Hypoxia-induced switch in SNAT2/SLC38A2 regulation generates endocrine-resistance in breast cancer

Tumor hypoxia is associated with poor patient outcomes in estrogen receptor-α (ERα) positive breast cancer. Hypoxia is known to affect tumor growth by reprogramming metabolism and regulating amino acid (AA) uptake. Here we show that the glutamine transporter, SNAT2, is the AA transporter most frequently induced by hypoxia in breast cancer and it is regulated by HIF1α both in-vitro and in-vivo in xenografts. SNAT2 induction in MCF7 cells was also regulated by ERα but it became predominantly a HIF-1α-dependent gene under hypoxia. Relevant to this, binding sites for both HIF-1α and ERα overlap in SNAT2’s cis-regulatory elements. In addition, the downregulation of SNAT2 by the ER antagonist fulvestrant was reverted in hypoxia. Overexpression of SNAT2 in-vitro to recapitulate the levels induced by hypoxia caused enhanced growth, particularly after ERα inhibition, in hypoxia, or when glutamine levels were low. SNAT2 upregulation in-vivo caused complete resistance to anti-estrogen and, partially, anti-VEGF therapies. Finally, high SNAT2 expression levels correlate with HIF-1α and worse outcome in patients given anti-estrogen therapy. Our findings show a switch in regulation of SNAT2 between ERα and HIF-1α, leading to endocrine resistance in hypoxia. Development of drugs targeting SNAT2 may be of value for a subset of hormone-resistant breast cancer.


Introduction
The estrogen receptor positive (ERα+) subtype accounts for approximately 70% of all newly diagnosed cases of breast cancer in Europe and the USA (1). The majority of these tumors depend on estrogen signaling, thereby providing the rationale for using anti-estrogens as adjuvant therapy to treat breast cancer. Endocrine agents targeting estrogen receptor alpha (ERα) such as tamoxifen, fulvestrant/faslodex or inhibiting estrogen biosynthesis, such as aromatase inhibitors, represent the cornerstone of systemic treatment of this breast cancer subtype, for both early and metastatic disease (2). However, despite recent therapeutic advances, poor response and resistance limit the effectiveness of these agents in up to 30 % of patients, manifesting as either relapse after therapy or as disease progression in the metastatic setting (3).
Recent randomized trials confirmed that first line endocrine therapy with Fulvestrant alone or in combination with palbociclib, a Cyclin-dependent kinase 4 and 6 inhibitor, improve clinical outcomes in ER+ metastatic breast cancer patients with disease progression or early relapse following anti-estrogen treatment (4-6).
However, as seen for other hormonal treatments, resistance eventually develops and leads to disease progression.
Many mechanisms have been proposed to account for endocrine resistance such as genetic (loss of ERα expression, mutation, expression of truncated ER isoforms), post-translational modification of ERα and activation by peptide growth factors and epigenetic (methylation, promoter inhibition) changes within the tumor that activate hormone-independent mitogenic pathways (7). In addition to cancer cell-autonomous factors, the host microenvironment can contribute to endocrine resistance (8).
Hypoxia is a key physiological and microenvironmental difference between tumor and normal tissues and is related to poor clinical prognosis and resistance to therapies in many solid tumors, including breast cancer (9, 10). Indeed hypoxia is associated with resistance to chemoendocrine treatment in breast cancer patients (11).
Furthermore, increased expression of hypoxia-inducible factor 1α (HIF-1α), the key transcription factor mediating hypoxia response, is associated with tamoxifen resistance in neoadjuvant, primary therapy of ERα+ breast cancers (12) and is associated with resistance to primary endocrine therapy in primary ER+ breast cancers (13). Moreover, we recently found that the HIF-1α gene bears a canonical ER-binding element that responds to estrogen signaling, demonstrating a direct regulatory link between the ERα and HIF-1α pathways in breast cancer (14). HIF-1α function conferred resistance to fulvestrant and could therefore compensate for estrogen signaling when ERα function is compromised during hormone therapy, and thus cause resistance to tamoxifen or fulvestrant treatment (14).
A crucial mechanism by which the growth of cancer cells is promoted in hypoxic microenvironments is by metabolic reprogramming (15). The adaptation to hypoxia is coordinated by HIF-1α, which induces metabolic genes involved in increasing glycolytic flux (16). There is also upregulation of glutamine metabolism to support proliferation, lipid biosynthesis and protection from free radical stress (17,18). In addition to pyruvate derived from glycolysis, hypoxic cancer cells can supply substrates to the tricarboxylic acid (TCA) cycle to sustain mitochondrial ATP production (anaplerosis) through the uptake of amino acids (AAs), such as glutamine, glycine and serine (19). In particular glutamine, the most abundant AA in the plasma, can fuel the TCA cycle through a series of biochemical reactions termed glutaminolysis (20).
This gives a strong rationale to identify hypoxia-induced metabolic alterations, particularly regarding glutaminolysis. These have the potential to provide induced essentiality opportunities in combination with other drugs (21). Metabolomic analysis of breast cancer patients showed that cancer cells have increased glutamine metabolism compared to normal cells. Compared with the ratio of glutamate to glutamine (indirect assessment of glutaminolysis) in normal tissues, 56% of the ER+ tumor tissues were glutamate enriched (22). Several studies have shown that endocrine therapy resistance in breast cancer cells is modulated by metabolic rewiring and tamoxifen-resistant cells are characterized by HIF-1α hyper-activation via modulation of Akt/mTOR, thus resulting in enhanced aerobic glycolysis and mitochondrial metabolism (23,24). These data highlight the importance of metabolic adaptability of cancer cells for endocrine therapy resistance and suggest that targeting glutamine metabolism could be a novel approach to overcome resistance to endocrine therapies.
Because tumor cells under hypoxia have a high demand for AA, we hypothesized that AA transporters may be upregulated selectively to meet this demand. Thus, we investigated the role of HIF-1α regulated AA transporters and in endocrine therapy resistance.

Identification of hypoxia-induced amino acid transporters by RNA-sequencing
To define the AA transporters that are involved in hypoxic adaptation, a panel of As AAs are crucial for maintenance of the TCA (Krebs) cycle throughout anaplerosis, we applied in silico hierarchical clustering and supervised exploratory analysis in order to evaluate which AA transporters, promoting the uptake of specific substrates in the TCA, might be crucial for maintaining anaplerotic pathways in hypoxia. We clustered the AA transporters based on substrate transport: -uptake of glucogenic versus ketogenic AA; essential versus non-essential AA and the entry of their substrates into the TCA cycle.
Our data show that several AA transporters are upregulated under hypoxia in breast cancer cell lines (Supplementary Table 1 figure 1D). A quantitative PCR assay for specific amplification only of the two main isoforms (1-2) was used, as isoform 3 was not detected. In the cell lines where SNAT2 was induced, there was an increase of expression of both SNAT2 mRNA isoform 1 and 2 after 24 hours of 0.1% O 2 (Supplemental figure 1E).

SNAT2 protein is upregulated under hypoxia in vitro in a HIF-1α dependent manner
We examined SNAT2 protein expression ( reduced by HIF-1α siRNA (P < 0.01, n = 3) but not HIF-2α, although a nonsignificant reduction in SNAT2 induction of ~50% was seen ( Figure 2C). The siRNA knockdown was validated by immunoblot of the expression of both HIF-1α and HIF-2α and by measuring carbonic anhydrase 9 (CA9) protein levels, which has been previously shown to be dependent on HIF-1α but not HIF-2α ( Figure 2D) (30). To further test the importance of HIF-1α for SNAT2 induction, we stably overexpressed HIF-1α by retroviral vector-mediated transduction in MCF7 (MCF7-HIF1α-o).
Measurement of SNAT2 and CA9 mRNA levels in normoxia showed an increased SNAT2 expression in MCF7-HIF1α-o, while no further increase was seen in hypoxia (Supplemental figure 2E).

SNAT2 is upregulated under hypoxia in vivo
To determine whether hypoxic induction of SNAT2 occurs also in solid tumors, two breast cancer cell lines, MCF-7 and MDA-MB-231, were injected into nude mice and grown as xenografts. MDA-231 cells were used to assess effect of hypoxia independently of ER expression. The mice were treated with either saline control or the VEGF inhibitor, Bevacizumab, which slows tumor growth temporarily and induces tumor hypoxia (30). Treatment with Bevacizumab increased SNAT2 protein levels in xenografts derived from both cell lines. In MCF7, SNAT2 protein expression was 10.5% after treatment compared to 6.9% in controls (p<0.05 n=5 per group). In MDA-MB-231, SNAT2 expression was 14.2% after treatment compared to 9.6% in the controls (p<0.05 n=5 per group; Figure 3A). Human SNAT2 mRNA and CA9 mRNA were upregulated also (Supplemental figure 3A-B). The levels of CD31 (blood vessels), H/E (necrosis) and CA9 protein (hypoxia) were measured in the same tumors (Supplemental Figure 3C-D). Bevacizumab decreased the percentage of CD31-positive cells. Accordingly, the proportion of necrosis was significantly higher and the percentage of CA9 positive cells was greater in Bevacizumab-treated xenografts compared to PBS (Supplemental figure 3C-D).
Furthermore, we found that SNAT2 colocalized in the same area where CA9 was expressed (hypoxic areas) (Fig 3B). To further investigate the effect of HIF-1α on SNAT2 expression in vivo an orthotopic xenograft tumor was established using MCF7-HIF1α-o. Compared with the control, HIF-1α overexpression resulted in an increased SNAT2 expression ( Figure 3C-D). These data show that SNAT2 is regulated by hypoxia via HIF-1α in vivo, as well as in vitro.

ERα and HIF-1α bind to overlapping sites in the SNAT2 promoter in MCF7 cells, but do not act synergistically
Previous studies showed that the expression of SNAT2 was increased in ER+ positive breast cancer cell lines after 17β-estradiol (E 2 ) stimulation and an estrogen response element (ERE) was described in the SNAT2 promoter in rat mammary glands during gestation (31, 32). Because hypoxia is related to adverse outcome in tamoxifentreated breast cancer patients (14) and SNAT2 was among the 202 genes bound by HIF-1α and ERα and regulated by fulvestrant (14), we investigated the regulation of SNAT2 expression by estrogen and hypoxia in ER+ breast cancer cell lines. Utilizing RNA-seq and HIF-1α and HIF-2α ChIP-seq data in MCF-7 cells, we confirmed the presence of HIF-binding site (HRE) sequences (RCGTG) both for HIF-1α and HIF-2α just upstream of the promoter region of SNAT2 ( Figure 4A)(33).
We aligned the ChIP results of HIF-1α in MCF7 with previous publicly available ERα-ChIP-seq data from ENCODE in the same cell line (34). Both binding sites for HIF-1α and ERα are aligned and overlapping in the genome in cis-regulatory elements, suggesting a potential interaction between these two transcription factors for SNAT2 induction ( Figure 4A, Supplemental figure 4A). Interestingly SNAT1, another system A transporter, showed ERα binding sites but not HIFs binding sites (Supplemental figure 4B).
We decided to investigate if an interaction existed between HIF-1α and ERα in driving SNAT2 expression. We cultured MCF7 and T47D cells in normal DMEM medium and then placed them in normoxia or hypoxia and treated with or without E 2 .
Estrogen supplementation induced SNAT2 expression in normoxia. Small additive effects were seen in hypoxia in MCF7 but levels of ERα protein significantly decreased ( Figure 4B)(35). A similar induction of SNAT2 under estradiol and hypoxia, but with a different magnitude, was also seen for T47D cell line.
Time course experiments of estradiol supplementation in MCF7 in normoxia and hypoxia confirmed the estradiol dependency in normoxia with no significant additive effects in hypoxia at mRNA and protein level (Supplemental figure 4C-D). To evaluate if these different stresses (estradiol and hypoxia) might be responsible for different localization of SNAT2, we examined SNAT2 distribution using immunofluorescence microscopy after 24 hours of hypoxia (0.1% O 2 ) and estradiol treatment (10nm). Cells were stained for SNAT2 (green) and the phalloidin (f-actin, red). None of the treatments used significantly affected SNAT2 distribution. SNAT2 was located on punctate structures and the plasma membrane, but the majority was on an intracellular compartment (perinuclear structures), which has been reported to be trans-Golgi network (36, 37). Hypoxia and estradiol resulted in a significant increase in SNAT2 staining intensity but not a clear redistribution of SNAT2 to other structures ( Figure 4C).

ERα and HIF-1α signaling regulate SNAT2 expression independently
To further confirm SNAT2 as an ERα and HIF-1α dependent gene, we cultured MCF7 cells for 5 days in charcoal-stripped phenol-free medium and treated with E 2 with or without fulvestrant (ICI182780, which induces ERα degradation) in normoxia and hypoxia. We found that fulvestrant treatment abolished estradiol-dependent SNAT2 induction in normoxia but only a small effect was seen in hypoxia ( Figure   5A), suggesting a major role for HIF-1α in this condition.
To further test the predominant effect of HIF-1α, MCF7-HIF1α-o cells were treated or not with estradiol in normoxia and hypoxia. SNAT2 was induced in normoxia and hypoxia in MCF7-HIF1α-o cells. When E 2 was supplemented to WT cells, an increase of SNAT2 was seen only in normoxia but not in hypoxia ( Figure 5B).
Interestingly when E 2 was supplemented to MCF7-HIF1α-o an increased level of SNAT2 was seen in both conditions, suggesting that HIF1-o might enhance ER signaling as previously reported ( Figure 5B) (14).
To further assess if this overlapping in the genome between HIF-1α and ERα binding sites is only related to SNAT2, we evaluated the 202 genes that we previously showed to have both binding sites for . Using different public available experiments (33, 34), we confirmed 179 genes with both HIF-1α and ERα binding sites. We found only 31 of these genes (31/179, 17.3%) with HIF-1α and ERα binding sites located in the same genomic region (Supplementary Table 1).
Most of these 31 genes, such as ALDOA, NEAT1 and GAPDH are well known HIF-1α or ERα targets.
An example of a gene with overlapping binding sites is QSOX1 (Supplemental figure   5A). To assess if some of these genes had the same pattern of SNAT2 under fulvestrant and hypoxia, we measured mRNA levels in normoxia and hypoxia with or without estradiol and fulvestrant (ICI182780) for GAPDH, ALDOA and NEAT1. We confirmed also for those genes the estradiol dependency in normoxia and the HIF-1αdependency in hypoxia and resistance to fulvestrant (Supplemental figure 5B).
This list of 31 genes was selected for further general and metabolic pathway enrichment analysis conducted by MetaCore GeneGo pathway. We found enrichment in glycolysis, HIF-1α and Notch pathways (Supplemental figure 5C). The most relevant GO process was the vesicle-transport pathway (Supplemental figure 5D).
These data collectively demonstrate that SNAT2 induction, as well other genes, in MCF7 is regulated by ERα but it becomes predominantly a HIF-1α-dependent gene under hypoxia, due to a potential compensation by HIF-1α when ERα is degraded.

SNAT2 is increased in a tamoxifen-resistant breast cancer cell line in normoxic conditions
SNAT2 expression was analyzed in the tamoxifen-resistant MCF7 cell line (MCF7/TamR), which has been developed through continuous passage of MCF-7 cells for 6 months in the presence of 4-hydroxytamoxifen (4-OHT) (38). SNAT2 protein was higher in MCF7/TamR compared to parental MCF7 in normoxia, but not in hypoxia ( Figure 5C).
When MCF7-TamR cells were grown as a 3D model (spheroids) an increase in cell growth was seen when compared to MCF7 parental. Interestingly when cells were treated with the SNAT2 inhibitor (MeIAB 10mM), MCF7-TamRs were more sensitive to the treatment (maximum volume for MCF7 0.56 mm 3 vs. 0.35 mm 3 for MCF7-TamR, two-way ANOVA, p<0.001). This corresponds to a 37.5% decrease in cell growth for MCF7-TamR after MeIAB treatment, versus a 21% decrease in cell growth after MeIAB treatment for parental MCF7 (t-test, p<0.001)( Figure 5D).

SNAT2 knockdown sensitizes breast cancer cells to anti-estrogen treatment and decreases glutamine consumption, mitochondrial respiration and regulates mTORC1 signaling
To investigate the role of SNAT2 in anti-estrogen resistance, we knocked down SNAT2 with a siRNA pool (double transfection). Cells were treated with E 2 (10 nM) and/or fulvestrant (10 μ g/ml) in normoxia and hypoxia ( Figure 6A). Cell growth was followed over five days. Consistent with their ER positivity MCF7 cells were dependent on E 2 for their growth in normoxia and sensitive to fulvestrant in normoxia and hypoxia ( Figure 6A). SNAT2 knockdown reduced cell proliferation in normoxia but the inhibitory effect was greater in hypoxia ( Figure 6A).
We then investigated if reduced proliferation was linked to reduced glutamine intake.
We found that glutamine consumption was increased after E 2 supplementation in normoxia and also decreased after fulvestrant treatment as previously described (24).
Glutamine consumption was also increased under hypoxia, but no effects were seen with E 2 supplementation or fulvestrant treatment in hypoxia. Interestingly we found that SNAT2 knockdown did not significantly decrease glutamine consumption in normoxia, but a clear effect was seen under hypoxia. An additive inhibitory effect on reducing glutamine consumption was seen in MCF-7 treated with SNAT2 siRNA (versus scrambled controls) and fulvestrant in hypoxia ( Figure 6B). These data indicate that in MCF7, SNAT2 inhibition sensitizes to hypoxia and anti-estrogen treatment and reduces glutamine uptake.
As glutamine is a key metabolite to maintain anaplerosis and mitochondrial function, we used FACS analysis with the cationic dye JC-1 staining to examine the change in mitochondrial membrane potential (ΔΨm) following SNAT2 knockdown in normoxia and hypoxia. Upon knockdown of SNAT2 and hypoxia, the percentage of cells with loss of mitochondrial membrane potential increased from 14.9% of the total to 26.1% of the total, respectively (p<0.01). An additive effect compared to control was seen when SNAT2 knockdown was performed in hypoxia, with 35.8% of the cells showing a loss of mitochondrial membrane potential (p<0.001) ( Figure 6D).
The effect of SNAT2 knockdown on mitochondria respiration was measured by the SeaHorse analyzer. SNAT2 knockdown reduced basal OCR in normal and glutamine-free medium ( Fig 6C). Moreover, SNAT2 knockdown in glutamine-free medium had a more pronounced effect in decreasing basal respiration (p<0.05), ATP production (p<0.01) and maximal respiration (p<0.01) compared to full medium ( Figure 6C). Thus, SNAT2 knockdown affects mitochondrial respiration and function in MCF7 breast cancer cells in both normoxic and hypoxic conditions.
We then assessed the SNAT2 knockdown effect on the mTORC1 pathway (a key AA-sensing pathway). SNAT2 depletion both in normoxia and hypoxia reduced the level of phosphorylated mTOR (p-mTOR) its downstream targets, p-pS6 and p-p70 (only in hypoxia) when compared with their total protein levels (Supplemental figure   6A). These results are consistent with previous studies that showed the ability of SNAT2, similar to other AA transporters, to modulate the mTORC1 signaling cascades (39, 40).
Recent research suggested that mTORC1 is localized also at the level of Trans-Golgi network (TGN) and can be activated by specific TGN-AA transporter (41, 42). As SNAT2 is also localized in the TGN (37) and has been suggested to act as a transceptor (40), we performed confocal microscopy of SNAT2, TGN and mTOR in MCF7 cells.
Confocal Immunofluorescence staining of MCF7 cells revealed that SNAT2 is concentrated on the TGN (Supplemental figure 6B). Interestingly, although most mTOR is localized in the cytoplasm, some colocalization with the TGN was also observed, raising the possibility that it might be associated with SNAT2 in this compartment.

SNAT2 is induced by amino acid deprivation and its overexpression promotes resistance to glutamine starvation, hypoxia and anti-estrogen treatment in vitro
To assess if the shortage of specific AAs can upregulate SNAT2, we incubated MCF7 cells in the culture medium depleted of several gluconeogenic and ketogenic AAs in normoxia and hypoxia. Glutamine, serine and glycine (SNAT2 substrates) increased SNAT2 expression in normoxia. Interestingly, when MCF7 cells where incubated under hypoxia, the SNAT2 upregulation became more profound and independent of single AA deprivation (Supplemental figure 6C) We investigated the growth inhibition after SNAT2 knockdown during metabolic stress (glucose free or glutamine free medium) compared to the nutrient-rich condition in normoxia and hypoxia in MCF7 cell line. SNAT2 knockdown had an effect on 2D cancer cell growth only under hypoxia in full medium (Supplemental figure 6D). SNAT2 knockdown sensitized MCF7 to glutamine deprivation in 2D growth, particularly under hypoxia. A small additive effect was also seen when SNAT2 was knocked-down in glucose-free medium.
Reciprocally, MCF7-SNAT2-overexpressing spheroids grew faster than their parental controls in full and, particularly, in glutamine-deprived medium (Supplemental figure   6G). These data indicate that in MCF7, SNAT2 inhibition sensitizes to glutamine deprivation and increased expression enhances growth in low glutamine conditions.

SNAT2 overexpression promotes resistance to antiangiogenic and antiestrogenic treatment in vivo
To further assess the role of SNAT2 in resistance to anti-estrogen treatment and hypoxia, MCF7 and MCF7-SNAT2-o were grown as xenografts with and without bevacizumab or fulvestrant treatment ( Figure 6E). MCF7-SNAT2-o xenografts grew faster than the empty vector clone (64.9% the growth rate of the empty vector; *, p < 0.01, n = 7). Bevacizumab treatment decreased the growth of empty vector tumors in a first phase and then tumor became resistant. Fulvestrant treatment also reduced xenograft growth rate (55.4% the rate of empty vector control; *, p < 0.01, n = 7).
The SNAT2 expressors grew more rapidly than the controls and there was no impact on this increased growth by either treatment.
The MCF7-SNAT2-o clones treated with bevacizumab grew significantly faster than MCF7 empty vector treated with bevacizumab (11.2% the growth rate of treated empty vector; *, p = 0.038, n =7). More strikingly MCF7-SNAT2-o clones treated with fulvestrant grew significantly faster than MCF7 empty vector and became completely resistant to fulvestrant (267% the growth rate of treated empty vector, p < 0.001, n =7) with a curve similar to untreated MCF7-SNAT2-o.
SNAT2-o xenografts, independently of the treatments, had increased proliferation compared to controls as determined by Ki67 staining (p<0.001). Interestingly no significant differences were seen in Ki67 staining amongst SNAT2-o xenografts treated or not with bevacizumab or fulvestrant (Supplemental figure 6F).

SNAT2 expression correlates with tumor hypoxia and is associated with a poorer recurrence free survival in endocrine treated breast cancer patients
To evaluate the clinical relevance of SNAT2 in breast cancer, we first assessed if SNAT2 expression correlates with hypoxia in vivo. Analyzing gene expression data from 2433 breast cancer patients using the Metabric cohort (43), we found that SNAT2 mRNA abundance significantly correlated with the expression of many genes in our previously reported in vivo hypoxia signature, (44) but not with the whole signature itself (Supplemental figure 7A). For confirmation SLC7A5 (a HIF-2α regulated gene) showed a correlation with the hypoxia signature (Supplemental figure   7A). Interestingly SNAT2 mRNA levels did not correlate with c-Myc copy number, a key regulator of glutamine metabolism (Supplemental figure 7B).
We used additional expression data derived from different breast cancer gene-array cohorts (4132 patients)(45) and we tested if the SNAT2 levels correlated with worse outcomes in ER+ patients treated with all adjuvant endocrine treatments [including aromatase inhibitors] or tamoxifen only. High SNAT2 expression correlated with low recurrence free-survival in patients who received endocrine or tamoxifen treatments ( Figure 7A). Moreover, when we looked at the tamoxifen-treated cohort we found that high SNAT2 levels correlated with worse outcome in tamoxifen-treated luminal B but not luminal A patients ( Figure 7B).

Discussion
Hypoxia is a cause of tumor aggressiveness and resistance to treatments, including endocrine therapy. Although the role of hypoxia and the HIF-1α transcriptional response in promoting tumor progression and metastasis is well established, the direct contribution of the HIF family to the regulation of AA transporters has been less studied, apart from LAT1 and glutamate transporters (25). Here we describe a new mechanism by which hypoxia can produce resistance to anti-endocrine therapy, by substituting for ERα in regulating the AA transporter SNAT2, a glutamine transporter. We found that despite the induction of several AA transporters in hypoxia, SNAT2 expression was key and able to control the growth response to glutamine and anti-estrogen treatment in vitro and markedly so in vivo.
Both main SNAT2 isoforms were hypoxia-regulated. Until now the main isoform has been investigated while for other isoforms their role has not been studied yet.
Recently, it has been shown that different SLC1A5 isoforms are upregulated and needed for SLC1A5 activity under AA deprivation, suggesting that the isoforms might form a complex necessary for AA transporter activity (46). Therefore, the role of isoform 2 now requires further investigation. The maturation of SNAT2 protein and transmembrane localization requires its glycosylation (47) which was maintained in hypoxia.
Accumulating data suggest a significant interplay between hypoxia and estrogenmediated pathways in breast cancer cells (14,48,49). Hypoxia down-regulates ERα in human breast cancer cells via a proteasome pathway (49). We found that both HIF- The xenograft models demonstrated a far more profound effect than the in vitro data.
It is well known that metabolic modifications are often more profound in vivo because of continuous poor nutrient conditions and hypoxic, acidic and glucose gradients. Additionally, SNAT2 expression is tightly controlled also at the posttranslational level (50). AA starvation results in uncharged tRNAs, activating GCN2 and ATF4, which in turn will cause translation of the abundant SNAT2 mRNA by a cap-independent mechanism (51). SNAT2 protein is also degraded after ubiquitination (52), and this may occur under nutrient-replete conditions, but is halted when amino acids are depleted. This could explain the more rapid growth rate of SNAT2 overexpressing tumors and potentially related to scavenging glutamine.
SNAT2 knockdown shifted the mitochondrial membrane potential both in normoxia and hypoxia, suggesting the decrease in cell growth might be mediated by mitochondrial impairment. These data suggest that the dependency on SNAT2 in hypoxia to maintain the TCA cycle cannot be compensated by other AA transporters (such SNAT1 or ASCT2), which are not induced by hypoxia in breast cancer cells.
c-Myc is a well-known downstream effector of ER-α (53). It is upregulated by E 2 and it plays a critical role in modulating resistance to endocrine therapies, by the unfolded protein response (UPR), in ERα-positive breast cancer (54-56). Although SNAT2 is induced by the UPR (57) and Myc was found to selectively bind to the promoter regions of SNAT2 (58), we did not find any correlation between c-Myc copy number and SNAT2 mRNA levels in breast cancer patients (Supplemental figure 7B).
Finally, our analysis of clinical data available from large RNA expression cohorts suggests that high baseline SNAT2 expression correlates with reduced recurrence free survival in ER+ endocrine-treated breast cancer patients, specifically in luminal B subtype. This is particularly interesting as the luminal B subtype (ER+ subtype) has aggressive clinical behavior, with prognosis similar to that of HER2-enriched and basal-like group and a lower sensitivity to endocrine treatment compared to luminal A ER+ breast cancer (59). Interestingly, recent studies showed that a high expression of different AA transporters, such as SLC3A2 or SLC7A5, is related to poor outcomes in luminal B ER+ breast cancer subtype (60, 61). These data suggest that this subtype might be particular vulnerable to glutamine depletion.
Clinical selection may be possible, as recently an amino acid-based PET radiotracer, 18 F-fluciclovine, has been investigated in the imaging of breast cancer (62).
As targeting glutamine metabolism is the subject of intensive research due to its potential clinical applications (20), we propose that SNAT2 should be investigated as a predictive biomarker and a potential target for well-defined molecular subtypes (luminal B) of ER+ breast cancer patients and as a combination approach to overcome endocrine-therapy resistance.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed by the authors.    w  s  k  i  D  ,  G  u  a  n  B  J  ,  W  u  J  ,  Z  h  e  n  g  Y  ,  P  a  t  t  a  b  i  r  a  m  a  n  P  P  ,  J  o  b  a  v  a  R  ,  e  t  a  l  .  G  A  D  D  3  4  F  u  n  c  t  i  o  n  i  n  P  r  o  t  e  i  n  T  r  a  f  f  i  c  k  i  n  g  P  r  o  m  o  t  e  s  A  d  a  p  t  a  t  i  o  n  t  o  H  y  p  e  r  o  s  m  o  t  i  c  S  t  r  e  s  s  i  n  H  u  m  a  n  C  o  r  n  e  a  l  C  e  l  l  s  .  C  e  l  l  R  e  p  .  2  0  1  7  ;  2  1  (  1  0  )  :  2  8  9  5  -9  1  0  .  3  7  .  H  a  t  a  n  a  k  a  T  ,  H  a  t  a  n  a  k  a  Y  ,  T  s  u  c  h  i  d  a  J  ,  G  a  n  a  p  a  t  h  y  V  ,  a  n  d  S  e  t  o  u  M  .  A  m  i  n  o  a  c  i  d  t  r  a  n  s  p  o  r  t  e  r  A  T  A  2  i  s  s  t  o  r  e  d  a  t  t  h  e  t  r  a  n  s  -G  o  l  g  i  n  e  t  w  o  r  k  a  n  d  r  e  l  e  a  s  e  d  b  y  i  n  s  u  l  i  n  s  t  i  m  u  l  u

RNA sequencing and bioinformatics
RNA-sequencing was performed as previously described (33

RNA extraction and RT-PCR
Cells were lysed in Trizol reagent (Invitrogen), and RNA was extracted using ethanol precipitation.
RNA quality and quantity were confirmed using the NanoDrop ND-1000 spectrophotometer

Cell Proliferation
Approximately 100,000 cells per well (MCF7 and T47D) were seeded in triplicate on 6-well tissue

Immunofluorescence
Immunostaining were performed as previously reported (64). Samples were incubated overnight with SNAT2 primary antibodies and then incubated with fluorescently conjugated donkey secondary antibodies and DAPI for 1 h at 37°C followed by a further three 10-min washes in PBST.
All samples were imaged on a Zeiss LSM 780 confocal microscope. Magnification images were obtained using the x40 objective (1.4NA, Oil DIC, Plan-APOCHROMAT). Image J software was used to quantify SNAT2 intensity as previously described (65).

Immunohistochemistry
Immunohistochemistry was carried out as previously described (66 Secondary-only control staining was performed routinely. Necrosis was quantified histologically on haematoxylin-stained sections as previously described (66). Expression of CA9 and necrosis was quantified on whole sections by using ImageJ software (US National Institutes of Health, Bethesda, MD, USA). Expression of CD31 was quantified by using 15 random fields of slides at × 100 magnification.

Determination of glutamine consumption
Cells (1×104/well) in a 24 well plate were cultured for 24 hours in medium without phenol red, medium was collected, and cells were lysed with RIPA buffer (Sigma-Aldrich). Concentrations of glutamine in the medium and in the cell lysate were determined with the Glutamine Detection Assay Kit (Abcam, ab197011). A standard curve was determined for each experiment to calculate the concentration of glutamine in samples as per manufacture guideline. Glutamine levels were calculated and normalized to total protein levels. The glutamine level of normal culture medium was also measured, and the glutamine consumption was calculated as (glutamine in normal medium-glutamine in medium after culturing cells) and normalized to protein level.

Seahorse XF-24 metabolic flux analysis.
Oxygen consumption rates and extracellular acidification rate were measured at 37 °C using an All xenografts were also implanted with estrogen as seen above. Tumor growth was monitored three times per week measuring the length (L), width (W) and height (H) of each tumor using calipers. Volumes were calculated from the formula 1/6 × π × L × W × H.

Bioinformatics and statistical analysis
For heatmap and hierarchical clustering, unsupervised exploratory analysis RNA transcriptomic data k-mean was utilized to determine the optimal number of clusters.