WNT signalling promotes NF-κB activation and drug resistance in KRAS-mutant colorectal cancer

Abstract

Approximately 40% of colorectal cancer (CRC) cases are characterised by KRAS mutations, rendering them insensitive to most therapies. While the reasons for this resistance remain incompletely understood, one key aspect is genetic complexity: in CRC, oncogenic KRAS is most commonly paired with mutations that alter WNT and P53 activities (“RAP”). Here, we demonstrate that elevated WNT activity upregulates canonical NF-κB signalling in both Drosophila and human RAS mutant tumours. This upregulation was enhanced by P53 loss and required immune-associated factors Toll-1 and Toll-9. These changes reduced efficacy of Ras pathway-targeting drugs such as trametinib due to NF-κB-dependent enhancement of the glucuronidation detoxifying pathway, likely through modulating gene transcription and glucose uptake. Inhibiting WNT activity pharmacologically suppressed trametinib resistance in RAP tumours and more genetically complex ‘patient avatar’ models. The efficacy of WNT/MEK drug inhibitor combinations was further enhanced by targeting brm, shg, ago, rhoGAPp190, and upf1, potential biomarkers for patients responsive to this dual therapeutic approach. These findings shed light on how genetic complexity impacts drug resistance and a strategy to overcome it.

Synopsis

Elevated WNT signalling triggers a wound-like response in RAS cancer cells, activating TLRs-dependent canonical NF-κB signalling, which enhances glucuronidation pathway activity to rapidly clear trametinib from cells and promote drug resistance.

• Wnt signalling reduced trametinib efficacy by elevating TLRs-dependent canonical NF-κB signalling in RAP tumours.

• Canonical NF-κB activity impacted trametinib response by altering its glucuronidation.

• Combining trametinib and Wnt inhibitor PNU-74654 or LF3 suppressed tumour progression in RAP tumour and genetically complex ‘avatar’ tumours.

• Elevated WNT plus KRAS is associated with increased canonical NF-κB signalling in human CRC tumour samples.

Elevated WNT signalling triggers a wound-like response in RAS cancer cells, activating TLRs-dependent canonical NF-κB signalling, which enhances glucuronidation pathway activity to rapidly clear trametinib from cells and promote drug resistance.

Introduction

Colorectal cancer (CRC) remains a leading cause of cancer-related deaths worldwide. Progressive disease is characterised by uncontrolled cell growth, primarily driven by a complex array of genetic mutations (Sung et al, 2021; Stratton et al, 2009; Stratton, 2011). CRC cases featuring mutations in the RAS family of small GTPases have proven especially problematic due to a striking insensitivity to most targeted therapies in the clinics (Yaeger et al, 2018; Cremolini et al, 2015; Wang et al, 2022). Despite the development of successive generations of inhibitors targeting the RAS pathway that demonstrate promise in pre-clinical studies, these inhibitors have shown minimal or transient activity in RAS-mutant CRC patients. While resistance mechanisms such as elevated KRAS or EGFR activity can lead to emergent drug resistance in some tumours (Huang et al, 2021; Hofmann et al, 2022; Zhang et al, 2018), the primary mode(s) of resistance for many CRC tumours remain unclear.

One source of drug resistance is genetic complexity, a consistent and key hallmark of CRC (Caponigro and Sellers, 2011; Bangi et al, 2016; Wang et al, 2022). Experiments in Drosophila and mouse models report that oncogenic RAS/KRAS mutations alone are only sufficient to initiate benign tumours (Pagliarini and Xu, 2003; Calcagno et al, 2008). The progression to malignancy in human CRC requires acquisition of additional mutations, most commonly in P53 and the WNT regulator APC (Fearon and Vogelstein, 1990; Boutin et al, 2017; Bangi et al, 2016). Our recent studies involving Drosophila CRC models have also demonstrated that genetic complexity amplifies metastatic potential and, key to this report, fosters drug resistance (Bangi et al, 2016). The precise mechanisms that connect genetic complexity to drug resistance remain poorly understood.

Here, we examine the impact of mutated apc and p53 on Ras^G12V-expressing tumours in the Drosophila hindgut. We found that elevated WNT (Wg in Drosophila) activity led to upregulation of canonical nuclear factor-kappa B (NF-κB) signalling in Ras^G12V tumours, which in turn led to emergent resistance to drugs such as the MEK inhibitor trametinib by elevating glucuronidation pathway activity. This resistance was reversed by reducing WNT activity with inhibitor compounds PNU-74654 or LF3, restoring trametinib-mediated rescue. P53 also contributed to drug response by regulating tumour growth and NF-κB signalling in high Wnt /Ras^G12V hindgut tumours. Consistent with this data, human CRC samples with high WNT activity and oncogenic KRAS were associated with elevated canonical NF-κB signalling. Together, our data identify NF-κB as a key mediator of drug resistance in RAS-WNT-P53 CRC tumours, and suggest a novel approach to the treatment of KRAS-mutant CRC.

Results

Wnt signalling reduced trametinib efficacy by elevating canonical NF-κB signalling

RAS family proteins regulate key cellular processes through multiple pathways, including the canonical Raf-MEK-ERK (MAPK) signalling pathway. The FDA-approved drug trametinib is a potent and precise MEK inhibitor that, despite strong preclinical activity, failed to demonstrate significant efficacy in CRC patients (Nalli et al, 2021; Infante et al, 2012). Consistent with these reports, we observed that trametinib—delivered orally in the food—strongly rescued tumour-induced lethality in Drosophila CRC models that target Ras^G12V to the hindgut (Ras^G12V; Fig. 1A). However, trametinib failed to rescue a multi-targeting CRC model that combined oncogenic Ras^G12V plus RNAi (“i”) knockdown of Apcⁱ and P53ⁱ (RAP; Fig. 1A; all transgenes in this study are targeted to the hindgut via the byn-GAL4 transcriptional driver). Trametinib did not affect control animals (Fig. EV1A). These data indicate an emergent resistance to trametinib in RAP hindgut tumours.

Figure 1. WNT signalling induced trametinib resistance through enhancing canonical NF-κB signalling.

(A) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM). Control (DMSO, n = 18), Ras^G12V (DMSO, n = 42; tram, n = 39), RAP (DMSO, n = 27; tram, n = 56). (B) Expression levels of drosomycin were quantified for each genotype by quantitative RT-PCR. Control (n = 3) and RAP (n = 3). (C–F) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM). (C) RAP;difⁱ (DMSO, n = 9; tram, n = 9) and RAP;dlⁱ (DMSO, n = 11; tram, n = 16); (D) Control (DMSO, n = 17), difⁱ (DMSO, n = 10; tram, n = 10) and dlⁱ (DMSO, n = 16; tram, n = 13); (E) Ras^G12V;GFP (control (DMSO, n = 18; tram, n = 28)) and Ras^G12V;cactⁱ (DMSO, n = 16; tram, n = 20); (F) GFP (DMSO, n = 20; tram, n = 20) and cactⁱ (DMSO, n = 20; tram, n = 18). (G) Expression levels of drosomycin were quantified for each genotype by quantitative RT-PCR. Ras^G12V (n = 3), arm^S10 (n = 3), and Ras^G12V;Arm^S10 (n = 3). (H–J) Images of the digestive tract of third instar larvae in the presence of trametinib (1 µM), which include the hindgut proliferation zone (HPZ). Nuclei are visualised with 4′,6-diamidino-2-phenylindole (DAPI) staining, and the hindgut is visualised by GFP. Scale bar 200 µm. (K) The average size of the hindgut proliferation zone (HPZ) size was measured by Fiji ImageJ and quantified as a relative size to the wild-type hindgut. All transgenes were expressed in the hindgut using byn-GAL4. RAP (DMSO, n = 13; tram, n = 13), RAP;dlⁱ (tram, n = 10), RAP;difⁱ (tram, n = 7). (A, B, E–J) The experiments were conducted at 27 °C. (C, D) The experiments were conducted at 29 °C. The statistical tests used to calculate the P value are as follows: (A, C–G, K) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. Source data are available online for this figure.

Figure EV1. Overgrowth of HPZ was driven by combined Ras, Wg and p53 alterations.

(A) Percent survival of control flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM). Control (DMSO, n = 12; tram, n = 12). (B–D) The average of the hindgut proliferation zone (HPZ) size was measured by Fiji ImageJ and quantified as relative size to the control hindgut. Control (DMSO, n = 12), Ras^G12V (DMSO, n = 13; tram, n = 12) (B); Control (DMSO, n = 15), Ras^G12V (DMSO, n = 13), Arm^S10 (DMSO, n = 8), Ras^G12V;Arm^S10 (DMSO, n = 17) (C); Control (DMSO, n = 15), Ras^G12V (DMSO, n = 13), RAP (DMSO, n = 13), Ras^G12V;p53ⁱ (DMSO, n = 15) (D). (E–K) Images of the digestive tract of third instar larvae in the presence or absence of trametinib (1 µM). Control (E), Ras^G12V (F, G), Arm^S10 (H), Ras^G12V;Arm^S10 (I), Ras^G12V;p53ⁱ (J) and RAP (K). Scale bar 200 µm. (A–K) The experiment was conducted at 27 °C. The statistical tests used to calculate the P value are as follows: (A) Mann–Whitney test; (B–D) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates.

In a screen for pathways such as ABC transporters, JNK signalling and autophagy that distinguish Ras^G12V and RAP models, we found that canonical NF-κB pathway activity—also known as the Toll pathway in Drosophila (Meng et al, 1999) was increased in RAP models (Fig. 1B). Previous studies have shown that inhibition of NF-κB signalling increases sensitivity of HTC15 human colon cancer cells to the chemotherapeutic daunomycin by modulating drug uptake (Bentires-Alj et al, 2003). Interestingly, we found that inhibition of canonical NF-κB signalling by targeted knockdown of NF-κB family members dorsal-related immunity factor (dif) or dorsal (dl) (Minakhina and Steward, 2006) strongly improved trametinib’s ability to rescue RAP-induced lethality (Fig. 1C). Knockdown did not affect the survival of control animals or RAP animals in the absence of trametinib (Fig. 1C,D). Conversely, elevating canonical NF-κB activity by targeted knockdown of the NF-κB inhibitor cactus (cact, an IκB orthologue) (Geisler et al, 1992; Kidd, 1992) reduced trametinib efficacy in Ras^G12V tumours (Fig. 1E). Knockdown of cact did not affect lethality in the absence of trametinib or in control animals (Fig. 1E,F), indicating the impact of NF-κB was trametinib-specific.

In our previous study, we reported that cooperative activation of Wnt and Ras led to trametinib resistance by enhancing glucose uptake in a Pi3k/Akt dependent manner (Cong et al, 2025). Here, we observed that the canonical NF-κB pathway reporter drosomycin was elevated when Ras and Wnt activities were elevated together in the hindgut but not when either gene was activated alone (Fig. 1G). These data indicate that Wnt activity induces trametinib resistance at least in part by elevating canonical NF-κB signalling in Ras^G12V tumours. We also observed that loss of function p53 increased the level of canonical NF-κB signalling (compare Fig. 1B to Fig. 1G), suggesting that p53 also contributes to the regulation of canonical NF-κB signalling in RAP tumours.

To gain a deeper understanding of the drug impact on Ras^G12V vs. RAP tumours, we assessed the effect of trametinib on tumour growth in the Drosophila hindgut proliferative zone (HPZ). Trametinib exhibited a near-complete suppression of HPZ tumour overgrowth in Ras^G12V tumours (Fig. EV1B,E–G). However, trametinib only partially suppressed HPZ tumour overgrowth of RAP tumours (compare Fig. EV1K to Fig. 1H, quantified in 1K). Inhibition of canonical NF-κB activity by knockdown of dl or dif did not significantly suppress RAP tumour overgrowth in the presence of trametinib (compare Fig. 1H to 1I,J, quantified in 1K). Moreover, elevating Wnt signalling did not alter tumour overgrowth in Ras^G12V tumours or control animals (compare Fig. EV1H,I to EV1F,E, quantified in EV1C). However, in byn-Ras^G12V tumours, HPZ overgrowth was significantly enhanced by both elevated Wnt activity and, more weakly, by loss of p53 (compare Fig. EV1J to EV1F,H, quantified in EV1D). Together, these data indicate that (i) Wnt activation and p53 loss cooperate in promoting overgrowth in Ras^G12V tumours and (ii) canonical NF-κB signalling reduces trametinib efficacy by primarily modulating host survival rather than tumour growth in Drosophila.

Of note, we observed that elevated Wnt signalling led to melanisation of Ras^G12V tumours (compare Fig. EV1I to EV1F or EV1H); melanisation also was observed in RAP tumours (Fig. EV1K). Melanisation is an immune response triggered locally by injury, suggesting that co-activation of Wnt and Ras contributes to a ‘wound-like’ response in the hindgut. Previous work reported that local wounding led to increased canonical NF-κB activity in Drosophila (Capilla et al, 2017), consistent with a model in which elevated Wnt signalling leads to a wound response, upregulation of canonical NF-κB activity, and emergent drug resistance in Ras^G12V tumours.

Toll-1 and Toll-9 are required for upregulation of NF-κB activity in RAP tumours

Toll-like receptors activate canonical NF-κB signalling by binding their ligands (Imler and Hoffmann, 2001). In RAP flies, knockdown of Toll-1 and Toll-9 robustly enhanced trametinib rescue of tumour-induced lethality 48 and 43%, respectively, in the presence of trametinib; knockdown of Toll-6 more weakly (29%) impacted trametinib response, while Toll proteins 3, 4, 5, 7 and 8 had no significant effect (Figs. 2A and EV2A). Importantly, knockdown of Toll-1 or Toll-9 did not significantly affect tumour-induced lethality in the absence of trametinib or control animals in the presence of trametinib (Fig. 2A,B), mirroring our results with NF-κB.

Figure 2. Toll-1 and Toll-9 are required for upregulation of canonical NF-κB activity in RAP tumours.

(A, B) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM). (A) RAP;Toll1ⁱ (DMSO, n = 18; tram, n = 17) and RAP;Toll9ⁱ (DMSO, n = 16; tram, n = 16); (B) Control (tram, n = 12), Toll1ⁱ (tram, n = 14) and Toll9ⁱ (tram, n = 11). (C–F) Control (C), RAP (D), RAP;Toll1ⁱ (E), RAP;Toll9ⁱ (F) transgenes were induced in hindguts and were stained with anti-dorsal antibody. DNA were visualised with propidium iodide (PI), Scale bar 200 µm. (A–F) The experiment was conducted at 29 °C. The statistical tests used to calculate the P value are as follows: (A, B) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. Source data are available online for this figure.

Figure EV2. Administration of NF-κB inhibitors did not suppress drug resistance in RAP tumours.

(A) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM), RAP;Toll3ⁱ (DMSO, n = 12; tram, n = 12), RAP;Toll4ⁱ (DMSO, n = 6; tram, n = 6), RAP;Toll5ⁱ (DMSO, n = 4; tram, n = 4), RAP;Toll6ⁱ (DMSO, n = 18; tram, n = 18), RAP;Toll7ⁱ (DMSO, n = 5; tram, n = 4), RAP;Toll8ⁱ (DMSO, n = 12; tram, n = 12). (B, C). Images of the digestive tract of third instar larvae in the presence of trametinib (1 µM), Scale bar 200 µm. (D) Average hindgut proliferation zone (HPZ) size was measured by Fiji ImageJ and quantified as relative size to the control hindgut. RAP;GFP (tram, n = 13), RAP;Toll1ⁱ (tram, n = 14), RAP;Toll9ⁱ (tram, n = 10). (A–D) The experiment was conducted at 29 °C. The statistical tests used to calculate the P value are as follows: (A, B) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates.

These data prompted us to test whether Toll-1 and Toll-9 are essential for inducing canonical NF-κB activity in RAP tumours. A recent study has demonstrated that Toll-9 can elevate canonical NF-κB activity to control proliferation and apoptosis in Drosophila imaginal discs in a Toll-1-dependent manner (Shields et al, 2022). We observed that nuclear translocation of Dorsal was strongly increased in RAP tumours compared to control (Fig. 2C,D), an indication of elevated pathway activity. Canonical NF-κB activity was suppressed by knockdown of Toll-1 or Toll-9 in RAP tumours (compare Fig. 2E,F to Fig. 2D), indicating that Toll-1 and Toll-9 is indeed required for enhancing canonical NF-κB activity in the context of RAP hindgut tumours. Also consistent with the results above, knockdown of Toll-1 or Toll-9 did not significantly suppress RAP hindgut proliferation in the presence of trametinib (compare Figs. 1H to EV2B,C, quantified in EV2D).

Canonical NF-κB activity impacted trametinib response by altering its glucuronidation

Our data suggest a model in which combined alterations in Ras, APC, and P53 lead to Toll-mediated activation of canonical NF-κB signalling, which in turn promotes resistance to trametinib. To better understand the link between NF-κB activity and drug resistance, we used RNA sequencing (RNA-Seq) analysis to compare RAP;dlⁱ and RAP tumours treated with trametinib. We identified 1032 significantly altered genes: 446 downregulated and 566 upregulated in RAP;dlⁱ tumours when compared to RAP tumours (Fig. 3A and Dataset EV1; padj.<0.05, |log2(fold change)| >0.3).

Figure 3. Canonical NF-κB activity affects trametinib response by regulating its glucuronidation.

(A) Volcano plot showing log2 fold change (x-axis) and −log10 adjusted p values (y-axis) of genes differentially expressed between RAP;dlⁱ and RAP in the presence of trametinib (1 µM). (B) Examples of significantly altered pathways. (C, H) The levels of trametinib glucuronidation measured by released UDP level in RAP;GFP (tram, n = 14), RAP;dlⁱ (tram, n = 14) (C) or RAP;GFP (tram, n = 16), RAP;cht4ⁱ (tram, n = 13) (H). (D–F) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM). (D) RAP;blanksⁱ (DMSO, n = 16; tram, n = 14), RAP;cht4ⁱ (DMSO, n = 18; tram, n = 16), RAP;mfs14ⁱ (DMSO, n = 17; tram, n = 17); (E) RAP;cht5ⁱ (DMSO, n = 16; tram, n = 13); (F) Control (tram, n = 15), blanksⁱ (tram, n = 12), cht4ⁱ (tram, n = 12), mfs14ⁱ (tram, n = 12). (G) Gene expression (blanks, cht4, cht5 and mfs14) was detected by RNA sequencing in RAP;GFP (n = 3) and RAP;dlⁱ (n = 3). (A, C, G, H) The experiments were conducted at 27 °C. (D–F) The experiment was conducted at 29 °C. The statistical tests used to calculate the P value are as follows: (A) Wald test; (C, F, H) Mann–Whitney test; (D, E) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. Source data are available online for this figure.

Knockdown of dl led to (i) a reduction in Toll signalling targets, including drs, atta and wntd as expected. Reduced dl also led to reduced Imd-associated immune signalling (dptb, pirk), as well as signalling pathways Jak/Stat (upd1, upd2, upd3, dome socs36e) and Jnk (kay, puc, mmp1; Fig. 3B and Dataset EV1). However, reducing activity of Imd (relishⁱ), Jak/Stat (domelessⁱ) or Jnk (basketⁱ) did not significantly reduce trametinib resistance in RAP tumours (Fig. EV3A). We therefore focused on other factors highlighted by our RNA-Seq analysis.

Figure EV3. Screening for factors that influence drug response.

(A–E) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM). All flies driven by byn-GAL4: (A) RAP;relⁱ (DMSO, n = 11; tram, n = 10), RAP;domeⁱ (HMJ21208) (DMSO, n = 11; tram, n = 7), RAP;domeⁱ (HMS01293) (DMSO, n = 10; tram, n = 10), RAP;bskⁱ (DMSO, n = 6; tram, n = 6); (B) RAP;GFP (tram, n = 14), RAP;CG9360ⁱ (tram, n = 5), RAP;ref(2)pⁱ (tram, n = 10), RAP;CG32302ⁱ (tram, n = 6), RAP;CG17104ⁱ (tram, n = 8), RAP;mec2ⁱ (tram, n = 10), RAP;CG1698ⁱ (tram, n = 5), RAP;CG15739ⁱ (tram, n = 9), RAP;ag5rⁱ (tram, n = 11), RAP;arcⁱ (tram, n = 6), RAP;CG2065ⁱ (tram, n = 6), RAP;CG10182ⁱ (tram, n = 8), RAP;18473ⁱ (tram, n = 12); (C) RAP;GFP (tram, n = 14), RAP;cdc23ⁱ (tram, n = 3), RAP;rpt3rⁱ (tram, n = 8), RAP;ter94ⁱ (tram, n = 6), RAP;prosalpha4ⁱ (tram, n = 4), RAP;alhⁱ (tram, n = 5), RAP;punchⁱ (tram, n = 6), RAP;CG4502ⁱ (tram, n = 5), RAP;CG12493ⁱ (tram, n = 5), RAP;visⁱ (tram, n = 12), RAP;rpt4ⁱ (tram, n = 5), RAP;rpn3ⁱ (tram, n = 2); (D) RAP;GFP (tram, n = 14), RAP;cyp4p1ⁱ (tram, n = 6), RAP;hsp23ⁱ (tram, n = 6), RAP;mal-a6ⁱ (tram, n = 4), RAP;fngⁱ (tram, n = 6), RAP;CG30427ⁱ (tram, n = 3), RAP;prosbeta3ⁱ (tram, n = 4), RAP;ldhⁱ (tram, n = 4), RAP;CG8036ⁱ (tram, n = 5), RAP;CG32365ⁱ (tram, n = 4), RAP;CG4733ⁱ (tram, n = 4), RAP;CG14395ⁱ (tram, n = 4); (E) Ras^G12V;GFP (tram, n = 13), Ras^G12V;nd-pdswⁱ (tram, n = 6), Ras^G12V;cda4ⁱ (tram, n = 6), Ras^G12V;mucⁱ (tram, n = 2), Ras^G12V;CG4459ⁱ (tram, n = 5), Ras^G12V;cox5aⁱ (tram, n = 6), Ras^G12V;CG32564ⁱ (tram, n = 6), Ras^G12V;nd-b22ⁱ (tram, n = 6), Ras^G12V;cox7aⁱ (tram, n = 6). (F–I) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM), QNZ (EVP4593) or JSH-23. (F) RAP;dl[1] (DMSO, n = 16; tram, n = 18) and RAP (DMSO, n = 14; tram, n = 14); (G–I) RAP (DMSO, n = 12; tram, n = 12; 1 µM QNZ, n = 12; 5 µM QNZ, n = 12; tram + 1 µM QNZ, n = 12; tram + 5 µM QNZ, n = 12) (G); control (tram, n = 20; tram + 1 µM QNZ, n = 12; tram + 5 µM QNZ, n = 12) (H); RAP (tram, n = 12; tram + 1 µM JSH-23, n = 12; tram + 5 µM JSH-23, n = 12; tram + 10 µM JSH-23, n = 12) (I). (J) Expression levels of drosomycin were quantified for each condition by quantitative RT-PCR, QNZ (5 µM), JSH-23 (5 µM). (A–D, F) The experiments were conducted at 29 °C. (E, G–I) The experiments were conducted at 29 °C. The statistical tests used to calculate the P value are as follows: (A, F–I) one-way ANOVA; NS P(>0.12), *P(0.033) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates.

Recently, we reported that RAP tumours displayed emergent upregulation of the glucuronidation pathway, a detoxification pathway known to directly inactivate many cancer drugs, including trametinib (Cong et al, 2025). We further demonstrated that the glucuronidation pathway was enhanced by the pentose phosphate pathway, linking detoxification to circulating glucose (Cong et al, 2025). Here, we observed that reduced dl (RAP;dlⁱ) led to reduced expression of key glucuronidation pathway enzymes, including sgl (human ortholog: UGDH) as well as pentose phosphate pathway enzymes, including pgd (PGD), rpi (RPIA), and zw (G6PD; Fig. 3B; Dataset EV1). Further, knockdown of dl significantly suppressed levels of glucuronidated trametinib in RAP;dlⁱ tumours as assessed by UDP release (Fig. 3C). This data indicates that canonical NF-κB signalling promotes resistance to trametinib at least in part by regulating genes that control glucuronidation, in turn helping tumours directly inactivate and expel drugs.

To identify novel canonical NF-κB downstream factors that influence trametinib response, we built on results from our RNA-Seq screen by performing a focused RNAi-based screen of 45 genes, comparing RAP vs. Ras^G12V tumours. RAP response to trametinib was strongly impacted by knockdown of blanks (human orthologue ADARB1), chitinase 4 (cht4, human orthologue AMCase or CHIA), major facilitator superfamily transporter 14 (mfs14, human orthologues SLC17A2, A3, A5, A7 and A8); chitinase 5 (cht5, human orthologue CHIA) weakly impacted trametinib response in RAP tumours (Figs. 3D,E; EV3B–E). Of note, knockdown of blanks, cht4 or mfs14 did not affect control flies in the presence of trametinib (Fig. 3F). Knockdown of cht4 did not affect tumour development in the absence of trametinib, while blanks and mfs14 weakly affected tumour development (compare Fig. 3D to EV3F): RAP alone displayed 7% survival compared to RAP;cht4ⁱ (1%), RAP;blanksⁱ (17%) and RAP;mfs14ⁱ (23%). Also, in the context of RAP, knockdown of dl strongly reduced blanks, cht4 expression and, more weakly, mfs14 and cht5 (Fig. 3G).

Previous work has shown that overexpression of the human cht4 orthologue CHIA upregulated PI3K/AKT activation in human lung epithelial cells in vitro (Hartl et al, 2009). Our previous study in Drosophila showed that elevated Pi3k/Akt signalling enhanced glucuronidation by increasing glucose uptake in RAP tumours, and the level of trametinib glucuronidation can be measured by the released UDP levels (Cong et al, 2025). Here, we found that knockdown of cht4 significantly suppressed the level of released UDP in the presence of trametinib in RAP tumours (Fig. 3H), suggesting that canonical NF-κB activity also influences trametinib response by regulating Pi3k-mediated glucose uptake.

To assess canonical NF-κB pathway signalling as a therapeutic target, we removed one functional copy of dl in the context of the whole animal using a dl[1] loss of function allele (RAP;dl^1/+). Subtle reduction of dl was sufficient to strongly enhance the ability of trametinib to reduce RAP lethality (Fig. EV3F), supporting canonical NF-κB pathway components as adjunct therapies. Administering NF-κB inhibitors QNZ (EVP4593) or JSH-23 reduced NF-κB pathway activity as assessed with drosomycin expression (Fig. EV3G–J). JSH-23 trend towards improved survival did not rise to significance, while QNZ reduced survival: in addition to reducing NF-κB pathway activity, QNZ also inhibits production of TNF-α (Tobe et al, 2003), suggesting that off-targets may contribute to reduced survival.

Wnt inhibitors increased trametinib efficacy on RAP tumours

Our results indicate that Wnt activation plays a role in regulating host lethality by increasing canonical NF-κB activity and, in turn, reducing trametinib’s ability to suppress tumour progression in Ras^G12V animals. We therefore next assessed whether inhibiting WNT activity would reverse drug resistance in RAP tumours by feeding RAP flies one of several (Figs. 4A and EV4A–D) WNT inhibitors plus trametinib. The result was emergent rescue: in particular, Wnt pathway inhibitors PNU-74654 (Trosset et al, 2006) and LF3 (Fang et al, 2016)—which act by suppressing the interaction between β-Catenin and TCF—demonstrated strong efficacy in reducing RAP tumours when paired with trametinib (Fig. 4B,C). This suppression led to increased animal survival in the presence of trametinib, without impacting survival in the absence of trametinib or in control animals (Fig. 4D). Consistent with our results, combining trametinib plus PNU-74654 strongly suppressed canonical NF-κB activity in RAP tumours (Fig. EV4E).

Figure 4. PNU-74654 and LF3 suppressed trametinib resistance in RAP tumours.

(A) Summary of the rescue rate of trametinib and WNT inhibitor drug combinations in RAP hindgut tumours. (B–D) Percent survival of RAP flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM), PNU-74654 (1 µM) or LF3 (10 µM). RAP (DMSO, n = 20; tram, n = 19; PNU, n = 18; tram+PNU, n = 24) (B); RAP (DMSO, n = 12; tram, n = 34; LF3, n = 12; tram + LF3 n = 33) (C); Control (DMSO, n = 17; tram+PNU, n = 11; tram + LF3, n = 14) (D). (E) The average hindgut proliferation zone (HPZ) size was measured by Fiji ImageJ and quantified as relative size to wild-type (WT) hindguts, RAP (DMSO, n = 13; tram, n = 8; PNU, n = 7; tram + PNU, n = 7). (F, G) Images of the digestive tract of third instar larvae in the presence or absence of trametinib (1 µM) or PNU-74654 (1 µM), Scale bar 100 µm. Scale bar 200 µm. (A–G) The experiment was conducted at 27 °C. The statistical tests used to calculate the P value are as follows: (B–E) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. Source data are available online for this figure.

Figure EV4. A screening for trametinib and WNT inhibitors drug combination in RAP hindgut tumours.

(A–D) Percent survival of transgenic RAP flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM), iCRT3, IWP-01, XAV-939 or Capmatinib. RAP (tram, n = 11; tram + 1 µM iCRT3, n = 11; tram + 5 µM iCRT3, n = 10; tram + 10 µM iCRT3, n = 11) (A); RAP (tram, n = 28; tram + 5 µM IWP-01, n = 18; tram + 10 µM IWP-01, n = 22; tram + 15 µM IWP-01, n = 21) (B); RAP (tram, n = 12; tram + 1 µM XAV-939, n = 12; tram + 5 µM XAV-939, n = 12; tram + 10 µM XAV-939, n = 12) (C); RAP (tram, n = 12; tram + 1 µM Capmatinib, n = 12; tram + 5 µM Capmatinib, n = 12; tram + 10 µM Capmatinib, n = 12) (D). (E) The expression levels of drosomycin among each genotype in the presence or absence of trametinib (1 µM) or PNU-74654 (1 µM) were detected by quantitative RT-PCR, n = 3. (E) control and RAP. (A–E) The experiment was conducted at 27 °C. The statistical tests used to calculate the P value are as follows: (A–D) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. (F) A summary of mutations in patient-specific CRC models. These patient-specific fly avatars come from a fly-to-bedside study (CPCTs). and TCGA (RAPp1, RAPp2).

Regarding tumour progression, PNU-74654 plus trametinib suppressed tumour overgrowth in the HPZ, while PNU-74654 alone did not affect tumour overgrowth (compare Figs. 1H to 4F,G, quantified in 4E). These data indicate that targeting Wnt activity pharmacologically is effective at reducing trametinib resistance in RAP tumours.

Combining trametinib and PNU-74654 suppressed tumour progression in genetically complex ‘avatar’ tumours

In previous work (Bangi et al, 2016), we found that fly CRC models targeting three to four genes responded poorly to trametinib as a single agent, requiring drug combinations for efficacy. We therefore assessed whether combining trametinib plus PNU-74654 could effectively suppress tumour progression in still more genetically complex CRC lines. We tested seven ‘patient-specific fly avatar’ lines, each targeting 6-10 genes to more fully model the mutation profile of individual CRC patients (Fig. EV4F). Each exhibited minimal response to trametinib alone (Fig. 5). In contrast, five CRC avatar lines designed as part of a clinical trial (Bangi et al, 2019) responded significantly to oral trametinib plus PNU-74654 (Fig. 5A–E), while two additional lines developed from the TCGA cancer database (Network, 2013) did not significantly respond to the cocktail (Fig. 5F).

Figure 5. Combination of trametinib and PNU-74654 suppressed tumour progression in various genetically complex tumours.

(A–F) Percent survival of transgenic patient-specific avatar fly lines to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM) or PNU-74654. (A) CPCT006 (DMSO, n = 33; tram, n = 26; tram + 1 µM PNU, n = 13; tram + 5 µM PNU, n = 15; tram + 10 µM PNU, n = 15); (B) CPCT018 (DMSO, n = 24; tram, n = 21; tram + 0.5 µM PNU, n = 18; tram + 1 µM PNU, n = 33; tram + 5 µM PNU, n = 26); (C) CPCT045 (DMSO, n = 33; tram, n = 24; tram + 5 µM PNU, n = 22; tram + 10 µM PNU, n = 17; tram + 15 µM PNU, n = 12); (D) CPCT050 (DMSO, n = 21; tram, n = 21; tram + 5 µM PNU, n = 12; tram + 10 µM PNU, n = 20; tram + 15 µM PNU, n = 12); (E) CPCT029 (DMSO, n = 28; tram, n = 28; tram + 0.5 µM PNU, n = 12; tram + 1 µM PNU, n = 20; tram + 5 µM PNU, n = 23). (F) trametinib and/or PNU-74654 (5 µM) exhibited poor rescue of RAPp1 and RAPp2. RAPp1 (DMSO, n = 7; tram, n = 8; PNU, n = 9; tram + PNU, n = 9); RAPp2 (DMSO, n = 12; tram, n = 11; PNU, n = 7; tram + PNU, n = 8). (A) The experiment was conducted at 25 °C. (B) and (D–F) The experiments were conducted at 28 °C. (C) The experiment was conducted at 27 °C. The statistical tests used to calculate the P value are as follows: (A–F) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. Source data are available online for this figure.

Mediators of trametinib/PNU-74654 efficacy in CRC tumours

To identify factors that mediate the efficacy of trametinib plus PNU-74654 on RAP flies, we performed a genetic screen (Figs. 6A and EV5A). We identified five genes that, when targeted for knockdown by RNA-interference (RNAi), enhanced adult eclosion of RAP flies in the presence of trametinib plus PNU-74654 (or LF3): brahma (brm, orthologue of human SMARCA2 and SMARCA4), shotgun (shg, CDH1), archipelago (ago, FBXW7), rhoGAPp190 (rhoGAPp190, ARHGAP5, ARHGAP35) and upf1 RNA helicase (upf1, UPF1) (Figs. 6A and EV5B). Importantly, none of these five gene knockdowns impacted survival of control animals or untreated RAP flies (Figs. 6B,C and EV5C). These data suggest that brm, shg, ago, rhoGAPp190 and upf1 help mediate the efficacy of trametinib plus a WNT pathway inhibitor in CRC tumours. Testing single drugs with each of the five loci, we found that brmⁱ and agoⁱ significantly increased the sensitivity of RAP tumours to trametinib (Fig. 6E, schematised in 6D); rhoGAPp190ⁱ sensitised RAP tumours to PNU-74654 (Fig. 6F, schematised in 6D).

Figure 6. Regulators of the combination of trametinib and PNU-74654 in CRC tumours.

(A–C) Survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM) or PNU-74654 (1 µM). (A) Knockdown of brm, shg, ago, rhoGAPp190, or upf1 improved the survival to adulthood of RAP flies in the presence of trametinib plus PNU-74654. RAP;GFP (tram + PNU, n = 11), RAP;brmⁱ (tram+PNU, n = 14), RAP;agoⁱ (tram+PNU, n = 11), RAP;shgⁱ (tram + PNU, n = 13), RAP;upf1ⁱ (tram + PNU, n = 8), RAP;rhoGAPp190ⁱ (tram + PNU, n = 10). (B) GFP (tram + PNU, n = 11), brmⁱ (tram + PNU, n = 12), agoⁱ (tram + PNU, n = 14), shgⁱ (tram + PNU, n = 11), upf1ⁱ (tram + PNU, n = 12), rhoGAPp190ⁱ (tram + PNU, n = 13). (C) RAP;GFP (DMSO, n = 12), RAP;brmⁱ (DMSO, n = 6), RAP;agoⁱ (DMSO n = 10), RAP;shgⁱ (DMSO, n = 5), RAP;upf1ⁱ (DMSO, n = 6), RAP;rhoGAPp190ⁱ (DMSO n = 10). (D) Summary of five loci found to impact RAP response to trametinib and/or PNU-74654. (E–H) Survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM) or PNU-74654 (1 µM except where noted). (E) RNAi-mediated knockdown of brm or ago improved the survival of RAP flies treated with trametinib. RAP;GFP (tram, n = 10), RAP;brmⁱ (tram, n = 11), RAP;agoⁱ (tram, n = 11), RAP;shgⁱ (tram, n = 11), RAP;upf1ⁱ (tram, n = 11), RAP;rhoGAPp190ⁱ (tram, n = 11). (F) rhoGAPp190 knockdown improved the survival of RAP flies treated with PNU-74654. RAP;GFP (PNU, n = 15), RAP;brmⁱ (PNU, n = 12), RAP;agoⁱ (PNU, n = 13), RAP;shgⁱ (PNU, n = 13), RAP;upf1ⁱ (PNU, n = 20), RAP;rhoGAPp190ⁱ (PNU, n = 20). (G, H) Knockdown of brm, shg, ago, rhoGAPp190, or upf1 rescued more genetically complex avatar lines RAPp1 and RAPp2 when treated with trametinib plus PNU-74654 (5 µM); note shg rescue of RAPp1 did not rise to the level of statistical significance. RAPp1;GFP (tram+PNU, n = 20), RAPp1;brmⁱ (tram+PNU, n = 11), RAPp1;agoⁱ (tram + PNU, n = 7), RAPp1;shgⁱ (tram + PNU, n = 12), RAPp1;upf1ⁱ (tram + PNU, n = 12), RAPp1;rhoGAPp190ⁱ (tram + PNU, n = 15) (G); RAPp2;GFP (tram + PNU, n = 22), RAPp2;brmⁱ (tram + PNU, n = 12), RAPp2;agoⁱ (tram + PNU, n = 12), RAPp2;shgⁱ (tram + PNU, n = 11), RAPp2;upf1ⁱ (tram + PNU, n = 16), RAPp2;rhoGAPp190ⁱ (tram + PNU, n = 14) (H). The statistical tests used to calculate the P value are as follows: (A–C and E–H) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. Source data are available online for this figure.

Figure EV5. Regulators for the combination of trametinib and LF3 in CRC tumours.

(A–C) Percent survival of transgenic flies to adulthood relative to control flies was quantified in the presence or absence of trametinib (1 µM), PNU-74654 (1 µM) or LF3 (10 µM). RAP;GFP (tram + PNU, n = 7), RAP;DNApol-etaⁱ (tram+PNU, n = 6); RAP;lrp1ⁱ (tram + PNU, n = 4) (A). RAP;GFP (tram + LF3, n = 19), RAP;brmⁱ (tram + LF3, n = 12), RAP;agoⁱ (tram + LF3, n = 11), RAP;shgⁱ (tram + LF3, n = 17), RAP;upf1ⁱ (tram + LF3, n = 12), RAP;rhoGAPp190ⁱ (tram + LF3, n = 12) (B); control (tram + LF3, n = 12), brmⁱ (tram + LF3, n = 12), agoⁱ (tram + LF3, n = 12), shgⁱ (tram + LF3, n = 12), upf1ⁱ (tram + LF3, n = 12), rhoGAPp190ⁱ (tram + LF3, n = 12) (C). (D) The graph showed the frequency of mutated genes in human CRC. These data from cBioPortal include 594 patients. (A–C) The experiment was conducted at 29 °C. The statistical tests used to calculate the P value are as follows: (A–C) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates.

As noted above, two of seven tested ‘patient-specific fly avatar’ lines, RAPp1 and RAPp2, were resistant to combined trametinib plus PNU-74654 (Fig. 5F). Knockdown of brm, shg, ago, rhoGAPp190 and upf1 each significantly enhanced tumour sensitivity to trametinib plus PNU-74654 in both multigenic tumours, with the exception that shg had only weakly improved RAPp1 drug response (Fig. 6G,H). Finally, we note human orthologues of these five loci are altered in a subset of human CRC patients (Fig. EV5D), suggesting they could serve as biomarkers for patients that would be especially sensitive to trametinib plus PNU-74654.

Elevated WNT plus KRAS is associated with increased canonical NF-κB signalling in human CRC tumour samples

To help assess whether our Drosophila data were relevant to human CRC, we examined human KRAS-mutant CRC tissue sections to determine whether high WNT activity is associated with high NF-κB activity. We used IKK isoforms to identify canonical vs. non-canonical NF-κB signalling. IKKβ serves as the primary catalytic subunit of IKK, activating canonical NF-κB signalling by proinflammatory cytokines such as TNFα, IL-1 and LPS. IKKα activates non-canonical NF-κB signalling activated by other members of the TNFR superfamily (Häcker and Karin, 2006; Scheidereit, 2006): for example, IKKα is phosphorylated by NIK at Ser-176 to promote release of the non-canonical NF-κB factor RelB-p52 into the nucleus to activate target genes (Ling et al, 1998).

WNT activity was assessed in human CRC patient samples by anti-β-catenin antibody (Fig. EV6A,B). We observed that expression of the IKKβ protein was significantly higher in CRC patient samples with both elevated WNT activity plus oncogenic KRAS when compared to those with only elevated WNT activity or KRAS mutations (Fig. 7A–D, quantified in 7I). In contrast, we found no significant differences in the levels of IKKα or Ser-176 phosphorylated IKKα in samples with (i) high WNT/oncogenic KRAS samples vs. (ii) high WNT activity or oncogenic KRAS (Fig. 7E–H, quantified in 7J; Fig. EV6C–F, quantified in EV6G). That is, CRC tumours with oncogenic KRAS plus high WNT activity are associated with significantly elevated canonical NF-κB signalling, consistent with our Drosophila data. Of note, we observed upregulation of glucuronidation in the presence of trametinib in human T84 colon cancer cells; JSH-23 administration led to a trend of enhancing trametinib efficacy that only rose to the level of significance in 1/3 experiments (Figs. 8A,B and EV6H). In contrast, SW620 colon cancer cells were significantly more sensitive to treatment with JSH-23-plus-trametinib (Figs. 8C and EV6I). Both cell lines contain KRAS and APC mutations, suggesting that additional genetic mutations can influence drug response.

Figure EV6. IHC staining of phospho-IKKα^s176 protein in human CRC.

(A–F) IHC staining of β-catenin or phospho-IKKα^s176 in each genotype CRC sample. (A) KRAS^WT (wild-type); (B) KRAS^MT (mutation in position G12/G13); (C) KRAS^WT plus low β-catenin; (D) KRAS^MT plus low β-catenin; (E) KRAS^WT plus high β-catenin; (F) KRAS^MT plus high β-catenin expression in CRC samples. Scale bar 50 µm. (G) The graph shows the mean of expression of phospho-IKKα^s176 in each different mutated human CRC, determined by IHC intensity values. Patients were grouped into four categories based on KRAS status and β-catenin expression. Both KRAS^WT and β-catenin low (n = 316); KRAS^M and β-catenin low (n = 164); KRAS^WT and β-catenin high (n = 59); Both KRAS^M and β-catenin high (n = 29). Each dot displays an individual sample. (H, I) Cell viability assay for T84 or SW620 colon cancer cells, showing the results of three experimental replicates, treatment with 0.1% DMSO, 50 nM trametinib or 10 µM JSH-23. The statistical tests used to calculate the P value are as follows: (G) one-way ANOVA; (H, I) 2-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates.

Figure 7. Co-activation of WNT and KRAS was associated with an upregulation of canonical NF-κB activity in human CRC.

(A–H) Representative immunohistochemical (IHC) staining for IKKβ and IKKα in stage 2–3 colorectal cancer patient samples. (A, E) KRAS^WT (wild-type) plus low β-catenin expression CRC samples; (B, F) KRAS^MT (mutation in position G12/G13) plus low β-catenin expression CRC samples; (C, G) KRAS^WT plus high β-catenin expression CRC samples; (D, H) KRAS^MT plus high β-catenin expression CRC samples. The black arrow highlights the high signal area. Scale bar 50 µm. (I, J) The graph shows the mean of expression of IKKβ (I) or IKKα (J) in each different mutated human CRC, determined by IHC intensity values. Patients were grouped into four categories based on KRAS status and β-catenin expression. (I) Both KRAS^WT and β-catenin low (n = 311); KRAS^M and β-catenin low (n = 164); KRAS^WT and β-catenin high (n = 56); Both KRAS^M and β-catenin high (n = 28). (J) Both KRAS^WT and β-catenin low (n = 291); KRAS^M and β-catenin low (n = 158); KRAS^WT and β-catenin high (n = 57); Both KRAS^M and β-catenin high (n = 31). Each dot displays an individual sample. The statistical tests used to calculate the P value are as follows: (I, J) one-way ANOVA; NS P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All statistical data are summarised in Table EV1. The error bar is a standard deviation (SD), with each point representing biological replicates. Source data are available online for this figure.

Figure 8. The efficacy of trametinib or JSH-23 on human colon cancer cells.

(A) The levels of trametinib glucuronidation measured by released UDP level in T84 colon cancer cells, n = 9. (B, C) Cell viability assay for T84 colon cancer cells (B) or SW620 (C) treatment with 0.1% DMSO, 50 nM trametinib or 10 µM JSH-23, n = 9. The error bar is a standard deviation (SD), with each point representing biological replicates and numbers (n), including three technical replicates. (D) Schematic summary. We previously reported that pairing activated RAS and WNT activities leads to increased glucose flux into cells in a PI3K/AKT pathway-dependent manner, leading to elevated glucuronidation and elimination of the potent MEK inhibitor trametinib in cancer cells (Cong et al, 2025). Here, we provide evidence that elevated WNT signalling leads to a wound-like response in RAS cancer cells that induced upregulation of canonical NF-κB activity by Toll-like receptors (TLRs). This upregulation enhances glucuronidation pathway activity by increasing gene expression of related enzymes including controlling the CHIA/AKT glucose uptake axis. Source data are available online for this figure.

Discussion

Colorectal tumours that contain oncogenic RAS isoforms have proven resistant to most targeted therapies. In this study, we conduct an in-depth analysis of ‘RAP’ tumours that contain the three genes most commonly mutated in human CRC: RAS (typically KRAS), APC, and P53. We found that pairing oncogenic Ras^G12V with elevated Wnt activity led to upregulation of canonical NF-κB signalling, both in Drosophila hindgut tumours and human tumours. This emergent canonical NF-κB activity has consequences: it reduced the efficacy of the MEK inhibitor trametinib in RAP tumours at least in part by increasing glucuronidation of the drug (Fig. 8D). This resistance to trametinib was strongly reversed when WNT inhibitors were included. Oncogenic RAS isoforms are present in approximately half of all CRC tumours, and our work provides new pathways towards benefiting these patients.

Most WNT pathway inhibitors currently undergoing clinical trials act by promoting degradation of β-catenin, including inhibitors of PORCN (e.g. ETC-1922159, WNT974 and XNW7201) and Frizzled receptors (e.g. vantictumab, ipafricept) (Neiheisel et al, 2022). However, our data suggest that the class of WNT inhibitors that disrupt the interaction between β-catenin and TCF—including PNU-74654 and LF3—are particularly effective when used in combination with trametinib for treating CRC. Further, this drug combination proved effective even in more genetically complex CRC avatar lines that were especially resistant to a broad palette of drugs, including trametinib. To extend our drug resistance work, we identified five genes that further enhanced the efficacy of trametinib/PNU-74654: brm, shg, ago, rhoGAPp190 and upf1. These genes are mutated in a subset of patients, identifying a cohort that may prove especially responsive to MEK/WNT inhibitor drug combinations. Two genes—brm and ago—enhanced the efficacy of trametinib alone, identifying a candidate biomarker for trametinib response. That is, our work identifies a path to matching drugs to specific subsets of RAS-mutant CRC patients.

The NF-κB pathway has been widely linked to cancer, including impacting drug resistance by regulating the survival of cancer cells. For example, NF-κB activity is reported to inhibit the response of HTC15 human colon cancer cells to daunomycin by controlling drug uptake (Bentires-Alj et al, 2003). NF-κB activity is also linked to sorafenib resistance in CD13⁺ hepatocellular carcinoma cell lines by controlling genes that regulate cell cycle and apoptosis (Hu et al, 2020). Pharmacologically blocking the NF-κB pathway sensitises tumour cells to doxorubicin in Dll1+ mouse breast cancer cells by promoting cell death (Kumar et al, 2021). In this whole animal study, we demonstrate that canonical NF-κB-mediated drug resistance—through upregulation of glucuronidation pathway activity—is an emergent property of CRC tumours that combine high WNT activity with oncogenic Ras. This may have therapeutic implications, as we demonstrate.

We previously showed a role for TNF signalling in regulating tumour progression in a Ras-dependent Drosophila cancer model (Cordero et al, 2010) and, indeed, removing just one genomic copy of the TNF pathway mediator dl was sufficient to significantly suppress tumour-induced animal lethality. However, pharmacological inhibition of canonical NF-κB pathway activity by JSH-23 did not strongly suppress RAP tumour-induced lethality, while the NF-κB signalling inhibitor QNZ (EVP4593) enhanced trametinib resistance in RAP tumours. QNZ (EVP4593) targets TNFα production, suggesting that systemic TNFα production plays a role in inhibiting tumour progression in the presence of trametinib. Indeed, it has been reported that TNFα renders tumour vessels more permeable, facilitating the delivery of anticancer drug agents to solid tumours (Seynhaeve et al, 2007). Together, pharmacological inhibition—through canonical NF-κB pathway activity inhibitor JSH-23 or signalling inhibitor QNZ (EVP4593)—did not significantly suppress RAP tumour-induced lethality below toxic levels. This suggests that broad targeting of NF-κB may be problematic from a whole-body standpoint, not surprising given the large number of processes controlled by NF-κB activity.

Genetic complexity is a common clinical feature of tumours. Here we demonstrate that a common version of this complexity linked to aggressive CRC disease—combining alterations in RAS, APC, and P53—is sufficient to direct overgrowth of the hindgut proliferative zone (HPZ) and promote emergent drug resistance. Currently, patient tumours with these three altered genes have few second-line therapeutic options, as RAS pathway inhibitors have failed to provide durable responses. Gaining a deeper understanding of how this combination directs drug resistance, through factors such as NF-κB, will provide new candidate avenues towards therapeutics.

Methods

Reagents and tools table

Reagent/resource	Reference or source	Identifier or catalogue number
Experimental models
byn-gal4	V. Hartenstein	N/A
UAS-RasG12V	G. Halder	N/A
tub-gal80TS	BDSC	7017
w1118	BDSC	3605
UAS-mCD8-GFP	BDSC	5137
UAS-dli	BDSC	36650
UAS-difi	BDSC	30513
UAS-cacti	BDSC	37484
UAS-ArmS10	BDSC	4782
UAS-Toll1i	BDSC	35628
UAS-Toll3i	BDSC	28526
UAS-Toll4i	BDSC	28543
UAS-Toll5i	BDSC	29533
UAS-Toll6i	BDSC	56048
UAS-Toll7i	BDSC	30488
UAS-Toll8i	BDSC	28519
UAS-Toll9i	BDSC	34853
UAS-brmi	BDSC	35211
UAS-shgi	BDSC	38207
UAS-agoi	BDSC	34802
UAS-rhoGAPp190i	BDSC	43987
UAS-upf1i	BDSC	64519
UAS-CG13344i	BDSC	41831
UAS-p38ai	BDSC	35244
UAS-fti	BDSC	34970
UAS-puti	BDSC	39025
UAS-dnapol-etai	BDSC	33410
UAS-lrp1i	BDSC	44579
UAS-tefui	BDSC	44073
UAS-neji	BDSC	37489
UAS-nosi	BDSC	33973
UAS-pci	BDSC	36070
UAS-rad51ci	BDSC	67355
dl[1]	BDSC	3236
UAS-blanksi	BDSC	33667
UAS-cht4i	BDSC	65001
UAS-mfs14i	BDSC	33999
UAS-cht5i	BDSC	57512
UAS-ref(2)pi	BDSC	36111
UAS-CG32302i	BDSC	67239
UAS-CG17104i	BDSC	42925
UAS-mec2i	BDSC	61259
UAS-CG15739i	BDSC	57216
UAS-ag5ri	BDSC	67225
UAS-arc1i	BDSC	25954
UAS-CG2065i	BDSC	55283
UAS-CG10182i	BDSC	61954
UAS-CG18473i	BDSC	57524
UAS-cdc23i	BDSC	61982
UAS-rpt3ri	BDSC	58140
UAS-ter94i	BDSC	35608
UAS-prosalpha4i	BDSC	65161
UAS-alhi	BDSC	39057
UAS-punchi	BDSC	41998
UAS-CG4502i	BDSC	35489
UAS-CG12493i	BDSC	42791
UAS-visi	BDSC	35738
UAS-rpt4i	BDSC	32874
UAS-rpn3i	BDSC	34561
UAS-cyp4p1i	BDSC	67349
UAS-hsp23i	BDSC	82961
UAS-mal-a6i	BDSC	60398
UAS-fngi	BDSC	25947
UAS-CG30427i	BDSC	58271
UAS-prosbeta3i	BDSC	34868
UAS-idhi	BDSC	41708
UAS-CG8036i	BDSC	60371
UAS-nd-pdswi	BDSC	29592
UAS-cda4i	BDSC	65909
UAS-muci	BDSC	44439
UAS-CG4459i	BDSC	61228
UAS-cox5ai	BDSC	58282
UAS-CG32564i	BDSC	58342
UAS-nd-b22i	BDSC	65011
UAS-cox7ai	BDSC	57572
UAS-reli	BDSC	33661
UAS-domeHMJ21208i	BDSC	53890
UAS-domeHMS10293i	BDSC	34618
UAS-bski	BDSC	36643
UAS-CG9360i	VDRC	v13189
UAS-CG1698i	VDRC	v101947
UAS-CG32365i	VDRC	v104119
UAS-CG4733i	VDRC	v34894
UAS-CG14395i	VDRC	v17517
Antibodies
Anti-dorsal antibody (Drosophila)	DSHB	7A4
Anti-mouse Alexa Fluor 633	Thermo Fisher Scientific	A-21126
Anti-mouse Alexa Fluor 546	Thermo Fisher Scientific	A-11004
DAPI-containing SlowFade Gold Antifade Reagent	Molecular Probes	S36939
Anti-β-catenin (human)	Dako, CA, USA	M3539
Anti-IKKα (human)	Genway, CA, USA	GWB-662250
Anti-IKKβ (human)	Abcam, Cambridge, UK	ab32135
Anti-IKKαs176 (human)	Abcam, Cambridge, UK	ab138426
Oligonucleotides and other sequence-based reagents
TRIzol®	Invitrogen™, Life Technologies	15596018
iScriptTM gDNA Clear cDNA Synthesis Kit	Bio-Rad Laboratories Ltd	1725035
iTaq™ Universal SYBR® Green Supermix kit	Bio-Rad Laboratories Ltd	1725124
Rp49_forward: CGCTTCAAGGGACAGTATCTG	Suzawa et al, 2019	N/A
Rp49_reverse: AAACGCGGTTCTGCATGA	Suzawa et al, 2019	N/A
Drosomycin_forward: CTCTTCGCTGTCCTGATGCT	Kleino et al, 2017	N/A
Drosomycin_reverse: ACAGGTCTCGTTGTCCCAGA	Kleino et al, 2017	N/A
Chemicals, enzymes and other reagents
Trametinib	Selleckchem or biorbyt	S2673 or ORB546250-BOR
QNZ (EVP4593)	Selleckchem	S4902
iCRT3	Selleckchem	S8647
IWP-01	Selleckchem	S8645
XAV-939	Selleckchem	S1180
Capmatinib	Selleckchem	S2788
JSH-23	Selleckchem	S7351
PNU-74654	Selleckchem	S8429
Propidium iodide	Selleckchem	S6874
CellTiter-Fluor (TM) Cell Viability Assay	Promega	G6082
UDP-Glo™ Glycosyltransferase Assay	Promega	V6961
Software
RStudio	Posit	N/A
GPT-3.5	Open AI	N/A
GraphPad Prism 10	GraphPad Software	N/A
Microsoft Word	Microsoft	N/A
Microsoft PowerPoint	Microsoft	N/A
Other
T84	ATCC	CCL-248™
SW620	ATCC	CCL-227

Drosophila strains and genetics

Fly lines were cultured at room temperature or 25–29 °C on standard fly food or food-plus-compound. Fly food contained agar 10 g, soya flour 5 g, sucrose 15 g, glucose 33 g, maize meal 15 g, wheat germ 10 g, treacle molasses 30 g, yeast 35 g, nipagin 10 ml, propionic acid 5 ml in 1000 ml water. Transgenes used (Bloomington Drosophila Stock Centre number): byn-gal4 (hindgut-specific line, V. Hartenstein), UAS-Ras^G12V (second chromosome, G. Halder), tub-gal80^TS (#7017), w¹¹¹⁸ (#3605), UAS-mCD8-GFP (#5137), UAS-dlⁱ (#36650), UAS-difⁱ (#30513), UAS-cactⁱ (#37484), UAS-Arm^S10 (#4782), UAS-Toll1ⁱ (#35628), UAS-Toll3ⁱ (#28526), UAS-Toll4ⁱ (#28543), UAS-Toll5ⁱ (#29533), UAS-Toll6ⁱ (#56048), UAS-Toll7ⁱ (#30488), UAS-Toll8ⁱ (#28519), UAS-Toll9ⁱ (#34853), UAS-brmⁱ (#35211), UAS-shgⁱ (#38207), UAS-agoⁱ (#34802), UAS-rhoGAPp190ⁱ (#43987), UAS-upf1ⁱ (#64519), UAS-CG13344ⁱ (#41831), UAS-p38aⁱ (#35244), UAS-ftⁱ (#34970), UAS-putⁱ (#39025), UAS-dnapol-etaⁱ (#33410), UAS-lrp1ⁱ (#44579), UAS-tefuⁱ (#44073), UAS-nejⁱ (#37489), UAS-nosⁱ (#33973), UAS-pcⁱ (#36070), UAS-rad51cⁱ (#67355), dl[1] (#3236), UAS-blanksⁱ (#33667), UAS-cht4ⁱ (#65001), UAS-mfs14ⁱ (#33999), UAS-cht5ⁱ (#57512), UAS-ref(2)pⁱ (#36111), UAS-CG32302ⁱ (#67239), UAS-CG17104ⁱ (#42925), UAS-mec2ⁱ (#61259), UAS-CG15739ⁱ (#57216), UAS-ag5rⁱ (#67225), UAS-arc1ⁱ (#25954), UAS-CG2065ⁱ (#55283), UAS-CG10182ⁱ (#61954), UAS-CG18473ⁱ (#57524), UAS-cdc23ⁱ (#61982), UAS-rpt3rⁱ (#58140), UAS-ter94ⁱ (#35608), UAS-prosalpha4ⁱ (#65161), UAS-alhⁱ (#39057), UAS-punchⁱ (#41998), UAS-CG4502ⁱ (#35489), UAS-CG12493ⁱ (#42791), UAS-visⁱ (#35738), UAS-rpt4ⁱ (#32874), UAS-rpn3ⁱ (#34561), UAS-cyp4p1ⁱ (#67349), UAS-hsp23ⁱ (#82961), UAS-mal-a6ⁱ (#60398), UAS-fngⁱ (#25947), UAS-CG30427ⁱ (#58271), UAS-prosbeta3ⁱ (#34868), UAS-idhⁱ (#41708), UAS-CG8036ⁱ (#60371), UAS-nd-pdswⁱ (#29592), UAS-cda4ⁱ (#65909), UAS-mucⁱ (#44439), UAS-CG4459ⁱ (#61228), UAS-cox5aⁱ (#58282), UAS-CG32564ⁱ (#58342), UAS-nd-b22ⁱ (#65011), UAS-cox7aⁱ (#57572), UAS-relⁱ (#33661), UAS-domeHMJ21208ⁱ (#53890), UAS-domeHMS10293ⁱ (#34618), UAS-bskⁱ (#36643), UAS-CG9360ⁱ (Vienna Drosophila Resource Centre, #v13189), UAS-CG1698ⁱ (#v101947), UAS-CG32365ⁱ (#v104119), UAS-CG4733ⁱ (#v34894), UAS-CG14395ⁱ (#v17517).

Chemicals

Drugs and compounds were used as follows: trametinib (Selleckchem or Biorbyt), QNZ (EVP4593), iCRT3, IWP-01, XAV-939, Capmatinib, JSH-23, PNU-74654 and propidium iodide (PI) purchased from Selleckchem. Drug and compound stocks were diluted in DMSO or water; drugs were then mixed into standard fly food with final DMSO concentration 0.1% to prevent toxicity.

Statistical analysis

Eggs were collected for 24 h in drug-containing food at 18 °C to minimise transgene expression during embryogenesis to prevent embryonic effects or lethality. After 3 days, the tubes were transferred to the appropriate temperature (25–29 °C) to induce transgene expression; the number of surviving Drosophila adults was quantified after 2 weeks. w¹¹¹⁸ served as a control in this study. Transgenic flies and non-transgenic flies’ pupae (TP and no-TP (endogenous control)), respectively, were counted for each test tube, and the percentage of adult survival to control was calculated using the formula [(TP/no-TP) × 100]. Each point on the survival graph represents data from a test tube.

Statistical analysis was performed using Prism 10. Based on the examined statistical distribution and variance, we have chosen the appropriate statistical test, e.g. if the data were not normally distributed and multiple comparisons were required, we used a nonparametric statistical test (e.g. “Dunnett’s multiple comparisons test”). N.S P(>0.12), *P(0.033), **P(0.002) and ***P(<0.001). All experiments were reproduced at least three times. All statistical data are summarised in Table EV1. All detailed genotypes are summarised in Table EV2.

Imaging of the digestive tract of third instar larvae

Third instar larvae were dissected in 1x PBS and fixed with 4% paraformaldehyde for 30 min at room temperature, then washed 3 × 15 min in PBT (0.1% Triton X in 1x PBS). Samples were incubated in anti-dorsal primary antibody (#7A4, DSHB, 1:100); the secondary antibody used was anti-mouse Alexa Fluor 546 or 633 (Invitrogen, 1:250). Samples were mounted with DAPI-containing SlowFade Gold Antifade Reagent (#S36939, Molecular Probes). Fluorescence images were visualised on a Leica TSC SPE confocal microscope.

RNA isolation and quantitative real-time PCR (Drosophila)

Total RNA from 30 hindguts was isolated using TRIzol® according to the manufacturer’s protocol (cat.15596018, Invitrogen™, Life Technologies). mRNA was reverse transcribed using iScriptTM gDNA Clear cDNA Synthesis Kit (cat# 1725035, Bio-Rad Laboratories Ltd).

For quantitative Real-Time PCR (qPCR), iTaq™ Universal SYBR® Green Supermix kit (cat. #1725124, Bio-Rad Laboratories Ltd.) was used according to the manufacturer’s recommendation with cDNA (diluted 1:10–20) as a template. RT-qPCRs were performed with three biological replicates. Relative expression values were determined by the 2^−ΔΔCt method using rp49 as an endogenous control. The RT-qPCR primers used are as follows: rp49 (forward: CGCTTCAAGGGACAGTATCTG; reverse: AAACGCGGTTCTGCATGA), drosomycin (forward: CTCTTCGCTGTCCTGATGCT; reverse: ACAGGTCTCGTTGTCCCAGA).

RNA isolation and RNA sequencing (Drosophila)

RNA sequencing was run and analysed by the CRUK Beatson Institute. The reference genome used was Drosophila melanogaster. BDGP6.46.110 (Ensembl genome). Reads were quality checked using FastQC version 0.11.8 and then trimmed with TrimGalore version 0.6.4 to remove adaptors and low-quality reads (Phred score <20). Then, aligned to the reference above using Hisat2 version 2.1.0, and gene-level counts were determined using FeatureCounts version 1.6.4. The differential expression analyses were done in R using DESeq2 version 1.22.2, which uses a Wald test to assess significance between groups. Graphs were drawn by using ggplot2 package of R, downregulated genes marked by blue (adjusted p values <0.05 and log2(fold change <−0.3); upregulated genes marked by red (adjusted p values <0.05 and log2(fold change >0.3).

Endogenously released UDP assay (Drosophila)

Eggs were collected for 24 h in drug-containing food at 18 °C to minimise transgene expression during embryogenesis to prevent embryonic effects or lethality. After 3 days, the tubes were transferred to the appropriate temperature to induce transgene expression. After 4 days, third instar larvae were dissected in 1x PBS, and the hindgut was assayed in 1x PBS. Cell number (CN) was measured by CellTiter-Fluor^TM Cell Viability Assay kit (Promega). Total endogenous UDP release (TEUDP) in the hindgut was measured by UDP-Glo^TM Glycosyltransferase Assay kit (Promega). The released UDP level in each cell was calculated using the formula [TEUDP/CN].

Immunohistochemistry for detection of β-catenin, IKKβ, IKKα and phospho-IKKα^s176

Samples from a retrospective cohort of 787 stage 2–3 colorectal cancer patients were stained via immunohistochemistry (IHC) for β-catenin, IKKβ, IKKα and phospho-IKKα serine 176 (IKKα^s176). Staining was performed on a previously constructed tissue microarray (TMA), which consisted of CRC tissue from patients undergoing surgery with curative intent within Greater Glasgow and Clyde hospitals between 1997 and 2013. Data were stored within the Glasgow Safehaven (GSH21ON009), and ethical approval was in place for the study (MREC/01/0/36).

IHC was performed as previously described (Al‐Badran et al, 2021). Briefly, TMA sections were dewaxed and then rehydrated through a series of alcohols. Antigen retrieval was performed using citrate buffer (pH 6) for β-catenin, IKKβ and IKKα, and Tris EDTA (pH 9) for IKKα^s176. Endogenous peroxidases were blocked in 3% hydrogen peroxide. Tissue was blocked using 10% casein (SP-5020, Vector Laboratories, CA, USA) for β-catenin and IKKα^s176, and 5% horse serum (S-2000, Vector Laboratories, CA, USA) for IKKβ and IKKα, incubating for 1 h at room temperature. Sections were incubated in primary antibody β-catenin ((M3539, Dako, CA, USA, 1:600), IKKα (GWB-662250, Genway, CA, USA, 1:4000), IKKβ (ab32135, Abcam, Cambridge, UK, 1:200) and IKKα^s176(ab138426, Abcam, Cambridge, UK, 1:150) overnight at 4 ^oC. Sections were washed in tris-buffered saline (TBS), incubated in Impress secondary antibody (MP-7500, Vector Laboratories, CA, USA) for 2 h at room temperature. Sections were washed in TBS and incubated for 5 min in 3,3′-diaminobenzidine (DAB) (SK-4105, Vector Laboratories, CA, USA). Slides were rinsed in water, counterstained and dehydrated before mounting with Pertex (00801-EX, Histolab products, Askim, Sweden). Stained sections were imaged using a Hamamatsu NanoZoomer (Hamamatsu Photonics, Shizuoka, Japan) onto an NZ Connect viewing platform (Hamamatsu Photonics, Shizuoka, Japan).

Staining quantification of β-catenin, IKKβ, IKKα and phospho-IKKα^s176 (human CRC)

Staining intensity was assessed semi-quantitatively by weighted histoscore using QuPath® software in the tumour cell cytoplasm for β-catenin, IKKβ, IKKα and IKKα^s176 (Bankhead et al, 2017). Continuous scores ranging from 0 to 300 for β-catenin were dichotomised into high and low expression groups using the Survminer package in RStudio (version 1.4, RStudio, Boston, MA, USA).

Mutational profiling and analysis (human CRC samples)

CRC tissue from the patient cohort was profiled for the presence of KRAS mutation by BioClavis (BioClavis Ltd, Glasgow, UK). Patients were grouped into three categories based on KRAS status and β-catenin expression. Group 1 patients were wild type for KRAS and low for β-catenin, Group 2 patients were either KRAS mutant or high for β-catenin and Group 3 patients were both KRAS mutant and high for β-catenin. These groups were then assessed for association with IKKβ, IKKα and IKKα^s176 expression using T-tests in GraphPad Prism (GraphPad Software, La Jolla, CA, USA).

Endogenously released UDP assay and cell viability assay (colon cancer cells)

Seed 3000 cells per well (T84, ATCC, CCL-248™ or SW620, ATCC, CCL-227) in a 96-well plate for each cell line. After 24 h, treat the cells with the respective drugs or 0.1% DMSO as a control. Incubate for 48, 96 and 144 h under standard conditions. At each time point, assess cell viability (CV) by adding CellTiter-Fluor Reagent according to the manufacturer’s instructions. To evaluate total endogenous UDP release (TEUDP), used the UDP-Glo™ Glycosyltransferase Assay Kit (Promega). The released UDP levels for each condition were calculated using the formula [TEUDP/CV].

Use of a large language model

Some sentences were revised with the aid of GPT-3.5, strictly to improve clarity. The authors take full responsibility for the accuracy of all prose in the manuscript.

Author contributions

Bojie Cong: Conceptualisation; Data curation; Formal analysis; Investigation; Methodology; Writing—original draft; Writing—review and editing. Evangelia Stamou: Investigation. Kathryn Pennel: Data curation; Formal analysis; Investigation. Teena Thakur: Formal analysis; Investigation. Molly McKenzie: Data curation; Formal analysis; Investigation. Amna Matly: Data curation; Formal analysis; Investigation. Kathryn Gilroy: Data curation; Formal analysis; Investigation. Harshit Shah: Data curation; Formal analysis; Investigation. Sindhura Gopinath: Resources. Joanne Edwards: Conceptualisation; Data curation; Supervision; Project administration. Ross Cagan: Conceptualisation; Supervision; Funding acquisition; Writing—original draft; Project administration; Writing—review and editing.

Source data underlying figure panels in this paper may have individual authorship assigned. Where available, figure panel/source data authorship is listed in the following database record: biostudies:S-SCDT-10_1038-S44319-025-00588-1.

Data availability

No primary datasets have been generated or deposited.

The source data of this paper are collected in the following database record: biostudies:S-SCDT-10_1038-S44319-025-00588-1.

Disclosure and competing interests statement

The authors declare no competing interests.

Supplementary information

Expanded view data, supplementary information, appendices are available for this paper at 10.1038/s44319-025-00588-1.