Dysregulated amino acid sensing drives colorectal cancer growth and metabolic reprogramming leading to chemoresistance

Abstract

Background and Aims:

CRC is a devastating disease highly modulated by dietary nutrients. mTORC1 contributes to tumor growth and limits therapy responses. Growth factor signaling is a major mechanism of mTORC1 activation. However, compensatory pathways exist to sustain mTORC1 activity following therapies that target oncogenic growth factor signaling. Amino acids potently activate mTORC1 via amino acid sensing GTPase activity towards Rags complexes (GATOR). The role of amino acid sensing pathways in CRC is unclear.

Methods:

Human colon cancer cell lines, preclinical intestinal epithelial specific GATOR1 and GATOR2 knockout mouse subjected to colitis induced or sporadic colon tumor models, siRNA screening targeting regulators of mTORC1, and CRC patient tissues were used to assess the role of amino acid sensing in CRC.

Results:

We identified loss-of-function mutations of the GATOR1 complex in CRC and show that altered expression of amino acid sensing pathways predict poor patient outcomes. We show that dysregulated amino acid sensing induced mTORC1 activation drives colon tumorigenesis in multiple mouse models. We found amino acid sensing pathways to be essential in the cellular reprogramming of chemoresistance, and chemotherapeutic resistant colon cancer patients exhibited deregulated amino acid sensing. Limiting amino acids in in vitro and in vivo model (low protein diet) reverted drug resistance revealing a metabolic vulnerability.

Conclusions:

Our findings suggest a critical role of amino acid sensing pathways in driving CRC and highlights translational implications of dietary protein intervention in CRC.

Introduction

Colorectal cancer (CRC) progress through adenoma-carcinoma sequence via an initiating driver mutation in APC and subsequent mutations in oncogenes such as PIK3CA, KRAS, BRAF, and loss of tumor suppressors TP53, SMAD4^1–3. Hyperactive mTORC1 drives CRC and its inactivation inhibits colon tumor growth⁴. mTORC1 is a central node in regulating anabolic metabolism, cellular growth, survival and proliferation⁵. PI3K/Akt is the most common growth factor signaling cascade dysregulated in cancers and a potent mTORC1 activator⁵. Animal studies show that inhibition of the PI3K/Akt/mTORC1 axis inhibits tumor growth⁶. Therapeutic targeting of PI3K/Akt/mTORC1 signaling using PI3K inhibitors have demonstrated clinical success⁷. However, patients develop intrinsic or adaptive drug resistance⁷. Importantly, persistent mTORC1 signaling is a major driver of drug resistance to PI3K inhibitors^8,9. Single use of rapalogs have limited therapeutic efficacy and pose serious side effects¹⁰. As mTORC1 is central in controlling diverse metabolic processes driving tumor growth and drug resistance, there is a pressing need to understand mechanisms deployed by cancer cells to chronically activate mTORC1 signaling in the tumor microenvironment.

Amino acids are known to activate mTORC1¹¹ and dietary protein restriction inhibit tumor growth via reduced mTORC1^12,13. Restriction of specific amino acids exert deleterious impact on cancer cell metabolism revealing tumor specific metabolic vulnerabilities¹⁴. Discoveries of amino acid sensors and downstream effectors have shed light into mechanisms regulating mTORC1¹¹. GTPase activity towards Rags (GATOR) 1 and 2 complexes are negative and positive regulators of mTORC1, respectively¹⁵. GATOR2 is a pentameric protein complex composed of Wdr59, Wdr24 (WD Repeat Domain 24), Mios, Seh1l and Sec13. GATOR2 forms a protein complex with GATOR1, a trimeric complex containing Depdc5 (DEP domain containing protein 5), Nprl2 and Nprl3. In the absence of leucine or arginine, GATOR2 is inhibited which in turn relieves the repression on GATOR1. GATOR1 inhibits Rags A/B and subsequently mTORC1 and cell growth¹⁵. The role of amino acid sensing induced mTORC1 in CRC is unclear.

Here, we report dysregulated amino acid sensing pathway decrease survival of CRC patients and drive CRC tumor growth in multiple mouse models. We found that dysregulated amino acid sensing induced mTORC1 activation drives drug resistance which could be reversed by limiting amino acids, revealing a tumor vulnerability. Our findings suggest a crucial role of amino acid sensing pathways in driving human CRC and supports dietary intervention in CRC.

Methods

Animals and treatments

All animal studies were approved by the IACUC at the University of Michigan. For all experiments, 6 – 8-week-old male and female mice on C57BL/6 background were used. PIK3CAα^f/f, PI3KCBβ^f/f were crossed with inducible villin-Cre (Vil^ERT2) to produce intestinal epithelial-specific inducible PI3K knockout (PI3K^ΔIE) mice. For syngeneic model, 2 million MC38 cells were subcutaneously injected in both flanks of C57BL/6 mice, placed on control diet: TD.180201 or 4% protein diet: TD.93032 (Envigo) when tumors became palpable. After 2 weeks, mice were sacrificed and tumors were collected, weighed, and processed. For sporadic colon tumor studies, CDX2^ERT2 Apc^−F/F mice were kept on control or 4% protein diet for 2 weeks, treated with one dose of 50-mg/kg tamoxifen (HY-13757A, MedChemExpress, NJ) and were euthanized after 30 days. Depdc5^F/F and Wdr24^F/F mice were obtained as described in¹⁶ Depdc5^F/F and Wdr24^F/F mice were crossed to villin-Cre to produce intestinal epithelial-specific Depdc5^ΔIE and Wdr24^ΔIE knockout mice respectively. For colitis-associated colon tumor model, Depdc5^ΔIE and Wdr24^ΔIE and their respective littermates were i.p injected with azoxymethane (Sigma A5486 – 10mg/kg). After 3 days, mice were cycled on and off 2.5% dextran sulfate sodium (MP Biomedicals,160110) in their drinking water for 7 days. Mice were sacrificed after 4 weeks following last cycle of DSS. Tumors were counted, measured and swiss-rolled for histological analysis. For sporadic tumor model, Depdc5^F/F mice were crossed with CDX2^ERT2 Apc ^F/F mice to generate double knockout mouse model (CDX2^ERT2 Apc^F/F Depdc5^F/F). CDX2^ERT2 Apc^F/F and CDX2^ERT2 Apc^F/F Depdc5^F/F mice were injected one dose of 50mg/kg tamoxifen, euthanized after 30 days. For metabolic vulnerability in vivo, CDX2^ERT2 Apc^F/F Depdc5^F/F mice were injected one dose of 50-mg/kg tamoxifen, placed on control diet or 4% protein diet, euthanized after 30 days.

Cells and treatments.

All the human and mouse colon cancer cell lines were purchased from ATCC. Wdr24^−/− and Depdc5^−/− CRC cells were generated by CRISPR/ Cas9. The primer sequences are given in Table 1. For Insulin and EGF experiments, colon cancer cells were incubated in serum free media for 16 hours followed by amino acid free media (Dulbecco’s MEM, D9800-13) for 2 hours and 1X amino acid mixture (GIBCO 11130 - MEM Amino Acids Solution) was added for 1 hour. Next, cells were stimulated with Insulin (10nmol/ml, Sigma I9278) or EGF (50ng/ml, R&D Systems 236EG200) at indicated time points. For Wdr24^−/− experiments, Wdr24^+/+ or Wdr24^−/− cells were incubated in serum free or serum (10% FBS) containing media and amino acid free or amino acid containing media for 16 hours. For insulin experiments in Wdr24^+/+ & Wdr24^−/−, cells were treated as above. For Depdc5^−/− experiments, Depdc5^+/+ and Depdc5^−/− cells were incubated in amino acid free or different amino acid concentrations (0.1X, 0.2X, 0.5X, 1X), all conditions being supplemented with 10% FBS for 24 hours. For generating resistant clones, CRC cells were cultured in increasing doses of 5FU (100nM – 5µM for HCT116, SW480, DLD1) and (100nM – 2µM) for MC38 or MK2206 (50 nM – 2µM for HCT116, SW480, DLD1, MC38) until resistance was achieved and maintained in 5FU or MK2206. (Note: Experiments conducted with Depdc5^+/+ and Depdc5^−/− CRC cells used 2µM concentration of 5FU for HCT116, SW480, DLD1 cells, 200 nM for MC38 cells. 1µM of MK2206 was used for EV and Depdc5^−/− CRC (HCT116, SW480, DLD1, MC38). IC_50s for respective cell lines were calculated as described previously¹⁷ and are given in Table 2. For CFUs, cells were plated in biological triplicates at 400 cells in a 12 well plate. After 24h, cells were treated with different amino acid concentrations or drugs (5FU, MK2206, Rapamycin). Media was changed every 3 days. Assays were concluded at 10 days by fixation in cold 10% buffered formalin for 15 min and staining with 0.5% crystal violet, 20% methanol solution for 30 min. Plates were scanned using (BioTek Cytation5 Imaging system) and colonies were graphed as measure of total number x area of colonies. Apc^−/− (single mutant), Apc^−/−/p53^−/− (double mutant) colonoids were cultured ( LWRN medium) and plated in Matrigel (Corning, 354230) and allowed to establish for at least 3–5 days as previously described¹⁸. Enteroids generated were then incubated in amino acid free, 0.1X and 1X media supplemented with 10% FBS for 2 hours. Depdc5 mutants were generated using site directed mutagenesis kit (QuikChange II XL, 200521). The primers used are in Table 1.

siRNA screen.

HCT116 cells were subjected to small interfering RNA (siRNA)-based screen targeting known regulators of mTORC1 as we have previously described¹⁹.

Immunoblotting.

Whole cell extracts and immunoblotting was performed as previously described²⁰. Primary antibodies used from cell signaling were pAkt-T308 (13038), pAkt-S473(4060S), tAkt (4691S), pErk1/2 (9101S), tErk1/2(4695T), pS6 (2215S), tS6 (2317S), pS6K (9234S), tS6K (2708S), cleaved caspase 3 (9664S), β-Actin (60008-1-Ig) and sestrin2 (66297-I-Ig) were from Proteintech, Wdr24 (244614, Santacruz), Antibodies for Depdc5 were generated as previously described¹⁶. HRP-conjugated secondary antibodies were purchased from Cell signaling (7076, 7074) and ThermoFisher ((A15999) Chemiluminescence was detected using ECL (34095, ThermoFisher) and processed using Bio-Rad ChemiDoc imaging systems.

RNA isolation and qPCR.

qPCR was performed using SYBR green (Life Technologies, Carlsbad, CA) as previously described²⁰. The primers used for quantitative PCR are listed in Table 1.

Histological analysis.

Duodenum and colonic tissues were rolled and fixed with PBS-buffered formalin. H&E and immunostaining was performed in paraffin-embedded tissue sections (5 microns) as previously described²⁰. Primary antibodies used are Ki67 (D3B5), pS6 and Sesn2. Secondary antibodies used are anti-rabbit IgG Alexa Fluor-488 (A-11008, Invitrogen) and sections were mounted using the Prolong Gold Antifade Reagent with DAPI (P36931, ThermoFisher). For cleaved caspase 3, IHC was performed using the Vecstatin Elite ABC kit (Vector Labs). Images were quantified using ImageJ

Statistical analysis.

Significance among two (student’s t-test) or more groups were analyzed using a 1-way ANOVA followed by Tukey’s post hoc test for multiple comparisons. P < 0.05 was considered significant. Graph Pad Prism 7.0 was used to conduct the statistical analyses.

Results

Dietary protein restriction reduces tumorigenesis in a sporadic colon tumor model.

In CRC-cells, we evaluated time dependent mTORC1 activation by PI3K/Akt by insulin in the presence and absence of amino acids. We observed significantly higher induction of S6K and S6 phosphorylation when amino acids were present regardless of insulin induced activation of Akt (Fig 1A). Similar results were observed in mouse CRC cells (Fig S1A) and following EGF treatment (Fig S1B). Next, mice disrupted for p110α and p110β subunits of PI3K (PI3K^ΔIE) were generated using a tamoxifen-inducible intestinal epithelial-specific Cre as described²⁰ and were confirmed by assessing Akt activation (Fig 1B, S1C). Interestingly, mTORC1 activation was marginally reduced in the colon (Fig 1B) and small intestine (Fig S1C) of PI3K^ΔIE mice. As amino acid withdrawal reduced mTORC1 activation in vitro, we next examined the impact of amino acids on mTORC1 activation in vivo. We employed a low dietary protein approach (4% protein compared to 21% protein in control diet). Strikingly, low protein diet reduced mTORC1 activation in the colonic epithelial scrapes in the PI3K^ΔIE and littermate controls (Fig 1C) suggesting that amino acids are essential in modulating mTORC1 activation in vivo. To assess the impact of low protein diet on tumor growth, a syngeneic model was used (Fig S1D). Interestingly, tumor growth (Fig S1D) and tumoral pS6 levels (Fig S1E) were unaltered on a low protein diet. However, a significant decrease in mTORC1 activation was observed in colon (Fig S1F) small intestine (Fig S1G), liver, kidney (Fig S1H) on low protein diet in this model. Based on these observations, we speculate that the depletion of tissue amino acids is relatively more rapid than subcutaneous tumors at the times assessed.

Figure 1: Low dietary protein decreases tumorigenesis in a colon tumor model.

(A) Human CRC cells were serum starved followed by amino acid starvation for 2 hours and 1X amino acids were added for 1 hour. CRC cells were stimulated with insulin (10nmol/ml) at indicated time points (B) PI3K^+/+ and PI3K^ΔIE mice aged 6–8 weeks were induced with three doses of tamoxifen (100mg/kg) and were sacrificed a week later, colonic scrapes collected (representative westerns shown, n = 3 – 5 mice per group). (C) PI3K^+/+ and PI3K^ΔIE mice were induced with tamoxifen, kept on control (21% protein) or low protein (4%) diet and were sacrificed a week later and colonic scrapes collected (D) Schematic of dietary regimen for the sporadic colon tumor model. (E) Gross colonic, (F) H&E images and (G) total and (H) stratified dysplasia scores of CDX2^ERT2 Apc^F/F mice on control (n = 8) or 4% protein diet (n = 11). (I) pS6 staining and quantitation in CDX2^ERT2 Apc^F/F mice on control (n = 3) or 4% protein diet (n = 4). (J) Ki67 staining and quantitation of in CDX2^ERT2 Apc^F/F mice on control (n = 8) or 4% protein diet (n = 11). All pictures shown are Representative. *P < .05, ***P < .001. Data are presented as mean ± SEM

Next, we assessed the impact of low protein diet on colon tumor growth in a sporadic model. Apc is the initiating mutation observed early in CRC¹. We have generated a CRC model driven by a biallelic loss of Apc under the tamoxifen inducible CDX2^ERT2 recombinase (CDX2^ERT2Apc ^F/F)¹⁸. CDX2^ERT2Apc ^F/F mice were kept on low protein or control diet for 2 weeks followed by tamoxifen induction and monitored for 1-month (Fig 1D). Mice on low protein diet displayed significantly less early tumorigenesis (polyposis) with less high-grade dysplasia compared to control group (Fig 1E–1H). Early tumorigenesis was evaluated based on the extent and grade of dysplasia (epithelial defects, edema, % region of tissue affected, and hyperplasia). These histological observations were further confirmed with reduced mTORC1 activation (Fig 1I) and reduced proliferation (Fig 1J). Together, these findings suggest that amino acids are the critical regulators of mTORC1 activity and limiting amino acids is sufficient to inhibit early tumorigenesis in a sporadic CRC model.

Amino acid sensing pathway is altered in CRC and associated with poor outcomes.

We sought to examine the role of amino acid sensing pathway in CRC (Fig 2A). CRC cells were disrupted for Wdr24, an essential component of GATOR2, by CRISPR/ Cas9 (Fig S2A). Wdr24^+/+ and Wdr24^−/− CRC cells were subjected to serum free or rich media (10% FBS) and amino acid free or rich conditions and mTORC1 activation was assessed. mTORC1 activity was significantly decreased in Wdr24^−/− CRC cells under serum free/rich containing media (Fig 2B) and amino acid free/rich conditions (Fig S2B). mTORC1 activation was also significantly reduced in Wdr24^−/− CRC cells treated with insulin in the presence of amino acids (Fig S2C) suggesting that amino acids primarily act through GATOR2 affecting mTORC1 activity in CRC cells. We further established CRC cells disrupted for Depdc5, an essential component of GATOR1, by CRISPR/Cas9 (Fig S2D). Knockout of Depdc5 resulted in constitutive mTORC1 activation which persisted under low amino acid conditions (Fig 2C). To investigate if amino acid sensing was dysregulated in human CRC, expression, and mutational status of Depdc5 was analyzed. Notably, data mining from The Cancer Genome Atlas (TCGA) revealed several point mutations (Q63R, G578V, S692Y, D1357G, L1472V, K917R) and truncated mutations (R423*) in Depdc5 (Fig 2D). We further mapped the mutations onto the known 3D structure of Depdc5 with R423* mutation highlighted in green and Q63R, D1357G, L1472V highlighted in red (Fig 2E). We identified mutants of Depdc5 as loss-of-function mutants as reflected by increased mTORC1 activation following overexpression (Fig 2F and 2G). Depdc5 expression levels were significantly reduced in tumor tissues of CRC patients using TCGA RNA-sequencing dataset (Fig 2H) and was validated in a small cohort of tumor tissues collected at the University of Michigan (Fig 2I). Lastly, reduced expression of Depdc5 was associated with diminished overall survival in the cohort (Fig 2J). Collectively, these findings demonstrate an essential role of amino acid sensing pathway in human CRC.

Figure 2: GATOR complexes are dysregulated in CRC and are essential in mTORC1 regulation.

(A) Schematic of amino acid sensing pathway. (B) Wdr24^+/+ & Wdr24^−/− CRC cells were incubated in serum free or 10% FBS containing media for 24 hours (C) Depdc5^+/+ and Depdc5^−/− CRC cells were incubated in amino acid free, 0.1X, 0.2X, 0.5X and 1X of total amino acids for 24 hours (D) Depdc5 mutants in CRC. (E) R243*, Q63R, D1357G, L1472V mutants were modeled and visualized in PyMOL (Molecular Graphics System Version 2.0 Schrödinger LLC). Green arrows indicate R243* mutant, red arrows indicate Q63R, D1357G, L1472V mutants. (F) HEK293T cells were transfected with 1 ug of Empty vector, R243* mutant and Depdc5 plasmids, and (G) D1357G, G578V, K917R, L1472V, S692Y, Q63R mutants and Depdc5 plasmids, 48 hours later cells were lysed and probed (H) Depdc5 expression was analyzed in adjacent normal and tumor tissues of CRC patients using RNA-sequencing dataset from The Cancer Genome Atlas (TCGA) and (I) in tumor tissues of a small cohort of CRC patients at UM. (J) CRC patient survival (Km plotter, statistical test) stratification based on Depdc5 expression. (Red indicates high expression of Depdc5 and black indicates low expression of Depdc5). **P < .01, ****P < .0001 Data are presented as mean ± SEM.

Inhibition of intestinal epithelial specific amino acid sensing decreases colon tumors.

Mice with intestinal epithelial-specific knockout of Wdr24 (Wdr24^ΔIE, GATOR2) using the villin cre recombinase were generated to evaluate its role in an in vivo model. (Fig 3A). A basal characterization of Wdr24^ΔIE mice demonstrated no morphological differences in the small intestine (Fig S3A) and colon (Fig 3B) of Wdr24^ΔIE mice. No significant difference in proliferation in the small intestine (Fig S3B) and colon (Fig 3C) were noted between the groups. However, mTORC1 activation was significantly decreased in the small intestine (Fig S3C) and colon (Fig 3D and 3E) of Wdr24^ΔIE mice. Further, differentiated or stem cell markers were not impacted by the loss of intestine epithelial Wdr24 (Fig S3D). To examine the role of intestinal epithelial Wdr24 in amino acid sensing mechanisms of colon cancer, we used a colitis-associated cancer (CAC) colon tumor model. Wdr24^ΔIE and littermate controls were treated with azoxymethane and cycling of dextran sulfate sodium (AOM/DSS) which recapitulates many features of human CAC²¹ (Fig 3F). Wdr24^ΔIE mice had reduced tumor number, size (Fig 3G, 3H and 3I) and tumor burden (Fig 3J) compared to littermates. Tumor tissue proliferation (Fig 3K) and tumoral mTORC1 activation (Fig 3L) were significantly decreased in the Wdr24^ΔIE mice compared to littermate controls. Interestingly, in an acute colitis model, loss of Wdr24 increased susceptibility to colitis (Fig S3E–3I). As mTORC1 is important in intestinal repair responses²², these findings suggest that inactivating the amino acid sensing pathway potentiates colitis but is sufficient to reduce colon tumors.

Figure 3: Disruption of intestinal epithelial Wdr24 decreases mTORC1 and colon tumors.

(A) Wdr24 expression in scraped duodenum (small intestine) and colons of Wdr24^+/+ and Wdr24^ΔIE mice by qPCR (n = 3). (B) Colonic H&E images of Wdr24^+/+ and Wdr24^ΔIE mice aged 6–8 weeks (n = 3 – 5 mice per group). (C) Ki67 staining and quantitation in colons of Wdr24^+/+ and Wdr24^ΔIE mice (n = 3). (D) pS6 staining and quantitation of colons of Wdr24^+/+ and Wdr24^ΔIE mice (n = 3). (E) Wdr24^+/+ and Wdr24^ΔIE mice aged 6–8 weeks were sacrificed and colonic scrapes were probed (F) Schematic of azoxymethane/dextran sulfate sodium (AOM/DSS) colon tumor model (G) Gross colonic tumors and (H) H&E images of Wdr24^+/+ (n = 7) and Wdr24^ΔIE mice (n = 14) subjected to AOM/DSS model. (I) Quantitation of tumor numbers, size and (J) tumor volume in Wdr24^+/+ and Wdr24^ΔIE mice. (K) Ki67 staining and quantitation in tumor tissues of Wdr24^F/F and Wdr24^ΔIE mice subjected to AOM/DSS model (n = 3) (L) pS6 staining and quantitation in tumor tissues of Wdr24^+/+ and Wdr24^ΔIE mice subjected to AOM/DSS model (n =3). All pictures shown are Representative. *P < .05, **P < .01. Data are presented as mean ± SEM.

Dysregulation of intestinal epithelial amino acid sensing increases colon tumor growth.

As reduced expression of Depdc5 was associated with poor survival, we further examined the role of epithelial Depdc5 (GATOR1) in vivo. Mice with intestinal epithelial-specific knockout of Depdc5 were generated (Depdc5^ΔIE, Fig 4A). Similar to Wdr24^ΔIE mice, no histological differences were observed in the small intestine (Fig S4A) and colon (Fig 4B) of Depdc5^ΔIE mice compared to littermates. No significant differences in proliferation were observed in the small intestine (Fig S4B) and colon (Fig 4C). However, mTORC1 activation was significantly increased in the small intestine (Fig S4C and S4D) and colon (Fig 4D and 4E) of Depdc5^ΔIE mice. Lastly, no changes in the differentiated or stem cell markers were observed (Fig S4E). In a CAC model, Depdc5^ΔIE mice had increased tumor size (Fig 4F, S4F and 4G) and tumor burden (Fig 4H) compared to littermates. Tumor cell proliferation (Fig 4I) and tumoral mTORC1 activation (Fig 4J) were significantly increased in Depdc5^ΔIE mice. To further evaluate the role of Depdc5 in a sporadic colon tumor model, we crossed Depdc5^F/F with CDX2^ERT2 Apc^F/F (Fig 1D) to generate double knockout mouse model (CDX2^ERT2 Apc^F/FDepdc5^F/F) (Fig 4K). Increased early tumorigenesis throughout the entire colon in CDX2^ERT2 Apc^F/FDepdc5^F/F mice compared to proximal dysplasia in CDX2^ERT2 Apc^F/F was observed (Fig 4L, 4M and 4N). Tumor cell proliferation (Fig 4O) and tumoral mTORC1 activation (Fig S4G) were significantly increased in CDX2^ERT2 Apc^F/FDepdc5^F/F mice compared to CDX2^ERT2 Apc^F/F. In a tumor syngeneic model, disruption of Depdc5 increased mTORC1 and tumor weight (Fig S4H–S4K). These findings strongly indicate that disruption of GATOR1 drives colon early tumorigenesis.

Figure 4: Disruption of intestinal epithelial Depdc5 increases mTORC1 and colon tumors.

(A) Depdc5 expression in scraped duodenum (small intestine) and colons of Depdc5^+/+ and Depdc5^ΔIE mice by qPCR (n = 3). (B) Colonic H&E images of Depdc5^+/+ and Depdc5^ΔIE mice aged 6–8 weeks (n = 3 – 5). (C) Ki67 staining and quantitation in colons of Depdc5^+/+ and Depdc5^ΔIE mice (n = 3–4). (D) pS6 staining and quantitation in colons of Depdc5^+/+ and Depdc5^ΔIE mice (n = 3–4). (E) Depdc5^+/+ and Depdc5^ΔIE mice aged 6–8 weeks were sacrificed and colonic scrapes were probed. (F) Gross colonic tumors of Depdc5^+/+ (n = 7) and Depdc5^ΔIE mice (n = 9) subjected to AOM/DSS model. (G) Quantitation of tumor numbers, size and (H) tumor burden of Depdc5^+/+ (n = 7) and Depdc5^ΔIE mice (n = 9) subjected to AOM/DSS model. (I) Ki67 staining and quantitation in tumor tissues of Depdc5^+/+ and Depdc5^ΔIE mice (n = 4) (J) pS6 staining and quantitation of tumor tissues in Depdc5^+/+ (n = 3) and Depdc5^ΔIE mice (n = 4). (K) Schematic of sporadic colon tumor model in CDX2^ERT2 Apc^F/F and CDX2^ERT2 Apc^F/F/Depdc5^F/F mice. (L) Gross colonic and (M) H&E images (N) Total and stratified Dysplastic scores of CDX2^ERT2 Apc^F/F (n = 5) and CDX2^ERT2 Apc^F/F/Depdc5^F/F mice (n = 4). (O) Ki67 staining and quantitation of dysplastic tissue area in CDX2^ERT2 Apc^F/F (n = 3) and CDX2^ERT2 Apc^F/F/Depdc5^F/F mice (n = 4). All pictures shown are Representative. *P < .05, **P < .01, ***P < .001, ****P < .0001. Data are presented as mean ± the SEM

Dysregulated amino acid sensing leads to drug resistance in CRC.

5-fluorouracil (5FU) is the first line of chemotherapeutics in CRC²³. 5FU upregulates sestrins, which mediates DNA damage-dependent mTORC1 silencing through the amino acid sensing²⁴. To assess whether disruption of amino acid sensing affects 5FU response, Depdc5^−/− and control cells were treated with 5FU, and growth was assessed. Depdc5^−/− CRC cells displayed resistance to growth inhibition to 5FU treatment (Fig 5A). Pan-Akt inhibitor, MK2206 is in clinical trials for several cancers including metastatic CRC²⁵. Depdc5^−/− CRC cells displayed resistance to growth inhibition by MK2206 compared to control cells (Fig S5A). Remarkably, sustained Akt inhibition had minimal impact on mTORC1 levels in Depdc5^−/− cells (Fig S5B). To characterize the role of amino acid sensing in drug resistance, we generated CRC cells resistant to 5FU (Fig 5B) and MK2206 (Fig S5C) and assessed the levels of mTORC1 activity in 5FU and MK2206 resistant clones. Sensitive and resistant cells were exposed to different amino acid concentrations and examined for pS6K and pS6 levels. 5FU resistant (Fig 5C) and MK2206 resistant (Fig S5D) cells displayed higher mTORC1 activation. Moreover, Sestrin2 (Sesn2) and Depdc5 were markedly reduced in 5FU resistant clones compared to sensitive clones (Fig 5D). Interestingly, short exposure to 5FU (24h) did not alter Sesn2 levels in CRC cells (Fig 5E) suggesting that resistant cells may have deregulated the amino acid sensing pathway to drive mTORC1 and resistance. A similar finding was observed in MK2206 resistant cells (Fig S5E and S5F). These results suggest that colon tumors deregulate amino acid sensing mTORC1 pathway affecting chemotherapy response.

Figure 5: Drug resistance alters cellular amino acid sensing leading to constitutive mTORC1 activation in CRC.

(A) Depdc5^+/+ and Depdc5^−/− CRC cells treated with 5FU, and growth assessed by live cell imaging. (B) Schematic of generating 5FU resistant colonies. 5FU sensitive and resistant cells were treated with 5FU and growth was assessed by live cell imaging. (C) 5FU sensitive and resistant CRC cells were incubated in amino acid free, 0.1X, 0.2X, 0.5X and 1X of total amino acids for 24 hours (D) 5FU sensitive and resistant CRC cells were lysed and probed (E) CRC cells were treated with 5FU for 24 hours (F) OncoPrint displaying mutually exclusive genetic alterations of amino acid sensing and TP53 pathways. (G) Colonic enteroids from Apc^−/− and Apc^−/−/p53^−/− mice were incubated in amino acid free, 0.1X, 1X hours for 2 hours, probed **P < .01, ***P < .001, ****P < .0001. Data are presented as mean ± SEM.

TP53, Sestrin2 and amino acid-sensing components are in the same tumorigenic pathway

Sesn2²⁴ and Depdc5 (Fig 2H, 2I) expression are strongly downregulated in colon cancer, thus, we examined the frequency of their mutations in cancer genome. Sesn2 mutations are rare in human cancer, with only 1.5% of total colon cancer samples possessing detectable mutations in Sesn2 (Fig 5F) and nonsense truncation mutations of Sesn2 were around 40% of cases of the total Sesn2 mutations. Similarly, Wdr24 and Depcd5 mutations were found in 2.9% and 6% of human colon cancers, respectively (Fig 5F). Interestingly, alterations in these genes were mutually exclusive to TP53 mutations (Fig 5F). In contrast, traditional TP53 targets controlling cell cycle or apoptosis (CDKN1A, GADD45, BAX, BBC3 and TP53Aip1) did not exhibit significant mutual exclusiveness to TP53 (Fig 5F). To further understand the role of TP53 in amino acid sensing, colonic enteroids from Apc^−/− and Apc^−/−/p53^−/− mice were assessed for mTORC1 activation. Remarkably, mTORC1 activation was significantly induced under amino acid free or low amino acid conditions (0.125X) in Apc^−/−/p53^−/− compared to Apc^−/− and littermate controls (Fig 5G). These findings put the amino acid sensing-mTORC1 pathway into the context of classical tumor suppressive p53.

mTORC1 activation driven by amino acid sensing pathways lead to synthetic vulnerability under amino acid limiting conditions.

In the settings of limited nutrient supply, inhibiting anabolic outputs downstream of aberrant mTORC1 signaling creates a metabolic imbalance inducing cell death^26–28. We reasoned if dysregulated amino acid sensing driven aberrant mTORC1 activation could lead to cell death under nutrient stress. Impact of different total amino acid concentrations on the growth of CRC cells using colony forming assay were assessed. A dose response of total amino acids revealed 0.025X amino acids to be the lowest concentration that could sustain cell growth (Fig S6A). To test hypothesis that aberrant mTORC1 activation could lead to a cell death under amino acid limiting conditions (0.025X), we used small interfering RNA (siRNA)-based screening approach targeting genes known to activate mTORC1 (Fig 6A). Targeting Tsc1,Tsc2, Depdc5, Nprl2and Sesn2 significantly reduced growth under low amino acid (0.025X) conditions (Fig 6B). Cell death was induced in Depdc5^−/− CRC cells in 0.025X amino acid conditions, validating the data obtained from the screen (Fig 6C). Using colony forming growth assays, we observed that Depdc5^−/− cells displayed significant growth reduction compared to control cells in limiting amino acids (Fig 6D and S6B). Chronic activation of mTORC1 induces ER-stress and unfolded protein response (UPR)²⁹. Thus, we examined the expression level of markers of UPR signaling and observed a significant increase in ER-stress markers in Depdc5^−/− compared to controls under amino acid limiting conditions (Fig S6C). To investigate if aberrant mTORC1 activation could lead to vulnerabilities under amino acid limiting conditions in an in vivo model, we assessed CDX2^ERT2 Apc^F/FDepdc5^F/F sporadic model. Mice were induced with tamoxifen (50mg/kg) and randomly assigned control or low protein diet (Fig 6E). Remarkably, low protein diet significantly reduced early tumorigenesis in CDX2^ERT2 Apc^F/FDepdc5^F/F compared to control diet where tumorigenesis was observed through entire colon (Fig 6F–6H ). Tumor cell proliferation was significantly decreased (Fig 6I) while apoptosis being significantly increased (Fig 6J). As dysregulated amino acid sensing driven aberrant mTORC1 are insensitive to amino acid levels, mTORC1 activation was found to be unaltered in CDX2^ERT2 Apc^F/FDepdc5^F/F regardless of control or low protein diet (Fig S6D). As restriction of specific amino acids exert deleterious impact on tumor metabolism¹⁴, we observed that restriction of leucine and cystine exhibited reduced growth in Depdc5^−/− CRC compared to controls (Fig S6E). Altogether, these data indicate that aberrant mTORC1 activation via dysregulated amino acid sensing drives metabolic vulnerabilities to nutrient stress.

Figure 6: Disruption of Depdc5 leads to a metabolic vulnerability under amino acid limiting conditions.

(A) Schematic of small interfering RNA (siRNA)-based screen targeting known regulators of mTORC1. (B) Quantitation of cell growth to low amino acids (0.025X) by small interfering RNA (siRNA)-based screen. (C) Depdc5^+/+ and Depdc5^−/− CRC cells were incubated in low amino acids (0.025X) for 5 days and were imaged (D) Depdc5^+/+ and Depdc5^−/− CRC cells were incubated in amino acid free, 0.025X, 0.1X, 0.2X,0.5X and 1X and growth was assessed using colony forming assays at 10 days. (E) Schematic of sporadic colon tumor model and dietary regimen. (F) Gross colonic and (G) H&E images and (H) quantitation of total and stratified dysplastic tissue of CDX2^ERT2 Apc^F/F/Depdc5^F/F mice on control and low protein diet (n = 5 mice per group). (I) Ki67 staining image and quantitation of CDX2^ERT2 Apc^F/F/Depdc5^F/F mice on control and low protein diet (n = 3) (J) Cleaved caspase 3 staining image and quantitation of CDX2^ERT2 Apc^F/F/Depdc5^F/F mice on control and low protein diet (n = 3). All pictures shown are Representative. *P < .05, **P < .01, ***P < .001, ****P < .0001. Data are presented as mean ± SEM

CRC drug resistance provides a therapeutic metabolic vulnerability to low amino acids.

Next, we reasoned if limiting amino acids will render resistant cells susceptible to cell death. As expected, dysregulated amino acid sensing (Depdc5^−/−) CRC cells were resistant to 5FU as Depdc5^+/+ which could be reversed by addition of rapamycin or under 0.025X amino acids (Fig 7A, 7B). Importantly, resistant cells were rendered susceptible by rapamycin or low amino acid conditions following 5FU (Fig 7A, Fig 7B, S7A) or MK2206 treatment (Fig S7B and S7C). Sesn2 is an essential executioner of altering GATOR2 activity in the presence or absence of amino acids²⁴. Thus, we examined the correlation between Sesn2 expression and clinical response to standard of care chemotherapeutics in human CRC biopsies. Tumor biopsies were evaluated and stratified into poor responders or responders. We observed Sesn2 expression to be significantly increased in CRC tumors that responded to chemotherapeutics as opposed to patients who were poor responders (Fig 7C). Taken together, these results suggest dysregulated amino acid sensing drive drug resistance and tumor progression, however this can also lead to metabolic vulnerability under amino acid limiting conditions (Fig 7D).

Figure 7: Low amino acids induce cell death in drug resistant CRC cells.

(A, B) Depdc5^+/+ and Depdc5^−/− or 5FU sensitive and resistant CRC cells were incubated in 1X amino acid media, rapamycin (1μM) or 0.025X amino acid media with 5FU and growth was assessed using colony forming assay at 10 days. (C) Human colorectal cancer patients treated with chemotherapeutics were classified as responder’s (n = 11) vs non responders (n = 9) depending on treatment effect and were assessed for Sesn2. (D) Model of dysregulated amino acid sensing pathway driving growth but offering vulnerability under low dietary protein. All pictures shown are Representative. *P < .05, **P < .01, ***P < .001, ****P < .0001. Data are presented as mean ± SEM

Discussion

Here, we demonstrate an essential role of the amino acid sensing induce mTORC1 activation driving CRC. mTORC1 is a potent anabolic protein kinase and is hyperactivated in ~50% of all human cancers including CRC². Integrated inputs from both growth factor signaling, and nutrient availability is essential for its activation³⁰. The role of oncogenic growth factor signaling in mTORC1 activation and cancer progression is extensively characterized⁵ and therapeutic targeting of PI3K/Akt/mTORC1 signaling using PI3K inhibitors have demonstrated clinical success⁷. However, these therapies are associated with intrinsic or adaptive resistance limiting their efficacy in many cancers. Sustained mTORC1 activity is a major driver of resistance to PI3K/Akt inhibition^8,9. In line with this, we observed that loss of intestinal epithelial PI3K/Akt signaling (PI3K^ΔIE) in vivo marginally reduced mTORC1 activation suggesting existence of other mechanisms to sustain mTORC1. Amino acids potently activate mTORC1 via amino acid sensing complexes¹¹ and our findings suggest that modulating GATOR complexes is sufficient to impact cancer growth. As sustained mTORC1 drives drug resistance, these findings also imply that colon tumor cells may deregulate amino acid sensing pathway to activate mTORC1 and drive resistance. Strikingly, dysregulated amino acid sensing CRC cells (Depdc5^−/−) were resistant to 5-fluorouracil and Akt inhibitor MK2206. Furthermore, 5FU and MK2206 resistant CRC cells displayed reduced Sesn2 levels and heightened mTORC1. It was previously shown that 5FU can produce heterogeneous transcriptome response³¹, which may have contributed to the selection of cells that are resistant to cell growth by downregulating Sesn2. These findings also extended to CRC patients whereby drug responders displayed increased tumoral Sesn2 expression compared to poor responders.

Dietary changes can alter nutrients accessible to tumors, impacting tumor cell metabolism and affecting tumor growth, proliferation and survival¹⁴. Protein restriction inhibits tumor growth^12,13. Accordingly, our data suggests that dietary protein restriction reduced early tumorigenesis in a sporadic colon tumor model. Although, our findings indicate that this effect is likely mediated via reduced tumoral mTORC1 signaling, the impact of low protein diet on different components of the tumor microenvironment cannot be ruled out. mTORC1 impacts immune cell metabolism and function³². A defective amino acid induced mTORC1 signaling in Tregs was found to reduce accumulation of Tregs in tumors, enhancing antitumor immune responses³³. Low protein diet was shown to induce antitumor immune responses via IRE1α-dependent UPR in cancer cells³⁴. Thus, it is tempting to speculate that the beneficial effects of low protein diet in reducing tumor growth are due to combined effect on tumor and immune cell metabolism with tumoral mTORC1 being the major contributor.

Mutations in the amino acid sensing arm have been observed in human cancers such as glioblastoma, ovarian and gastric, they are less frequent than signaling pathways which induce mTORC1^24,35. Sesn2 and Depdc5 account for less than 8% of all mutations observed in CRC. This brings about an interesting question of why there is a low frequency of mutations in amino acid sensing pathways compared to oncogenic growth factor signaling? One possible explanation is that it would to deleterious for tumor cells to mutate Depdc5 and render cancer cells completely insensitive to amino acid limiting conditions of the tumor microenvironment. Moreover, deregulated amino acid sensing (mutant Depdc5) driven hyperactive mTORC1 could potently suppress activation of cellular autophagic pathways³⁶ and nutrient scavenging pathways³⁷, known to drive cell survival under nutrient stress. Other possible explanation is that both Sesn2 and Depdc5 are downstream to tumor suppressor TP53 mutations. A potent transcriptional regulator of Sesn2 is TP53^24,38. TP53-mutation in colon cancer cell line strongly downregulated Sesn2 expression²⁴ This suggest that amino acid sensing pathways are aberrantly upregulated in many cancers upon loss of TP53, due to the subsequent downregulation of Sesn2. This strategy might be particularly useful for tumor cells as it would allow sustained mTORC1 activity albeit at lower threshold to drive anabolic processes. This would also maintain the tumor ability to inhibit mTORC1 as opposed to mutating Depdc5 where tumors may lose this ability.

Tumors with heightened mTORC1 signaling are sensitive to rapalogs¹⁵ suggesting that rapalogs could be effective to treat tumors with amino acid sensing dysregulation. However, single use of rapalogs have shown limited therapeutic efficacy with patients requiring to remain on the regimen for long periods to prevent relapse and use of rapalogs may pose serious side effects¹⁰. Alternative strategies to rapalogs are needed to selectively target and kill tumor cells with aberrant mTORC1 signaling. We found that dysregulated amino acid sensing driven aberrant mTORC1 activation led to increased cell death under amino acid limiting conditions in vitro and in vivo. Furthermore, drug resistance could be reverted by limiting amino acids. We found these detrimental effects of low amino acids to be likely mediated by ER stress induced apoptosis. We speculate that aberrant mTORC1 activation in the setting of amino acid limiting conditions may put significant burden on protein synthesis machinery inducing unresolved ER stress and apoptosis. Thus, our findings suggest that low dietary protein induced alterations in tumor metabolism could be utilized as metabolic vulnerabilities in hyperactive mTORC1 driven tumors.

Active mTORC1 promotes anabolic processes via ribosome biogenesis, nucleotide, protein synthesis to drive growth and drug resistance⁵, the exact molecular underpinnings driven by deregulated amino acid sensing driven mTORC1 remains to be determined. The question remains what advantage cancer cells gain by deregulating amino acid sensing driven mTORC1 in the presence of oncogenic growth signaling? A recent study found genomic alterations of amino acid sensing genes (Depdc5, Nprl2 and Nprl3) in PIK3CA-mutated ER⁺ breast cancer limited drug responses³⁹. Interestingly, we observed Depdc5^−/− cells to be extremely resistant to 5FU. As 5FU targets nucleotide metabolism, one might argue that tumor cells may further boost nucleotide biosynthesis by deregulating amino acid sensing driven mTORC1 and limit 5FU responses. Further studies will be needed to dissect differences in oncogenic growth factor or amino acid sensing driven mTORC1 and pinpoint the exact molecular pathways driving tumor growth and resistance.

Dietary interventions such as ketogenic diet and caloric restriction inhibit tumor growth in preclinical models and offer benefit in clinical trials, however patient compliance remains a challenge¹⁴. Moreover, protein restriction increases risk of malnutrition and weight loss in cancer patients. Thus, one way to circumvent the use of low protein diets would be to restrict specific amino acids rather than entire protein content. Restriction of specific amino acids exerts deleterious impact on tumor metabolism revealing tumor specific metabolic vulnerabilities¹⁴. In line with this, our in vitro assays suggest that Depdc5^−/− cells are sensitive to amino acids leucine and cystine deprivation. As leucine is important in the amino sensing, diets low in leucine may mimic low protein diets and may further allow to precisely modulate amino acid dietary content. Tumoral pS6 levels predict drug sensitivity in patient biopsies^8,9, thus our findings suggest that evaluating amino acid sensing genes such as Sesn2, Wdr24 or Depdc5 would inform whether a patient would benefit by addition of low dietary protein to the therapy regimen. In summary, these are first studies showing an essential role of amino acid sensing pathways and further reinstate the importance of dietary protein intervention in CRC.

Abbreviations:

CRC

colorectal cancer

AOM

azoxymethane

DSS

dextran sulfate sodium

mTORC1

mechanistic target of rapamycin complex 1

GATOR

GTPase activity towards Rags complexes

Wdr24

WD Repeat Domain 24

Depdc5

DEP domain containing protein 5

5FU

5-Fluororuracil