149)]. Anti-inflammatory M2-like macrophages which also play an important role in metastasis at distant organs [reviewed in (150)] are polarized to the M2 phenotype by STAT3 activation [reviewed in (149)]. Furthermore, in addition to STAT3’s role in immune suppression, STAT3 activation was recently shown to also directly mediate dormant DTC outgrowth. DDR1 interaction with Col-I at the permissive niche induced non-canonical signaling converging on activation of STAT3 in dormant breast DTCs, which culminated in their outgrowth at multiple organ sites (92). Notably, in paraffin-embedded tissue microarray sections of human breast cancers there was a significant increase in the expression of phosphorylated STAT3 (pSTAT3) in lung metastases compared to their matched primary tumors and the highest levels of pSTAT3 were present in those that had recurred after a short disease-free interval (92). These results suggest a role of activated STAT3 in metastatic outgrowth of dormant breast DTCs in the lungs. Therefore, targeting STAT3 may prevent both dormant breast DTCs outgrowth at the permissive site in the lungs and the formation of an immune suppressive niche.

Hence, STAT3 may serve as an attractive clinical target given that inhibitors of STAT3 have already entered clinical trials along with newer inhibitors at the preclinical stage [reviewed in (151)]. Moreover, several FDA-approved drugs were shown already to inhibit STAT3 signaling and thus may be repurposed to potentially prevent the outgrowth of dormant breast DTCs (152). However, it is important to keep in mind STAT3’s central role in signaling networks and therefore targeting it may lead to toxicity.

Of note, STAT3’s role in cancer recurrence is just beginning to unravel, and a recent study in contrast demonstrated that dormant breast DTCs residing in the BM niche remain dormant upon activation of STAT3 via the activation of leukemia inhibitory factor (LIF) receptor (53). Furthermore, inhibition of STAT3 led to bone osteolysis resulting in the outgrowth of dormant DTCs (53). Hence, LIFR : STAT3 signaling appears to confer a dormancy phenotype in breast cancer cells disseminated to bone. In addition, STAT3 was shown to be part of a pro-dormancy gene signature for estrogen receptor positive breast tumors (153)

Therefore, further research needs to be conducted in order to clarify the role of STAT3 in dormancy and outgrowth. It may be that the outcome of targeting STAT3 may depend on breast cancer subtype, the site of hibernation of the dormant DTCs and the precise timing of such intervention.

Another gatekeeper that holds great promise is NR2F1. Reinstating NR2F1 expression in the BM may prevent dormant DTCs from awakening by promoting cell-intrinsic dormancy programs in prostate and/or HNSCC cells while also reinstating the dormancy niche in the BM by secretion of BMP7 and TGFβ2 (69, 70). Furthermore, SPARC, which is a target of NR2F1 (58), was shown to regulate tumor dormancy of prostate cancer cells by promoting the expression of BMP7 in BM stromal cells.

Indeed, recent studies demonstrated that combining the epigenetic regulating drug 5-AZA with the differentiation agent trans-retinoic acid (ATRA), reinstated in part NR2F1 expression (70), while 5-AZA by itself reinstated SPARC expression in bone DTCs (58). This combination is now part of an ongoing Phase II clinical trial (https://clinicaltrials.gov/ct2/show/NCT03572387) to study combined 5-AZA and ATRA treatment on top of standard of care in recurrent prostate cancer patients based on rising prostate-specific antigen (PSA) only.

Overall, these studies may open up novel avenues to maintain DTCs dormancy.

As noted herein, dormant DTCs are undetectable at distant sites and reside as single/small cellular quiescent cells and or as small indolent micrometastases. Until it becomes possible to directly (e.g., via imaging or biopsy), or indirectly (e.g., via a biomarker) detect and follow such dormant DTCs and indolent micrometastases, we must rely on traditional clinical end points to assess the activity of drugs that may target and eliminate or control dormant DTCs. Unfortunately, there is no established trial design that completely accomplishes this task.

This is because clinical trials have several flawed assumptions in the context of examining effects on minor populations (180). The first is that clinical trials consider a cancer to be a single entity. Growth of existing lesions or development of new lesions is unequivocally considered progressive disease and thus failure. The second assumption that limits detection of effects on minor populations such as DTCs follows: trial endpoints are indifferent to whether failure (new lesion growth) was due to the outgrowth of dormant DTCs or outgrowth of small tumors below the detection limits of modern imaging. Finally, there is currently no method in clinical trials to distinguish between preventing the outgrowth of small tumor populations associated with dormant DTCs vs eradicating dormant DTCs, since both of these phenotypes would be considered a complete response if accomplished for a long enough duration of follow up.

Herein lies the opportunity for novel trial endpoints to assess agents that affect dormant DTCs and contribute to improved outcomes. Osteosarcoma (OS) provides an excellent disease model for this discussion. OS recurrence is observed in about 40% of patients presenting with localized disease who are treated with standard of care, most commonly to the lungs within a few years of completing therapy (181). About 80% of patients presenting with metastatic disease and who achieve a second complete remission by surgical resection will have further additional recurrences in the lungs. It is unclear if systemic therapy provides benefit when recurrent disease is amenable to resection and thus there is not a standard of care for systemic therapy either before or after surgical resection of lung metastases (181, 182). Thus, the clinical trial community has adopted historically controlled trial designs for this population whereby the null hypothesis is that 20% of patients after resection of the lung lesion will remain without disease at 12 months and an active agent being defined by doubling this to 40% of patients remaining without detectable lung lesions at 12 months (183). ( Figure 3 ). Controlled trials in which a therapeutic agent is compared to clinical choice can also be utilized in Phase II or in pivotal or post approval studies with the above time lines informed by historical data. While this trial design captures failure of both growth of subclinical cancer populations and recurrence from DTCs, it is unable to distinguish between these modes of failure.

The hypothesis that we believe needs to be tested clinically involving dormant DTCs posits a limited number of dormant DTCs and/or dormant indolent tumors either residing alone or in the presence of undetectable micrometastases. Thus, in the presence of undetectable micrometastases even a 100% effective agent at either eliminating dormant DTCs or preventing their outgrowth may fail since the preformed micrometastases may progress clinically. Thus, in order to demonstrate the effectiveness of dormant DTCs targeting therapies in OS, novel trial designs are required. This study design would be based on the hypothesis that some undetectable tumors are too advanced to be affected by drugs that target dormant DTCs. In other words, if microscopic micrometastases resulting from outbreaking dormant DTCs are already present at the time of initial recurrence, they will emerge as additional lesions with or without DTCs treatment. Such a recurrence could be considered an “early failure” and its presence will not truly reflect the efficacy of an agent preventing the outgrowth of dormant DTCs or eliminating dormant DTCs ( Figure 4 ). We thus propose an innovative trial design, namely, adding a primary aim of improving disease remission at a later time point (e.g., 2 years following first recurrence in addition to the initial EFS at 12 months post-surgery per point 2 above). This could be an open label study informed by historical data or a controlled study in osteosarcoma. In this scenario, an agent controlling or eliminating dormant DTCs, which would fail by any current trial methodology that removes patients from study at first progression (if any non-dormant cells are present), will have a chance to demonstrate important longer term disease control and clinical benefit missed by current trials.