Background
Breast cancer patients are at risk of developing disease relapse decades after curative treatment because of the presence of minimal residual disease [
1]. Minimal residual disease is caused by spread of invasive tumor cells from the primary tumor through the circulation to distant sites [
2,
3]. Once in circulation, interaction with platelets seems to contribute to circulating tumor cell (CTC) survival as well as enhanced extravasation, especially to the bone microenvironment [
4‐
6]. In breast cancer, tumor cells often migrate to the bone marrow (BM) where these disseminated tumor cells (DTCs) can survive for years by entering a dormant state. This is a prolonged quiescent state, in which tumor cells are present, but disease progression is not clinically manifested. Two mechanisms are believed to maintain tumor cell dormancy: single-cell dormancy and/or micrometastatic dormancy [
7]. Single-cell dormancy, or cellular dormancy, is characterised by a state in which the tumor cells are non-proliferative and thus assumed to be resistant to traditional chemotherapeutics targeting proliferating cells [
7]. In contrast, the micrometastatic dormancy model, which is supposed to be linked to more aggressive breast cancer, involves slowly proliferating tumor cells that are counterbalanced by cell death from impaired vascularisation or immunesurveillance, which prevents tumor growth [
7]. However, dormant DTC survival in the BM depends on pro-survival signals from the microenvironment, as well as on development of complex immune evasion mechanisms in which interference with major histocompatibility complex–mediated antigen presentation seems to be important [
8‐
11]. Breast cancer recurrence after a long asymptomatic period, even more than 20 years after the initial diagnosis, is believed to arise from an interruption of this dormant DTC state, possibly initiated by microenvironmental factors in the colonised tissue [
12]. Thus, accurate identification of breast cancer patients at risk for late recurrences is needed to identify candidates for extended therapy and improve survival. Unfortunately, few studies report sufficiently long follow-up of breast cancer patients.
The prognostic relevance of DTC detection has been demonstrated by us and others [
13‐
18], but only one other study has investigated the clinical significance of DTC detection for prediction of late recurrences by including long-term follow-up data [
19]. For this reason, we evaluated the capacity of DTC detection prior to and after surgery to predict long-term outcome in 191 operable, prospectively recruited breast cancer patients in comparison to other well-established prognostic markers such as mitotic activity index (MAI) and lymph node status in a retrospective analysis [
14‐
16]. To our knowledge, this study involves the longest follow-up reported in breast cancer studies investigating the clinical significance of DTC detection.
Discussion
We and others have previously shown that detection of DTCs in BM samples from operable breast cancer patients is associated with adverse clinical outcomes [
14‐
16,
20,
26,
27]. In this study, we revisited DTC detection data from a previously described cohort, and have now collected extended long-term patient follow-up data. In addition to analysis of all data, we restricted some analyses to include only patients who were alive and recurrence-free after 5 years from surgery. A prognostic impact of DTC detection, but not MAI, could be demonstrated with regard to prediction of late disease relapse.
About 20% of clinically disease-free breast cancer patients experience disease recurrence 7–25 years after their initial diagnosis, even after mastectomy [
28,
29]. In this respect, the average recurrence risk has been calculated at 4.3% per year between 5 and 12 years after postoperative adjuvant therapy [
29], while the relapse risk between 10 and 20 years is about 1.5% per year [
28,
29]. Although the biology behind the late recurrences in breast cancer patients is still not clearly defined, evidence indicates that BM is a common homing target for tumor cells in many types of carcinoma before they re-circulate into other distant sites [
30,
31]. However, for tumor cells to survive in the BM, this microenvironment needs to be permissive for them (the metastatic niche model) [
32]. To facilitate this permissiveness, the primary tumor and secondary sites seem to communicate through exosomes and direct organ tropism, modulate immune evasion, and support mesenchymal-to-epithelial transition, thus contributing to enhanced metastasis by influencing the fate of DTCs (reviewed in [
33‐
35]). Following arrival at the BM, cellular and molecular cross-talk between the DTCs and the microenvironment further directs the DTCs into various native BM niches that also promote cell survival and dormancy [
36,
37]. Recently autophagy has also been revealed as a critical mechanism for both the survival and outgrowth of the DTCs [
38]. In addition to this, a large proportion of DTCs display stem cell–like features, which confer resistance to cytostatic therapy and contribute to enhanced cell survival [
39]. When considering that all these factors contribute to survival of latent/dormant DTCs it reveals the molecular complexity of breast cancer, which further challenges both the choice and the success of adjuvant treatment for long-term survival.
To our knowledge, our study has the longest median follow-up reported in an analysis of DTCs in breast cancer. Only one other study has analysed the prognostic value of DTC status with regard to relatively long follow-up. In that study, 189/350 (54%) patients, of whom 31% were DTC-positive, relapsed during a median 12.5 years of follow-up [
17]. This is comparable to 33% of the DTC-positive patients relapsing in our study. However, in their study, DTC detection was a significant independent prognostic factor for prediction of early relapses occurring 0–5 years from surgery and not for late relapses (> 5 years from surgery) [
17], in contrast to our findings.
We show that the presence of DTCs in BM before surgery is a significant predictor of late recurrences and thus reduced systemic recurrence-free and breast cancer–specific survival in operable breast cancer patients. Stratification further demonstrated that pre-operative DTC status was particularly predictive of reduced survival in postmenopausal women (
p = 0.027), patients with ER-positive disease (
p = 0.007), lymph node involvement (
p = 0.008), and large tumors (
p < 0.001) by the log-rank test. Hence, our results support the fact that ER-positive patients are at particular risk of experiencing late recurrences, and extended endocrine therapy from 5 to 10 years is now recommended for this patient group [
40]. A recent report from the International Breast Cancer Study Group, in which the hazard rates of breast cancer recurrence were estimated from 4105 breast cancer patients and 24 years of follow-up, also showed that the hazard for experiencing late relapse remains elevated and fairly stable beyond 10 years in ER-positive patients [
41]. Other studies also support this conclusion (reviewed in [
36,
42,
43]). However, because the ER-positive patient group is heterogeneous, differences have further been demonstrated between pre- and post-menopausal women based on molecular characteristics. In this respect, it has been specified that most cases of late recurrences arise in postmenopausal ER-positive women aged 60 years or older [
42,
44]. Nevertheless, our study did not confirm ER-status as an independent prognostic factor, in contrast to pre-operative DTC status.
Previously, we and others have shown that both MAI scoring and DTC detection give independent prognostic information in operable breast cancer patients [
24]. Because proliferation is a key driver for cancer progression, and substantial variability in Ki67 scoring is well known to occur [
45], we wanted to extend our study to include investigations of MAI as a marker for prediction of late recurrences. High proliferation is associated with a more aggressive disease, so one prediction would be a higher relapse rate among these patients, as well as more frequent systemic disease and thus a positive DTC status. Our data support a significant association between high MAI score and relapse (Fig.
3) but no significant association between high MAI and DTC positivity (data not shown). This finding is probably because patients with high proliferation in the primary tumor largely seem to experience early disease relapse, within 5 years from diagnosis, in contrast to DTC-positive patients, who experience both early and late relapses [
13]. Proliferation as a marker for prediction of early relapse has also been shown in other studies [
46,
47]. Moreover, several molecular multi-gene assays including proliferation markers, among other markers (such as Mammaprint [
48], Oncotype Dx Recurrence Score [
49], Genomic Grade Index [
50], Prosigna PAM50 Risk of Recurrence Score [
51], Breast Cancer Index [
52] and EndoPredict [
53]) also support this. These assays were developed originally to give an overall risk assessment of recurrence by providing prognostic information not contained in the clinicopathological parameters. However, with a few exceptions, these multi-gene assays provide prognostic information restricted only to the first 5 years after the diagnosis (reviewed in [
54]). This situation illustrates the challenges of accurate classification of primary breast tumors for prediction of late recurrences and suggests that DTC assessment may supplement primary tumor diagnostics in prognostic stratification. A few studies comparing the risk assessment by multi-gene assays and presence of DTCs in the bone marrow of operable breast cancer patients also support this as they did not find any association between them [
55,
56]. On the contrary, another study did show that DTC detection was associated with a high Oncotype DX recurrence score [
57]. Further studies are warranted.
Characterisation of DTCs and CTCs has revealed a challenge in current adjuvant treatment: the choice of targeted/adjuvant therapy in almost every solid cancer is largely based on an initial tissue biopsy obtained from the primary tumor. However, primary tumor characteristics do not necessarily reflect the characteristics of the metastasising DTCs and CTCs due to tumor cell heterogeneity and acquired evolutionary changes in the DTCs/CTCs during treatment. This has been demonstrated in several breast cancer studies, especially with regard to HER2 and ER status [
58‐
61]. HER2-positive DTCs/CTCs have been detected in patients with an apparently HER2-negative primary tumor, resulting in patients who in fact are eligible for HER2-based treatment [
62,
63]. The same has been shown for ER status [
60,
61,
64,
65]. DTCs and CTCs may be ER-negative and PR-negative despite originating from a hormone receptor–positive tumor, possibly explaining the failure of endocrine therapy in a subset of ER-positive patients and vice versa. To overcome the issue of tumor cell heterogeneity, it has in the recent years been much focus on detection of circulating cell-free tumor DNA (ctDNA) from plasma of cancer patients. ctDNA is extracellular DNA that may originate from apoptotic and necrotic tumor cells in the primary tumor, metastatic lesions, or CTCs/DTCs in the circulation. In this respect, the ctDNA pool should be representative of the total tumor burden in an individual cancer patient, and several studies have shown both a prognostic and a predictive value of ctDNA detection in breast cancer patients (e.g. [
66,
67]). Because ctDNA analysis is easily performed, without the need for enrichment and isolation of rare cancer cells, it is likely to be the preferred option for genotyping and monitoring of treatment response in the future. Further investigations are, however, needed to elucidate whether ctDNA assessment can predict late recurrences of breast cancer similarly to or better than DTC detection. Nevertheless, analyses of CTCs and DTCs provide a unique opportunity for in-depth assessment of viable metastasising tumor cells and their interaction with the tumor microenvironment, providing access to information that cannot be revealed using ctDNA.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.