Metabolic diversity within breast cancer brain-tropic cells determines metastatic fitness

SUMMARY

HER2+ breast cancer patients are presented with either synchronous (S-BM), latent (Lat) or metachronous (M-BM) brain metastases. However, the basis for disparate metastatic fitness among disseminated tumor cells of similar oncotype within a distal organ remains unknown. Here, employing brain metastatic models, we show metabolic diversity and plasticity within brain-tropic cells determines metastatic fitness. Lactate secreted by aggressive metastatic cells or lactate supplementation to mice bearing Lat cells limits innate immunosurveillance and triggers overt metastasis. Attenuating lactate metabolism in S-BM impedes metastasis, while M-BM adapt and survive as residual disease. In contrast to S-BM, Lat and M-BM survive in equilibrium with innate immunosurveillance, oxidize glutamine and maintain cellular redox homeostasis through the anionic amino acid transporter xCT. Moreover, xCT expression is significantly higher in matched metachronous brain metastatic samples compared to primary tumors from HER2+ breast cancer patients. Inhibiting xCT function attenuates residual disease and recurrence in these preclinical models.

Graphical Abstract

eTOC Blurb:

The basis for disparate metastatic fitness among disseminated tumor cells of similar oncotype within a distal organ is unknown and vital to understand for better clinical management. Here, Parida et al. develop models of synchronous, latent residual and metachronous brain metastatic disease and reveal distinct metabolic states associated with HER2+ breast cancer brain-tropic cells. Metabolic diversity and plasticity dictate cellular ability to survive and initiate metastasis.

INTRODUCTION

Brain metastases develop following the spread of cells from the primary tumor to the brain through vasculature (Achrol et al., 2019). The majority of disseminated tumor cells in the brain parenchyma perish. The surviving few may initiate synchronous metastases that are detected along with primary tumor or adapt and stay latent for months to years before triggering a metachronous metastatic outbreak (Kienast et al., 2010; Kim et al., 2019; Massague and Obenauf, 2016). The survival dependencies of cancer cells with a similar genomic profile that have differentially adapted to the brain parenchyma are unknown. Understanding these differences is vital to devise effective strategies that identify and treat patients presented with synchronous or delayed metachronous metastases.

Brain metastatic incidence in breast cancer patients varies with disease subtype. Patients with human epidermal growth factor receptor 2 (ERBB2; HER2) amplification or with triple negative hormone receptor status (TN) are at a higher risk of developing brain metastases than those with ER positive and/or PR positive breast cancer. Moreover, metastatic relapses in brain are commonly observed in patients after primary adjuvant therapy (Olson et al., 2013). Patients with HER2+ breast cancer and synchronous brain metastases, although relatively rare, have a median overall survival of around 6 months (Ho et al., 2015). Metachronous brain metastases are observed in fifty percent of HER2+ breast cancer patients considered disease free after a variable length of time post primary diagnosis and treatment (Kabraji et al., 2018; Kodack et al., 2015; Kodack et al., 2017; Lin, 2015; Lin et al., 2017; Olson and Mullins, 2013; Olson et al., 2013). Systemic anti-HER2 therapies are highly effective for extracranial metastasis but ineffective on brain metastases, despite adequate delivery and activity in the brain parenchyma (Kabraji et al., 2018). Although small molecule brain permeable tyrosine kinase inhibitors are approved for treating HER2+ breast cancer patients with intracranial metastasis, overall survival benefit to patient is short lived (Maher et al., 2009; Morikawa et al., 2015; Saleem et al., 2015).

Reversible epigenetic and metabolic adaptations are likely to be responsible for the observed variability in metastatic fitness of tumor cells disseminated to distal organs (Bergers and Fendt, 2021; Faubert et al., 2020; Li and Simon, 2020). Few elegant studies have identified the role of the brain microenvironment and nutrient availability in shaping disseminated tumor cell metabolism and growth (Ciminera et al., 2017). Neuregulins expressed in the brain microenvironment promote survival of tumor cells in the brain (Kodack et al., 2017). The ability to utilize acetate in addition to glucose as a carbon source is reported to provide greater metabolic flexibility to breast, lung and skin cancer cells, enabling their survival in the brain microenvironment (Mashimo et al., 2014). Brain-tropic melanoma and triple negative (TN) breast cancer cells can take advantage of polyunsaturated fatty acids released from astrocytes to proliferate (Jin et al., 2020; Zou et al., 2019). Increased dependence on oxidative phosphorylation was also observed in brain tumors and melanoma brain metastasis (Chen et al., 2007; Fischer et al., 2019; Molina et al., 2018). Breast cancer cells that mimic neuronal gene expression and utilize gamma-aminobutyric acid (GABA) to augment citric acid cycle analogous to neurons survive better in the brain microenvironment (Neman et al., 2014). Brain metastatic TN breast cancer cells enhance gluconeogenesis to survive (Chen et al., 2015) and have increased expression of glycolytic enzymes (Kim et al., 2014; Palmieri et al., 2009). Likewise, limited microenvironmental serine and glycine results in selection of brain metastatic cells with increased dependency on de novo serine synthesis (Ngo et al., 2020). Molecular and metabolic adaptations within disseminated tumor cells that result in either synchronous, and residual or metachronous brain metastatic disease presentation have not yet been defined.

Through a phenotypic screen in mice, we isolated HER2+ synchronous (S-BM), latent residual (Lat) and metachronous (M-BM) brain metastatic cells. By investigating these phenotypically distinct brain-tropic S-BM, Lat and M-BM cells, we uncovered the impact of metabolic diversity and adaptations on metastatic fitness and identified metabolic vulnerabilities in these cell populations. Genetic or pharmacologic inhibition of these vulnerabilities limits residual and established HER2+ breast cancer brain metastases, identifying a potential therapeutic opportunity.

RESULTS

Phenotypic selection of synchronous, latent and metachronous HER2+ breast cancer brain metastatic cells

To isolate isogeneic synchronous (S-BM), latent (Lat) and metachronous (M-BM) HER2+ brain metastatic cells, we performed a phenotypic screen in vivo. As primary tumor sources, we used HER2+ breast adenocarcinoma cells, HCC1954, derived from a primary stage IIA, grade 3 invasive ductal carcinoma with no lymph node metastasis, and SKBR3, derived from a pleural effusion in a patient with metastatic disease (Table S1). HCC1954 and SKBR3 cells transduced with GFP-luciferase and an antibiotic resistance vector were orthotopically implanted in athymic mice. Tumor progression and metastatic incidence was tracked weekly by bioluminescent imaging (BLI). We observed overt synchronous brain metastatic lesions (S-BM) in 2/10 mice bearing primary HCC1954 tumors and none in mice bearing SKBR3 tumors 5 weeks after implantation. We generated HCC1954 S-BM cell lines from these metastatic lesions. In the majority of HCC1954 and SKBR3 mice, despite tumor progression, animals were healthy with no overt metastasis after 5 weeks. Mimicking clinical practice, we surgically resected the primary tumor and tracked mice for metastatic incidence and progression. Delayed metachronous brain metastases (M-BM) were observed approximately 2-3 months post resection in mice implanted with HCC1954 (n=3) and SKBR3 (n=4). We generated HCC1954 and SKBR3 M-BM cell lines from these metastatic lesions. Brains from mice with no BLI signal, approximately 3 months post tumor resection, were dissociated into single cell suspensions; HCC1954 Lat (n=4) and SKBR3 Lat (n=3) cells were recovered using antibiotic selection (Table S1).

Next, we assessed the phenotypic stability of the isolated isogenic HCC1954 (S-BM; Lat and M-BM) and SKBR3 (Lat and M-BM) derivatives by injecting them intracardially into athymic mice. Early metastatic incidence was detected in HCC1954 S-BM (~3 weeks) by BLI (Fig 1A-B; Left panel). Mice bearing HCC1954 Lat cells were predominantly metastasis-free 2 months post injection (Fig 1A-B; middle panel), while HCC1954 M-BM bearing mice developed metastasis within 5-7 weeks post-injection (Fig 1A-B; right panel). Metastasis to the brain were predominant with occasional spinal metastasis and very rarely to other distal organs. Twelve weeks post-injection majority of mice bearing S-BM (14/15) and M-BM cells (10/14) had succumbed to HER2 brain metastatic disease, while only 1/10 Lat-injected mice developed overt brain metastasis (Fig 1A, S1A). Nonetheless, we were able to detect residual latent cells in the brains of these mice with no magnetic resonance imaging (MRI) and BLI signal (Fig 1C; middle panel, 1D). Similar observations were made with SKBR3 Lat and M-BM injected mice: M-BM injected mice developed metastases approximately 7-8 weeks post-injection, while majority of Lat-injected mice (5/6) did not develop overt metastasis but had residual Lat cancer cells in the brain (Fig S1B-C). Analogous to HCC1954 metastatic models, approximately 75% of mice bearing SKBR3 M-BMs developed brain metastasis 9 weeks post injection and succumbed to death (Fig S1D). We selected these phenotypically stable isogenic brain metastatic cells from HCC1954 (S-BM, Lat, M-BM) and SKBR3 (Lat, M-BM) for further analysis.

Figure 1. Synchronous, latent and metachronous HER2+ breast cancer brain metastatic cells are metabolically distinct.

(A) Bioluminescence imaging (BLI) analysis showing whole body photon flux of mice with intracardiac injection of HCC1954 S-BM (Left panel), Lat (Middle panel) and M-BM (right panel) cells. Each dot represents an individual mouse. *p < 0.05, **p < 0.01, ***p < 0.001, and ns = not significant: One-way ANOVA.

(B) Whole mice BLI images on day 0, 21 and 63 showing different brain metastatic potential of S-BM (Left panel), Lat (Middle panel) and M-BM (right panel).

(C) Magnetic resonance imaging (MRI), ex vivo brain BLI of mice injected with S-BM, Lat, and M-BM cells.

(D) Immunofluorescence (IF) image shows latent residual metastatic cells in brain with no BLI signal.

(E) Bar graph showing ImageJ quantified oncosphere of indicated brain tropic cells from HCC1954 and SKBR3. ****p < 0.0001, One-way ANOVA.

(F) Principal component analysis (PCA) of differentially expressed genes in isogenic brain-tropic cells.

(G and H) Nova Biomedical BioProfile 400 Analyzer (BioProfile 400) data showing secreted lactate and glutamate in HCC1954 (S-BM, Lat and M-BM) and SKBR3 (Lat and M-BM) in oncosphere cultures. (G) One-way ANOVA and (H) Student’s t test, *p < 0.05, **p < 0.01.

Synchronous, latent and metachronous metastases are metabolically distinct

Isolated isogeneic HCC1954 and SKBR3 latent and metastatic derivatives readily form 3D tumor oncospheres, an assay used to measure tumor/metastasis initiating capacity by culturing tumor cells in non-adhesive plates under serum free conditions. Latent derivatives formed oncospheres of smaller size compared to S-BM and M-BM from HCC1954 and SKBR3 (Fig 1E). However, no significant difference in number of oncospheres was observed implying they all had metastasis initiating capabilities with varied potency in agreement with the observed phenotypic differences between these isogenic lines. Moreover, in 2D culture or upon orthotopic implantation no significant difference in growth was observed between S-BM, Lat and M-BM cells (Fig S1E-G). Likewise, brain metastatic lesions in mice bearing S-BM and M-BM had no significant differences in Ki-67 staining (Fig S1H). Principal component analysis of the differentially expressed genes, as measured by RNA-seq, highlighted differences in the transcriptomic profiles of HCC1954 derivatives. Lat and M-BM clustered together away from S-BM (Fig 1F). Gene set enrichment analysis revealed enrichment for glycolysis and fatty acid metabolism in S-BM and M-BM compared to Lat cells (Fig S1I-J).

As transcriptomic profiles and phenotypic traits suggested Lat and metastatic derivatives (S-BM and M-BM) are metabolically distinct, we estimated their ability to catabolize glucose and glutamine, prominent sources of bioenergetics and biosynthesis in cancer cells (DeBerardinis and Cheng, 2010). Metabolism of glucose to pyruvate can support the tricarboxylic acid (TCA) or result in conversion to lactate. Glutamine is an important source of carbon and nitrogen for a variety of pathways that support bioenergetics, biosynthesis and redox homeostasis. Secreted lactate levels were significantly higher in aggressive HCC1954 and SKBR3 metastatic derivatives (S-BM and M-BM) compared to Lat cells in 2D and 3D oncosphere cultures (Fig 1G-H, S1K-L).

High lactate secretion was also observed in syngeneic MMTV-HER2 brain metastatic derivatives compared to their parental counterparts (Valiente et al., 2014) (Fig S1M). Similar observations were made in invasive and therapy resistant MMTV-HER2-PIK3CA^H1047R cells (Hanker et al., 2013) (Fig S1N). We observed a significant increase in glutamate secretion in Lat and M-BM compared to S-BM in HCC1954 cells (Fig 1G, S1K). Likewise, glutamate secretion was also high in both HCC1954 and SKBR3 M-BM compared to Lat (Fig 1G-H, S1K-L) cells. In addition, secreted glutamate was significantly higher in one of the MMTV-HER2 brain metastatic derivatives compared to primary MMTV-HER2 cells and in MMTV-HER2-PIK3CA^H1047R cells (Fig S1M-N).

Distinct glutamine utilization patterns in disseminated brain metastatic breast cancer cells

Next, to assess glucose and glutamine utilization patterns and their contribution to the observed differences in metabolite pools, we performed ¹³C₆ glucose and ¹³C₅ glutamine isotope tracing analyses on BLI⁺ brain metastatic lesions with similar signal intensities in mice injected with HCC1954 S-BM and M-BM (Fig 2A). Mice bearing BLI⁺ M-BM and S-BM lesions were anesthetized and infused intravenously with a bolus of either ¹³C₆ glucose or ¹³C₅ glutamine, followed with continuous infusion for either 3 hours or 5 hours, respectively, to achieve steady state serum levels (Faubert et al., 2017; Marin-Valencia et al., 2012). Next, we performed BLI imaging and identified metastatic lesions by ex vivo imaging, which we then collected for metabolite analysis. Labeling of TCA intermediates from glucose was not significantly different between the M-BM (n=6) and S-BM (n=6) lesions (Fig S2A). A marked increase in labeling of the TCA intermediates succinate (M+4), malate (M+4), aspartate (M+4) and citrate (M+4) from glutamine was observed in M-BM (n=6) compared to S-BM (n=6) brain lesions (Fig 2B). Reductive carboxylation was prominent in S-BM brain metastasis with a significantly high citrate (M+5/M+4) ratio in S-BM lesions (Fig 2C).

Figure 2. Distinct glutamine utilization patterns in phenotypically distinct brain-tropic breast cancer cells.

(A) Cartoon, depicting process of in vivo glucose ¹³C₆ and glutamine ¹³C₅ isotope tracing in BLI positive brain metastatic lesions.

(B) Total labeling of indicated metabolites in brain metastatic lesions analyzed by mass spectrometry and normalized to enriched ¹³C₅-glutamine. **p < 0.01, ***p < 0.001, Student’s t test.

(C) M+5/M+4 citrate ratios highlighting reductive carboxylation of glutamine in S-BM and oxidative carboxylation in M-BM. **p < 0.01, Student’s t test.

(D) ¹³C₆-glucose tracing in isogenic brain-tropic HCC1954 cells shows differential contribution of glucose to TCA metabolites.

(E) ¹³C₅-glutamine tracing in isogenic brain-tropic HCC1954 cells highlighting differential enrichment of metabolites in S-BM, Lat and M-BM cells.

(F and G) M+5/M+4 (Citrate) and M+3/M+4 (malate, aspartate and fumarate) ratio indicating enriched reductive carboxylation of glutamine in S-BM and oxidative carboxylation of glutamine in Lat, M-BM.

(H and I) Representative graph showing enrichment of M+5-glutamate and M+3 lactate in media supernatant of S-BM, Lat, and M-BM. (D-I) Statistics-**p < 0.01 ***p < 0.001, ****p < 0.0001: One-way ANOVA followed by Dunnett’s test.

As mice bearing Lat cells are BLI negative, identifying and isolating latent cells for rapid metabolic analysis presented technical challenges. Therefore, we resorted to in vitro tracing experiments using both metastatic and latent derivatives. We performed two complementary labeling experiments by culturing brain-tropic cells with ¹³C₆ glucose and unlabeled glutamine, and unlabeled glucose with ¹³C₅ glutamine. Metabolites were extracted after 8 hours and analyzed by gas-chromatography-mass spectrometry (GC-MS) to obtain mass isotopologue distributions of tricarboxylic acid cycle (TCA) intermediates. Metastatic derivatives (M-BM and S-BM) showed increased contribution of glucose towards the TCA intermediates citrate, fumarate, malate, aspartate and glutamate compared to latent derivatives (Fig 2D). In contrast, glutamine tracing revealed similarities between HCC1954 latent and M-BM that were distinct from S-BM (Fig 2E, S2B). Augmented fumarate (M+4), malate (M+4), aspartate (M+4) and citrate (M+4) pools were observed in HCC1954 latent and M-BM suggestive of oxidative metabolism, while S-BM cells displayed relative increases in citrate (M+5/M+4), malate (M+3/M+4), aspartate (M+3/M+4), and fumarate (M+3/M+4) pools, suggesting increased reductive carboxylation (Fig 2E-G), in accordance with our in vivo findings. Analogous observations were made in SKBR3 Lat and M-BM model systems (S2C-D). Moreover, increased levels of secreted ¹³C₅ glutamate were observed in HCC1954 Lat and M-BM derivatives in agreement with results obtained from Nova Biomedical BioProfile 400 Analyzer (Fig 2H).

Significantly higher enrichment of M+3 lactate from¹³C₅ glutamine was observed in HCC1954 M-BM and Lat cells compared to S-BM (Fig 2I). Comparing Lat and M-BM cells from HCC1954 and SKBR3, enrichment of M+3 lactate was significantly higher in M-BM cells (Fig 2I, S2E). Malic enzyme (ME), that drives reversible oxidative decarboxylation of malate to pyruvate could contribute to the observed increase in secreted lactate by M-BMs (DeBerardinis et al., 2007). Indeed, western blot analysis showed increased expression of malic enzyme in M-BM relative to Lat (Fig S2F). siRNA depletion of malic enzyme in M-BM resulted in a significant decrease in secreted lactate in M-BM cells (Fig S2G-H, Table S2). Taken together, these data highlight differential glutamine utilization patterns in phenotypically distinct disseminated brain resident HER2+ breast cancer cells.

Lactate supplementation triggers overt metastasis in mice with residual disease

Extracellular lactate could be reutilized by tumor cells as a source of fuel to support citric acid cycle and/or could alter the activity of stromal cells in the brain microenvironment that ultimately results in an established metastasis. Expression of lactate uptake transporter MCT1 was low in brain-tropic S-BM, Lat and M-BM cells compared to parental HCC1954 cells, while the expression of lactate efflux transporter MCT4 was unaltered. To test whether exogenous lactate is sufficient to promote brain metastasis, we administered lactate, sodium lactate or PBS intraperitonially three times a week to athymic mice bearing HCC1954 Lat cells (Fig 3A). Lactate or sodium lactate administration gave rise to overt brain metastasis and conferred poor overall survival relative to controls (Fig 3B, S3A-B). Similar observations were made in athymic mice bearing SKBR3 Lat cells (Fig 3C, S3C).

Figure 3. Lactate supplementation triggers overt metastatic outbreak in mice with residual disease.

(A) Schematic illustrating experimental design. HCC1954 or SKBR3 Lat cells were intracardially injected. Two weeks post-injection lactate (Lac), or sodium lactate (Na-Lac) was administrated intraperitoneally (100μl of 1mM Lac or Na-Lac). Metastatic incidence was tracked by BLI.

(B) Brain metastatic incidence as a measure of photon flux in mice bearing HCC1954 Lat cells administrated with either PBS, Lac or Na-Lac. One-way ANOVA: **p < 0.01, ***p < 0.001.

(C) Brain only photon flux showing metastatic burden in mice bearing SKBR3 Lat cells administrated with either PBS or Lac. **p < 0.01: Student t-test.

(D) Altered neutrophil to NK cell ratio upon lactate treatment in mice bearing HCC1954 Lat cells.

(E) Flow cytometry analysis of peripheral blood showing circulating NK cells as a percentage of CD45+ cells. **p < 0.01: Student t-test.

(F) Immunofluorescent image of mouse brain sections highlighting decreased number of NK cells (Red: NKp46) in proximity of HCC1954 Lat cells (Green: GFP) upon lactate treatment.

(G) Metastatic progression of HCC1954 Lat cells in NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ mice (NOD/SCID Gamma, NSG mice) upon PBS and lactate treatment.

Analysis of blood collected prior to euthanization revealed no persistent steady-state differences in serum lactate levels (Fig S3D). Peripheral blood analysis showed an increase in circulating neutrophil to Natural Killer (NK) cell ratio in mice systemically treated with lactate (Fig 3D). A significant decrease in circulating NK cells was observed with no apparent differences in neutrophils and macrophages (Fig 3E, S3E-G). Furthermore, a significant reduction in number of NK cells were observed in metastatic lesions from mice supplemented with lactate compared to micro-metastatic lesions in PBS injected mice bearing Lat cells (Fig 3F). Analogous to these observations, depletion of NK cells in athymic mice bearing HCC1954 Lat cells or injecting Lat cells into NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ mice (NOD/SCID Gamma, NSG mice) that lack functional NK cells results in metastatic outbreaks (Er et al., 2018; Malladi et al., 2016). As latent cell bearing mice depleted of NK cells will result in metastatic outbreaks irrespective of lactate treatment, we assessed the impact of lactate administration in NSG mice bearing Lat cells and find no significant difference in metastatic outgrowth upon lactate supplementation (Fig 3G, S3H). Taken together, these observations suggest exogenous lactate aids latent cell escape from NK cell-mediated innate immune surveillance.

Lactate limits NK cell clustering and cytotoxicity

Next, we assessed the impact of lactate on NK cell activity and its ability to induce cell death. IL-2 activated human NK cells derived from peripheral blood mononuclear cells form clusters and this priming is critical for NK cell cytotoxicity (Harmon et al., 2019; Kim et al., 2017). NK cell cluster size and numbers were significantly reduced in the presence of lactate or the supernatant from metastatic derivatives and latent cells (Fig 4A-B, S4A). Significant reduction in NK cell cluster size was also observed with NK cells isolated from mouse spleen upon exposure to lactate or supernatant from metastatic cells (Fig S4B). Secretion of effector molecules required for target cell killing, interferon gamma (IFN-γ) and granzyme-b (Gzmb), were also significantly reduced in the presence of lactate or supernatant from brain-tropic cells (Fig 4C-D, S4C-D). Supernatant from the brain metastatic derivatives (S-BM and M-BM) significantly decreased both human and mouse NK cell clustering and cytotoxic response relative to supernatant from latent derivatives (Fig 4A-D, S4A-D).

Figure 4. Lactate limits NK cell clustering and cytotoxicity.

(A) Representative images from InCucyte highlighting the effect of lactate (10mM) or media supernatant from S-BM, Lat and M-BM on human NK cell clustering.

(B) Graph showing quantification of human NK cell clusters under indicated conditions.

(C) Secreted IFN-γ levels by human NK cells in the presence of lactate or supernatant from S-BM, Lat and M-BM. Tumor cell secreted IFN-γ from each media condition were subtracted to estimate NK cell secreted IFN-γ. One-way ANOVA followed by Turkey’s test.

(D) Secreted IFN-γ from mice NK cells in the presence of lactate or supernatant from S-BM, Lat and M-BM.

(E and F) Representative images and quantification of NK cell mediated cytotoxicity upon supplementation of lactate or supernatant from S-BM and M-BM to Lat cells.

(G) Schematic model highlighting the role of lactate in promoting escape from innate immune surveillance in metastatic derivatives compared to Lat cells that survive in equilibrium with innate immune surveillance.

In vitro cytotoxicity assays were performed to assess the ability of latent and metastatic derivatives to survive in the presence of NK cells. Freshly isolated IL-2 activated NK cells from mouse spleens and a human NK cell line were incubated with isogenic metastatic and latent derivatives. Compared to latent cells, S-BM and M-BM were resistant to NK cell-mediated cell death (Fig 4E-F, S4E-F). Lactate supplementation or addition of supernatant from S-BM and M-BM to Lat cells resulted in reduced NK cell cytotoxicity and increased survival compared to control (Fig 4E-F, S4E-F). These data suggest that, while latent cells survive in equilibrium with NK cell-mediated immune surveillance, lactate secreted by brain metastatic cells limits NK cell function and aids escape from innate immune surveillance (Fig 4G).

Differential effect of limiting lactate metabolism on S-BM and M-BM

Increased expression of glycolytic enzymes, secreted lactate and glucose oxidation in metastatic derivatives (Fig 1 G-H, S1K-L, 2D, S2C, S5A) and the ability of exogenous lactate to limit NK cell cytotoxicity and abet overt metastasis in mice with residual disease (Fig 3, 4) prompted us to explore the therapeutic potential of sodium oxamate, a pyruvate analog that inhibits conversion of pyruvate to lactate by lactate dehydrogenase.

Sodium oxamate treatment, as predicted, reduced the amount of secreted lactate in metastatic derivatives compared to Lat cells (Fig 5A). Likewise, administration of stiripentol that inhibits LDH activity (Sada et al., 2015), also resulted in reduced lactate secretion in metastatic derivatives (Fig S5B). Next, we assessed the impact of sodium oxamate on metastatic incidence in mice injected with HCC1954 M-BM and S-BM by administering it intraperitoneally. Sodium oxamate was well tolerated and no difference in body weight was observed relative to controls (Fig S5C). To our surprise, a significant reduction in metastatic incidence was observed only in S-BM (Fig 5B, S5D). Although there was a trend towards reduced metastatic incidence, no significant difference was observed in mice injected with M-BM (Fig S5D).

Figure 5. Differential effect of limiting lactate metabolism on S-BM and M-BM.

(A) Effect of sodium oxamate (10mM) on lactate secretion in LDHA high S-BM and M-BM cells and LDHA low Lat cells.

(B) Metastatic outgrowth as measured by brain only photon flux in in S-BM injected mice treated with vehicle or sodium-oxamate (I.P: 750mg/kg b.w).

(C) shRNA mediated depletion of LDHA (KD^LDHA) attenuates lactate secretion in metastatic derivatives (S-BM and M-BM).

(D) Graph showing changes in oncosphere number LDHA depletion in HCC1954 S-BM and M-BM cells.

(E and F) Effect of doxycycline inducible shRNA mediated depletion of LDHA in S-BM (S-BM-KD^LDHA) and M-BM (M-BM-KD^LDHA) on brain metastatic incidence.

(G) Increased residual disease in mice bearing M-BM depleted of LDHA (M-BM-KD^LDHA) compared to S-BM-KD^LDHA.

(H) Cell ROX Deep red staining followed by flow cytometric analyses highlights increased ROS in KD^LDHA S-BM but not in M-BM.

(I) N-Acetyl-L-cysteine (NAC;1mM) rescues S-BM-KD^LDHA cells ability to form oncospheres.

To directly assess the impact of limiting lactate production and secretion, we transduced S-BM and M-BM with Doxycycline (dox)-inducible shRNAs targeting lactate dehydrogenase A (LDHA), the enzyme responsible for conversion of pyruvate to lactate (Fig S5E). A significant reduction of extracellular lactate production was observed relative to control shRNA in HCC1954 S-BM and M-BM lines (Fig 5C). Reduced extracellular acidification rate (ECAR) and increased oxygen consumption rate (OCR) was observed in LDHA depleted metastatic derivatives (Fig S5F-I). No effect on cell proliferation was observed in 2D cultures; however, LDHA depletion significantly reduced oncosphere formation in S-BM relative to M-BM (Fig 5D, S5J). Consistent with our earlier findings, significant increase in secreted IFN-γ by NK cells was observed in the presence of supernatant from LDHA depleted S-BM and M-BM (Fig S5K-L).

Next, to assess the effect of LDHA depletion on metastatic potential, we intracardially injected mice with S-BM and M-BM cells transduced with dox-inducible shRNAs targeting LDHA and induced depletion of LDHA post extravasation. Reduced brain metastatic incidence was observed in S-BM and M-BM as shown by ex vivo brain imaging (Fig 5E-F). Similar to our earlier observations with sodium oxamate treatment and oncosphere formation, the difference in metastatic incidence was striking in HCC1954 S-BM compared to M-BM. To thoroughly assess the impact of LDHA depletion on S-BM and M-BM, we sectioned the brains and screened for residual disease. Strikingly, more residual LDHA depleted M-BM cells were observed in mouse brains compared to LDHA depleted S-BM cells (Fig 5G, S5M).

Increased oxygen consumption rate observed in both S-BM and M-BM upon LDHA depletion (Fig S5H-I) could result in increased oxidative stress and promote cell death. Relative to S-BM, Lat and M-BM had comparatively low cellular reactive oxygen species (ROS) and high GSH/GSSG ratio (Fig S5N-P). In vivo steady state metabolite analysis also showed increased cysteine and glutathione levels in M-BM lesions compared to S-BM (Fig S5Q-R). LDHA depletion results in further increase of cellular ROS in S-BM, while no such increase was observed in M-BM (Fig 5H), suggesting M-BM are better equipped at coping with oxidative stress. Indeed, antioxidant N-Acetyl-L-cysteine (NAC) administration rescued the observed reduction in oncosphere size upon LDHA loss in S-BM (Fig 5I). Overall, these data suggest that an inability to cope with altered cellular redox upon LDHA depletion in S-BM cells results in their demise and reduced metastasis. M-BM cells, on the other hand, were better at modulating cellular redox status and thus able to survive as residual disease.

xCT mediates metastatic latency

Several observations suggested that exchange of intracellular glutamate was critical to maintain cellular redox balance in Lat and M-BM cells: (1) Reduced cellular ROS with high GSH/GSSG ratio (Fig S5N-P) and increased glutamate secretion was observed in both Lat and M-BM derivatives (Fig 1G-H, S1K-L, 2H). (2) LDHA depletion in M-BM cells resulted in residual disease with no significant difference in cellular ROS levels (Fig 5G-H, S5M). (3) The xCT cystine/glutamate antiporter subunits SLC7A11 and SLC3A2, which import cystine into cells while exporting glutamate, were enriched in Lat and M-BM (Fig S6A). (4) Lat and M-BM cells cultured in media without cystine had reduced glutamate secretion, glutamine consumption, cell viability, and increased ROS (Fig S6B-F). Moreover, limiting xCT function has been reported to attenuate tumor growth and progression (Chen et al., 2009; Conti et al., 2020; LeBoeuf et al., 2020; Liu et al., 2020). Therefore, we examined the role of xCT in promoting survival of latent residual and brain metastatic HER2+ breast cancer cells.

First, we assessed cellular ROS and secreted glutamate levels in HCC1954 and SKBR3 Lat and M-BM derivatives in the presence of erastin, a known pharmacological inhibitor of xCT (Dixon et al., 2012). A significant reduction in glutamate secretion, glutamine consumption and cell viability was observed in disseminated latent and M-BM cells (Fig 6A-B, S6G-H). Decreased GSH/GSSG ratio and a concomitant increase in ROS were also observed in these cells upon erastin treatment (Fig 6C-D, S6I). The observed increase in secreted lactate by M-BM was also attenuated in the presence of erastin (Fig 6E). Moreover, administration of erastin to mice bearing M-BMs resulted in significant reduction of metastatic incidence (Fig 6F, S6J) with no significant side effects (Fig S6K).

Figure 6. xCT mediates metastatic latency.

(A) HCC1954 Lat or M-BM cells were treated with 20μM erastin, pharmacological xCT inhibitor or ctrl (DMSO,0.2%) and secreted glutamate was measured by BioProfile 400.

(B) MTT assay data showing effect of Erastin (20μM, 24 hours) on HCC1954 Lat and M-BM cells.

(C) Representative graph highlighting significant decrease in GSH/GSSG ratio upon erastin treatment (24 hours) in Lat and M-BM cells.

(D) Flow cytometry analysis of Cell ROX deep red staining shows increased reactive oxygen species (ROS) on erastin treatment.

(E) BioProfile 400 analysis shows reduction in secreted lactate in erastin treated M-BMs.

(F) Brain only photon flux and ex vivo brain images showing metastatic burden in mice bearing M-BMs treated with erastin (Oral administration: 30mg/kg b.w Erastin in 5% DMSO, 0.2% methyl cellulose, once daily; two days post intracardial injection) for 5 weeks. Images were acquired 6 weeks post injection. **p < 0.01: Student t-test.

(G and H) BioProfile 400 data showing differential effect of xCT depletion on glutamate and lactate secretion in HCC1954 Lat and M-BM cells.

(I) Showing brain only photon flux and ex vivo brain image of mice injected with M-BM (shCtrl, KD^xCT and KD^xCT+lac) and treated with doxycycline.

(J) Significantly reduced number of GFP positive Lat cells in mouse brains upon xCT depletion and in mice injected with xCT depleted Lat cells supplemented with lactate two weeks post injection.

Next, we depleted xCT in HCC1954 Lat and M-BM cells using Dox-inducible shRNAs. xCT loss resulted in reduced glutamate secretion, lessened glutamine consumption and increased cellular ROS (Fig 6G, S6L-M). Moreover, depletion of xCT in malic enzyme high M-BM cells resulted in reduction of secreted lactate levels (Fig 6H), similar to erastin treatment (Fig 6E). Depletion of xCT also attenuated oncosphere formation and size in Lat and M-BM cells (Fig S6N). Administration of the antioxidant NAC resulted in increased oncosphere formation and size in xCT depleted Lat and M-BM cells (Fig S6N). These studies show limiting xCT function results in increased cellular ROS in both Lat and M-BM cells and attenuates lactate secretion in M-BM cells, suggesting xCT mediated cellular redox homeostasis promotes survival and metastatic fitness of latent residual and M-BM cells.

To assess the effect of xCT on disseminated cancer cell survival in the brain, we administered dox to induce xCT depletion in Lat and M-BM derivatives. Reduced metastatic incidence was observed in xCT depleted M-BM cells compared to their controls as shown by brain only photon flux (Fig 6I, S6O). We observed a significant reduction in the number of residual cells in the whole brain. Likewise, the number of disseminated GFP positive cancer cells was significantly reduced in mice bearing xCT depleted Lat cells compared to control shRNA (Fig 6J). Moreover, as lactate supplementation limits NK cell activity and facilitates metastatic outbreaks, we administered lactate to mice two weeks post injection of xCT depleted Lat and M-BM cells. No overt metastasis or increase in the number of residual tumor cells was observed, affirming elimination of residual disease (Fig 6I-J, S6O).

Inhibiting xCT sensitizes Lat and M-BM to HER2 inhibitors

Next, we performed immunohistochemical evaluation of xCT expression in seven HER2 breast cancer patient primary tumor/lymph node and metachronous metastatic lesions collected from over a year to over a decade after primary treatment (Table S3). Histoscore analysis showed significantly high xCT expression in brain metastatic lesions compared to matched primary tumors (Fig 7A). Conversely, low xCT expression was observed in metastatic lung and neck lymph node samples obtained from two HER2 breast cancer patients (Fig 7B). As depletion or pharmacologic inhibition of xCT resulted in reduced metastasis, we assessed the potential of targeting xCT in combination with the HER2 tyrosine kinase inhibitors (TKIs) lapatinib or tucatinib, approved systemic therapies for metastatic HER2+ breast cancer patients. Both HCC1954 and SKBR3 Lat and M-BMs were resistant to lapatinib and tucatinib as single agents relative to their parental cells (Fig 7C-D, S7A). In contrast, HCC1954 S-BM was sensitive to both lapatinib and tucatinib as single agents and formed fewer oncospheres relative to control (Fig 7C). Erastin failed to block the growth of HCC1954 S-BM oncospheres and had no added benefit in the presence of HER2 TKIs (Fig 7C).

Figure 7. Inhibiting xCT sensitizes Lat and M-BM to HER2 inhibitors.

(A) Immunohistochemical (IHC) staining for xCT (1:100, DAB, 20X) in matched human HER2+ primary tumor and brain metastases (left panel). Representative graph showing xCT histoscore in HER2+ primary tumor and brain met samples (right panel).

(B) xCT staining in lung and neck lymph node metastatic samples obtained from HER2+ breast cancer patients.

(C and D) Bar graphs showing differential sensitivity of HER2 inhibitors (lapatinib, 2μM and tucatinib, 3μM) and xCT inhibitor (erastin, 5μM) alone or in combination on HCC1954 (S-BM, Lat and M-BM), SKBR3 (Lat and M-BM) oncospheres.

(E) Illustration highlighting metabolic differences in S-BM, Lat and M-BM cells along with their sensitivities to HER2 inhibitors - lapatinib, tucatinib and xCT inhibitor erastin as single agent or in combination (Sen: Sensitive, NE: Not effective, Res: Resistant, NAB: No additional benefit).

Both HCC1954 and SKBR3 Lat and M-BM derivatives formed significantly reduced oncospheres in the presence of erastin; this number was further reduced in the presence of lapatinib or tucatinib (Fig 7C-D). We extended these studies to additional HER2+ breast cancer cell lines. HER2 gene amplified BT474 and HCC1569 cells were sensitive to lapatinib and tucatinib, while MDA-361 and HCC202 were resistant. Administration of erastin sensitized all cell lines to HER2 TKIs. (Fig S7B-C).

DISCUSSION

Disseminated tumor cells undergo metabolic adaptations to survive in distal organs (Bergers and Fendt, 2021; Elia et al., 2018; Faubert et al., 2020; Lehuede et al., 2016; Li and Simon, 2020). Here, we have uncovered the metabolic diversity and targetable vulnerabilities of brain-tropic HER2-overexpressing breast cancer cells with varied metastatic fitness (Fig 7E, S7D). Incorporating this knowledge and developing precise therapeutic regimens for cancer patients presented with synchronous, latent residual or metachronous brain metastases has the potential to improve disease free survival.

Metabolic plasticity shapes metastatic fitness

Glucose and glutamine are essential nutrients required by proliferating cancer cells to support cellular bioenergetics and biosynthesis (DeBerardinis and Chandel, 2016). The mammalian brain depends on glucose as its main source of energy; therefore, glucose metabolism is tightly regulated compared to glutamine, which is abundantly available (Chen et al., 2015; DeBerardinis and Cheng, 2010). Our data reveal that metabolic plasticity in utilizing glucose and glutamine by disseminated brain-tropic breast cancer cells shapes metastatic outcome.

These data support a model in which nutrient availability and utilization capacity of disseminated tumor cells (DTCs) in the brain determines metastatic outcome. Upon extravasation, DTCs that cannot outcompete the proximate cells for glucose uptake or efficiently metabolize glutamine perish. Brain-tropic HER2+ cancer cells that outcompete the proximate cells for glucose uptake metabolize lactate and negate innate immune surveillance can initiate synchronous brain metastases. In contrast, disseminated cells that utilize abundantly available glutamine as a primary energy source survive as latent entities; these cells can adapt and reprogram the surrounding microenvironment to initiate delayed asynchronous or metachronous metastasis. Additional metabolic conduits dependent on abundantly available fatty acids, ketone bodies and acetate could also be exploited by residual HER2+ brain metastatic cells and warrant further investigation.

xCT mediated cellular redox homeostasis is critical for latent residual brain metastatic cells

To our surprise, pharmacological or genetic interference of lactate metabolism had differential effect on synchronous and metachronous metastatic cells. A significant increase in cellular ROS and death was observed upon targeting LDHA in synchronous metastatic cells. In contrast, LDHA depleted M-BMs were able to maintain cellular redox status and survive as residual disease. We demonstrated xCT dependent cellular redox in latent residual and metachronous metastatic cells is critical for their survival. Moreover, xCT expression was high in matched metachronous brain metastatic samples from HER2 breast cancer patients. Administration of the antioxidant NAC is effective in limiting cellular ROS and promoting oncosphere formation in LDHA depleted brain-tropic cells, while limiting xCT function results in attenuated metastasis and improved survival. Our findings are in line with studies showing oxidative stress is a deterrent to metastatic progression (Le Gal et al., 2015; Piskounova et al., 2015). Improved understanding of the metabolic-redox circuits in cancer cells is indispensable to limit metastatic disease (Wang et al., 2019).

Lactate metabolism and compromised innate immune surveillance

Cancer cells that enter latency post extravasation survive in equilibrium with innate immune surveillance by downregulating cell surface NK cell activating ligands and cell death receptors (Malladi et al., 2016). Lactate efflux in synchronous and metachronous metastatic cells attenuated NK cell cytotoxicity and accelerated escape from innate immune surveillance, while depleting or pharmacologically targeting LDHA to limit lactate metabolism reduced metastases. Moreover, increased expression of malic enzyme and malate dehydrogenase in metachronous metastasis is critical for lactate metabolism. Overall, lactate efflux from cancer cells can modulate immune response or be reutilized by the tumor cells (Brand et al., 2016; Chen et al., 2016; Colegio et al., 2014; Dietl et al., 2010; Faubert et al., 2017; Li and Simon, 2020; Nakajima and Van Houten, 2013). Our data strongly suggest that limiting lactate metabolism in aggressive HER2+ brain metastatic cells could either promote their demise or enforce metastatic latency. Comprehensive metabolic profiling of the metastatic cancer cell secretome may identify additional immunomodulatory metabolites that aid metastatic colonization.

Therapeutic implications

Current systemic therapies targeted towards HER2+ brain metastatic disease are not curative and patients with synchronous, latent residual or metachronous brain metastases are treated with the same regimens despite differences in clinical presentation. Our study suggests several potential approaches to limit residual disease and outgrowth of brain metastasis. First, targeting lactate metabolism resurrects innate immune surveillance and results in elimination of aggressive metastatic cells that lack redox modulating capabilities. Second, pharmacological inhibition of xCT (with erastin) or glutamine metabolism in combination with current standard of care anti-HER2 drugs is therapeutically beneficial to limit residual disease, potentially delaying metastatic relapse. The erastin analogue PRLX 93936 is currently undergoing clinical testing in multiple myeloma (NCT01695590); clinical use or trials of this strategy to delay brain metastasis in HER2+ metastatic breast cancer patients are warranted. Finally, these findings may also have broader therapeutic applicability to other cancers with a propensity to disseminate to the brain.

Limitations of the study

Employing novel brain synchronous, latent and metastatic models we defined metabolic determinants of metastatic fitness in HER2 breast cancer brain metastatic cells and their ability to survive in equilibrium or escape innate immune surveillance. Development of genetically engineered breast cancer brain metastasis models mimicking human disease are likely to provide a more comprehensive analysis of metastatic fitness in the presence of active innate and adaptive immune surveillance. Erastin administration had impressive potency in mouse-xenograft models. As a prospective clinical-trial candidate, administration of erastin or its analogue PRLX 93936 needs further optimization and testing in brain metastatic patient derived xenografts in combination with current standard of care.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and request for resources and reagents should be directed to the lead contact, Srinivas Malladi (srinivas.malladi@utsouthwestern.edu)

Materials availability

Cell lines generated in this study are available from the lead contact upon request.

Data and code availability

• The raw RNA-seq data for this manuscript are available through Gene Expression Omnibus (GEO) under the accession number GSE180098.

• This paper does not report original code

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Patients and tumor specimens

Female breast cancer patient specimens (HER2+) were obtained from Breast Cancer Research Program, Vanderbilt Ingram Cancer Center and Department of Pathology, UT Southwestern Medical Center with voluntary patient consent. Age of the patients were undisclosed. Unstained paraffin-embedded 5μm tissue sections were obtained and stained for xCT for subsequent analysis.

Mice

All animal studies were performed in accordance with UTSW Institutional Animal Care and Use Committee guidelines (Animal protocol number #2017-102099). 4-5 weeks old female Athymic mice (Hsd: Athymic nude mice-Foxn1^nu) and NSG mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ) were purchased from Envigo (#069) and Jackson laboratory (#005557) were allowed to acclimatize to housing condition in animal facility for 1 week before use for the experiments. Animals used for all the experiment were aged between 5-6 weeks. All mice were housed in the UT Southwestern animal facility room maintained in a 12/12-hour light/dark cycle at a temperature of 20-26°C with 30-50% humidity and fed with standard Teklad diet (Envigo, #2916). In doxycycline inducible shRNA mediated knockdown, mice were fed with doxycycline diet (Envigo, #TD.08541). Ad libitum access to food and water was provided to mice at all times. Animal health was monitored once daily throughout the timeline of experiment.

Cell lines

HCC1954 (PA, S-BM, Lat, M-BM), SKBR3 (PA, Lat, M-BM), BT474, HCC1569, MDA-361, HCC202 cells were cultured in RPMI 1640 and MMTV-HER+2 (PA, BM1, BM2) cells were grown in F/12 DMEM media. Media was supplemented with 10% FBS, 2mM glutamine, 100 units/L penicillin, 100 mg/L streptomycin and 1μg/ml amphotericin B. MMTV-PIK3CA^H1047R cells were grown in DMEM F12 supplemented with 10% FBS, 2mM glutamine, 100 units/L penicillin, 100 mg/L streptomycin and 1μg/ml amphotericin B along with 10ng/ml EGF, 5μg/ml Insulin 5μg/ml dexamethasone, 2nM progesterone. Cells were maintained in 37°C incubator with 5% CO₂ and split every 3 days at 1:4 dilution.

METHOD DETAILS

Isolation of synchronous, latent and metachronous HER2+ breast cancer brain metastatic cells

Athymic mice were used to isolate isogeneic synchronous(S-BM), residual latent (Lat) and metachronous metastatic (M-BM) cells from HCC1954 and SKBR3 HER2+ breast adenocarcinoma cell transduced with GFP-luciferase and antibiotic resistance vectors. Briefly, 2.5 x 10⁶ cells were resuspended in PBS and Matrigel (1:1 ratio) in 1ml. Mice were anaesthetized by controlled isoflurane administration through a nose cone in a sterile hood. An incision was made between the fourth and fifth nipple of the mouse to expose the mammary fat pad, and 100μl of the cell suspension was injected using a 28G insulin syringe. Tumor progression and metastatic incidence was tracked weekly by bioluminescent imaging (BLI) using IVIS (In Vivo Imaging System) Spectrum (PerkinElmer). Bioluminescence signaling intensity was measured using ROI tool in living image software. Overt metastasis was identified by BLI signal and HCC1954 S-BM, HCC1954 M-BM and SKBR3 M-BM cell lines were generated. From mice brains with no BLI signal latent residual cells (HCC1954 Lat; SKBR3 Lat) were isolated as previously described (Malladi et al., 2016). For the MRI, mice were anaesthetized using isoflurane and the respiratory rate was monitored during the procedure. The imaging was done using 7T Bruker Animal MRI Scanner.

Intracardiac injections

1.0 × 10⁵ cells were resuspended in 100μl 1X PBS were intracardially injected into the left ventricle of mice with the help of 26G tuberculin syringe. Tumor progression and metastatic incidence was tracked weekly by BLI. For dox inducible gene depletion experiments in mice Tripz control or shRNA-KD LDHA and xCT cells were intracardially injected. 2 days post-injection mice were fed with doxycycline and monitored for metastatic incidence by BLI.

In vivo and in vitro isotope tracing

For isotope tracing experiments, mice were fasted for overnight before infusion, then a 27-gauge catheter was placed in the lateral tail vein under anesthesia. ¹³C₆- glucose (CLM-1396, Cambridge Isotope Laboratories) was intravenously infused as a bolus of 0.6 mg/g body mass over 1 min in 150μl of saline, followed by continuous infusion of 0.0138 mg/g body mass per min for 3 hours (in a volume of 150μl/hour) and ¹³C₅-glutamine (CLM1822, Cambridge Isotope Laboratories) initial infused bolus was 0.2125mg/g body mass over 1 min in 150μl of saline, followed by continuous infusion of 0.004 mg g⁻¹ body mass per min for 5 hours in a volume of 150μl/hour (Faubert et al., 2017; Marin-Valencia et al., 2012). Blood was collected every hour to monitor the successful infusion of metabolites. Post infusion, mice were deeply anesthetized with ketamine, intracardially perfused with 1x PBS and brain metastatic lesions tracked by BLI were resected and flash frozen in liquid nitrogen and stored at −80°C. 30-50mg of metastatic lesion was homogenized using an electronic tissue disruptor (Qiagen) in 1ml chilled 80% methanol (GC-MS grade, Fisher Scientific) followed by three freeze-thaw cycles using liquid nitrogen and 37°C water bath. Supernatant was collected and lyophilized using a Speed-Vac (Thermo) before running in GC-MS.

For in vitro isotope tracing experiments, 1x10⁶ cells were grown in 6cm dishes with 4ml of R3F media for overnight. Brain-tropic cells were cultured with either ¹³C₆ glucose and unlabeled glutamine (¹³C₆ glucose labelling), or unlabeled glucose with ¹³C₅ glutamine (¹³C₅ glutamine labelling) for 8 hours. Post incubation, media supernatant was collected, and cells were rapidly washed using 300μl of ice-cold 0.9% NaCl and fixed with 250μl of 80% methanol was added to the plates on the dry ice. Collection and processing of sample was done similar in vivo isotope tracing before analyzed in GC-MS or LC-MS.

In vivo metabolite analysis

For in vivo metabolite analysis, metastatic lesions in mice brain were tracked by bioluminescence imaging using IVIS spectrum (Perkin Elmer) and specific metastatic lesions were surgically removed. The tumors were flash frozen in liquid nitrogen and stored at −80°C until further processing. Metastatic lesions weighing 30-50mg were homogenized using an electronic tissue disruptor (Qiagen) in 1ml chilled 80% methanol (LC-MS grade, Fisher Scientific) followed by three freeze-thaw cycles using liquid nitrogen and 37°C water bath. Supernatant was collected after a 10min centrifugation at 13,000g. The supernatant was collected and lyophilized using a Speed-Vac (Thermo) before running on LC-MS.

Extracellular lactate and glutamate measurement

For the measurement of extracellular lactate and glutamate in oncospheres, 5000 cells/MW-6 were resuspended in oncosphere media (HuMEC media with bFGF 10ng/ml, EGF 20ng/ml, Insulin 5μg/ml and 1x B27 supplement) and incubated at 37°C and 5% CO2 for 7-8 days. Media supernatant was collected and centrifuged before analysis in BioProfile 400. Oncospheres were dissociated with trypsin and cells were counted for normalization. For 2D cultures, 0.5x10⁶ cells were grown for 8 hours in 6cm plate with 4ml of R3F media (RPMI 1640 media supplemented with 3% FBS,10mM glucose, 2mM glutamine, 100 IU/ml penicillin-streptomycin). At the end of 8 hours, media was discarded, and cells were replenished with 1ml fresh R3F and kept overnight. Media supernatant was collected and processed as described earlier.

Oncosphere Assays

100-400 cells were resuspended in oncosphere media (HuMEC serum free media with bFGF (10ng/ml), EGF (20ng/ml), Insulin (5μg/ml) and 1XB27 supplement) and incubated at 37°C and 5% CO₂ for 7-8 days. The oncospheres were imaged using EVOS microscope and quantified using ImageJ software.

Cell Viability Assays

In order to determine cell proliferation in various conditions, MTT assay (Parida et al., 2018) or Cell Titer-Glo luminescent cell viability assay (Promega) was performed in 96-well, flat, clear-bottom, opaque wall microplates according to the manufacturer’s protocol.

Western Blotting

Cells grown in R3F were harvested and lysed with RIPA buffer with protease and phosphatase inhibitor. For immunoblotting, 20–35 μg of protein was resolved in 7.5–15% SDS-PAGE, transferred onto nitrocellulose membrane (Millipore, USA), using a semi-dry transfer apparatus and blocked in 5% TBST-milk for 1h followed by incubation with primary antibody in 1% BSA in TBST overnight at 4°C. Membranes were then washed in TBST and incubated with secondary antibody followed by washing and developed using West Femto super signal (Thermofisher) and Bio-Rad imager.

Seahorse Assays

Mito Stress and Glucose Stress Kits were used to compare the differences in energy usage between different cell lines. Cells were plated in 96 well plates and incubated in 37°C and 5% CO₂ for 24 hours. The probes were primed overnight according to manufacturer’s recommendation. After 24 hours the plate was washed in the buffer and the assay was performed on Sea horse XF.

RNA extraction and quantitative RT-PCR

Total RNA was extracted using RNeasy Kit (Qiagen). 0.2-1 μg of RNA was subjected to reverse transcription using Bio-Rad iScript reagents. To determine relative mRNA expression quantitative RT-PCR was performed using SYBR green Supermix Bio-Rad. All mRNA quantification data were normalized to β-actin. Primers used for q-PCR are listed in Table S2.

Lentiviral shRNA knockdown

To generate shRNA knockdown cells, HEK293T cells were prepared and co-transfected with either pTRIPZ or pTRIPZ-shRNA of target gene with packaging plasmid PAX (7μg) and envelop plasmid MD2.G (2.7μg) by lipofection method using lipofectamine 3000 (Invitrogen). 18 hours post lipofection media was changed with fresh media and virus particle were harvested post 48 hours incubation. Cancer cells were transduced with 1:1 ratio of harvested lentivirus to Opti-MEM Pro SFM along with polybrene to a final concentration of 5μg/ml. Following 6 hours incubation cells were replenished with 2-3ml of fresh growth media and incubated overnight and selected by flow sorting and knock down was induced by doxycycline (1μg/ml) supplementation.

siRNA transfection

For this experiment 2x10⁶ cells were grown in 10cm plate for 24 hours prior to transfection. Cells were then transfected with 10nM scrambled or target siRNA by lipofectamine 3000 in opti-MEM for 6 hours and then replaced with complete RPMI-1640.

Differential counting of WBCs in mice peripheral blood by flow cytometry

Blood was collected from submandibular vein of athymic nude mice using EDTA tubes and lysed with ACK lysis buffer. Washed with FACs buffer (PBS, 3% FBS and 0.5M EDTA). In all experimental condition, dead cells were excluded using zombie NIR fixable staining reagent and doublets were excluded by FSC-A versus FSC-H and SSC-A versus FSC-H gating. Following directly labeled anti-mouse primary antibodies were used:CD45-AF 488, CD11b-BV785, F4/80-PerCP/Cy5.5, Ly6G-APC, CD335/NKp46-BV510. Flow cytometry gating and analysis was done as shown in Fig S3E using FlowJo software.

NK Cytotoxicity Assays

Isolated mice splenocytes were mechanically dissociated in a sterile condition and suspended in splenocyte isolation buffer (1XPBS, 0.5% BSA,2mM EDTA, pH7.2). NK cells were isolated from these suspensions by depletion of non-NK cells using magnetic assisted cell sorting (MACS) using NK Cell Isolation Kit (Miltenyi Biotec-negative selection method). Isolated NK cells were expanded in RPMI 1640 containing 15% FBS, 50uM β-2ME, 2 mM L-glutamine and 500 U/ml recombinant mouse-IL-2(Roche) overnight. For co-culture experiments, target cells labeled with Calcein AM-Red dye were incubated with NK cells (effector to target ratio of 1:10) for 24 hours at 37°C with time lapse imaging with IncuCyteS3 imager.

Human NK cells were cultured in media consisted of [αMEM: without ribonucleosides and deoxyribonucleosides, 2mM L-glutamine, 1.5g/L sodium bicarbonate, 0.2mM inositol; 0.1mM β-mercaptoethanol; 0.02mM folic acid] + IL-2 (200U/ml). For co-culture experiments, NK cells were labelled with Calcein AM-Red and target cancer cells labeled with Green cell tracker dye were incubated with a 1:5 ratios for 6 hours at 37°C with time lapse imaging with IncuCyteS3 imager.

IFN-γ and Granzyme-B ELISA assay

Human and mice NK cells were cultured with filtered supernatant from S-BM, Lat, M-BM in 1:1 (NK media: Supernatant) ratio for 6 hours and 24 hours respectively. Supernatant were collected, and ELISA was performed according to manufacturer’s protocol (R&D Biosystem).

GSH/GSSG analysis

3x10⁶ cells were grown in 10cm dish in R3F media for 24 hours at 37°C and 5% CO₂. The level of reduced (GSH) and oxidized glutathione (GSSG) were measured using a kit (Cat. K264, BioVision, Inc, CA, USA) according to the manufacturer’s protocol. All data presented were normalized to 10⁶ cells.

Fluorescence ROS measurement

3x 10⁵ cells were plated in 6-well plates for 24 hours. Cells were collected and washed with 3% FBS in PBS and incubated with 5μM Cell ROX® Deep Red reagent for 30 min. After centrifugation at 1,000 rpm for 5 min, the cell pellets were washed twice and resuspended in 3% FBS in PBS before analyzed in FACS Calibur flow cytometer (Becton-Dickinson).

Immunostaining-Staining of paraffinized brain tissue sections

Brains collected from mice were fixed in 10% normalized buffered formalin (NBF) over 48h before transferring to 1X PBS. The brains were then put in paraffin blocks and 5-10 μm sections were cut and mounted on slides. During staining the slides were de-paraffinized and rehydrated by serial incubations in xylene (2X for 3 min), 100% ethanol (2X for 3 min), 95% ethanol (2X for 3 min), 70% ethanol (2X for 3 min) and PBS (4X for 3 min). To prevent sections from folding or falling, they were incubated in 10%NBF for 30 min. Following this, antigen retrieval was carried out by placing slides in 700ml of Antigen retrieval buffer (10mM Tris HCl, 1mM EDTA, 10% glycerol, pH9.0) at 95°C for 25 min. The slides were then cooled to room temperature for 30 min and rinsed with PBS (2X for 2 minutes) before being blocked in 10% horse, donkey or goat serum (depending on the species of the secondary antibodies) in 0.1% PBS-T for 1 hour at room temperature. The slides were incubated with primary antibodies in 2% serum PBST overnight at 4°C. The next day, the slides were washed in PBS (3X for 10 min) and incubated with secondary antibodies for 1 hour at room temperature. For fluorescence detection, fluorophore conjugated antibodies were used for secondary staining.

Immunofluorescence staining of frozen brain tissue sections

Isolated brain tissues were fixed 4% paraformaldehyde overnight at 4°C, washed in PBS (2X for 15 min each) and cryoprotected in 30% sucrose for 3 days or until the organs settle to the bottom. Tissues were then frozen in optimal cutting temperature compound (OCT compound; Fisher healthcare) at −20°C. Using a cryostat, 16-20 μm sections were cut, washed 3 times in PBS for 5 min, permeabilized in 0.25% PBS-Tween 20 for 45 min before blocked in 10% horse serum in PBST (1X PBS, 0.05% Tween20) for 1h at room temperature. Sections were then stained with primary antibodies (1:100) overnight and washed three times in PBST for 10 min each followed by secondary antibodies (1:500) staining for 1 hour in the dark at room temperature. Sections were washed three times in PBST for 5 min each then stained with DAPI (1:1,000) and mounted with prolong gold for confocal imaging.

RNA-Sequencing and Bioinformatic Analysis

Total RNA isolated from HCC1954 (S-BM, Lat, M-BM) cells grown in R3F conditions. Quality of RNA was checked with an Agilent BioAnalyzer 2000. RNA with an integrity number of greater than 9.5 were used for subsequent analyses. Libraries were prepared with TruSeq RNA Sample Prep Kit v2 (Illumina). RNA libraries were prepared for sequencing using standard Illumina protocols. Single-end 76 bp read length Fastq files were checked for quality using fastqc (v0.11.2) and fastq screen (v0.4.4). Fastq files were mapped to human reference genome(hg19) using STAR (v2.5.3a). Read counts were generated using feature Counts (subread) for coding genes from gencode(v19) and igenomes. Differential expression analysis was performed using edgeR. Principal component analyses were generated using Clustvis web tool (Metsalu and Vilo, 2015). Ranked Gene Set Enrichment Analysis (GSEA) was performed with software version 4.0.3. Default setting was used except for “Collapse/Remap dataset to gene symbols” set to “No Collapse” and “Phenotype” permutation type to “Gene set”.

QUANTIFICATION AND STATISTICAL ANALYSIS

Sample sizes were predetermined as per the requirement of the statistical analysis. All cell culture experiments included at least three biological replicates and in animal experiments each group contained at least 4-15 mice. Individual data points were represented as dots in graphs. The data showed normal distribution either in the D’Agostino-Pearson omnibus or Shapiro-Wilk test. Statistical significance between two comparative groups was determined using Student’s t-test. For statistical comparisons between multiple study groups, one-way ANOVA was used, followed by Dunnett’s test or Turkey test. GraphPad prism 8 software was used for all statistical analysis. Values are expressed as mean ± SEM for control and comparative groups. The values were considered statistically significant as per p values (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 and ns = not significant).

KEY RESOURCES TABLE

REAGENTS or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal anti-xCT/SLC7A11	Cell Signaling Technology	Cat# 12691S; RRID: AB_2687474
Rabbit monoclonal anti-SLC3A2	Cell Signaling Technology	Cat# 13180S; RRID: AB_2687475
Rabbit monoclonal anti-SLC1A3	Cell Signaling Technology	Cat# 14501S; RRID: AB_2798499
Rabbit monoclonal anti-LDHA	Cell Signaling Technology	Cat# 3582S; RRID: AB_2066887
Rabbit polyclonal anti-ME	Cell Signaling Technology	Cat# 12399S; RRID: AB_2797899
Rabbit polyclonal anti-ENO1	Cell Signaling Technology	Cat# 3810S; RRID: AB_2246524
Mouse monoclonal anti-beta Actin	Abcam	Cat# ab49900; RRID: AB_867494
Rabbit polyclonal anti- Firefly Luciferase	Abcam	Cat# ab21176, RRID: AB_446076
Rabbit poly clonal anti-MCT4	Millipore	Cat# AB3314P; RRID: AB_2286063
Rabbit polyclonal anti-GFP	Thermofisher	Cat# PA5-22688; RRID: AB_2540616
Goat polyclonal anti- NKp46	R & D Systems, Inc.	Cat# AF2225; RRID: AB_355192
Rabbit polyclonal anti-Ki67	Abcam	Cat#ab15580, RRID: AB_443209
Mouse anti-CD45-Alexa Fluor 488	Bio-Legend	Cat#103121, RRID: AB_493532
Mouse anti-CD11b-BV785	Bio-Legend	Cat#101243, RRID: AB_2561373
Mouse anti-Ly6G-APC	Bio-Legend	Cat#127613, RRID: AB_1877163
Mouse anti-F4/80- PerCP/Cy5.5	Bio-Legend	Cat#123127, AB_893496
Mouse anti-CD335/NKp46-BV510	Bio-Legend	Cat#137623, AB_2563290
Biological samples
Breast cancer patient specimens	Vanderbilt Ingram Cancer Center	N/A N/A
Breast cancer patient specimens	UT Southwestern Medical center
Chemicals, peptides, and Recombinant proteins
RPMI-1640	Sigma-Aldrich	Cat# R8758
RPMI-1640 - Glutamine, - Glucose	Biological Industries	Cat# SKU: 01-101-1A
Penicillin-Streptomycin	Sigma-Aldrich	Cat# P0781
L-Glutamine	Sigma-Aldrich	Cat# G7513
DMEM F12	Gibco	Cat#12634-010
Opti-MEM™	Gibco	Cat# 31985-070
D-(+) Glucose	Sigma-Aldrich	Cat# G8270-1KG
Pepstatin A	Tocris Bioscience	Cat# 1190
Leupeptin hemisulfate	Tocris Bioscience	Cat# 1167
Aprotinin	Tocris Bioscience	Cat# 4139
Sodium orthovanadate	Sigma-Aldrich	Cat# 450243
Sodium fluoride	Sigma-Aldrich	Cat# 450022
Thiazolyl Blue Tetrazolium Bromide	Sigma-Aldrich	Cat# M5655-1G
Interleukin-2, human (hIL-2)	Sigma-Aldrich	Cat# 10799068001
Interleukin-2, mouse (mIL-2)	Sigma-Aldrich	Cat# 11271164001
2-Mercaptoethanol	Life Technologies	Cat# 21985-023
ACK Lysing Buffer	Thermo Scientific	Cat# A1049201
Calcein AM	Thermo Scientific	Cat# 65-0853-78
Dimethyl sulfoxide (DMSO)	Sigma-Aldrich	Cat# D2650
D-Glucose (13C6)	Cambridge Isotope lab.	Cat# CLM-1396-1
Glutamine(13C5)	Cambridge Isotope lab.	Cat# CLM-184161-19-1
CellROX™ Deep Red Reagent	Thermo Scientific	Cat# C10422
Doxycycline Hydrochloride	Sigma-Aldrich	Cat# D3072
Methyl cellulose	Sigma-Aldrich	Cat# M6385-100G
L-Cystine	Sigma-Aldrich	Cat# C7602-25G
RPMI-1640	Sigma-Aldrich	Cat# R7513
L-Methionine	Sigma-Aldrich	Cat# M9625-25G
FBS, dialyzed, US origin	Thermo Scientific	Cat# A3382001
Oxamic acid	Sigma-Aldrich	Cat# O3750-25G
Sodium lactate	Sigma-Aldrich	Cat# L7022
Lactic acid	Sigma-Aldrich	Cat# L6661
Erastin	Selleck USA	Cat# S7242
Erastin	Medchem express	Cat# HY-15763
Stiripentol	Selleckchem	Cat# 5266
Tucatinib (Irbinitinib, ONT-380)	Selleckchem	Cat# S8362
Lapatinib (10mM/1mL in DMSO)	Selleckchem	Cat# S2111
Neratinib (HKI-272)	Selleckchem	Cat# S2150
N-Acetyl-L-cysteine	Sigma-Aldrich	Cat# A8199-10G
4% Paraformaldehyde in PBS	Santacruz	Cat# sc-281692
hEGF	Sigma-Aldrich	Cat# E9644-.5MG
bFGF	Merck	Cat# GF003-AF
HuMEC Basal Serum-Free Medium	Thermo Scientific	Cat# 12753018
Recombinant Human Insulin Protein	Novus Biologicals	Cat# NBP2-26530
B-27™ Supplement (50X), serum free	Thermo Scientific	Cat# 17504044
Opti-MEM™	Thermo Scientific	Cat# 31985070
Lipofectamine 3000	Thermo Fisher (Invitrogen)	Cat# L3000015
Seahorse XF RPMI Medium (500ml)	Agilent Technologies	Cat# 103336-100
Tissue-Plus™ O.C.T. Compound	Fisher healthcare	Cat# 23-730-571
D-Luciferin potassium salt	Gold Biotechnology	Cat#LUCK-5G
Critical Commercial Assays
Mito Stress Test Kit	Agilent Technologies	Cat# 103015-100
Glycolytic Rate Assay Kit	Agilent Technologies	Cat# 103344-100
Glycolysis Stress Test Kit	Agilent Technologies	Cat# 103020-100
Glutathione Fluorometric Assay Kit	Biovision	Cat# K-264
CellTiter-Glo® assay kit	Promega	Cat# G7571
Human Granzyme B DuoSet ELISA	R&D systems	Cat# DY2906-05
Mouse Granzyme B DuoSet ELISA	R&D systems	Cat# DY1865-05
DuoSet ELISA Ancillary Reagent Kit 2	R&D systems	Cat# DY008
Human IFN-gamma ELISA Kit	R&D systems	Cat# DIF50
Mouse IFN-gamma ELISA Kit	R&D systems	Cat# MIF00
NK Cell Isolation Kit, mouse	MACS Miltenyi Biotec	Cat# 130-115-818
IQ SYBR green supermix	Bio-Rad	Cat# 1708880
Zombie NIR™ Fixable Viability Kit	Bio-Legend	Cat# 423106
Experimental Models: Cell Lines
HCC1954 PA	ATCC	CRL-2338
HCC1954 S-BM, Lat and M-BM	This paper	N/A
SKBR3 PA	ATCC	HTB-30
SKBR3 Lat and M-BM	This paper	N/A
MMTV PA, BM1, BM2	Er et al., 2018
MMTV-HER2-PIK3CAH1047R	Hanker et al.,2013
HCC202	ATCC	CRL-2316
HCC1569	ATCC	CRL-2330
BT474	ATCC	HTB-20
MDA-361	ATCC	HTB-27
HEK293	ATCC	CRL-1573
Experimental Models: Organisms/Strains
Mouse:Hsd:Athymic nude mice-Foxn1nu	Envigo	Cat#069
Mouse: NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ	Jackson laboratory	Cat#005557
Oligonucleotides
shRNA: human LDHA-KD1	Dharmacon	Cat# RHS4696-200774439
shRNA: human LDHA-KD2	Dharmacon	Cat# RHS4696-200778151
shRNA: human SLC7A11-KDxCT#1	Dharmacon	Cat# RHS4696-200759835
shRNA: human SLC7A11-KDxCT#2	Dharmacon	Cat# RHS4696-200710181
shRNA: human SLC7A11-KDxCT#3	Dharmacon	Cat# RHS4696-200758664
siRNA: human Scramble	Dharmacon	Cat# D-001810-01-05
siRNA: human ME1	Dharmacon	Cat# L-009348-00-0005
siRNA: human ME2	Dharmacon	Cat# L-009461-01-0005
q-PCR primers, see Table S2	This paper	NA
Software and Algorithms
GraphPad Prism version 8.0.0	GraphPad Software Inc.	https://www.graphpad.com/scientific-software/prism/
ClustVis (Beta)	ClustVis	https://biit.cs.ut.ee/clustvis/
Living image 4.5	PerkinElmer	https://www.perkinelmer.com/
Biorender	Biorender	https://biorender.com/
Gene Set Enrichment Analysis	Broad Institute	https://www.gsea-msigdb.org/

Highlights:

Metabolically distinct HER2+ brain-tropic cells determine metastatic fitness

Tumor cell secreted lactate modulates NK cell cytotoxicity

xCT mediated cellular redox homeostasis promotes metastatic latency and relapse

Limiting xCT function attenuates metastatic latency and late recurrences