In this study, we showed that HFD restricted to peripuberty reduced the latency of DMBA-induced mammary tumors and led to tumors with characteristics very similar to those occurring in mice fed a continuous HFD. Notably, the tumors occurring after HFD limited to peripuberty shared characteristics with basal-like human breast cancers. The incidence of basal-like, triple-negative, adenosquamous tumors was significantly increased among the short-latency tumors. On the basis of genomic and immunohistochemical analysis, adenosquamous human breast carcinomas are a rare variant of basal-like carcinomas [
16]. It is notable that the incidence of adenosquamous tumors in young, ovary-intact mice temporally parallels the increased occurrence of human basal-like breast cancer in younger women [
3]. The short-latency tumors showed elevated rates of proliferation, increased numbers of macrophages, and enhanced vascularization. Increased proliferation, increased hyperplasia, and increased numbers of macrophages were observed in mammary glands before the occurrence of tumors, implicating these characteristics as plausible effectors of tumor promotion as a result of peripubertal HFD exposure.
HFD and tumor latency
Apart from adenosquamous tumors, all other histotypes also had shorter latency in HFD mice than in LFD mice. Glandular, cribriform, and papillary tumors did not show reduced latency in HFD-LFD mice and had a similar incidence in mice on continuous HFD and LFD. Because both continuous HFD and LFD-HFD mice received quite lengthy HFD exposure, the length of treatment is unlikely to be the explanation for this difference in latency for nonadenosquamous tumors in HFD mice. Rather, the reduced latency of nonadenosquamous tumors in HFD mice is likely dependent upon a puberty-specific effect of HFD that additionally requires adult HFD exposure to promote these tumors. We speculate that the increased proliferation observed in normal mammary epithelium of both continuous HFD and HFD-LFD mice, coupled with elevated angiogenesis that is observed only in continuous HFD, is responsible. In our earlier studies [
6], we found that
Vegfa expression was increased in the mammary glands of mice fed continuous HFD, and this may play a role in elevated angiogenesis.
Interestingly, LFD-HFD tumors showed trends toward increased angiogenesis and macrophage recruitment that did not reach statistical significance, which suggests that HFD exposure limited to adulthood also affected these tumor characteristics, although to a lesser extent than peripubertal HFD exposure and not translating to increased tumor incidence or reduced tumor latency.
Gene expression characteristics of tumors
qRT-PCR showed similar patterns of gene expression in both early HFD-LFD and HFD tumors, in accord with their similar phenotypic characteristics. Regulation of the same genes was observed across adenosquamous and epithelial histotypes. Microarray analysis also strongly supported the similarity of early HFD and HFD-LFD tumors, showing that all of these tumors cluster together, in contrast to tumors from mice on other dietary regimens and/or tumors that occur with longer latency. Notably, only mice exposed to HFD during puberty developed early tumors. Interestingly, although peripubertal HFD particularly promoted the occurrence of adenosquamous tumors, the gene clusters associated with early tumors in the microarray analysis were drawn from epithelial as well as adenosquamous tumors. Indeed, epithelial carcinomas also showed reduced latency in mice fed continuous HFD. Only one late-occurring tumor clustered with an otherwise uniform collection of early tumors. This tumor had an adenosquamous phenotype, whereas two other adenosquamous tumors clustered with late tumors. It is possible that the adenosquamous phenotype contributes to the observed early signature, but the uniform clustering of early tumors vs. the more diverse clustering of adenosquamous tumors suggests a temporal rather than a histological signature for this cluster. Collectively, the data from qRT-PCR of specific RNAs and from microarrays are most consistent with a gene expression signature for early-occurring tumors, regardless of histology. It remains to be determined whether the observed pattern of gene expression is reflective of peripubertal HFD or early occurrence of the tumors.
Examination of the existing literature reveals that, in addition to DMBA-induced mammary tumors, other mouse models that show a similarly high presentation of squamous-like mammary tumors are the
Brg1
+/−
[
17] and
Pik3ca-H1047R [
18] models, the squamous tumors of which were shown by gene set analysis to be similar to claudin-low human breast tumors [
19]. It remains to be determined whether the adenosquamous tumors identified in our study are similar to claudin-low human breast tumors in their pattern of gene expression.
HFD and HFD-LFD tumors showed significant similarities with regard to increased proliferation, increased angiogenesis, and increased macrophage recruitment, which indicates that peripubertal HFD treatment had a lasting effect on tumor phenotype. Consistent with enhanced proliferation, Ingenuity Pathway Analysis of the upregulated microarray gene cluster associated with the early occurring HFD and HFD-LFD tumors highlighted canonical pathways associated with proliferative processes (i.e., G
1/S checkpoint regulation, G
2/M DNA damage checkpoint regulation, cyclins and cell cycle regulation, antiproliferative role of TOB in T cell signaling, mTOR signaling, purine nucleotides de novo biosynthesis II, cell cycle control of chromosomal replication, and molecular mechanisms of cancer). Also consistent with enhanced proliferation, downregulated pathways associated with suppressing proliferation and/or increasing apoptosis of breast cancer cells (i.e., FXR/RXR activation and LXR/RXR activation) [
20‐
22] were also identified. Other downregulated pathways are consistent with anti-inflammatory processes (i.e., coagulation system and acute-phase response signaling). The early HFD tumors showed higher levels of Arg1-positive macrophages [
6], indicative of M2 or alternative anti-inflammatory activation. Although early HFD-LFD tumors did not show elevated levels of M2 macrophages, these pathways were not as robustly downregulated in those tumors. Additionally, pathways associated with genotoxic stress (i.e., GADD45 signaling, DNA damage-induced 14-3-3σ signaling, and EIF2 signaling) were identified among upregulated genes. This may be associated with DNA damage resulting from exposure to the mutagenic carcinogen DMBA.
qRT-PCR showed upregulation of
Ntf3,
Trp53,
Ccnd2,
Ctnnb1,
Brca1,
Apaf1, and
Bmp7 RNAs and downregulation of
Bmp10 RNA in both early HFD and early HFD-LFD tumors. We previously found that
Trp53,
Bmp7,
Ctnnb1, and
Bmp10 identified Ingenuity canonical pathways for basal cell carcinoma signaling and role of NANOG in mammalian embryonic stem cell pluripotency [
6]. This suggests that the similarity between HFD early tumors and basal-like breast cancer [
23] is reiterated in the HFD-LFD early tumors. In our prior studies [
6],
Trp53,
Ctnnb1, Ccnd2,
Brca1, and
Apaf1 also identified Ingenuity canonical pathways for p53 signaling and GADD45 signaling. This confirms our findings in Ingenuity Pathway Analysis of the microarray clusters.
Trp53,
Ctnnb1,
Bmp7,
Bmp10,
Ccnd2,
Apaf1, and
Brca1 additionally identified the Ingenuity canonical pathway for molecular mechanisms of cancer, again confirming our findings from Ingenuity Pathway Analysis of the microarray clusters.
Ntf3 and its receptor
Ntrk3, though they do not identify a canonical pathway, are overexpressed in a significant proportion of human breast cancers, particularly in basal-like breast cancers (11 % amplified + upregulated in basal-like PAM50 breast cancers vs. 5 % in all breast cancers, 6 % in luminal A/B, and 0 % in Her2+) [
24,
25]. Neurotrophins and the p75 neurotrophin receptor are expressed in human breast cancers and are implicated in promoting angiogenesis, tumor growth, invasion, resistance to apoptosis, and resistance to anoikis in triple-negative breast cancer [
26‐
28]. Interestingly, neurotrophin expression is increased in the brains of mice fed 60 % HFD, suggesting its upregulation in early-occurring continuous HFD and HFD-LFD tumors may also be diet-induced [
29].
The increased expression of
Ccnd2 (cyclin D2) is discordant with enhanced tumorigenesis of early continuous HFD and HFD-LFD tumors, as loss of cyclin D2 expression is frequent in breast cancers [
30] and cyclin D2 has been considered to be a tumor suppressor. However, transgenic overexpression of cyclin D2 does block lobuloalveolar development [
31], and perhaps
Ccnd2 overexpression in our system could be viewed as suppressing differentiation. It is noteworthy that
Ccnd2 expression may be specifically elevated in poorly differentiated breast cancer cells that exhibit features of epithelial-mesenchymal transition and a higher potential for metastasis [
32]. Examination of The Cancer Genome Atlas database revealed that
Ccnd2 expression was altered (mainly amplified or upregulated) in 16 % of basal-like PAM50 breast cancers vs. altered (mainly downregulated) in 4 % of luminal A/B and 0 % in Her2+ breast cancers [
24,
25].
In regard to increased
Ctnnb1 (β-catenin) expression, the Wnt/β-catenin pathway is involved in normal mammary gland proliferation and development and is associated with poor prognosis in breast cancer [
33]. Elevated
Ctnnb1 expression may activate this pathway. We found that β-catenin protein and activation levels were elevated in all adenosquamous tumors regardless of diet, and thus this is unlikely to be a factor in their shortened latency on continuous HFD and HFD-LFD. β-catenin levels were also elevated in the nonadenosquamous continuous HFD tumors, as well as in nonadenosquamous switched-diet tumors. Thus, elevated β-catenin is associated with adenosquamous tumors regardless of diet, and HFD exposure in any life period results in elevated β-catenin across the various other nonadenosquamous tumor histopathologies. Because both HFD-LFD and HFD elevate β-catenin but HFD-LFD does not shorten the latency of nonadenosquamous tumors, it is unlikely that β-catenin is key to driving shortened latency in the nonadenosquamous tumors from HFD mice. As mentioned above, increased angiogenesis is a more likely mechanism, as it requires continuous HFD exposure.
Characteristics of the mammary gland before the occurrence of tumors
We previously reported that, before tumor development, continuous HFD mice exhibited increased proliferation in normal mammary gland structures and hyperplastic lesions, as well as increased incidence of abnormal hyperplasia, associated with increased tumorigenesis [
6]. The pretumor mammary glands of HFD-LFD mice were similar in all of these characteristics, and these factors were all likely important contributing factors in promoting tumor development after peripubertal HFD exposure. In contrast, macrophage recruitment was increased in all treatment groups that received HFD, regardless of timing, whereas increased angiogenesis required continuous HFD exposure. These latter results suggest that, if macrophage recruitment plays a role in HFD tumor promotion, it is likely through interaction with an effect specific to peripubertal exposure or with a property of the gland at this stage of development. With regard to angiogenesis, only continuous exposure to HFD was sufficient for this, and thus it was not a contributing factor to the peripubertal HFD promotional effects.
Gene expression analysis in pretumor mammary glands showed that all growth factors and chemokines observed to be elevated before tumor development in continuous HFD mice at 13 weeks of age (i.e.,
Tgfa,
Ccl1,
Ccl17,
Ccl20,
Ccl22, and
Tgfb1) were elevated in LFD-HFD mammary glands at this time. Because LFD-HFD mammary glands do not develop early tumors, this indicates that these factors are unlikely to be factors specific to peripubertal exposure or to be responsible for the enhanced proliferation observed in normal HFD-LFD mammary glands. Only
Tgfa was elevated in HFD-LFD mammary glands at a time after the switch to LFD. If increased
Tgfa levels were sustained throughout the tumorigenesis period (up to 45 weeks of age) in the HFD-LFD group and did not require continued HFD exposure, this could explain, at least in part, the promotional effect of limited peripubertal exposure to HFD.
Ctnnb1 (β-catenin) expression was also increased in pretumor mammary glands of HFD-LFD mice and showed a trend toward increased expression in continuous HFD mice, indicating another early effect of peripubertal exposure to HFD. It is noteworthy that TGFα can activate β-catenin [
34]. This further suggests a plausible role for TGFα in HFD-enhanced proliferation. However, TGFα itself is certainly not sufficient to promote proliferation, as elevated
Tgfa expression and proliferation are dissociated in mice exposed to HFD only in adulthood (i.e., LFD-HFD). It may be that peripubertal HFD exposure induces another growth factor not assayed here, perhaps through the action of TGFα itself; that peripubertal HFD exposure induces long-lasting changes in the regulation of proliferation (e.g., epigenetic effects); or that peripubertal TGFα interacts with a specific population of cells not present in the adult gland (e.g., stem cells). The length of exposure to elevated levels of TGFα is a less likely explanation, as brief peripubertal exposure and lengthy continuous exposure to HFD elicit similar increases in
Tgfa expression. The pubertal window of exposure seems critical regardless of whether elevated
Tgfa expression is essential for enhanced proliferation. It is noteworthy that there was no overlap between the genes and pathways identified in pretumor mammary glands by expression analyses at 3 or 4 weeks on HFD and those identified at 10 weeks on HFD, as well as between those identified at any pretumor time point and in early HFD tumors [
6].
Interaction of DMBA and HFD
We previously reported that in the absence of DMBA, at 3 weeks on HFD, there was a significant transient increase in eosinophil recruitment to the periepithelial stroma, as well as transient increases in
Ccl3,
Ccl24, and
Il4 gene expression; a twofold increase in mammary epithelial cell proliferation; and robust upregulation of
Tnfs11 (receptor activator of nuclear factor κB ligand) gene expression at 4 weeks on HFD [
6]. However, in the present study, in the absence of DMBA at 10 weeks on their diets (13 weeks of age), only increased macrophage recruitment was observed among continuous HFD and HFD-LFD mice. Importantly, this highlights the long-lasting effect of peripubertal HFD to cause higher levels of macrophages that are maintained after a switch to LFD. None of the genes identified as HFD-regulated in DMBA-treated mice were regulated by HFD in the absence of DMBA. Among those genes that were regulated by only HFD with DMBA treatment were
Tgfb1,
Ccl1,
Ccl17, and
Ccl22, whose products are all associated with the recruitment and function of immunosuppressive Treg cells [
35‐
37]. Thus, the interaction of DMBA with HFD may influence tumorigenesis by immune modulation in addition to its activity as a mutagen. We also noted that DMBA, independent of diet, had a profound effect on mammary gland development. DMBA-treated mammary glands retained a pubertal morphology, evidenced by the presence of numerous terminal end buds and limited ductal growth. Because the pubertal gland is undergoing rapid proliferation, it is likely that the retention of a pubertal developmental state resulting from DMBA treatment contributed to the increased proliferative effects of HFD at 13 weeks of age, as previously reported [
6].
An additional consideration regarding the interaction of HFD and DMBA is the possibility that HFD could increase the metabolism and activation of DMBA, thereby increasing the “effective dose” of DMBA. This could in part contribute to the increased tumorigenesis observed with HFD. It is also of interest to note that increasing doses of DMBA can also increase the proportion of adenosquamous mammary tumors [
38]. However, it is noteworthy that the increased incidence of early tumors and adenosquamous tumors occurred mainly as a result of peripubertal exposure to HFD, indicating an important life-stage period of increased susceptibility to an HFD’s effects.
Another consideration regarding HFD effects in the present study are its potential contribution to tumorigenesis through increased caloric density. Lard is the major animal fat in our HFD, and it contributes to caloric density. Researchers in other studies have compared 45 % and 60 % lard HFD, which differ in caloric density, on tumorigenesis in C3(1)-T
Ag mice [
39], and they found the same increases in tumorigenesis for both 45 % and 60 % lard HFD compared with 10 % LFD. This suggests that increased caloric density per se was not the only contributor to increased mammary tumorigenesis. However, regardless of caloric density, excess lard is apparently a risk factor for the mice; it may be the fat, or it may be the extra calories.
It is noteworthy that mice in the present study which were started on HFD in peripuberty did not exhibit a significant increase in body weight. Thus, HFD had a promotional effect on tumorigenesis in normal-weight mice. Interestingly, mice started on HFD in adulthood did gain significant body weight. However, despite the increase in body weight, this did not promote tumor development as measured by increased incidence, decreased latency, or increased tumor cell proliferation in the LFD-HFD tumors. Also to be considered are the metabolic consequences of HFD with regard to the development of prediabetic or diabetic conditions. In this regard, despite the increase in body weight observed in the LFD-HFD mice, there were no significant differences in blood glucose or insulin levels between diet regimens at 4 weeks after diet switches or in tumor-bearing HFD-LFD and LFD-HFD mice.
Studies on the effect of HFD on tumor development without obesity have been investigated in other mouse mammary cancer models. Results vary by age at diet initiation and by tumor model. In two studies of the effects of HFD initiated at 4 weeks of age in mice overexpressing
HER2/Neu in the mammary gland [
40,
41], HFD promoted tumor development by increasing tumor incidence without increasing tumor cell proliferation, and there was no insulin resistance or hyperinsulinemia. In contrast,
HER2/Neu-transgenic mice fed HFD starting in adulthood showed no difference in tumor latency, incidence, or metastasis [
42]. In the BALB/c 4 T1 tumor transplant model, mice were started on HFD at 4 weeks of age and tumor cells were transplanted after 16 weeks on diet [
43]. Tumor weight and number of metastases were significantly increased by HFD. In contrast, there was no promotional effect when HFD was initiated in 10-week-old adult mice. The promotional effect observed when diet was started at 4 weeks of age is similar to our present results, showing an association of HFD with increased macrophage infiltration, angiogenesis, and cellular proliferation, as well as increased levels of a number of inflammatory factors.