Contributed by Myles Brown, January 18, 2017 (sent for review September 8, 2016); reviewed by Mitchell A. Lazar and Donald P. McDonnell)
Epidemiological evidence suggests that vitamin D can protect women from developing breast cancer (BC). This study reveals that vitamin D and its receptor regulate autophagy in both normal mammary epithelial cells and luminal BCs, and suggests a potential mechanism underlying the link between vitamin D levels and BC risk. In addition, this work suggests that vitamin D receptor ligands could be exploited therapeutically for the treatment of a significant subset of BCs.
Women in North America have a one in eight lifetime risk of developing breast cancer (BC), and a significant proportion of these individuals will develop recurrent BC and will eventually succumb to the disease. Metastatic, therapy-resistant BC cells are refractory to cell death induced by multiple stresses. Here, we document that the vitamin D receptor (VDR) acts as a master transcriptional regulator of autophagy. Activation of the VDR by vitamin D induces autophagy and an autophagic transcriptional signature in BC cells that correlates with increased survival in patients; strikingly, this signature is present in the normal mammary gland and is progressively lost in patients with metastatic BC. A number of epidemiological studies have shown that sufficient vitamin D serum levels might be protective against BC. We observed that dietary vitamin D supplementation in mice increases basal levels of autophagy in the normal mammary gland, highlighting the potential of vitamin D as a cancer-preventive agent. These findings point to a role of vitamin D and the VDR in modulating autophagy and cell death in both the normal mammary gland and BC cells.
Breast cancer (BC) is the most common nonskin cancer diagnosed in North American women, accounting for 29% of newly diagnosed cancers in the United States (1). Despite considerable recent progress in the development of BC therapeutics, a significant proportion of patients with BC will eventually develop resistance to therapy and relapse (2), making BC the second most common cause of cancer-related deaths in US women. Several molecular subtypes of BC have been identified: luminal A and B (accounting for 50–60% of BC), basal-like or triple-negative (10–20% of BC cases), and human epidermal growth factor receptor 2 (HER2)-enriched (10–15% of cases) (3). In the United States alone, nearly 41,000 BC-related deaths are estimated for 2015 (1). The risk for developing BC is modulated by a number of factors (age, early menarche, family cancer history, ethnicity, and late menopause). However, modifiable risk factors, resulting from the patient’s behavior and environment, can also impact the relative risk for developing BC. These risk factors include postmenopausal hormone use, alcohol and tobacco consumption, exercise, obesity and weight gain, and dietary components such as vitamin D.
Epidemiological data indicate that circulating vitamin D levels (serum levels of ≥45 ng/mL; achieved with a daily intake of ∼4,500 international units (IU) in winter) protect against BC (4) (1). Furthermore, a prospective cohort study showed that sufficient vitamin D serum levels in patients with luminal BC correlate with better response to therapy as well as improved disease-free survival (5). However, because of the lack of randomized clinical trials on the use of vitamin D to prevent cancer (6), the Institute of Medicine stated in their most recent controversial report (7⇓–9) that no evidence exists to justify an increase in the recommended levels of vitamin D supplementation (600 IU) and called for additional mechanistic studies of vitamin D in cancer models (10). Because most women in North America are deficient in vitamin D (11), elucidating the functional mechanistic roles of vitamin D and of the vitamin D receptor (VDR) in BC models and prevention is of paramount importance.
Vitamin D is a secosteroidal prohormone, and arguably not a true vitamin, because it can be synthesized at sufficient levels in skin, given adequate skin exposure to UV B radiation from sunlight; in fact, humans naturally can make up to 140 ng/mL in their skin per day (12). The “vitamin” misnomer originates from its discovery as the key antirachitic compound in cod liver oil in 1922. Cod liver oil had been used effectively to prevent and treat rickets, a childhood disease characterized by defective calcification of bones (13). Since its discovery, vitamin D is still best known for its role as a key regulator of the homeostasis of calcium and a requirement for bone health. However, growing experimental and epidemiological data point to a wide range of biological properties of vitamin D. The VDR is expressed in a wide variety of tissues unrelated to calcium homeostasis, and vitamin D has pleiotropic effects that range from immune system modulation, to nervous and muscular system regulation, and (notably) to arrest of cancer cell proliferation (14). However, numerous studies have suggested that vitamin D serum levels required to achieve most of these physiological benefits are higher (30–60 ng/mL) than levels required for achieving bone health (∼20 ng/mL) (4, 15).
Vitamin D modulates its biological effects by directly regulating target gene expression through the VDR, a ligand-regulated transcription factor and a member of the nuclear receptor superfamily. Whether synthesized in the skin or ingested, vitamin D requires two hydroxylation steps to become the biologically active hormone, 1,25-dyhydroxyvitamin D3 [1,25(OH)2D3], a form that signals through the VDR. The hormone-bound VDR modulates target gene transcription in response to vitamin D. In cancer models, such changes in gene expression lead to modulation of the cell cycle; arrest of cell proliferation; cell differentiation; and regulation of programmed cell death, including apoptosis and autophagy (16, 17).
Autophagy (macroautophagy) or “self-eating” is a conserved lysosomal event that consists of autophagosome formation by encapsulation of intracellular components by a double membrane; the autophagosomes are then delivered to and fuse with lysosomes, forming autolysosomes, wherein their contents are digested and recycled back into the cytosol (18, 19). Autophagy is multifunctional: It serves as a mechanism for clearing lipids, toxic protein aggregates, and damaged or defective organelles. It is also a prosurvival mechanism in response to starvation and hypoxia. Additionally, it is a mechanism of cell death. The physiological consequences of induction of autophagosomes can vary, depending on whether the induction process is acute or chronic. The role of autophagy in cancer appears to be context-specific: In some cases, autophagy is antineoplastic, whereas in other instances, it seems to promote cancer growth and survival (20⇓–22). However, in normal tissue, autophagy generally serves to protect the tissue from damage and cancer initiation (18). In various cell models, 1,25(OH)2D3 can modulate autophagy (17, 23). We have found that 1,25(OH)2D3 regulates cell death by modulating autophagy in luminal-like BC-cell models, and also in normal mammary gland in mice. This mechanistic work underlines the importance of 1,25(OH)2D3 signaling in normal mammary gland and the role of the VDR as a fine-tuned modulator of autophagy.
Antiproliferative functions for vitamin D have been reported in a variety of cancer cell models, including BC cells (24). We treated different BC-cell lines corresponding to the different molecular subtypes of BC with 1,25(OH)2D3 and observed that it has antiproliferative properties in luminal-like models, but not in mesenchymal basal-like BC or immortalized basal cell models (Fig. 1A). Also, following vitamin D treatment, 1,25(OH)2D3-sensitive cells appeared granular by phase-contrast microscopy, suggesting the presence of autophagy (Fig. S1A, arrows). Because autophagy induction has been reported following 1,25(OH)2D3 treatment in MCF-7 cells (24), we tested for potential induction of autophagy by staining the cellular acid compartments with LysoTracker Red. Fluorescence microscopy confirmed that only the luminal-like models exhibited increased acidic compartments, consistent with autophagy (Fig. 1B). Importantly, our luminal-like panel included the estrogen receptor (ER)-negative cell line MDA-MB-453, indicating that this 1,25(OH)2D3-mediated effect is ER-independent. We transfected MCF-7 luminal-like model cells with an EGFP-LC3 expression vector and confirmed the presence of increased punctae following 1,25(OH)2D3 treatment (Fig. S1B). Sorted MCF-7 EGFP-LC3 cells treated with 1,25(OH)2D3 showed significantly (P ≤ 0.05) higher GFP intensity relative to control cells (Fig. S1B). Western blotting was used to document increased expression and cleavage of the autophagy marker LC3 following 1,25(OH)2D3 treatment, along with a modest increase in Beclin1 (Fig. S1C). Transmission electron microscopy (TEM) confirmed that the increased acidic structures (with a dark appearance following citrate counterstain) were autolysosomes (Fig. 2). Taken together, these data confirm that 1,25(OH)2D3 induces autophagy in luminal-like BC-cell models.
Vitamin D increases the acid compartments and inhibits proliferation specifically in luminal-like BC cells. (A) Cell counts of luminal-like and nonluminal-like BC-cell lines with 10 nM, 100 nM, and 1 μM 1,25(OH)2D3. *P ≤ 0.05 and **P ≤ 0.01 as determined by one-way ANOVA. (B) LysoTracker Red staining of live luminal and nonluminal cell lines. (Magnification: 20×.) The positive control is 8 h of serum starvation, the negative control is carrier control, and treatment is 5 d of 100 nM 1,25(OH)2D3. ctrl, control.
The vitamin D-induced acidic compartment are autolysosomes. Autolysosome formation following 1,25(OH)2D3 treatment exhibiting defining autolysosomal features: double membrane (a), lysosome (b), partially digested organelle (mitochondria) (c), and cytoplasm (*).
Vitamin D increases the acidic compartments and inhibits proliferation specifically in luminal-like BC cells. (A) MCF-7 cells treated with control and 100 nM 1,25(OH)2D3. Arrows point to smaller, granular-looking cells. (Magnification: 20×.) (B) MCF-7 cells transfected with GFP-LC3 expression vector were FACS-sorted for equal GFP expression, plated, and serum-starved for 4 h (positive control), carrier control, or 48 h with 100 nM 1,25(OH)2D3. GFP fluorescence was measured by FACS after 48 h of treatment. *P ≤ 0.05 relative to control as determined by one-way ANOVA. (Magnification: 40×.) (C) Western blot of MCF-7 cells treated for 1, 2, 3, 4, and 5 d with 100 nM 1,25(OH)2D3 blotted against Beclin-1 and LC3A/B antibodies. (D) Autolysosome formation following 1,25(OH)2D3 treatment exhibiting defining autolysosomal features: double membrane (a), lysosome (b), partially digested organelle (mitochondria) (c), and cytoplasm (*). (E) Gene ontology (GO) analysis of the top 800 1,25(OH)2D3-regulated genes in MCF-7 cells treated for 16 h. (F) Gene expression profile of autophagy genes regulated upon 1,25(OH)2D3 treatment in MCF-7 cells. (G) The 1,25(OH)2D3-induced autophagy network in MCF-7 cells. (H) Protein role (proportion) of the autophagy genes induced upon 1,25(OH)2D3 treatment in MCF-7 cells. (I) Autophagy induction is not VDR-level dependent.
Expression analysis was used to identify genes differentially regulated by 1,25(OH)2D3 in MCF-7 cells. Gene ontology analysis of the top 800 regulated genes showed metabolism and cell death to be significantly present (1.07E-13 and 4.32 E-11), along with other biological processes reported to be regulated by vitamin D, such as cellular differentiation and regulation of proliferation (Fig. S1E). We built the 1,25(OH)2D3-induced autophagy network by using the Dijkstra algorithm to calculate the shortest direct path connecting our regulated gene set using MetaCore software (Fig. S1G). The resulting 1,25(OH)2D3-specific autophagy signature present in treated 1,25(OH)2D3 MCF-7 cells was compared against the expression profiles of patients with invasive ductal and lobular breast carcinoma relative to the normal mammary gland [The Cancer Genome Atlas (TCGA) breast dataset]. We found that the induced autophagy profile was present in the normal mammary gland, but was significantly lost (P ≤ 0.05) upon BC progression (Fig. S2A). Moreover, expression of some of the genes from this signature was also modestly, yet significantly, correlated with long-term survival of all patients with BC (Fig. S2B). The interaction between these genes was graphed with the use of Ingenuity Pathway Analysis (Fig. S2C). Interestingly, this signature was also present in human breast stroma (Finak Breast; 59 samples), not just in the ductal or luminal cell compartment, and it was also significantly lost upon cancer progression in stroma (Fig. S2D).
Vitamin D-induced autophagy signature is present in normal mammary gland, but it is lost during BC progression. (A) The 1,25(OH)2D3-specific autophagy signature present in treated 1,25(OH)2D3 MCF-7 cells was compared against the expression profiles of patients with invasive ductal and lobular breast carcinoma relative to normal mammary gland (TCGA breast dataset; Oncomine). (B) The 1,25(OH)2D3-regulated autophagy genes in MCF-7 that correlate with patient survival (TCGA data; Oncomine). (C) The 1,25(OH)2D3-regulated autophagy gene interaction (IPA) in MCF-7 cells. (D) The 1,25(OH)2D3-specific autophagy signature present in normal breast stroma and lost upon BC progression.
Motivated by these findings, we fed a vitamin D-supplemented diet ad libitum to mice (11 IU/g and 23 IU/g vitamin D3 added; TestDiet) to determine whether autophagy levels in the normal mammary gland could be modulated by chronic, orally administered noncalcemic doses of vitamin D (Fig. S3A). A vitamin D-supplemented diet significantly (P ≤ 0.05) modulated basal levels of autophagy in wild-type mice in a dose-responsive manner in the mammary gland at doses that do not induce hypercalcemia (Fig. 3 and Fig. S3). These results also suggest that the levels of dietary vitamin D in the control diet, although sufficient to maintain bone homeostasis, are not sufficient to induce autophagy in the mammary gland.
Vitamin D modulation of autophagy in mouse mammary gland. (A) Immunohistochemistry of mouse mammary gland fed with increasing doses of vitamin D-supplemented diet stained for cleaved LC3. (Magnification: 60×.) (B) LC3 + mammary gland alveolae and ducts significantly increase in the vitamin D-supplemented diet. *P ≤ 0.05 as determined by the stratified random approach (Braun–Blanquet) sampling method followed by Kruskal–Wallis one-way analysis of variance. (C) Punctate LC3 staining pattern. (Magnification: 100×.)
Vitamin D (21 IU/kg) diet supplementation fed ad libitum to mice for up to 6 mo does not induce hypercalcemia. (A) Serum calcium level measurements of control and vitamin D-supplemented (21 IU/kg) mice. (B) Body weight of control and vitamin D-supplemented (21 IU/kg) mice. No significant differences were determined by unpaired t test.
We also characterized the mammary glands of VDR knockout (VDRKO) and wild-type littermate virgin mice synchronized and vaginally staged (25) in all four stages of the estrous cycle: proestrus, estrus, metestrus, and diestrus (Fig. S4). As previously reported for apoptosis (26), autophagy was found to cycle according to the estrous phase in the mammary glands of wild-type virgin mice. In VDRKO (B6.129S4-Vdrtm1Mbd/J) mice, the periodicity and amplitude of the autophagy cycle were significantly (P ≤ 0.05) disrupted (Fig. S5).
Mammary glands of VDRKO and wild-type (WT) littermate virgin mice synchronized and vaginally staged (20× magnification) in all four stages of the estrous cycle: proestrus, estrus, metestrus, and diestrus duct measurements and counts of alveolae of wild-type mice fed with control diet, wild-type mice fed with Ca++ rescue diet, and VDRKO mice fed with Ca++ rescue diet (15× magnification). No statistical significance of VDRKO was found relative to wild-type Ca++ diet control mice as determined by unpaired t test.
VDR regulates the periodicity and amplitude of the autophagy cycle in the mammary gland in mice. (A) Percentage of LC3-positive alveolae in the mammary glands from wild-type mice fed with control diet, wild-type mice fed with Ca++ rescue diet, and VDRKO mice fed with rescue Ca++ diet immunostained against cleaved LC3. (Magnification: 40×.) (B) Percentage of LC3-positive duct cells in the mammary glands from wild-type mice fed with control diet, wild-type mice fed with Ca++ rescue diet, and VDRKO mice fed with rescue Ca++ diet immunostained against cleaved LC3. (Magnification: 20×.) *P ≤ 0.05 relative to control WT Ca++ mice as determined by the stratified random (Braun–Blanquet) sampling approach followed by one-way ANOVA.
We next examined the biological function and cellular localization of the autophagy genes regulated by 1,25(OH)2D3 (Ingenuity Pathway Analysis software) in MCF-7 cells. We noted that over 55% of the proteins encoded by these genes are involved in autophagosome formation (Fig. S6). Because autophagosome accumulation is toxic to cells (27, 28), we hypothesized that inducing autophagosome formation with 1,25(OH)2D3 and then blocking autophagosome degradation would be detrimental to cell survival. Therefore, we tested whether adding the 4-aminoquinoline compound chloroquine (CHQ), an inhibitor of autophagosome acidification, which is the last step in autophagosome degradation, would augment the antiproliferative actions of 1,25(OH)2D3 in MCF-7 cells. We found that CHQ significantly (P ≤ 0.001) synergizes with the antiproliferative actions of 1,25(OH)2D3 in MCF-7 cells (Fig. 4A, Left). TEM demonstrated that cells treated with both 1,25(OH)2D3 and CHQ were saturated with vacuoles as expected (Fig. 4B). Next, to assess the potential additional therapeutic benefit of an autolysosome inhibitor to mice fed a vitamin D-supplemented diet, we treated mice harboring MCF-7 xenografts with a combination of the 4-aminoquinoline hydroxychloroquine (HCQ; 10 mg/kg i.p. every 3 to 7 d) and a vitamin D-supplemented diet ad libitum, and then compared them with MCF-7 xenograft-bearing mice treated with HCQ or vitamin D individually and with controls. The size of the xenografts was substantially smaller in animals treated with the combination therapy relative to animals receiving the individual therapies and significantly smaller (P ≤ 0.05) compared with control animals (Fig. 4A, Right). These findings suggest that the induction of autophagy by vitamin D, together with an inhibitor of autolysosomal acidification, may be an effective therapeutic strategy in luminal BCs.
1,25(OH)2D3 shows antiproliferative benefit with 4-aminoquinoline compounds in vitro and in vivo. (A) MCF-7 cell counts treated with control, 10 nM 1,25(OH)2D3, and 500 nM CHQ alone and combined. MCF-7 side-flank xenograft measurements from mice treated with control, mice fed add libitum vitamin D-supplemented diet (23 IU/kg), mice treated with HCQ (10 mg/kg i.p. three times per week), or mice fed the vitamin D diet and treated with HCQ. *P ≤ 0.05 and **P ≤ 0.01 as determined by one-way ANOVA. (B) TEM of MCF-7 cells treated with a combination of 1,25(OH)2D3 and CHQ show a higher vacuole accumulation than control or either treatment alone. al, autolysosomes; ap, autophagosomes; g, Golgi apparatus; m, mitochondria, N, nucleolus; rER, rough endoplasmic reticulum. Undigested vacuoles are shown by an asterisk.
Cellular localization of the autophagy genes regulated by 1,25(OH)2D3 (Ingenuity Pathway Analysis software) in MCF-7 cells.
Because the hormonal form of vitamin D mediates its effects through transcriptional regulation via the VDR, we silenced the VDR in MCF-7 cells by RNA interference, and used LysoTracker Red to monitor autophagy. Surprisingly, VDR knockdown led to much higher levels of autolysosome formation than did 1,25(OH)2D3 treatment, and there was no appreciable effect of the addition of 1,25(OH)2D3 in the setting of VDR knockdown. Importantly, the induction of autolysosome formation by silencing VDR was rescued by transfection of a small interfering VDR-resistant construct (Fig. 5A). TEM demonstrated that the total volume of autolysosomes following VDR knockdown had increased by ∼100-fold relative to cells treated with 1,25(OH)2D3 (P ≤ 0.01) (Fig. 5B). The increased net volume of autophagolysosomes was accompanied by an enhanced autophagic flux. Autophagic flux was determined by stably expressing mCherry-EGFP-LC3B in MCF-7 cells (29) for 1,25(OH)2D3 treatment and VDR knockdown (Fig. 5C). Finally, we compared the levels of autophagy observed in the mammary glands of VDRKO mice at estrus with the levels of autophagy in wild-type littermates fed a vitamin D-supplemented diet and in control mice. LC3B quantification revealed significantly (P ≤ 0.05) higher levels of autophagy in the mammary glands of the VDRKO mice compared with control mice following vitamin D supplementation (Fig. 5D). Collectively, these results show that the absence of VDR induces higher levels of autophagy than does 1,25(OH)2D3 treatment alone, and suggest that the VDR acts as a constitutive repressor of autophagy and that this repression is partially relieved upon 1,25(OH)2D3 stimulation.
Absence of VDR results in higher autophagy levels than the levels seen with vitamin D treatment alone. (A) Live MCF-7 cells transfected with small interfering VDR (siVDR) and stained with LysoTracker Red show a higher level of acidic compartment than MCF-7 cells treated with 1,25(OH)2D3. (Magnification: 20×.) This effect can be rescued by cotransfecting an siVDR-resistant construct. (B) TEM of control, 1,25(OH)2D3, and VDR knockdown in MCF-7 cells showing significant differences on autolysosome volumes. *P ≤ 0.05 as determined by Kruskal–Wallis one-way analysis of variance. (C) MCF-7 cells stably expressing mCherry-EGFP-LC3B exhibit higher levels of autophagy flux upon VDR knockdown than MCF-7 cells following 1,25(OH)2D3 treatment. (Magnification: 40×.) (D) Immunohistochemistry of mammary gland alveolae immunostained against cleaved LC3 (40× magnification) in control mice, mice fed a vitamin D-supplemented diet, and the VDRKO mice (punctate stain, 60× magnification) *P ≤ 0.05 and **P ≤ 0.01 as determined by the stratified random approach (Braun–Blanquet) sampling method followed by Kruskal–Wallis one-way analysis of variance.
To identify VDR direct target genes, we performed ChiP sequencing (ChIP-Seq) of endogenous VDR in MCF-7 cells in the presence and absence of 1,25(OH)2D3. Model-based analysis of ChIP-Seq was used to call the VDR peaks, and it revealed 2,278 VDR-binding sites in the absence of 1,25(OH)2D3: a total of 7,418 sites seen only following 4 h of ligand stimulation and 660 sites that remained unchanged in the presence or absence of 1,25(OH)2D3 (Fig. 6A). Consistent with other VDR and nuclear receptor cistromes, most of the VDR-binding sites were detected in enhancers located in intergenic regions, and rarely (∼5%) near promoters (Fig. S7A). A motif-based sequence analysis tool (Tomtom) was used to identify enriched motifs, which revealed significant enrichment of the consensus VDR-RXR binding motif (DR3) in both the presence (P = 2.76 e-9) and absence (P = 1 e-30) of 1,25(OH)2D3 (Fig. S7A). Also, other specific transcription factor-binding motifs were differentially enriched in peaks found in the control-specific group, in the 1,25(OH)2D3 stimulation-specific group, and in common peaks (Fig. S7B). We found evidence of constitutive VDR binding in the first intron of the key autophagy gene MAP1LC31B (Fig. 6B). Interestingly, although VDR was found constitutively bound at this site, we could demonstrate by directed ChIP-quantitative PCR for HDAC3 and p300 that there was an exchange of coregulators induced by 1,25(OH)2D3 (Fig. 6B). Furthermore, addition of 1,25(OH)2D3 activated transcription of the LC3B gene (MAP1LC31B). Treatment with the HDAC inhibitor Trichostatin A (TSA) also significantly induced MAP1LC31B expression, and the combined treatment of 1,25(OH)2D3 with TSA synergized to induce MAP1LC31B expression to similar levels as found following VDR knockdown (Fig. 6C). These experiments demonstrate a constitutive repression of the MAP1LC31B gene by VDR that is partially relieved upon 1,25(OH)2D3 stimulation and that this repression is mediated, in part, by HDAC-associated corepressors. This mode of gene regulation was not recapitulated at all vitamin D target genes, such as CYP24A1 (Fig. S7C). In addition, we identified the genes located near VDR-bound enhancers (±30 kb from promoter regions) and performed pathway analysis on these genes. Interestingly, the pathway analysis of the genes found near VDR enhancers consistently yielded apoptotic and metabolic (autophagy) pathways to be significantly enriched in the three groups of VDR peaks: control, 1,25(OH)2D3-stimulated, and common peaks (Fig. S7D).
Autophagy regulation by the VDR. (A) MCF-VDR ChIP-Seq of control and 4 h of 1,25(OH)2D3 treatment (0.05 false discovery rate). (B) Time course of directed ChIP-quantitative PCR on a VDRE located in the first intron of autophagy gene MAP1LC3B (LC3B). (C) LC3B expression following control, 1,25(OH)2D3, TSA, combined 1,25(OH)2D3 and TSA, and siVDR knockdown. *P ≤ 0.05 and **P ≤ 0.01 as determined by one-way ANOVA. (D) Venn diagram of expression analysis comparison of genes significantly up-regulated (Top) or down-regulated (Bottom) in the presence of 1,25(OH)2D3 relative to control (black) and VDR knockdown relative to 1,25(OH)2D3. (P ≤ 0.01 and Benjamini–Hochberg correction). (E) ChIP-Seq and expression array integration indicating the percentage of expressed genes with VDREs (direct targets). (F) Pathway analysis of the VDR protein corepressor interactome (RIME) in the presence of 1,25(OH)2D3 in MCF-7 cells. (G) Integration of 1,25(OH)2D3-induced autophagy genes with ChIP-Seq data.
VDR directly modulates autophagy in MCF-7 BC cells. (A) VDR ChIP-Seq (MCF-7) revealed that most VDR-binding sites are detected in enhancers located in intergenic region, and that only a small percentage of peaks are located in promoter regions. The consensus VDR-RXR (DR3) motif was enriched in the presence (P = 2.76 e-9) and absence (P = 1 e-30) of ligand. (B) Different transcription factors enriched in peaks found in control, ligand, or common peaks. (C) Directed ChIP-quantitative PCR (q-PCR) of p300 and HDAC3 at VDRE located in the first intron of the MAP1LC3B gene following 1, 5, and 48 h of 100 nM 1,25(OH)2D3 compared with control. (D) MAP1LC3B and CYP24A1 RT q-PCR of control, 1,25(OH)2D3, TSA, 1,25(OH)2D3 + TSA, and VDR knockdown (48-h treatment). (E) ChIp-Seq expression integration analysis indicating the percentage of expressed genes containing VDREs within 30 kb (direct target genes). (F) The 1,25(OH)2D3-specific autophagy signature genes present following VDR knockdown. (G) MCF-HDAC3 ChIP-Seq of control and 4 h of 1,25(OH)2D3 treatment (0.05 false discovery rate) overlapped with VDR ChIP-Seq. (H) Chromatin remodeling proteins significantly present in the VDR interactome (RIME).
To investigate the mechanisms involved in the increased autophagy levels upon VDR knockdown further, we performed gene expression analysis comparing knockdown control with no-ligand, knockdown control treated with 1,25(OH)2D3 for 16 h, and VDR knockdown (Fig. 6D and Dataset S1). This analysis revealed that 1,271 of the total 1,689 genes induced by 1,25(OH)2D3 are down-regulated following VDR knockdown. Likewise expression of 2,581 of the total 3,450 genes repressed by 1,25(OH)2D3 is enhanced following VDR knockdown. This result indicates, as expected, that regulation of most genes by 1,25(OH)2D3 is dependent on the VDR. Interestingly, this analysis shows that the largest group of regulated genes is up-regulated in the absence of the VDR, consistent with a constitutive gene repression mechanism mediated by the VDR that we observed in LC3B. Therefore, the constitutive repression of autophagy by VDR we observe is a widespread mechanism of gene regulation. Similar to 1,25(OH)2D3-induced genes (45%), nearly 40% of the genes up-regulated following VDR knockdown contain at least one VDR-binding site within 30 kb, suggesting direct VDR regulation (Fig. 6E, Fig. S7E, and Dataset S2). Interestingly, upon VDR knockdown, the 1,25(OH)2D3-induced autophagy signature is lost and only four autophagy-related genes remain present: FOS, CXCR4, VEGFA, and MAP1LC3B (Fig. S7F); however, only LC3B was directly up-regulated by 1,25(OH)2D3 [with vitamin D response elements (VDREs) located ±30 kb from the gene] and further up-regulated following VDR knockdown. Given that our phenotype suggests constitutive repression of autophagy by the VDR, and as a proof of principle for constitutive VDR repression, we did ChIP-Seq of HDAC3 and confirmed VDR overlap in the absence of 1,25(OH)2D3 (Fig. S7G). Next, we performed rapid immunoprecipitation mass spectrometry (RIME) of endogenous proteins in the presence of 1,25(OH)2D3 to identify the proteins associated with VDR when it is bound to DNA (Dataset S3). Ingenuity Pathway Analysis of the VDR interactome revealed highly significant (P ≤ 0.01 × 10−5) enrichment of proteins involved in transcriptional repression and DNA methylation, as well as chromatin remodeling and gene silencing (Fig. 6F and Fig. S7H), providing a mechanistic insight on VDR constitutive repression of gene expression. These results indicate a role of the ligand-bound VDR as regulator of autophagy in normal mammary gland and BC cells by regulated expression of autophagy-related genes by both activation of gene transcription and relief of constitutive repression (Fig. 7A) of the key autophagy gene LC3B, which is highly expressed in the absence of VDR (Fig. 7B) and de-repressed in the presence of 1,25(OH)2D3 (Fig. 7C).
Model for VDR regulation of autophagy in luminal-like models. (A) Upon 1,25(OH)2D3 binding, the VDR activates gene transcription to target genes to vitamin D by derepression (Left) or induction of transcription (Right). (B) In the absence of VDR, there is a full de-repression of autophagy and 100-fold higher expression levels of MAP1LC3B than with 1,25(OH)2D3 alone. (C) In the presence of VDR, 1,25(OH)2D3 induces a constitutive de-repression of autophagy gene LC3B and regulation of transcription by the VDR, resulting in autophagy induction following vitamin D treatment.
In summary, our results show that in response to 1,25(OH)2D3, luminal-like BC cells enter autophagy and create a transcription signature that is present in normal breast and lost during cancer progression. We have shown that vitamin D in the diet can increase basal levels of autophagy in the normal breast, and that modulation of autophagy by vitamin D can provide a potential therapeutic opportunity as a combinational therapy with autophagy inhibitory chemotherapeutic drugs.
Our study shows that 1,25(OH)2D3 modulates autophagy via the VDR, by direct regulation of transcription. Moreover, treatment with 1,25(OH)2D3 leads to an autophagy-related gene expression signature, which is present in normal mammary gland but is no longer detected as the cancer progresses, offering a potential mechanism for the epidemiological observation that serum vitamin D levels ≥45 ng/mL are protective against BC. Mechanistically, the VDR constitutively represses autophagy; upon 1,25(OH)2D3 stimulation, basal levels of autophagy increase by de-repression of the key autophagy gene MAP1LC3B (LC3B).
The role of autophagy in cancer is context-dependent, and extensive reports document a prosurvival role of autophagy in cancer. However, in some systems, autophagy also reportedly inhibits cancer growth (20). Our data indicate that 1,25(OH)2D3-induced autophagy plays an antisurvival role. Upon stimulation with 1,25(OH)2D3, we observed increased levels of autophagy, exclusively in BC-cell models (luminal-like models), which also exhibit an antiproliferative response. The landscape of gene expression following 1,25(OH)2D3 treatment in our model, as well as in many others (30), also reflects inhibition of proliferation, cell cycle arrest, and cell death, as opposed to a proliferative or prosurvival gene transcription profile.
Vitamin D induces autophagy in macrophages, neurons, and models of skin and BC (31, 32). Høyer-Hansen et al. (24) described a 1,25(OH)2D3-induced autophagy mechanism involving AMP-activated protein kinase (AMPK) activation triggered by activation of calcium/calmodulin-dependent protein kinase kinase 2 (CAMKK2) activation in MCF-7 cells. We also observed increased activation of AMPK that was not accompanied by a change in protein levels. In contrast, however, no changes were detected in levels of CaMKK transcript, its protein levels, or its activation upon 1,25(OH)2D3 treatment. Unexpectedly, levels of autophagy in the absence of VDR were much higher than with 1,25(OH)2D3 treatment in MCF-7 cells that expressed VDR. Mirroring these in vitro findings, we found that in vivo, VDR KO mice have higher basal levels of autophagy in their mammary glands than do their wild-type littermates. Vitamin D supplementation also increases basal autophagy levels in the mammary gland, consistent with our in vitro findings. Collectively, the above findings suggest a constitutive VDR-mediated repression of autophagy that is partially relieved upon stimulation with 1,25(OH)2D3.
Canonically, upon ligand binding, VDR regulates transcription as an RXR heterodimer that is bound to DNA. In the presence of ligand, this VDR-RXR heterodimer mediates gene activation by recruiting coactivators of transcription, but in the absence of ligand, VDR instead binds corepressors (33). However, gene expression studies show a very similar number of genes being up-regulated and down-regulated across a number of different cancer models upon 1,25(OH)2D3 treatment. In addition, Meyer and Pike (34, 35) used ChIP-Seq studies in a colon cancer model (LS180) to show that the recruitment of corepressors NCoR and SMRT overlaps with VDR binding in 1,25(OH)2D3-activated enhancers. To confirm the presence of corepressors associated with DNA-bound VDR after 1,25(OH)2D3 stimulation, we performed RIME of endogenous protein experiments. Our RIME data show that VDR interacts with corepressor complexes following 1,25(OH)2D3 treatment, including the characterized VDR corepressors SIN3/NurD/CoREST, PRC2, TFTC, and SWI/SNF, as well as coactivator complexes INO80, CBP, and SRC3/ncoa3 and WTAP-SFRS (a complete list of the VDR interactome is provided in Dataset S3). Although we did not find p300 or HDAC3 in the VDR interactome (perhaps due to stringent analysis cutoffs and extensive washing in the RIME protocol), ChIP-Seq HDAC3 confirmed VDR overlap in both the presence and absence of 1,25(OH)2D3. Canonical pathway analysis of our data revealed that in the presence of vitamin D, the DNA-bound VDR is significantly associated with protein groups that are involved not only with transcription and translation initiation but also with DNA methylation and transcription repression. Our RIME studies offer a global view of DNA-bound 1,25(OH)2D3-activated VDR functions as sophisticated transcription node fine-tuning gene expression regulation via multiple concurrent mechanisms.
We focused on autophagy regulator genes that were activated by 1,25(OH)2D3 but showed higher activation upon VDR knockdown: Candidates were narrowed down with regard to the VDR enhancers found in the absence and presence of 1,25(OH)2D3. We identified LC3B, the most extensively studied member of the Atg8 protein family (36), which plays a key role in autophagosome biogenesis. LC3B is essential for autophagy, and has a critical function in autophagosome elongation and autophagy flux. LC3B is believed to control autophagosome size on account of its hemifusion and membrane tethering activities (37, 38). Overexpression of LC3B alone increases both formation and elongation of autophagosomes by more than 60% in human HeLa cells. Furthermore, LC3 proteins serve as docking sites for adaptor proteins, and thereby play a key role in the selective recruitment of autophagy cargoes into autophagosomes (37). MCF-7 treatment with 1,25(OH)2D3 results in LC3B induction. Moreover, VDR knockdown results in a 100-fold up-regulation of LC3B; this much higher up-regulation can be partially mimicked by cotreatment with 1,25(OH)2D3 and the HDAC class I and II inhibitor TSA, suggesting that LC3B is modulated by a constitutive de-repression of the VDR at the MAP1LC3B VDRE. As with 1,25(OH)2D3, LC3B not only plays a role in autophagy but is also heavily involved in the immune response, particularly during phagocytosis (39). Ramagopalan et al. (40) used ChIP-Seq to characterize the VDR-binding elements in lymphoblastic cell models (GM10855 and GM10861), and also identified VDR cis-regulatory sites near the MAP1LC31B gene. These observations support the possibility that VDR modulation of LC3 plays a wider role than in autophagy in the mammary gland, but also in innate and adaptive immunity, all of which are among the broad spectrum of biological functions of vitamin D.
Serum levels of vitamin D correlate with a positive prognosis in patients with BC (5). We show that 1,25(OH)2D3 induces an autophagy gene expression signature: Comparison against TCGA human BC data reveals that this signature is present in normal mammary gland and lost upon cancer progression. Interestingly, we have shown that dietary supplementation of vitamin D can increase basal levels of autophagy in the normal mammary gland in mice. Modulation of autophagy, although still premature, is an attractive venue in BC therapeutics (41), given that autophagy plays a role in the different stages of BC (42).
Vitamin D and its analogs are known to reduce the growth of luminal BC xenografts in vivo. As a proof of concept that vitamin D cotreatment amplifies potential therapeutic properties of autophagy-modulating drugs, we treated mice harboring MCF-7 xenografts with vitamin D alone, HCQ alone, or a combination of the two. We observed added benefit from the combination compared with the use of either compound alone, showing added therapeutic gain from concurrent induction of autophagy and inhibition of the last step of autolysosome acidification. However, HCQ is not highly specific in inhibiting autophagy. Further studies are indicated, with better inhibitors of autophagy, such as Lys05, along with low-calcemic vitamin D analogs, to determine the therapeutic potential of vitamin D analogs and autophagy-modulating drugs. Rexinoids constitute another interesting noncalcemic venue that is actively under investigation for prevention of mammary cancers (43), which could potentially be enhanced by autophagy manipulation. Autophagy induction, along with inhibition for cancer therapeutics, is currently under study for melanoma and other solid tumors (44). Our results suggest that vitamin D serum levels and intake should be monitored in patients as they are recruited to clinical trials for autophagy-modulating therapeutics.
In terms of cancer risk, epidemiological studies show that vitamin D reduces the risk of developing BC, but only when serum levels are ≥45 ng/mL (4). Although levels up to 140 ng/mL can be synthesized in skin (12), the overall incidence of vitamin D deficiency (serum levels ≤20 ng/mL) is strikingly high in the United States: 41.6% of Americans show this deficiency, with the highest rate seen in African Americans (81.2%) (7, 11). Several factors contribute to these low levels of vitamin D: geographical location with respect to the equator, an indoor lifestyle, and sunscreen protection. Because there is naturally very little vitamin D in food, including in vitamin D-supplemented products (45), we rely mostly on supplementation for the fulfillment of our vitamin D needs. Our findings show that only vitamin D diet supplementation at higher levels than required for bone health increases basal autophagy in the mouse mammary gland. Notably, mouse mammary tissue in which autophagy is compromised exhibits hyperproliferative and preneoplastic changes: These changes are likely due to a compromised ability to cope with metabolic stress and overall cell damage, resulting in DNA instability and further damage, which, in turn, may accelerate tumor progression (20, 42). Further studies are needed to determine whether vitamin D supplementation induces changes in basal autophagy in human breast tissue, and whether it produces the cancer-protective gene signature that we report.
The cell lines MCF-7, ZR-75-1, MDA-MB-453, MCF-12A, MDA-MB-231, and HMEC were purchased from the American Type Culture Collection (ATCC) and authenticated. Cells were cultured and kept in recommended media and conditions (ATCC).
The 1,25(OH)2D3 (D1530; Sigma) treatments were done at 100 nM concentrations unless specified otherwise in Results. HCQ sulfate (H0915) and CHQ salt (C6628) were purchased from Sigma. Antibodies for LC3 immunohistochemistry cleaved LCA3A (AP1805a; Abgent). For Western blotting, LC3B (2775; Cell Signaling), Beclin-1 (3738; Cell Signaling), and VDR (D-6) (sc-13133; Santa Cruz Biotechnology) were used. For ChIP applications, we used VDR (C-20) (sc-1008; Santa Cruz Biotechnology), p300 (N-15) (sc-534; Santa Cruz Biotechnology), and HDAC3 (ab7030; Abcam) antibodies.
Information on microscopy, data analysis, and animal studies is provided in SI Materials and Methods. All animal studies were approved by the Institutional Animal Care and Use Committee at the Dana–Farber Cancer Institute.
MCF-7 cells grown in full media were treated with 100 nM 1,25(OH)2D3 for 4 h before VDR immunoprecipitation (C-20; Santa Cruz Biotechnology). Four independent immunoprecipitations and libraries were prepared for ChIP-Seq, and three independent RIME samples were prepared. VDR ChIP-Seq samples were prepared according to the laboratory protocol of Brown and co-workers (46), and RIME sample preparation and analysis were performed according to Mohammed et al. (47).
LysoTracker Red (L-7528; ThermoFisher Scientific) experiments were performed exactly as described (48) in live MCF-7 cells, and samples were viewed using a Nikon wide-field inverted live-cell imaging system. A small interfering (si_rVDR) construct was made by introducing a silent mutation per the manufacturer’s instructions (Phusion Site-Directed Mutagenesis Kit; ThermoFisher Scientific) using a PSG5-VDR construct (graciously donated by John White, McGill University, Montreal, QC, Canada), making it resistant to siVDR 1 (ThermoFisher Scientific) siRNA location 443.
MCF-7 cells stably expressing mCherry-EGFP-LC3B (22418; Addgene) (49) were assayed as described (29). Samples were viewed using a Nikon wide-field inverted live-cell imaging system.
EGFP–LC3–MCF-7 cells were transiently transfected with EGFP-LC3 plasmid (50) (11546; Addgene). Samples were viewed using a Nikon wide-field inverted live-cell imaging system. In parallel, EGFP-LC3–transfected cells sorted using a BD FACSAria II system and equally expressing EGFP-LC3 cells were sorted and seeded (n = 15,000). Following 48 h of treatment, fluorescence intensity was measured using a BD LSRFortessa analyzer on 10,000 cells. Data were analyzed using a one-way ANOVA test.
TEM samples were fixed in 1:1 4% paraformaldehyde and 25% glutaraldehyde. Samples were followed by osmication and uranyl acetate staining, dehydration in alcohols, and embedding in Taab 812 Resin (Marivac Ltd). Samples were counterstained with 0.2% lead citrate, and then viewed and imaged under a Philips Technai BioTwin Spirit Electron Microscope. Due to high variance, 1,500 autolysosmes were counted and measured for autolysosome volume calculations. Data were analyzed using the Kruskal–Wallis test (nonparametric one-way ANOVA).
All animal studies were approved by the Institutional Animal Care and Use Committee at the Dana–Farber Cancer Institute.
VDRKO mice (B6.129S4-Vdrtm1Mbd/J) and their wild-type parental strain (B6;129) were used (The Jackson Laboratories). Animals were fed ad libitum with regular control chow (8604; Teklad) or with AIN-76A with no added vitamin D3 (TestDiet), AIN-76A with 11.5 IU/g vitamin D3 (TestDiet), or AIN-76A with 23 IU/g vitamin D3 (TestDiet). VDRKO mice were kept on a high-calcium, high-lactose rescue diet (TD.120014; Teklad). For autophagy in mouse mammary gland studies, mice were synchronized and vaginally staged as described (26). Fat pads were dissected from four different animals at all stages of the estrous cycle, and whole mounts were prepared for characterization before immunohistochemistry, as described by Moser et al. (51). We observed the mammary gland being a highly dynamic and complex organ not just in the different stages of the estrous cycle but that the branches grow and develop in a caudal cranial direction as the cycle progresses. For accuracy in the comparison between animals and within stages of the cycle, we decided to adapt the Braun–Blanquet method (stratified random approach), the gold-standard method used for vegetation mapping and sampling (52). We created a squared quadrant of 12 × 12 mm2 (each sampling unit = 1 mm2) and determined the sampling unit by random number generation that 16 random counts gave a SE ≤ 5% within the mammary gland and between the two mammary glands (left and right) of the same mouse. The quadrant was always located such that the lymph node of the fourth mammary gland was in the center, and measured and quantified ducts and alveolae (primary and secondary) in the mammary gland. From each mammary gland, four random 5-mm3 cubes were subsampled caudal, central, and cervical; embedded; and sectioned for immunohistochemical staining of cleaved LC3. Samples were viewed using a Nikon wide-field inverted live-cell imaging system. Ten ducts and 10 alveolae each were scored per animal for LC3 staining, and four animals per treatment per estrous stage were studied. Data were analyzed using the Kruskal–Wallis test (nonparametric one-way ANOVA).
NOD-scidIL2Rgnull mice were used (The Jackson Laboratories). Two low-dose, 90-d release estrogen pellets (Innovative Research of America) were implanted in mice harboring xenographs. Mice harboring MCF-7 xenografts we treated alone or with a combination of 4-aminoquinoline HCQ (10 mg/kg i.p. every 3–7 d) as described (53) and a vitamin D-supplemented diet ad libitum. There were five animals per group, and the study was replicated once. Data were analyzed using one-way ANOVA.
We used a QuantiChromTM Calcium Assay Kit (DIC-500) from BioAssay Systems following the manufacturer’s instructions.
The 1,25(OH)2D3 was measured from mice serum samples by liquid chromatography/mass spectrometry (API4000; Applied Biosystems).
We used the R programming language to analyze microarray data. To compute expression values for probe sets, we used the robust multi-array average (RMA) algorithm as implemented in the Bioconductor package. To determine differential expression, we used the limma package to fit linear models to the RMA expression values and used empirical Bayes statistics to decide differential expression using an F-test with a P value of 0.01 and the Benjamini–Hochberg correction.
Pathways were built using MetaCore software, and networks were calculated by the Dijkstra algorithm using version 6.3, selecting for direct interactions with a maximum of two steps in the path but including disconnected seed nodes.
Read libraries were mapped to the human genome (hg18) using Bowtie v1.0. Analysis was performed using the Cistrome analysis pipeline (54). VDR peaks were called using model-based analysis of ChIP-Seq (MACS) version 1 with default settings (Poisson threshold: P = 10−5).
For RT-PCR, total RNA was isolated using TRIzol (ThermoFisher Scientific), followed by RNeasy (Qiagen). For cDNA preparation, a OneStep RT-PCR Kit (Qiagen) was used following the manufacturer’s protocol. The expression of MAP1LC3B and CYP24A1 was measured using Power SYBR Green PCR Master Mix (Applied Biosciences) with the following primers: MAP1LC3B (forward) 5′-CGGAGAAGACCTTCAAGCAG-3′ and MAP1LC3B (reverse) 5′-TGACATGGTCAGGTACAAGGAA-3′, and CYP24A1 (forward) 5′-CACAAGAGCCTCAACACCAA-3′ and CYP24A1 (reverse) 5′-TCTTTGAAAGCCGATTCTGG-3′. For directed ChIP-quantitative PCR, we used MAP1LC3_MACS 2714 (forward) 5′-GTCACCTTCAGGGCTGGTT-3′ and MAP1LC3_MACS 2714 (reverse) 5′-AGACAGGAACTGGCAACACC-3′.
We thank Ana Tavera Mendoza for illustration (Fig. 7) design and drawing; Christine Unitt for help with LC3 immunohistochemistry; Louise M. Trakimas for TEM assistance; Melinda Baker (Thomson Reuters) for invaluable advice in creating the autophagy pathway; the Dana–Farber Animal Facility staff (Michael Terrasi, Antonia Garcia, Daisy Moreno, Carmen Da Silva, Elsy Moreno, and Catherine A. Sypher) for assistance with animal experiments and overall excellence in animal care; and Sonal Jhaveri-Schneider for editing help. This work was supported by National Cancer Institute Grant P01 CA080111 (to M.B.).
↵1L.E.T.-M. and T.W. contributed equally to this work.
Author contributions: L.E.T.-M. and M.B. designed research; L.E.T.-M., A.M., L.C., L.A.C., S.B.F., and J.K. performed research; A.M., L.A.C., and J.K. contributed new reagents/analytic tools; L.E.T.-M., T.W., E.L., L.C., R.C., S.B.F., J.A.M., and M.B. analyzed data; and L.E.T.-M. and M.B. wrote the paper.
Reviewers: M.A.L., Perelman School of Medicine at the University of Pennsylvania; and D.P.M., Duke University School of Medicine.
The authors declare no conflict of interest.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1615015114/-/DCSupplemental.
Freely available online through the PNAS open access option.
