JNK and Yorkie drive tumor malignancy by inducing L-amino acid transporter 1 in Drosophila

Identifying a common oncogenesis pathway among tumors with different oncogenic mutations is critical for developing anti-cancer strategies. Here, we performed transcriptome analyses on two different models of Drosophila malignant tumors caused by Ras activation with cell polarity defects (RasV12/scrib-/-) or by microRNA bantam overexpression with endocytic defects (bantam/rab5-/-), followed by an RNAi screen for genes commonly essential for tumor growth and malignancy. We identified that Juvenile hormone Inducible-21 (JhI-21), a Drosophila homolog of the L-amino acid transporter 1 (LAT1), is upregulated in these malignant tumors with different oncogenic mutations and knocking down of JhI-21 strongly blocked their growth and invasion. JhI-21 expression was induced by simultaneous activation of c-Jun N-terminal kinase (JNK) and Yorkie (Yki) in these tumors and thereby contributed to tumor growth and progression by activating the mTOR-S6 pathway. Pharmacological inhibition of LAT1 activity in Drosophila larvae significantly suppressed growth of RasV12/scrib-/- tumors. Intriguingly, LAT1 inhibitory drugs did not suppress growth of bantam/rab5-/- tumors and overexpression of bantam rendered RasV12/scrib-/- tumors unresponsive to LAT1 inhibitors. Further analyses with RNA sequencing of bantam-expressing clones followed by an RNAi screen suggested that bantam induces drug resistance against LAT1 inhibitors via downregulation of the TMEM135-like gene CG31157. Our observations unveil an evolutionarily conserved role of LAT1 induction in driving Drosophila tumor malignancy and provide a powerful genetic model for studying cancer progression and drug resistance.

Introduction Cancer development is achieved by the accumulation of oncogenic mutations that promote cell proliferation, survival, invasion, and metastasis. Mutations that drive tumor growth and malignancy are different between tumors and thus identification of a common oncogenesis pathway among tumors with different oncogenic alterations is crucial for establishing effective anti-cancer strategies.
In recent years, amino acid transporter, especially L-amino acid transporter 1 (LAT1), has been attracting attention as a potential therapeutic target for cancer. LAT1 is a plasma membrane transporter for branched-chain amino acids (BCAAs) such as leucine and isoleucine [1], thereby promoting tumor growth by activating mTOR signaling [2]. LAT1 acts as a protein complex composed of LAT1 covalently bound to 4F2 Heavy Chain Antigen (CD98/SLC3A2) [3]. CD98 promotes LAT1 protein stability and mediates the translocation of LAT1 to the cell membrane [3]. Studies in mammalian cells have shown that LAT1 is upregulated in neuroblastoma and Burkitt's lymphoma cells via Myc [4] and in breast cancer cells via aryl hydrocarbon receptor (AHR)-mediated signaling [5]. Given that LAT1 expression is elevated in a variety of cancers, it is thought to be an ideal therapeutic target as a component of a common oncogenesis pathway and is thus currently under clinical trial in cancer patients [2,6]. However, genetic complexity and heterogenous nature of cancer have hindered progress in understanding the mechanism of the common genetic pathway of tumor growth and malignancy in mammalian systems.
The genetic mosaic technique available in Drosophila provides an ideal model system to study tumor growth and progression with genetically traceable oncogenic mutations. Indeed, previous studies in Drosophila imaginal epithelium have identified critical mechanisms by which accumulation of distinct oncogenic alterations drives tumor malignancy. For instance, clones of Ras-activated benign tumors are transformed into malignant tumors when simultaneously mutated for an apicobasal polarity gene such as scribble (scrib) [7,8]. In addition, clones of cells activating Ras and Src signaling develop into invasive tumors under high-sugar diet condition [9].
Here, through a transcriptome analysis combined with Drosophila genetics, we searched for a common pathway of oncogenesis among different types of Drosophila malignant tumors. By an unbiased transcriptome analysis, we found that LAT1 expression was commonly elevated in Drosophila malignant tumors with different oncogenic mutations. Genetic or pharmacological inhibition of LAT1 significantly blocked tumor growth and malignancy in these tumors. Our findings unveil an evolutionarily conserved role of LAT1 induction in tumor progression and provide a novel genetic model for analyzing cancer progression and drug resistance.

JhI-21/LAT1 is required for tumor growth and invasion in Drosophila
In Drosophila imaginal epithelia, clones of cells overexpressing oncogenic Ras V12 with simultaneous mutations in apico-basal polarity genes such as scribble (scrib) or discs large (dlg) result in tumorous overgrowth and metastatic behavior, the best-characterized model of Drosophila malignant tumors [7,8] (Fig 1b, compare to Fig 1a). To study a common pathway of oncogenesis among different types of malignant tumors, we first tried to establish another model of Drosophila malignant tumors using different oncogenic mutations. As a result, we found that lossof-function mutations in a tumor-suppressor gene rab5 [10], a small GTPase essential for generating early endosomes [11], in clones of cells overexpressing a pro-growth microRNA bantam (bantam/rab5 -/cells), a target of the Hippo pathway effector Yorkie (Yki), in the eye discs resulted in drastic tumor growth and malignant invasion to adjacent organ ventral nerve cord (VNC) (Figs 1c and S1e, quantified in S1f Fig). Notably, overexpression of bantam alone or rab5 mutation alone caused neither tumor growth (S1a and S1c Using these two malignant tumor models with distinct oncogenic mutations, we performed RNA sequence (RNA-seq) analyses from GFP-labeled fluorescence-activated cell sorting (FACS)-sorted Ras V12 /scrib -/or bantam/rab5 -/cells compared to GFP-labeled wild-type cells (Fig 1d). We identified 4,553 and 2,471 genes that are significantly upregulated or downregulated in Ras V12 /scrib -/and bantam/rab5 -/cells, respectively, and 1,734 genes that are commonly upregulated (1,028 genes) or downregulated (706 genes) in both tumors compared to wild-type cells (Fig 1e and S1 Table, false discovery rate (FDR)< 0.05). Among 1,028 commonly upregulated genes, we found JNK signaling targets such as mmp1 [12] and puckered (puc) [13], JAK/STAT signaling targets such as upd1, upd2, and upd3 [14], and Yki targets such as expanded, cycE, and crb [15] (Fig 1f and S1 Table), which are consistent with previous reports for the upregulated genes in Ras V12 /scrib -/tumors [16][17][18]. These data validate our experimental conditions and suggest that these malignant tumors with distinct mutant origins share common downstream signaling to induce tumor growth and invasion.

Activation of JNK and Yki upregulates JhI-21
LAT1 is a plasma membrane transporter for BCAAs such as leucine and isoleucine and is often upregulated in tumor cells, thereby promoting tumor growth [6]. However, the mechanism by which LAT1 is upregulated in various cancer cells is still unclear. We thus investigated the mechanism by which JhI-21 is upregulated in Ras V12 /scrib -/or bantam/rab5 -/tumor cells. Consistent with the RNA-seq data, immunostaining for JhI-21 protein in the eye-antennal discs showed upregulation of JhI-21 in Ras V12 /scrib -/-. (Fig 2a, Fig 2g), suggesting that additional factor is required for JhI-21 induction in conjunction with JNK activation. Interestingly, we found that overexpression of a Hippo pathway component Warts (Wts, a Lats homolog that suppresses Yki/YAP activity [22]) in Ras V12 /scrib -/tumors abolished the upregulation of JhI-21 (Fig 2d, quantified in Fig 2g), suggesting that Yki activity is also required for JhI-21 induction. Indeed, Yki activation was observed in both Ras V12 /scrib -/and bantam/rab5 -/tumors, as visualized by Yki activity reporter expanded (ex)-lacZ or four-jointed (fj)-lacZ (Figs 2f and S2g) as well as by the RNA-seq data showing upregulation of Yki targets expanded, cycE, crb, and upd1 (Fig 1e and S1 Table). However, Yki activation alone by overexpressing an activated form of Yki (Yki S168A ) did not cause JhI-21 induction (Fig  2e, quantified in Fig 2g). Significantly, we found that co-activation of JNK and Yki caused JhI-21 induction (Fig 2f, quantified in Fig 2g). These data indicate that activation of JNK and Yki in Ras V12 /scrib -/or bantam/rab5 -/tumors cause upregulation of JhI-21 expression.

PLOS GENETICS
Tumor malignancy by induction of L-amino acid transporter 1 in Drosophila blocked mTOR signaling activation in Ras V12 /scrib -/tumors (Fig 3g, quantified in S3h Fig). In addition, blocking mTOR signaling by knocking down of an upstream regulator of mTOR signaling Rheb [24] significantly suppressed RpS6 phosphorylation in Ras V12 /dlg -/tumors (Fig  3e, compare to S3f Fig, quantified in Fig 3f) and tumor growth (Fig 3i, compare to Fig 3j, quantified in Fig 3k), while Rheb knockdown alone did not affect tissue growth (Fig 3h, quantified in Fig 3k). Together, these data suggest that JhI-21 upregulation promotes tumor growth by activating mTOR-S6 signaling in Drosophila.

Administration of LAT1 inhibitors reduces growth of Ras V12 /scrib -/tumors
We next examined whether pharmacological inhibition of JhI-21 activity could suppress growth of malignant tumors in Drosophila. It is known that LAT1 inhibitors, 2-amino-2-Norbornanecarboxylic Acid (BCH) and KYT0353 (JPH203), suppress the activity of LAT1 in mammalian cells [6]. KYT0353 is currently being evaluated in a Phase 2 clinical trial in patients with advanced biliary tract cancers (UMIN Clinical Trials Registry UMIN000034080). Notably, we found that feeding BCH or KYT0353 to larvae bearing Ras V12 /scrib -/tumors in the eye-antennal discs significantly reduced tumor growth, while these drugs did not affect growth of wild-type clones (Figs 4a, 4b, S4a and S4b). Furthermore, BCH treatment significantly suppressed mTOR signaling activity in Ras V12 /scrib -/tumors (S4c Fig). These data indicate that pharmacological inhibition of JhI-21 activity suppresses growth of Ras V12 /scrib -/malignant tumors by downregulating mTOR signaling.

MicroRNA bantam renders malignant tumors resistant to LAT1 inhibitors
To our surprise, BCH and KYT0353 did not suppress growth of bantam/rab5 -/tumors (Fig 4a  and 4b). Consistent with this result, BCH treatment did not suppress mTOR signaling activity in bantam/rab5 -/tumors (S4c Fig). Notably, overexpression of bantam in Ras V12 /dlg -/tumors abolished the suppressive effect of LAT1 inhibitors on their growth (Figs 4c, 4d, S4d and S4e). These data suggest that bantam renders malignant tumors resistant to LAT1 inhibitory drugs.
To identify gene(s) responsible for the drug resistance against LAT1 inhibitors upon bantam expression, we performed an RNA-seq analysis of FACS-sorted bantam-overexpressing cells compared to wild-type cells in the eye-antennal discs. Expression levels of 42 genes were significantly altered in bantam-overexpressing cells compared to wild-type cells, and among these 10 genes were commonly upregulated (3 genes) or downregulated (7 genes) in both bantam cells and bantam/rab5 -/tumors but not in Ras V12 /scrib -/tumors (Fig 4e and S2 Table, false discovery rate (FDR)<0.05). Crucially, we found that knockdown of CG31157 (GD3640), one of the commonly downregulated 7 genes encoding a TMEM135-like protein, in Ras V12 / dlg -/tumors abrogated tumor-suppressive effect of BCH (Fig 4g, S4h Fig). Furthermore, bantam/rab5 -/tumors overexpressing CG31157 transgene became sensitive to BCH treatment (Fig 4j), indicating a critical role of CG31157 in drug responsiveness. Notably, CG31157 expression was significantly upregulated in Ras V12 /scrib -/tumors (~1.7 fold, S2 Table), suggesting that CG31157 expression is critical for cells to acquire sensitivity to LAT1 inhibitors. These data suggest that bantam renders malignant tumors resistant to LAT1 inhibitory drugs via downregulation of TMEM135-like gene CG31157.

Discussion
Our genetic study using Drosophila tumor models revealed that activation of JNK and Yki drives tumor growth and malignancy by inducing JhI-21, a fly homolog of LAT1 (Fig 5). It has previously been shown that JhI-21 acts as an amino acid transporter that uptakes leucine into insulin producing cells (IPCs) in Drosophila larva and is required for leucine-dependent secretion of Drosophila insulin-like peptide 2 (Dilp2) from IPCs [23]. In this study, we found that JhI-21 is commonly upregulated in Drosophila malignant tumors and contributes to tumor growth and malignancy via activation of the mTOR-S6 pathway. Similar to mammalian systems, knockdown or pharmacological inhibition of JhI-21/LAT1 significantly reduced growth of malignant tumors. We also found that mnd was commonly upregulated in Ras V12 /scrib -/and bantam/rab5 -/tumors and knockdown of mnd suppressed tumor growth. Mnd is an amino acid transporter belonging to LAT1 family, which catalyzes the cross-membrane flux of large neutral amino acids by forming heterodimers with CD98 [19,27]. Thus, our data suggest that similar to mammalian cancers, elevation of LAT1 activity is critical for tumor growth and progression in Drosophila. This indicates that future studies on the common oncogenesis pathway in Drosophila could provide novel therapeutic strategies against human cancer.
In this study, we found that co-activation of JNK and Yki leads to upregulation of JhI-21 in Drosophila imaginal discs. The mechanism by which JNK and Yki induce JhI-21 expression is currently unknown, which should be addressed in the future studies. Nonetheless, JNK activation has been shown to be essential for tumor growth and invasion in Drosophila malignant tumors [16] and JNK activation has long been implicated in tumor growth and progression in mammalian systems [30][31][32], underscoring the critical role of JNK in tumor progression. Indeed, hyperactivation of JNK signaling, as well as elevated YAP activity, have been reported in many human cancers [24,32]. Given that signaling molecules identified in this study are all conserved in humans, similar tumor progression mechanism via JNK and YAP-mediated LAT1 induction could regulate human cancers.

Drosophila strains and genetics
Fly stocks were cultured at room temperature or 25˚C on standard fly food. Fluorescentlylabeled mitotic clones [33,34] were produced in larval imaginal discs using the following strains: CG31157 was PCR-amplified from cDNA of w 1118 adult flies using primers designed to append restriction sites for enzymes EcoR1 and Xbal to the 5' and 3' end of the product. Sequence of the products were confirmed and then the product was cloned into the pUAST vector. Transgenic flies were generated by WellGenetics.

RNA-seq data analysis
Reads were trimmed to 75 nucleotides length by fastx_trimmer in FASTX Toolkit (v0.0.14), and further quality-filtered by trim_galore (v0.5.0) with default setting to remove the adaptor sequence and the low quality reads. The reads passing filters were mapped to the Drosophila melanogaster Ensembl BDGP6 obtained from illumine iGenomes by STAR (v2.7.0e) [37]. >80% of reads were uniquely mapped for each experiment. The number of reads that map to each gene was counted by htseq-count (HTSeq v0.11.2) with -s reverse option [38]. Normalization was carried out using calcNormFactors function (edgeR) [39,40]. Differentially expressed genes were identified using glmQLFit and glmQLFTest function in edgeR at a FDR threshold 0.05. R version 3.