Background
The clinical outcomes for gastroesophageal cancer are poor. One year survival is only 40–50 % and 5 year survival 10–20 % [
1]. At the time of clinical diagnosis only 30–40 % patients have loco-regionally confined disease that is amenable to potentially curative therapy and the majority of patients relapse systemically after such treatment [
1].
These outcomes are largely the consequence of systemic dissemination at a very early stage and indicate the importance of systemic therapies in disease management [
2,
3]. Accordingly, cytotoxic chemotherapy has value as neo-adjuvant, adjuvant and palliative treatment [
2‐
4]. Cisplatin, oxaliplatin and docetaxel are amongst the most active cytotoxics and key components of combination chemotherapy regimens [
2,
5]. Nevertheless, resistance to cytotoxic drugs is common and severely limits the effectiveness of these treatments by resulting in the delivery of ineffective and toxic therapy.
Accordingly, identification of predictive biomarkers for chemotherapy in gastroesophageal cancer are urgently needed in clinical practice and would enable a stratified approach to treatment selection, and optimise clinical and cost effectiveness. Despite extensive investigation there are no predictive biomarkers for chemotherapy that are recommended for clinical use in gastroesophageal cancer. More recently the use of global molecular analysis tools such as gene expression profiling, array-CGH, exome and whole genome sequencing, has provided more promising leads for predictive biomarkers for chemotherapy in gastroesophageal cancer [
6,
7].
Predictive biomarkers for chemotherapy resistance may also have value as therapeutic targets for agents that would combine effectively with cytotoxic drugs. A clinical proof of principle for the safety, tolerability and effectiveness of combining targeted agents with chemotherapy as part of a biomarker directed stratified therapy approach, has been demonstrated recently in gastroesophageal adenocarcinoma, combining trastuzumab with cisplatin and 5FU in patients whose tumours are HER 2 positive [
8]. However only 10–15 % of gastroesophageal adenocarcinomas are HER2 positive and the identification of clinically effective targeted agents has proven challenging in gastroesophageal cancer, with Phase III trials evaluating the addition of targeted therapies against Epidermal Growth Factor Receptor (EGFR), Vascular Endothelial Growth Factor (VEGF), Mammalian Target of Rapamycin (mTOR) Mamalian mTOR, to cytotoxic chemotherapy, not demonstrating any benefit [
9‐
12], and there are no targeted therapy options at all for squamous cell carcinoma of the esophagus. More recently, the addition of the VEGFR-2 targeting agent Ramicurumab to paclitaxel chemotherapy has been shown to be beneficial in a phase III randomised controlled trial, but as yet there are no predictive biomarkers for Ramicurimab, which is likely to significantly limit the cost effectiveness of this treatment [
13]. Overall, there is a clear ongoing clinical need to identify further new targets and biomarker combinations for gastroesophageal cancer, in particular those which might combine effectively with cytotoxic chemotherapy.
In order to address this we utilised gastroeosphageal cancer cell lines selected for resistance to cisplatin, oxaliplatin and docetaxel as models for the identification of new markers of drug resistance and candidate novel therapeutic targets. Such models have been widely used and have provided new insights into mechanisms of drug action and resistance, but translation from such studies to clinically useful targets or biomarkers has been more limited [
14]. In light of this, and the more recent demonstration of the usefulness of global molecular profiling tools with gastroesophageal cancer cell line models to identify predictive markers and targets [
6,
7], we used global gene expression profiling on our cytotoxic resistant cell lines to identify lead molecules for further investigation. To further determine their clinical utility as predictive biomarkers and/or novel therapeutic targets leads were validated by quantitative real-time polymerase chain reaction (qRT-PCR), assay of relevant tumour metabolites in key biological pathways, pharmacological inhibition of an identified target, and evaluation of predictive and prognostic value in an independent panel of gastric cancer cell lines and tumour tissues from gastroesophageal cancer patients.
Discussion
We approached the clinical need to identify predictive biomarkers for cytotoxic chemotherapy and new therapeutic targets in gastroesophageal cancer by using a hypothesis generating approach with global gene expression profiling of a panel of cell lines selected for resistance to clinically used cytotoxics. In order to enable rapid clinical translation, we prioritised further investigation of identified lead candidates for which agents already exist, as these would provide new strategies for combinations of targeted and cytotoxic therapies to overcome resistance and increase clinical effectiveness.
Sphingosine metabolism was identified as a lead candidate target following gene expression profiling and biological pathway mapping. In drug resistant gastroesophageal cancer cell lines we observed increased levels of sphingosine metabolite, S1P and increased cisplatin sensitivity in response to pharmacological inhibition of SPHK1, a kinase required for metabolism of sphingosine to S1P, as well as correlation between SPHK1 protein expression and cisplatin sensitivity in an independent gastric cancer cell line panel. Furthermore, SPHK1 protein expression was associated with worse survival in a cohort of patients with gastroesophageal cancer who received cytotoxic neo-adjuvant chemotherapy. Our data demonstrate an inverse relationship between the expression of SPHK1 (increased) and SGPL1 (decreased) in resistant cell lines and we propose this leads to increased cellular S1P and cytotoxic drug resistance.
S1P is a phospholipid with many functions, formed intra-cellularly through the phosphorylation of sphingosine by sphingosine kinases (2 isoforms SPHK1 and SPHK2) [
39]. S1P is actively transported out of the cytosol to act via membrane S1P receptors, although receptor independent effects, including intracellular targets are also recognised [
39]. Alternatively, irreversible cleavage of S1P by SGPL1 can occur in the cytosol. A role for SPHK1 and S1P in drug resistance in gastroesophageal cancer is consistent with many previous investigations, which have suggested a pathogenic role in several cancer types including gastric cancer [
21], where SPHK1 overexpression is observed in tumour cells and associated with increased stage and poor survival [
20]. In addition, investigations in cancer and non-cancer cells demonstrate that increased SPHK1 is associated with increased production of S1P in cells and S1P promotes cell proliferation, angiogenesis and inhibits cell death all of which could promote cell survival following cytotoxic drug insult and hence induce resistance [
22‐
29,
37,
40,
41]. In addition SPHK1 activity and levels of S1P have been demonstrated to be involved in resistance to cytotoxic and targeted agents in a variety of cancer types, although not oesophageal or gastric adenocarcinoma drug resistance [
30,
31,
34‐
36,
40,
42]. Here we provide the first evidence for the importance of sphingosine metabolism, and in particular SPHK1 and S1P, in resistance to cytotoxics in gastroesophageal cancer.
Accordingly S1P could lead to cytotoxic drug resistance in gastroesophegal cancer acting in an autocrine or paracrine manner via cell surface S1P receptors following transportation out of the cytosol. Alternatively S1P may mediate cytotoxic drug resistance acting intracellularly by counteracting apoptosis mediated by its pro-apoptotic precursor ceramide or interaction with known intracellular targets involved in cancer pathogenesis and cytotoxic drug resistance such as Histone deacetylase 1 (HDAC1) and Histone deacetylase 2 (HDAC 2) to which S1P directly binds and inhibits, and TNF Receptor-Associated Factor 2 (TRAF 2), or Protein Kinase C (PKC) [
39]. Further investigation is required to determine which of these potentially pleomorphic mechanisms is important for cytotoxic drug resistance in gastroesophageal cancer, and the key mechanisms may be different for different drugs. Nevertheless our data demonstrates that S1P production controlled by SPHK1 and SGPL1 are key determinants of cytotoxic drug resistance and that decreasing S1P production in cancer cells could lead to increased cytotoxic sensitivity.
Rex et al. demonstrated no effect on the viability of cancer cell lines (cultured with and without serum starvation), nor on the growth of xenografts, with the use of highly specific SPHK1 and SPHK2 inhibitors [
43]. However, the effect of SPHK1/2 inhibition under conditions of exposure to cytotoxic drugs was not investigated. Accordingly, it is not possible to rule out the importance of S1P and SPHK1 on tumour cell viability in specific circumstances not tested in their experiments- which would include exposure to cytotoxics. In addition it is possible that SPHK1 may have different effects in different types of cancer, and while Rex et al., provide data on a variety of cancer types, they have not provided data on gastroesophageal cancer cell lines. Our experiments with gastric adenocarcinoma cell lines suggest that safingol, a SPHK1 inhibitor, has cytotoxic effects as a single agent as well as acting synergistically with cisplatin.
Our finding of an inverse relationship between SPHK1 and SGPL1 expression in gastroesophageal cancer associated with increased levels of S1P and cytotoxic drug resistance is consistent with a recent report in prostate cancer, where a similar inverse relationship between expression of SPHK1 and SGPL1 was noted leading to increased production of S1P and an association with resistance to docetaxel [
37]. In addition, increased S1P in glioblastoma multiforme tissues is associated with increased SPHK1 and decreased expression of S1P phosphatase (SGPP2) [
22]. Together with our data, these observations suggest that increased SPHK1 and decreased SGPL1 or SGPP2 may be a relatively common pathogenic mechanism that could also be involved with therapy resistance in several cancer types.
Our data suggest that increased SPHK1 expression in gastroesophageal cancers has predictive impact indicating chemotherapy resistance in gastroesophageal cancer patients, consistent with our cell line findings, but not therapy independent prognostic value. A previous study by Li et al., demonstrated that high SPHK1 expression in resected gastric cancers was associated with worse survival [
20]. Li et al., noted that the relationship between increased SPHK1 and survival was only significant for patients with higher stage (III and IV) disease who were often given adjuvant chemotherapy (chemotherapy details not provided) after surgery and there was no significant correlation between SPHK1 and survival for early stage (I and II) disease patients who were not given adjuvant chemotherapy after surgery. Given our findings reporting an effect of SPHK1 on resistance to chemotherapy, this interaction between SPHK1 and chemotherapy may explain the contrasting findings by Li and colleagues in early versus late stage disease. Pan et al., reported worse survival in esophageal squamous cell carcinoma patients treated with surgery alone (no neo-adjuvant chemotherapy) who had high SPHK1 protein expression [
19]. Therefore, the impact of SPHK1 may vary according to histological sub-type of gastroesophageal cancer. Our investigation included 1 squamous cell carcinoma oesophagus cell line (OE21), but only a minority (18 %) of patients in our clinical cohort had squamous cell carcinomas. There was no observed association between SPHK1 IHC and histology and multivariate analysis did not suggest and differential effect of SPHK1 in squamous versus adenocarcinoma (although numbers were small in this analysis). Further prospective studies are required to determine the interaction between SPHK1 (SGPL1 and/or S1P) and histological subtype and their therapy independent prognostic versus predictive impact.
Here we have investigated the relationship between resistance and benefit from neo-adjuvant chemotherapy in gastroesophageal cancer patients. It would be valuable to investigate the relationship between SPHK1 expression and response to palliative chemotherapy to determine if there is any differential impact according to the stage of the disease. This would be useful in planning future clinical trials of therapies targeting SPHK1. The small numbers of patients in our surgical cohort who would have recurred and received palliative chemotherapy mean that such an analysis in our cohort would not be informative. In addition there may be a change in the SPHK1 expression from the primary tumour that was resected and the recurrent disease at a later date when palliative chemotherapy might be administered. Investigation in a large cohort of patients with advanced gastroeosphageal cancer that have received palliative chemotherapy would be worthwhile. Further investigation of the relationship between the ratio of SPHK1 and SGPL1 protein expression determined by immunohistochemistry or other methods, and the response/benefit from both neo-adjuvant and palliative chemotherapy in clinical cohorts of patients would also be valuable.
There are a number of SPHK1 and S1P inhibitors in clinical development as anti-cancer agents that confer increased sensitivity to cytotoxic drugs, targeted agents or radiotherapy and/or have single agent activity in preclinical cancer models [
30,
31,
38,
44‐
49] (Table
2). Therefore our findings in gastroesophageal cancer could be readily translatable to the clinic. SPHK1 may be a useful biomarker to identify patients who are likely to be resistant to cytotoxic chemotherapy and who would benefit from the addition of a SPHK1 or S1P inhibitor to a cytotoxic chemotherapy regimen. Alternatively, SPHK1 and S1P inhibitors may be useful as part of second line therapies in patients who have clinical resistance to cytotoxic chemotherapy, a situation where current therapies have limited efficacy. We investigated this
in vitro by examining the effect of safingol on cisplatin resistance in gastric cancer. We chose safingol (L-theo dihydrospingosine) rather than a more specific SPHK1 inhibitor or SPHK1 knock down by siRNA mediated knock down of SPHK1, since it is the most developed SPHK1/S1P inhibitor as an anti-cancer agent and in particular safingol has demonstrated safety in combination with cisplatin in a phase I clinical trial and lead to decreased serum S1P in treated patients [
38]. Accordingly using safingol would provide useful data to facilitate more rapid translation to a clinical therapy and similar to the phase I trial, we investigated the combination of cisplatin safingol in gastrooesophegal cell lines. Safingol is a potent competitive inhibitor of SPHK1 with a Ki <0.4 μM, it is a substrate for but not inhibitor of SPHK2, and it is also known to inhibit PKC but the Ki is considerably higher at 33–40 μM [
50‐
53]. In addition safingol treatment of colon cancer cells leads to decreased S1P and increased sphingosine [
46]. In our evaluation of safingol in the cisplatin resistant cell line AGS
CIS4 and the gastric cancer cell line N87 we used safingol at concentrations ranging from 0.375–12 μM, which is well within the readily achievable serum plasma concentrations of safingol in cancer patients and at which significant decreases in serum S1P are reported [
38]. As well as synergy with cisplatin our experiments demonstrated cytotoxic activity of safingol, which together with the safety data from the clinical phase I trial with concomitant cisplatin administration suggest that this would be a feasible and appropriate combination to investigate in early phase trials in gastroesophageal cancer patients. Further pre-clincial experiments with specific SPHK1 inhibitors and siRNA mediated knockdown of SHK1 would allow additional exploration of the potential of SPHK1 as a target and the usefulness of combining with other cytotoxic and targeted drugs.
Table 2
SPHK1 and S1P inhibitors in clinical development as anti-cancer therapies
Safingol | SPHK1 inhibitor | Demonstrated safety in combination with cisplatin in Phase I trial in solid tumours | |
Fingolimod | S1P receptor antagonist | Licenced for use in multiple sclerosis. No data on clinical use in cancer | |
(FTY720) |
Sonepcizumab | Humanised S1P neutralising monoclonal antibody | Demonstrated safety in Phase I trial in solid tumours | |
(LT1009) |
Competing interests
RP has received research funding for projects in gastroesophageal cancer distinct from the data presented in this manuscript from, Astra Zeneca, and Roche and received reimbursement for advice provided regarding the clinical management of gastric cancer from Eli Lily and reimbursement for attendance at scientific congresses from Pfizer, and Eli Lilly.
Author Contributions
RP is the chief investigator of the study, responsible for the inception of the study and its overall conduct, overall experimental design, and provided patients and tumour materials for the study, performed or supervised all data analysis and it’s interpretation, and wrote the manuscript. KM performed cell line experiments, assisted with their design, analysed and interpreted the data generated and assisted with manuscript writing. ECD assisted with experimental design, data analysis, data interpretation and manuscript writing. GM provided patients tumour materials for the study, assisted with experimental design and data interpretation. KP performed cell line experiments and assisted with data analysis and interpretation. HG and PT provided cell line materials for the study, performed cell line experiments and assisted with their design, data analysis, and data interpretation . SL and RG performed cell line experiments, and undertook data analysis. YO performed immunohistochemistry of patient samples, and performed analysis of this data. GB,ADS,GU and MF provided, patients and tumour materials for the study, and performed data analysis and interpretation of patient related data. JB performed S1P assays, and analysed and interpreted this data. All authors contributed to revision of the manuscript and final approval of the manuscript.