Progestins counteract estrogens to shift breast cancer cells into a quiescent metabolic profile CURRENT STATUS: REVIEW

Progestins generally suppress the growth of estrogen receptor (ER) and progesterone receptor (PR) positive breast cancers and are occasionally used as treatments. However, progestins also increase populations of therapy-resistant cancer stem cells. We speculated that the downstream effects of progestins on cell metabolism might help explain its unusual impact on cell phenotype. Thus, in this study we investigated how progestins, in the absence or presence of estrogens, affect cell metabolism in ER+PR+ breast cancer. Metabolites were quantified and compared from ER+PR+ breast cancer cell lines T47D and UCD65 treated with vehicle, estrogen only, progestin only, or the combination using ultra-performance liquid chromatography coupled with mass spectrometry (UPLC-MS). Metabolic flux analysis was performed on cells given the same hormone treatments. Likewise, the influence of treatments on mitochondrial morphology was measured using transmission electron microscopy in cell lines and patient-derived xenograft tumors and mitochondrial-targeted GFP in cell lines. Mitochondrial biogenesis was measured via fluorescence shift in the photoconvertible MitoTimer reporter coupled with confocal microscopy, and biogenesis regulators were measured by qPCR. Select metabolites and ATP were measured using fluorometric assays.

3 resulted in a more aged and less elongated mitochondrial population. Notably, progestins blocked the estrogen-induced expression of mitochondrial biogenesis regulators PGC1α and PGC1β and their downstream targets. While progestin treatment reduced total intracellular amino acid and glutathione pools, a subpopulation of progestin-inducible cancer stem cells revealed a dependence on glutathione.

Conclusions
These findings indicate that progestins antagonize estrogen's effects on cellular metabolism and shift cells to a more quiescent phenotype, with reduced mitochondrial functional capacity, more reliance on glycolysis, and increased cell survival traits. Our results have implications for current clinical studies testing selective PR modulators in ER+ breast cancers.

Background
In the normal breast, the female hormones estrogen and progesterone act in coordination to facilitate growth and development. In breast cancer, however, hormone signaling becomes aberrant and uncoupled from its normal tissue function. Nearly three-quarters of breast cancers are estrogen receptor alpha (ER) positive and depend on estrogens for growth (1). As most diagnoses occur in postmenopausal women with low circulating estrogen levels, tumor growth in these women relies on local adipose tissue production of estrogen (2,3). Progesterone dissipates more completely at menopause (4), and its reintroduction in hormone therapy (HT) with synthetic analogs (progestins) in combination with estrogens increases breast cancer incidence (5). However, as a treatment in advanced breast cancers, progestins have response rates similar to endocrine therapies such as tamoxifen and are sometimes used upon endocrine failure (6)(7)(8). The dichotomy of action of progestins in the biology of breast cancer cells remains an impediment to their guided use in treatments. 4 Receptors for estrogens and progestins, ER and PR, are closely related members of the nuclear receptor transcription factor family. Crosstalk between family members is emerging as prototypical in controlling gene regulation and cell phenotype (reviewed in (9)). Proteomic and genomic studies in ER + PR + breast cancer cell lines and patientderived xenografts (PDX) have found that unliganded PR is a cofactor in the ER transcription complex, and that progestins redirect ER-PR chromatin binding away from core estrogen responsive genes (10)(11)(12). Thus, transcriptional interference of ER driven genes is one mechanism by which PR suppresses the mitogenic activities of ER (10)(11)(12).
However, the downstream consequences of ER-PR crosstalk on breast cancer cell physiology, particularly cellular bioenergetics and metabolism, are relatively unexplored.
Understanding these changes is important since cancer cells adopt distinct metabolic phenotypes that modulate their longevity and survival (reviewed in (13)).
Estrogens stimulate breast cancer cell metabolism to satisfy the increased energy and biomass needs of growing cells. Energy-generating capacity for processes of oxidative phosphorylation (oxphos) and glycolysis is generally enhanced in the presence of estrogen through genomic and non-genomic mechanisms (reviewed in (14,15)). Estrogen upregulates glucose transporters GLUT-1, -3, and − 4, and increases glucose uptake and consumption along with lactate production in breast cancer cells and xenografts (16,17).
Glutamine uptake and consumption also increases with estrogen treatment of ER + MCF7 breast cancer cells (18). Tricarboxylic acid (TCA) cycle enzymes including citrate synthase (12), aconitase (19), and isocitrate dehydrogenase are all increased by estrogen (20,21). Likewise, estrogen treatment increases expression of mitochondrial-encoded electron transport chain subunits, including increased NADH dehydrogenase subunit 1 (ND1) and cytochrome oxidases I and II (COI, COII) (22)(23)(24). Both ERα and ERβ can localize to the mitochondrial matrix in the estrogen-responsive cell line MCF7 (25,26). Collectively, ER 5 targets bioenergetics processes with a notable impact on mitochondrial adaptation to meet the demands of continuous breast cancer cell growth.
In contrast to E2, progestin effects on breast cancer cell metabolism are less studied. PR regulates genes involved in cholesterol and steroid, fatty acid and lipid, and nucleotide and amino acid metabolism (27). In breast cancer cells, progestins increase de novo lipid synthesis and accumulation of lipid droplets (28). A truncated isoform of PR (mPR) was also reported to localize to the mitochondria of both ER + and ER − breast cell types, with proposed effects on mitochondrial membrane potential and functional capacity; however, the biological importance of mPR in cells with functional nuclear PR is unclear (29,30).
These few studies were done in the absence of estrogens and/or in ER − breast epithelial cells. The impact of progestins on estrogen-driven metabolism in ER + PR + breast cancer cells remains largely untested, while highly relevant since it represents the typical physiological and clinical context.
In this study, we evaluated the effect of estrogens, progestins, and the combination on modulation of cell metabolism machinery in ER + PR + breast cancer cells. We found that progestins perturb estrogen-driven mitochondrial metabolism by suppressing mitochondrial oxidative respiratory capacity and preventing estrogen-induced mitochondrial elongation. This likely occurs by blocking mitochondrial biogenesis regulators. Overall, we conclude that progestins exert a dominant effect on estrogendriven energy production in breast cancer cells, prospectively as part of a broader strategic shift from rapid tumor growth to cytostasis and metabolic dormancy. The consequences of a progestin-induced metabolic shift are unclear, but could have implications for the safety and efficacy of progestin-based breast cancer therapies. 6 Cell lines and cell culture The breast cancer cell line T47D (ER + PR+) was obtained from the University of Colorado Cancer Center Tissue Culture core and was maintained in minimal Eagle's medium, 5% fetal bovine serum (FBS), 1X non-essential amino acids, 1 × 10 − 9 M insulin, 0.1 mg/mL sodium pyruvate, and 2 mM L-glutamine. Development of ER + PR + breast cancer PDX UCD4 and UCD65 has been previously described (12,31). The UCD65 cell line was derived from the UCD65 PDX and is ER + PR + and Her2 unamplified. The UCD65 cell line was maintained in DMEM/F-12 1:1 with 10% FBS, 1 × 10 − 9 cholera toxin, 1 × 10 − 9 hydrocortisone, and 1 × 10 − 9 M insulin. Cell lines were authenticated using short tandem repeat (STR) analysis using the University of Colorado Genetics core. For UCD65, cells were matched to the original PDX and not to any other cell lines in the database. All cell lines were routinely tested for mycoplasma contamination using the MycoAlert mycoplasma detection kit (Lonza, Basel, Switzerland).

Methods
In vitro hormone experiments were performed using phenol red-free media with the same additives described above. Hormone treatment was used as follows: E2 = 17-β-estradiol,
For T47D xenografts, 1 × 10 6 cells were injected into the mammary fat pad of female NSG mice supplemented with E2 or E2 + P pellets. Tumors were measured weekly with a digital 7 caliper and volume was estimated by the formula lw 2 /2. All animal experiments were performed under a protocol approved by the University of Colorado Institutional Animal Care and Use Committee.

Metabolomics
Metabolites were extracted from T47D and UCD65 cell pellets in quadruplicate using icecold extraction buffer (5:3:2 methanol:acetonitrile:water) at a concentration of 2 × 10 6 cells per mL as previously described (34,35). Supernatants (10 uL) were analyzed on a Thermo Vanquish ultra-high performance liquid chromatography column coupled online to a Thermo Q Exactive mass spectrometer (UHPLC-MS) in positive and negative ion modes (separate runs) using a 5 min C18 gradient. Untargeted data acquisition, quality control, and targeted data analysis were performed as previously described (36). Precipitated protein was reconstituted in PBS and measured using BCA protein assay (Pierce, Thermo Fisher, Waltham, MA). Metabolomics intensity signals were normalized to sample protein concentration.
Normalized data was imported into MetaboAnalyst software (37,38), where data was logtranformed and autoscaled (39). Partial least squares discriminant analysis (PLS-DA) was performed on all samples within cell lines for visual inspection of clustering patterns and outlier detection. Heatmaps were constructed using Pearson distance with average linkage and depict non-scaled PLS-DA variable importance in projection (VIP) averaged across replicates (N = 4) within treatment groups.
Metabolites from random forest variable importance analysis with mean decrease in accuracy > 0 were evaluated for fold-change direction (lower in E2 + P vs. E2, called 8 "down"; higher in E2 + P vs. E2, called "up"). These subsets of metabolites were submitted to MetaboAnalyst Pathways Analysis (MetPA) and can be found in Additional file 1: Table   S1 for T47D cells and Additional file 2: Table S2 for UCD65 cells. Pathways were identified using default settings; specifically, the hypergeometric test for overrepresentation analysis and relative betweenness centrality was used for pathway topology analysis, with pathways mapped to the Homo sapiens KEGG reference library.

Immunocytochemistry
Immunohistochemistry and ICC were performed as previously described (31,40). Cells

Inhibition Of Glutathione Synthesis
The small molecule buthionine sulfoximine (BSO) inhibits the catalytic subunit of the glutamate-cysteine ligase enzyme (GCLC), which catalyzes the rate-limiting step in glutathione synthesis (41). BSO was purchased from Cayman Chemical (#14484, Ann Arbor, MI) and was reconstituted fresh in sterile PBS immediately prior to use at a final concentration of 10 − 4 M.

Transmission Electron Microscopy
Cells were cultured on PermaNox 60-cm dishes (VWR, Radnor, PA). Excised tumors were cut into approximately 1 mm 3 pieces. Cultured cells and tumor pieces were fixed with 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer and then post-fixed with reduced osmium (1.5% potassium ferrocyanide + 1% osmium tetroxide) followed by 2% osmium tetroxide. Samples were dehydrated with a graded series of ethanol and embedded in a thin layer of Epon. Following Epon curing, small pieces were cut out and reembedded in blocks that were sectioned at 65 nm on an ultramicrotome, collected on formvar coated slot grids, and post-stained with 2% osmium tetroxide and lead citrate.
At least 10 fields per treatment were imaged and blinded prior to analysis. Mitochondrial length along the longest axis was measured using Fiji and plotted via histogram, with bin mode indicated on the X-axis. Outliers greater than 3 standard deviations outside the mean of the full dataset were excluded. Differences in distributions were analyzed using the Kolmogorov-Smirnov test for frequency distributions; comparisons for cells were 1) vehicle vs. E2 and 2) E2 vs. E2 + P, while tumor samples were pooled by hormone treatment and compared E2 pellets vs. E2 + P pellets.

Confocal Microscopy
T47D cells were labeled overnight with CellLights BacMam 2.0 MitoGFP (Thermo Fisher, Waltham, MA) according the manufacturer's protocol. Following treatment, cells were fixed in 4% paraformaldehyde, counterstained with DAPI, and mounted on coverslips.
Images were collected using confocal laser scanning microscopy (Zeiss LSM 780) with 40X objective. Mitotimer T47D and UCD65 cells were transduced to express the MitoTimer construct from Addgene (Watertown, MA) as previously described (42). Following stable transduction with the pLenti-CMV-rtTA3 Blast (w756-1) (Plasmid #26429) with > 2 weeks blasticidin selection, cells were transiently transfected with pTRE-Tight-MitoTimer (Plasmid #50547) using Lipofectamine 3000 (Thermo Fisher, Waltham, MA). MitoTimer expression was induced by treatment for 1 hour with doxycycline (4 ug/ml, Cayman Chemical, Ann Arbor, MI), followed by washout and hormone treatment for 48 h, at which point most mitochondria should be yellow to red. A second 1 h dox pulse, followed by washout, was used to label a new wave of mitochondria corresponding to cells undergoing active biogenesis. Following an additional 6 h of hormone treatment, live cells (via IncuCyte ZOOM at 20X magnification or fixed cells (4% paraformaldehyde with DAPI counterstain) were imaged for analysis.

Real-time Quantitative Pcr (qpcr)
RNA was harvested using QIAzol lysis reagent (Qiagen, Venlo, the Netherlands) and converted to cDNA using the Verso cDNA kit (Thermo Fisher, Waltham, MA). qPCR was performed on cDNA using ABsolute Blue Sybr Green (Thermo Fisher, Waltham, MA) and normalized to β-actin using the Pfaffl method (43). qPCR primers are provided in Additional file: Table S3. Results are representative of 3-4 experiments.

Statistical Analyses
Statistics were performed using GraphPad Prism 8.3.0 (GraphPad Software, San Diego, CA), with the exception of metabolomics data, which was analyzed using MetaboAnalyst (see Metabolomics section for details). Two-tailed Student's t-tests, or one-way ANOVA followed by Tukey post hoc multiple comparison tests were used as indicated with significance set a priori at P < 0.05. Equivalent non-parametric tests were used in the event of unequal variance between groups.

Results
Progestins counter-regulate estrogen-driven metabolic reprogramming 12 To globally evaluate the combined effects of estrogen plus progestin on the bioenergetics of breast cancer cells, we performed metabolomics on two ER + PR + breast cancer cell lines (T47D and PDX-derived cell line UCD65) treated with vehicle (Veh), estrogen (17βestradiol, E2), the synthetic progestin R5020 (P), or combination (E2 + P) for 24 h. The four treatment groups were readily distinguished based on their metabolite signatures, with E2 treatment causing a notable metabolic shift in both cell lines ( Fig. 1A and 1B).
Heatmaps of the top 50 features by variable importance (VIP) scores demonstrated widescale reductions in metabolites in cells treated with P, either alone or in combination with E2 ( Fig. 1C and Fig. 1D).
We used MetaboAnalyst Pathways Analysis (MetPA) to identify the most significant pathways altered in T47D and UCD65 cells, shown in Fig. 1E. We focused our analysis on E2 vs. E2 + P which captures the most common physiological scenario. In both cell lines, many metabolic pathways were down-regulated in E2 + P compared to E2, including the TCA cycle, amino acid metabolism, and the glutathione antioxidant defense pathways.
Only the pentose phosphate pathway was stimulated by E2 + P vs. E2. These results suggest P may impact the overall energetic phenotype of estrogen-driven cells in a manner consistent with the preferential use of carbohydrates for glycolytic energy production.

Progestins Shift Cells To A More Quiescent Bioenergetic Phenotype
Since TCA cycle was a key affected pathway in each cell line, we speculated that a reduction in TCA cycle metabolites might indicate a shift in the energetic profile. Seahorse flux analysis revealed that hormone treatment of T47D and UCD65 cells altered the basal usage ratio of oxphos to glycolysis, called 'energetic phenotype' (Fig. 2A). When mitochondrial stress was induced by inhibiting ATP synthase (oligomycin) and uncoupling membrane potential (FCCP), the combination of E2 + P significantly decreased the 13 maximal oxygen consumption rate (OCR) compared to E2 alone with no compensatory change in glycolysis. Thus, in both cell lines, E2 + P effectively shifted cells from an "Energetic" phenotype, with high oxphos/high glycolysis, toward a "Quiescent" phenotype, with lower oxphos/lower glycolysis. In T47D but not UCD65 cells, E2 + P treatment reduced maximal respiratory capacity compared to P alone and compared to vehicle, suggesting that combined hormone treatment may exert a greater effect on respiratory capacity than either hormone alone (Fig. 2B).
In agreement with observed changes to energetic phenotype, we observed a timedependent increase in ATP with E2 treatment, increasing to approximately 140% or 125% of Vehicle-treated ATP by 72 h in T47D and UCD65 cells, respectively (Fig. 2C). P alone did not impact ATP in either cell line. The E2 dependent ATP increase in intracellular ATP was significantly blocked by combination E2 + P in T47D and trended downward in UCD65 cells (P = 0.056) (Fig. 2C) These data suggest that P co-treatment blocks the E2-driven increase in energy production.

Progestin-treated Cells Have A More Aged Mitochondrial Network
To visualize the entirety of the mitochondrial network within cells, mitochondria labeled with a baculovirus GFP construct demonstrated that cells treated with E2 had enhanced phenotypic mitochondrial signal compared to cells treated with E2 + P (Fig. 4A). We speculated that P may activate mitochondrial fission to block effects of E2; however, we did not find evidence of activated fission with P treatment, determined by qPCR for fission proteins and Western blot for total dynamin-related protein (Drp-1) and its phospho-Ser 616 isoform, which represents activated Drp-1 (Additional file: Fig. S1).
Therefore, we evaluated mitochondrial turnover to interrogate processes of mitochondrial birth (biogenesis) and death using the inducible MitoTimer system developed by Hernandez et al (42), in which newly made mitochondria fluoresce green but photoconvert to red fluorescence over approximately 48 h. Confocal microscopy revealed that the majority of mitochondria in E2-treated cells were green and yellow, indicative of continual mitochondrial biogenesis, whereas mitochondria from P or E2 + P-treated cells were relatively more orange or red, indicative of older mitochondria (Fig. 4A). Quantitation of MitoTimer demonstrated an increased ratio of green:red fluorescence (more cells with predominantly green "new" mitochondria than red "old") with E2 treatment (Fig. 4B).

Progestins Block The E2-induced Pgc1α Mitochondrial Biogenic Signaling
Cascade Mitochondrial biogenesis is regulated by the peroxisome proliferator activated receptor gamma coactivator (PGC) family of proteins, particularly PGC1α. Since P treatment resulted in a more aged mitochondrial network, we evaluated the mitochondrial biogenesis signaling cascade depicted in Fig. 5A for response to hormone treatment. E2 increased transcript expression of Nrf2 (NFE2L2) (Fig. 5B) and its downstream targets PGC1α (PPARGC1A) (Fig. 5C) and PGC1β (PPARGC1B) (Fig. 5D). Conversely, the time-dependent induction of PGC1α by E2 was entirely mitigated by co-treatment with P; this was observed to a lesser extent in PGC1β. Downstream effectors of PGC1α include the Nrf1 (NFE2L1) transcription factor and mitochondrial transcription factor A (TFAM) which activate mitochondrial biogenesis and trigger replication of the mitochondrial genome, respectively. E2 induction of these targets was abolished by E2 + P treatment ( Fig. 5E and Fig. 5F, respectively). This reduction in transcript levels of key mitochondrial biogenesis genes in response to progestin co-treatment, with the exception of PGC1β, was largely recapitulated in UCD65 cells (Additional file: Fig. S2). Detection of PGC1α protein level is complex, due to the presence of more than 10 isoforms with variable stability, biological activity, and tissue expression (45). However, immunoblot against Nrf2 as the inducer of the downstream signaling cascade suggested a decrease in Nrf2 protein expression with P alone or E2 + P (Additional file: Fig. S3).
Figures 1-5 demonstrated that P co-treatment slows cell metabolism compared to E2 alone: by lowering oxphos, progestins shift cells to a glycolytic energy profile. While this shift could be favorable for a quiescent cytostatic phenotype, it could also potentially prolong cancer cell survival. To evaluate other potential metabolic contributors to cytostasis, we returned to our initial metabolomics findings to interrogate amino acid regulation and the glutathione pathway, both shown to be down-regulated in E2 + P vs. E2 alone.

Glutathione Dependence
Beyond their role as metabolic substrates for the TCA cycle and protein biosynthetic precursors, amino acids play a key role in other cellular signaling cascades, including redox homeostasis. We used biochemical assays to validate our metabolomics findings of progestin effects on intracellular amino acids and glutathione. Co-treatment with progestins elicited a striking reduction in the L-amino acid pool compared to E2 alone ( Fig. 6A). Likewise, we validated a reduction in the total glutathione pool in response to progestin treatment (Fig. 6B).
As many types of chemotherapy exert cytotoxicity via induction of oxidative stress, we found the limited glutathione defense capacity of progestin-treated cells intriguing from a potential therapeutic perspective. We and others have shown that both natural and synthetic progestins induce a subpopulation of cells to acquire cancer stem cell (CSC) properties including relative quiescence, tumorsphere formation, increased tumor initiating capacity, and resistance to endocrine and chemotherapies (46)(47)(48)(49)(50). Progestininducible CSCs lose expression of ER and PR while activating expression of the basal cytokeratin, cytokeratin 5 (CK5) (46). As glutathione is a vulnerability of many types of CSCs (eg, (51-53)), we used the small molecule buthionine sulfoximine (BSO) to block the rate-limiting step in glutathione synthesis. As shown in Fig. 6C and quantified in Fig. 6D, P treatment induces approximately 15% of T47D cells to express CK5. Upon co-treatment with P + BSO, the CK5 subpopulation is largely absent, suggesting that targeting the glutathione dependency of P-inducible CSCs may be a viable strategy to mitigate one negative consequence of P treatment.
Together these data suggest that P, alone or in combination with E2, antagonizes mitochondrial respiration to shift the energetic potential of ER + breast cancer cells towards a low-energy more quiescent phenotype with reduced mitochondrial activity.
Since these cells are cytostatic yet viable, the consequence and permanence of the metabolic dormancy remain key questions.

Discussion
Disruption of normal metabolic pathways is a hallmark of cancer. In the majority of breast cancers, aberrant estrogen signaling drives metabolic changes. This is the first study to our knowledge that demonstrates progestins antagonize the metabolic effects of estrogen, including inhibition of ATP production, disruption of mitochondrial function and biogenesis, and depression of amino acid levels for reduced anabolic capacity. Concurrently, progestins alter the cell's antioxidant defense system through the glutathione pathway, another important hallmark of cancer and cell survival. These findings have implications for pre-and postmenopausal women taking progestins for HT and breast cancer treatments.
It is well established that synthetic progestins in combination with estrogen are tumorigenic in women without a history of breast cancer. A 2019 meta-analysis by the Collaborative Group on Hormonal Factors in Breast Cancer confirmed increased breast cancer risk with nearly every type of synthetic progestin containing HT compared to estrogen-only HT (5). Some theories suggest progestins activate pre-existing occult cancer stem cells in these women (54, 55). In fact, it is well known that in established breast cancer models, progestins increase populations of cancer stem cells, while slowing overall cell/tumor growth (eg, (12,46,49,50)). Interestingly, both quiescence and stem cell maintenance require low cellular energy production. In normal tissues, embryonic and induced pluripotent stem cells generally fulfill energy needs through glycolysis while differentiation reduces glycolytic rate with increased oxphos (56). Energetic processes in CSCs, where defined, depend on tumor type and show considerable metabolic plasticity (reviewed in (57)). Given the unique hormone dependency of most breast cancers, CSC metabolism is likely to be influenced by hormonal milieu.
In our studies, progestins had a profound impact on mitochondrial morphology and function. First, progestins changed the energetic phenotype of cancer cells by suppressing oxphos while maintaining glycolysis (Fig. 2), suggesting that progestin treatment reduces mitochondrial activity. Second, progestins notably altered mitochondrial morphology (in the absence or presence of estrogens) transitioning to less elongated mitochondria (Fig. 3,   Fig. 4). Mitochondrial morphology is tightly linked to energy metabolism: a highly interconnected mitochondrial network and enlarged cristae are associated with enhanced respiration, whereas low oxphos and high glycolysis correlates with smaller mitochondria displaying reduced intracristae space (58). Furthermore, disrupting or modulating mitochondrial dynamics can have profound impacts on stem cell behaviors (59). For example, deletion of PGC-1α and PGC-1β in pre-adipocytes prevents their differentiation into mature adipocytes, whereas stimulation of mitochondrial biogenesis promotes differentiation of embryonic stem cells and induced pluripotent stem cells (59). Curiously, progestins potently decreased PGC1α and PGC1β in breast cancer cells (Fig. 5, Additional file: Fig. S2). We speculate this in addition to regulation by other genes contributes to the observed switch in mitochondrial phenotype and function. Whether this progestin-induced metabolic response is transient or persistent will be important to know, since a recent meta-analysis found that tumors expressing high levels of glycolytic proteins were more aggressive and corresponded to shorter overall survival of breast cancer patients, as well as other types of cancer (60).
The metabolic phenotype of cancer cells that have been forced into cytostasis is virtually untested. Studies by Havas et al (61), using primary murine mammary organoid cultures with inducible oncogene expression noted metabolic shifts in residual cells following oncogene withdrawal that were important for recurrence. Noticeably, many of the pathways identified by Havas as distinct between induced vs. regressed organoids overlapped with metabolic pathways we identified as counter-regulated by E2 + P vs. E2 including reduced aminoacyl-tRNA biosynthesis, TCA cycle, glyoxylate/ dicarboxylate metabolism, and Ala/Asp/Glu metabolism (61). Progestin and endocrine therapies induce a cytostasis that is thought to keep occult tumor cells dormant for long duration or even permanently. Neither cytostasis, nor the factors that trigger exit from dormancy, are well understood and it is conceivable that manipulation of the described progestin induced metabolic profile could act as a treatment target. Interestingly, Havas found that PR antagonists reduced recurrence of their organoid models upon re-induction of oncogenes (61). This has relevance to ongoing studies of PR agonists and antagonists for treatment of ER + breast cancer, and which types of ligands are able to reduce growth and prevent recurrence.
In conclusion, our studies support that progestins induce a metabolic transition in ER + breast cancer cells away from oxphos towards a glycolytic and stress-avoiding phenotype.
We speculate that this shift is important for the dual actions of progestins in subduing estrogen-driven growth and promoting expansion of CSCs. Furthermore, our studies 20 emphasize that the type and duration of progestin use is similarly important for breast cancer treatment as it is for HT. Further studies into hormone-regulated metabolic phenotypes may help design strategies to prolong benefits of relatively non-toxic PR targeted therapies.

Conclusions
Our understanding of the metabolic phenotype of hormone-dependent breast cancer cells is limited, particularly in relation to estrogens, the underlying driver, and progestins, often used in medicines and treatments. Here we report that progestins impair estrogeninduced energy production by disrupting the function of existing mitochondria and blocking creation of new mitochondria. Progestins cause breast cancer cells to undergo a metabolic switch from a high-energy mitochondria-powered phenotype to a low-energy glycolysis-powered phenotype. Additionally, we find that a key intracellular stress manager, the glutathione pathway, is altered by progestins and may help sustain a subpopulation of treatment-resistant cancer cells. Our work has implications for progestins and antiprogestins that are being tested in clinical trials, suggesting that immediate benefits may be counteracted by long term adeptness of cytostatic cells. Furthermore, some identified features could be potential vulnerabilities of progestin treated cells.

Consent for publication
Not applicable.     replicates normalized to DNA content following the assay. *P < 0.05, **P <0.01, ***P <0.001 for the indicated comparisons using one-way ANOVA.