Hedgehog-GLI signalling promotes chemoresistance through the regulation of ABC transporters in colorectal cancer cells

Colorectal cancer (CRC) is a leading cause of cancer death. Chemoresistance is a pivotal feature of cancer cells leading to treatment failure and ATP-binding cassette (ABC) transporters are responsible for the efflux of several molecules, including anticancer drugs. The Hedgehog-GLI (HH-GLI) pathway is a major signalling in CRC, however its role in chemoresistance has not been fully elucidated. Here we show that the HH-GLI pathway favours resistance to 5-fluorouracil and Oxaliplatin in CRC cells. We identified potential GLI1 binding sites in the promoter region of six ABC transporters, namely ABCA2, ABCB1, ABCB4, ABCB7, ABCC2 and ABCG1. Next, we investigated the binding of GLI1 using chromatin immunoprecipitation experiments and we demonstrate that GLI1 transcriptionally regulates the identified ABC transporters. We show that chemoresistant cells express high levels of GLI1 and of the ABC transporters and that GLI1 inhibition disrupts the transporters up-regulation. Moreover, we report that human CRC tumours express high levels of the ABCG1 transporter and that its expression correlates with worse patients’ prognosis. This study identifies a new mechanism where HH-GLI signalling regulates CRC chemoresistance features. Our results indicate that the inhibition of Gli1 regulates the ABC transporters expression and therefore should be considered as a therapeutic option in chemoresistant patients.

www.nature.com/scientificreports/ is composed of secreted ligands (Sonic Hedgehog-SHH, Desert Hedgehog-DHH and Indian Hedgehog-IHH) that bind to and inactivate the transmembrane receptor Patched (PTCH), which in turn relieves its repression on a second transmembrane receptor Smoothened (SMO) 5 . This activation triggers intracellular molecular events that end up with the activation of the transcription factor GLI1. The disruption of regulatory mechanisms in the HH-GLI pathway is linked to tumorigenesis, tumour maintenance and cancer stem cell phenotype 5 . Of note, GLI1 can also be activated by non-canonical intracellular signalling 5 , referred as "oncogenic load", such as Neuropilin2 in non-small cell lung cancer 7 and KRAS in pancreatic ductal adenocarcinoma 8 . HH-GLI signalling has been described as a major signalling in CRC maintenance and it was recently shown that it mediates anticancer drug resistance in patient-derived organoid cultures 9 . According to this, the present study aimed at analysing the role of the HH-GLI signalling in colorectal cancer (CRC) chemoresistance. Indeed, ABCG2, one of the most intensely studied ABC transporters, is a GLI1 transcriptional target and has been linked to HH-GLI1 dependant drug sensitivity 7, 10, 11 .
The above-described findings prompted us to investigate whether HH-GLI1 contributes to chemoresistance by regulating the expression of ABC transporters.

HH-GLI sustains cell growth and GLI inhibition sensitizes CRC cells to chemotherapy.
We aimed at understanding the role of the HH-GLI signalling in colorectal cancer (CRC) chemoresistance. To this end we first investigated the role of the HH-GLI signalling in the apoptosis and cell growth of CRC cells Colo205 treated with 5-fluorouracil (5-FU) and Oxaliplatin, cytotoxic drugs used in the treatment of advanced CRC 12 .
5-FU and Oxaliplatin treatment of Colo205 cells for 48 h resulted in the up-regulation of GLI1 protein expression with a dose of up to 5 μM, while this was not evident at 10 μM treatment (Fig. 1A). 5-FU and Oxaliplatin treatment at 5 and 10 μM was also accompanied by a significant increase in apoptosis, measured by cleaved PARP (c-PARP) (Fig. 1A).
Coherently, Colo205 showed significantly enhanced cell viability when exposed to up to 5 μM of 5-FU and Oxaliplatin compared to untreated cells (Fig. 1B). However, if cells were previously exposed to the GLI inhibitor GANT61 13 , cell viability was significantly impaired, indicating that GLI inhibition can overcome the growthpromoting effects of a low dose of 5-FU and Oxaliplatin (Fig. 1B). Since 5-FU and Oxaliplatin treatment resulted in GLI1 expression induction, we then proceeded to investigate the subcellular localization of the transcription factor GLI1 in Colo205 treated with 5 μM of 5-FU and Oxaliplatin. While in untreated cells GLI1 was localized both in the cytoplasm and in the nucleus (Fig. 1C), in treated cells GLI1 staining was essentially localized in the nucleus with significantly more cells displaying nuclear GLI1 (Fig. 1D), an observation coherent with an enhanced GLI1 transcriptional activity. GANT61 treated cells showed a weaker staining of GLI1 protein with fewer cells showing nuclear GLI1 (Fig. 1C, D). We then investigated whether the effects of 5-FU plus Oxaliplatin and GANT61 on Colo205 growth could be attributable to modulation in cell proliferation or apoptosis. Proliferation, measured through PCNA protein levels, was impaired by GANT61 treatment and enhanced by 5-FU plus Oxaliplatin treatment (Fig. 1E). Pre-treating cells with GANT61 prevented the proliferative effect of 5-FU plus Oxaliplatin on cells (Fig. 1E). Interestingly, we observed that apoptosis was slightly albeit significantly affected by 5-FU plus Oxaliplatin treatment and the combination with GANT61 was able to significantly increase this effect (Fig. 1E).
ABC transporters are transcriptionally regulated by HH-GLI. Since drug resistance could be associated with modified expression of ABC transporters, we wondered whether HH-GLI signalling was involved in their regulation.
To this end, we performed a qPCR-based array of all ABC transporters in Colo205 after short hairpin mediated GLI1 silencing (shGLI1) ( Fig. 2A). ABCG2 was already demonstrated to be a transcriptional target of GLI1 11 , so we used it as a positive control in our experiments.
To this end, we searched for GLI1 putative binding sites on the proximal promoter region of the ABC transporters. We analysed the region encompassing the 1,000 bases upstream the transcription starting sites (TSS) for the presence of canonical (GAC CAC CCA) and non canonical (CGC CTC CAG) GLI consensus sequences 14 Indeed, we found GLI consensus sequences for six transporters (ABCA2, ABCB1, ABCB4, ABCB7, ABCC2 and ABCG1) ( Fig. 3A and supplementary Fig. 1). The rationale behind the selection of the investigated ABC transporters is depicted in Supplementary Fig. 2.
Therefore, we performed chromatin immunoprecipitation (ChIP) experiments to investigate the recruitment of GLI1 protein on the promoter of the six selected ABC transporters.
We found recruitment of GLI1 on the promoter of all six ABC transporter analysed with a significant downmodulation in shGLI1 Colo205 cells (Fig. 3B); moreover, we observed a significant down-modulation of the transcriptional activation marker acetyl histone H3 recruitment on the ABC transporters promoters after silencing of GLI1 (Fig. 3B).
To validate our results, we overexpressed GLI1 in HCT15, a cell line characterized by low levels of GLI1, and we observed the recruitment of GLI1 on the promoters of ABCB1, ABCB7, ABCG1 and ABCG2, along with higher acetyl histone H3 recruitment (Fig. 3C). The following set of experiments was designed to validate www.nature.com/scientificreports/ our results. First of all, we analysed mRNA expression of the ABC transporters in Colo205 after drug-mediated GLI inhibition. Indeed, GANT-treated Colo205 cells showed decreased mRNA levels of all six ABC transporters analysed along with ABCG2 ( Fig. 4A) and decreased protein levels of ABCA2, ABCB1, ABCC2, ABCG1 and ABCG2 (Fig. 4B). 5-FU and Oxaliplatin treatment caused an increase of mRNA levels of ABCB1, ABCB4, ABCG1 and ABCG2 (Fig. 4A) and of protein levels of ABCB4, ABCB7 and ABCC2 (Fig. 4B). Interestingly, when we combined short-term treatment of GANT61 and 5-FU plus Oxaliplatin, GANT61 treatment was able to overcome the inducing effects Moreover, in HCT15 cells over-expressing GLI1 we observed increased mRNA levels of ABCA2, ABCB1, ABCB7, ABCC2, ABCG1 and ABCG2 (Fig. 4D) and an increase of protein levels of ABCA2, ABCB1, ABCB7, ABCG1 and ABCG2 (Fig. 4E).
Finally, GLI1 over-expression was able to increase cell growth in HCT15 cells, while 5-FU and Oxaliplatin treatment impaired cell growth (Fig. 4F); the combination of GLI1 over-expression and 5-FU and Oxaliplatin www.nature.com/scientificreports/ www.nature.com/scientificreports/ treatment resulted in an increase in cell proliferation ( Fig. 4F) with respect to 5-FU and Oxaliplatin, therefore ectopic GLI1 was able to overcome 5-FU and Oxaliplatin effects on cell growth.

Role of HH-GLi signalling in in-vitro induced chemoresistance.
To further investigate the role of HH-GLI signalling in CRC chemoresistance, we induced chemoresistance in CRC cells by treating cells with 5-FU and Oxaliplatin, as described previously 15 , carrying out the treatment for 5 weeks (supplementary Fig. 4A, B). To verify the resistance, we compared the number of cells after 48 h of 5-FU and Oxaliplatin treatment in parental and resistant cells and we found that resistant cells were significantly more than parental cells, both in Colo205 and in HCT15 (supplementary Figs. 4C, 5D). We then analysed the expression of GLI1 and of the six GLI1-regulated ABC transporters in Colo205 and HCT15 cells during the induction of drug resistance (Fig. 5). Expression of both GLI1 and of the ABC transporters ABCA2, ABCB1, ABCB4, ABCG1 and ABCG2 increased with drug treatment in Colo205 (Fig. 5A). In HCT15, we observed that the treatment up-regulated the expression of GLI1 and of the ABC transporters ABCB1, ABCB4, ABCC2, ABCG1 and ABCG2 (Fig. 5B). Interestingly, if we added GANT61 to the chronic 5-FU and Oxaliplatin treatment in order to prevent GLI1 up-regulation, ABC transporters failed to increase; ABCC2 in HCT15 and ABCG2 in both cell lines were still significantly up-regulated with respect to the starting population, however expression levels were significantly lower with respect to the relative 5-FU plus Oxaliplatin treated cells (Fig. 5A, B).
Taken together, our experiments demonstrated that six ABC transporters were regulated by the HH-GLI signalling, their expression levels were increased in a chemoresistance promoting condition, and this modulation was HH-GLI dependent. We observed that, among the 14 ABC transporters that were modulated by GLI1 silencing, not all resulted to be directly targeted by GLI1 in ChIP experiments, a plausible observation since HH-GLI can regulate their expression levels in an indirect fashion.
In silico analyses of ABC transporters in CRC patients. To validate our results from CRC cellular models, we analysed GLI1 regulated ABC transporters in samples from CRC patients by re-interrogating publicly available CRC datasets using the R2 platform 16 . The following datasets were selected for our analyses: Watanabe 17 , that investigated gene expression in 53 samples of non-neoplastic rectal mucosa and 67 samples of CRC; Galamb 18 , that investigated gene expression in 6 microdissected samples from CRC and 6 microdissected samples from non-neoplastic tissues; Carmical 19 , that investigated gene expression in colorectal CD133+ cancer cells and in cancer-associated fibroblasts (CAFs).
Differential expression and p values of the analysed transporters are reported in Table 1. Among Gli1 regulated ABC transporters, ABCG1 was expressed at significantly higher levels in cancer tissues respect to non-tumour tissues (Fig. 6A, B) and in CD133+ cancer cells respect to CAFs (Fig. 6C).
Moreover, we investigated if there was any association between the expression levels of the ABC transporters of interest and the overall survival of patients. The following datasets were queried: TCGA mixed colon adenocarcinoma, that investigated overall survival probability in 174 samples; Tumor Colon (Core-Transcript) Sveen, that investigated event-free survival probability in 333 samples, Tumor Colon CIT (Combat) Marisa that investigated relapse-free survival probability in 566 samples and Tumor Colon MVRM-SieberSmith, that investigated relapse free survival probability in 345 samples. The results and statistical significance associated with expression levels of ABC transporters of interest are reported in Table 2.
High expression levels of ABCG1 transporter were significantly associated with worse prognosis in three out of four datasets that we investigated, namely TCGA, Sveen and Marisa, as shown in the Kaplan Meier curves (Fig. 6D-F).
We observed that the dataset reporting discordant results (Sieber Smith) combines data from 2 smaller datasets, Sieber (290 samples) and Smith (55 samples), and the Kaplan Meier analysis of each dataset alone did not result in a significant association of ABCG1 and prognosis. Even though not statistically significant, www.nature.com/scientificreports/ www.nature.com/scientificreports/ the Smith dataset showed a concordant trend with those of TCGA, Sveen and Marisa. Unfortunately, we were not able to rule out that the difference in the Kaplan Meier trend could be ascribed to the characteristics of the Sieber cohort patients. Importantly, the three datasets showing significant association of high levels of ABCG1 and worse prognosis (TCGA, Sveen and Marisa) use different platforms and therefore different probes for the detection of ABCG1, and we believe that a concordance of results using different probes strengthens our results.

Discussion
Despite recent advances in cancer therapy (i.e. screening programs allowing the surgical treatment of early-stage CRC patients and targeted therapies) CRC is still among the prevalent causes of cancer-related death 1 . Failure of therapeutic strategies is mainly linked to drug resistance that can be determined by an enhanced ability of cancer cells to detoxify by actively pumping cytotoxic drugs out of the cells, mediated by ABC transporters 3 . Interestingly, increased activity and expression of efflux pumps and detoxifying machinery has been detected in cancer stem cells, the subpopulation of cancer cells mainly responsible for drug resistance and recurrence 3,20 .
The expression and role of ABC transporters in CRC have been intensely investigated, in vitro and in cohorts of patients. The expression of ABCB1 (MDR1/P-GP) has been previously reported as induced by chemotherapeutic agents in CRC cell lines 21 . ABCG2 was also associated with reduced toxicity and intracellular concentration of different types of chemotherapeutics 21 .
Signalling involved in CRC maintenance has been previously shown to regulate ABC transporters. Nuclear factor-kappaB (NF-κB) was shown to regulate ABCB1 and inhibition of NF-κB was able to sensitize colorectal cancer cells to chemotherapy in an ABCB1 dependent way 21 . ABCC3 was proved to be transcriptionally regulated by WNT signalling in CRC, thus likely contributing to acquired drug resistance 22 . Furthermore, MALAT1 silencing down-regulated the expression of ATP-binding cassette transporters (ABC), breast cancer resistance protein (BCRP), and multi-drug resistance proteins including MDR1 and MRP1, resulting in decreased resistance of CRC cells to 5-FU 23 .
Interestingly, given the wide overlap of the substrate, the potential of many ABC transporters involved in resistance to therapy is high. Moreover, the induction of multidrug resistance by treatment with a first drug could associate with the resistance to different chemotherapeutic agents, thus making therapy more difficult 21 . For example, CRC LoVo cells showed high levels of ABCG2, and consequently multidrug resistance, after oxaliplatin treatment and resulted resistant to 5-FU and doxorubicin, too 24 .
ABC transporters have been associated with CRC risk and toxicity in different studies. On one hand, both ABCB1 and ABCG2 haplotypes were associated with risk of CRC in a Danish cohort 25 . On the other hand, ABC transporters polymorphisms were also correlated with toxic side effects caused to healthy cells. In detail in two independent cohorts, ABCC5 and ABCG1 haplotypes were associated with increased severe toxicity 26 . In a different study, SNPs in ABCB1 and ABCC4 were associated with a severe adverse reaction, including haematological toxicity in CRC patients 27 .
Given their undeniable role in conferring resistance to chemotherapy, ABC transporters have been investigated as therapeutic targets. Unfortunately, three generations of inhibitors of the most studied ABC transporters ABCB1 (P-gp) and ABCG2 did not meet expectations in terms of successful clinical trials, mostly due to unforeseen pharmacokinetic interactions, poor selectivity, low potency and high toxicity 21 . Interestingly, the role of ABC transporters in resistance is generally attributed to increased expression and not to acquired mutations 3 . This opens possibilities to indirect targeting of ABC transporters through direct targeting of active signalling pathways in CRC.
Interestingly, the HH-GLI pathway is involved in intense cross-talk with numerous signalling pathways that promote cancer progression, activating GLI in absence of the canonical PTCH/SMO pathway 5 . HH-GLI has been shown to regulate CRC cells maintenance, including cancer stem cell phenotype, metastasis and chemoresistance. Zhang et al. 28 showed that GLI1 and GLI2 mediate 5-FU resistance in CRC cell line LoVo. GLI1 inhibition was shown to revert chemo-resistance in organoids from colorectal cancer patients, and concurrently downregulated cancer stem cells markers, such as c-Myc, CD44 and Nanog 9 . Our results showed a direct involvement of Gli1 in chemoresistance of CRC cells. Indeed, we demonstrated that GLI1 modulates the expression level of a subset of ABC transporters, and through the modulation of GLI1 activity it is possible to impair chemoresistance in CRC cells. www.nature.com/scientificreports/ www.nature.com/scientificreports/ Among the ABC transporters directly regulated by GLI1, ABCG1 is a good prognosis biomarker. Indeed, ABCG1 was expressed at high levels in CRC cells with respect to non-cancerous tissues and to CAFs, and its expression levels in primary tumours are associated with worse prognosis.
ABCG1 is a sterol transporter, thus contributing to cholesterol homeostasis in non-cancerous tissues 29 , however previous studies have described another role for ABCG1 in cancer. Indeed, ABCG1 was found to be involved in multidrug resistance in osteosarcoma, where it was up-regulated in cells with induced doxorubicin resistance 30 . Interestingly, this phenotype displayed also stem-like features, i.e. colony formation ability. Moreover, ABCG1 was recently reported to induce resistance to oxaliplatin and saracatinib and to be regulated by WNT signalling in hepatocellular carcinoma 31 .
Interestingly, SNPs correlating with higher levels of mRNA expression of ABCG1 were correlated with worse prognosis in non-small cell lung cancer 32 .
Of note, our results showed that 4 ABC transporters were up-regulated after GLI1 silencing in Colo205. Despite not being the focus of the current work, we noticed that ABCA1 was recently demonstrated to be a bona fide target of P53 and oncosuppressor in liver cancer 33 . Since P53 was shown to be a target of GLI1 34 and the status of P53 in Colo205 is believed to be wild-type 35 , we speculate that oncogenic GLI1 could indirectly target ABCA1 via P53.
The targeting of HH-GLI pathway in solid tumours has been the focus of previous and current clinical trials involving the targeting of Smoothened (SMO) 36 , however, results have been disappointing so far 37 ; this could be ascribed to the frequent activation of HH-GLI by non-canonical oncogenic pathways 5,7 .
On the other hand, GLI1 has been shown to be a target of arsenic trioxide (ATO), which is FDA approved for the therapy of adult patients with acute promyelocytic leukaemia (APL) 38 . A phase I trial investigating the co-administration of ATO and 5-Fluorouracil/Leucovorin in patients with advanced/relapsed colorectal cancer showed that ATO was well tolerated and that in some patients it was associated with down-regulated thymidylate synthase expression, indicating a therapeutic response, and increased survival 39 ; a later study investigated GLI1 levels in biopsies from the said clinical trial and found that it resulted to be down modulated after ATO administration 40 .
In conclusion, the present study provides a rationale for the consideration of HH-GLI pathway as a therapeutic target in CRC patients.
Indeed, our results indicate that the addition of GLI targeting drugs to CRC treatment strategies is a therapeutic option that could prevent the onset of chemoresistance.

Materials and methods
Cell culture, treatment, over-expression and drug resistance. Colo205 and HCT15 were obtained from the American Type Culture Collection and were grown in RPMI-1640 (supplemented with 10% (v/v) fetal bovine serum, 1% (v/v) penicillin (50 U ml −1 )-streptomycin (50 U ml −1 ) and 2 mM l-glutamine). Cells were routinely checked for mycoplasma contamination by testing with PCR Mycoplasma Detection Kit (ABM, Cat. No. G238).
Cells were treated with 10 μM GANT61 (ENZO Lifesciences), and equimolar concentration of Oxaliplatin (Selleckchem) and 5-Fluorouracil (5-FU) (Selleckchem) as described earlier 41  To induce chemoresistance, cells were treated with increasing and equimolar concentrations of Oxaliplatin and 5-FU. Cells were exposed to a starting concentration of 1 μM in RPMI plus 10% of FBS. The surviving population underwent a stepwise selection of resistant cells by increasing concentration every week to a final concentration of 10 μM. www.nature.com/scientificreports/ Cell viability assay. Cell viability was evaluated by using Celltiter-GLO luminescent cell viability assay (Promega, G7570). CTG reagent was mixed at a 1:1 ratio with supernatant from the treatment plate. The mix was incubated for 10 min at room temperature on the shaker, followed by luminescence measurement using GloMax (Promega).