Human colon cancer cells (primary adenocarcinoma, WiDr and metastatic adenocarcinoma, SW620; American Type Culture Collection, Manassas, VA, USA) and HMEC-1 cells (Centers for Disease Control and Prevention, Atlanta, GA, USA) were used (28). Cancer cell genotypes are listed in Table I and HMEC-1 cells were assumed to be wild-type (no contrasting evidence was reported). Cells were cultivated in high (25 mM) or normal (5.6 mM) glucose Dulbecco’s modified Eagle’s medium (DMEM; Gibco, Carlsbad, CA, USA) with 10% cosmic calf serum (CCS; Thermo Scientific HyClone, Logan, UT, USA) in standard conditions (humidified at 37°C in 5% CO₂), unless specified otherwise. Glucose concentrations in cell culture were monitored over time (Optium Xceed; Abbott Diabetes Care, Doncaster, Australia), demonstrating that glucose concentrations reduced by ~18% over 24 h in a confluent cell culture.

Cells were cultivated in DMEM with 10% CCS with high (25 mM) or normal (5.6 mM) glucose concentrations and incubated for 24 h. Plain media, IGF-1 (13 nM; Sigma-Aldrich, St. Louis, MO, USA) or insulin (10 nM; Invitrogen Life Technologies, Carlsbad, CA, USA) were added to the cells 24, 4 or 0 h prior to the addition of 5-FU (0.2–200 μM), oxaliplatin (0.001–100 μM) or irinotecan (0.001–100 μM). Each treatment was tested in four wells per experiment, with three independent experiments, and the cells were treated for 72 h. Cell viability was estimated by standard 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (29), calculated as a percentage of the controls (0 μM of chemotherapy drug) and adjusted for background absorbance. The concentration of drug able to reduce viability to 50% (IC₅₀) was calculated from equations obtained by model fitting. Although it is accepted that MTT, an indicator of metabolically active mitochondria, potentially overestimates the number of viable cells compared with several other viability methods (30), it remains widely used in drug discovery and allows for comparisons with previously published data.

Media was replaced with serum-reduced DMEM (0.1% CCS) 24 h prior to and throughout the experiment. The cells were treated for 4 h with IGF-1 (13 nM), insulin (10 nM), CoCl₂ (100 μM positive control) (31) or plain media (negative control). Nuclear protein fractions were extracted and analysed by western blot analysis following standard protocols (31). A total of 40 μg protein extract was loaded per well for the total and nuclear fractions. Anti-HIF-1α (1:250; clone 54/HIF-1α; BD Biosciences, San Jose, CA, USA) and anti-β-actin (1:2,000; clone AC-15; Sigma-Aldrich) were simultaneously used as primary antibodies to detect HIF-1α and to verify equal loading of protein. Horseradish peroxidase (HRP)-conjugated polyclonal goat anti-mouse antibody (1:1,000; DakoCytomation, Glostrup, Denmark) was used as a secondary antibody. For IGF-1R protein detection, total protein extracts were analysed, using anti-human IGF-1R (1:100; C-20; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA) as the primary antibody and HRP-conjugated polyclonal goat anti-rabbit antibody (1:1,000; DakoCytomation) as the secondary antibody.

Cells were plated into 96-well plates provided in the human/mouse total HIF-1α immunoassay kit (R&D Systems, Minneapolis, MN, USA) at recommended concentrations (10⁴/well) and cultivated under standard conditions. The media was replaced with serum-reduced DMEM (0.1% CCS) 24 h prior to and throughout the experiments. The cells were treated with CoCl₂ (100 μM positive control), plain media (negative control), IGF-1 (13 nM) and insulin (10 nM) for 4 h. The cells were then fixed with 4% formaldehyde and analysed immediately by immunoassay according to the manufacturer’s instructions.

SPSS 16.0 (SPSS, Inc., Chicago, IL, USA) and Microsoft Excel 2007 (Microsoft Corporation, Redmond, WA, USA) software were used for the statistical analysis and graphical presentation of the results.

For the MTT assay results, several linear regression models using ln transformation of drug concentration (μM) or cell viability (percent) or the two variables together were tested. The model was considered to be a good fit if the R² value was >0.8. Different models were allowed to be used for the various cell lines and chemotherapy drugs. However, within these, the same model was used across the various growth factors, pretreatments and media conditions. Selected models were used to calculate the IC₅₀ and ultra-low dose (ULD) values and for multiple regression analysis.

The following linear regression models were selected to fit the viability data according to the R² values: ln transformation of drug concentration for WiDr treated with 5-FU and oxaliplatin, for all treatments of SW620 and for HMEC-1 treated with 5-FU; and ln transformation of cell viability for WiDr treated with irinotecan and for HMEC-1 treated with oxaliplatin and irinotecan. IC₅₀ values were calculated from the equations obtained by model fitting. These values were used to compare the effect of growth factors on the cellular response to chemotherapy. Independent sample t-tests were used to compare IC₅₀, ULD and HIF-1α protein levels between the various treatments. In the multivariable regression analysis the effect of growth factors on the response to chemotherapy drugs was estimated by B coefficients (regression ‘slopes’).

Glucose concentrations were specifically selected to be clinically relevant and are those used widely in cancer cell culture studies. The lower glucose concentration (5.6 mM) approximates the lower threshold for normal fasting glucose and the high glucose concentration (25 mM) falls in the hyperglycemic range associated with diabetes (32). Specifically, high glucose concentrations are standard in cancer cell culture studies (33). Growth factor concentrations were selected from previously published patient data; 10 nM insulin (plasma, 2 nM) (34) and 13 nM IGF-1 (plasma, 109 ng/ml) (35).

IGF-1 and insulin increased the proportion of cells with metabolically active mitochondria (cell viability) of stromal and cancer cells by between 13 and 55% (HMEC-1 in high glucose with insulin and WiDr in high glucose with IGF-1, respectively). IGF-1 generally increased viability more than insulin (with the exception of WiDr under normal glucose conditions), and an increased viability was more apparent in high glucose than in normal glucose conditions (with the exception of SW620 with IGF-1 and HMEC-1 with insulin) (Table I).

Western blot analysis confirmed that the two cancer cell lines expressed IGF-1R (36), with levels not notably affected by glucose concentration (Fig. 1A). IGF-1R levels appeared higher in WiDr compared with SW620, as reported previously (36).

The concentrations of chemotherapy agents used in the current study were within the clinically relevant ranges: 5-FU, 0.2–200 μM (maximum plasma concentration, 426 μM); oxaliplatin, 0.001–100 μM (maximum plasma concentration, 3.3 mM); and irinotecan, 0.001–100 μM (maximum plasma concentration, 10 mM) (37–42).

The mean IC₅₀ and results of the t-tests for each condition in all cell lines are presented in Table II. For the majority of cell lines, no significant difference was identified in the concentrations of drugs required to reduce IC₅₀ between the growth factor-treated and control cells. The duration of incubation with growth factors did not consistently modify the drug response, nor did the glucose concentration. Only one set of data demonstrated significant differences; IGF-1-treated SW620 cells in high glucose were more resistant to irinotecan treatment compared with the controls (P=0.009). Treatment with irinotecan in the presence of insulin under the same conditions showed a similar trend, although a significant difference was not observed (P=0.096).

To compare entire response curves, as opposed to single data points (IC₅₀), a multivariable regression model was developed (Table III). As predicted, chemotherapy drug concentration exhibited a significant effect on cell viability in all cases (P<0.001). In the majority of cases, the presence or the duration of pretreatment with growth factors, or the glucose concentration of the media did not significantly change the chemoresponse.

Of the results that showed statistically significant changes, the addition of IGF-1 to tumour cell lines increased the resistance to chemotherapy: WiDr 5-FU in normal glucose at 24 h (P<0.001) and 4 h (P<0.001); WiDr oxaliplatin in normal glucose at 24 h (P<0.001) and 4 h (P<0.001); SW620 5-FU in normal glucose at 0 h (P=0.007); and SW620 irinotecan in high glucose at 4 h (P=0.015).

The addition of insulin to WiDr significantly increased sensitivity to chemotherapy: 5-FU in high glucose at 24 h (P<0.001) and in normal glucose at 0 h (P=0.006); and oxaliplatin in high glucose at 24 h (P=0.021) and in normal glucose at 0 h (P=0.015). In addition, insulin induced variable effects in the SW620 cells, such as increased sensitivity; 5-FU in high glucose at 24 h (P=0.004) and 4 h (P=0.042), and increased resistance; oxaliplatin in high glucose at 0 h (P=0.034) and irinotecan in high glucose at 4 h (P=0.011) and 0 h (P=0.016).

The impact of growth factors in the HMEC-1 endothelial cell line on the chemoresponse was variable; IGF-1 in high glucose marginally increased resistance (oxaliplatin at 24 h, P=0.043), but also sensitivity (5-FU at 4 h, P=0.008; oxaliplatin at 0 h, P=0.011; and irinotecan at 0 h, P=0.008). In normal glucose IGF-1 increased 5-FU sensitivity (0 h, P=0.021), but marginally decreased sensitivity to oxaliplatin (24 h, P=0.038). Insulin increased sensitivity (5-FU in high glucose at 24 h, P=0.01; and in normal glucose at 4 h, P=0.016), but also resistance slightly (oxaliplatin in normal glucose at 0 h, P=0.032).

WiDr cells showed significantly increased viability when treated with ULDs (defined as 1/1,000 of IC₅₀) of chemotherapy in normal glucose conditions with IGF-1, ranging between 182% (oxaliplatin at 4 h, P=0.003) and 240% (5-FU at 4 h, P=0.018), compared with WiDr in normal glucose without growth factors or chemotherapy (viability, 100%) (Table IV). Similar trends were observed at 24 h; WiDr viability in normal glucose with IGF-1 increased to 195% with ULDs of oxaliplatin (P=0.082) and to 283% with ULDs of 5-FU (P=0.088). The viability of cells at ULDs was calculated from the equations obtained by model fitting, and the values were used to compare the effect of growth factors on the cellular response to chemotherapy. No significant differences were identified in ULD response between growth factor-treated and control cells under high glucose conditions or insulin, and this effect was not observed in the SW620 or HMEC-1 cells.

Western blot analysis of the nuclear protein fractions of SW620, WiDr and HMEC-1 showed extremely low or undetectable basal levels of HIF-1α protein (Fig. 1B). As predicted, a marked increase in HIF-1α protein was observed in all cell lines in response to CoCl₂, an agent used as a positive control as it interferes with HIF-1 degradation (31). An increase in HIF-1α protein levels in response to IGF-1 and insulin treatment was observed in the SW620 cells, with a weaker increase due to IGF-1 and no increase due to insulin in the WiDr cells. No visible changes from the basal HIF-1α protein levels in response to IGF-1 or insulin were observed in HMEC-1.

The effect of IGF-1 and insulin on total HIF-1α protein levels was further quantified using a cell-based immunoassay, with basal levels defined as 100% (Fig. 1C). An increase in HIF-1α protein levels was observed in all 3 cell lines in response to CoCl₂ [SW620, 233% (P=0.031); WiDr, 126%; and HMEC-1, 136%]. HIF-1α protein levels appeared to be increased in response to IGF-1 in SW620 (132%; P=0.057) and to insulin and IGF-1 in WiDr (insulin, 119%; and IGF-1, 119%) and HMEC-1 (insulin, 121%; and IGF-1, 121%) cell lines, but the increases were not statistically significant.