Overexpression of sortilin is associated with 5‐FU resistance and poor prognosis in colorectal cancer

Colorectal cancer (CRC) is the third most common cancer worldwide. Even if 5‐fluorouracil (5‐FU) is used as the first‐line chemotherapeutic drug, responsiveness is only 20‐30%. Acquired resistance to 5‐FU contributes to both poor patient prognosis and relapse, emphasizing the need to identify biomarkers. Sortilin, a vacuolar protein sorting 10 protein (Vps10p), implicated in protein trafficking, is over expressed in CRC cell lines cultured 72 hours in presence of 5‐FU. This overexpression was also observed in 5‐FU‐resistant cells derived from these cell lines as well as in CRC primary cultures (or patients derived cell lines). A significantly higher expression of sortilin was observed in vivo, in 5‐FU‐treated tumours engrafted in Nude mice, as compared with non‐treated tumour. A study of transcriptional regulation allowed identifying a decrease in ATF3 expression, as an explanation of sortilin overexpression following 5‐FU treatment. In silico analysis revealed SORT1 expression correlation with poor prognosis. Moreover, sortilin expression was found to be positively correlated with CRC tumour grades. Collectively, our findings identify sortilin as a potential biomarker of 5‐FU resistance associated with poor clinical outcomes and aggressiveness in CRC. As a new prognostic factor, sortilin expression could be used to fight against CRC.

5-FU-based systemic chemotherapy. 6 The mechanisms involved in 5-FU resistance are numerous and complex. 7 These mechanisms remain at the centre of intensive research programmes to find the origin of the resistance 8 or to discover new molecules that are able to overcome this resistance. 9 Sortilin, which is ubiquitously expressed in mammalian cells, 10 acts as a regulator of intra-and extra-cellular protein sorting and trafficking, 11,12 as well as a co-receptor of neurotrophins (NTs). 13 The neurotrophin family consists of four members: nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), NT4/5 and NT3.
Each one of the four NTs strongly binds to a tropomyosin receptor kinase (Trk) receptor (NGF to TrkA, BDNF and NT4/5 to TrkB, and NT3 to TrkC) and promotes cell survival. All NTs also bind with low affinity to the neurotrophin receptor p75 (p75 NTR ) alone, then triggering cell death. NTs may also bind with higher affinity to the p75 NTR /sortilin/Trk or sortilin/Trk complexes, and then they promote cell survival. [13][14][15] The NTs pathways have been implicated in multiple types of cancer, notably digestive. 16 TrkB/BDNF is the only NTs complex which has been firmly shown to play oncogenic roles in CRC development, including tumour growth, progression and cell survival. [17][18][19][20] The roles played by sortilin in cancers are more complex: depending on the origin of the tumour, it has been described either as a good or a bad prognosis marker of tumour aggressiveness. In breast cancer, ovarian carcinoma and in neuroendocrine tumours, targeting sortilin has been reported to inhibit tumour metastasis and to promote tumour cell apoptosis; this observation underscores the bad prognosis linked to a high level of sortilin expression. [21][22][23][24] On the opposite, our team has recently reported that the loss of sortilin promoted lung cancer cell proliferation, in relation with the epidermal growth factor receptor (EGFR) signalling efficiency. 12 More especially, sortilin acts as a key regulator in balancing CRC cell survival and death through its association with, respectively, TrkB and p75 NTR . 20 Moreover, besides the role played by the full-length protein, sortilin can also be cleaved and released as a soluble form that promotes the survival of cancer cells and enhances their invasive potential. 25,26 Therefore, sortilin can be considered as a multifunctional protein, able to play various roles, as a receptor, a trafficking chaperone or a soluble signalling molecule. A link between altered SORT1 expression and cancer treatment has only been reported for B acute lymphoblastic leukaemia. 27 However, to date, a relationship between sortilin expression and chemotherapy efficiency and, more specially, its role in 5-FU resistance, has not been demonstrated yet.
This work was aimed at exploring a possible link between cancer therapy and sortilin expression in 5-FU-treated CRC cells. It was found that short-(72 hours) and long-term (one month) treatments resulted in sortilin overexpression. This enhancement was evidenced in CRC cell lines, as well as in primary cultures and even in tumours obtained after transplantation of CRC cells in Nude mice. We also found that sortilin enhancement was accompanied with a down-regulation of ATF3 expression. In silico analyses revealed that SORT1 expression was associated with poor clinical outcomes. Moreover, we found that sortilin overexpression was associated with higher CRC tumour grades in patient tissue samples.
To obtain long-term 5-FU-resistant cells (one month), a modified protocol was followed. 30 In, 30 it took nearby 32 weeks to obtain resistant cells at a final 5-FU concentration of 2µg/ml. Since the final concentration of 1 µg/mL, corresponding to the IC50 of the cell lines, was chosen, our revisited protocol allowed obtaining resistant cells in only 4 weeks. Cells were seeded in T75 flasks at 5 × 10 6 cells/10 mL of media (WiDr) or 2.5 × 10 6 cells/10 mL of media (SW620 and primary cultures) for untreated conditions, and 7 × 10 6 (WiDr) or 5 × 10 6 (SW620 and primary cultures) for 5-FU treatment.
During one month, 2 mL of media containing 5-FU (Sigma-Aldrich) was added every 72 hours to the cultures in order to reach a final concentration of 8 µmol/L, corresponding to IC50.

| Nude mice xenografts and pharmacological treatment
Animal study was conducted in accordance with guidelines established by the regional Institutional Animal Care and Use Committee

| Cellular metabolic activity
Cells were seeded in 96-well plates. The next day, they were exposed

| Indirect immunofluorescence staining
Fixation, permeabilization/saturation and staining were performed as previously described. 20  Images were analysed using Zeiss ZEN and ImageJ softwares (NIH, Bethesda, MD, USA).

| Real-time quantitative polymerase chain reaction (RT-q-PCR)
RNA extraction, reverse transcription and real-time quantitative PCR analyses were performed as previously described. 35 38 were extracted from SurvExpress-Biomarker validation for cancer gene expression.

| Statistical analysis
Results were analysed by Kruskal-Wallis test for immunohistochemistry, and by two-way ANOVA/ANCOVA for others (StatView

| A 72 hours 5-FU treatment correlates with sortilin overexpression in CRC cell lines
In previous work, we demonstrated that NTs are survival factors. 19,20 Since 5-FU responsiveness is only 20-30%, the possible link between NTs and 5-FU treatment efficiency was studied in two different CRC cell lines. First, the appropriate 5-FU concentration for treatment of WiDr and SW620 was determined.
WiDr and SW620 were treated with increasing 5-FU concentrations for 72 hours. As shown in Figure

SW620
F I G U R E 2 Reponses to 5-FU long-term treatment (8 µmol/L, one month) of WiDr, SW620 and primary cultures: cell death, insensitivity to 5-FU increasing concentrations and sortilin expression. After one month of 5-FU treatment, A, cells were analysed by flow cytometry for detection of living and dead cells (PI/Annexin-FITC) as described before. Histograms are means from at least three independent experiments. B, Metabolic activity of 5-FU-resistant cells was determined using MTT assay (Promega, France). Untreated cells (maintained in culture during one month) were also analysed and used as control (CTL). Significant P-values are indicated in the graphs: *P < .05; **P < .01; ***P < .001. Sortlin protein expression was analysed by Western blotting from whole-cell lysates of 5-FU-resistant (5-FU) and -untreated cells (CTL) in WiDr and SW620 C, and in primary cultures D,. Actin was used as loading control. Histograms are the means from at least three independent experiments. Significant P-values are indicated in the graphs *P < .05, **P < .01, ***P < .001. E, Sortilin transcriptomic expression of 5-FU-resistant cells (one month) and -untreated cells (CTL) was analysed by RT-q-PCR. Sortilin mRNA expression (SORT1) upon 5-FU was normalized to both HPRT transcriptomic expression and that of untreated cells. Histograms are the means from at least three independent experiments. Significant P-values are indicated in the graphs *P < .05, **P < .01, ***P < .001 In a previous study, we demonstrated the implication of NTs in the survival of CRC cells. 19,20 So, even if we did not see any lethal effect of 5-FU, we resolved to look after TrkB, p75 NTR and sortilin expression. After 72 hours of 5-FU treatment, we did not observe any effect on TrkB and p75 NTR expression ( Figure 1SC). However, a significant enhancement of sortilin expression was observed in both cell lines ( Figure 1C; P = .007 for WiDr and P = .036 for SW620). By examining the main proliferative pathways linked to NTs signalling, the P-Akt/Akt ratio increased in WiDr (P = 0,016), and significant changes were observed in SW620 ( Figure S1D) with a decrease of the P-Akt/Akt ratio (P = .004) and an increase of the P-Erk/Erk ratio (P = .0003). However, because the P-Erk/Erk signalling pathway seemed to increase in the two cell lines after 5-FU treatment, and more especially in SW620, this result could justify the persistence of the proliferative activity of this latter cell line ( Figure 1SB).

| Long-term (one month) 5-FU-resistant cells exhibit sortilin overexpression
This short-term 5-FU treatment (72 hours) had cytostatic rather than cytotoxic effects, since the percentage of apoptotic cells failed to vary between control and treated conditions in both cell lines ( Figure 1B). So, we resolved to extend the 5-FU exposure time to one month (8 µmol/L every 72 hours) based on IC50 for WiDr, SW620 and six primary cultures. Even if this prolonged 5-FU treatment induced an important drop in living cells (85% for WiDr, 88% for SW620 and 72% for primary cultures), it allowed the survival of more than 10% of the cells (Figure 2A).
At the end of the treatment, surviving cells were exposed to increasing doses of drug (0 to 64 µmol/L) for 72 hours and were shown do display a strong resistance to 5-FU. Metabolic activity of these 5-FU-resistant cells showed absolutely no sensitivity towards 5-FU, up to 64 µmol/L, contrary to control cells which had been maintained in drug-free culture during one month ( Figure 2B). This acquired resistance was obtained with the two cell lines and with primary cultures.
As before, we evaluated NTs receptor (TrkB, p75 NTR and sortilin) expression in 5-FU-resistant cell issued from the two lines and from primary cultures. Once again, after one month of 5-FU treatment, only the expression of sortilin was significantly up-regulated in each cell line (P = .008 for WiDr and P = .004 for SW620; Figure 2C) and in primary cultures as well (P = .002; Figure 2D). Similarly to the data obtained after 72 hours treatments, NTs receptor expression was unchanged (data not shown). Probably due to the low proportion of living cells, proliferative pathways were dramatically affected and were not exploitable (data not F I G U R E 3 Sortilin localization and transcriptional regulation upon long-term 5-FU treatment. A, To analyse sortilin localization, a subcellular fractionation was done on WiDr, SW620 and primary cultures treated by 5-FU. Extracted proteins from different cell component (C for cytoskeleton; M for membranes; N for nucleus; Chr for chromatin; Ctsk for cytosquelet) were subjected to Western blot analysis as described in materials and methods section. Anti-PARP antibody was used as a control for nuclear proteins, and anti-Tubulin antibody was used as a control of cytoplasmic proteins. B, Indirect immunofluorescence staining was analysed using confocal microscopy (LSM 510 META; Zeiss, Göttingen, Germany). Anti-GM130 was used as a Golgi marker and nuclei were stained with DAPI. Images were processed with the ZEN software application, and surface plots of the fluorescence data were generated with the image processing program ImageJ. C, ATF3 expression was analysed after subcellular fractionation of WiDr and SW620 as described just above

| 5-FU disrupts sortilin localization and transcriptional regulation
Because 5-FU interferes with numerous cellular processes notably implicated in cytoskeleton organization and function 39 and because sortilin is a key protein implicated in intracellular trafficking, 40 its localization was evaluated in 5-FU-resistant cells (WiDr, SW620 and primary cultures). After 5-FU long-term treatment (one month), sortilin was found to be accumulated in membranes and in the cytoskeleton, contrary to control cells ( Figure 3A). Moreover, sortilin localization is restricted to the Golgi as shown by its co-localization with GM130, a Golgi marker ( Figure 3B) which is more easily seen in SW620. Furthermore, transcriptomic regulation of sortilin expression was analysed in WiDr and SW620. Since it has been demonstrated that ATF3 is able to bind the SORT1 promoter and repress SORT1 transcription, 41

| 5-FU treatment increases sortilin expression in tumours from Nude mice xenografted with CRC cell lines
In order to explore the correlation between sortilin expression and the responsiveness of tumours to 5-FU, WiDr and SW620 were grafted into Nude mice. Once tumour volumes reached 100 mm 3 (day 10 for WiDr and day 17 for SW620), mice were treated (Treatment Initiation, T. I.) with 5-FU or vehicle (CTL) every 72 hours until sacrifice (day 28, ie one month). WiDr-derived tumours regressed significantly relative to control after 15 days of 5-FU treatment ( Figure 4A). Despite the evident sensitivity of these tumours to 5-FU, the regimen was not sufficient to induce total tumour regression, suggesting that a 5-FU-resistant subpopulation of cells survived the treatment. By contrast, SW620derived tumours were totally resistant to 5-FU and did not regress at all ( Figure 4A). Moreover, 5-FU-treated tumours from both WiDr and SW620 expressed elevated levels of the sortilin protein ( Figure 4B,C) contrary to other NTs expressions (data not shown). In vivo, even though cells from low-grade tumours (WiDr) responded better to 5-FU than those derived from high grade tumours (SW620), sortilin protein expression was significantly higher in either one of the persistent tumours following 5-FU treatment. These in vivo data are in complete agreement with the results obtained in vitro highlighting the role of sortilin as a marker of 5-FU resistance.

| SORT1 mRNA expression is associated with poor clinical outcomes and sortilin expression is positively correlated with CRC grades
Finally, once demonstrated the link between 5-FU resistance and sortilin, we assessed its expression level in relation with prognosis of CRC. Thus, we checked the expression of SORT1 mRNA in relation with patient status. The in silico analysis of data bases showed that SORT1 mRNA expression levels were significantly elevated in highrisk patient groups characterized by the worst survival, disease-free survival (DFS) and relapse-free survival (RFS), who are at the highest risk of developing new tumours after treatments ( Figure 5A-C).
Moreover, ex vivo quantification of sortilin in tissues from 133 patients with benign (n = 7) or colorectal tumours from three tumour grades (n = 8, 98, and 20 for I, II, and III, respectively) revealed a positive correlation of sortilin protein expression with tumour grade ( Figure 5D) while a simple upward trend was observed for CRC stages (data not shown). These observations have to be put together with those obtained in vitro: the sortilin level of the WiDr cell line, a reference for low tumour grade, was significantly lower than the level found in the SW620 cell line, a reference for high tumour grade ( Figure 5E and Figure 1C). These findings about sortilin expression in various environments indicated that not only is sortilin a 5-FU resistance marker but also a biomarker for CRC aggressiveness associated with poor clinical outcomes.

| D ISCUSS I ON
CRC is the second leading cause of cancer-related deaths. First-line chemotherapy of CRC uses 5-FU, a pyrimidine analogue that inhibits thymidylate synthase, and whose metabolic activation results in its incorporation into RNA, which interferes with RNA function and prevents DNA synthesis. Even if 5-FU remains the reference molecule used in combination with other chemotherapeutic drugs, it is highly toxic and can lead to patients' death. 42 This observation recently suggested a fine monitoring 43 implying assays of dihydropyrimidine dehydrogenase (DPD) activity, 44,45 the enzyme responsible for the catabolism of 5-FU. Moreover, resistance to 5-FU is a major clinical problem and its mechanism is highly complex. 7 Even if different combinations are currently used to fight CRC, as referred as cytotoxic chemotherapy, molecular-targeted therapy and immunotherapy, the need to find new and reliable biomarkers remains because resistance always comes soon or later and whatever combination used. 46 Pathology tumour, node, metastasis (pTNM) staging, currently the most significant prognostic factor in CRC, is highly relevant to therapeutic decision-making. 47 In this context, the multifunctional protein sortilin represents a promising candidate even though its function is cell-type-specific: in F I G U R E 5 High SORT1 expression is associated with poorer survival, DFS and RFS, and sortilin expression is associated with higher CRC tumour grades. SORT1 expression levels by risk group according to survival (n = 77; A), DFS (n = 545; B) and RFS (n = 37; C) of patients suffering from CRC, extracted from 'SurvExpress' database. Significant P-values are indicated in the graphs *P < .05, **P < .01, ***P < .001. D, Sortilin protein expression, observed by immunohistochemistry (magnification, 50×) on tissue microarrays including seven samples of benign tissue and 126 samples from CRC tumours of different grades (n = 8, 98, and 20 for grades I, II and III, respectively). Quantification was performed by calculating the immunochemistry intensity score described in materials and methods. E, Sortilin expression in whole-cell lysates of WiDr and SW620. Actin was used as a loading control breast cancer 23 and neuroendocrine tumours, 24 it is associated with tumorigenesis and poor prognosis, but the opposite is true for nonsmall cell lung cancer. 12 In this study, sortilin is overexpressed in vitro after 5-FU treatments whether short (72 hours) or long (one month) and also in in vivo-treated tumours. Elevated SORT1 mRNA expression was associated with poorer survival, DFS and RFS in high-risk CRC patients. Its overexpression is also associated with high CRC grades. These findings provide the first evidence that sortilin is a biomarker of 5-FU resistance in CRC, which is associated with poorer prognosis and clinical outcomes, and higher tumour grades.
Consistent with this, the cell lines used in this study were chosen on the basis of their (a) tumour grades (low for WiDr; high for SW620 29