Targeting GRP78 to enhance melanoma cell death


  • Shaun Martin,

    1.  Northern Institute of Cancer Research and Newcastle Cancer Centre, Newcastle University, Newcastle upon Tyne, UK
    2.  Dermatological Sciences, Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, UK
    Search for more papers by this author
  • David S. Hill,

    1.  Dermatological Sciences, Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, UK
    Search for more papers by this author
  • James C. Paton,

    1.  Research Centre for Infectious Diseases, School of Molecular and Biomedical Science, University of Adelaide, Adelaide, SA, Australia
    Search for more papers by this author
  • Adrienne W. Paton,

    1.  Research Centre for Infectious Diseases, School of Molecular and Biomedical Science, University of Adelaide, Adelaide, SA, Australia
    Search for more papers by this author
  • Mark A. Birch-Machin,

    1.  Dermatological Sciences, Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, UK
    Search for more papers by this author
  • Penny E. Lovat,

    1.  Dermatological Sciences, Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, UK
    Search for more papers by this author
  • Chris P.F. Redfern

    1.  Northern Institute of Cancer Research and Newcastle Cancer Centre, Newcastle University, Newcastle upon Tyne, UK
    Search for more papers by this author

Chris Redfern, email:


Targeting endoplasmic reticulum stress-induced apoptosis may offer an alternative therapeutic strategy for metastatic melanoma. Fenretinide and bortezomib induce apoptosis of melanoma cells but their efficacy may be hindered by the unfolded protein response, which promotes survival by ameliorating endoplasmic reticulum stress. The aim of this study was to test the hypothesis that inhibition of GRP78, a vital unfolded protein response mediator, increases cell death in combination with endoplasmic reticulum stress-inducing agents. Down-regulation of GRP78 by small-interfering RNA increased fenretinide- or bortezomib-induced apoptosis. Treatment of cells with a GRP78-specific subtilase toxin produced a synergistic enhancement with fenretinide or bortezomib. These data suggest that combining endoplasmic reticulum stress-inducing agents with strategies to down-regulate GRP78, or other components of the unfolded protein response, may represent a novel therapeutic approach for metastatic melanoma.


The activation of cell death pathways in response to cellular damage is a central paradigm of cancer treatment and relies on drug-induced damage exceeding cellular repair capacity, allowing pro-apoptotic signals to predominate (Bitomsky and Hofmann, 2009). Drugs which induce DNA damage are often the mainstay of cancer treatment but target proliferating normal and tumour cells indiscriminately. For malignant melanoma, the low 5-year survival rates for patients with metastatic disease may be due to an intrinsic resistance of these tumour cells to programmed cell death (Soengas and Lowe, 2003) and emphasizes the need for new therapeutic strategies. Chemotherapeutic drugs which target cytoplasmic rather than nuclear function are now available. In particular, the endoplasmic reticulum (ER), the site of synthesis of membrane and secretory proteins, is a target of fenretinide [N-(4-hydroxyphenyl) retinamide, a synthetic derivative of retinoic acid] and bortezomib (a 26S proteasome inhibitor), which induce melanoma cell death as a result of ER stress both in vitro and in vivo (Corazzari et al., 2007; Hill et al., 2009; Lovat et al., 2008). Some tumour types may be more susceptible than normal cells to these ER stress-inducing drugs, particularly bortezomib (Fernandez et al., 2005; Lovat et al., 2008; McCloskey et al., 2008; O’Donnell et al., 2002; Pigneux et al., 2007). As with nuclear damage, ER stress results in the induction of homeostatic repair pathways and in the induction of programmed cell death. Although melanoma cells appear to be adapted to withstand high levels of ER stress (Jiang et al., 2009), targeting the adaptation mechanisms may provide a new route to developing novel therapeutic strategies (Hersey and Zhang, 2008).

ER stress can result from the accumulation of unfolded or misfolded proteins in the lumen of the ER, which is counteracted by activation of the unfolded protein response (UPR) (Szegezdi et al., 2006). The primary sensor of ER stress is thought to be Glucose-regulated protein 78 (GRP78, also known as BiP), which interacts with PERK, ATF6 and IRE1 in the ER. Evidence suggests that the binding of GRP78 to exposed hydrophobic regions of unfolded proteins results in the active release of PERK, ATF6 and IRE1 (Bertolotti et al., 2000; Schindler and Schekman, 2009; Shen et al., 2005), which then activate signalling cascades resulting in the global attenuation of protein synthesis, the induction of UPR genes to restore ER function (Rao et al., 2004) and the induction of pro-apoptotic signalling pathways, which are activated under conditions of prolonged ER stress (Rutkowski et al., 2006). The levels of GRP78 also increase in response to ER stress (Corazzari et al., 2007; Lee, 2005) and this may be, at least in part, a negative-feedback mechanism to reset the stress sensor. GRP78 also has a key role as an ER chaperone (Paton et al., 2006), can promote cell survival by preventing caspase-7 activation (Reddy et al., 2003) and also reduces cell death by buffering oxyradical accumulation and stabilizing mitochondrial function (Liu et al., 1998; Yu et al., 1999).

The expression of GRP78 is increased in relation to malignant progression and drug resistance in melanoma and other cancers (Daneshmand et al., 2007; Dong et al., 2005; Kim et al., 2006; Lee et al., 2006; Zhuang et al., 2009) and the inhibition or down-regulation of GRP78 can increase ER stress-induced death (Rao et al., 2002; Reddy et al., 2003), inhibit tumorigenesis of prostate cancer cells (Fu et al., 2008) and prevent the development of resistance to cytotoxic T cells (Sugawara et al., 1993). Therefore, the abrogation of GRP78 function coupled with drugs to increase ER stress may represent an effective strategy for the treatment of melanoma. The aim of this study was to test this hypothesis using the ER stress-inducing agents fenretinide and bortezomib. To abrogate GRP78 we have compared small-interfering RNA (siRNA)-mediated knockdown of GRP78 with a novel approach based on the intracellular cleavage of GRP78 by the AB5 subtilase cytotoxin (SubAB) (Paton et al., 2006). The SubAB cytotoxin is a serine protease which cleaves GRP78 specifically as a result of an unusually deep cleft at the active site within the A subunit (Paton et al., 2006).

Results and Discussion

siRNA-mediated knockdown of GRP78 increases apoptosis in response to fenretinide or bortezomib

Clinically-achievable concentrations of either fenretinide or bortezomib induce ER stress and apoptosis of melanoma cells (Hill et al., 2009). Therefore, we tested the hypothesis that RNA interference (RNAi)-mediated knockdown of GRP78 would increase apoptosis of melanoma cells in response to fenretinide or bortezomib. After transfection of GRP78 siRNA into CHL-1 and WM266-4 cells, GRP78 protein was reduced by 56% (range 56.8–55.5, duplicate experiments) and 59% (range 58.5–66.6), respectively, compared to cells transfected with a control scrambled siRNA (Figure 1). These experiments also confirmed the induction of GRP78 (Corazzari et al., 2007) in response to fenretinide or bortezomib, with increases in GRP78 levels, respectively, of 38 and 9% in WM266-4 cells and 18 and 29% in CHL-1 cells treated with control siRNA (Figure 1). The induction of GRP78 in response to fenretinide or bortezomib was abolished with GRP78 siRNA and GRP78 protein levels in CHL-1 and WM266-4 cells treated with fenretinide or bortezomib were 50–54% lower than cells treated with control siRNA alone. After knockdown of GRP78 there was a significant two- to three-fold increase in apoptosis in response to fenretinide or bortezomib (Figure 1). Overall, WM266-4 cells were more resistant, with 32% apoptosis in response to fenretinide after GRP78 knockdown compared to 76% apoptosis of CHL-1 cells under similar conditions (Figure 1 legend). In CHL-1 and WM266-4 cells exposed to vehicle alone, knockdown of GRP78 doubled the background level of apoptosis to 19 and 8%, respectively. RNAi-mediated knockdown of GRP78 also decreased the viability of CHL-1 cell cultures in response to fenretinide or bortezomib and WM266-4 cell cultures in response to fenretinide (Figure 1). These data show that knockdown of GRP78 substantially increased apoptosis induced by fenretinide or bortezomib and suggest that GRP78 is a valid target for the development of drug combinations to improve the treatment of melanoma.

Figure 1.

 The consequence of GRP78 knockdown for fenretinide- and bortezomib-induced apoptosis. CHL-1 and WM266-4 cells were transfected with control or GRP78-specific siRNA prior to treatment with 10 μM fenretinide (FenR), 200 nM bortezomib (Bort) or control vehicle (Ctrl) for 24 h. (A,B) Western blot analysis to assess GRP78 induction and knockdown relative to a β-actin loading control. Induction of apoptosis (C,D) or reduction in culture viability (E,F) of CHL-1 and WM266-4 cells, respectively, in response to 10 μM fenretinide, 200 nM bortezomib or control vehicle after transfection with control or GRP78 siRNA. In these graphs, the ordinate scales are apoptosis and viability expressed as a fold-change relative to control vehicle-treated cells after transfection with control siRNA (relative apoptosis and relative viability, respectively). This representation is used to demonstrate the increase in %apoptosis after GRP78 knockdown relative to that in control-vehicle-treated cells. Actual mean %apoptosis values for CHL-1 cells (with 95% confidence interval) for cells treated with control vehicle, fenretinide or bortezomib are, respectively: (Control siRNA) 9.37 (0.07), 25.33 (3.52), 25.97 (0.17); (GRP78 siRNA, 18.70 (0.74), 76.17 (2.29), 72.93 (1.94). For WM266-4 cells the values are: (Control siRNA) 4.33 (0.69), 13.80 (0.41), 11.80 (1.86); (GRP78 siRNA) 7.80 (1.04), 32.27 (5.40), 27.13 (0.57). Each bar is the mean of three (apoptosis) or eight replicates (viability) ±95% confidence intervals. Data were analysed using two-way anova on siRNA type (control or GRP78) and treatment (control, fenretinide, bortezomib), followed by one-way anova to contrast siRNA levels (control and GRP78 siRNA) for control, fenretinide and bortezomib treatments. For all experiments, transfection with GRP78 siRNA produced a significant increase in apoptosis or reduction in culture viability (siRNA main effect, P ≤ 0.009). Probability levels for contrasts (comparisons between control siRNA- or GRP78 siRNA-transfected cells for the drug treatments control vehicle, fenretinide or bortezomib) are indicated by: ***P < 0.001; **P < 0.01; *P < 0.05. In the viability assays, there was no significant effect of GRP78 siRNA versus control siRNA for the control vehicle treatment (P > 0.9), and no significant effect of GRP78 siRNA compared to control siRNA for the bortezomib treatment of WM266-4 cells (P = 0.56).

Subtilase cytotoxin as a specific inhibitor of GRP78

To test the ability of SubAB to reduce cellular GRP78 levels and enhance ER stress-induced cell death, CHL-1 and WM266-4 cells were treated for 24 h with 0.5–1000 ng/ml of SubAB or the proteolytically inactive mutant SubAA272B (Paton et al., 2004; Wolfson et al., 2008). SubAB induced apoptosis and inhibited culture viability with near-maximal effects at 5 ng/ml, whereas the inactive mutant toxin had marginal effects only at maximum doses of 500–1000 ng/ml (Supporting Information Figure S1). For CHL-1 cells, SubAB was more effective at reducing culture viability (MTS assay), which was reduced by approximately 70%, than at inducing apoptosis, which increased to 10–15%. WM266-4 cells were more resistant to SubAB, at least in the viable-cell assays; this, and the lower induction of apoptosis in response to fenretinide or bortezomib (Figure 1 legend), may relate to the presence of the BRAFV600D mutation in the WM266-4 cell line (Davies et al., 2002). In the continuous presence of 5 ng/ml of SubAB, cell death increased with time in both melanoma cell lines, producing similar levels of apoptosis up to 24 h (Figure 2). SubAB-mediated cleavage of GRP78 was detected using a C-terminal-specific GRP78 antibody which detects the 27-kD cleavage product; this showed a time-dependent increase in GRP78 cleavage in both cell lines detectable within 1 h of exposure to SubAB, and a concomitant decrease in full-length GRP78 (Figure 2). The increase in GRP78 cleavage correlated with increased levels of cell death; the inactive SubAA272B mutant toxin had no effect on apoptosis and did not cleave GRP78.

Figure 2.

 Time-course of apoptosis in response to continuous (Time Course) or pulsed exposure (Wash Out) to SubAB toxin. CHL-1 and WM266-4 cells were treated with 5 ng/ml toxin for 1, 2, 4, 8 and 24 h and harvested for flow cytometry (A,B) and to assess GRP78 cleavage by Western blotting (C,D); blots were probed simultaneously for GRP78 and β-actin; the β-actin band is intermediate in size between full length GRP78 and the cleavage product. Densitometric data for full-length GRP78 (E) and the cleavage product (F), relative to the β-actin loading control from three independent experiments, were combined and are shown in panels (E) and (F). Error bars in (E,F) are ±1 SD to enable the curves for each cell line to be presented together. To investigate the length of exposure to toxin required to induce apoptosis after 24 h, cells were treated with 5 ng/ml toxin continuously for 24 h, or for 1, 2, 4 or 8 h, and then washed and cultured in the absence of toxin in the medium for the time remaining until 24 h after initial exposure; at 24 h the cells were harvested for measurement of apoptosis (G,H) or GRP78 cleavage as above (I,J). Flow cytometry data are means ±95% confidence intervals.

The length of exposure to SubAB required to induce apoptosis was assessed by treating cells with wild-type or mutant toxin for 1, 2, 4, 8 and 24 h, and then continuing incubation in the absence of SubAB for the time remaining to 24 h after initial exposure. There was a significant difference in response to SubAB between the two cell lines (two-way anova; effect of cell type F1,20 = 13.8, P = 0.001). Using the 1-h exposure time as a reference (similar levels of cell death to cells exposed to the inactive mutant toxin), for CHL-1 cells, cell death was only significantly increased after 24 h exposure (one-way anova, Dunnett’s test, up to 8 h, P ≥ 0.245; 24 h, P < 0.001). Conversely, for the WM266-4 cells, cell death was significantly elevated after 4 h exposure (one-way anova, Dunnett’s test, 4, 8 and 24 h, P ≤ 0.023). This difference between the cell lines was mirrored by higher levels of cleaved GRP78 in WM266-4 cells after even a short 1-h exposure to SubAB (Figure 2).

These data clearly demonstrate cleavage of GRP78 in both melanoma cell lines as well as increased cell death. Over the time scale used, cleavage of GRP78 was not complete, even at high toxin concentrations; nevertheless, cleavage occurred rapidly and the level of cleavage increased over 24 h. SubAB enters cells via a clathrin-dependent process and is trafficked specifically to the ER (Chong et al., 2008). As evidenced by the activation of classic ER stress markers (Lass et al., 2008; Wolfson et al., 2008), SubAB itself induces ER stress, perhaps as a consequence of the release of the ER-stress signalling activators PERK, IRE1 and ATF6, due to a reduction in levels of intact GRP78 within the ER (Wolfson et al., 2008).

To determine the effect of SubAB on sensitivity to fenretinide- or bortezomib-induced cell death, fixed-dose ratio experiments were carried out using fenretinide with SubAB (ratio of 2000:1, nM fenretinide:ng/ml SubAB) or bortezomib with SubAB (40:1, nM bortezomib:ng/ml SubAB). Under these conditions, SubAB increased apoptosis to a maximum of around 40% in combination with fenretinide or bortezomib (Figure 3). The mutant toxin had no effect on cell death in response to fenretinide or bortezomib (Supporting Information Figure S2). Combination indices (CI) were calculated to test for synergy, where for additivity, CI = 1 (representing the classical isobologram equation) and for synergistic interactions, CI < 1 (Chou, 1991). CI values for 10–20 μM fenretinide with SubAB and for 200–400 nM bortezomib with SubAB were indicative of mild synergy (Figure 3). In three independent experiments on CHL-1 and WM266-4 cells at mid-range doses (dose ratios as above), mean CI values for the induction of apoptosis by fenretinide (10 μM) with SubAB were 0.51 and 0.52, and for bortezomib (200 nM) 0.36 and 0.55, respectively, and were significantly different from 1 (1-sample t-test, P ≤ 0.024). Data for the cell viability assays were also indicative of synergy between SubAB and fenretinide or bortezomib (Supporting Information Figure S2). Western blot analysis confirmed the SubAB-mediated cleavage of GRP78 (Figure 3) but, in addition, treatment with bortezomib in the presence or absence of toxin resulted in the appearance of a lower molecular weight band detected with the COOH-specific GRP78 antibody; this band had an apparent molecular mass of 68 kDa (range 66–70), compared to 79 kDa (range 76–82) for full length GRP78 (Figure 3).

Figure 3.

 Effect of GRP78-specific SubAB on fenretinide- and bortezomib-induced apoptosis. CHL-1 (A–C) and WM266-4 (D–F) cells were treated with fenretinide, bortezomib and SubAB cytotoxin or SubAA272B mutant cytotoxin alone and in combination for 24 h at fixed-dose ratios of 2000:1 (nM fenretinide:ng/ml SubAB) and 40:1 (nM bortezomib:ng/ml SubAB), respectively. Note that fenretinide units are expressed in μM on the figure. Apoptosis data (A,B and D,E for CHL-1 and WM266-4 cells treated with fenretinide or bortezomib, respectively) are the mean of three experiments ±95% confidence intervals. Combination indices (Chou, 1991) ranged from 0.6 to 0.8 for 10–20 μM fenretinide with SubAB for CHL-1 and WM266-4 cells, indicative of mild synergy. Combination indices (range 0.5–0.6) for 200–400 nM bortezomib with SubAB also indicated mild synergy on WM266-4 cells, but rather more marked synergy on CHL-1 cells (combination indices 0.05–0.07). Western blots to confirm GRP78 cleavage were carried out on samples of CHL-1 (C) and WM266-4 (F) cells which had undergone treatment with 10 μM fenretinide, 200 nM bortezomib and 5 ng/ml SubAB cytotoxin or SubAA272B mutant cytotoxin alone and in combination after 24 h. The blots were probed sequentially for GRP78 and β-actin.

These data show that SubAB, as well as inducing cell death in its own right, acts synergistically to increase the death of melanoma cells in response to fenretinide or bortezomib. The inhibition of homeostatic responses to ER stress results in increased cell death (Corazzari et al., 2007; Lovat et al., 2008). As SubAB preferentially cleaves newly synthesized GRP78 (Hu et al., 2009), this may abrogate the pro-survival consequences of GRP78 induction and be the mechanism underlying the synergy between SubAB and fenretinide or bortezomib. The overexpression of GRP78 has been associated with melanoma and glioblastoma (Graner et al., 2009; Zhuang et al., 2009) and, therefore, targeting GRP78 specifically in combination with other ER-stress-inducing agents may be the most appropriate strategy to develop more effective therapies for these notoriously drug-resistant tumours. However, SubAB is lethal when injected intraperitoneally in mice and induces pathological changes characteristic of haemolytic uraemic syndrome (Wang et al., 2007). This, and the discovery that the cleaved GRP78 C-terminal substrate-binding domain blocks antibody secretion by B lymphocytes (Hu et al., 2009), may make the native SubAB toxin unsuitable for clinical use. In a recent important development, it has been shown that the active subunit, SubA, can be targeted to breast and prostate cancer cells as an epidermal-growth factor (EGF) fusion protein (Backer et al., 2009). This raises the prospect of targeting SubA to tumour cells in combination with drugs that are also relatively tumour-specific in their action. Thus, although bortezomib and fenretinide are licensed as clinical drugs for a restricted range of tumour types, combining them both together (Hill et al., 2009) and with drugs to abrogate stress-survival responses will open up new markets for their therapeutic development. With respect to SubAB, in vivo studies with EGFR-targeted SubA in combination with fenretinide or bortezomib will be important to assess the potential of this reagent for melanoma therapy.

Materials and methods

Tissue culture

Human metastatic melanoma cell lines CHL-1 and WM266-4 were cultured as described previously (Corazzari et al., 2007). Cell lines were verified by staining with melan A antibody to confirm tumour type (Flockhart et al., 2009) and BRAF mutational status: CHL-1 and WM266-4 were confirmed, using Custom TaqMan SNP genotyping assays (Applera Europe BV, Warrington, UK), to be BRAF wild-type and V600D, respectively (Davies et al., 2002). CHL-1 and WM266-4 cells were seeded at 0.15 × 106 and 0.25 × 106, respectively, in six-well flat-bottom tissue culture plates (Helena Biosciences, Gateshead, UK), in a final volume of 3 ml medium and allowed to attach overnight. For single-dose treatments, cells were subsequently treated with 10 μM fenretinide (Jansen-Cilag Ltd, Basserdorf, Switzerland) or 200 nM bortezomib (Millennium Pharmaceuticals, Cambridge, MA, USA) alone or in combination with 5 ng/ml SubAB cytotoxin or SubAA272B mutant cytotoxin (Paton et al., 2004) for 24 h at 37°C. For fixed-dose ratio experiments, fenretinide was combined with cytotoxin at a fixed ratio of 2000:1 (fenretinide:cytotoxin), and bortezomib was combined at a fixed ratio of 40:1 with cytotoxin.

Flow cytometry and viable cell assays

After treatment, cells and supernatant were harvested (Corazzari et al., 2007), resuspended in 500 μl of PBS and fixed with an equal volume of cold 4:1 methanol:acetone before storage at 4°C prior to analysis. The magnitude of the sub-G1 peak after flow cytometric analysis of propidium iodide-stained cells was used as a measure of cell death or apoptosis (Lovat et al., 2008). Flow cytometry data were analysed using Windows Multiple Document Interface (WinMDI). The MTS cell viability assay was used to measure changes in culture viability resulting from changes in cell number (proliferation) and cell viability (cell death): melanoma cells were seeded at 5 × 103 cells per well into 96-well (flat-bottom) tissue culture plates (Helena Biosciences) in 100 μl of tissue culture medium and allowed to attach overnight. Cells were treated for 24 h and viability assessed by staining with 20 μl Cell Titer 96 Aqueous One Solution Reagent (Promega, Southampton, UK) and the optical density measured at 490 nm using a SpectraMax 250 plate reader (Molecular Devices, Wokingham, UK) after 4 h incubation at 37°C, 5% CO2 in air.

Western blot

Samples were harvested by trypsinization, washed in PBS and stored as cell pellets at −20°C prior to analysis. Cell pellets were lysed, proteins (20 μg/track) separated by electrophoresis through 4–20% SDS–PAGE gels and blotted onto PVDF membranes (Armstrong et al., 2007). Blots were probed, either sequentially or simultaneously, with primary antibodies to GRP78 (goat polyclonal, diluted 1:1000; Santa Cruz biotechnology, Santa Cruz, CA, USA) and, as a loading control, β-actin (mouse monoclonal, diluted 1:5000; Sigma, Poole, UK). Detection of primary antibodies was done using peroxidase-conjugated anti-mouse antibody (diluted 1:2000; Upstate Biotechnology, Watford, UK ) and visualized using the ECL system (Armstrong et al., 2007).

GRP78 siRNA transfection

To achieve efficient knockdown of GRP78, cells were subjected to a double transfection procedure. Melanoma cells were seeded in six-well plates at a density of 0.3 × 106 per well in 3 ml of culture medium for flow cytometry and Western blot analysis, or 5 × 103 per well in 96-well plates in 100 μl of medium for determination of cell viability, and allowed to attach overnight at 37°C, 5% CO2. Cells were then transfected for 8 h with 40 nM GRP78 siRNA (Hs_HSPA5_6_HP validated siRNA; Qiagen, Crawley, UK) or control scrambled siRNA (Qiagen) using Lipofectamine 2000 (Invitrogen, Paisley, UK) according to the manufacturer’s instructions (Corazzari et al., 2007); transfection was terminated by changing the culture medium and the cells incubated for a further 16 h before repeating the transfection procedure. The cells were then treated with fenretinide or bortezomib to measure their response in terms of apoptosis, culture viability or protein expression.

Statistical analysis

Data were analysed using general linear models (GLM) in spss 15 (SPSS Inc., Chicago, IL, USA). Homogeneity of variances was assessed using Levene’s test; where variance heterogeneity could not be corrected by log transformation, weighted least squares analysis within the GLM procedure was used. Comparisons between treatments were performed using contrasts or Dunnett’s test within one-way anova. Fixed-dose ratio experiments were plotted within sigmaplot 11.0 (SPSS Inc.) and synergy analysed using calcusyn (Biosoft, Cambridge, UK); for independently replicated experiments at particular doses (constant fixed-dose ratio), one-sample t-tests were used to test against a mean of 1, the expectation for additivity.


We are grateful to Cancer Research UK and the British Skin Foundation for grant support for this study, and to Robbie Woodman and Phil Elstob of Cancer Research Technology Ltd. for their invaluable intellectual input.