The isomiR-140-3p-regulated mevalonic acid pathway as a potential target for prevention of triple negative breast cancer

We used an MCF10A-based model, a well-established model of TNBC progression developed by Dr Fred Miller (Karmanos Cancer Institute), which recapitulates four major steps of breast cancer progression and comprises normal-like mammary epithelium MCF10A(P), atypical hyperplasia (MCF10.AT1), ductal carcinoma in situ (MCF10.DCIS), and invasive carcinoma (MCF10.Ca1d), representing a stepwise progression in TNBC progression. Normal-like MCF10A (P) is a spontaneously immortalized cell like that was developed using mammary tissue from a women with fibrocystic breast disease. Premalignant MCF10.AT1 cells were obtained by transfection of MCF10A(P) cells with constitutively active oncogene H-ras, which form simple ducts in mice xenografts [3] . Two successive passages of a lesion formed by the MCF10.AT1 cells in xenografts gave rise to MCF10.DCIS cells that forms comedo DCIS lesions in mice xenografts and resembles human DCIS lesions [2]. MCF10.CA1d is a highly tumorigenic derivative of MCF10.AT1 cells [3]. We have tested and found these cells lines to be estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (Her2) negative using the same assays applied for ascertainment of biomarkers in patients’ tumors (data not shown). Others have similarly reported on the biomarker profile of this model, which resembles that of TNBC [9‐11]. We purchased MCF10A(P) cells from American Type Culture Collection (ATCC). MCF10.DCIS cells were obtained from Wayne State University, and MCF10.AT1 and MCF10.Ca1d from Karmanos Cancer Institute under a materials transfer agreement. All the cell lines used in the study were authenticated by the source agency and were used within the first 10 passages. Cell lines were periodically tested for mycoplasma and confirmed negative throughout the course of the work presented.

While some cells are inherently resistant to statins, others are initially sensitive to statins and eventually develop resistance to statins. Here, we recapitulated adaptive resistance by exposing otherwise sensitive MCF10.AT1 cells to 20 μM fluvastatin. After a period of massive cell death, a cell population that was resistant to fluvastatin emerged and was expanded. These cells were then grown in a maintenance dose of fluvastatin (10 μM). We determined the concentration needed to induce a 50% reduction in the viability (MTT50) of these cells and found that 8.5 μM fluvastatin killed 50% of cells. We designated these resistant cells as MCF10.AT1-R cells, which are four times more resistant than regular MCF10.AT1 cells (MTT50 = 2.1 μM).

Ten million exponentially growing MCF10.AT1 cells that were resuspended in 75 μl PBS and mixed with an equal volume of Matrigel were injected into the mammary fat pads of 5-week-old inbred female BALB/c Nu/Nu mice (Charles River Laboratories). All the mice were of the same age and randomly divided in two groups. We injected MCF10.AT1 cells into both flanks. One week after the cell injections, fluvastatin treatment (10 mg/kg body weight/day) was started, and continued for 16 weeks. Fluvastatin was mixed in the drinking water of mice (n = 30) and changed every other day. Control mice (n = 25) received plain water. Water intake was noted on every change of water, and the concentration of fluvastatin was adjusted to maintain a level of 10 mg/kg body weight/day, if needed. The body weight of mice was noted once a week, and no change in the body weight was observed with statin treatment. Lesion size was measured every week. At the end of 16 weeks, mice were euthanized, and the tissues were explanted from the site of injection. Half of the tissue was fixed in formalin and the other half was saved in TRIzol for RNA extraction. These formalin-fixed tissues were subsequently embedded in paraffin and stained with hematoxylin and eosin to determine their histological grading. Histological grading ranged from simple tubules with 1–2 cell layers (grade 0); simple tubules with > 2 cell layers but no architectural complexity (grade 1); complex hyperplasia (grade 2); atypical hyperplasia (grade 3); to ductal carcinoma in situ (grade 4) as described [10].

The reporter constructs of the 3′UTR of HMG-COA reductase (HMGCR) and HMG-COA synthase1 (HMGCS1) containing the wild-type seed sequence of the miR-140-3p-1 binding site along with 200 flanking nucleotides (both upstream and downstream) were generated from PCR-amplified human genomic DNA that was subsequently cloned downstream of the firefly luciferase open reading frame at the PmeI and XbaI sites in a pmiRGlo vector (Promega Corporation). Mutant versions of the HMGCR and HMGCS1 3′UTR reporter plasmids were generated by site-specific mutagenesis, as described previously [4]. Sequences of all the primers used are provided in Additional file 1: Table S1.

As described previously (17), MCF10.AT1 and MCF10.DCIS cells were transiently transfected using Lipofectamine 2000 (Invitrogen Technologies) following the manufacturer’s instructions. Cells were plated in 6-well/10-cm culture dishes and then transfected with miR-140-3p-1 mimic (Thermo Scientific) or scramble mimic (10 nM) with/without the pmiRGLo vector containing miR-binding sites. After 5-h incubation in Opti-MEM (Thermo Fisher Scientific), the medium was replaced with regular cell culture medium supplemented with 2X horse serum. Cells were lysed or plated for further assays at 48 h after the transfection.

Total cellular RNA was extracted from cells using an miRNeasy mini kit (Qiagen) that also preserves small RNAs. Complementary DNA (cDNA) was prepared using an iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer’s instructions. qPCR was performed in triplicate on each sample using an SYBR Green-based PCR assay as described previously [12]. The gene encoding ribosomal protein L19 (RPL19) was used as a control to ensure equal loading. All primer sequences are provided in Additional file 1: Table S1. Mature miR-140-3p-1 and miR-140-3p-2 were quantified by using TaqMan-based miRNA assays from Thermo Scientific according to the manufacturer’s instructions.

Luciferase activity was measured in cells that were transfected with empty pmiRGlo vector or 3′UTR-containing reporter vectors using the dual-luciferase reporter system (Promega).

Proliferation of MCF10.AT1 and MCF10.DCIS cells treated with miRNA mimic or drugs was measured either by the MTT dye uptake method or Ki67 antibody based-immunofluorescence assay as described elsewhere [13]. The intensity of Ki67 staining, representing the proliferation index of cells, was measured by counting the cells that expressed high levels (> 3 foci) of Ki67 staining (Ki67-positive cells) or low levels (0–2 foci) of Ki67 staining (Ki67-negative cells).

MCF10.DCIS cells were plated at sub confluent density in 60 mm or 100 mm dishes for western blotting experiments. The day after plating the cells, the regular growth medium was removed and cells were treated with the indicated dose of fluvastatin, aspirin and metformin for 48 h in low-glucose growth medium. Total cellular protein was subjected to SDS-PAGE and transferred to Hybond ECL nitrocellulose membranes (Sigma Aldrich), which were probed with AMPK, pAMPK (Thr 172), HMGCR, or loading control vinculin antibodies. Proteins were detected using an Odyssey Classic infrared imaging system (Li-Cor Biosciences) as described previously [14].

Top canonical pathways and their effector molecules were generated by IPA (QIAGEN Inc., https://www.qiagenbio- informatics.​com/​products/​ingenuity-pathway-analysis).

To identify miRNAs that drive normal-to-preneoplastic transition in TNBC progression, we grouped miRNAs according to their expression pattern across the continuum of cell lines in the MCF10A model of TNBC tumorigenesis. Next-generation small-RNA sequencing analyses of this breast cancer progression model, which we have previously published, placed miR-140-3p as one of the top deregulated miRNAs [4]. In order to validate the next-generation sequencing results, we performed qPCR assays using sequence-specific TaqMan-based primers for the canonical miR-140-3p (miR-140-3p-2) and its isomiR, miR-140-3p-1. miR-140-3p-1 is known to be generated by a 1-nucleotide (nt) shift in the cleavage of the miRNA processing enzyme DICER during its processing of pre-miRNA (Fig. 1a). Interestingly, we found miR-140-3p-1 to be expressed at 13-fold to 17-fold higher levels than canonical miR-140-3p-2 throughout the whole spectrum of breast cancer progression, from normal-like MCF10A (P) to preneoplastic MCF10.AT1, DCIS (MCF10.DCIS), and invasive MCF10.Ca1d cells (Fig. 1b). Although the ratio of miR-140-3p-1 relative to miR-140-3p-2 remained consistently higher, the absolute levels of both miR-140-3p-1 and miR-140-3p-2 decreased during TNBC progression, as indicated by qPCR results (Fig. 1b). We found however that the greatest decrease in both miR-140-3p-1 and miR-140-3p-2 occurred early (during the normal (MCF10A.P) -to-atypia (MCF10.AT1) transition) with 60% drop in the levels of both isoforms.

Although much is known about a myriad of biological functions performed by canonical mature miRNAs, understanding of the relevance of isomiRNAs remains elusive. Therefore, to investigate the role of miR-140-3p-1, we ectopically expressed this isomiRNA in breast preneoplastic (MCF10.AT1) and MCF10.DCIS cells and measured its effects on the colonizing ability of cells, as described in “Materials and methods”. We found ectopic expression of miR-140-3p-1 to preferentially inhibit the colonization ability of preneoplastic MCF10.AT1 cells (62% reduction, p < 0.05 compared to control) but did not have functional effect further down the disease progression spectrum in MCF10.DCIS cells (Fig. 2). Similarly, ki-67 staining indicated that miR-140-3p-1 restoration modestly inhibited the cell proliferation (by 9.8%, p = 0.08) of preneoplastic MCF10.AT1 cells but not in MCF10.DCIS cells (Additional file 2: Figure S1). In order to confirm that these changes represented biologically relevant differences across the cell lines rather than a consequence of variability in transfection efficiency, we transfected both MCF10.AT1 and MCF10.DCIS cells with miRNA-140-3p-1 or scramble control mimic. qPCR analysis of these cells suggested both the cell lines to be expressing more than 3500-fold expression of miR-140-3p-1 relative to scramble control (Fig. 2c), thus confirming that the growth inhibition seen with miR-140-3p-1 preferentially in MCF10.AT1 represents important context-specific effects of miR-140-3p-1 treatment.

To identify functional gene targets and driver pathways downstream of miR-140-3p-1, we integrated the expression pattern of miR-140-3p-1 with the gene expression data obtained by the next-generation RNA sequencing that we previously performed [4] on the MCF10A model system. In particular, we focused on the TargetScan predicted gene targets of miR-140-3p-1 that were upregulated significantly during the progression from the non-cancer parental line [MCF10A(P)] to MCF10.DCIS. The filters that we employed to obtain functionally relevant predicted gene targets of miR-140-3p-1 are listed in Additional file 3: Figure S2A. From these analyses, we identified 10 genes that have significant (p < 0.05) and reverse correlation with miR-140-3p. Ingenuity pathway analysis showed that the mevalonate/cholesterol biosynthesis pathway, through its key gene mediators HMGCR and HMGCS1, was the top predicted pathway to be deregulated (Additional file 3: Figure S2 B). As a first step, we validated the endogenous levels of HMGCR and HMGCS1 transcripts in the MCF10A progression model using qPCR. These analyses showed a steady increase in the levels of HMGCR (about 2.5-fold) and HMGCS1 (5.5-fold) in the cell lines from later stages of tumorigenic progression compared to the normal-like MCF10A(P) cell line (Additional file 3: Figure S2 C&D). Next, to test whether HMGCR and HMGCS1 are regulated by miR-140-3p-1, we restored levels of these in the preneoplastic MCF10.AT1 cell line by transfecting cells with miR-140-3p-1 mimic. These qPCR-based assays showed that indeed HMGCR and HMGCS1 were repressed (by 37% and 47%) with the addition of miR-140-3p-1 mimic relative to their expression upon transfection of scramble control mimic, confirming this predicted miR-gene relationship to be valid in the context of breast preneoplastic cells (Fig. 3a). Finally, to test whether miR-140-3p-1 directly binds to and regulates HMGCR and HMGCS1, we cloned a 500-bp fragment of the HMGCR and HMGCS1 3’UTR containing the miR-140-3p-1 binding site in pmiR-Glo, a luciferase vector. As expected, we found the miR-140-3p-1 mimic to repress the reporter luciferase activity of the wild-type construct (containing the intact miR-140-3p-1 binding site from the HMGCR and HMGCS1 3’ UTR) by 55% (Fig. 3c). As a control, we also studied the effect of miR-140-3p-1 mimic on luciferase activity of constructs that harbor a mutated miR-140-3p-1 binding site from the 3’ UTR of HMGCR and HMGCS1. As expected, the miR-140-3p-1 mimic failed to repress the luciferase activity of the mutant constructs (Fig. 3c), indicating that miR-140-3p-1 directly binds to its binding site in the 3’UTR of HMGCR and HMGCS1 and represses their activity.

Because statins inhibit the activity of HMG-CoA reductase, we next investigated whether targeting this pathway with fluvastatin impairs the growth of breast preneoplastic (MCF10.AT1) and DCIS (MCF10.DCIS) cells. Fluvastatin (5 μM and 10 μM) impaired the cell colonizing ability of both cell lines but was more effective in preneoplastic AT1 cells (by 72.42% and 93.9% relative inhibition for 5 μM and 10 μM doses, respectively, compared to vehicle control) than in DCIS cells (by 39.66% and 62.76% relative inhibition for 5 μM and 10 μM doses, respectively, compared to vehicle control) (p < 0.001) (Fig. 4a, b). Similarly, an MTT assay also revealed that fluvastatin preferentially inhibited the cell proliferation of MCF10.AT1 cells as indicated by an IC50 of 2.1 μM in MCF10.AT1 cells relative to half maximal inhibitory concentration (IC50) of 18 μM in MCF10.DCIS cells (p < 0.01), values that are derived from the fluvastatin dose response curves shown in Fig. 4c.

To test the efficacy of fluvastatin in inhibiting the progression of MCF10.AT1-driven lesions in mice xenografts, we injected MCF10.AT1 cells in both the left-side and right-side mammary glands of 55 inbred nude (BALB/c Nu/Nu homozygous) mice. Thirty mice were given fluvastatin (10 mg/kg body weight/day) in their drinking water. This dose was selected as the human equivalent of this dose (48 mg for an adult weighing 60 kg, using the BSA normalization method [15]) is well within the prescribed clinical dosing for fluvastatin. As a control group, the remaining 25 mice were given plain water. After 16 weeks of treatment, mice were euthanized, and the lesions were collected from the sites of the implant. Although we found fluvastatin-treated lesions to be 25% smaller than vehicle treated lesions, as indicated by their average weight (16.05 vs 12.48 mg, p = 0.03) (Fig. 5a), the histological findings were similar between these two groups. Specifically, using the grading system developed by Visscher et al. and as described in the “Materials and methods”, we saw no difference in the distribution of histologic grade between the treated and control xenografts (Fig. 5b), indicating that statin treatment did not inhibit the progression of MCF10.AT1-driven lesions. Given this lack of efficacy, we next investigated whether statins effectively inhibited the target mevalonate/cholesterol pathway. Although assessment of HMGCR inhibition is typically performed using blood assays, earlier studies suggest that in at-risk women, dysregulation of the MVA pathway occurs in the tissue independent of blood levels [16]. Therefore, in order to understand the local tissue effects of fluvastatin treatment, we chose to measure messenger RNA (mRNA) levels of HMGCR within the explanted xenografts. Statin inhibition of HMG-CoA reductase enzymatic activity in normal cells and statin-sensitive cancer cells [17] is known to activate a series of feedback responses, including modulation of AMPK, which in turn fine-tunes the levels and activity of HMGCR and Sterol regulatory element binding protein (SREBP)1 and SREBP2, leading to homeostatic levels of the cholesterol pathway [17‐19]. Analyses of HMGCR levels in the explanted xenografts, to our surprise, revealed that lesions from the fluvastatin treatment group of our mouse xenografts did not show uniform suppression of HMGCR mRNA (Fig. 5c). Rather we found a steady increase in HMGCR with increase in histologic grade of progression; the least progressed histological grade lesions (grade 0 and grade 1) expressed lower levels of HMGCR transcript, indicating more effective inhibition of the cholesterol pathway. Conversely, we found about 200-fold higher levels of HMGCR in higher-grade lesions (grade 2 and grade 3) in the fluvastatin-treated group. In contrast, the basal levels of HMGCR transcript was relatively uniform in the vehicle-treated xenografts (maximum 16-fold change across lesions compared to 200-fold in the statin-treated group), suggesting that the variation in HMGCR mRNA levels in the fluvastatin group reflects the normal homeostatic cellular response to inhibition of the cholesterol biosynthesis pathway.

If, indeed over activation of the MVA pathway feedback loop and thus insufficient suppression of HMGCR contributed to resistance to statins (Fig. 6a), we predicted, based on the known feedback mechanism that AMPK-activating drugs, such as aspirin and metformin, will inhibit this feedback response and potentiate the ability of statins to inhibit HMGCR (Fig. 6b), leading to more effective abrogation of cell growth. To test this, we treated MCF10.DCIS cells, which are relatively resistant to statins (IC50 of MCF10.DCIS cells was significantly higher than that of MCF10.AT1 cells, p < 0.01, Fig. 4c) with a combination of fluvastatin and varying concentrations of the AMPK-activating drugs aspirin and metformin for 48 h. Fluvastatin treatment induced HMGCR protein expression (Fig. 6c and d). As postulated, aspirin and metformin substantially abrogated the fluvastatin-induced HMGCR protein expression by 50% (2.8-fold increase in HMGCR with fluvastatin vs 1.3 to 1.4-fold increase with the combination therapy as compared to vehicle treatment, Fig. 6c and d). In tandem, we found that aspirin and metformin increased levels of pAMPK. This was confirmed by western blot analyses that demonstrated a dose-dependent increase in pAMPK levels with aspirin (2.77-fold to 5.18-fold) and metformin (5.83-fold to 7.8-fold) (Fig. 6c and d). Compound C blocked the aspirin-induced pAMPK activation and consistently led to a partial restoration in HMGCR levels (data not shown). These data are consistent with the model wherein the homeostatic feedback loop, restoring HMGCR levels, can be blocked by activation of pAMPK (Fig. 6b).

We next tested the functional significance of dual targeting of the cholesterol pathway. We performed colony formation assays, treating DCIS cells with aspirin (0.5 mM and 1 mM) in combination with fluvastatin (5 μM) and found that such dual treatment completely inhibited the ability of cells to form colonies at both 0.5 mM and 1 mM dose of aspirin (100% inhibition, Fig. 7a and b, p < 0.001), compared to single-agent treatments. Similarly, combination treatment with fluvastatin (5 μM) and metformin (5 mM) also substantially reduced colony formation more than fluvastatin and metformin alone (Fig 7a and b, p < 0.001). Although treatment with fluvastatin alone had modest efficacy (35% inhibition in total colonies), a combination of fluvastatin and aspirin or metformin was most effective (100% inhibition by aspirin and 99% inhibition by metformin (5 mM)) to overcome the inherent resistance to statin seen in DCIS cells. This is consistent with the reported complexities of MVA pathway regulation due to feedback activation loops, which leads to requirement of blockage of need to block this pathway at two nodes in order to completely inhibit this pathway, to inhibit cellular growth [20].

We next examined whether activating AMPK with aspirin was also effective in overcoming adaptive resistance to fluvastatin, a clinically relevant scenario that may have led to the ineffectiveness of fluvastatin in preventing the histological progression of MCF10.AT1-driven xenografts. We first generated a model of adaptive resistance to fluvastatin by continuously exposing MCF10.AT1 cells to increasing doses of fluvastatin (up to 20 μM) to create an MCF10.AT1-R line. MCF10.AT1-R cells showed significantly higher resistance relative to parental MCF10.AT1 cells with IC50 of 8.5 μM compared to 2.1 μM in the parental AT1 line (Fig. 8a). We next tested whether adaptive resistance of MCF10.AT1-R cells to fluvastatin can be overcome with AMPK-activating drugs. Cell proliferation was assayed by performing the MTT assay with a range of aspirin alone, metformin alone and finally fluvastatin with or without simultaneous exposure to aspirin or metformin. Proportions of cell death under each experimental condition were calculated and entered into Calcusyn, software that determines combined drug effects by taking into account the entire shape of the growth inhibition curve and performs multiple drug dose-effect calculations as described by Chou and Talalay [21]. The output of this assay is a “combination index (CI)” that is calculated by the median drug-effect analysis method and suggests whether a drug combination is synergistic (CI < 1), additive (CI = 1) or antagonistic (CI > 1). Through these analyses, we identified a range (that included IC25, IC50, IC75) of aspirin (0.5–10 mM) or metformin (0.5–10 mM) to have CI < 1 and thus be clearly synergistic with fluvastatin (5–100 μM) in MCF10.AT1-R cells (Fig. 8b, and by normalized isobolograms in Additional file 4: Figure S3 A). For generating these combination dose-response curves, the IC50 of aspirin alone, and of metformin alone were also determined from their drug-response curves (Additional file 4: Figure S3 B and C).

Last, we tested whether aspirin or metformin in combination with fluvastatin also inhibits the colonizing ability of MCF10.AT1-R (adaptive resistant) cells (Fig. 9). Similar to assays in the inherently resistant MCF10.DCIS line, and consistent with the synergistic interaction between fluvastatin and aspirin/metformin, these assays showed that combination therapy with fluvastatin/aspirin to completely inhibit colony formation (100% inhibition by both 0.5 mM and 1 mM) or fluvastatin/metformin (100% inhibition by 5 mM and 82% inhibition by 1 mM) to be more effective at overcoming adaptive resistance than each drug alone (51% inhibition by fluvastatin, 30% by aspirin and 24% by metformin).