Drosophila Lgl and its mammalian homologues, LLGL1 and LLGL2, are scaffolding proteins that regulate the establishment of apical–basal polarity in epithelial cells1,2. Whereas Lgl functions as a tumour suppressor in Drosophila1, the roles of mammalian LLGL1 and LLGL2 in cancer are unclear. The majority (about 75%) of breast cancers express oestrogen receptors (ERs)3, and patients with these tumours receive endocrine treatment4. However, the development of resistance to endocrine therapy and metastatic progression are leading causes of death for patients with ER+ disease4. Here we report that, unlike LLGL1, LLGL2 is overexpressed in ER+ breast cancer and promotes cell proliferation under nutrient stress. LLGL2 regulates cell surface levels of a leucine transporter, SLC7A5, by forming a trimeric complex with SLC7A5 and a regulator of membrane fusion, YKT6, to promote leucine uptake and cell proliferation. The oestrogen receptor targets LLGL2 expression. Resistance to endocrine treatment in breast cancer cells was associated with SLC7A5- and LLGL2-dependent adaption to nutrient stress. SLC7A5 was necessary and sufficient to confer resistance to tamoxifen treatment, identifying SLC7A5 as a potential therapeutic target for overcoming resistance to endocrine treatments in breast cancer. Thus, LLGL2 functions as a promoter of tumour growth and not as a tumour suppressor in ER+ breast cancer. Beyond breast cancer, adaptation to nutrient stress is critically important5, and our findings identify an unexpected role for LLGL2 in this process.
LLGL1 levels in a panel of breast cancer cell lines did not differ significantly from those in the non-transformed cell line, MCF-10A; however, LLGL2 levels were 3–5 times higher in cells expressing both ER and progesterone receptors (ER+/PR+ cells) than in other cells analysed (Fig. 1a, b, Extended Data Fig. 1a). Immunohistochemical analysis showed that LLGL2 was strongly expressed in 70% (24 of 34) of ER+ breast cancer tissue samples, but only 14% (5 of 34) of ER− breast cancer tissue samples (Extended Data Fig. 1b–d). The LLGL2 gene was amplified in 6% of patients with breast cancer and this amplification correlated with mRNA levels (Extended Data Fig. 1e). High levels of LLGL2 (but not LLGL1) mRNA expression correlated with poor patient survival in patients with ER+/PR+ breast cancer, but not in those with ER−/PR− breast cancer (Fig. 1c, Extended Data Fig. 1f). These observations identify an unexpected association between LLGL2 (but not LLGL1) and poor prognosis in ER+ breast cancer.
Overexpression of similar levels of LLGL1 or LLGL2 in MCF-7 or T47D breast cancer cell lines did not affect cell proliferation under normal culture conditions (Fig. 1d, Extended Data Fig. 2a, b). However, under serum-free conditions with medium supplemented with epidermal growth factor (EGF) and B27 supplement, overexpression of LLGL2, but not LLGL1, promoted cell proliferation in both adherent and non-adherent cultures (Fig. 1e, Extended Data Fig. 2c, d).
Next, we investigated how LLGL2 promotes cell proliferation. Knockdown of LLGL2 expression (LLGL2-KD) by RNA interference (RNAi) impaired the proliferation of both MCF-7 and T47D cells under adherent or non-adherent conditions (Fig. 1f, g, Extended Data Fig. 2e–h), and proliferation was rescued by re-expression of RNAi-resistant LLGL2 (Extended Data Fig. 2i–p). Knockdown of LLGL2 did not affect cell viability (Extended Data Fig. 2q), so LLGL2 is a regulator of cell proliferation and not cell survival under serum-free conditions. Furthermore, knockdown of LLGL2 substantially impaired in vivo growth, compared to control MCF-7 cells (Fig. 1h), when cells were orthotopically injected into the mammary fat pads of immunocompromised mice. Thus, LLGL2 regulates the proliferation of ER+ cells both in culture and in vivo.
To better understand the results under serum-free conditions, we used fetal bovine serum (FBS) dialysed with a 3,500-Dalton (Da) cut-off that retains growth factors (molecular mass 6,400–84,000 Da) but depletes amino acids and other low-molecular-mass nutrients6. Proliferation of LLGL2-KD cells was impaired in 10% dialysed FBS medium (Extended Data Fig. 3a), showing that knockdown of LLGL2 creates a dependency for low-molecular-mass nutrients in serum for efficient cell proliferation. Thus, we refer to our serum-free medium supplemented with B27 and EGF as a nutrient stress condition.
To investigate how LLGL2 promotes adaptation to nutrient stress, we analysed changes in cellular metabolites using targeted mass spectrometry7 in three biological replicates of LLGL2-KD and control MCF-7 cells in culture and in vivo. Among the top 50 metabolites that differed substantially between LLGL2-KD and control MCF-7 cells, essential amino acids were highly represented, being lower in LLGL2-KD cells than in control cells (Extended Data Fig. 3b). Among the nine essential amino acids, leucine, isoleucine and tryptophan were lower in LLGL2-KD cells than in control cells both in culture and in tumours in vivo (Fig. 1i, Extended Data Fig. 3c), identifying an unexpected relationship between LLGL2 and changes in intracellular levels of the essential amino acids Leu-Ile and Trp.
We tested whether supplementation with an excess of the limiting amino acid can rescue growth under nutrient stress. Some amino acid transporters need to function as antiporters for Gln in order to facilitate import of essential amino acids8,9, so Gln was included in all media. Excess Gln alone was not sufficient to rescue proliferation of cells (Fig. 1j). Only excess Leu rescued proliferation of LLGL2-KD cells, which suggests that LLGL2 supports cell proliferation by promoting uptake of Leu (Fig. 1j). Compared to LLGL2-KD cells, control cells were insensitive to excess Leu and Gln, ruling-out non-specific effects (Extended Data Fig. 3d). Overexpression of LLGL2 (LLGL2-OE) was sufficient to promote cell proliferation in LQ stress medium (containing one-tenth the normal concentration of Leu and Gln; Fig. 1k), which shows that LLGL2 overexpression is sufficient for cells to adapt to LQ stress. Furthermore, a tenfold excess of Leu and Gln potentiated the growth of LLGL2-OE but not control cells (Extended Data Fig. 3e), identifying a quantitative relationship between cellular LLGL2 levels and LQ-regulated cell proliferation.
LLGL2-KD cells were less effective at depleting extracellular leucine than parental or LLGL2-OE cells (Fig. 1l, Extended Data Fig. 3f), which shows that cellular levels of LLGL2 affect leucine consumption. LLGL2-OE or LLGL2-KD cells in 10× LQ medium consistently showed either an increasing trend or a marked increase in intracellular Leu; however, levels of other amino acids showed varying patterns, requiring further analysis (Extended Data Fig. 3g, h).
As there is no precedent for a cell polarity protein affecting uptake of Leu, and recognizing that LLGL2 is a scaffolding protein, we performed proximity-dependent biotinylation (BioID)10,11 analysis under polarized and physiological conditions. Like wild-type LLGL2, Flag-tagged biotin ligase (BirA*)-fused LLGL2 retained the ability to promote proliferation of MCF-7 cells under nutrient stress (Extended Data Fig. 4a, b). Notably, in addition to known LLGL2-binding proteins, we identified several solute carrier (SLC) family proteins as interacting with LLGL2 (Fig. 2a, Extended Data Fig. 4c). Among the five SLC proteins, SLC7A5 showed the strongest interaction according to the number of peptides detected and high probabilistic scoring of affinity purification-mass spectrometry data (SAINT) score12. High levels of SLC7A5 mRNA, but not of SLC1A5, SLC4A7, SLC38A1 or SLC7A2 mRNA, correlated with poor clinical outcome in patients with ER+/PR+ breast cancers (Fig. 2b, Extended Data Fig. 4d). SLC7A5, also referred to as LAT1 (L-type amino acid transporter), is the primary leucine transporter in cells and is overexpressed in multiple cancers including breast cancer13,14,15,16,17,18 (Extended Data Fig. 5a, b); however, the pathological relevance of SLC7A5 in ER+ breast cancer and its relationship to LLGL2 are unknown.
Co-immunoprecipitation analysis revealed that LLGL2 interacted with SLC7A5 and SLC1A5 under nutrient stress conditions (Fig. 2c, Extended Data Fig. 5c–e), consistent with the fact that SLC7A5 and SLC1A5 function as heterotrimers to regulate amino acid transport8. Furthermore, SLC7A5 and LLGL2 co-localized at cell junctions (Extended Data Fig. 5f), supporting the hypothesis that LLGL2 interacts with SLC7A5 to promote cell proliferation under nutrient stress.
Consistent with a role in amino acid transport, phosphorylation of the mTOR pathway target S6 kinase8 by amino acid stimulation was reduced by knockdown of LLGL2 compared to control MCF-7 cells (Fig. 2d), which shows that LLGL2 is a regulator of amino-acid-induced activation of mTOR in ER+ breast cancer cells.
Knockdown of SLC7A5 (SLC7A5-KD) in MCF-7 or T47D cells impaired cell proliferation in both adherent and non-adherent conditions, and expression of RNAi-resistant SLC7A5 rescued cell proliferation (Fig. 2e, f, Extended Data Fig. 5g–k). Orthotopically transplanted SLC7A5-KD cells did not form tumours, unlike control MCF-7 cells (Fig. 2g). In addition, parental MCF-7 tumours regressed markedly when treated with JPH203, an selective inhibitor of SLC7A519 (Fig. 2h). Treatment with JPH203 also caused a decrease in total body mass of the mice during the nine-day treatment period, with recovery to control levels upon cessation of treatment (Extended Data Fig. 5l). Furthermore, overexpression of SLC7A5 was sufficient to support cell proliferation under nutrient stress (Fig. 2i, j, Extended Data Fig. 5m, n). Thus, SLC7A5 is a key regulator of proliferation of ER+ tumour cells in culture and in vivo.
Although total protein levels of SLC7A5 were not affected in LLGL2-KD cells, SLC7A5 was lost from cell–cell junctions (Fig. 2k, l, Extended Data Fig. 6a–c). In addition, cell surface levels of SLC7A5, but not E-cadherin, were substantially lower in LLGL2-KD cells than in control or rescued cells, as determined by biotin labelling of surface proteins (Fig. 2m, Extended Data Fig. 6d, e). Immunohistochemical staining of breast cancer tissues confirmed membrane localization of SLC7A5 in LLGL2high ER+ tumours but not in LLGL2low ER− tumours (Extended Data Fig. 6f). Thus, LLGL2 is required for localization of SLC7A5 to cell junctions and membranes in ER+ breast cancer cells.
In contrast to LLGL2, total SLC7A5 expression did not vary with ER status in breast cancer cells (Extended Data Fig. 6g). Notably, growth of ER+ cells, but not ER− cells, was substantially impaired in response to treatment with 2-aminobicyclo-(2,2,1)-heptane-2-carboxylic acid (BCH)20, a selective SLC7A5 inhibitor that impairs leucine transport (Fig. 2n, Extended Data Fig. 6h). This is consistent with previous reports that ER−/PR−/HER2− cancer cells are not sensitive to Leu stress21,22. However, complete depletion of Leu inhibited the growth of ER− cells (Extended Data Fig. 6i), consistent with the fact that Leu is an essential amino acid. ER− breast cancer cells had eightfold lower levels of SLC7A5 at the cell surface (Fig. 2o, Extended Data Fig. 6j) compared to ER+ cells, and re-expression of LLGL2 in ER− cells did not restore the surface levels of SLC7A5 (Extended Data Fig. 6k, l) suggesting that the relationship between LLGL2 and cell surface levels of SLC7A5 is specific to ER+ cells. Thus, unlike ER+ cells, ER− cells have acquired the ability to grow in low levels of Leu and with low surface levels of SLC7A5.
Although the interaction between LLGL2 and SLC7A5, and cell surface SLC7A5, were detectable under nutrient-replete conditions, both were induced by nutrient stress (Fig. 3a, b, Extended Data Fig. 7a, b), which suggests that cells adapt to nutrient stress by increasing surface SLC7A5. The N-terminal domain of LLGL2 was sufficient to interact with SLC7A5 (Extended Data Fig. 7c, d). Notably, the LLGL2-Polybasic (LLGL-Pb) mutant23, which is defective in membrane localization, retained the ability to interact with SLC7A5 (Extended Data Fig. 7d) but failed to rescue growth of LLGL2-KD cells under nutrient stress (Fig. 3c, d, Extended Data Fig. 7e). Thus, LLGL2 interacts with SLC7A5 in the cytosol but this interaction is not sufficient to promote proliferation under nutrient stress.
To identify other members of the LLGL2–SLC7A5 complex, we investigated the role of YKT6, a soluble NSF attachment protein receptor (SNARE) protein that has been implicated in multiple membrane transport steps in the secretory pathway24 and was shown to interact with LLGL2 in our BioID analysis (Extended Data Fig. 4c). Endogenous YKT6 formed a trimeric complex with SLC7A5 and LLGL2 in response to nutrient stress (Fig. 3e, f, Extended Data Fig. 7f–j). Knockdown of YKT6 suppressed cell proliferation under nutrient stress and reduced surface levels of SLC7A5, phenocopying knockdown of LLGL2 (Fig. 3g–i, Extended Data Fig. 7k–q). Thus, YKT6 is part of the LLGL2–SLC7A5 pathway that regulates SLC7A5 surface levels and adaptation to nutrient stress.
In contrast to SLC7A5, YKT6 interacted with the C-terminal half of LLGL2 (Extended Data Fig. 7r) and the LLGL2-Pb mutant failed to interact with YKT6, showing that LLGL2 interacts with YTK6 at cell membranes. Knockdown of LLGL2 had no effect on membrane or cytosolic localization of YKT6; thus, unlike SLC7A5, localization of YKT6 is independent of LLGL2 (Extended Data Fig. 7s). Together, our results support a mechanism in which LLGL2 interacts with its cargo, SLC7A5, in the cytoplasm and transports it to the membrane, where it interacts with the SNARE protein YKT6 to facilitate membrane fusion and increase SLC7A5 levels at the cell surface.
Stimulation of MCF-7 and T47D cells with oestrogen (E2) induced an increase in expression of LLGL2 (but not LLGL1) that was similar to the increase in cyclin D1, a known target of ER signalling25 (Fig. 4a, Extended Data Fig. 8a, b). We investigated whether LLGL2 is an ER target and is involved in E2-induced cell proliferation. Both control and LLGL2-KD cells failed to proliferate in medium lacking oestrogenic activity (Fig. 4b, Extended Data Fig. 8c). E2-induced proliferation was substantially impaired in LLGL2-KD cells compared to control cells (Fig. 4b, Extended Data Fig. 8c), indicating that LLGL2 is required for E2-induced cell proliferation. E2 stimulation was sufficient to restore cell proliferation under LQ stress conditions (Fig. 4c, Extended Data Fig. 8d, e). Notably, knockdown of LLGL2 blocked the ability of E2 to rescue cell proliferation under nutrient stress (Fig. 4c, Extended Data Fig. 8d, e), demonstrating that E2-induced cell proliferation relies on LLGL2. Furthermore, LLGL2 overexpression enhanced E2-induced cell proliferation compared to control MCF-7 cells (Extended Data Fig. 8f), suggesting that there is a quantitative relationship between LLGL2 levels and E2-regulated cell proliferation.
We observed an ER-binding site (ERE) at intron 2 of LLGL2 that was coincident with an open chromatin region characterized by histone H3 acetylation (H3K27ac; Fig. 4d) in chromatin immunoprecipitation sequencing (ChIP–seq) data sets from E2-stimulated MCF-7 (Gene Expression Omnibus accession number GSM1534722) and E2/progesterone-treated T47D (GSM1669014) cells and from breast cancer samples (GSE32222)26 (Extended Data Fig. 9a), suggesting that LLGL2 may be a direct target of ER-mediated transcription. To test the specificity of LLGL2 in regulating E2-induced proliferation under nutrient stress conditions, we eliminated a 523-bp ERE in intron 2 of LLGL2 gene using CRISPR–Cas9 gene editing (Fig. 4e, Extended Data Fig. 9b, d). A 567-bp region in intron 1 of LLGL2 was deleted to serve as a non-specific control (Fig. 4e, Extended Data Fig. 8b, c). E2 stimulation failed to induce the expression of LLGL2 in cells lacking the ERE site in LLGL2, but had no effect in control cells (Fig. 4f), showing that the ERE in intron 2 is a critical mediator of E2-induced LLGL2 expression. Deletion of the ERE, and not the control region, abolished E2-induced cell proliferation (Fig. 4g), demonstrating that E2-mediated adaptation to nutrient stress requires expression of LLGL2.
High expression of LLGL2 or SLC7A5 correlated with poor survival in 799 patients with ER+ breast cancer who had been treated with tamoxifen (Fig. 4h), suggesting that LLGL2 and SLC7A5 might be involved in tamoxifen resistance. Total protein and cell surface levels of SLC7A5 were markedly upregulated in tamoxifen-resistant (TamR) MCF-7 cells without any significant change in the levels of LLGL2 (Fig. 4i, j). Notably, TamR cells were insensitive to LQ stress, unlike control cells (Fig. 4k). Knockdown of LLGL2 decreased cell surface levels of SLC7A5 (Fig. 4l, m, Extended Data Fig. 10a). Furthermore, knockdown of LLGL2 or SLC7A5 was sufficient to restore tamoxifen sensitivity to TamR cells under nutrient stress (Fig. 4l, n–p) and the SLC7A5 inhibitor BCH suppressed growth of TamR cells under nutrient stress (Extended Data Fig. 10b). Strikingly, overexpression of SLC7A5 in parental MCF-7 cells was sufficient to induce tamoxifen resistance (Fig. 4q, Extended Data Fig. 10c). Thus, LLGL2–SLC7A5 is a previously undescribed regulator of tamoxifen resistance (Fig. 4r).
We have described an unexpected relationship between the cell polarity protein LLGL2 and leucine transport and have shown that LLGL2 functions as a tumour promoter, and not a tumour suppressor, by helping cancer cells to overcome nutrition stress. In addition, we have shown that adaptation to nutrition stress and regulation of the LLGL2 amino acid transport pathway are regulated by E2, contribute to tamoxifen resistance and represent potential therapeutic targets for overcoming endocrine treatment resistance in breast cancer.
The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.
MCF-7 cells were cultured in Eagle’s minimum essential medium (EMEM) (ATCC 30-2003) supplemented with 10% FBS and 10 µg/ml insulin. T47D cells, AU565 cells, HCC1569 cells and BT474 cells were cultured in RPMI 1640 (GIBCO 11875119) supplemented with 10% FBS. HEK293T cells, MDA-MB-231 cells and Hs578T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; high glucose, GIBCO 11-965-118) supplemented with 10% FBS. Tamoxifen-resistant MCF-7 cells and the parental MCF-7 cells were kindly provided by R. Schiff (Baylor College of Medicine). Parental MCF-7 cells were cultured in RPMI 1640 supplemented with heat-inactivated FBS. Tamoxifen-resistant MCF-7 cells were cultured in phenol-red-free RPMI 1640 (GIBCO 11835030) supplemented with 10% charcoal-dextran-treated FBS and 100 nM tamoxifen (Sigma). Charcoal-dextran-treated FBS was prepared as described27. Dialysed FBS was prepared as described6. All cell lines were monitored by mycoplasma PCR testing and maintained in mycoplasma-free conditions.
Nutrient stress medium
DMEM/F12 containing 1× B27 supplement (minus vitamin A) and 20 ng/ml EGF. The absolute concentration of leucine in this medium was about 0.96 mM, as determined by quantitative mass spectrometry, whereas the regular growth medium with 10% FBS contains about 0.88 mM leucine/isoleucine (Extended Data Fig. 3i). Comparative amounts of metabolites in the culture media that were used in this study are shown as a heatmap in Extended Data Fig. 3j.
LQ stress medium
Leucine/glutamine-free DMEM/F12 was prepared from DMEM/F12 without amino acids (USBiological, #D9807-10) by adding amino acids (0.25 mM glycine, 0.05 mM l-alanine, 0.7 mM l-arginine hydrochloride, 0.05 mM l-asparagine H2O, 0.05 mM l-aspartic acid, 0.1 mM l-cysteine hydrochloride H2O, 0.1 mM l-cysteine 2HCl, 0.05 mM l-glutamic acid, 0.15 mM l-histidine hydrochloride H2O, 0.42 mM l-isoleucine, 0.5 mM l-lysine hydrochloride, 0.116 mM l-methionine, 0.215 mM l-phenylalanine, 0.15 mM l-proline, 0.25 mM l-serine, 0.45 mM l-threonine, 0.0442 mM l-tryptophan, 0.214 mM l-tyrosine disodium salt dihydrate, 0.45 mM l-valine), 17.5 mM d-glucose, 12.3 mM sodium bicarbonate, 15 mM HEPES, 0.5 mM sodium pyruvate, 0.015 mM hypoxanthine and 0.0015 mM thymidine. l-leucine and l-glutamine were added before use at concentrations of 0.45 mM and 2.5 mM, respectively.
10× LQ medium
DMEM/F12 with 20 ng/ml EGF and B27; Leu and Gln added at concentrations of 4.5 mM and 25 mM, respectively.
Generation of viral vector
To produce cells in which LLGL2 was specifically knocked down, we prepared pLKO.1 vectors (shGFP for control, LLGL2-KD and LLGL2-KD_2). The sequences are 5′-TACAACAGCCACAACGTCTAT-3′ for targeting GFP sequences, 5′-TAGGTGTCAGCAAAGTACAGG-3′ for LLGL2-KD (TRCN0000116433), 5′-AATCGCTTTGCAAGGAAAGGG-3′ for LLGL2-KD_2 (TRCN0000116434), 5′-CTAGATCCCAACTTCTCATTT-3′ for SLC7A5 (TRCN0000043009), 5′-GCCGAACTAGATGAGACCAAA-3′ for YKT6 (TRCN0000059763) and 5′-CGCATACGATGTGTCTTCCTT-3′ for YKT6 (TRCN0000059764). For overexpression of LLGL, each LLGL gene was sub-cloned into pMSCV vectors. For the rescue of LLGL2 expression, shRNA-resistant LLGL2 was prepared for each LLGL2 shRNA and sub-cloned into a pLJM1 vector (Addgene 19319) at the AgeI–EcoRI site. The LLGL2 cDNA sequences were modified at the shRNA targeting sites to 5′-TTATATTTCGCGGATACATAT-3′ (for LLGL2-KD; TRCN0000116433) or 5′- CCTTTCCCGTGTAAGGCTATC-3′ (for LLGL2-KD_2; TRCN0000116434). The SLC7A5 cDNA sequences are modified at shRNA targeting site to 5′-TTGGACCCAAATTTTAGCTT-3′. Each lentivirus was produced by transfection with pCMV delta R8.2 (Addgene 12263) and pCMV-VSV-G (Addgene 8454) using Lipofectamine 2000 (Invitrogen) in HEK293T cells.
CRISPR–Cas9-mediated deletion of ERE
A 2,000-bp genomic region surrounding the ER binding site in intron 2 of LLGL2 was used to find guide RNAs using the Cistrome Database (http://www.cistrome.org/Cistrome/Cistrome_Project.html). Two guide RNAs flanking a 523-bp genomic region of LLGL2 surrounding ERE were selected for deletion. To knock out the oestrogen binding site in LLGL2, the guide RNAs were cloned into pLentiCRISPR V2-Puromycin (Addgene 52961) and pLentiCRISPR V2-Blasticidin vectors. MCF-7 cells were infected with both lentiviruses and selected for resistance to puromycin and blasticidin. Resistant cell populations were pooled and used for this study. To confirm deletion of the 523-bp region containing ERE, genomic DNA from the resistant pool was used to amplify a genomic region of 1,596 bp flanking the ERE (Extended Data Fig. 9).
To control for non-specific effects of deleting intronic sequences in LLGL2, we used a 2,000-bp genomic region in intron 1 to find guide RNAs using the Cistrome Database. Two guide RNAs flanking a 567-bp genomic region in intron 1 of LLGL2 were selected for deletion. These guide RNAs were cloned into a pLentiCRISPR lentivirus vector and puromycin- and blasticidin-resistant populations were generated as outlined above. To confirm deletion of the 567-bp region in intron 1, genomic DNA from the resistant pool was used to amplify a genomic region of 1,615 bp in intron 1 (Extended Data Fig. 9).
Genomic PCR was performed to confirm the targeted genomic deletion in LLGL2. The sequences of each guide RNA and PCR primer are given in Extended Data Fig. 9. Genomic DNAs were extracted using PureLink Genomic DNA Mini Kit (Invitrogen K182001). We used 100 ng genomic DNA for the template. PCR was conducted using KOD Xtreme Hot Start DNA Polymerase (EMD Millipore 71975) according to the manufacturer’s protocol. Thermal steps were 94 °C for 2 min, and 35 cycles of 98 °C for 10 s, 57 °C for 30 s and 68 °C for 3 min. PCR products were subjected to agarose gel electrophoresis.
2D cell proliferation assay
One million cells were plated on a 6-cm dish. After 48 h culture, cells were trypsinized and the cell number was counted. This cell number was set as day 0. After rinsing with pre-warmed PBS, culture medium was replaced with serum-free medium (DMEM/F12 with B27 supplement (Life technologies 12587-010) and 20 ng/ml EGF (Peprotech AF-100-15)). On the indicated days, the cells were stained with 0.4% trypan-blue and counted. The results were confirmed with more than two independent experiments with three biological replicates.
Cells were plated into 96-well plates. After 48 h culture, the culture medium was replaced with serum-free medium (DMEM/F12 with B27 and 20 ng/ml EGF). On day 4, an MTT assay was conducted using Vybrant MTT Cell Proliferation Assay Kit (Invitrogen V13154) according to the manufacturer’s protocol.
Antibodies and reagents
Anti-LLGL1 (D2B5A), anti-E-cadherin (24E10, 3195), anti-LAT1 (5347), anti-ASCT2 (D7C12), anti-p-S6 kinase (9205), anti-p70-S6 kinase (9202) and anti-HA (C29F4) antibodies were purchased from Cell Signalling. Anti-LLGL2 (A-4), anti-LLGL2 (H-58), anti-actin (C-11), anti-ERK2 (C-14), anti-cyclin D1 (H-295), anti-v-SNARE YKT6P (E-2) and anti-v-SNARE YKT6P (FL-198) antibodies were purchased from Santa Cruz. The anti-Flag antibody (M2) was purchased from Sigma. Anti-SLC7A5 (BMP011) and anti-DDDDK-tag (M185) antibodies were purchased from MBL. Antibodies were used with 1/1,000 dilution for western blot and 1/100 dilution for immunofluorescence and immunohistochemistry. The SLC7A5 inhibitor BCH was purchased from Sigma (A7902) and used at 10 mM final concentration. The SLC7A5 inhibitor JPH203 was purchased from Sigma (SML1892).
Western blotting was conducted as previously described28. In brief, cells were lysed with lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, protease inhibitors and phosphoSTOP tablets (Roche)). The supernatant was subjected to SDS–PAGE and proteins were transferred onto PVDF membrane (Millipore #IPVH00010). After blocking with 1% BSA/TBS-0.1% Tween 20, the primary antibody was reacted overnight at 4 °C. HRP-conjugated secondary antibodies (GE healthcare) and HRP substrate (Pierce) were used for detection. Membranes were exposed to Hyperfilm (GE healthcare). Films were scanned into tiff format and signal intensity was quantified using ImageJ64 software.
Cells were cultured in DMEM/F12 medium supplemented with B27 minus vitamin A and 20 ng/ml EGF for 3 days. Cells were collected and lysed with lysis buffer (50 mM Tris pH7.4, 100 mM NaCl, 5 mM EDTA pH7.4, 1% Triton-X, 2 mM sodium orthovanadate, protease inhibitor and phosphoSTOP tablets (Roche)). Cell lysates were pre-cleared with 20 µl Protein G Sepharose beads (GE Healthcare, 17-0168-01) for 10 min and the cleared cell lysates were transferred to a new tube. Four microlitres of anti-DDDDK-tag (M185) antibody, anti-LAT1 antibody (Cell Signaling 5347) or anti-YKT6 antibody (SantaCruz sc-365732) was added to cell lysates and incubated for 1 h at 4 °C with rotation. LLGL2, SLC7A5 or YKT6 complexes were pulled down with 20 µl Protein G Sepharose beads (GE Healthcare, 17-0168-01) for 45 min. After being washed five times with lysis buffer, precipitants were eluted with 1× SDS sample buffer and subjected to western blotting. For investigation of the sites of LLGL2 interactions with SLC7A5 or YKT6, HEK293T cells were transfected with vectors using Lipofectamine2000 (Invitrogen). After 48 h, immunoprecipitation was performed as described above.
Amino acid stimulation
Cells were spread one day before experiments. Cells were rinsed more than three times with amino-acid-free RPMI (USBiological #R8999-04A) medium supplemented with 2 g/l sodium bicarbonate and 0.8 g/l sodium phosphate dibasic. Cells were cultured in amino-acid-free medium for 50 min at 37 °C. Then, cells were stimulated with complete RPMI medium for 10 min at 37 °C. The level of phosphorylation of S6K was examined by western blot.
MCF-7 cells or T47D cells were cultured in phenol-red-free MEM (GIBCO #51200-038) supplemented with 10% dextran-charcoal-treated FBS and l-glutamine for 3 days. Cells were collected using phenol-red-free 0.05% trypsin and plated onto a 6-cm dish in phenol-red-free MEM. On the next day, 100 nM E2 was added to the culture medium and cells were collected at the indicated times.
Sphere formation assay
The sphere formation assay was conducted as previously described29. Cells were trypsinized and single cell suspensions were prepared in culture medium. Seven thousand, seven hundred single cells were suspended in DMEM/F12 containing 1× B27 supplement (minus vitamin A), 20 ng/ml EGF and 0.5% methylcellulose (Fisher HSC001) and were spread onto a non-tissue-culture-coated 6-well plate (Falcon, 351146). The cells were cultured and the numbers of spheres of MCF-7 and T47D cells were counted on days 4 and 7, respectively. Each sphere was imaged using a phase contrast microscope (Leica) and the diameter was measured using ImageJ software. Spheres with diameters over 50 µm were counted.
For E2 stimulation, cells were cultured in phenol-red-free MEM (GIBCO 51200-038) supplemented with 10% dextran-charcoal-treated FBS and l-glutamine for 3 days and were collected with phenol-red-free 0.05% trypsin. Seven thousand, seven hundred single cells were suspended in phenol-red-free DMEM/F12 containing 10 nM E2, 1× B27 supplement (minus vitamin A), 20 ng/ml EGF and 0.5% methylcellulose (Fisher HSC011) and were spread onto a non-tissue-culture-coated 6-well plate (Falcon, 351146).
The effects of LLGL genes and SLC family genes on the survival of patients with breast cancer were analysed using Kaplan–Meier Plotter (http://kmplot.com/analysis/). To assess the effects of genes of interest, we analysed data from n = 577 patients with ER+/PR+ cancer and n = 298 patients with ER−/PR− cancer. We analysed n = 799 patients with ER+ cancer who had been treated with only tamoxifen.
MCF-7 cells (3 × 107 cells) were cultured on a 15-cm dish for 4 days in EMEM supplemented with 10% heat-inactivated FBS. One day before cell collection, biotin (final concentration of 50 µM) was added to the culture medium. Cells were collected and BioID was performed as described previously10,11.
Biotin–streptavidin affinity purification
The cell pellet was resuspended in 10 ml lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 0.1% SDS, 1:500 protease inhibitor cocktail (Sigma-Aldrich), 1:1,000 benzonase nuclease (Novagen)), incubated on an end-over-end rotator at 4 °C for 1 h, briefly sonicated to disrupt any visible aggregates, then centrifuged at 45,000g for 30 min at 4 °C. The supernatant was transferred to a fresh 15-ml conical tube, 30 µl packed, pre-equilibrated streptavidin-sepharose beads (GE) were added, and the mixture was incubated for 3 h at 4 °C with end-over-end rotation. Beads were pelleted by centrifugation at 2,000 r.p.m. for 2 min and transferred with 1 ml of lysis buffer to a fresh Eppendorf tube. Beads were washed once with 1 ml lysis buffer and twice with 1 ml 50 mM ammonium bicarbonate (pH 8.3). Beads were transferred in ammonium bicarbonate to a fresh centrifuge tube, and washed two more times with 1 ml ammonium bicarbonate buffer. Tryptic digestion was performed by incubating the beads with 1 µg MS grade TPCK trypsin (Promega, Madison, WI) dissolved in 200 µl 50 mM ammonium bicarbonate (pH 8.3) overnight at 37 °C. The following morning, an additional 0.5 µg trypsin was added, and the beads incubated for 2 h at 37 °C. Beads were pelleted by centrifugation at 2,000g for 2 min, and the supernatant was transferred to a fresh Eppendorf tube. Beads were washed twice with 150 µl 50 mM ammonium bicarbonate, and these washes were pooled with the first eluate. The sample was lyophilized, and resuspended in buffer A (0.1% formic acid). One-fifth of the sample was analysed per mass spectrometry (MS) run.
Mass spectrometry analysis
High-performance liquid chromatography (HPLC) was conducted using a 2-cm pre-column (Acclaim PepMap 50 mm × 100 μm inner diameter (ID)), and 50-cm analytical column (Acclaim PepMap, 500 mm × 75 μm ID; C18; 2 μm; 100 Å, Thermo Fisher Scientific, Waltham, MA), running a 120-min reversed-phase buffer gradient at 225 nl/min on a Proxeon EASY-nLC 1000 pump in-line with a Thermo Q-Exactive HF quadrupole-Orbitrap mass spectrometer. A parent ion scan was performed using a resolving power of 60,000, then up to the twenty most intense peaks were selected for MS/MS (minimum ion count of 1,000 for activation), using higher-energy collision-induced dissociation (HCD) fragmentation. Dynamic exclusion was activated such that MS/MS of the same m/z (within a range of 10 ppm; exclusion list size, 500) detected twice within 5 s were excluded from analysis for 15 s. For protein identification, Thermo .RAW files were converted to the .mzXML format using Proteowizard30, then searched using X!Tandem31 and Comet32 against the Human RefSeq Version 45 database (containing 36,113 entries). Search parameters specified a parent ion mass tolerance of 10 ppm, and an MS/MS fragment ion tolerance of 0.4 Da, with up to two missed cleavages allowed for trypsin. Variable modifications of +16@M and W, +32@M and W, +42@N terminus, and +1@N and Q were allowed. Proteins identified with an iProphet cut-off of 0.9 (corresponding to ≤1% false discovery rate, FDR) and at least two unique peptides were analysed with SAINT Express v.3.3. Sixteen control runs (four each from MCF10A, MGH7, HEK293 and HeLa cells expressing no bait or the FlagBirA* epitope tag) were collapsed to the four highest spectral counts for each prey, and compared to the two technical replicates of each bait analysis. High-confidence interactors were defined as those with Bayes false discovery rate (BFDR) ≤ 0.01.
Biotin-labelling surface proteins and pulldown assay
Cells were rinsed with ice-cold PBS and surface proteins were labelled with 400 µM EZ-link-sulfo-NHS-SS-Biotin/PBS for 30 min at 4 °C with gentle rocking. The reaction was quenched with 2 ml 150 mM glycine. After two rinses with ice-cold PBS, cells were lysed in lysis buffer (50 mM Tris (pH 7.4), 100 mM NaCl, 5 mM EDTA (pH 7.4), 1% Triton-X-100, 5 mM NaF, plus protease inhibitors and phosphoSTOP tablets (Roche)). Fifty micrograms of protein lysate was incubated with 20 µl Streptavidin Sepharose High Performance Beads (GE Healthcare) and rotated for 1.5 h at 4 °C. Streptavidin beads were washed three times with lysis buffer and proteins were eluted with 1× sample buffer.
Orthotropic tumour xenograft
Ten-week-old female mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) were purchased from Jackson Laboratory. Ninety-day slow-release oestrogen pellets were subcutaneously implanted into 10-week-old mice two days before cell injection (0.72 mg, Innovative Research of America). Cell suspension (1 × 105 cells) in 50 µl DMEM/F12 with 20 ng/ml EGF and 1× B27 minus vitamin A were mixed with 50 µl of matrigel. One hundred microlitres of cell mixture was injected into the fourth mammary fat pad using 27 G 1/2 inch 1-ml syringe as described33. The tumour size was measured every 5 days from day 15 (for LLGL2-KD cells) or every 3–4 days from day 14 (for SLC7A5-KD cells). For JPH203 treatment, 50 mg/kg body weight of JPH203 was intraperitoneally injected on the indicated days for 2 weeks. Tumour volumes were estimated with the formula: volume = (2a × b)/2, where a is the shortest and b is the longest tumour measurement in millimetres. The two-way ANOVA and Tukey’s multiple comparison tests were conducted using Prism 7.0 software. All mouse experiments were approved by the BIDMC institutional animal care and use committee (IACUC) (protocol 017-2016), and were carried out in accordance with the ‘Guide for the Care and Use of Laboratory Animals’. The approved maximum tumour size was 10% of weight or 2 cm in diameter (single or cumulative diamater if there was more than one tumour). All relevant procedures were compliant with ethical regulations.
Tissue array slides, in which ER status was qualified, were purchased from USBiomax (BC081120b). Slides were baked at 60 °C for 30 min. After deparaffinizing, antigen retrieval was conducted for 15 min using a pressure boiler in 10 mM citrate buffer containing 0.05% Tween-20 (pH 6.0) for the anti-SLC7A5 antibody and 10 mM citrate buffer for the anti-LLGL2 antibody. After cooling down to room temperature, endogenous HRP was inactivated with 3% H2O2 for 10 min. For blocking we used the VECTASTAIN UNIVERSAL Elite ABC KIT (Vector; PK-6200) and the ImmPACT DAB Peroxidase Substrate Kit (Vector; SK-41050) according to the manufacturers’ protocols.
ChIP–seq data for ESR1 were analysed from the Cistrome database (http://www.cistrome.org/Cistrome/Cistrome_Project.html) as previously described34.
For in vitro cultured cells, we changed the medium 2 h before metabolite extraction. Polar metabolites were extracted with 80% methanol as previously described7. For in vivo tumour samples, tumours were homogenized with pre-cooled 80% methanol in HPLC-grade water (1 ml 80% methanol per 100 mg sample) in dry ice. Homogenate (200 μl) was dried using SpeedVac. Metabolite extracts were suspended in 20 μl LC-MS grade water and 5 μl was analysed by targeted liquid chromatography with tandem MS (LC–MS/MS) via selected reaction monitoring (SRM) for 300 targets with positive/negative ion polarity switching using a 5500 QTRAP hybrid triple quadrupole mass spectrometer (A/B SCIEX) with amide HILIC chromatography (Waters). SRM peaks were integrated using MultiQuant 2.1 software (A/B SCIEX) and statistical analysis and generation of a heatmap were performed using MetaboAnalyst 3.0 software.
Measurement of Leu–Ile concentration
Concentrations of Leu–Ile in culture medium were determined by LC–MS/MS and comparison to a known quantity of spiked stable isotope-labelled Leu. In brief, 250 µl of each culture medium was mixed with l-leucine (13C6, 99% purity; Cambridge Isotope Laboratories) at 4.5 mM final concentration. The metabolites in the mixture were extracted with 1 ml cooled methanol (HPLC grade) overnight at −80 °C. After centrifugation at 14,000g for 10 min, the supernatant was transferred to a new 1.5-ml microcentrifuge tube and dried with SpeedVac. Labelled Leu was monitored with a Q1/Q3 SRM transition of 138.05/91.0 to account for six labelled carbons compared to the unlabelled Leu SRM transition of 132.1/86.0.
All statistical analysis were conducted using Prism 7.0 (GraphPad Software). Statistical significances are shown as NS (not significant; P>0.05), *P < 0.05, **P < 0.001, ***P < 0.0001. The numbers of experiments are noted in figure legends. The sample size used in each experiment was not predetermined or formally justified for statistical power. To assess the statistical significance of a difference between two treatments, we used two-tailed Student’s t-tests. To assess the statistical significance of differences between more than two treatments, we used one-way or two-way ANOVA with Tukey’s multiple comparison test or Bonferroni’s multiple comparison test.
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.
The Kaplan–Meier plot data that support the findings of this study are available in Kaplan–Meier Plotter (http://kmplot.com/analysis/) with the identifier 10.18632/oncotarget.1033735. The ChIP–seq data GSM153472236 and GSM166901437 that support the findings of this study are available from Cistrome Data Browser (http://cistrome.org/db/#/). The ChIP–seq data from breast cancer patients (GSE32222) have been previously published26. Source data are provided. All other relevant data are available from the corresponding author on reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank members of the Muthuswamy laboratory for discussions, R. Schiff for TamR cells, T. Xiao for advice on CRISPR–Cas9, J. Zoeller for advice on mouse experiments, F. Au-Yeung for assistance with BioID experiments, and M. Yuan for assistance with mass spectrometry. The work was supported by funding from the National Institutes of Health (NIH) (grant 5P01CA120964; J.M.A.), a long term postdoctoral fellowship (LT000091/2014) from the Human Frontier Science Program and research funds from the Yamagata prefectural government and the City of Tsuruoka (Y.S.), and the Breast Cancer Research Foundation (S.K.M.).
Nature thanks Jason Carroll, Zach Schafer and the other anonymous reviewer(s) for their contribution to the peer review of this work.