Background
Breast cancer (BC) is the most common and fatal malignancy in women [
1]. Accumulating evidence supports the presence, within BC, of a subpopulation of tumor cells, named cancer stem cells (CSCs). These cells exhibit stem-like features, such as self-renewal, differentiation capacity, and are believed to represent the subpopulation responsible for the tumor-initiating activity and the resistance to antineoplastic agents [
2,
3].
In vivo, CSCs sustain tumor growth, reproducing the heterogeneity of the original tumor from which they are derived [
4]. According to the current carcinogenesis theory, BC development and recurrence is driven by CSCs [
5], and these cells represent the main pharmacological target for tumor eradication. Breast CSCs were initially characterized from surgically removed human tumors, although their isolation was possible only in a small percentage of postsurgical specimens [
6]. However, since this first seminal study, most of the research on breast CSCs was carried out in established cancer cell lines [
7,
8], which were reported to contain putative CSC subpopulations. Conversely, only few studies were performed using cells isolated from tumor samples [
9,
10]. This limitation was likely a consequence of the CSC rarity within the tumor mass and the usually extremely small post-surgical specimens available for
in vitro studies. A possible pitfall using cells expressing CSC signatures but isolated from continuous BC cell lines, is that they might include subsets of cells adapted to prolonged
in vitro culture in the presence of high serum concentration that, overtaking the majority of the tumorigenic subpopulations, inadequately represent cancer cell heterogeneity. Moreover, due to genotypic and phenotypic alterations, these cells often show different drug responsivity from tumors
in vivo [
3,
11].
The human BC cell subpopulation identified as CSCs is characterized by CD44
+/CD24
low/− phenotype, the ability to grow
in vitro as mammospheres maintaining a constant percentage of stem cells, high tumorigenicity
in vivo [
6,
9], developing serially transplantable tumors in immunodeficient mice [
12], indicative of long-term self-renewal ability [
13,
14]. Moreover, several BC CSC features are also relevant to metastasis, such as high motility, invasiveness, and resistance to apoptosis and drug treatments.
Recently, comparative oncology emerged as a relevant tool for pharmacological development in human cancer research. Spontaneous pet tumors represent important pre-clinical models of human cancers retaining the heterogeneous nature of tumors and allowing the validation of treatment strategies that will result beneficial to both human and animal patients [
15,
16]. These tumors, which develop in immunocompetent animals, at odd with those experimentally induced in laboratory rodents, display genetic, histopathological and biological features similar to the human counterpart, as well as the metastatic pattern and the response to therapy [
17]. For example, spontaneous canine mammary carcinomas (CMCs) retain inter- and intra-tumor heterogeneity, as human cancer [
18-
20] but, due to the shorter life-span of dogs, they allow the evaluation of the natural course of the tumor and its pharmacological modulation after a shorter lag time than that required in human clinical trials. Thus, CMC is considered a reliable comparative model for human BC [
21]. CMC is the most common neoplasm of female dogs, representing 50-70% of all tumors [
22], and multiple deregulated genes and signaling pathways (PI3K/AKT, KRAS, PTEN, Wnt-beta catenin, MAPK, etc.) identified as responsible for its development, nicely resemble those observed in humans [
19]. For example, the expression level of epidermal growth factor receptor (EGFR) in CMCs affects clinical prognosis [
23]; HER-2 overexpression, occurring in about 20% of CMCs as in BC [
24], or the loss of estrogen (ER) and progesterone (PR) receptors [
25] are related to tumor progression. Moreover, triple-negative CMCs (lacking ER, PR and HER-2) show clinical-pathological characteristics associated with unfavorable prognosis, similarly to the triple-negative phenotype in women [
26].
Because of the limited source of primary human BC tissues due to early diagnosis and multiple histopathological analysis required during and after surgery, and the lack of in vivo preclinical models that accurately reflect patients’ tumor biology, the study of pet spontaneous tumors may represent an innovative approach. However, this model is still underused and, in particular, studies on the role of CSCs in tumor development and treatment are lacking.
In veterinary research, putative CSCs have been identified in canine osteosarcoma, glioblastoma, acute myeloid leukemia, hepatocellular carcinoma [
27-
31], as well as in feline mammary carcinomas [
32]. CSC-like subpopulations were isolated and partially characterized from canine mammary cancer continuous cell lines [
33-
35], mainly relying on
in vitro observations, such as spheroid formation, cell surface antigens and aldehyde dehydrogenase (ALDH) activity, whereas isolation of CSCs from spontaneous canine mammary tumors have been described only in few studies [
36]. Immunodetection of cells with CD44
+/CD24
− phenotype in canine mammary tumor tissues, similarly to human BC CSCs, has been also reported [
37], and CD44 expression has been associated with proliferation of cultured canine cancer cells [
38]. Moreover, canine CSCs, isolated from the REM134 cell line, are resistant to common chemotherapeutic drugs and radiation, exhibiting epithelial-mesenchymal transition (EMT) phenotype [
34].
Metformin is the first-line hypoglycemizing agent used for the treatment of type 2 diabetes (T2D) due to its efficacy and safety profile [
39]. Epidemiological studies reported that metformin-treated T2D patients show reduced cancer incidence and mortality; furthermore metformin therapy seems to improve the clinical outcome of diabetic patients with cancer and to exert a protective anticancer effect in non-diabetic patients [
40,
41]. Thus metformin’s antitumor properties are currently tested in several clinical trials, mainly focusing on BC [
42,
43]. Preclinical
in vivo studies reported that metformin reduces growth of BC xenografts in mice [
44,
45], and directly inhibits the proliferation of several BC [
46,
47] and other tumor [
48] continuous cell lines, mainly interfering with CSC proliferation. However, in all these studies the effects of metformin, alone or in combination with doxorubicin or trastuzumab, were mainly evaluated in CSC-like derived from established lines [
49-
51].
Thus, the evidence of metformin activity in human BC CSCs is still limited, and a comparative approach studying CSCs from spontaneous dog tumors presents several advantages, including the retention of intra-tumor cell heterogeneity, an extremely relevant issue to identify pharmacological approaches with higher predictive validity when translated from preclinical to clinical setting. Moreover, since these tumors are often not treated before surgery, comparative oncology provides the unique opportunity in a preclinical model to map the nascent BC biology, without modifications induced by therapy pressure. Since CSCs are generally highly resistant to chemotherapy, drugs that successfully target this subpopulation may represent an effective therapeutic approach, and the analysis of efficacy on CMC may pave the way to the identification of clinically useful compounds in humans.
The aim of this study was to establish cell cultures enriched in CSCs from spontaneous CMCs, in order to provide a cellular model that may better reflect BC heterogeneity, pathogenesis and drug responses. Moreover, we tested the effects of metformin on CSCs isolated and characterized from spontaneous CMCs, providing evidence that these cells are highly responsive to in vitro and in vivo metformin treatment.
Discussion
The hierarchical model of carcinogenesis implies that only a small subset of tumor cells, named CSCs or tumor-initiating cells, drives tumor development and progression, determining drug responsiveness. Conversely, differentiated tumor cells, incapable of long-term self-renewal but representing the majority of the cells within hematologic and solid cancers, are not tumorigenic. Although most studies addressing this issue relied on established cancer cell lines rather than primary cultures, it is now clearly evident that putative CSC identification and usefulness is strictly coupled with powerful and systematic
in vitro isolation and
in vivo transplant of CSCs derived from fresh tumor tissues [
64].
In the present study, we isolated and characterized CSCs from CMC surgical tissues, by selection of primary CMC cultures in stem cell permissive, serum-free medium for several passages in vitro. CMC-derived stem cell-like spheroids express CD44, as well as other BC related proteins, such as EGFR and CXCR4, and showed high resistance to DOX when compared to the corresponding CMC differentiated cells. Xenografts in NOD/SCID mice demonstrated that CMC CSCs derived from mammosphere disaggregation, successfully initiate tumor formation.
Breast cancer remains a major clinical challenge with high mortality both in humans and companion animals [
1,
25]. Advancement in understanding tumor biology and public awareness campaigns led, in most cases, to early BC diagnosis and neo-adjuvant treatments were developed in women. While the better clinical management significantly improved the overall prognosis of BC patients, only small fragments of fresh tissues, frequently already exposed to intensive cytotoxic therapy, are available from excised human tumors for experimental purposes, also because large amounts are required for intra- and post-surgery histopathological analyses. On these premises, a major strength of our study is the use of spontaneous dog tumors as a source of biological material necessary to isolate mammary CSCs. Indeed, to improve the knowledge of
in vitro and
in vivo biological features of CSCs and their drug responses, allowing the translation of preclinical findings into effective human clinical trials, cell models that faithfully mimic BC cell heterogeneity are an absolute requirement. Spontaneous tumors of companion animals, such as CMC, are rather frequent in the clinical veterinary practice [
22] and thus may provide a unique opportunity as a model for human cancer translational research [
15]. In contrast to experimental tumor models in mice, CMCs develop naturally, reproducing the same environmental and genetic aetiology as occurs in humans, grow in immunocompetent organisms [
67] and share strong clinical (e.g. hormonal dependence, age of onset, histological appearance, prognostic factors, course of the disease) and molecular (e.g. tumor genetics, overexpression of steroid receptors, proliferation markers, EGF, p53 mutations, metalloproteinases, cyclooxygenases) similarities to human BC [
20,
22].
Additionally, human and canine mammary tumors show similar responses to conventional anticancer agents and, more importantly, both display inter/intra-patient tumor cell heterogeneity [
68]. The enrichment in CSCs has been largely demonstrated in established human and canine mammary cancer cell lines [
5]. Here, we reinforce the hypothesis of the stem cell basis for mammary tumors, achieving isolation of such a population from surgical samples. Features of CSCs include: (i) self-renewal capacity, (ii) stem marker expression, (iii) ability to reproduce the original tumor after xenografting, and (iv) chemo-resistance. Therefore, we verified these prerequisites in CSCs derived from CMCs. Cell surface marker CD44 [
6], and the expression of the enzyme aldehyde dehydrogenase (ALDH), partially overlapping within this subpopulation [
69], were assumed as criteria for CSC-enrichment and stem signatures in human BC, correlating to the tumor-initiating ability. However, it is currently debated whether these markers univocally identify CSCs in all BCs [
70,
71]. Breast CSC plasticity induced by tumor microenvironment, which allows these cells to undergo reversible EMT, may influence distinct marker profiles [
72]. Nevertheless, CD44 was identified as major target using CD44 antibodies [
73] or inducing CD44 down-regulation [
74,
75], to inhibit CSC proliferation and migration and overcome drug resistance.
In dogs, limited studies on mammary CSCs are available, thus their phenotyping is mainly derived from commonly employed human markers [
33,
36], reproducing similar controversies: spheres from primary tumors show the expression of CD49f (integrin α6) and CD29 [
36] while spheres derived from canine cell lines express CD44 and CD133 and/or Sox2 and Oct4 [
33]. Moreover, CD44 was associated to proliferative activity of cultured canine cancer cells [
38,
71]. CMC cultures, analysed after
in vitro enrichment in stem-like cells, display CD44 expression in a high number of cells, while, in the tumor of origin, CD44
+ cells are confined in randomly spotted tissue areas, as described for human samples [
71], explaining, as previously discussed, the apparent high percentage of CD44 negative tumors. However, although this is a non-defining criterion for CSC identification [
76] and we did not select CMC CSCs on the basis of markers’ expression, we observed a depletion/decrease of CD44
+ cells upon CMC CSC differentiation (shifting cell cultures in serum-containing medium), confirming that this marker labels some CSC-like populations in culture. In addition, we analysed the expression of genes associated, in human and pet mammary cancers, with stem/progenitor cell survival, self-renewal and tumor aggressiveness, such as EGFR, ER-α and CXCR4 [
54,
77,
78]. Similarly to the original tumors, CMC mammospheres quite homogenously express EGFR confirming its relevance in mammary cancer biology, and CXCR4, whose signalling regulates BC stem cell activities and metastatic potential [
54,
79]. Conversely, ER-α positivity was less commonly detected in isolated CMC stem-like cells and in tumor sections. The pattern of expression of these proteins did not change in the corresponding differentiated CMC cultures, showing that both the bulk of CMC cells and CSC-enriched cultures retain
in vitro the same distinctive factors, as master regulators of tumor formation and growth. Thus, CSC cultures isolated from CMCs reproduce the tissue heterogeneity, covering different histopathological types and maintaining in culture the phenotype observed
in vivo.
CSC biomarker expression [
80] and the ability to growth in serum-free medium as non-adherent spheroids [
81] are variably identified features in human and canine mammary tumors, reflecting intra- and inter-tumor heterogeneity but do not systematically define CSCs [
82]. Thus, functional validation, by means of
in vivo tumorigenicity experiments, is actually the more reliable tool to corroborate the definition of cell populations as CSCs (tumor-initiating cells). We demonstrate that CSCs isolated from CMC retain tumor-initiating activity, by serial transplantation in NOD-SCID mice (CSCs were xenografted, recovered and re-transplanted to form new tumors in new recipient mice) achieving about 100% of take rate after both 1
st and 2
nd injection, reforming the heterogeneous population of CMC cells within the xenografts. Another biological property generally ascribed to CSCs is drug resistance, as they more efficiently survive therapy than differentiated cancer cells. Indeed, the major challenge in BC treatment is the targeting of CSCs usually refractory to conventional drugs both in humans and in dogs [
34]. This underlines the need of novel preclinical models to test drug effects on CSC biological features such as increased drug-efflux and DNA repair ability, and resistance to apoptosis, that contribute to drug resistance. In fact, using these mechanisms CSCs survive therapies contributing to recurrence and progression, after the initial remission caused by the differentiated tumor cell death [
83]. Drug-resistance of CMC CSCs is therefore similar to that observed in human BC stem cells, whose content within the tumor mass is increased by chemotherapy [
83], and in CSCs isolated from a CMC cell line [
34]. Therefore, although cytotoxic treatments reduce the bulk of the tumor, they may not affect the most important target: the CSCs. In this study we developed a preclinical model reproducing this
in vivo condition. CMC CSCs survived DOX treatment, at odd of differentiated cultures isolated from the same tumors, likely due to transporters that pump out the drug from the nuclei or outside the cells.
The need of drugs effectively targeting CSCs inspired new approaches for anti-cancer drug discovery and in particular the so-called “drug repositioning”. It was reported that the biguanide metformin specifically inhibits self-renewal and proliferation of CSC from several tumors [
66,
84,
85], including BC [
49,
51,
86,
87]. Starting from early evidence using different biguanides [
88-
90], the antitumor potential of metformin is now well-established [
65,
91]. Epidemiological studies showed that diabetic cancer patients may benefit of metformin treatment and, on these bases, several clinical trials are ongoing [
92]. Recently, the inhibition of chloride intracellular channel 1 (CLIC1) activity was identified as a specific molecular mechanism by which metformin affects only CSC viability, sparing normal stem cells [
93], making this molecule particularly interesting as novel anticancer drug. We show that canine CSCs are not sensitive to DOX cytotoxicity, but highly responsive to metformin
in vitro, while differentiated tumor cells are only partially affected by metformin, although highly responsive to DOX. These data confirm that combined treatment with metformin and conventional cytotoxic drugs may provide a therapeutic advantage. In fact, while metformin is clearly preferentially active on CSCs, a successful therapy will require targeting of both undifferentiated CSCs and differentiated non-CSCs [
94], since the reverse transition of non-CSCs into CSC subpopulation was also reported [
95]. Mean metformin IC
50 values obtained in our experiments is about 10 mM, ranging from 0.40 to 31 mM in cells from different tumors, evidencing significant variability in metformin activity among individual dog-derived cells, as expected for the inter-patient tumor heterogeneity, although we remark that viability of CSCs from all 13 CMCs analyzed was impaired. These IC
50 values are in line, and in some cases lower, with those of most of the previous studies reporting antitumor activity of metformin (range 1–30 mM) [
44,
87,
96,
97]. However, metformin concentrations used
in vitro, exceeding those achieved
in vivo in T2D patients (10–30 μM range) [
39], are still a debating issue, due to the concern whether these
in vitro data might be relevant for translation to clinics [
98]. We acknowledge that the concentrations used are higher than metformin steady state plasma levels in T2D patients, however, the discrepancy between clinical and
in vitro conditions could be less significant considering the following factors: (a) metformin concentrations in tissues are several-fold higher than those in blood because of tissue accumulation [
99], thus actual intra-tumor concentrations should compare with
in vitro results; (b) medium supplements, required to maintain tumor cell proliferation in culture (i.e. high concentrations of glutamine and glucose), reduce cell sensitivity to metformin [
4,
100,
101]; (c) tumors often show increased cationic transporters compared their normal counterparts [
91] further favoring tissue accumulation, (d) longer exposure of cell cultures to metformin (15–18 days) shift metformin-antitumor effect to a lower threshold [
44]. In fact, it was reported that in human glioblastoma CSCs, short-term experiments (24-72 h, as here reported) show metformin anti-proliferative activity in the mM range [
44,
93,
102], but increasing drug exposure to 15 days metformin efficacy was evident already at 10 μM concentration [
93]. This latter point might justify higher IC
50 in
in vitro studies, supporting time-dependent mechanisms that significantly differ between
in vitro and
in vivo experimental conditions. This hypothesis was indeed supported by our demonstration that prolonged metformin treatment of CMC xenografts resulted in reduced tumor size and growth arrest and depletion of CSC content, for much lower blood concentrations (about 40 μM) than required
in vitro (about 10mM). The plasma levels obtained treating mice with metformin dissolved in drinking water to attain the dosage of 360 mg/kg body weight/day, which is comparable to therapeutic doses used for T2D as translated to humans (human equivalent dose [
103]: corresponding to ~1,750 mg/day in an average-sized person of 60 kg), are therefore lower than the maximum recommended dose in humans (
http://www.fda.gov/ohrms/dockets/dailys/02/May02/053102/800471e6.pdf).
Thus,
in vivo dosing that induces a highly significant reduction of tumor growth is within human therapeutic range, even if the high safety profile and negligible side-effects of metformin could allow experimental doses over pharmacological concentrations. In very recent phase II and III clinical trials, non-diabetic women with BC received metformin at high dose (2 g/day) as adjuvant therapy [
43,
92], demonstrating that higher blood concentrations can be safely achieved.
Notably, our study demonstrates that the clonogenic potential of cells surviving long-term in vivo exposure to metformin, was markedly reduced as compared to untreated controls, indirectly confirming that metformin preferentially kills CSCs. Moreover, desensitization induced by prolonged exposure to metformin does not occur in CMC CSCs after prolonged in vivo treatment, since metformin still exhibits a strong and consistent antiproliferative action in vitro on cells derived ex vivo from both treated and untreated mice tumors. Above observations support metformin as an attractive agent for chemoprevention and use of low-dose for long period in combination with cytotoxic agents like DOX to kill both CSCs and the bulk of differentiated cancer cells.