Colorectal cancer is driven by mutations that activate canonical WNT/β-catenin signaling, but inhibiting WNT has significant on-target toxicity, and there are no approved therapies targeting dominant oncogenic drivers. We recently found that activating a β-catenin–independent branch of WNT signaling that inhibits GSK3-dependent protein degradation induces asparaginase sensitivity in drug-resistant leukemias. To test predictions from our model, we turned to colorectal cancer because these cancers can have WNT-activating mutations that function either upstream (i.e., R-spondin fusions) or downstream (APC or β-catenin mutations) of GSK3, thus allowing WNT/β-catenin and WNT-induced asparaginase sensitivity to be unlinked genetically. We found that asparaginase had little efficacy in APC or β-catenin–mutant colorectal cancer, but was profoundly toxic in the setting of R-spondin fusions. Pharmacologic GSK3α inhibition was sufficient for asparaginase sensitization in APC or β-catenin–mutant colorectal cancer, but not in normal intestinal progenitors. Our findings demonstrate that WNT-induced therapeutic vulnerabilities can be exploited for colorectal cancer therapy.
Solid tumors are thought to be asparaginase-resistant via de novo asparagine synthesis. In leukemia, GSK3α-dependent protein degradation, a catabolic amino acid source, mediates asparaginase resistance. We found that asparaginase is profoundly toxic to colorectal cancers with WNT-activating mutations that inhibit GSK3. Aberrant WNT activation can provide a therapeutic vulnerability in colorectal cancer.
See related commentary by Davidsen and Sullivan, p. 1632.
This article is highlighted in the In This Issue feature, p. 1611
Colorectal cancer remains the second leading cause of cancer-related death in the United States, and outcomes are dismal for patients with metastatic disease (1, 2). An estimated 96% of colorectal cancers have mutations that activate canonical WNT/β-catenin signaling (3), and these mutations promote intestinal transformation (4, 5). Despite a compelling rationale for therapeutic inhibition of this pathway (6), oncogenic β-catenin (also known as CTNNB1) activity is difficult to inhibit directly (7). The discovery that approximately 15% of colorectal cancers have mutations that drive ligand-dependent activation of WNT signaling, such as R-spondin (RSPO) fusions and RNF43 mutations (8–13), prompted considerable interest in therapeutic inhibition of WNT ligand activity. This is much more tractable pharmacologically, and a number of approaches targeting ligand-induced WNT pathway activation have been developed (reviewed in ref. 7). However, inhibiting WNT ligand activity leads to significant bone toxicity with pathologic fractures (14). This is an on-target toxicity also seen in patients with germline mutations of WNT ligands (15, 16). Although efforts to mitigate this toxicity are ongoing, whether inhibition of WNT signaling has a sufficiently favorable therapeutic index for cancer therapy remains unclear.
Asparaginase, an antileukemic enzyme that degrades the nonessential amino acid asparagine (17), has little activity in unselected patients with colorectal cancer (18–20), most of whom have APC mutations (3). We recently found that activation of WNT signaling upstream of GSK3 induces potent sensitization to asparaginase in drug-resistant acute leukemias, but not in normal hematopoietic progenitors (21). WNT-induced signal transduction is mediated by inhibition of the kinase GSK3 (22–24), and GSK3 inhibition was sufficient for asparaginase sensitization in leukemias. However, this effect appeared to be independent of APC or β-catenin. Instead, asparaginase sensitization was mediated by WNT-dependent stabilization of proteins (WNT/STOP), a β-catenin independent branch of WNT signaling that inhibits GSK3-dependent protein ubiquitination and proteasomal degradation (25). Proteasomal protein degradation is a catabolic source of amino acids (26, 27) required for asparaginase resistance in leukemia (21), and this adaptive response is blocked by WNT-induced inhibition of GSK3.
Colorectal cancer provides a unique experimental context in which to test predictions from our model, because mutations that arise spontaneously in colorectal cancer are predicted to unlink WNT/β-catenin from WNT-induced sensitization to asparaginase. Approximately 10% to 15% of colorectal cancers have mutations that activate WNT signaling upstream of GSK3, such as RSPO fusions (8, 9). These mutations are predicted to stimulate WNT ligand–induced inhibition of GSK3, thus resulting in activation of both β-catenin and WNT-induced sensitization to asparaginase. In contrast, approximately 85% of human colorectal cancers have mutations of genes such as APC or β-catenin (also known as CTNNB1; refs. 3, 28), which we predicted would selectively activate the β-catenin branch of WNT signaling downstream of GSK3, without activating WNT/STOP or inducing asparaginase sensitivity. The objective of this study was to test these predictions in the context of mutations that arise spontaneously in colorectal cancer.
WNT Pathway Activation Upstream of GSK3 Induces Asparaginase Hypersensitivity
To test whether ligand-induced WNT pathway activation induces asparaginase hypersensitivity in colorectal cancer, we began with the human APC-mutant colorectal cancer cell lines HCT15 and SW480 (Supplementary Table S1; refs. 29, 30). Treatment with asparaginase revealed that both of these cell lines were refractory to asparaginase monotherapy, but treatment with the recombinant ligands RSPO3 and WNT3A induced significant sensitization to asparaginase (Fig. 1A). WNT-induced signal transduction is mediated by inhibition of the kinase GSK3 (22–24), and treatment of these cells with CHIR-99021, a small-molecule inhibitor of both GSK3α and GSK3β (31) was sufficient to induce asparaginase sensitivity (Fig. 1B). Importantly, the combination of GSK3 inhibition and asparaginase had little toxicity to CCD841 cells derived from normal human colonic epithelium (Fig. 1C; ref. 32).
Mammalian cells have two GSK3 paralogs (GSK3α and GSK3β) that are redundant for regulation of canonical WNT/β-catenin signaling in several experimental contexts (33–35). However, we found that knockdown of GSK3α was sufficient for asparaginase sensitization in colorectal cancer, whereas GSK3β knockdown had little effect (Fig. 1D; Supplementary Fig. S1A and S1B). Treatment of HCT15 or SW480 cells with asparaginase in combination with GSK3α shRNA knockdown led to induction of caspase 3/7 activity, a marker of apoptosis induction (Supplementary Fig. S1C and S1D). We then used recently described isoform-selective GSK3 inhibitors to validate these findings pharmacologically (35). Treatment with the GSK3α-selective inhibitor BRD0705 sensitized both HCT15 and SW480 cells to asparaginase-induced cytotoxicity, whereas the GSK3β-selective inhibitor BRD3731 had little effect (Fig. 1E).
We then asked whether these findings are relevant in the context of endogenous mutations that arise spontaneously in colorectal cancer. Thus, we turned to genetically engineered mouse intestinal organoids designed to recapitulate the genetics of human colorectal cancer (6, 9, 36). The combination of KRAS, p53 (also known as TP53 in humans or Trp53 in mice), and a WNT/β-catenin–activating mutation is a common genotype in metastatic colorectal cancer (3); thus, we leveraged triple-mutant organoids harboring these mutations. The WNT-activating mutations were of distinct types: (i) Apc deficiency or a β-catenin (Ctnnb1) activating mutation (S33F), both of which we predicted would activate WNT/β-catenin without inhibiting GSK3 or activating WNT/STOP, and thus have no effect on asparaginase sensitivity; or (ii) an endogenous Ptprk–Rspo3 fusion that potentiates WNT ligand-induced inhibition of GSK3, which we predicted would activate both β-catenin and WNT/STOP, leading to asparaginase sensitivity. Treatment of these organoids revealed that asparaginase monotherapy was highly toxic to Rspo3 fusion organoids, whereas it had little activity against those that were Apc-deficient or β-catenin mutant (Fig. 1F–H). However, Apc or β-catenin mutant organoids were sensitized to asparaginase by cotreatment with the GSK3α inhibitor BRD0705 (Fig. 1G and H), indicating that inhibition of GSK3α is sufficient for asparaginase sensitization.
Asparaginase Sensitization Is Mediated by WNT-Dependent Stabilization of Proteins
We then asked how WNT pathway activation induces sensitivity to asparaginase. The E. coli–derived asparaginase we utilized has potent asparaginase activity and also degrades glutamine with lower affinity (approximately 2% of asparaginase activity; ref. 37). Thus, we first assessed whether WNT signaling represses expression of relevant amino acid metabolic enzymes or transporters. However, we found that treatment with WNT-activating ligands, the pan-GSK3 inhibitor CHIR-99021, or the GSK3α-selective inhibitor BRD0705 had no consistent effect on expression of asparagine synthetase, glutamine synthetase, or relevant amino acid transporters (ref. 38; Supplementary Fig. S2A–S2F).
We previously showed that drug-resistant leukemias tolerate asparaginase therapy by relying on GSK3-dependent protein ubiquitination and proteasomal degradation as a catabolic source of asparagine. This adaptive response is blocked by WNT-dependent stabilization of proteins (WNT/STOP; ref. 21), a β-catenin–independent branch of WNT signaling that inhibits GSK3-dependent protein degradation (23, 25, 39). Consistent with the β-catenin independence of asparaginase sensitivity, we found that the three approaches we used to trigger asparaginase sensitivity (WNT-activating ligands, the pan-GSK3 inhibitor CHIR-99021, or selective GSK3α inhibition; see Fig. 1) had disparate effects on activation of β-catenin (Supplementary Fig. S3) in SW480 cells, which express an APC allele that is partially but not completely impaired in its ability to inhibit β-catenin (29, 36).
To assess whether these perturbations activate WNT/STOP, we focused on its cellular hallmarks, which are an increase in cell size and an increase in total cellular protein half-life (25). Our model is that these effects should be particularly striking in the context of asparaginase therapy, when catabolic protein degradation is mediating asparaginase resistance. Indeed, we found that asparaginase therapy significantly decreased cell size in the human colorectal cancer cell line HCT15, and this effect was reversed by treatment with WNT-activating ligands (Fig. 2A). Asparaginase also reduced cell size in Apc-deficient; Kras; Trp53-mutant mouse intestinal organoids, and this effect was blocked in organoids expressing an Rspo3 fusion (Fig. 2B). We also asked whether expression of an Rspo3 fusion increases total cellular protein half-life in organoids, using a pulse-chase experiment with the methionine analogue azidohomoalanine (AHA). During the pulse period, there was no significant difference in the rate of labelled methionine incorporation in Rspo3-fusion versus Apc-deficient organoids (Supplementary Fig. S4). However, total cellular protein half-life was increased by approximately 1.8-fold in Rspo3 fusion versus Apc-deficient organoids in the context of asparaginase therapy (Fig. 2C). To test whether WNT ligand–induced sensitization to asparaginase is mediated by WNT/STOP, we first leveraged the fact that overexpression of the E3 ubiquitin ligase FBXW7 restores the degradation of a subset of proteins stabilized by WNT-induced inhibition of GSK3 (25). We found that the toxicity of asparaginase combined with the GSK3α inhibitor BRD0705 to Apc-mutant organoids was reversed by overexpression of wild-type FBXW7, but not by an FBXW7 R465C point mutant allele that is impaired in its ability to bind its protein substrates (Fig. 2D; ref. 40). In addition, the toxicity of this combination was reversed by expression of a hyperactive mutant of the proteasomal subunit PSMA4 (Fig. 2E), which directly stimulates proteasomal degradation of a range of proteasomal substrates (41). Thus, activation of WNT/STOP induces asparaginase sensitization in colorectal cancer.
If WNT/STOP activation induces sensitivity to asparaginase by impairing access to amino acids via catabolic protein degradation, then the toxicity of this combination should be rescued by replenishing the relevant amino acid(s). The asparaginase used in our studies (pegaspargase) is a PEGylated form of E. coli asparaginase, which has potent asparaginase activity and low but not absent glutaminase activity (37). Treatment of HCT15 cells with asparaginase led to profound depletion of asparagine, but had no significant effect on glutamine levels (Supplementary Fig. S5). We then asked whether cell death in response to GSK3α inhibition and asparaginase is caused by depletion of asparagine. HCT15 colorectal cancer cells were transduced with a GSK3α-targeting shRNA and treated with asparaginase, and we then added back a 10-fold excess of asparagine, glutamine, or vehicle control every 12 hours. This revealed that replenishing asparagine completely rescued colorectal cancer cells from the toxicity of GSK3α inhibition and asparaginase (Supplementary Fig. S6). In contrast, adding glutamine alone had no effect, and the combination of asparagine and glutamine was no better than asparagine alone. Thus, the combination of WNT/STOP activation and asparaginase is toxic to colorectal cancer cells due to depletion of asparagine.
To gain insights into the biological basis for the tumor-selective toxicity of this combination, we first assessed expression of asparagine synthetase and glutamine synthetase, but found no differences in their basal expression levels in normal versus malignant intestinal cells (Supplementary Fig. S7). We then leveraged an allelic series of mouse intestinal organoids that either were wild-type, had a single WNT-activating mutation of Apc or Rspo3 without other oncogenic mutations, or had these WNT-activating mutations in combination with a Kras-activating (G12D) and a Trp53-inactivating mutation. Treatment with the combination of asparaginase and GSK3α inhibition revealed little toxicity to wild-type organoids or to those harboring single WNT-activating mutations of either Apc or Rspo3, but the combination was significantly more toxic to those that also had Kras and Trp53 mutations (Supplementary Fig. S8A–S8C). Kras activation and Trp53 loss can both negatively regulate autophagy (42–44), raising the possibility that normal cells may tolerate asparagine starvation by relying on autophagy-mediated protein degradation as an alternative catabolic source of asparagine. This adaptive response may be impaired in colorectal cancer cells as a result of oncogenic mutations. To assess this possibility, we first measured levels of the autophagy marker P62 (also known as SQSTM1) by Western blot analysis in APC-deficient organoids that were either wild-type or mutant for KRAS and p53. P62 is degraded by autophagy; thus, levels of this protein are inversely correlated with the rate of autophagy (45). P62 levels were markedly increased in KRAS/p53–mutant organoids, with or without asparaginase therapy, suggesting that these mutations impair autophagy in colorectal cancer cells (Supplementary Fig. S8D). Moreover, treatment with inhibitors of lysosomal protein degradation, which block autophagy-induced protein degradation (45), had no effect on sensitivity to GSK3α inhibition and asparaginase in colorectal cancer cells, but did significantly sensitize CCD841 cells derived from normal human intestinal epithelium (Supplementary Fig. S8E–S8G). Inhibiting lysosomal protein degradation blocks not only autophagy but also macropinocytosis, which provides an alternative catabolic source of amino acids via endocytosis and degradation of extracellular proteins (46, 47). To distinguish these possibilities, we used shRNA knockdown of Beclin-1 to inhibit autophagy genetically (45), and the small-molecule EIPA as a selective inhibitor of macropinocytosis (46, 48). This revealed that Beclin-1 knockdown phenocopied the ability of lysosomal protein degradation inhibitors to stimulate asparaginase sensitivity in CCD841 cells, whereas inhibiting pinocytosis with EIPA had a more modest effect (Supplementary Fig. S8H–S8J). These data suggest that normal cells rely at least in part on autophagy as an alternative catabolic source of asparagine to tolerate asparaginase therapy, an adaptive response impaired in colorectal cancer cells.
Therapeutic Activity of Asparaginase in Colorectal Cancers with Upstream WNT Pathway Mutations
We then asked whether asparaginase has selective in vivo toxicity to colorectal cancers with “upstream” WNT pathway mutations that stimulate WNT-induced signal transduction upstream of GSK3, which inhibits GSK3 (22–24). We generated subcutaneous tumors in immunodeficient nude mice injected with triple-mutant mouse intestinal organoids that had Kras and Trp53 mutations, together with either Apc deficiency or an Rspo3 fusion. Once tumors engrafted (defined as growth to a volume >100 mm3), mice were randomized to treatment with vehicle or a single dose of asparaginase (Fig. 3A). Asparaginase had little effect on Apc-deficient tumors, but had significant therapeutic activity against Rspo3 fusion tumors. Indeed, asparaginase therapy not only markedly delayed disease progression in Rspo3 fusion tumors (Fig. 3B), but also induced tumor regression in most treated mice (Fig. 3C), and prolonged progression-free survival (Fig. 3D), without inducing appreciable weight loss (Supplementary Fig. S9).
We noted that the Rspo3 fusion tumors progressed approximately 2 weeks after the asparaginase dose, which coincides with the waning of asparaginase activity following a single dose in mice (49). To distinguish whether tumor progression reflected loss of asparaginase activity or the development of resistance, mice with Rspo3; Kras; Trp53 tumors were rechallenged with a second dose of asparaginase after tumor regrowth, which retained activity (Supplementary Fig. S10A). Furthermore, treatment of a cohort of mice engrafted with new Rspo3; Kras; Trp53 fusion tumors using three doses of asparaginase dosed every 12 days revealed that repeated asparaginase dosing can provide sustained disease control (Supplementary Fig. S10B).
GSK3a Inhibition and Asparaginase for Colorectal Cancers with APC or b-Catenin Mutations
Our model predicted that tumors with WNT pathway mutations that selectively activate β-catenin without directly inhibiting GSK3 should exhibit in vivo asparaginase resistance, unless GSK3α was also inhibited. To test this prediction, we began by generating subcutaneous tumors from mouse intestinal organoids harboring a β-catenin activating (S33F) mutation (36), as well as Kras and Trp53 mutations. At the time of tumor engraftment, mice were randomized to treatment with vehicle, asparaginase, the GSK3α inhibitor BRD0705, or asparaginase in combination with BRD0705 (Fig. 4A). These β-catenin–mutant tumors proved refractory to monotherapy with either asparaginase or BRD0705, despite effective inhibition of GSK3α autophosphorylation by BRD0705 (Supplementary Fig. S11A). However, the combination had significant activity (Fig. 4B–E).
We then asked whether GSK3α inhibition would induce asparaginase sensitization in the setting of metastatic colorectal cancer, which most commonly involves the liver (50, 51). Thus, we leveraged a model of liver metastatic colorectal cancer generated via intrasplenic injection of genetically engineered mouse intestinal organoids (52). Triple-mutant mouse intestinal organoids harboring a WNT-activating mutation of either Apc, β-catenin, or Rspo3, together with mutations of Kras and Trp53, were injected intrasplenically into distinct cohorts of mice. Five days after injection, mice were randomized to treatment with either vehicle, asparaginase, the GSK3α inhibitor BRD0705, or both drugs in combination (Fig. 4F). Metastatic engraftment to the liver was assessed by measuring liver weights in sentinel mice, which were euthanized 28 days post-injection with Apc; Kras; Trp53-mutant organoids (Fig. 4G). We then followed mice in each treatment cohort for survival. Although the Rspo fusion organoids failed to engraft in any mouse (data not shown), Apc; Kras; Trp53 organoids yielded efficient tumor engraftment, and the combination of GSK3α inhibition and asparaginase had significant therapeutic activity in this model (Fig. 4F–H). β-catenin; Kras; Trp53 organoids yielded tumor engraftment in approximately 50% of injected mice, which impaired our statistical power; nevertheless, all of the mice treated with the combination of BRD0705 and asparaginase were alive and well at day 40 post-injection, whereas half of mice in each other treatment condition succumbed to disease by day 25 (Fig. 4I and J).
Therapeutic Activity of GSK3a Inhibition and Asparaginase in Colorectal Cancer Patient-Derived Xenografts
We then asked whether GSK3α inhibition could induce asparaginase sensitization in vivo using patient-derived xenograft (PDX) models of APC-mutant colorectal cancer. We engrafted nude immunodeficient mice with a human patient-derived colorectal cancer xenograft termed COCA8, which had a biallelic APC mutation as well as a KRAS-activating mutation (Supplementary Table S1). Following tumor engraftment, mice were randomized to treatment with vehicle, asparaginase (1,000 U/kg × 1 dose), the GSK3α inhibitor BRD0705 (15 mg/kg every 12 hours × 21 days), or the combination of asparaginase and BRD0705 (Fig. 5A). We confirmed that BRD0705 treatment inhibited GSK3α in vivo (Supplementary Fig. S11B), as assessed by GSK3 autophosphorylation (53). The combination treatment was well tolerated, with no appreciable weight loss and no clinical or laboratory evidence of common asparaginase toxicities such as hepatic or pancreatic toxicity (Supplementary Fig. S12A–S12H), although these immunocompromised mice were ill-suited for assessing risk of hypersensitivity reactions to asparaginase. Monotherapy with asparaginase or BRD0705 had no significant therapeutic activity, but treatment with the combination of asparaginase and BRD0705 had a potent effect on tumor growth, including tumor regression in all treated mice, and significant prolongation in progression-free survival (Fig. 5B–E). To confirm the generalizability of these findings, we leveraged a distinct PDX model, termed COCA9, which had mutations of APC, KRAS, and TP53 (Supplementary Table S1). Consistent with predictions from our model, monotherapy with either asparaginase or BRD0705 had little activity, whereas both drugs in combination had significant therapeutic activity, as assessed by tumor size and progression-free survival (Fig. 5F–J).
We show here that the WNT-induced therapeutic vulnerability to amino acid starvation can be exploited for colorectal cancer therapy (Fig. 6A and B). Inhibition of GSK3 is a key mediator of WNT-induced signal transduction (22–24); thus, this kinase is predicted to be endogenously inhibited in a subset of colorectal cancers as a consequence of mutations that stimulate WNT ligand–induced pathway activation. On the basis of our previous work in leukemia, we predicted that these cases would be selectively sensitized to asparaginase. Indeed, we found that colorectal cancers with RSPO3 fusions, a recurrent oncogenic alteration in colorectal cancer that potentiates WNT ligand activity (8–10, 13, 54–56), were profoundly sensitive to asparaginase monotherapy. Our model predicts that colorectal cancers with other upstream WNT-activating mutations, such as RSPO2 fusions or mutations of the RSPO receptor RNF43, should also be asparaginase sensitive, as long as these mutations result in effective inhibition of GSK3. These findings also suggest the need to test asparaginase in other tumor types with mutations predicted to stimulate WNT-induced inhibition of GSK3, such as RNF43-mutant pancreatic, endometrial, and gastric cancers (11, 57, 58) and G9a-mutant melanomas (59).
We found that APC-mutant colorectal cancers were refractory to asparaginase monotherapy, unless GSK3 function was inhibited in these tumors. Selective inhibition of GSK3α was sufficient for this effect. Although the ATP-binding pockets of GSK3α and GSK3β differ by a single amino acid, isoform-selective inhibitors of GSK3α can be developed (35), which provides a strategy to leverage this therapeutic interaction for the majority of patients with colorectal cancer, who have mutations of APC or β-catenin. Selective inhibition of GSK3α is expected to provide a significant safety advantage because GSK3α and GSK3β are redundant for regulation of β-catenin in several experimental contexts (33–35), and GSK3β is the predominant β-catenin regulator in some contexts (60). Thus, small molecules that inhibit both GSK3α and GSK3β are expected to have toxicity due to widespread activation of β-catenin signaling, which is oncogenic. Moreover, GSK3β deficiency is embryonic lethal due to liver degeneration (61), and the liver is a target organ of asparaginase toxicity (62). In contrast, GSK3α-deficient mice are viable and have no known tumor predisposition (63). Thus, we expect isoform-selective inhibition of GSK3α to be better tolerated in combination with asparaginase.
We found that the combination of GSK3α inhibition and asparaginase was potently toxic to oncogenically transformed colorectal cancer cells but had little toxicity to normal intestinal cells or to mouse intestinal organoids harboring a single WNT-activating mutation. In contrast, this combination was more toxic to intestinal organoids that also had mutations of KRAS and p53. These findings suggest that KRAS and p53 mutations impair one or more adaptive responses that allow normal cells to tolerate the combination of GSK3α inhibition and asparaginase. One possibility was autophagy, which allows cells to tolerate starvation via lysosomal degradation of organelles and macromolecules (45), and which can be negatively regulated by KRAS activation and p53 loss (42–44). Indeed, small-molecule inhibitors of lysosomal protein degradation, which block autophagy-induced amino acid release, or knockdown of the autophagy factor Beclin-1, sensitized cells derived from normal human intestine to the toxicity of GSK3α inhibition and asparaginase. These findings suggest that the ability of normal cells to tolerate asparaginase is in part due to their ability access asparagine via autophagy. However, additional factors may also contribute to the resistance of normal cells to asparagine starvation, such as checkpoints that might trigger proliferative arrest in response to asparagine depletion or an improved capacity for de novo asparagine synthesis. Defining the precise molecular mechanisms that account for the tumor-selective toxicity of this therapeutic combination is of interest for future investigation.
Taken together with our recent work in leukemia (21), our data indicate that GSK3α-dependent protein degradation is required for asparaginase resistance in colorectal cancer and in drug-resistant leukemia. Given that excessive protein degradation is likely to antagonize cell growth in nutrient-rich conditions, tumor cell fitness may be maximized by selective induction of protein degradation in response to amino acid starvation. Deciphering the molecular regulation of this adaptive response will require additional investigation. These studies lead to the unexpected conclusion that mechanisms of intrinsic asparaginase resistance in solid tumors can overlap with those of acquired resistance in leukemia, and are fundamentally distinct from those that allow normal cells to tolerate asparaginase. Given that the therapeutic interaction of GSK3α inhibition and asparaginase is selectively toxic to tumors derived from cellular lineages that are as diverse as intestinal epithelium and hematopoietic cells (21), this approach could have meaningful therapeutic activity in a broad range of human cancers, as long as these rely on GSK3α-dependent protein degradation to tolerate treatment with asparaginase.
All asparaginase experiments were performed using pegaspargase (Oncaspar; Shire Pharmaceuticals), an FDA-approved PEGylated form of E. coli asparaginase. BRD0705 and BRD3731 were synthesized as described previously (35). CHIR-99021 was obtained from Selleckchem. Bafilomycin, chloroquine, EIPA, and ammonium chloride were purchased from Sigma-Aldrich.
Specimens were collected from patients with APC-mutant colon cancer (per methods described in ref. 64), with written informed consent in accordance with the Declaration of Helsinki, and approval of the Dana-Farber Cancer Institute Institutional Review Board. PDXs were generated by subcutaneous implantation into immunodeficient mice, as described in ref. 64. Mouse studies were performed in accordance with all regulatory standards and approved by the Boston Children's Hospital Institutional Animal Care and Use Committee. PDX models were genotyped using a clinical genotyping assay based on exon/fusion capture and next-generation sequencing, as described previously (65).
Cell Lines, Cell Culture, and Organoid Culture
293T cells, colorectal cancer cell lines, and normal colon cells were purchased from ATCC or DSMZ and cultured in DMEM, RPMI 1640, or Leibovitz's L-15 media (Thermo Fisher Scientific) with 10% or 20% FBS (Sigma-Aldrich) or TET system approved FBS (Clontech) and 1% penicillin/streptomycin (Thermo Fisher Scientific) at 37°C, 5% CO2.
Organoids carrying Ptprk–Rspo3 fusions and LSL–KrasG12D were derived from transgenic mice (9). Ptrpk–Rspo3 rearrangement was selected by culturing organoids without exogenous RSPO for 7 days, and validated by sequencing of the Ptprk–Rspo3 fusion junction. Targeted Apc (Q884X) and Ctnnb1 (S33F) mutations were generated by base editing, as described previously (36). KRASG12D activation and p53 loss were induced by delivery of a Cas9-P2A-Cre lentivirus (66) and p53-targeting guide RNA (67). For selection of p53 loss, organoids were cultured with 5 μmol/L Nutlin-3 for 7 days, and KRASG12D activation was indirectly selected during this selection. p53 loss was validated by sgRNA targeting site sequencing and Western blot analysis, and KRASG12D activation was validated by RNA sequencing.
For maintenance of mouse intestinal organoids in culture, organoids were resuspended in Matrigel composed of 25% advanced DMEM/F12 (Gibco) and 75% Matrigel (Corning), and after the Matrigel polymerized, organoids were cultured in organoid basal medium supplemented with growth factors (complete organoid growth medium). Organoid basal medium was advanced DMEM/F-12 (Thermo Fisher Scientific) with 1% penicillin/streptomycin, 2 mmol/L l-Glutamine (Sigma-Aldrich), 1 mmol/L N-acetylcysteine (Sigma-Aldrich), and 10 mmol/L HEPES (Sigma-Aldrich). Apc-deficient organoids were cultured in complete large-intestinal organoid growth medium, consisting of organoid basal medium supplemented with murine WNT3A (50 ng/mL, Merck Millipore), murine Noggin (50 ng/mL), murine EGF (R&D Systems, 50 ng/mL), and human RSPO1 (R&D Systems), as described previously (68). RSPO3-fusion organoids were cultured in complete small-intestinal organoid growth medium, consisting of organoid basal medium supplemented with murine Noggin (50 ng/mL), murine EGF (R&D Systems, 50 ng/mL), and human RSPO1 (R&D Systems), as described previously (68). Because our hypothesis was that WNT and RSPO ligands would induce asparaginase sensitivity, all experiments involving asparaginase treatment were performed using organoid basal medium, without growth factor supplementation.
Cell line identities were validated using short tandem repeat profiling at the Dana-Farber Cancer Institute Molecular Diagnostics Laboratory (most recently in December 2018), and Mycoplasma contamination was excluded using the MycoAlert Mycoplasma Detection Kit according to the manufacturer's instructions (most recently in November 2018).
Nude (J:NU) mice were purchased from the Jackson Laboratories (Stock # 0007850). Seven- to 9-week-old male nude mice were used for experiments and littermates were kept in individual cages. Mice were randomly assigned to experimental groups and handled in strict accordance with Good Animal Practice as defined by the Office of Laboratory Animal Welfare. All animal work was done with Boston Children's Hospital (BCH) Institutional Animal Care and Use Committee approval (protocol # 18-09-3784R).
Lentiviral Transduction of Colon Cancer Cell Lines
Lentiviruses were generated by cotransfecting pLKO.1 plasmids of interest together with packaging vectors psPAX2 (a gift from Didier Trono; Addgene plasmid # 12260) and VSV.G (a gift from Tannishtha Reya; Addgene plasmid # 14888) using OptiMEM (Invitrogen) and and Fugene (Promega), as described previously (69).
Lentiviral infections were performed by spinoculating colorectal cancer cell lines with virus-containing media (1,500 × g × 90 minutes) in the presence of 8 μg/mL polybrene (Merck Millipore). Selection with antibiotics was started 24 hours after infection with puromycin (1 μg/mL for a minimum of 48 hours; Thermo Fisher Scientific) or blasticidin (15 μg/mL for a minimum of 5 days; Invivogen).
Lentiviral Transduction of Mouse Intestinal Organoids
Prior to lentiviral transduction, a full 0.95 cm2 well of mouse intestinal organoids was harvested by pipetting up and down the Matrigel and complete organoid growth medium. Briefly, disrupted organoids were centrifuged at 300 × g for 5 minutes and the cell pellet was resuspended in 250 μL of cold 0.25% trypsin (Thermo Fisher Scientific) and incubated for 5 minutes at 37°C. Subsequently, trypsin was inactivated by adding 750 μL complete organoid growth medium and centrifuged (300 × g × 5 minutes). Cells were resuspended in 250 μL of concentrated lentivirus supplemented with 8 μg/mL polybrene (Merck Millipore). For lentiviral infection, the organoid virus mixture was incubated for 12 hours at 37°C, 5% CO2. Subsequently, 750 μL of complete organoid growth medium was added to the well, and the mixture was centrifuged at 300 × g for 5 minutes. The pellet was resuspended in 40 μL of ice-cold Matrigel, and 250 μL of complete organoid growth medium was added to each well after Matrigel solidification. Selection with antibiotics was started 24 hours after infection.
shRNA and Expression Plasmids
The following lentiviral shRNA vectors in pLKO.1 with puromycin resistance were generated by the RNAi Consortium library of the Broad Institute and obtained from Sigma-Aldrich: shLuciferase (TRCN0000072243), shGSK3α#1 (TRCN0000010340), shGSK3α#4 (TRCN0000038682), shGSK3β#2 (TRCN0000039564), shGSK3β#6 (TRCN0000010551).
Expression constructs expressing wild-type FBXW7 (also known as CDC4) or its R465C mutant were synthesized by gene synthesis and cloned into the pLX304 lentiviral expression vector by GeneCopoeia. A hyperactive open-gate mutant of the human proteasomal subunit PSMA4, termed ΔN-PSMA4, was designed by deleting the cDNA sequences encoding amino acids 2 to 10 (SRRYDSRTT) of PSMA4 isoform NP_002780.1 (encoded by the transcript NM_002789.6), based on the data of Choi and colleagues (41). This ΔN-PSMA4 coding sequence was synthesized by gene synthesis and cloned into the pLX304 lentiviral expression vector in-frame with the C-terminal V5 tag provided by this vector, by GeneCopoeia.
Assessment of Chemotherapy Response and Apoptosis in Colon Cancer Cell Lines
Cells (100,000 per well) were seeded in 1 ml of complete growth medium in 12-well plates and incubated with chemotherapeutic agents or vehicle. Cells were split every 48 hours and cell viability was assessed by counting viable cells based on Trypan blue vital dye staining (Invitrogen), according to the manufacturer's instructions. Chemotherapeutic drugs included asparaginase (pegaspargase), CHIR99021 (Selleckchem), recombinant human WNT3A protein (R&D Systems), and recombinant human R-Spondin3 protein (R&D Systems). BRD0705 and BRD3731 were synthesized as described previously (35). Caspase 3/7 activity was assessed using the Caspase Glo 3/7 Assay (Promega) according to the manufacturer's instructions.
Assessment of Chemotherapy Response in Mouse Intestinal Organoids
For assessment of chemotherapy response in mouse intestinal organoids, Apc-deficient and Rspo3-fusion organoids were cultured in organoid basal medium (in the absence of WNT3A, murine Noggin, and human RSPO1 protein). A full 0.95-cm2 well of organoids was split into new wells, aiming to obtain approximately 25 organoids per well. Organoids were split according to previously published protocols (68). Matrigel and basal organoid medium were supplemented with vehicle or chemotherapeutic agents and split every 48 hours. After 10 days in culture, total organoid numbers per well were counted manually by light microscopy. Dying organoids were distinguished by the drastic change in organoid morphology with loss of epithelial integrity and impaired lumen formation. Organoids touching the edge of a well were excluded from counting. Microscopy was performed using an 100× objective on an Axio Imager A1 microscope (Zeiss), with images captured using a CV-A10 digital camera (Jai) and Cytovision software (Leica Biosystems). Images were taken from a representative of three independent experiments and analyzed with ImageJ Software (70).
Assessment of Cell Size
Briefly, HCT15 cells (100,000 per well) were plated in 1 mL of complete growth medium, containing a final concentration of 100 U/L asparaginase or 100 ng/mL WNT3A ligand and 75 ng/mL RSPO3 ligand in a 12-well format. Indicated organoids were seeded in 250 μL of organoid growth medium supplemented with 100 U/L of asparaginase. After 48 hours of treatment, forward scatter height (FSC-H) was assessed by flow cytometry on a Beckton-Dickinson (BD) LSR-III or a BD FACS DIVA instrument.
Assessment of Response to Autophagy Inhibition
Cells (100,000 per well) were plated in 1 ml of complete growth medium, supplemented with vehicle (PBS) or 100 U/L asparaginase and 1 μmol/L BRD0705. Growth medium contained a final concentration of 100 nmol/L bafilomycin (Sigma Aldrich), 10 μmol/L chloroquine (Sigma Aldrich), or 20 mmol/L ammonium chloride (Sigma Aldrich). Cells were split after 48 hours, and cell viability was assessed after 5 days of treatment by counting viable cells based on Trypan blue vital dye staining (Invitrogen), according to the manufacturer's instructions.
Quantitative Reverse Transcriptase PCR
RNA was isolated using RNeasy Kit (Qiagen) and cDNA was made using SuperScript III First-Strand cDNA Synthesis Kit (Thermo Fisher Scientific). qRT-PCR was performed using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) and 7500 Real-time PCR System (Applied Biosystems). Primers used are described in Supplementary Table S2.
Western Blot Analysis
Cells were lysed in RIPA buffer (Merck Millipore) supplemented with cOmplete protease inhibitor (Roche) and PhosSTOP phosphatase inhibitor (Roche). Laemmli sample buffer (Bio-Rad), and β-mercaptoethanol (Sigma-Aldrich) were mixed with 20 μg of protein lysate before being run on a 4% to 12% bis-tris polyacrylamide gel (Thermo Fisher Scientific). Blots were transferred to polyvinylidene difluoride membrane (Thermo Fisher Scientific) and blocked with 5% BSA (New England Biolabs) in PBS with 0.1% Tween (Boston Bioproducts) and probed with the following antibodies: Non-phospho (active) β-catenin (Ser33/37/Thr41) antibody (1:1,000, Cell Signaling Technologies #8814), P62 antibody (1:1,000, Cell Signaling Technologies, #5114), phospho-GSK3 (Tyr279/216; 1:1,000, Thermo Fisher Scientific #OPA1-03083), or GAPDH (1:1,000, Cell Signaling Technologies #2118). Detection of horseradish peroxidase–linked secondary antibodies (1:2,000, Cell Signaling Technologies #7074S) with horseradish peroxidase substrate (Thermo Fisher Scientific) was visualized using a Bio-Rad GelDoc XR+ Imaging System.
Assessment of Protein Stability
Protein degradation was assessed using a nonradioactive quantification of the methionine analogue AHA AlexaFluor488 (Thermo Fisher Scientific), as described previously (71). Briefly, organoids were seeded in 250 μL of methionine-free DMEM. After 30 minutes, the pulse step was performed by replacing this medium with 250 μL DMEM supplemented with AHA at a final concentration of 50 μmol/L for 18 hours. In the chase step, cells were released from AHA by replacing media with DMEM containing 10× L-methionine for 2 hours. Subsequently, media were replaced with regular organoid growth medium and cells were treated with a final concentration of 100 U/L asparaginase, followed by fixation of cells. AHA-labeled proteins were tagged using TAMRA alkyne click chemistry, and fluorescence intensity was measured by flow cytometry. A sample without AHA labeling but TAMRA alkyne tag was included as a negative control to account for background fluorescence.
Amino Acid Quantification
HCT15 cells (100,000 per well) were seeded in 1 mL of complete growth medium in a 12-well format. The growth medium was supplemented with final concentrations of 100 U/L asparaginase. After 24 hours of treatment, medium was collected and stored at −80°C until amino acid quantification.
For intracellular amino acid extraction, cells were harvested and resuspended in 0.16 mol/L potassium chloride (Sigma Aldrich). After 10 minutes, a mix of leupeptin (1 μmol/L, Sigma Aldrich), Pepstatin (1 μmol/L, Sigma Aldrich), phenylmethylsulfonylfluoride (1 mmol/L, Sigma Aldrich), and EDTA (1 mmol/L, Sigma Aldrich) was added to the suspension and incubated on ice for an additional 10 minutes. Subsequently, cells were lysed by thaw-freeze cycling (15 minutes at −80°C followed by 60 minutes at 4°C). After 4 cycles, the suspension was deproteinized by adding sulfosalicylic acid (7 mg/mL, Sigma Aldrich) and stored at −80°C until amino acid quantification.
The entire amino acid profile was determined by means of LC/MS-MS.
Rescue of Asparaginase Sensitization with Amino Acid Supplementation
HCT15 cells were transduced with shRNAs (100,000 per well) and were seeded in 1 mL of complete growth medium, containing a final concentration of 100 U/L asparaginase (or PBS vehicle control) in a 12-well format. Growth medium (RPMI 1640 with 10% FBS) was supplemented with l-asparagine (Sigma-Aldrich) at a final concentration of 3.78 mmol/L (10×), l-glutamine at a final concentration of 20.5 mmol/L (10×), or with both 10× asparagine and 10× glutamine. Every 12 hours, 500 μL of complete growth medium was removed and replaced with 500 μL fresh growth medium, supplemented with the appropriate concentration of asparaginase. After 72 hours of treatment, viability was assessed by Trypan blue viable cell staining.
In Vivo Drug Treatment of Subcutaneous PDX and Organoids
For implantation of APC-mutant human colorectal cancer PDX, patient tumor material was collected in PBS and kept on wet ice for engraftment within 24 hours after resection. Upon arrival, necrotic and supporting tissues were carefully removed using a surgical blade. Approximately 1 mm × 1 mm tissue fragments were implanted subcutaneously into the flank region of male nude mice, as described previously (64). For injection of intestinal organoids (Rspo3; Trp53; Kras, or Apc; Trp53; Kras), per mouse, one full 9.5 cm2 well of organoids was injected subcutaneously.
Treatment was started when tumors reached a volume of approximately 150 mm3. For the APC-mutant human colorectal cancer PDX, a single dose of asparaginase (1,000 U/kg) or PBS was injected by tail-vein injection on day 1 of treatment, and BRD0705 (15 mg/kg) or vehicle was given every 12 hours for 21 days by oral gavage. Vehicle was formulated as described previously (35).
After start of treatment, body weight and tumor size were evaluated every other day. Tumor size was assessed by caliper measurements and the approximate volume of the mass was calculated using the formula (l × w × w) × (π/6), where l is the major tumor axis and w is the minor tumor axis. The response was determined by comparing tumor volume change at time t to its baseline: % tumor volume change = 100% × ((Vt – Vinitial)/Vinitial), as described previously (72). Mice were euthanized as soon as they reached a tumor volume of 1,500 mm3, developed weight loss greater than 15%, and/or showed signs of tumor ulceration. For Figs. 3B, 4B, and 5D and G, the last recorded tumor size of mice that died due to tumor progression was used for volume plots until the last mouse of the treatment group reached a tumor volume of 1,500 mm3.
To assess the potential hepatic or pancreatic toxicity of the combination of BRD0705 and asparaginase treatment, retro-orbital blood collections were performed, and liver function and pancreatic enzyme levels were measured in the Boston Children's Hospital clinical laboratory.
In Vivo Drug Treatment of Metastatic Mouse Intestinal Organoids
For injection of intestinal organoids (Apc; Trp53; Kras), per mouse one full 9.5 cm2 well of organoids was injected into the spleen, as described previously (73). Briefly, organoids were collected, resuspended in PBS, and kept on ice. Mice were anesthetized with isoflurane and the left subcostal area was prepped with 70% ethanol and iodine. A left subcostal incision was made in line with the left ear through the skin and the peritoneum using scissors. The spleen was expressed by pulling the caudal aspect of the spleen through the incision using tweezers, and a total volume of 50 μL of organoids in PBS was slowly injected into the exposed part of the spleen. Subsequently, the spleen was placed back into the peritoneum by applying digital pressure. The peritoneum was closed with 2 sutures, and the skin incision was closed by applying one skin clip.
Five days post-injection of organoids, treatment was started and consisted of one single dose of asparaginase (1,000 U/kg) or PBS given by tail-vein injection on day 1 of treatment, and BRD0705 (15 mg/kg) or vehicle was given every 12 hours for 21 days by oral gavage.
After start of treatment, body weight was evaluated every other day. Mice were euthanized as soon as they developed weight loss greater than 15% and/or showed signs of disease progression. Mice were harvested for postmortem analysis of liver weights.
Quantification and Statistical Analysis
For two-group comparisons of continuous measures, a two-tailed Welch unequal variances t test was used. For three-group comparisons, a one-way ANOVA was performed and a Dunnett adjustment for multiple comparisons was used. For analysis of two effects, a two-way ANOVA model was constructed and included an interaction term between the two effects. Post hoc adjustment for multiple comparisons for two-way ANOVA included Tukey–Kramer adjustment. The log rank test was used to test for differences in survival between groups, and the method of Kaplan–Meier was used to construct survival curves. Data shown as bar graphs represent the mean and SEM of a minimum of three biologic replicates, unless otherwise indicated. All P values reported are two-sided and considered as significant if <0.05.
Disclosure of Potential Conflicts of Interest
L. Hinze reports grants from NIH/NCI R01, Dana-Farber Cancer Institute, and ERA-NET Transcan/European Commission under the 7th Framework Programme (FP7) during the conduct of the study; in addition, L. Hinze has a patent for Methods of Treating Cancer pending. J. Degar reports grants from Dana-Farber Cancer Institute and grants from NIH during the conduct of the study. J.R. Sacher reports a patent for WO2018187630 issued. F. Wagner reports personal fees from Biotechnology company (as a consultant on a GSK3-related project) outside the submitted work; in addition, F. Wagner has a patent for PCT/US2013/064716 pending, a patent for US 9,096,594 issued, a patent for US 10,137,122 issued, and a patent for US 16/525,494 pending. K. Ng reports grants from NCI, Department of Defense, and Cancer Research UK during the conduct of the study, Revolution Medicines, Genentech, Gilead Sciences, Tarrex Biopharma; grants and nonfinancial support from Evergrande Group; nonfinancial support from Pharmavite; personal fees from Bayer, Seattle Genetics, and Array Biopharma outside the submitted work. A. Gutierrez reports grants from NCI and grants from Dana-Farber Cancer Institute during the conduct of the study; in addition, A. Gutierrez has a patent for PCT Patent Application PCT/US2019/041555 pending (patent submitted by Boston Children's Hospital) and a patent for US Provisional Patent Application 62/930,258 pending (patent submitted by Boston Children's Hospital). No potential conflicts of interest were disclosed by the other authors.
L. Hinze: Conceptualization, formal analysis, investigation, visualization, methodology, writing-original draft, writing-review and editing. R. Labrosse: Formal analysis, investigation, visualization, writing-review and editing. J. Degar: Formal analysis, investigation, writing-review and editing. T. Han: Methodology. E.M. Schatoff: Methodology. S. Schreek: Investigation. S. Karim: Investigation. C. McGuckin: Investigation. J.R. Sacher: Methodology. F. Wagner: Methodology. M. Stanulla: Formal analysis, investigation. C. Yuan: Formal analysis. E. Sicinska: Methodology. M. Giannakis: Formal analysis, investigation, writing-review and editing. K. Ng: Formal analysis, investigation, methodology. L.E. Dow: Investigation, methodology, writing-review and editing. A. Gutierrez: Conceptualization, formal analysis, supervision, visualization, writing-original draft, writing-review and editing.
We thank Kimberly Stegmaier, Daniel Bauer, Alex Kentsis, Scott Armstrong, Gabriela Zurek, Nikolaus Kuehn-Velten, Mark Kellogg, Timothy Hagan, Otari Chipashvili, and Sung-Yun Pai for advice and discussion, and Meaghan McGuinness and Casey O'Brien for experimental assistance. This work was supported by NIH/NCI R01 CA193651, the Boston Children's Hospital Translational Investigator Service, a Dana-Farber Cancer Institute Medical Oncology Translational Grant Award, and the ERA-NET Transcan/European Commission under the 7th Framework Programme (FP7). L. Hinze was supported by the German National Academic Foundation and the Biomedical Education Program. M. Giannakis was supported by a Conquer Cancer Foundation of ASCO Career Development Award, the Project P-Fund, the Cancer Research UK C10674/A27140 Grand Challenge Award, and a Stand Up To Cancer Colorectal Cancer Dream Team Translational Research Grant (grant number: SU2C-AACR-DT22-17). Stand Up To Cancer (SU2C) is a division of the Entertainment Industry Foundation, and research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C. E.M. Schatoff was supported by a Medical Scientist Training Program grant from the National Institute of General Medical Sciences of the NIH under award number T32GM07739 to the Weill Cornell/Rockefeller/Sloan-Kettering Tri-Institutional MD/PhD Program, and an F31 Award from the NCI/NIH under grant number 1 F31 CA224800-01. K. Ng was supported by NIH R01 CA205406 and DOD CA160344. A. Gutierrez was supported by a CHPA Investigatorship at Boston Children's Hospital.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.