Targeting cancer cell dormancy

Abstract

The field of tumour dormancy, originally defined as a clinical phenomenon of late recurrence after a long, apparently disease-free period, has seen significant advances that now allow us to think about monitoring and targeting dormant tumour cells to prevent relapse. In this Viewpoint article, we asked experts to share their views on the steps that are needed to translate dormancy research into the clinic.

Can we accurately detect and quantify dormant tumour cells in patients with cancer? What are the major methods for sampling and the most sensitive techniques to monitor dormancy?

Mickie Bhatia: Within patients with acute myeloid leukaemia (AML), relapse is highly prevalent and heterogenous both within and among patients. Methodologies to identify dormant tumour cells while the patient is in clinical ‘remission’ after a successful treatment have been developed over the years. The presence and frequency of these cells, often referred to as minimal residual disease (MRD), can be assayed by both digital droplet polymerase chain reaction (ddPCR) and flow cytometry. MRD identification by flow cytometry relies on tracking patient-specific leukaemia immunophenotypes, and ddPCR relies on tracking patient-specific genetic mutations that are harboured by leukaemic cells. As both techniques rely on the precise characterization of patient-specific factors, they may not be applicable to all patients with AML. Furthermore, although both are considered sensitive techniques, patients that maintain an MRD-negative state through ddPCR and flow cytometric testing can still relapse, highlighting the need for more effective methodologies and approaches.

Julio A. Aguirre-Ghiso: Yes, disseminated cancer cells (DCCs) can be detected and quantified at the time of surgery in patients without metastasis (M0) in sites that can be sampled such as in lymph node biopsies and bone marrow aspirates. However, this is not without limitations as the markers used may only detect subpopulations and not all sites can be probed. However, the sites and detection methods can be leveraged; for example, bone marrow DCC detection can be used for monitoring response to treatment¹. In epithelial cancers, DCCs can be detected using antibodies against markers of epithelial lineage such as cytokeratins (including cytokeratin 7, 8, 18 and/or 19 by antibodies A45-B/B3, AE1/AE3 or 2E11) or sialomucin (E29). It is true that some DCC subpopulations may lose these markers but more testing should be done because, at least in our experimental models, epithelial identity is never fully lost. Still, even if suboptimal, it may be more informative than not detecting and profiling the MRD at all. In melanoma, gp100 is also a reliable marker. A significant proportion of patients positive for DCCs may develop distant metastasis after 5 years, suggesting that these DCCs were clinically dormant (asymptomatic)². Dormancy markers are only recently being tested. Markers obtained from experimental models, such as nuclear receptor subfamily 2 group F member 1 (NR2F1), provided evidence that, compared with metastatic masses, solitary or small clusters of DCCs in the bone marrow and lymph nodes upregulate NR2F1 and transforming growth factor-β2 (TGFβ2). In patients with breast cancer, the frequent detection of NR2F1 in bone marrow DCCs was associated with delayed or no bone metastatic relapse and this was inversely correlated with the proliferation marker Ki67 (ref. 3); those patients with DCCs that are low for NR2F1 relapsed quickly. Such markers may be useful in new clinical trials using anti-HER2 therapies in the adjuvant setting that appear to prevent DCCs or micro-metastasis from progressing to overt lesions and, thus, may activate dormancy mechanisms². Although promising, more testing of these or new model systems-derived markers in the clinic is needed to identify reliable markers and accurately detect and quantify dormant DCCs in patients.

Christoph A. Klein: I will focus my discussion on dormancy of solid epithelial cancers and not touch hematopoietic malignancies. For solid cancers, the most honest answer to this question is a simple ‘No’, for two main reasons. First, detection in patients is currently still based on the markers originally defined by G. Riethmüller over three decades ago⁴: epithelial cytokeratins and EpCAM, which are useful for the detection of DCC in mesenchymal organs such as bone marrow or lymph nodes. However, given the enormous insights that have been gained about epithelial cell plasticity in the past decades, we have currently no clue whether, or to what extent, detection of DCC by these markers underestimates systemic spread in patients. The second reason is that many have equated DCC detection with dormant DCC (dDCC) detection, which is certainly incorrect. We still do not know the percentage of dDCC among DCC, and this leads me to the major problems of the field: what is the clinical relevance of dDCC and how can we determine it? Obviously, a dDCC is benign as long as it remains dormant. However, in patients, it is, in my opinion, almost impossible to prove that metastases arise from dDCC by awakening and not from non-dormant DCC that form metastatic colonies via slow but steady growth during clinical latency. I am afraid that such a proof is only possible ex juvantibus by successful selective eradication of dDCC that translates into improved outcome. I am not sure how such a study would ever be undertaken.

Judith Agudo: Unfortunately, dormant micro-metastases or single DCCs are too small to be detected by current technologies such as PET scan or MRI. Moreover, recovering them from organs like the lungs or liver by a biopsy is very difficult; it is like finding a needle in a haystack. However, bone marrow aspirates can be successfully used to quantify DCCs in patients with breast and prostate cancer. This has opened the possibility of investigating whether abundance of DCCs in the bone marrow can predict relapse and to better understand dormancy in patients. Nonetheless, performing bone marrow biopsies as a routine procedure to detect DCCs is challenging.

Tumour dormancy can result from two different biological processes: either tumour cells enter a state of quiescence or micro-metastases are kept in an equilibrium state of growth and destruction by the immune system. In the latter, it has been hypothesized — and there are some indications — that immune cells can prevent outgrowth while not completely eradicating the totality of tumour cells⁵. Something that we have been debating in our laboratory is whether it could be possible to use this ongoing immune attack as a surrogate to detect the presence of dormant indolent disease. The idea would be to detect whether this constant immune recognition is happening to conclude the presence of micro-metastases.

Ana Luísa Correia: It is possible to detect dormant DCCs in patients with cancer, yet we are currently unable to do it systematically or to distinguish between DCCs that require vigilant management from those that will not pose a threat. First, no current clinical imaging modality has the sensitivity to detect solitary DCCs or small clusters (2–24 DCCs) among oodles of other cells, or to provide predictive information. Second, the gold standard diagnostic tool of tissue biopsy becomes extremely challenging in metastatic sites other than the bone marrow, owing to inaccessibility, sampling bias not portraying DCC heterogeneity and impracticalities (such as medical contraindications, risk of complications and hesitation of patients), all enhanced when considering longitudinal sampling.

Liquid biopsy derived from body fluids has emerged as a promising alternative to improve upon these shortcomings. Because every organ has access to systemic distribution of these fluids, liquid biopsies have the potential to reflect the full landscape of metastatic progression within a patient, while being minimally invasive and, thus, amenable to longitudinal tracking. Despite the promise as dormancy diagnostic tools, detection and enrichment of fluid-based biomarkers (which are circulating tumour cells (CTCs), circulating tumour DNA (ctDNA), protein and extracellular vesicles) remain limited by the physiological scarcity and insufficient standardization and specificity of analytes, and pend validation in clinical trials before introduction into clinical practice.

Beyond assessing dormancy in life, rapid autopsies of deceased individuals with cancer reveal the burden of disseminated disease at its end stage, offering an opportunity to understand disease kinetics retrospectively and pinpoint DCC growth restraints across multiple sites. Increased access to and use of this admittedly tedious approach, combined with new imaging-based developments (for example, radiomics and label-free contrasting agents) and the virtue of liquid biopsies, are expected to unlock the full potential of early detection of metastatic disease and, consequently, comprehensive adjuvant cancer care.

Lewis A. Chodosh: Methods to detect dormant MRD in patients with common solid tumours are far from where they need to be. Because reversible quiescence is a functional definition and cannot be demonstrated for individual DCCs in a clinical setting, the greatest impediment to detecting dormancy in patients is the paucity of reliable surrogate markers. Although the absence of Ki67 identifies quiescent cells, exceedingly few markers have been identified whose presence (rather than absence) robustly and specifically identifies the dormant state.

The detection of DCCs in bone marrow aspirates is a strong independent risk factor for cancer recurrence across secondary sites, despite the fact that these methods do not evaluate dormancy. However, test sensitivity is poor, particularly for cytokeratin-based immunohistochemistry which typically detects only one or two DCCs in positive patients. This limits utility for assessing response to DCC-targeted therapies, evaluating dormancy state or otherwise analysing DCC phenotype, which likely holds important prognostic information beyond that provided by DCC number. Indeed, most DCC detection methods rely upon epithelial markers, thereby precluding detection of tumour cells with mesenchymal phenotypes and almost certainly underestimating DCC burden.

Additional markers for MRD can be detected in the bloodstream, including CTCs and ctDNA; however, their relevance to dormancy is uncertain given the inability to discriminate whether they are derived from dormant versus proliferating tumour cells. Indeed, the observation that patients with early-stage breast cancer with detectable CTCs or ctDNA have a high risk of recurrence over relatively short periods of time suggests that these markers likely originate from proliferating populations of tumour cells and, as such, may actually reflect subclinical recurrence rather than dormancy.

What host components determine the rate of relapse? Are there any biomarkers that allow the prediction of time to relapse in patients with cancer?

Mickie Bhatia: Positive MRD is a fair indicator of relapse likelihood in adult patients with AML; however, this offers little information on the onset of relapse. Furthermore, MRD status yields both false-negative and false-positive outcomes. There are many biomarkers in the literature that have been retrospectively applied to predict survival in patients with AML. These comprise markers across the leukaemic hierarchy, including those of quiescence, leukaemia stem cells (LSCs), monocytes and leukaemia progenitors. It is important to note that these biomarkers are often assayed at diagnosis, and not from inpatient cohorts who have undergone a wide array of clinical treatments and presenting varying outcomes. As such, these are not specifically applicable as relapse biomarkers. LSCs are often cited and theorized as the cause of relapse in AML. Yet, after decades of investigation and studies of putative LSCs, this concept has yet to be applied to the clinic in a biologically meaningful way, nor has it diminished the dismal relapse rate of patients with AML or changed clinical practice so far.

Christoph A. Klein: In my opinion, the host component is of outstanding importance. All of our patient-derived data suggest that DCC leave their primary tumour early after transformation, when the lesion is still small and early in ‘genomic time’ — usually long before diagnosis of the disease. At the distant sites, molecular evolution continues and depends on the specific niche involved⁶. The individual outcome may further be influenced by systemically acting factors or altered systemic conditions that may include factors released from the as-yet undiagnosed primary tumour, which may support niche formation, or factors resulting from organismal ageing occurring during year-long occult disease, but also factors administered or produced during anaesthesia and surgery. To identify reliable prognostic biomarkers in this complex situation is difficult. The focus of my laboratory is on DCC, in future combined with some surrogate for the patient-specific microenvironment. Whether this surrogate will be a genetic, cellular or functional assay is not clear yet; however, it should represent microenvironmental determinants driving or blocking DCC progression. I expect that careful molecular characterization of DCC will eventually allow us to predict much better who is at risk and who is not. Our recent data show that DCC constitute a heterogeneous cell population, whose members differ in their potential to act as candidate metastasis founder cells (MFC).

Julio A. Aguirre-Ghiso: It is postulated that niches that control quiescence and growth arrest of adult stem cells via cues, such as bone morphogenic proteins (BMPs) and TGFβs among others, may determine the rate of relapse as these cues can induce cancer cell dormancy; functional niches result in low relapse; dysfunctional niches lose control over DCCs and lead to high relapse⁷. The detection of pro-dormancy niche signals such as TGFβ2 and BMP7 in bone marrow supernatants of patients at risk of metastatic relapse was informative. Patients, for example, with detectable BMP7 displayed delayed metastasis whereas those with no BMP7 recurred early. Ageing and markers of aged niches that are dysfunctional may also inform on relapse. For example, niches that upregulate vitamin K-dependent protein S (PROS1) ligand may indicate earlier relapse in older patients (~65 years old). Loss of WNT5A expression and function in melanoma DCCs in aged niches may also mark for early relapse. Although promising, more clinical studies incorporating the detection of such biomarkers and their relationship to the timing of relapse is needed. Having host-derived markers of relapse is also important if DCCs are hard to detect and profile or whether their presence or absence is not informative. It is thought that functional status of the immune system may also control the rate of break from dormancy, but more work is needed in this space⁷. Overall, much more work pairing basic scientists and clinicians is needed to test these potential biomarkers. This needs team science with basic scientists that are not afraid to work with clinicians and clinicians that learn the biology and jointly design innovative trials. Unless this happens, the experimental data will remain in the basic science space.

Lewis A. Chodosh: Current clinical tests cannot reliably identify patients with MRD or accurately determine who will relapse, much less when. Clearly, the immune system has a role in determining the probability and timing of tumour recurrence, as illustrated by immunosuppressed transplant recipients who develop cancer in the transplanted organ soon after receiving it from an apparently cancer-free donor. On a population level, obesity is associated with an acceleration of breast cancer recurrence, with multiple mechanisms proposed relating to obesity-associated changes in the host microenvironment. More broadly, the influence of host microenvironment on recurrence is underscored by the impact of additional factors on dormancy, including cytokines, integrins, angiogenic regulators, oxidative stress and ligand-activated growth factor receptors.

Host-related factors notwithstanding, the properties of tumour cells themselves are a major determinant of recurrence. For example, oestrogen receptor (ER) status in breast cancer is a reasonably reliable predictor of recurrence kinetics, with most ER-negative tumours that recur doing so within 5 years whereas ER-positive tumours relapse at a relatively constant rate of 1–2% per year over the course of several decades. Moreover, subgroups of late-recurring ER-positive and ER-negative breast cancers have been identified based on the presence of characteristic genomic alterations in tumour cells.

Judith Agudo: What leads to the awakening of dormant DCCs and how to predict relapse are the Holy Grail questions in the dormancy field. In one way or another, we are all trying to address this. In my case, as an immunologist, I am focused on immune surveillance of dormant cancer cells. I think that many cases of relapse are triggered by a failure of the immune system to keep micro-metastases in check. This failure can result from immune cells being less effective owing to changes in lifestyle or the overall health of the patient, for example, sudden stress, the use of anti-inflammatory drugs or even an immune decline related to ageing. Alternatively, micro-metastases may acquire immune evasion mechanisms and, for example, lose a recognized neo-antigen or acquire mutations in the antigen presentation machinery. These genetic or epigenetic changes would then facilitate tumour escape from cellular immunity and, hence, outgrowth. It is hard to predict when a given tumour cell will acquire the right mutation to cloak from an ongoing antitumour immune response, but maybe, in the not-too-distant future, we can predict relapse by monitoring systemic immune function.

Ana Luísa Correia: Perhaps the most striking evidence that the host microenvironment controls relapse is the involuntary transmission of metastasis by organ transplantation from apparently disease-free donors into immunosuppressed recipients. A body of literature has cemented this predominant role of the immune system in timing relapse across several types of cancer, and we are just beginning to appreciate how site-specific differences in immune cell composition and spatial organization disproportionally impact DCC progression⁸. Within the liver, for instance, breast DCCs settled in natural killer (NK) cell-abundant microenvironments remain dormant, whereas those experiencing a local shortage of NK cells initiate metastases⁹. In extension, a drop of NK cells in patients with cancer who are apparently disease-free might serve as a biomarker to identify individuals who are at risk of relapse.

Besides immune cells, DCCs are locally conditioned by non-immune tissue-resident cells (for example, fibroblasts and endothelial and nerve cells), extracellular matrix, metabolites and soluble factors, all of which with tissue-specific subtleties yet to be fully understood¹⁰. Additionally, each individual is exposed to a unique combination of challenges throughout life (including ageing, circadian rhythm, the gut microbiota, therapy and certain pathological states) that tweak tissue physiology and catalyse relapse. Our capacity to expand our understanding of the rich and ever-changing ecosystem hosting DCCs will underpin efforts to predict which patients will tend to relapse and when.

What are the most promising therapeutic strategies or molecular targets to address tumour dormancy in the clinic? What are the patient populations that would most probably benefit from such treatment?

Lewis A. Chodosh: Depleting the population of dormant cells that survive therapy or otherwise blocking their reactivation is an exciting new approach to prevent cancer recurrence and its associated mortality. It might also be the case that depletion of MRD itself — dormant or otherwise — can delay or prevent recurrence. Inducing dormant cells to proliferate as a means to increase their sensitivity to therapies preferentially active against proliferating cells has also been proposed, though this approach seems precarious given that even a small number of therapy-resistant proliferating cells would ostensibly give rise to an incurable recurrent cancer.

Targeting dormant DCCs in patients requires a deep understanding of the biology of tumour dormancy, particularly as dormant residual DCCs may use pathways of survival and escape that differ from those of actively proliferating tumour cells. Unfortunately, although most mechanistic insights have been gleaned from model systems, the extent to which these operate in human cancers is largely unknown.

Given the unique biology of dormancy, therapeutic agents that are effective against dormant tumour cells may be ineffective against actively proliferating tumour cells, and vice versa. Consequently, a therapy that has failed in clinical trials in the metastatic setting may nonetheless show activity when administered in the dormant phase. Critically, the tacit assumption on which this approach is based is that it is possible to know that residual tumour cells in a particular patient reside in a dormant state and have not yet reactivated.

In this context, the possibility that different MRD markers reflect different stages of the dormancy-to-recurrence cascade becomes especially important. If detection of ctDNA and/or CTCs in a patient reflects the presence of proliferating tumour cells, dormancy-directed agents will probably be ineffective. Conversely, detecting DCCs in the absence of ctDNA or CTCs may identify patients that may be more likely to benefit from such agents. This, in turn, indicates that MRD testing focused solely on circulating markers may miss patients harbouring dormant MRD. Accordingly, developing improved methods for detecting MRD dormancy and reactivation is equally important as identifying dormancy-specific targets for therapeutic intervention.

Christoph A. Klein: We can either eradicate DCC, dDCC and MFC directly or decelerate or prevent their maturation and progression. For both approaches, we need to gain much better insights into their nature, gene and protein expression programmes, expression pattern of potential targets, functional phenotypes, cellular interactions and accessibility of their niches. As mentioned, I would not focus on dDCC (which by definition do not form metastases while dormant) but on MFC, because I expect all cancer types and patients to benefit from such a treatment. We have to consider that the statistical association of DCC detection and outcome is found for relapses that occur mostly within the first 5 years after surgery for all cancer types studied. Given the classical clinical definition of dormancy by Hadfield^11,12 (relapse later than 5 years), we currently do not identify dormancy-associated DCC.

Julio A. Aguirre-Ghiso: No FDA-approved standard-of-care strategies that specifically address dormant cancer cell biology are available. However, several strategies are being tested. Experimental agonists for NR2F1 have shown promising anti-metastatic effects in mice by turning on a stable dormancy transcriptional programme in head and neck squamous cell carcinoma (HNSCC) models. The combination of two FDA-approved drugs, azacytidine and retinoic acid, that induce a dormancy-like transcriptional programme in HNSCC and other cancer models and suppress metastasis was recently tested in patients with prostate cancer with biochemical relapse (NCT03572387); the results should inform on how to design dormancy-inducing clinical trials. Drugs in a phase I safety trial (NCT04834778) that target the endoplasmic reticulum kinase PERK, which serves as a survival kinase in dormant and stressed cancer cells, also have anti-metastatic effects in mice. Blocking specific adhesion signalling survival pathways may also sensitize cells emerging from dormancy to chemotherapy. These approaches have not reached clinical trials that test the maintenance or elimination of residual disease and those tested need to reach a higher power to be fully informative. In breast cancer, the use of bisphosphonates has shown promising results to eradicate DCCs, but these trials did not incorporate dormancy markers.

It is possible that for any cancer that under-goes an M0 phase or even in patients resected for metastasis, dormancy-inducing or anti-dormant cancer cell strategies might provide a benefit. This requires testing if the therapies delay the emergence of the first metastasis after adjuvant treatments or in patients with stage IV cancer that underwent metastasis resection, if the therapies delay the appearance of the next metastasis. It is possible, although not tested, that anti-hormone therapies in hormone receptor-positive breast cancer and in prostate cancer are inducing dormancy-like responses. It is also possible that anti-HER2 therapies in the adjuvant setting might be inducing dormancy of DCCs or delaying reactivation as concluded from recent clinical trials². It is also possible that the mechanisms by which kinase inhibitors induce a ‘persister’ state overlap with niche-induced dormancy and DCC survival mechanisms. Thus, understanding the mechanisms of maintenance and survival of persister cells as mapped on spontaneous dormancy mechanisms might prove useful to eliminate persister cancer cells or prevent them from re-entering proliferation. These questions would need innovative trials to test the fate of DCCs and this again required better pairing of clinicians and scientists.

Judith Agudo: Several companies and academic laboratories are attempting to develop treatments to retain dormant DCCs in their state of cell cycle arrest. Obviously, achieving this could save many lives and it is a promising avenue for prevention of cancer recurrence. However, in my opinion, eradicating these DCCs and micro-metastases should be the preferable solution. Quiescent cells are inherently resistant to traditional systemic therapies, such as chemotherapy; precisely, this intrinsic resistance is the major roadblock for prevention of relapse. But I think that immunotherapy can change this landscape. I believe that immune cells can find and eliminate dormant DCCs if we learn how to unleash the right response. The literature on immune surveillance of dormant disease is still very limited and some early studies show that cancer cells that enter a quiescent state escape T cell attack^13,14. Thus, the key is to identify the mechanisms of immune evasion that quiescent cancer cells evolve or, alternatively, exploit other immune cells for which escape has not happened; for example, if dormant tumour cells downregulate MHCI, then we could either develop a therapy to restore MHCI or entirely switch to exploiting NK cells or tumoricidal macrophages. Thus, immune-mediated eradication of dormant micro-metastases and DCCs would be the most effective in patients with early-stage cancer before these cells evolve multiple escape mechanisms under immune-editing pressure.

Ana Luísa Correia: I believe that a remarkable opportunity resides in harnessing site-specific immunotherapy to either eradicate DCCs or, most probably, maintain metastatic disease in a long-term dormant state. One strategy would be to directly sustain pre-existing anti-metastatic resident immune populations, for example, by boosting liver NK cells with adjuvant interleukin-15 (IL-15)-based immunotherapy to prevent breast cancer hepatic metastases, or by activating bone γδ T cells through adjuvant administration of bisphosphonates to maintain breast cancer dormancy in the bone. The converse strategy would be to suppress pro-metastatic immune subsets at specific sites, for instance, by blocking high neutrophil density and activity, either by dis-solving neutrophil extracellular traps or by inhibiting S100A8–S100A9-mediated accumulation of oxidized lipids, to prevent relapse in the lung. Because site-specific immunity is heavily imprinted by non-immune resident cells, targeting the latter may be another prospect to push tissue immunity to effectively control dormant DCCs. One example would be to attempt reverting activated hepatic stellate cells into quiescence, either through the use of all-trans retinoic acid, vitamin D or nitric oxide, as a means to reset liver-resident NK cell proliferation, ensuring sufficient numbers to maintain DCC dormancy at this site.

With a main goal to reliably prevent relapse in patients who present with a primary tumour of virtually any cancer type, dormancy-targeted immunotherapy has the highest tendency to benefit patients in remission at risk of recurrence. Additionally, the notion that dormant DCCs coexist with growing detectable metastases opens an opportunity for dormancy-targeted immunotherapy to provide long-term benefit even in patients with stage IV cancer by preventing additional relapses.

Mickie Bhatia: Currently, the only curative treatment of AML is hematopoietic stem cell transplant (HSCT). The HSCT uses a two-pronged approach to target AML: the healthy hematopoietic stem cells (HSCs) can outcompete LSCs in the bone marrow niche to render them ineffective, and the T cells coming from the new healthy HSCs will target all remaining dormant leukaemic cells. The difficulty with HSCT is that only a handful of patients are eligible, as it is a relatively risky procedure which may not be well tolerated by already weakened patients. Even this has a relationship to MRD and is more successful in reduced MRD conditions

What are the main barriers to the translation of dormancy-targeting therapies into the clinic and what are the next steps to move these therapies into the clinic?

Ana Luísa Correia: The first challenge is to improve detection of DCCs and accurately assess metastatic dynamics in patients throughout treatment, which will be greatly facilitated by label-free imaging and machine learning-fuelled pathological assessment, the use of liquid biopsy-derived biomarkers, and increased access to rapid autopsies. The second challenge is to continue expanding our understanding of the properties of individual DCCs and of the distant microenvironment in which they have lodged that determine the risk and timing of site-specific relapse; here, the widespread implementation of orthogonal spatial technologies, combined with improved computational approaches to analyse multiple large datasets, will bring spatiotemporal resolution of relapse full circle. The third challenge lies in pairing clinical data with preclinical models that most accurately represent the response of patients and, thus, guide rational combination therapy to be tested in clinical trials. Ultimately, translating dormancy-targeted therapies to the clinic will require large and long-enough trials able to support the complexity of precision immuno-oncology protocols, and to extend clinical endpoints beyond the traditional progression-free survival to also include intermediate readouts (such as DCC burden, and fluid-based biomarkers) that might anticipate metastatic dynamics (Box 1). Convening a broader community of basic researchers, clinician–scientists, technology entrepreneurs and patient advocates is our best way to overcome these challenges and accelerate delivery of long-lasting cures to patients with cancer.

Box 1. Challenges and opportunities of translating dormancy research into the clinic.

Clinical relevance of DCC

• Timing of relapse: At the moment, we are unable to reliably predict precisely which patients will relapse and, even less so, when. More accurate markers, methodologies and approaches to predict relapse in patients based on MRD markers or specific primary tumour information are needed to identify patient populations that would benefit from dormancy-targeting therapies.

• Contribution of dormant DCCs to relapse: Determining whether dormant DCCs are benign or require management to prevent relapse remains challenging. The relationship of biomarkers to the timing of relapse needs to be determined. DCC-targeting therapy that improves outcomes would best serve to prove the clinical relevance of dormant DCCs.

Detection, sampling and monitoring

• Identification of DCCs and detection of dormancy markers: We require accurate markers and sensitive methods for the detection of DCCs and also for determining whether they are resuming proliferation or remaining dormant. A few experimental markers potentially suitable to determine the dormant state of a DCC have been identified and await further validation in human patients with cancer. Next to the detection of DCCs in mesenchymal organs such as bone marrow or lymph nodes using epithelial markers, identifying markers to pinpoint DCCs of mesenchymal-like or hybrid origin is needed.

• Sampling challenges: Biopsies from metastatic sites other than the bone marrow or lymph nodes are often impractical or risky, limiting the ability to sample and analyse dormant DCCs in certain locations and at the appropriate frequency during patient management. The value of fluid-based biomarkers such as ctDNA and CTCs in determining recurrence and dormancy requires further analysis. Pairing ctDNA and CTCs with DCC or other markers of relapse might aid estimating the state of the residual cancer. Rapid autopsy samples allow for characterizing dormant and proliferating DCC of multiple organs of the same individual.

• Limitations to monitoring: Longitudinal tracking of dormant cells is challenging owing to the limitations of current imaging modalities and the challenges of obtaining repeated biopsies. Innovative clinical trial designs are needed that enable determining relevant biomarkers, either in the bone marrow, blood borne or tracked with non-invasive tools such as imaging modalities.

Therapeutic targets and clinical testing

• Therapeutics: Developing therapies that specifically target the unique biology of dormant tumour cells while they are well-tolerated will be crucial. This will require biological characterization and a distinction between dormant DCCs and other cell types to avoid toxicity and off-target effects.

• Clinical trials: When testing dormancy-targeting drugs, intermediate endpoints beyond traditional progression-free survival, such as DCC burden, and ctDNA paired with fluid-based biomarkers that inform on the host need to be considered to anticipate metastatic dynamics and therapy efficacy.

Community and regulatory affairs

• Challenging existing paradigms: For targeting dormant DCCs, shifting from ‘watchful waiting’ treatment paradigm to a more proactive intervention is needed. Similarly important is to test prevailing assumptions that lack thorough investigation and verification in human patients with cancer. Achieving consensus on terminology and concepts related to dormancy is necessary to avoid confusion and guide future research that will eventually lead to candidate methods in patient trials.

• Collaboration: A close collaboration between scientists (including cancer biologists, immunologists and other experts in different fields), clinicians and patient advocates in a focussed manner will be required to address the issues that would have immediate need for relapsed patients without other alternatives.

• Translational and clinical programmes: Funding agencies and research programmes should focus on translational and clinical studies related to dormancy biology and use inter-tumour and intra-tumour approaches to best determine generalizable versus tissue-specific features of dormant DCCs.

Judith Agudo: A major limitation to bring treatments for tumour dormancy to the clinic is the inability to detect and quantify dormant micro-metastases and DCCs in patients. Without a reliable assessment, it is impossible to confidently guarantee that a therapy has indeed eliminated these cells. Thus, the approach to test efficacy of a treatment would inevitably be to measure whether such therapy has reduced the number of patients that relapse. Some cancers can recur relatively fast, but others have dormancy periods of decades. This means that the results of a clinical trial for such diseases would need to wait for 20 or 30 years (Box 1). Obviously, this reduces attractiveness for industry and makes it economically very challenging for academic institutions.

Another roadblock, in my opinion, is the limited availability of robust animal models of tumour dormancy, mostly immune-competent models. Most preclinical models of cancer have been selected for fast tumour growth to accelerate research. But studies of dormancy require models in which DCCs are in long-term quiescence and micro-metastases enter in a long-lasting balance with the immune system.

Christoph A. Klein: The first barrier is our nomenclature (Box 1). The field must first agree on an understanding of the terms to convince authorities to permit much-needed new avenues of treatment. We should differentiate DCC-targeting, dDCC-targeting and MFC-targeting therapies. Then, we should convince clinicians and regulatory authorities that for prevention of metastatic relapse, drug efficacy in patients with manifest metastasis cannot be an essential prerequisite, as this is limiting the selection of approaches. Using overall or progression-free survival outcomes as endpoints in (neo)adjuvant therapy studies, however, decelerates clinical development. Using MFC phenotypes or MFC quantification as surrogate endpoints may be useful for clinical testing of novel therapies that either target MFC directly or indirectly via their microenvironmental interactions.

Mickie Bhatia: The main barrier of targeting dormant cells is the lack of understanding of what this means (Box 1). The term dormancy has become muddled and is arguably contrived within the field depending on context and experimentalisms. In part, this stems from commonplace assumptions that have been passively reiterated for decades and sequentially referenced in the literature in the absence of new primary supportive results. Some of the assumptions rely on logical fallacies that need to be addressed. Cellular dormancy such as quiescence grants protection to anti-proliferative therapeutics. This has led to the false equivalency of dormancy and resistance, in which cancer cells at remission are considered exclusively non-replicative. However, dormancy cannot be the only form of leukaemia cell resistance, exemplified by the high percentage of patients that do not respond to initial treatment, and it is unfounded to suggest that these refractory aggressive tumours are non-proliferative. Therefore, it is logical to postulate that remaining cells during remission are not exclusively dormant.

Another layer of complexity arises when discussing the concept of LSC quiescence and dormancy. Many cells of the leukaemic system are not active in the cell cycle, which include subtypes of LSCs. This claim has begun to be refuted, as it appears LSCs are more proliferative than initially hypothesized^15,16. Neverthe-less, this has led to the belief that dormancy markers are LSC markers, a fallacy of division. Both the false equivalency and fallacy of division drive the assertion that LSCs are present during clinical remission and are the origin of relapse, with the only evidence being that non-cycling cells are not targeted by chemotherapy. As the evidence and the assertion differ, this model should be revised as to allow the elaboration of new therapeutic strategies unrestrained by this dogma.

LSCs have also been cited as the driver of relapse owing to the retrospective predictive capacity of LSC-associated gene signatures on patient survival. This is a perfect reminder that correlation with overall survival is not causation of treatment failure, as there is increasing evidence that LSCs are effectively depleted by chemotherapy and are not relevant to the regeneration process in causal studies^15,17,18. These studies have proposed alternative mechanisms to leukaemic regeneration through gene expression and bonified functional assays inpatient tissues and a variety of modelling systems. Yet, proposed alternative theories have caught little traction, and LSC dormancy is consistently cited as the culprit of treatment failure and the prudent cellular target for increasing patient survival. Although the concept of reversable quiescence is possible, it is important to recognize that within the context of LSCs and treatment failure, it forms the basis of an unfalsifiable hypothesis. The regenerative nature of LSCs supports the disease when a patient is refractory to treatment, whereas the quiescent nature of these same LSCs ensures survival to treatment, which allows them to drive relapse. Whichever ‘characteristic’ of the LSCs fits better the current observation is cited. As these characteristics are mutually exclusive, this makes the hypothesis unfalsifiable.

As a field, it is time we consider the complexity of regeneration that is going unexplored, and effectively study and analyse this process without an LSC-shaped lens. This lens can lead to passive reiteration through ‘begging the question fallacy’, in which studies will assume that LSCs are the known driver of relapse, use that to orient their data analysis and experiments, and conclude that LSCs drive relapse. Passive reiteration in this manner can be exponential and could be a reason for the widespread acceptance of this idea. This makes it difficult to make any clinical or scientific progress.

With nearly three decades of studying LSC biology without a change in clinical treatment efficacy, it is time to reconsider the theories of dormancy and LSCs and how it relates to resistance and relapse: Dormant when? To what extent of depth in the cell cycle? Halted by what processes — the niche or chemical intervention? The next steps are to re-investigate dormancy and leukaemia both thoroughly and thoughtfully and consider alternative mechanisms of leukaemic dormancy, resistance and regeneration, in hopes of finally helping patients with AML and use logical approaches rather than concepts to development new drugs and new treatments that prevent relapse.

Lewis A. Chodosh: Several major barriers to translating dormancy-targeted therapies to the clinic exist. The first is strict adherence to a ‘watchful waiting’ treatment paradigm, an approach with little active intervention during the surveillance period that risks forfeiting the possibility of cure, because recurrent cancers are often incurable and dormancy-targeted approaches will be ineffective once DCCs have resumed proliferation (Box 1). A second barrier is the lack of therapeutics that leverage the unique biology of dormant DCCs while being sufficiently well-tolerated to justify treating patient populations with MRD who might or might not recur. A third barrier is the absence of sensitive assays to (1) identify patients with dormant MRD who are at elevated risk of recurrence, (2) provide molecular information to enable selection of personalized therapies and (3) monitor the impact of selected therapies on dormant DCC number and phenotype as a pharmacodynamic marker.

Encouragingly, two early clinical trials aimed at targeting breast cancer DCCs using chemotherapy¹ (NCT00248703) or zoledronic acid¹⁹ (NCT00172068) have yielded promising results. More recently, screening studies to identify patients with DCC-positive breast cancer²⁰ (NCT02732171), coupled with phase II trials to eradicate DCCs using the autophagy inhibitor hydroxychloroquine (HCQ) and/or the mTOR inhibitor everolimus²¹ (NCT03032406), the CDK4/6 inhibitor abe-maciclib with or without HCQ (NCT04523857), or the checkpoint inhibitor avelumab or HCQ with or without the CDK4/6 inhibitor palbociclib (NCT04841148), have been initiated. These initial therapeutic forays into the MRD and dormancy ‘window’ should help further establish the feasibility and acceptability to patients with cancer of serial bone marrow sampling in the MRD setting, while informing our understanding of how best to approach this new paradigm for recurrence prevention.

Julio A. Aguirre-Ghiso: Clearly, we need to improve the detection of DCCs and MRD and determine their dormant or non-dormant state; this will be necessary for testing pro-dormancy therapies, for example, although for dormant DCC-eradicating therapies, enumeration should be sufficient. However, these efforts cannot move forward unless we overcome in parallel other main barriers, such as the absence of translational and clinical programmes to develop and test new drugs and biomarkers (Box 1). Funding agencies such as the National Cancer Institute (NCI), Department of Defense (DoD), and Cancer Research UK (CRUK) among others have incorporated the concepts of cancer dormancy in their programmes. However, these remain focused mainly on basic science. More focused programmes on translational and clinical studies exclusively aimed at the wealth of dormancy biology being published would accelerate the translation of this science. Furthermore, work-shops or training programmes that are funded by the above agencies, for example, and integrating patients, scientists and physicians on how to implement testing biomarkers and drugs in the dormancy space would help all those parties bridge the gap from bench to bedside. Scientists need to better understand the clinical trial space and physicians the dormancy biology, so together with the patient advocates, all parties can contribute to innovative clinical trials. This is because the current model for testing new drugs and analysing endpoints does not allow clearly addressing the biology of cancer dormancy.

Biographies

Judith Agudo: Judith Agudo is an Assistant Professor in the Department of Cancer Immunology and Virology at the Dana-Farber Cancer Institute and the Department of Immunology at Harvard Medical School. She is also a member researcher in both the Parker Institute for Cancer Immunotherapy at Dana-Farber Cancer Institute and the Ludwig Cancer Center at Harvard, and an affiliated member of the Harvard Stem Cell Institute and a New York Stem Cell Foundation–Robertson Investigator. The Agudo laboratory investigates mechanisms of immune evasion of cancer stem cells and quiescent cancer cells with the goal of identifying immune-based strategies to prevent metastasis.

Julio A. Aguirre-Ghiso: Julio A. Aguirre-Ghiso is the Rose Falkenstein Chair in Cancer Research, Professor of Cell Biology and Director of the Cancer Dormancy & Tumor Microenvironment Institute at Albert Einstein College of Medicine. His work led a paradigm shift, revealing novel cancer biology that diverges from the notion that cancer is perpetually proliferating. His team discovered that reciprocal crosstalk between disseminated tumour cells and the microenvironment regulates the inter-conversion between dormancy and metastasis initiation. His laboratory also provided a mechanistic understanding of the process of early dissemination and revealed how it contributes to dormancy and metastatic progression and how stress-adaptive pathways allow cancer cells to persist while quiescent.

Mickie Bhatia: Mickie Bhatia is a Professor at McMaster University, the Canada Research Chair in Human Stem Cell Biology, the Michael DeGroote Chair of Human Stem Cell Biology and Cancer Research and currently the Program Director of Experimental Therapeutics in human leukaemias. His program is integrated with several clinical sites and forms the hub for chemical genomics and drug discovery targeting putative cancer cells with stem cell properties. His laboratory was the first to culture pluripotent stem cell in Canada and leverages stem cell technologies and novel assay development for preclinical and translational phase I study in patients.

Lewis A. Chodosh: Lewis A. Chodosh is Perelman Professor and Chair of the Department of Cancer Biology at the Perelman School of Medicine at the University of Pennsylvania, Associate Director for Basic Science at the Abramson Cancer Center and, along with Angela DeMichele, co-founder and co-director of the 2-PREVENT Translational Center of Excellence, which is focused on developing novel approaches to prevent tumour recurrence in patients with breast cancer by targeting dormant minimal residual disease. His laboratory focuses on understanding mechanisms of tumour dormancy and recurrence, developing improved methods for detecting and characterizing disseminated tumour cells in patients, and leveraging those advances through clinical trials. He has been elected to the National Academy of Medicine, Association of American Physicians and American Society for Clinical Investigation.

Ana Luísa Correia: Ana Luísa Correia leads the Cancer Dormancy & Immunity Lab at the Champalimaud Foundation. Her group strives to understand what makes a tissue favourable or not to metastasis and leverages this biology into therapeutic interventions that reliably prevent the emergence of metastases in patients with cancer. Ana has received a few international awards (2021 Metastasis Research Prize, 2022 Pfizer Oncology and AACR 2022 NextGen Stars) and is an EMBO Young Investigator.

Christoph A. Klein: Christoph A. Klein studied Medicine at the Ludwig-MaximiliansUniversity in Munich. He is heading two laboratories, one at the University of Regensburg for basic metastasis research and one at the Fraunhofer Society for clinical translation of the results. The major focus of the said laboratories is on the earliest beginnings and mechanisms of metastatic spread and colonization.