Amino acid transporter LAT1 in tumor-associated vascular endothelium promotes angiogenesis by regulating cell proliferation and VEGF-A-dependent mTORC1 activation

Tumor angiogenesis is regarded as a rational anti-cancer target. The efficacy and indications of anti-angiogenic therapies in clinical practice, however, are relatively limited. Therefore, there still exists a demand for revealing the distinct characteristics of tumor endothelium that is crucial for the pathological angiogenesis. L-type amino acid transporter 1 (LAT1) is well known to be highly and broadly upregulated in tumor cells to support their growth and proliferation. In this study, we aimed to establish the upregulation of LAT1 as a novel general characteristic of tumor-associated endothelial cells as well, and to explore the functional relevance in tumor angiogenesis. Expression of LAT1 in tumor-associated endothelial cells was immunohistologically investigated in human pancreatic ductal adenocarcinoma (PDA) and xenograft- and syngeneic mouse tumor models. The effects of pharmacological and genetic ablation of endothelial LAT1 were examined in aortic ring assay, Matrigel plug assay, and mouse tumor models. The effects of LAT1 inhibitors and gene knockdown on cell proliferation, regulation of translation, as well as on the VEGF-A-dependent angiogenic processes and intracellular signaling were investigated in in vitro by using human umbilical vein endothelial cells. LAT1 was highly expressed in vascular endothelial cells of human PDA but not in normal pancreas. Similarly, high endothelial LAT1 expression was observed in mouse tumor models. The angiogenesis in ex/in vivo assays was suppressed by abrogating the function or expression of LAT1. Tumor growth in mice was significantly impaired through the inhibition of angiogenesis by targeting endothelial LAT1. LAT1-mediated amino acid transport was fundamental to support endothelial cell proliferation and translation initiation in vitro. Furthermore, LAT1 was required for the VEGF-A-dependent migration, invasion, tube formation, and activation of mTORC1, suggesting a novel cross-talk between pro-angiogenic signaling and nutrient-sensing in endothelial cells. These results demonstrate that the endothelial LAT1 is a novel key player in tumor angiogenesis, which regulates proliferation, translation, and pro-angiogenic VEGF-A signaling. This study furthermore indicates a new insight into the dual functioning of LAT1 in tumor progression both in tumor cells and stromal endothelium. Therapeutic inhibition of LAT1 may offer an ideal option to potentiate anti-angiogenic therapies.


Background
Therapeutic intervention in tumor angiogenesis is one of rational strategies for anti-cancer treatment. Various agents including neutralizing antibodies and decoy receptors for proangiogenic factors, as well as antibodies and inhibitors for the receptor tyrosine kinases (RTKs), have been developed to target angiogenic signaling pathways in endothelial cells. Their efficacy and indications in clinical practice are, however, relatively limited [1,2]. The redundancy in proangiogenic growth factor signaling with compensatory functions is one of the mechanisms accounting for the insufficient responsiveness and resistance to anti-angiogenic therapy [1,2]. It was reported that treatment of rectal cancer patients with bevacizumab, an anti-VEGF antibody, increased the PlGF in plasma [3]. FGF-2 and PlGF were increased in glioblastoma multiforme patients treated with cediranib, a pan-VEGF receptor tyrosine kinase inhibitor [4,5]. Similar upregulation of pro-angiogenic factors was also observed in mouse models of pancreatic islet tumor treated with anti-VEGFR2 antibody, where the expression of Ang-1, Ephrin-A1, Ephrin-A2, FGF-1, and FGF-2 was increased [6,7]. The resultant tumor growth suppression was only transient with modest prolongation of survival [6,7]. These results clearly indicate that the inhibition of a specific pro-angiogenic signaling pathway per se in endothelial cells is not sufficient to control the aberrant angiogenic activity in tumor.
To improve the clinical benefits of anti-angiogenic therapy, it is fundamental to understand the molecular signature of tumor-associated endothelium involved in the pathological blood vessel formation. L-type amino acid transporter 1 (LAT1) forms heterodimeric complex with its ancillary protein 4F2hc, and preferentially transports most of the essential amino acids [8,9]. LAT1 is known to be upregulated in a wide spectrum of primary tumors and metastatic lesions from over 20 tissue/organ origins [10][11][12]. Furthermore, correlations between the LAT1 expression with poor prognosis have been indicated in various tumors including, but not limited to, triple negative breast cancer [13], highly proliferative ER + subtype of breast cancer [14], bladder cancer [15], lung adenocarcinoma [16], lung neuroendocrine tumor [17], pancreatic ductal adenocarcinoma [18,19], and biliary tract cancer [20]. LAT1 in cancer cells has, thus, been recognized as an emerging molecular target for anti-tumor therapy. Several LAT1selective inhibitors have been synthesized [21][22][23], including JPH203 that showed prominent anti-tumor effects in preclinical animal models [21,[24][25][26][27]. The first-in-human phase I clinical trial was recently conducted in patients with advanced solid tumors, and reported that JPH203 appeared to be well-tolerated and to provide promising activity against biliary tract cancer [28].
Besides its well-recognized function in tumor cells, a yet unclarified role of LAT1 in tumor biology has been its implication to endothelial cell functions in tumors. An elevated expression of LAT1 in tumor-associated microvasculatures was reported in N-butyl-N-(4-hydroxybutyl) nitrosamineinduced rat bladder carcinoma model [29]. A clinicopathological study on human glioma showed LAT1 expression in both vascular endothelial cells and tumor cells, demonstrating significant correlations of LAT1 expression with the pathological grade and the intratumoral microvessel density [30]. These observations prompted us to hypothesize that LAT1 mediates amino acid supply not only to tumor cells, but also to tumor-associated endothelial cells, thereby promoting cellular functions related to angiogenesis. Here, we demonstrate the LAT1 expression is upregulated in tumorassociated blood vessels but not in the blood vessels of normal tissues in general. Functional relevance of endothelial LAT1 in tumor angiogenesis was investigated, pursuing the possibility of obtaining anti-angiogenic effects by targeting endothelial LAT1.

Antibody production
A GST-fused recombinant protein of mouse LAT1 Nterminal 53 amino acids was expressed in E.coli BL21(DE3), and purified by Glutathione Sepharose 4B (GE Healthcare) affinity column chromatography. For rabbit antibody production (anti-mLAT1(R) antibody), a New Zealand White rabbit was intramuscularly immunized with the purified recombinant protein (200 μg in Freund's complete adjuvant for the initial injection, followed by three times injection of 200 μg in incomplete Freund's adjuvant with 2-week intervals). For chicken antibody production (anti-mLAT1(C) antibody), a White Leghorn chicken was immunized with the purified recombinant protein (200 μg in Freund's complete adjuvant for the initial injection, followed by four times injection of 100 μg in incomplete Freund's adjuvant with 2-week intervals). One week after the final injection, antisera were collected, passed through a GST-coupled Affi-Gel 10 column (Bio-Rad) for absorption of anti-GST antibody, and then subjected to purification by antigen-coupled Affi-Gel 10 column chromatography.

Effect of VEGF-A and FGF-2 stimulation on LAT1 expression in HUVECs
HUVECs were seeded in collagen-coated 6 cm dish (1.0 × 10 4 cells/dish). Two days later, cells were starved for VEGF-A and FGF-2 for 6 h, and then stimulated with either VEGF-A or FGF-2 alone, or in combination (10 ng/mL each). Total RNA and cell lysate were prepared and subjected to real-time PCR and western blotting, respectively.

Wound healing assay
HUVECs were seeded at 2.8 × 10 4 cells in 70 μL of EGM-2 medium/well in 2-well silicone culture insert (ibidi GmbH) settled in 24-well plate, and incubated for 18 h. After starvation for serum and growth factors in EBM-2B medium (EBM-2 medium supplemented with 0.1% BSA) for 6 h, cell migration was initiated by removing the inserts and adding 1 mL/well of EBM-2B medium containing 10 ng/mL VEGF-A. DIC images were acquired immediately after removing the inserts to locate the initial edges of cell-free gaps. Cells were incubated for 12 h for migration, fixed by 4% paraformaldehyde (PFA), stained with crystal violet, and subjected to image acquisition. Bright field images were acquired using an inverted microscope (DMi1, Leica Microsystems). The number of migrated cells were counted by using Cell Counter plugin for ImageJ software (NIH).

Invasion assay
HUVECs starved for serum and growth factors were seeded at 5 × 10 4 cells in 250 μL of EBM-2B in the upper chamber of BioCoat Angiogenesis system: Endothelial Cell Invasion (Corning). Lower chamber was filled with 750 μL of EBM-2B medium containing 10 ng/mL VEGF-A. Cells were incubated for 12 h for invasion, stained by 0.5 μM Calcein-AM in HBSS for 1 h, and subjected to image acquisition from the bottom of chamber by a bright-field/fluorescence microscope (BZ-9000, Keyence). Area covered with invaded cells were calculated from binarized images by using ImageJ software.

Tube formation assay
HUVECs starved for serum and growth factors were seeded at 1.0 × 10 4 cells/well in 96-well plate coated with 50 μL/well of growth factor-reduced Matrigel. In each well, 100 μL of EBM-2B medium containing VEGF (10 ng/mL) was added. After incubation for 8 h, cells were stained with 3 μM Calcein-AM at 37°C for 20 min, and subjected to image acquisition using a fluorescent microscopy (EVOS FL, Thermo Fisher Scientific). Total branching length was quantified by ImageJ software with Angiogenesis Analyzer plugin (http://image.bio.methods. free.fr/ImageJ/?Angiogenesis-Analyzer-for-ImageJ).

Cell proliferation assay
HUVECs were seeded in collagen coated 96-well plates (1.0 × 10 3 cells/well) in EGM-2 medium. BCH or JPH203 was added on the next day (Day 0). For LAT1 knockdown, cells were seeded at 48 h after siRNA transfection (Day 0). Cell proliferation was measured every 24 h for 3 days by CCK-8 kit (Dojindo).

Aortic ring assay
Aortic ring assay was performed as described previously [32]. Serum-starved aortic rings from C57BL/6 J female mice were embedded in growth factor-reduced Matrigel, and cultured in the presence of 2.5% fetal bovine serum and 30 ng/mL VEGF. When indicated, BCH or JPH203 dihydrochloride was added into the medium. Doxycycline (DOX, 100 ng/mL) was added into the medium throughout the assays using DOX-inducible conditional LAT1knockout mice. The numbers of sprouting microvessels were manually counted at 5 days after embedding.

Construction of animal tumor models and quantification of blood vessels
Human pancreatic cancer MIA PaCa-2 cells (JCRB0070, JCRB) and lung cancer H520 cells (HTB-182, ATCC) were grown in DMEM (SIGMA-Aldrich) supplemented with 10% FBS (Gibco), and 100 units/mL penicillin -100 μg/mL streptomycin (Nacalai Tesque). Before inoculation, cells were suspended in filtrated PBS, and mixed with growth factor-reduced Matrigel in a 1:1 volume ratio to give a final concentration of 2.5 × 10 7 cells/mL. The cell suspension was subcutaneously injected into the lower flank of 6-week-old BALB/c-nu/nu female mice (5.0 × 10 6 cells, 0.2 mL/animal). When indicated, the size of tumor was measured by caliper to calculate volumes using the formula: Tumor volume (mm 3 ) = (length × width 2 )/2, where length and width are the longest and shortest dimensions of the tumor, respectively. Seven days later, when the tumor volume reached to 100~250 mm 3 , mice were divided into two groups (n = 5 for each group), and treated everyday with either JPH203-SBECD in saline (25 mg/kg/day, i.v.) or equivalent amount of placebo control. After 14 days of consecutive injection, tumors were excised and subjected to immunofluorescence analysis against CD34. From the acquired immunofluorescence images, binary images were generated by manual thresholding and used for the quantification of blood vessel density by "Analyze Particles" plugins of ImageJ software. Images were acquired from at least five randomly selected fields on each section, and 10 sections were analyzed for each tumor (50 100 pictures per tumor). Averaged numbers of blood vessels per mm 2 tissue area for each tumor were used for statistical analysis.
An orthotopic syngeneic tumor model was constructed by subcutaneous inoculation of B16-F10 mouse melanoma cells (CRL-6475, ATCC) into Lat1 fl/fl /Tek-Cre or control Lat1 fl/fl mice. B16-F10 cell suspension in PBS were mixed with growth factor-reduced Matrigel in a 1:1 volume ratio to give a final concentration of 2.5 × 10 6 cells/mL. The cell suspension was subcutaneously injected into the lower flank of 6-to 8-week-old mice (0.5 × 10 6 cells, 0.2 mL/animal). Tumor volumes were calculated every day as described above. Ten days after the implantation, the tumors were collected for the quantification of blood vessel formation using paraffin sections. To label intratumoral blood vessels, 100 μL of FITC-Dextran (2,000,000 MW, Invitrogen) solution (5 mg/mL in saline) was intravenously injected 35 min before the collection of tumors. Entire sections were analyzed to quantify the blood vessel area. From the acquired fluorescence images of FITC-dextran, binary images were generated by manual thresholding and used for the quantification of blood vessel area by "Analyze Particles" plugins of ImageJ software. Eight sections were analyzed for each tumor to calculate averaged blood vessel areas (μm 2 /mm 2 tissue area), and used for statistical analysis. Experiments were performed with n = 4 for each group (Lat1 fl/fl /Tek-Cre mice and control Lat1 fl/fl mice).

Transgenic mice
Lat1 fl mice harboring floxed Lat1 gene for conditional knockout were generated by Unitech Co., Ltd. Targeting construct was designed to excise exon 3 of Lat1 gene (Supplementary Figure 2). A 1.2 kb-genomic region containing exon 3 was replaced by the corresponding genomic sequence flanked with a pair of loxP sequences. An FRT site-flanked neomycin resistance gene cassette was also inserted into the downstream of exon 3. Long and short arms (5.4 kb and 2.3 kb, respectively) were added for homologous recombination. All the genomic sequences were amplified from BAC clone RP23-46D12. A diphtheria toxin A-fragment (DTA) under thymidine kinase promoter was used for negative selection. The targeting construct was electroporated into mouse Bruce-4 ES cells derived from C57BL/6 J. After selection with 200 μg/ml of G418, successfully targeted ES clones were screened by PCR. Homologous recombination was further confirmed by Southern blot analysis using two external probes (5′-and 3′ probes against SpeI-digested genomic DNA) and an internal probe (Neo probe against EcoRV-digested genomic DNA). Positive ES clones were then injected into Balb/c blastocysts to obtain chimeric mice. Germ line transmission was established by crossing the chimeric mice with C57BL/6 J mice, and obtained heterozygous founder mice were further crossed with CAG-FLP mice expressing Flprecombinase under the control of the CAG-promoter, to excise the FRT site-flanked neomycin resistance cassette. After confirming the removal of neomycin resistance gene cassette by PCR, the resultant Lat1 fl mice were maintained with C57BL/6 J genetic background.
Whole-mount immunofluorescence of aortic rings was performed as described previously [38] with minor modifications. Serum-starved aortic rings were embedded into Matrigel on 35 mm glass-bottom dish. After incubation for 3 days, aortic rings were fixed in 4% PFA at 4°C overnight, washed twice in PBS, and permeabilized with 0.5% Triton X-100/PBS for 1 h. Blocking was performed for 2 h in PBS containing 0.5% Triton X-100 and 1% BSA at room temperature. Incubation with primary-[anti-mLAT1 (C), and anti-Claudin5 (sc-28670, Santa Cruz Biotechnology)] and secondary antibodies [Alexa Fluor 488-conjugated donkey anti-chicken IgY, and Alexa Fluor 568-conjugated goat anti-rabbit IgG] were performed at 4°C overnight, followed by washing with PBS for 30 min for three times. DAPI was used for nucleus staining. Stained samples were observed under an inverted confocal laser scanning microscope (FV-1000; Olympus).

Real-time PCR
Total RNA from HUVECs and that from mouse aorta were extracted using Isogen II (Nippon Gene) and Agencourt RNAdvance Tissue Kit (Beckman Coulter), respectively. Quantitative real-time PCR was performed as described previously [31].

Western blotting
Total cell lysates of HUVECs were prepared as described previously [39]. Crude membrane fractions were prepared as previously [40], and solubilized on ice for 30 min with 1% NP-40. After mixing with Laemmli buffer, SDS-polyacrylamide gel electrophoresis and western blot analysis were performed [39]. The antibody-treated PVDF membrane was developed with ECL Prime Western Blotting Detection System and imaged by Amersham Imager 680 (GE Healthcare).

LAT1 is expressed in tumor-associated endothelial cells
Upregulation of LAT1 has been reported in cancers of various tissue origins including pancreatic ductal adenocarcinoma (PDA) [18,19]. Consistently, a high expression of LAT1 was detected in cancer cells of PDA tissue in our immunohistochemistry (Fig. 1a). Intriguingly, we also noticed a significant expression of LAT1 in the stromal cells that are positive for an endothelial cell marker CD31 (Fig. 1a). Endothelial cells in normal pancreatic tissue were, in contrast, mostly negative for LAT1 staining. The colocalization of LAT1 and CD31 in the tumor-associated endothelial cells of PDA tissue was further demonstrated by immunofluorescence (Fig. 1b). This observation was confirmed in a larger number of samples on tissue microarray by immunohistochemistry (Fig. 1c). Only a minor fraction (25.0%) of normal pancreatic tissues exhibited positive LAT1 staining in endothelial cells: 4 showed low-to-moderate, and 1 showed strong staining among 20 analyzed tissue spots. In contrast, a majority of PDA tissues (81.4%) exhibited positive LAT1 expression in endothelial cells: 26 showed low-to-moderate, and 35 showed strong staining among 75 analyzed tissue spots.
The expression of LAT1 in tumor-associated blood vessels was further examined in human cancer-cell xenograft tumor models in athymic nude mice. To detect mouse LAT1 in blood vessels surrounded by cancer cells highly expressing human LAT1, we generated mouse LAT1-specific antibodies (Supplementary Figure 1). Using the obtained antibody, mouse LAT1 was detected in CD34-positive endothelial cells in the tumors of pancreatic cancer MIA PaCa-2 cells (Fig. 1d) and of non-small cell lung cancer H520 cells (Fig. 1e). In contrast, no clear LAT1 staining was detected in the blood vessels of normal tissues except brain capillaries, where the expression of LAT1 has been reported previously [41,42] (Supplementary Figure 4). The expression of LAT1 in the endothelial cells of tumor-associated blood vessels was, thus, recapitulated in the xenograft tumor models of distinct tissue origins.

Endothelial LAT1 contributes to angiogenesis in ex-and in vivo assays
The results above prompted us to investigate the functional relevance of endothelial LAT1 in angiogenesis. We first performed aortic ring assay, in which endothelial microvessel-like sprouts grew out from the slice of aorta in the Matrigel. Expression of LAT1 in the endothelial sprouts were confirmed by whole-mount immunofluorescence with an endothelial marker Claudin-5 (Fig. 2a). To examine the effects of pharmacological inhibition of LAT1, Matrigel-embedded aortic rings were cultured in the presence of JPH203 or BCH. Both of the compounds suppressed the outgrowth of endothelial sprouts to 10~20% of control ( Fig. 2b and Supplementary Figure 5A).
We then performed Matrigel plug assay, in which Matrigel was subcutaneously injected into mice and analyzed for blood vessel formation. Immunofluorescence revealed that LAT1 is expressed in the CD31-positive endothelial cells within the Matrigel plugs (Fig. 2c). Proangiogenic growth factors, VEGF-A and FGF-2, mixed with Matrigel induced vascularization, as demonstrated by the higher fluorescence of intravenously injected FITC-dextran. LAT1 inhibition by BCH and JPH203 mixed in Matrigel reduced the fluorescence of the plugs, indicating a decreased angiogenesis (Fig. 2d and Supplementary Figure 5B).
To obtain further evidence for the roles of endothelial LAT1 in angiogenesis, we generated conditional knockout mice harboring exon 3-floxed Lat1 allele (Supplementary Figure 2). The deletion of exon 3 in mouse Lat1 gene results in an early frameshift and creates a premature stop codon. As a consequence, an N-terminal fragment of LAT1 composed of 227 amino acids (containing TM1-TM5) followed by two unrelated amino acids (−Thr-Ile) would be potentially expressed. The corresponding LAT1 fragment (LAT1-Δex3) co-expressed with 4F2hc did not exhibit any L-[ 14 C] leucine transport function in Xenopus oocytes (Supplementary Figure 3A). The protein expression of LAT1-Δex3 in oocyte membrane fraction was markedly lower than that of wild type LAT1. LAT1-Δex3 did not form a heterodimer with 4F2hc (Supplementary Figure 3B).
The doxycycline (Dox)-inducible conditional knockout Lat1 fl/fl /rtTA3/TetO-Cre mice were generated from the Lat1 fl mice (Supplementary Figure 6A). In the aortic rings isolated from the Lat1 fl/fl /rtTA3/TetO-Cre mice, DOXtreatment decreased the LAT1 mRNA expression to4 0% of the control without DOX-treatment on the day of embedding (Day 0, after overnight DOX-treatment with serum starvation), and to~20% of the control 3 days after embedding (Day 3) (Fig. 3a). We found that the outgrowth of endothelial sprouts was suppressed by DOX-treatment in the aortic rings prepared from Lat1 fl/fl /rtTA3/TetO-Cre mice, whereas not in those from control Lat1 fl/fl /rtTA3 and Lat1 fl/fl /TetO-Cre mice (Fig. 3b). To exclude the contribution of non-endothelial cells, we used endothelial cell-specific knockout mice (Supplementary Figure 6B). Depletion of endothelial LAT1 protein in Lat1 fl/fl /Tek-Cre mice was evidenced by the immunofluorescence using brain sections (Supplementary Figure 6C). LAT1 mRNA was decreased in the aortic rings of the Lat1 fl/fl /Tek-Cre mice to~70% of the control Lat1 fl/fl mice on Day 0 and Day 3 (Fig. 3c). The endothelial sprouting was significantly suppressed in the aortic rings of Lat1 fl/fl /Tek-Cre mice, further supporting the contribution of endothelial LAT1 (Fig. 3d).

Genetic and pharmacological inhibition of endothelial LAT1 suppress angiogenesis and tumor growth
It has been demonstrated that LAT1 inhibitor JPH203 suppresses the xenograft tumor growth [21,[24][25][26][27]. In the present study, intravenous administration of JPH203 drastically suppressed the growth of MIA PaCa-2 xenograft tumors (Fig. 4a-c). Concomitantly, the intratumoral blood-vessel density was reduced in JPH203-treated tumors to~45% of the placebo-treated control (Fig. 4d  and e), which let us speculate that the reduction of tumor angiogenesis could contribute to the anti-tumor effects of JPH203.
We then examined whether the depletion of endothelial LAT1 suppresses tumor angiogenesis and, consequently, suppresses tumor growth. In the orthotopic syngeneic tumor model of B16-F10 mouse melanoma cells, expression of LAT1 was confirmed in tumorassociated endothelial cells as well as tumor cells (Fig.  4f). When the tumor was constructed in Lat1 fl/fl /Tek-Cre mice, the growth was significantly suppressed compared to that in control mice ( Fig. 4g and h). The analysis of the intratumoral blood vessel density revealed that the blood-vessel area in tumors was decreased in Lat1 fl/fl / Tek-Cre mice to~50% of that in the control mice ( Fig.  4i and j). These results demonstrate that LAT1 in tumor-associated endothelial cells plays essential roles in tumor angiogenesis, and that the suppression of its function or expression could contribute to exert anti-tumor effects.

LAT1 supports proliferation of endothelial cells via mTORC1-and GAAC pathways
Because LAT1 preferentially transports many essential amino acids, we investigated the importance of LAT1 in the endothelial cell proliferation using human umbilical vein endothelial cells (HUVECs). The culture medium of HUVECs is supplemented with major pro-angiogenic factors, VEGF-A and FGF-2. As shown in Fig. 5a, VEGF-A or FGF-2 alone, as well as their combination induced LAT1 mRNA expression in HUVECs. The combinational effect of VEGF-A and FGF-2 on LAT1 mRNA expression peaked at 2~4 h, and was sustained as long as 16 h (Fig. 5b). A consistent increase in the LAT1 protein amount was observed at 8 and 24 h after the stimulation with VEGF-A and FGF-2 (Fig. 5c). These results suggest that VEGF-A and FGF-2 could contribute to the induction of endothelial LAT1 expression under pro-angiogenic conditions.
The knockdown (KD) of LAT1 by siRNAs, that reduced the LAT1 protein amount to 15~25% of the control (Fig.  5d), impaired the proliferation of HUVECs (Fig. 5e). Similarly, LAT1 inhibition by JPH203 or BCH suppressed the proliferation of HUVECs in concentration dependent  Figure 5C). These results showed that LAT1 plays a crucial role for the endothelial cell proliferation.
Amino acids are essential signaling molecules to activate a serine/threonine kinase complex mTORC1 (mechanistic target of rapamycin complex 1) that integrates nutrient-and growth factor signaling to support cell growth and proliferation [43]. Most wellcharacterized downstream effectors of mTORC1 include ribosomal protein S6 kinase p70S6K, a regulator of translation initiation. The accumulation of uncharged tRNAs under amino acid deficiency also activates the other signaling pathway, known as general amino acid control (GAAC) pathway [44,45]. Uncharged tRNAs activate general control nonderepressible 2 (Gcn2) kinase and induce phosphorylation of eIF2α, which triggers a global down-regulation of translation by inhibiting the recruitment of initiator methionyl-tRNA to ribosome. As shown in Fig. 5g and h, LAT1 KD as well as LAT1 inhibition by JPH203 in HUVECs markedly reduced the phosphorylation of p70S6K and its substrate ribosomal protein S6. The phosphorylation of eIF2α was also increased, indicating the activation of GAAC pathway by amino acid deficiency. Collectively, these results indicate that LAT1-mediated amino acid transport in HUVECs is an essential prerequisite to activate translation initiation. The inhibition of endothelial LAT1 could globally downregulate translation by suppressing mTORC1 activity and activating GAAC pathway.

LAT1 is involved in migration, invasion and tubular network formation of endothelial cells in angiogenic cellular processes
Angiogenesis involves multiple cellular processes such as proliferation, migration, invasion, morphological change, and differentiation of endothelial cells. Because we confirmed LAT1 is essential for the proliferation of HUVECs (Fig. 5 and Supplementary Figure 5C), we further examined whether LAT1 is also involved in the other angiogenic cellular processes of HUVECs. In wound healing assay, LAT1 KD suppressed the migration (Fig. 6a), where the number of migrated cells was reduced to 50~60% of the control (Fig. 6b). LAT1 inhibitors, JPH203 and BCH, also suppressed the cell migration in concentration dependent manners ( Fig. 6c and Supplementary Figure 5D). In the transwell invasion assay, LAT1 KD as well as LAT1 inhibitors reduced the number of cells migrated through a Matrigel layer (Fig.  6c, d and Supplementary Figure 5E). In the tube formation assay, LAT1 KD strongly disturbed the formation of tubular networks (Fig. 6e and f). Treatment with JPH203 also exhibited a reduction of tube formation. Effects of BCH, in which high concentration is required due to its lower affinity, were not evaluated because the tube formation was highly sensitive to the osmolality of culture medium. These results indicate that endothelial LAT1 contributes not only to proliferation but also to multiple angiogenic cellular processes, including migration, invasion, and tubular network formation.

LAT1-mediated amino acid transport is indispensable for VEGF-A-dependent activation of mTORC1
In the signaling pathways regulating angiogenesis, VEGF-A and its cognate receptor VEGFR2 are known to play a central role [46,47]. In our ex-and in vivo assays, VEGF-A was utilized as an angiogenic stimulant (Figs. 2 and 3). We also revealed that LAT1 in HUVECs is essential for angiogenic processes induced by VEGF-A stimulation (Fig. 6). We thus examined the contribution of LAT1 to VEGF-A-mediated pro-angiogenic intracellular signaling pathways under the condition comparable to that of in vitro assays shown in Fig. 6, i.e., HUVECs were starved for serum and growth factors (VEGF-A, FGF-2, EGF, and IGF-1), and then stimulated by VEGF-A.
As shown in Fig. 7a, stimulation of the starved HUVECs with VEGF-A resulted in a transient increase of VEGFR2 phosphorylation at 20 min. The phosphorylation decreased with longer incubation time (> 1 h), but was sustained at a higher level than that before the treatment. Major downstream factors of VEGF-A/ VEGFR2, including Erk1/2, Akt, p38, Src, FAK, p70S6K, and S6 ribosomal protein, also exhibited similar transient time courses in their phosphorylation, except PLCγ that showed a relatively delayed response. Treatment with JPH203 did not influence the phosphorylation of VEGFR2 and the downstream factors except for p70S6K and S6. The phosphorylation of p70S6K and S6 was drastically suppressed by JPH203 as early as 20 min after the stimulation, revealing that the VEGF-A-induced activation of mTORC1 is highly dependent on LAT1. It is especially of note that the phosphorylation of Akt at Thr308, locating in the upstream of mTORC1 [43], was less affected by JPH203. The decreased mTORC1 activity is, thus, most likely due to the reduced input of amino acid signaling mediated by Ragulator-Rag complex, which recruits mTORC1 onto lysosomal surface and facilitates its interaction with kinase activator Rheb in a manner generally independent of RTK-PI3K-Akt axis [43].
The inhibitory effects of JPH203 on VEGF-A-dependent mTORC1 activation were comparable to that of mTORC1 inhibitor rapamycin (Fig. 7b). Even though a residual phosphorylation of p70S6K was detected in JPH203treated cells, the increase of phosphorylation in response to VEGF-A stimulation was mostly abolished. Furthermore, the phosphorylation of ribosomal S6 protein, the downstream of p70S6K, was suppressed to a similar extent by JPH203 and rapamycin. As shown in Fig. 7c, LAT1 KD also impaired the VEGF-A-dependent activation of Fig. 7 Effects of LAT1 knockdown and inhibition on VEGF-A-dependent signaling pathways. a HUVECs starved for serum and growth factors were stimulated with VEGF-A (10 ng/mL) in the presence or absence of JPH203 (50 μM). b and c Effects of JPH203 and rapamycin, and LAT1 KD on mTORC1-and GAAC pathways under stimulation with VEGF-A. HUVECs starved for serum and growth factors were stimulated with VEGF-A (10 ng/mL) for 20 min in the presence or absence of JPH203 (50 μM) or rapamycin (10 nM) (b). HUVECs transfected with control siRNA (NC) or LAT1targeting siRNA (LAT1 siRNA #1) were starved and stimulated with VEGF-A (10 ng/mL) for 20 min (c). d Proposed molecular mechanism for the essential role of endothelial LAT1 in the VEGF-A-dependent activation of mTORC1. The LAT1-mediated amino acid signaling is independent of the PI3K-Akt axis in the downstream of VEGFR2, and is possibly mediated by Ragulator-Rag GTPase heterodimer complex that recruits inactive mTORC1 onto lysosomal surface for its interaction with kinase activator Rheb. The input of amino acid signaling is indispensable for the proangiogenic VEGF-A signaling to induce activation mTORC1 and subsequent angiogenesis p70S6K and S6, without affecting th`e phosphorylation levels of VEGFR2 and Akt (Thr308).

Discussion
Our study revealed that amino acid transporter LAT1 expressed in tumor-associated endothelial cells is a novel key molecule in tumor angiogenesis. Extending previous studies with limited observations on a rat bladder carcinoma model [29] and human glioma tissues [30], we established the upregulation of LAT1 expression as a general characteristic of tumor-associated endothelial cells. The functional relevance of endothelial LAT1 to tumor angiogenesis was demonstrated in in vivo models by genetic and pharmacological inhibition of LAT1 (Fig.  4). Even though our study do not completely exclude a possibility that endothelial LAT1 also contributes to angiogenesis in certain physiological contexts, the endothelial cell-specific knockout of LAT1 in mouse strongly indicate that endothelial LAT1 is, at least, not essentially required for angiogenesis related to growth and survival. Furthermore, although LAT1 is expressed in brain epithelial cells as shown in Supplementary Figure 4 and in previous study [41], no obvious neurological adverse effects of JPH203 have been reported in previous studies using animal models [21,[24][25][26][27] and the first clinical trial [28]. We also did not observe any apparent neurological symptoms of mice in the present study. One possible explanation is that the inhibition of LAT1 can be compensated by the function of other amino acid transporters at blood brain barrier, the substrate specificity of which is overlapped with LAT1 [48]. Therefore, LAT1 seems to be a novel promising target in anti-angiogenic therapy. A strong anti-proliferative effect supported by a global down-regulation of translation could be achieved by endothelial LAT1 inhibition, not only by blocking the supply of amino acids as building blocks for protein synthesis, but also by interfering with amino acid signaling that regulates the initiation of translation ( Fig. 5g and h). Such predominant inhibitory effects on translation are specific to LAT1 inhibition, clearly differentiating the mechanisms of action of LAT1 inhibitors from that of existing anti-angiogenic agents.
Intrinsic and acquired resistances against anti-angiogenic therapy often limit the benefits for patients [1,2]. The multiple redundant and compensatory pro-angiogenic signaling pathways present in endothelial cells are supposed to play a crucial role in the resistance. A promising strategy to overcome the resistance would be to target multiple signaling pathways simultaneously. Accordingly, combination of FGFR inhibitor and bevacizumab in mouse tumor models almost completely suppressed tumor growth [49]. In pancreatic islet mouse tumors, resistance to VEGFR2 inhibitor was successfully impaired by the soluble decoy FGF receptor [6]. In this study, we demonstrated that LAT1 is indispensable for VEGF-A-dependent activation of mTORC1 (Fig. 7), which plays key roles in the cellular processes such as migration and tube formation in vitro as well as in in vivo angiogenesis [50][51][52][53]. Our results suggest that the roles of LAT1 in the activation of mTORC1 is mediated by Ragulator-Rag complex that is independent of RTK-PI3K-Akt axis. The amino acid signaling mediated by LAT1 seems to behave as a "gate-control" signal to permit the passage of pro-angiogenic VEGF-A signaling through mTORC1 to its downstream (Fig. 7d). Similar to the VEGFR signaling, multiple other proangiogenic RTKs including FGFR and TIE-2 share the PI3K-Akt axis that activates mTORC1 [46]. Therefore, the therapeutic inhibition of LAT1 with JPH203 could simultaneously interfere with not only VEGF-A/VEGFR2 signaling but also other pro-angiogenic signaling pathways at mTORC1, offering a possibility to circumvent the resistance resulting from the compensatory function of pro-angiogenic growth factor signaling.
While LAT1 is well-known as a "tumor cell-type transporter" highly and broadly upregulated in tumor cells to support their growth and proliferation, our study indicates a new insight into the dual functioning of LAT1 in tumor progression both in tumor cells and stromal endothelium. In this regard, we also would like to emphasize the unique dual mechanisms of action of LAT1 inhibitor JPH203 as anti-tumor agents, i.e. the well-established direct anti-proliferative effects on tumor cells through the inhibition of LAT1 in tumor cells and the anti-angiogenic effect through the inhibition of endothelial LAT1. A tempting speculation is that, when combined with other anti-angiogenic agents, administration of LAT1 inhibitors would suppress the compensatory paracrine secretion of pro-angiogenic factors from tumor cells, through the down-regulation of protein synthesis in tumor cells. Therefore, combinational therapies of LAT1 inhibitors with anti-angiogenic agents may show beneficial synergic anti-tumor effects with a lower risk of developing resistance.
Several lines of evidence indicate that tumor-associated endothelial cells are distinct from their normal counterparts in the expression of characteristic proteins [54][55][56]. Our present study indicates the increased LAT1 expression is also a part of such tumor endothelium-specific characteristics. We detected the endothelial LAT1 expression not only in tumor tissues but also in in vitro HUVEC cultures and in the endothelial cells from ex/in vivo angiogenesis assays, in which VEGF-A and FGF-2 were supplemented to culture media (Figs. 2, 3, and 5). These proangiogenic factors induced the expression of LAT1 in HUVECs at both mRNA and protein levels (Fig. 5a-c). It was previously reported that LAT1 is a direct target gene of oncogenic c-Myc [57,58]. In the ontogenetic development, the expression of c-Myc in endothelial cells is regulated by VEGFR2 [59] and FGFR [60]. Even though further studies are awaited to elucidate the details, the tumor microenvironment rich in VEGF-A and FGF-2 may partly account for the upregulation of LAT1 in tumorassociated endothelium.

Conclusion
In summary, we demonstrate that an amino acid transporter LAT1 is upregulated in tumor endothelium and plays fundamental roles in tumor angiogenesis. We revealed a cross-talk between LAT1-mediated amino acid signaling and growth factor-dependent pro-angiogenic signaling, converging on nutrient-sensing hub kinase mTORC1 to regulate angiogenesis. LAT1-targeting therapy may offer an ideal option to potentiate current cancer treatments especially for anti-angiogenic therapies.