Atorvastatin facilitates chemotherapy effects in metastatic triple-negative breast cancer

Abstract

Background

Metastatic triple-negative breast cancer (mTNBC) is treated mainly with chemotherapy. However, resistance frequently occurs as tumours enter dormancy. Statins have been suggested as effective against cancer but as they prolong and promote dormancy, it is an open question of whether the concomitant use would interfere with chemotherapy in primary and mTNBC. We examined this question in animal models and clinical correlations.

Methods

We used a xenograft model of spontaneous metastasis to the liver from an ectopic tumour employing a mTNBC cell line. Atorvastatin was provided to sensitise metastatic cells, followed by chemotherapy. The effects of statin usage on outcomes in women with metastatic breast cancer was assessed respectively by querying a database of those diagnosed from 1999 to 2019.

Results

Atorvastatin had limited influence on tumour growth or chemotherapy effects in ectopic primary tumours. Interestingly, atorvastatin was additive with doxorubicin (but not paclitaxel) when targeting liver metastases. E-cadherin-expressing, dormant, breast cancer cells were resistant to the use of either statins or chemotherapy as compared to wild-type cells; however, the combination of both did lead to increased cell death. Although prospective randomised studies are needed for validation, our retrospective clinical analysis suggested that patients on statin treatment could experience prolonged dormancy and overall survival; still once the tumour recurred progression was not affected by statin use.

Conclusion

Atorvastatin could be used during adjuvant chemotherapy and also in conjunction with metastatic chemotherapy to reduce mTNBC cancer progression. These preclinical data establish a rationale for the development of randomised studies.

Background

TNBC is associated with early distal recurrence and rapid progression as compared to other types of breast cancer (BC). Visceral metastases are also more common in this group, with shorter survival times from recurrence until death [1]. Given the propensity for early metastatic relapse, adjuvant chemotherapy is often indicated, even in the absence of evident metastatic spread [2]. This toxic treatment reduces the recurrence rates by approximately one-third, by targeting of primarily cycling cells. BC cells that disseminate to distant sites may enter into a dormant state, rendering chemotherapy ineffective [3].

Prior epidemiologic studies of women with breast cancer reported a correlation between statin use for cardiovascular indications, and slightly delayed mortality [4–9], without reducing the incidence of primary BC. Subset analyses linked these effects to lipophilic statin use, as BC cells lack the transporters that hydrophilic statins use for entering hepatocytes and block 3-hydroxy-3-methyl-glutaryl CoA reductase (HMG-CoA), the rate-limiting step on cholesterol synthesis [10, 11]. These human correlations dovetail nicely with recent preclinical studies. Lipophilic, but not hydrophilic statins have been shown to limit the growth of aggressive, mesenchymal phenotype cancer cells in culture but not so much of those of epithelial features [12–14]. Further, in mouse models of spontaneous BC metastasis, statins had no effect on the growth of primary tumours but effectively delayed metastatic outgrowth [13]. For instance, statins may suppress breast cancer cell proliferation and prolong the dormant stage, delaying or preventing recurrences [13].

Thus, statins could perhaps reduce the efficacy of chemotherapies by limiting the fraction of cells in cycle, they also act by inhibiting prenylation of isoprenyl groups that activate small G proteins. These participate in proliferation, migration, and survival [13–15]. Therefore, describing the therapeutic efficacy of conventional chemotherapy in the presence of statins may guide the optimisation of therapies to prevent and treat metastatic spread. For instance, would lipophilic statins interfere with chemotherapies or augment their effects on prolonged overall survival?

Herein, we outline that statins downregulate cell cycle of metastatic BC cells in vitro and in vivo; this is further reduced by chemotherapy without significantly affecting cell apoptosis. In addition, metastatic cells could upregulate E-cadherin and become resistant to statins and chemotherapy when used separately but may be susceptible when used concomitantly. Finally, through retrospective analyses of a metastatic BC patient database, we found that patients on statins through the course of their disease experienced delayed first metastasis from time of the diagnosis of the primary tumour, and overall survival, but were not distinguishable from non-statin-treated patients once metastatic progression/recurrence clinically developed. These findings constitute preclinical and clinical basis for the development of randomised trials that evaluate statin maintenance or incorporation for women with TNBC when starting chemotherapy in the adjuvant or recurrent disease.

Methods

Cell lines and culture methods

MDA-MB-231 ± red fluorescent protein (RFP) or RFP-E-cadherin, and MDA-MB-468 were maintained in Roswell Park Memorial Institute (RPMI) containing 10% fetal bovine serum (FBS) and 0.5% penicillin/streptomycin. RFP and RFP-E-cadherin were stably transfected as previously described [16]. Puromycin (5 μg/mL for MDA-MB-231-RFP) and G418 (900 μg/mL for MDA-MB-231-RFP-Ecad) were used to maintain transfection cassettes. Experiments were performed using 2% FBS.

Human hepatocytes were obtained from excess specimens through the Liver Tissue Cell Distribution System, Pittsburgh, PA (NIH #HSN276201200017C). Hepatocyte maintenance media consisted of a base of William’s E Medium supplemented with Hepatocyte Maintenance Supplement Pack.

Reagents

Atorvastatin was dissolved in dimethyl sulfoxide (DMSO) at a stock concentration of 50 mM. Starting concentrations in DMSO were prepared for atorvastatin (50 mM) and paclitaxel (100uM).

Cell viability

MDA-MB-231 and MDA-MB-468 were cultured in 5% FBS and seeded in 48 well plates at 2 × 10⁴ cells/well and starved overnight. On day 1 atorvastatin was added at 0.5 μM and 5 μM, respectively. On day 2, chemotherapy was added with or without atorvastatin for a further 24 h. Cells were then fixed in 3.7% formaldehyde and stained with 0.5% crystal violet. Excess crystal violet was removed using water. Once dried, 0.2% sodium dodecyl sulfate (SDS) was added to the wells to extract absorbed dye and then transferred to a 96-well plate. Absorbance was measured at 560 nm using a Tecan SpectraFluor microplate reader (Durham, NC, USA).

Immunoblotting

For protein extraction, cells were lysed using RIPA buffer containing 1:100 protease inhibitor cocktail V and 1 mM Na₃VO₄. After 10 min, cells were scraped, sonicated 3 times for 3 s each and centrifuged at 12,000 rpm × 12 min at 4 °C. Samples were added 5× laemmli buffer prior to boiling for 5 min and loading into 8–10% Tris Bis-acrylamide gels. Samples ran at 93–108 V, and then transferred to nitrocellulose membrane at 300 mA for 75 min. Blocking was performed with 5% non-fat milk (5% BSA for phosphorylated protein) in tris buffered saline with 0.5% Tween-20 (TBST).

Animal studies and drug administration

Eight-week-old NOD-SCID gamma (NSG) female mice were purchased from the Jackson Laboratories and maintained in the animal facility of the VA Pittsburgh Healthcare System (Animal Biosafety Level 2) for one week before starting the experiment. Mice were maintained in individual cages and euthanized on the 33rd day of the experiment, using American Veterinary Medical Association (AVMA)-approved methods. Mice were taken individually into a separate room and introduced in a chamber with exposure to gradual increase in CO₂ (approximately 30% vol/min) until complete lack of breathing occurred. Spleen, liver and serum were harvested for analysis. There were five experimental and a negative control group. Between 3 and 4 mice were randomly allocated (doxorubicin, paclitaxel and atorvastatin + paclitaxel were 4 mice on each group, the rest 3 mice/group).

Tissue histology

Harvested tissue was either fixed in 10% neutral formalin and then paraffin embedded or maintained in liquid nitrogen and stored at −80 °C until further use.

Paraffin-derived slides were stained for cCasp 3, Ki-67, pan-Akt, pan-Erk, p-Akt and p-Erk, hematoxylin and eosin at the PBC. TUNEL and EdU staining was performed in house. For quantification and comparison of Ki-67, EdU, TUNEL or active caspase 3 expression among treatment groups, positive cells/tumour area were quantified and normalised using ImageJ.

Immunofluorescence

EdU staining was completed following the manufacturers protocol. Then, coverslips were blocked with 3% bovine serum albumin (BSA) in PBS for 45 min, followed by overnight, 4 °C primary antibody incubation. After washing and secondary antibody incubation for 1 h, Hoechst (5 μg/mL) was added for 15 min.

For frozen immunostaining, slides were fixed with 100% methanol for 20 min at −20 °C and posteriorly air-dried. Cells were washed and permeabilized with 0.1% Triton-X-100 in PBS and blocked with 3% BSA in PBS; the EdU+ Cocktail was then added for 30 min. Blocking with 20% goat serum in 0.5% BSA in PBS (PBB) was performed, followed by primary antibody incubation in 10% goat serum in PBB. Secondary antibodies were provided by the Center of Biologic Imaging.

Hepatocyte cocultures

Twelve-well plates with Nunc Thermanox Plastic coverslips (#174950, ThermoFisher) were coated with 0.1% collagen, and maintained at 37 °C for 1 hour. Primary human hepatocytes were seeded at 6 × 10⁵ cells/mL in Williams E Medium (WEM). After 24 h, 1000 MDA-MB-231-RFP cells were added to the wells.

Image analysis

Positive cells were analysed per field area and normalised to the control group. QuPath was used to semi-quantify whole scans of Ki-67-stained slides. For analyses of positive cells/metastatic area, the number of positive cells expressing the antigen of interest were quantified, divided by the metastatic nodular area and normalised to their control group [17].

For image analysis of immunohistochemical staining, DAB areas were measured after colour deconvolution and thresholding using ImageJ v1.51.

To distinguish E-cadherin(+) cancer cells, RFP cells with surrounding E-cadherin through z-stacked confocal images were considered to be positive.

Study population of cancer patients

The present study includes a retrospective analysis of the Magee Women’s Hospital Breast Cancer Program database, which includes 1749 patients with mBC treated from 1999 to 2019. The database contains demographic and clinical information of women from western Pennsylvania (in majority) who have undergone mBC clinical care at UPMC. HER2 positivity was based on +3 IHC or +2 IHC with positive HER2 FISH. Statins used, dose dates of birth and death, and dates of prescription were retrieved from medical records by the Research Informatics Office. This information was then merged with our current patient database. Access to the patient records was by honest broker and provided without PHI (personal health information), as approved by the University of Pittsburgh Institutional Review Board (IRB).

Statistical analysis

Statistical analyses for in vivo and in vitro data were performed using GraphPad Prism v8.0. In vitro data is represented as the mean from at least 3 independent experiments. Experiments were analysed using the non-parametrical Kruskal–Wallis ANOVA test with either Tukey’s, Sidak’s or Dunn’s multiple correction. For animal analyisis of metastasis, statistics were obtained from metastatic nodules as the sample size, as previously described [17].

Clinical data were analysed using R v3.6.3, including the packages survminer, survival, and ggpubr. T-tests with Welch’s correction or the Kruskal–Wallis test were performed to compute differences between statin and non-statin users. The Pearson’s Chi-squared test was used to calculate differences in the distribution of bisphosphonate use between statin and non-statin users. Univariate and multivariate analyses were done using the Cox proportional regression model. This model provided hazard ratios and confidence intervals, which are documented in the present manuscript (Table 1). Statistical significance for survival curves was computed using log-ranks. P < 0.05 was considered statistically significant.

Table 1.

Univariate and multivariate analysis based on overall survival, time to relapse and survival after metastasis, from the retrospective cohort.

	Overall survival						Time to relapse						Survival after metastasis
	Univariate			Multivariate			Univariate			Multivariate			Univariate			Multivariate
	Beta coef	p	CI	HR	p	CI	Beta coef	p	CI	HR	p	CI	Beta coef	p	CI	HR	p	CI
Statin use
Yes	0.7224	0.00135	(0.59–0.88)	0.7702	0.01	(0.62–0.95)	0.6899	0.003	(0.53–0.88)	0.63	<0.001	(0.49–0.82)	0.9	0.4	(0.75–1.12)
ER
Negative
Positive	0.53	<0.001	(0.46–0.61)	0.5137	<0.001	(0.44–0.6)	0.6302	<0.001	(0.54–0.72)	0.5833	<0.001	(0.49–0.68)	0.58	<0.001	(0.5–0.66)	0.53	<0.001	(0.44–0.61)
HER2
Negative
Positive													0.63	<0.001	(0.43–0.91)	0.47	<0.001	(0.31–0.69)
NDI
<65
≥65
Type of BC
Ductal	1.52	0.004	(1.14–2.02)				1.47	0.01	(1.09–1.98)
Lobular
Other	1.46	0.0372	(1.02–2.08)				1.5	0.03	(1.04–2.16)
De novo metastatic
No
Yes	2.915	<0.001	(2.41–3.52)
Type of AC
None
Anthracycline	0.7725	<0.001	(0.66–0.89)	0.7337	<0.001	(0.62–0.86)	0.666	<0.001	(0.57–0.77)	0.6412	<0.001	(0.54–0.75)
Non-anthracycline	0.6354	<0.001	(0.53–0.76)	0.6117	<0.001	(0.5–0.74)	0.5498	<0.001	(0.46–0.66)	0.5152	<0.001	(0.42–0.62)
TNM at diagnosis
1
2
3
4	1.499	<0.001	(1.38–1.62)	1.4837	<0.001	(1.36–1.6)	1.63	<0.001	(1.48–1.75)	1.617	<0.001	(1.16–1.56)	0.12	0.001	(1.05–1.22)
Age at diagnosis
≤50
>50	1.428	<0.001	(1.25–1.63)				1.36	<0.001	(1.19–1.56)	1.3528	<0.001	(1.48–1.76)	1.22	0.003	(1.07–1.38)

Covariates measured were Statin use, ER status, HER2 status, National Deprivation Index (NDI), type of breast cancer (BC), de novo metastatic, type of adjuvant chemotherapy (AC), TNM at diagnosis, and age at diagnosis.

Statistically significant p < 0.05 values are in bold.

Additional methods

Information on catalogue number, manufacturer and location, of antibodies, reagents, media and pharmaceuticals can be found in Supplementary Material.

Results

Efficacy of chemotherapy on ectopic primary tumours of a spleen-to-liver model is not affected by the use of atorvastatin

As chemotherapy is considered most effective when used on dividing cells [18], and statins have been shown to limit the fraction of cycling cancer cells [17], we investigated whether the response to chemotherapy would be diminished in the presence of atorvastatin, which was chosen for its lipophilicity [15, 17]. We used a TNBC cell line on a spleen-to-liver model of primary tumour and spontaneous metastases. Parental MDA-MB-231 cells seldom metastasise to the liver on mammary fat pad models [19]. Therefore, we chose to use spleen to liver spleen-to-liver for its consistent metastases to the liver, and its demonstrated applicability for therapeutic testing [17, 20–23].

We injected intrasplenically fifty thousand Red Fluorescent Protein (RFP)-labelled MDA-MB-231 cells (5 × 10⁵ cells/mL) diluted in sterile Hank’s Balanced Salt Solution (HBSS) (Fig. 1a), as previously described to establish an ectopic primary tumour [17, 20]. Intraperitoneal (IP) injections have previously been used in mice to assess statin effects [24–26]. Therefore, an IP dose of 10 mg/kg atorvastatin was administered every day starting on day 10 [17]. From day 24, IP mono-chemotherapy was given every other day until day 32 (5 total injections). The ectopic tumours were then challenged with either doxorubicin (2 mg/kg) or paclitaxel (10 mg/kg) [20, 27, 28], as they are standard therapies for mTNBC and have different mechanisms of action [2]. To ascertain the fraction of cycling cells, an EdU + solution was injected IP two days before euthanizing mice (Fig. 1a). Untreated mice developed protruded tumours that grew beyond the spleen (Fig. 1b). Atorvastatin alone was not sufficient to reduce primary tumour or absolute spleen size (Fig. 1b, c), nor to downregulate cell cycle in the primary site (Fig. 1d, e, g), as this group previously reported [17]. In both doxorubicin- and paclitaxel-treated mice there was reduced spleen area, but this was not altered by the addition of atorvastatin (Fig. 1b, c).

Fig. 1. Efficacy of chemotherapy on ectopic primary tumours of a spleen-to-liver model is not affected by the use of atorvastatin.

a Spleen-to-liver model of metastasis workflow. 5 × 10⁴ MDA-MB-231-RFP cells were inoculated intrasplenically on day 0, and on day 10 mice started receiving 10 mg/kg atorvastatin or 2% DMSO every day until day 32. On day 24 mice started receiving 2 mg/kg doxorubicin or 10 mg/kg paclitaxel every other day until day 32. Mice were euthanized on day 33. b Representative images from gross primary tumour formations in the spleen. c Spleens containing the primary tumour were cross-sectioned. Areas were then measured and normalised to the negative control. d Spleens were also stained for TUNEL, and positive nuclei/tumour area were quantified and normalised to the negative control. e Ki-67-stained slides from spleens were analysed by semi-automatic quantification using QuPath software and normalised to the negative control. f Thymidine uptake was quantified and measured as EdU + cells/area. g Representative images of EdU(+)-stained spleens. Images were taken using the 60X magnification on a widefield microscope. At least 3 mice were challenged in each treatment group. Kruskal–Wallis one-way ANOVA with Dunn’s multiple comparison was performed for all figures. Top bar represents median variation for ANOVA analysis. Data represented as mean with SD. *(<0.05), **(<0.01). ATO atorvastatin, DOX doxorubicin, PTX paclitaxel.

The same was true for DNA fragmentation and cell cycle progression reduction (Fig. 1d, Supplementary Fig. 1A). Although there was a numerical reduction in tumour cells in cycle with chemotherapy as measured by Ki-67, this did not reach statistical significance (Fig. 1e, Supplementary Fig. 1B). In summary, atorvastatin pretreatment did not influence the efficacy of either chemotherapy on the primary splenic tumour.

Efficacy of chemotherapy on metastases in the liver is enhanced in the presence of atorvastatin

We previously found that atorvastatin could reduce metastatic outgrowth [17]. Herein, we then wanted to evaluate if the chemotherapeutic efficacy on breast cancer metastases remained unchanged in atorvastatin-treated mice. In terms of toxicity derived from therapies, we found that mice did not evidence significant weight loss (>15% of total weight) (Supplementary Fig. 2A). Serum analyses showed elevated alanine (ALT) and aspartate (AST) aminotransferase levels in doxorubicin-treated mice, which could suggest compromised hepatocyte function [29]. Contrary to other studies, atorvastatin did not minimise doxorubicin hepatotoxicity (Supplementary Fig. 2B) [30]. All livers lacked macroscopically visible metastases (Supplementary Fig. 2C, D).

We looked at the antimetastatic effects of chemotherapy with or without atorvastatin. First, we found that atorvastatin alone reduced the fraction of cancer cells that had transited S-phase (EdU+) without changing the number of cells in cycle (Ki67+), validating prior findings [13, 31]. Different chemotherapies affected these metastases differentially in the face of atorvastatin pre treatment. EdU+ uptake was further reduced for those receiving doxorubicin, but not for mice treated with atorvastatin and paclitaxel, though this may be due to the more reduced levels in the face of paclitaxel (Fig. 2a, b). Cell growth also was faster in the metastatic nodules compared to the primary splenic tumours, and atorvastatin in combination with doxorubicin was more effective in the liver metastases relative to the primary splenic tumours (Supplementary Fig. 2E).

Fig. 2. Efficacy of chemotherapy on metastases in the liver is enhanced in the presence of atorvastatin.

a Slides were obtained from frozen specimens and stained for RFP and EdU+. Images were then obtained using a ×40 magnification through confocal microscopy. Proliferating tumour cells to be interpreted as EdU+ within RFP+ cells. b Proliferating cells per micronodule area were measured and quantified. c Quantification of dividing cells was also assessed by Ki-67 staining. Images were taken using the ×60 magnification through widefield microscopy. d Ki-67 staining was measured and quantified as number of positive ki-67 nuclei per nodular area, and then normalised to the negative control. e Apoptosis was also measured through active caspase 3 staining. Images were taken using the ×60 magnification through widefield microscopy. Active caspase 3 staining was measured and quantified as number of positive stained cells per nodular area, and then normalised to the negative control. Ki-67 (f) and active caspase 3. At least 3 mice were analysed for each treatment group. Kruskal–Wallis one-way ANOVA with Dunn’s multiple comparison was performed for (b, d, f), whereas the Tukey’s correction was performed for g–h. Data represented as min to max values. *(<0.05), **(<0.01), ***(<0.001), ****(<0.0001). ATO atorvastatin, DOX doxorubicin, PTX paclitaxel.

To place our findings of examining cells in cycle versus those having undergone mitosis, we also enumerated Ki-67 positivity among all treatment groups. Atorvastatin did not reduce Ki-67 expression, neither did doxorubicin alone, while a two-fold reduction was seen with paclitaxel (Fig. 2c, d). Interestingly, doxorubicin significantly reduced the Ki-67(+) fraction in mice treated with atorvastatin. Paclitaxel added to atorvastatin did not further reduce significantly the two-fold reduction of Ki-67 expression seen with single use of paclitaxel, but again this was at a very low level already.

Metastatic breast cancer cells subjected to paclitaxel or doxorubicin also undergo apoptosis [32]; this was assessed by cleavage of caspase-3. Any chemotherapy alone increased the fraction of apoptotic cells (Fig. 2e, f). Remarkably, atorvastatin further increased the fraction of apoptotic cells in doxorubicin-treated metastases. This difference was not seen when comparing paclitaxel with atorvastatin plus paclitaxel (Fig. 2c, d).

We then analysed the distribution of metastatic clusters by liver zones that differ in oxygenation levels [33], finding that tumour cells were evenly distributed throughout the liver and similarly in the cell cycle (Supplementary Fig. 2F). Mice treated with both statin and a chemotherapy, cells present around the pericentral vein (Zone 3) showed the lowest proliferation rates and highest apoptosis when compared to cells in pericentral zones from other treatment groups (Supplementary Fig. 2G, H). This suggests that doxorubicin may preferentially affect cells under hypoxic stress within the liver, though further exploration is needed into this question.

Statin sensitisation correlates with inhibition of Akt phosphorylation

To parse the biological mechanisms underlying the responses noted in vivo, we evaluated these effects in vitro. First, we found that inhibiting HMG-CoA reductase (HMGCR) in vitro decreased proliferation of two phenotypically distinct TNBC cell lines (MDA-MB-231 and MDA-MB-468) (Fig. 3a). Although statins may reduce cell cycle turnover, they also may disrupt pro-survival signalling pathways [15, 32], as cell cycle progression and survival are both maintained by the PI3K/Akt pathway [34], and its phosphorylation cascade is inhibited in response to statins in breast cancer [15, 35, 36]. As either effect could influence the cellular response to chemotherapy, we decided to test this rationale in vitro. As a proof of concept, we used different FBS concentrations to determine that chemotherapy is less effective in low-proliferative conditions (Supplementary Fig. 2A); this rationale was used to propose that statins could decrease the efficacy of chemotherapy through cell cycle arrest. Atorvastatin blunted EGF-induced Akt phosphorylation in both the basal B line MDA-MB-231 and the basal A line MDA-MB-468 (Fig. 3b, c), validating prior results [15]. We also utilised the PI3K inhibitor LY294002 to show that its cellular effects are directly comparable to those of atorvastatin. In this regard, the effects of 5 μM atorvastatin were similar to 5 μM LY294002 in inhibiting the PI3K pathway in MDA-MB-468 cells (Supplementary Fig. 3B). Inhibition of phosphorylation of Akt was also accompanied by decreased p-Erk in MDA-MB-231 cells (Fig. 3B, C), although whether this response accounted downstream of p-Akt inhibition is beyond the scope of this study. In contrast with other reports [37], unchanged beclin-1 reflected that atorvastatin did not alter autophagy of tumour cells (Fig. 3b, c, Supplementary Fig. 3C, D).

Fig. 3. Statin sensitisation correlates with inhibition of Akt phosphorylation.

a BC cells were seeded at a density of 2 × 10³ (MDA-MB-231) or 5 × 10³ (MDA-MB-468) on 96-well plates. Atorvastatin was given at doses between 0 and 100 μM one day after seeding. After 24 h, media was substituted for an EdU(+)-containing serum-free media for 5 h. Next, cells were fixed and EdU(+)-stained. Cells were then imaged using ×20 magnification on a widefield microscope and EdU/Hoechst ratios were obtained. Doses were transformed to log10 values and nonlinear fit curves were performed. IC50 values shown in the micromolar range. MDA-MB-231 (b) and MDA-MB-468 (c) were seeded on 6-well plates at a density of 3 × 10⁵ cells and 2 × 10⁵ cells, respectively, and treated with atorvastatin or 0.2% DMSO for 24 h and then subjected to 5 nM EGF for 1 h, after which lysates were obtained. Western blotting was performed to determine protein levels. d A multi-sourced survival curve plotter was used to estimate survival from TNBC based on HMGCR expression. e Gene expression of HMGCR across breast cancer metastases from a publicly available dataset was accessed. f Liver sections from untreated and atorvastatin-treated mice were immunostained; protein levels were calculated using ImageJ. ATO atorvastatin. Scale bar = 20 μm.

Next, we accessed a multi-sourced survival curve plotter (https://kmplot.com/analysis/) to evaluate impact on survival of HMGCR expression in TNBC tumours. We found that HMGCR was associated with increased mortality (hazard ratio (HR) = 1.87) (Fig. 3d). Next, we obtained the expression data from breast tumours on the met500 dataset [38], an integrative sequencing analysis of multiple metastatic cancers, to evaluate if HMGCR was differentially expressed in primary tumours vs distant organ metastases. All organs but the brain had numerically higher expression of HMGCR as compared to primary breast tumours. Differential expression in liver trended to statistical significance (p = 0.08) (Fig. 3e). We then found that atorvastatin reduced the expression of HMGCR of liver metastases in our mouse model (Fig. 3f), similar to a reported phase II window-of-opportunity trial that found that atorvastatin decreased proliferation of HMGCR-positive tumours before surgery [39]. Given that metastatic tumours may express increased HMGCR, atorvastatin could effectively reduce progression of metastatic disease.

Statins inhibit the PI3K/Akt signalling transduction cascade in vitro [15]; we also questioned if these sensitising effects could be observed in vivo. Overall, atorvastatin-treated mice had a numerically lower but nonsignificant expression of total (pan)-Akt, phosphorylated Akt, (p-Akt) pan-Erk, and p-Erk in the metastatic liver nodules (Fig. 3f).

We validated our findings using a mammary fat-pad-to-lung model of BC previously utilised in our laboratory [17]. Briefly, the model consists of a matrigel-based orthotopic inoculation of breast cancer cells into the fat pad of NOD-SCID mice, wherein lung metastasis eventually occurs stochastically [40]. Here, the levels of pan-Akt were reduced both in the primary site and the lung. In addition, while atorvastatin didn’t reduce the levels of p-Akt in the mammary xenograft, it reduced phosphorylation in the lung. Meanwhile, levels of total Erk were reduced in the lung, yet its phosphorylation was not diminished upon atorvastatin treatment. This pointed towards an Akt-associated growth of metastatic nodules, targetable by atorvastatin usage (Supplementary Fig. 3c, d).

Breast cancer cell sensitisation with atorvastatin enhances antiproliferative effects of chemotherapy in vitro

Some studies have investigated cancer cell response to statins in combination with chemotherapy in vitro [41]. To reflect the in vivo results, we wanted to investigate if the use of low-dose atorvastatin would synergise, antagonise, or not change the effects of chemotherapies used against the two aforementioned mTNBC cell lines. First, we concomitantly gave atorvastatin and either doxorubicin or paclitaxel for the course of 48 h. We found that in MDA-MB-231 cells, 0.5 μM atorvastatin did not shift the IC50 curve of doxorubicin (Supplementary Fig. 4A), whereas it just partially shifted leftwards the curve when used with paclitaxel, compared to paclitaxel alone (Supplementary Fig. 4B). When combining atorvastatin with paclitaxel or doxorubicin in MDA-MB-468 cells, a partial additive effect was seen (Supplementary Fig. 4C, D).

Next, we sought to investigate if chemotherapy effects were modified in previously statin-sensitised TNBC cells. Atorvastatin was given at a dose of 0.5 μM (MDA-MB-231) or 5 μM (MDA-MB-468) for 24 h, followed by atorvastatin + 1 μM doxorubicin or atorvastatin + 0.1 μM paclitaxel, for another day. Media was then changed for a serum-free, EdU-containing media for 5 h, followed by fixation. In either MDA-MB-231 and MDA-MB-468, cPARP staining showed minimal apoptotic activity for both chemotherapies at the chosen concentration, although a numerically higher level was seen with atorvastatin and paclitaxel in MDA-MB-231 cells (Fig. 4a, b). Intriguingly, passage through S phase when combining atorvastatin with doxorubicin both in MDA-MB-231 and MDA-MB-468 cells was almost absent, suggesting sensitisation to the use of anthracyclines (Fig. 4a–d). On the other hand, atorvastatin only partly decreased cell cycle fraction when used with paclitaxel (Fig. 4b, d).

Fig. 4. Breast cancer cell sensitisation with atorvastatin enhances antiproliferative effects of chemotherapy in vitro.

a MDA-MB-231 cells were seeded at a density of 2 × 10⁵ cells/well (day 0) on 6-well plates. On day 1, cells received 0.5 μM atorvastatin. After 24 h, chemotherapy (1 μM doxorubicin or 0.1 μM paclitaxel) was added for a day, with or without atorvastatin. Cells were imaged at ×20 magnification using widefield microscopy. b EdU+ and cPARP expression were quantified using ImageJ. The same experimental protocol was applied to MDA-MB-468 cells (5 μM atorvastatin) (c), as well as the image analysis (d). e Hepatocytes were cocultured with 1000 MDA-MB-231-RFP cells. Atorvastatin was administered for 24 h, followed by chemotherapy (1 μM doxorubicin or 0.1 μM paclitaxel) with or without atorvastatin. After 24 h, media was substituted for fresh, EdU(+)-containing media for 24 h. Cells were imaged using ×20 magnification through confocal microscopy. f EdU/RFP ratios were obtained by positive cell quantification using ImageJ. g Cell lysates were also obtained from MDA-MB-231 and MDA-MB-468 j cells after 24 h atorvastatin and 24 h more of atorvastatin + chemotherapy, and protein levels were then obtained. Kruskal–Wallis one-way ANOVA with Tukey’s correction was performed for (b, d, f). Data represented as mean with SEM of at least three independent experiments. *(<0.05), **(<0.01), ***(<0.001), ****(<0.0001). p = p-value for EdU, p^† = p-value for cPARP. ATO atorvastatin, DOX doxorubicin, PTX paclitaxel.

The metastatic microenvironment differs from that in the primary tumour, orchestrating unique response to therapies [42, 43]. Herein, to better mimic this context, we performed a 2D coculture of MDA-MB-231 cells with fresh human hepatocytes. On day 0, hepatocytes were seeded and maintained in WEM. On day 1, 1000 MDA-MB-231-RFP cells were added to the culture well. On day 3, atorvastatin was given at a dose of 1uM, and on day 4 chemotherapy was added (1 μM Doxorubicin and 0.1 μM Paclitaxel) with or without atorvastatin. On day 5, media was substituted for a drug-free, EdU-containing WEM media for 24 h. We found that both doxorubicin and paclitaxel reduced proliferation when in combination with atorvastatin, as compared to chemotherapy alone (Fig. 4e, f).

To investigate whether if atorvastatin enhanced the proapoptotic effect of doxorubicin and atorvastatin at the dosage utilised, we tracked caspase-3/7 activation throughout 48 h. As previously described, atorvastatin was given for 24 h, followed by atorvastatin + chemotherapy. Immediately thereafter, plates were inserted into an IncuCyte^® incubator. Forty-eight-hour live-cell imaging showed that pretreatment of BC cells with low-dose atorvastatin only minimally sensitised paclitaxel both in MDA-MB-231 and MDA-MB-468 cells (Supplementary Fig 4E, J).

Following the previous treatment protocol, cells were also lysed 24 h after atorvastatin + chemotherapy. We found that cell cycle was greatly reduced in atorvastatin-sensitised cells upon addition of both doxorubicin and atorvastatin, as seen by decreased levels of p-Rb and Cyclin-D1 (Fig. 4g). Remarkably, paclitaxel increased levels of both proteins, probably due to formation of multipolar spindles [44]. Lastly, atorvastatin and chemotherapy did not enhance the expression of caspase 9 compared to chemotherapy alone (Fig. 4g, Supplementary Fig. 4K, L).

Atorvastatin enhances partial cancer-associated mesenchymal-to-epithelial reverting transition (MErT)

Metastatic dormancy has been linked to an epithelial phenotype [45]. Loss and re-expression of E-cadherin constitutes the most relevant marker of epithelial-to-mesenchymal plasticity [46]. Therefore, we sought to investigate if epithelial transition of mesenchymal-like cancer cells could be facilitated by atorvastatin. Fluorescent staining for E-cadherin, RFP and EdU revealed that metastatic MDA-MB-231 cells presented higher levels of membrane E-cadherin (Fig. 5a, b). We proposed that atorvastatin-induced E-cadherin expression would halt the proliferation of metastatic cells. However, among statin-treated cells, we found no differences in proliferation between E-cadherin(+) and E-cadherin(−) (Supplementary Fig. 5A, B), suggesting that E-cadherin re-expression upon inhibition of the mevalonate pathway does not always lead to quiescence, and even could possibly constitute an early marker of resistance. In addition, no treatment condition modified the proliferation of epithelialized MDA-MB-231 cells (Supplementary Fig. 5C). We then performed a 2D hepatocyte coculture with MDA-MB-231stably transfected with E-cadherin, as previously described [16], and analysed the S-phase progression in response to different therapies. Here, we found that both doxorubicin and paclitaxel reduced EdU(+) uptake in half, which was further reduced by another half when prior sensitisation with atorvastatin (Fig. 5c, d).

Fig. 5. Atorvastatin enhances partial mesenchymal-to-epithelial reverting transition (MErT).

a Frozen slides were stained for EdU, E-cadherin and RFP. Imaging was then performed using ×40 magnification on a confocal microscope. Yellow arrows point RFP-negative, E-cadherin-positive cells. White arrows point RFP-positive (MDA-MB-231), E-cadherin-positive cells. b Z-stacked images were then analysed with ImageJ quantification of E-cadherin (+) BC cells. At least three mice were challenged in each group. A Kruskal–Wallis test with multiple Dunn’s correction was performed. *(<0.05), **(<0.01). c Hepatocytes were cocultured with 1000 MDA-MB-231-RFP cells transfected with E-cadherin. Atorvastatin was administered for 24 h, followed by chemotherapy (1 μM doxorubicin or 0.1 μM paclitaxel) with or without atorvastatin. After 24 h, media was substituted for fresh, EdU(+)-containing media for 24 h. Cells were imaged using ×20 magnification through confocal microscopy. d EdU/RFP ratios were obtained by positive cell quantification using ImageJ. e MDA-MB-231-RFP cells and MDA-MB-231-RFP-Ecad were seeded at a density of 3 × 10⁵ cells/well in 6-well plates. Atorvastatin (0.5 μM) was given one day after seeding. After 24 h, chemotherapy (1 μM doxorubicin or 0.1 μM paclitaxel) were administered with or without atorvastatin for 24 h more. Protein levels were then obtained. *(<0.05), **(<0.01). ATO atorvastatin, DOX doxorubicin, PTX paclitaxel.

To elucidate if mesenchymal breast cancer cells which undergo MErT respond differently to statin-containing treatments, we sought to investigate treatment response of E-cadherin(+) MDA-MB-231 BC cells as opposed to E-cadherin(−), wild-type MDA-MB-231. In MDA-MB-231 cells transfected with E-cadherin, immunoblotting of BC lysates showed that canonical Akt signalling in E-cadherin(+) cells was not inhibited by the use of 0.5 μM atorvastatin, as opposed to E-cadherin(−) cells (Fig. 5e, Supplementary Fig. 5D, E), consistent with previous data describing less sensitivity of E-cadherin(+) cancer cells to statins [15].

E-cadherin expression was associated with reduced expression of apoptotic mediators (cCasp-9 and cPARP), despite doxorubicin treatment. Interestingly, we found that only after atorvastatin treatment, doxorubicin induced Casp-9 and PARP cleavage in these E-cadherin-expressing cells. Higher cCasp-9 and cPARP levels were also seen in E-cad (+) cells treated with atorvastatin and paclitaxel, compared to paclitaxel alone (Fig. 5e, Supplementary Fig. 5D, E). In sum, statin use could help overcoming chemotherapy resistance in metastatic epithelial-reverting phenotypes. The molecular interactions underlying this response await further studies.

Use of lipophilic statins is associated to prolonged dormancy and overall survival in a retrospective real-world analysis

Prior epidemiological studies suggest that statins may reduce BC-specific mortality by increasing overall survival [47–49]. To ascertain the clinical effect of these drugs through different stages of progression, we investigated real-world use of two lipophilic statins (atorvastatin and simvastatin), among 1749 patients treated for metastatic breast cancer (mBC) from 1999 until 2019 at UPMC. To reduce the risk of selection bias, we selected patients who were deceased at the time of analysis (n = 1016). From these, 122 (12%) patients were or had been on statins at some point throughout the course of their breast cancer. As expected, statin users were in average older at the time of primary (60 vs 52 years of age, respectively; p < 0.001) and metastatic (66.5 vs 55.7 years of age, respectively, p < 0.001) BC diagnosis (Supplementary Table 1). Importantly, we also found a 2-fold reduction in the development of liver metastases as a first relapse in statin, as compared to non-statin users (5.7% vs 10.4%, respectively) (Supplementary Table 1).

Next, we classified patients according to their estrogen (ER) and HER2 receptor status. We measured (1) the overall survival from the date of diagnosis; (2) the time to relapse, measured as time from date of diagnosis until the date of first metastasis (patients who were de novo metastatic or who started statin treatment after metastasis, were not included); and (3) the overall survival from metastasis to death. Log-rank analyses revealed increased overall survival among patients on statins who were ER-positive (p = 0.016), and HER2-negative (p = 0.017) (Fig. 6a, b). In addition, statins were associated with an increase in time until relapse among ER-positive (p = 0.03), and HER2-negative patients (p = 0.022) (Fig. 6c, d). When measuring OS after relapse, we found no differences between statin users and non-users, herein suggesting that statins could no longer be effective once metastatic progression occurs (Fig. 6e, f).

Fig. 6. Use of lipophilic statins is associated to prolonged dormancy and overall survival in a retrospective real-world analysis.

Atorvastatin and simvastatin usage data was retrieved from medical records of women with mBC who underwent care at UPMC between 1999 and 2019. a Overall survival (OS) was measured based on ER status and statin use. b OS was measured based on HER2 status and statin use. c Time to relapse was measured based on ER status and statin use. d Time to relapse was measured based on HER2 status and statin use. e OS after relapse was measured based on ER status and statin use. f OS after relapse was measured based on ER status and statin use. Log-ranks were performed to compute statistical difference in survival. Dot lines indicate median survival.

Given that patients who had ER-positive and HER2-negative mBC seemed to benefit the most from statin use, we analysed OS, time to relapse and OS from relapse of ER-positive, HER2-negative patients, finding a non-statistically significant trend that benefited statin users in terms of OS and time to relapse (Supplementary Fig. 6E, G).

Due to the in vivo and in vitro findings of this study, we also analysed the clinical course of patients with TNBC, as they could be suitable candidates for statin use as a secondary preventive of mBC and improvement of chemotherapy efficacy. For instance, we looked at time to relapse in the ER-negative, HER2-negative subgroup. We saw a trend that favoured statin usage, but the results did not reach statistical significance, possibly due to small sample size (p = 0.09) (Supplementary Fig. 6D).

To ascertain if statins could influence the effect of metastatic chemotherapy, we measured progression, considered as the time from the start date of first metastatic chemotherapy until the diagnosis of second metastasis. Although there were trends of delayed progression among statin users both in patients with ER-negative and HER2-negative mBC, statistical significance was not achieved (Supplementary Fig. 6E, F).

To evaluate if a hazard model could determine if statin use correlates with better outcomes, we performed univariate analysis through the Cox proportional hazard model. After adjusting for covariates, we saw that statin use, ER-positive, and adjuvant chemotherapy were predictors of longer overall survival and time to relapse, whereas ductal and other non-lobular breast cancer, de novo metastatic, stage IV and older than 50 years of age at diagnosis were predictors of poorer outcome, when measuring overall survival and time to relapse (Table 1). Univariate analysis also showed that statins did not have an effect when the variable measured was survival after metastasis (Table 1). Upon t-test analysis, OS was found to be prolonged in ER-positive and HER2-negative tumours. Patients on statin use had longer OS both for ductal (IDC) and lobular (ILC) subtypes although the difference in OS between statin and non-statin users was only significant for ILC (Supplementary Table 2). When looking at type of adjuvant chemotherapy, OS difference was remarkably higher if patients were treated with anthracyclines. Statin users had prolonged OS when stratifying by all TNM stages but for stage 3. Women taking statins had longer overall survival both in the premenopausal (<50 years of age, p = 0.002) and postmenopausal groups (>50 years of age, p = 0.025) (Supplementary Table 2).

With the covariates that were statistically significant in univariate analysis, we performed multivariate analysis, and found that statin use, adjuvant chemotherapy, and ER-positive predicted better OS and time to relapse (Table 1). Thereupon, these findings propose the new consideration for statin usage as a secondary preventive to delay mBC and its safe maintenance while undergoing chemotherapies.

Discussion

Seeding of disseminated breast cancer cells as micrometastases frequently leads to recurrent active metastatic disease. Unfortunately, during dormancy, standard chemotherapies appear to be ineffective [50]. Among others reasons, this may be in part due to tumour cells not being in cycle [51], and upregulation of E-cadherin with downstream activation of survival pathways [16, 20]. Previous studies have remarkably pointed out the antitumor effects of statins, both alone or in combination with chemotherapy [13, 52]. However, these have been often limited to in high dose testing in vitro or preclinical models in which statins alone are cytocidal [53]. In real world usage for metabolic conditions, statins do not approach such levels. Thus, there is a gap in our experimentation about how lower levels of statins would affect metastatic cancers. Although we previously found that statins could prevent metastatic emergence, more knowledge is needed to ascertain if statins facilitate or blunt the effects of chemotherapy on metastasis. In this study, we integrated in vivo, in vitro and clinical retrospective data to further clarify this rationale.

Controlling the outgrowth of metastases in the liver is a therapeutic priority, as it constitutes the site of visceral metastases with the poorest prognosis and the first distant organ for relapse in 20% of TNBC cases [54]. Due to the limitations in spontaneous metastasis models, examining other distant sites, while important for human translation of these findings, is beyond the scope of this study. We used a spleen-to-liver metastasis model since parental MDA-MB-231 cells metastasise to the liver at a predictable rate [19], and its demonstrated applicability for therapeutic testing [17, 20–23] and prior usage as metastasis to the liver models [20, 55–57].

This study provides a mechanistic rationale for the use of statins as a secondary preventative after removal of the primary tumour. First, we validated prior data showing that atorvastatin prolonged dormancy and reduced visceral metastatic outgrowth [17]. We then considered that chemotherapy may not be effective due to the slow-growing pattern of statin-treated metastases. Atorvastatin did not modify the cytocidal effects of doxorubicin or paclitaxel in reducing growth of the primary tumour. In fact, in vivo atorvastatin exposure reduced the fraction of cycling cells as compared to mono-chemotherapy, and increased the proportion of apoptotic cells in the atorvastatin + doxorubicin group.

Statins limit cell growth through altering the signalling of Akt and translocation of Ras [15]. In addition, statins decreased the phosphorylation of the residue S70 of Bcl-2, which could suggest that inhibition of anti-apoptotic signalling is a possible sensitisation mechanism. Primary hepatocyte coculture with MDA-MB-231 cells proved that the liver environment allows for atorvastatin-induced sensitisation.

We also found that in vivo atorvastatin use correlated with increased metastatic E-cadherin expression. Two different possibilities are brought by HMG-CoA-reductase-derived inhibition and subsequent upregulation of E-cadherin. On the one hand, it has been proposed that E-cadherin upregulation could be induced by mevalonate inhibition [58]. Increased E-cadherin(+) may prompt cell resistance to statins [15] and therefore continued proliferation. On the other hand, it is likely that tumour cells that lacked E-cadherin were preferentially suppressed, and therefore E-cadherin(+) cells increased in relation to the metastatic population. Acquired E-cadherin expression confers chemotherapy resistance [20]. We found that Akt was not inhibited by atorvastatin; doxorubicin also lacked efficacy in epithelial cells, as noted by relative lower expression of cCasp9 and cPARP. However, the combination of these drugs seemed to promote apoptosis independently of Akt. Absence of canonical Akt inhibition on E-cadherin(+) cells could explain the lack of statin effects in cell division seen in E-cadherin(+) metastatic cells. It is also suggested that the effect of statins on MDA-MB-231 relied in Erk, rather than an Akt-mediated inhibition [59]. However, our data on wild-type MDA-MB-231 indicates that the antimetastatic effects of statins primarily affect Akt. Because Akt was not inhibited on statin-treated MDA-MB-231-Ecad(+) cells, E-cadherin re-expression may dictate an Akt-independent mechanism through which metastatic cells may respond to therapies. E-cadherin is also a survival factor and driver of progression [60]. Here, we propose that E-cadherin also constitutes a marker of chemotherapy resistance during metastatic relapse, which could be overcome by concomitant use of atorvastatin.

We then evaluated lipophilic statin use and mBC progression using clinical practice data from patients who were treated at UPMC from 1999 to 2019. We were able to provide a novel insight on the role of statins in cancer treatment. First, we show that statin use prolonged OS of patients with mBC, validating clinical outcomes from previous studies [48, 61]. Importantly, this study points out the role of statins in maintaining metastatic cells dormant and delaying relapse. In fact, our results suggest that while statins may be effective in prolonging dormancy, they could lack promise once metastatic progression develops. This study also suggests that although all molecular subtypes may experience longer latency periods from lipophilic statin use, patients with ER-positive and/or HER2-negative mBC would benefit the most in terms of OS and time to relapse.

Given that our experimental data focuses on mTNBC, we also investigated the outcomes of this subpopulation. This subtype tends to affect younger women, who usually suffer less comorbidities, making statin prescription less frequent. Nevertheless, we saw a trend that benefited statin users when looking at time until relapse. We also sought to investigate if statins could prompt improved chemotherapy outcomes. Additional 1.6 months in median progression-free survival was observed among statin-treated patients who had ER-negative mBC, but overall statins did not substantially alter the effects of chemotherapy in progression.

Our clinical analyses have limitations. First, the clinical correlation took place at a single institution. In addition, given the retrospective nature of the study, patients had different lengths from the onset of statin use until outcome. Furthermore, differences in OS are initially assumed to be due to breast cancer and no other causes of death. It is also possible that the effects seen might be due to confounding factors, such as high cholesterol levels. Use of drugs such as aspirin, which are used among populations with comorbidities requiring statin prescription, could constitute a confounding. We did not control for this covariate due to the widespread, over-the-counter use of the drug, which compromises the feasibility of a retrospective analysis. In addition, patients who were not on statins were also younger on average, and likely to have experienced more aggressive disease [62]. Data on length of prescription was not included. Nevertheless, these results need to be contrasted with those from ongoing clinical trials [63, 64].

In summary, our study supports that those persons on statins can and should be maintained on statins during chemotherapy, as the use of lipophilic statins will not blunt and may enhance the chemotherapeutic effect. The findings also suggest that one may consider adapting statin usage after primary tumour resection and maintaining statin utilisation for those women who undergo active therapies for mBC so as to prevent emergence of dormant micrometastases. This hypothesis encourages the development of prospective studies to evaluate if this approach could prevent recurrences and improve survival, herein adding a new use of an old drug to our armament against mBC.

Author contributions

JLGM, CHB and AW conceived the study, designed, and performed in vivo experiments. JLGM and AW wrote the manuscript. JLGM performed in vitro experiments and experimental and clinical analysis. AMC helped with hepatocyte culture and manuscript edition. All authors read, edited, and approved the final manuscript.

Data availability

Data supporting the conclusions of this manuscript are included within the article and the Supplementary Figures.

Ethics approval and consent to participate

Animal studies were approved and done in compliance with the Pittsburgh VA Institutional Animal Care and Use Committee (IACUC) and Institutional Biosafety Committee (IBC) under protocol name “Molecular Regulation of Breast Cancer Progression” (ID:03017). Cell lines used are obtained from the American Type Culture Collection. Clinical retrospective studies were IRB-approved (STUDY20030072) by the University of Pittsburgh Human Research Protection Office (HRPO).

Consent to publish

Not applicable.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-021-01529-0.