Aberrant regulation of argininosuccinate synthetase by TNF-α in human epithelial ovarian cancer

Authors

  • Peter W. Szlosarek,

    Corresponding author
    1. Cancer Research UK, Translational Oncology Laboratory, Barts and The London, Queen Mary's School of Medicine and Dentistry, London, United Kingdom
    2. Department of Medical Oncology, St. Bartholomew's Hospital, West Smithfield, London, United Kingdom
    • Cancer Research UK, Translational Oncology Laboratory, Barts and The London, Queen Mary's School of Medicine and Dentistry, John Vane Science Centre, Charterhouse Square, London EC1M 6BQ, UK
    Search for more papers by this author
    • Fax: +44-20-7882-6110.

  • Matthew J. Grimshaw,

    1. Breast Cancer Biology Group, Cancer Research UK, King's College London, School of Medicine, Guy's Hospital, London, United Kingdom
    Search for more papers by this author
  • George D. Wilbanks,

    1. Department of Obstetrics and Gynaecology, University of South Florida, H. Lee Moffitt Cancer Center & Research Institute, Tampa, Florida
    2. Department of Pathology, University of South Florida, H. Lee Moffitt Cancer Center & Research Institute, Tampa, Florida
    Search for more papers by this author
  • Thorsten Hagemann,

    1. Cancer Research UK, Translational Oncology Laboratory, Barts and The London, Queen Mary's School of Medicine and Dentistry, London, United Kingdom
    Search for more papers by this author
  • Julia L. Wilson,

    1. Cancer Research UK, Translational Oncology Laboratory, Barts and The London, Queen Mary's School of Medicine and Dentistry, London, United Kingdom
    Search for more papers by this author
  • Frances Burke,

    1. Cancer Research UK, Translational Oncology Laboratory, Barts and The London, Queen Mary's School of Medicine and Dentistry, London, United Kingdom
    Search for more papers by this author
  • Gordon Stamp,

    1. Experimental Pathology Laboratory, Cancer Research UK, London, United Kingdom
    2. Department of Histopathology, Imperial College, Faculty of Medicine, Hammersmith Campus, London, United Kingdom
    Search for more papers by this author
  • Frances R. Balkwill

    1. Cancer Research UK, Translational Oncology Laboratory, Barts and The London, Queen Mary's School of Medicine and Dentistry, London, United Kingdom
    Search for more papers by this author

Abstract

The pro-inflammatory cytokine, tumour necrosis factor-α, TNF-α, is dysregulated in malignant compared with normal ovarian surface epithelium (OSE). Several epidemiological studies have associated inflammation with ovarian tumorigenesis, with TNF-α playing a key role in modulating invasion, angiogenesis and metastasis. Here, we show that TNF-α also induces expression of arate-limiting enzyme in arginine synthesis, argininosuccinate synthetase (AS), thereby linking inflammation with several arginine-dependent metabolic pathways, implicated in accelerated carcinogenesis and tumour progression. Having identified AS mRNA induction in TNF-α-treated IGROV-1 ovarian cancer cells, using RNA-arbitrarily primed-PCR, we then observed differential regulation of AS mRNA and protein in malignant, compared with normal, OSE cells. A cDNA cancer profiling array with matched normal ovarian and ovarian tumour samples revealed increased expression of AS mRNA in the latter. Moreover, AS protein co-localised with TNF-α in ovarian cancer cells, with significantly higher levels of AS in malignant compared with normal ovarian tissue. Increased co-expression of AS and TNF-α mRNA was also observed in 2 other epithelial tumours, non-small cell lung and stomach cancer, compared with normal corresponding tissues. In summary, high levels of AS expression, which may be required for several arginine-dependent processes in cancer, including the production of nitric oxide, proline, pyrimidines and polyamines, is regulated by TNF-α and may provide an important molecular pathway linking inflammation and metabolism to ovarian tumorigenesis. © 2007 Wiley-Liss, Inc.

The molecular links between inflammation and tumorigenesis are currently under increasing scrutiny. This is driven by centuries-old evidence that tumours often arise at sites of chronic inflammation,1 while anti-inflammatory drugs, such as aspirin and celecoxib, reduce tumorigenesis in both animal models and defined human populations.2 A variety of pro-inflammatory molecules have been implicated in tumour development, including cytokines, adhesion molecules, free radicals and enzymes.3 Therapeutic strategies based on inhibiting these pathways are being tested in the clinic with some evidence for modulation of tumour biology, resulting in improvement in patient outcome.4

We have shown previously that expression and control of the pro-inflammatory cytokine TNF-α is dysregulated in malignant ovarian epithelial cells. Malignant, but not normal, ovarian surface epithelial (OSE) cells displayed increased TNF-α mRNA stability, production of autocrine TNF-α and regulation of IL-6 by endogenous TNF-α.5 Here, the aim was to analyse further the pro-inflammatory activity of TNF-α in malignant ovarian epithelial cells, using the differential display technique of RNA arbitrarily primed–PCR (RAP-PCR). In addition to the known TNF-α-inducible genes NF-κB and manganese superoxide dismutase in ovarian cancer,6 we identified regulation of the enzyme argininosuccinate synthetase (AS) by TNF-α. We provide evidence for differential regulation of AS by TNF-α in malignant, compared with normal, OSE cells. Moreover, AS co-localised with TNF-α in primary ovarian cancer, with significantly higher levels of AS mRNA and protein in malignant compared with normal ovarian tissue. TNF-α induced AS expression may be important for several arginine-dependent processes in ovarian cancer, such as nitric oxide, proline (a constituent of collagen) and polyamine production.7–9 The increased expression of AS protein provides a potential novel diagnostic and therapeutic marker in ovarian cancer, with our work highlighting the need for wider studies of AS in epithelial oncogenesis.

Material and methods

Cell lines and TNF-α stimulation

The following ovarian carcinoma cell lines were utilised: IGROV-110; OVCAR-311; PEO112; and SKOV-3 from the American Tissue Culture Collection (Rockville, MD). Cells were grown in a humidified atmosphere at 37°C and 5% CO2, in either endotoxin-free RPMI medium (IGROV-1, OVCAR-3 and PEO1 cell lines) or endotoxin-free DMEM medium (SKOV-3 cell line) supplemented with 10% fetal bovine serum (Sigma, Poole, UK). Insulin (2.5 μg/ml, Sigma) was added to the medium of OVCAR-3 and PEO1 cell lines. TNF-α (20 ng/ml) was added to 70% confluent cells, and analysis was performed at 0, 0.5, 1, 3, 6 and 24 hr after addition of the cytokine. The TNF-α concentration was selected from dose–response studies for TNF-α mRNA induction (data not shown).

Primary cell isolation, culture and TNF-α stimulation

The study was approved by the East London and the City Health Authority Research Ethics Committee; informed consent was obtained from patients attending the gynaecological oncology unit at St. Bartholomew's Hospital, London. Normal OSE cells, obtained by scraping the ovarian surface at the time of laparotomy, were cultured in a humidified atmosphere at 37°C and 5% CO2, in endotoxin-free MCDB105/M199 medium (Sigma) supplemented with 15% fetal bovine serum.13 Primary ovarian cancer cells from ascites were obtained from patients undergoing surgery for ovarian cancer and cultured in RPMI medium (as mentioned earlier). Malignant and normal OSE cells were identified on the basis of microscopic appearance, positive staining for cytokeratin and negative staining for Factor VIII. TNF-α was added to 70% confluent cells, and analysis was performed at regular intervals of 0, 1, 3, 6 and 24 hr after addition of the cytokine.

Quantitative real-time RT-PCR

DNase-treated RNA was reverse-transcribed with M-MLV reverse transcriptase (Promega, Southampton, UK) according to manufacturer's instructions. Multiplex real-time RT-PCR analyses were performed using AS and NF-κB1 (FAM) primers, and probes were designed using Primer Express v1.5a software: AS forward (CAAGCGCCTCCAGGTCTCTA); AS reverse (GGACCCCTTTTTTGAACTCGAT); and AS probe (AGACCCAGGACCCAGCCAAAGCC). NF-κB1 forward (GGCTACACCGAAGCAATTGAAG); NF-κB1 reverse (CAGCGAGTGGGCCTGAGA); and NF-κB1 probe (CAGGCAGCCTCCAGCCCAGTGA). AS and 18S rRNA (VIC) primers and probes were used with the ABI PRISM 7700 Sequence Detection System instrument and software (PE Applied Biosystems, Warrington, UK). Two microlitres of cDNA were used per 25 μl reaction, with cycling conditions as follows: 50°C for 2 min, 95°C for 10 min, and then 40 cycles of 15 sec at 95°C and 1 min at 60°C. Each sample was analysed in duplicate or triplicate and normalised (ΔCt) to 18S by removing the cycle threshold (Ct) value of 18S from the Ct value of the gene under investigation. The ΔCt for the control was subtracted from the ΔCt for TNF-α stimulation (i.e. ΔΔCt) and the fold difference was calculated by 2math image.

RNA arbitrarily primed-PCR

RAP-PCR and the subsequent analysis of differentially expressed bands was conducted according to the protocol described by Grimshaw and Mason.14

Northern blot analysis

[α-32P]dCTP-Labelled TNF-α, AS, manganese superoxide dismutase (MnSOD) and β-actin cDNA probes were obtained from Dr S.C. Robinson (TNF-α and β-actin) and the Human Genome Mapping Project Resource Centre (I.M.A.G.E. clones for AS and MnSOD), and were prepared and used for Northern analysis as described previously.15

Cancer profiling array

The Cancer profiling array II (Clontech, Cowley, Oxford, UK) comprises cDNA, from tumour and corresponding normal tissues from individual patients spotted on a nylon membrane. The array was hybridised with a specific radiolabelled TNF-α or AS probe, according to manufacturer's instructions. Hybridisation signals were detected by phosphorimaging and analysed using ImageQuant Version 1.11. The membrane was stripped and reprobed with a ubiquitin probe to normalise each signal.

Western blotting

Whole cell extracts were made from unstimulated cells or TNF-α stimulated cells. Cell extract (10 μg) was run on an SDS 12% acrylamide gel and transferred to a nylon membrane. The membrane was blocked overnight (4°C in PBS with 0.1% Tween (PBST) and 10% milk powder) and probed using an anti-AS antibody at 1:2,500 (BD Biosciences, Oxford, UK), in PBST and 10% milk powder at room temperature for 1 hr. After washing with PBST, the membrane was incubated in PBST and 10% milk powder with an HRP-conjugated secondary antibody (1:5,000 dilution; room temperature for 1 hr). The secondary antibody was detected using the Western Lightning Chemiluminescence kit (Perkin Elmer Life Sciences, Beaconsfield, UK).

Immunohistochemistry

Immunohistochemistry was performed on 34 normal and 34 ovarian tumour (papillary surface carcinoma, endometrioid carcinoma, adenocarcinoma and undifferentiated carcinoma) paraffin-embedded sections. Sections were first deparaffinized in xylene, rehydrated with graded ethanol–water mixtures, and then washed with distilled water. Blocking was achieved using normal anti-rabbit serum at 1:25 (Dako, Ely, Cambs, UK) followed by incubation with the monoclonal AS antibody at 1:100 (BD Biosciences). Both a negative (no primary antibody) and a positive control (liver and kidney) were employed. Following incubation with a secondary biotinylated anti-mouse antibody (Dako E354) and streptavidin-peroxidase (Dako P397), the sections were developed using diaminobenzidine (Sigma). The sections were washed with water, counterstained in haematoxylin and assessed by 2 of the authors (GS and PS), according to the intensity (0 = no staining, 1 = weak expression, 2 = moderate expression, and 3 = strong expression) and extent of staining (0 = no staining at all in the tumour, 1 = less than 25% of the cells, 2 = 25–50%, and 3 = more than 50%). Immunohistochemistry for TNF-α was performed as mentioned earlier using the antibody, MAB610 (R&D Systems).

Data analysis and statistics

NIH Image v.1.61 software was used to analyse Western blots by densitometry, and ImageQuant software (Molecular dynamics) v1.11 was used to analyze phosphorimaging data (Storm scanner) obtained from the cancer profiling array. InStat v2.01 software was used to test results for statistical significance (Student's t test and Bonferroni test).

Results

TNF-α induces AS in a subset of ovarian cancer cell lines

We employed RAP-PCR to investigate the pro-inflammatory activity of TNF-α in IGROV-1, an ovarian cancer cell line characterised by increased TNF-α mRNA stability and the production of autocrine TNF-α.5 Here, we detected 3 TNF-α-inducible genes involved in inflammation, namely MnSOD, NF-κB1 and AS (Fig. 1a). We validated MnSOD and NF-κB1 induction using Northern blotting and quantitative real-time RT-PCR (data not shown), confirming a previous report in the literature.6 AS mRNA increased 2- to 3-fold by 24 hr in both the IGROV-1 cell line, and a second ovarian cancer cell line OVCAR-3, following treatment with TNF-α (Fig. 1b). This correlated with a corresponding increase in AS protein levels as assessed by Western blotting (Fig. 1c). In contrast, although the ovarian cancer cell lines, PEO1 and SKOV-3, expressed AS in abundance, this was not TNF-α-inducible (data not shown). In summary, our data confirm that AS, a rate-limiting enzyme in arginine synthesis, is TNF-α-inducible in 2 of the 4 ovarian cancer cell lines known to secrete autocrine TNF-α.5

Figure 1.

TNF-α induces AS expression in ovarian cancer cell lines. (a) RNA was extracted from the TNF-α-treated ovarian cancer cell line, IGROV-1, at the indicated timepoints and subjected to RAP-PCR. Three TNF-α-inducible genes with roles in inflammation were detected using autoradiography: NF-κB1, MnSOD and AS. (b) AS mRNA induction was confirmed in a subset of ovarian cancer cell lines with known capacity for autocrine TNF-α production (n = 6, p < 0.05). (c) A 2- to 3-fold increase in AS protein was confirmed by 24 hr in the IGROV-1 (n = 6) and OVCAR-3 (n = 2) cell lines exposed to 20 ng/ml of TNF-α.

AS is differentially regulated in malignant compared with normal OSE cells

Next, we compared the effect of TNF-α stimulation on the expression of AS using cultures of malignant and normal OSE cells. Following TNF-α treatment, AS mRNA increased 2-fold by 24 hr in ovarian cancer cell lines (IGROV-1 and OVCAR-3) and primary malignant, but not normal, OSE cells (Fig. 2a). We observed a 1.5-fold induction of AS mRNA by TNF-α in normal OSE cultures by 6 hr, but this declined to unstimulated levels by 24 hr. There was a corresponding increase in AS protein levels by 24 hr in the primary malignant but not the normal OSE cells (Fig. 2b). In summary, TNF-α induced prolonged expression of AS mRNA and protein in ovarian tumour cells, contrasting with short-lived expression of AS mRNA in the normal OSE cells.

Figure 2.

Differential regulation of AS in malignant compared with normal OSE cells. Cultures of ovarian cancer cell lines (IGROV-1 and OVCAR-3), primary ascitic malignant ovarian cells and normal OSE cells were treated for 24 hr with 20 ng/ml of TNF-α. (a) RNA was extracted, reverse-transcribed and analysed for AS mRNA expression using quantitative real-time RT-PCR. AS mRNA increased2-fold by 24 hr in the ovarian cancer cell lines (data shown for OVCAR-3, n = 6, p < 0.05) and malignant OSE cell cultures (a representative experiment of three) treated with TNF-α, whereas the increased AS mRNA that was observed by 6 hr (n = 6, p < 0.05) in normal OSE cells declined to baseline by 24 hr. (b) Western blots showing differential regulation of AS by TNF-α in 2 sets of cultures of primary malignant and normal OSE cells. Densitometry confirmed elevated levels of AS protein in cultures of primary ovarian ascitic tumour cells (Asc1 = 1.5-fold; Asc 2 = 1.5-fold) but not in normal OSE cell cultures (OSE1 = 0.8-fold; OSE2 = 0.7-fold) exposed to TNF-α for 24 hr.

AS mRNA is increased in ovarian tumours compared with normal ovarian tissue

Given the in vitro observations, we proceeded to quantify the levels of AS mRNA and protein in primary malignant and normal ovarian tissue. The level of AS mRNA on the cDNA cancer tissue array was increased significantly in ovarian tumours compared with corresponding normal ovarian tissue (Figs. 3a and 3b). This paralleled the increase in TNF-α mRNA previously described in ovarian cancer using the Clontech array.5 Second, a search of Oncomine™ (www.oncomine.org) of available AS microarray data provided further evidence that this enzyme is up-regulated highly in ovarian cancer compared with both normal ovarian tissue and other tumour types (data not shown).

Figure 3.

AS mRNA expression in ovarian tumours compared with corresponding normal ovarian tissue. The Clontech cancer profiling array II was hybridised using a radiolabelled AS cDNA probe. The array was then stripped and probed using a radiolabelled ubiquitin cDNA probe to normalise the signal. Increased expression of AS mRNA was observed in ovarian tumours compared with normal ovarian tissue (p < 0.01).

AS protein is increased in primary epithelial ovarian cancer compared with normal OSE

We assessed AS protein expression by immunohistochemistry, using 34 paraffin blocks each of malignant and normal ovarian tissue. AS protein was detected in 24/34 (71%) paraffin-embedded tumour sections compared with 5/34 (15%) normal ovarian sections. AS protein was weakly expressed by less than 5% of the normal OSE, whereas it was seen in 5–95% of tumour cells varying from weak to strong expression (Figs. 4a4f). AS immunostaining was confined to the apical membrane of normal OSE cells, and was found in association with reactive epithelium and surface invaginations. In contrast, tumour cells expressed AS protein in various cell compartments, including the plasma membrane, cytoplasm and nucleus. Of note, we observed co-localisation of TNF-α protein in ovarian tumour cells expressing AS protein using serial sections (Figs. 5a and 5b).

Figure 4.

AS protein expression in epithelial ovarian cancer compared with the normal ovary. Thirty-four normal and 34 malignant ovarian paraffin sections were analysed using immunohistochemistry for AS protein. AS expression was increased markedly in malignant compared with normal OSE (p < 0.0001): (a) weak focal AS positivity of the surface epithelium of a normal ovary, (b) Grade 1 papillary serous ovarian carcinoma, (c) Grade III papillary serous ovarian carcinoma, (d) endometrioid ovarian carcinoma, (e) ovarian adenocarcinoma, (f) undifferentiated ovarian carcinoma.

Figure 5.

Co-localisation of TNF-α and AS expression in ovarian cancer. Immunohistochemistry confirmed positive TNF-α and AS expression in ovarian cancer. Serial sections of serous papillary adenocarcinoma showing positive expression for (a) TNF-α and (b) AS.

Differential AS expression in human epithelial and non-epithelial cancer

Finally, the cancer profiling array revealed evidence for complex regulation of AS mRNA expression according to tumour type (Fig. 6). Thus, as with ovarian cancer, there was increased expression of AS mRNA in non-small cell lung and stomach cancer when compared with matched normal tissues. Several groups, including ours, have documented increased expression of TNF-α mRNA in non-small cell lung and stomach cancer, indicating that this cytokine may be an important regulator of AS in vivo.5, 16 In contrast to the increased AS expression in epithelial cancer, we demonstrated reduced AS mRNA expression in melanoma and renal cancer compared with normal tissue, as described previously.17, 18 In summary, we have identified differential expression of AS mRNA in several cancers, with some epithelial tumours showing increased AS expression—that appears to be dependent, in part, on local production of the cytokine TNF-α—whilst other tumours, particularly those of non-epithelial origin, are characterised by a reduction in AS expression compared with the corresponding normal tissues.

Figure 6.

Histogram showing differential expression of AS in epithelial and non-epithelial cancer. The Clontech cancer profiling array II (Fig. 3) revealed differential expression of AS in several tumour types compared with matched normal tissue. There was an apparent increase of AS mRNA in 2 other epithelial tumour types, namely non-small cell lung cancer and stomach cancer. In contrast, and consistent with published data, there was a statistically significant reduction in AS mRNA in renal and skin (melanoma) tumours, suggesting complex roles of AS in tumour biology.

Discussion

Chronic inflammation is a recognised feature of many adult epithelial cancers, involving a variety of cellular and molecular mediators that play a role in tumour initiation, promotion and progression. In this study, we focused on TNF-α, a pro-inflammatory cytokine, that is dysregulated in ovarian cancer, which has the ability to influence all 6 ‘hallmarks’ of malignancy.19 We showed that TNF-α regulates AS, an enzyme that is up-regulated in a majority of ovarian tumours compared with normal OSE. AS is located at the cross-roads of inflammation and metabolism and, being considered a rate-limiting enzyme in arginine production, may play an important role in a variety of pathways including generation of nitric oxide, stroma (collagen) and polyamines.20

TNF-α induced AS in a subset of ovarian cancer cell lines characterised by autocrine TNF-α signaling. Although cultured normal OSE cells expressed AS at comparable levels to the primary malignant cells and ovarian cancer cell lines in vitro, the induction of AS mRNA and protein seen by 24 hr of stimulation with TNF-α was confined to the malignant cells. This has parallels with another inflammatory enzyme, COX-2, and its regulation by TNF-α in malignant and normal prostatic epithelium.21 The basis for the differential expression of inflammatory enzymes in response to TNF-α in malignant compared with normal OSE remains unclear, although NF-κB may be involved as suggested by the RAP-PCR gel. NF-κB, and another transcription factor, Sp1, have both been shown to mediate up-regulation of AS in response to IL-1β and glutamine, respectively, indicating that further validation of these pathways is warranted in ovarian (and other) tumours overexpressing TNF-α and AS.22, 23

The high levels of tumoral AS mRNA and protein expression, co-localising with TNF-α, suggest that inflammatory cytokines are involved in the regulation of this enzyme in vivo. Recent work identified both IL-1β and TNF-α as distinct regulators of AS in the Caco-2 epithelial cell line22; previously, studies employed combinations of inflammatory cytokines to show that AS is inducible maximally (2- to 3-fold) by 24 hr in malignant cell lines.24, 25 Presently, it remains unclear as to the role of AS in ovarian cancer, although several of the ‘downstream’ products have been implicated in ovarian tumorigenesis. Increased AS expression in ovarian cancer may reflect an ongoing requirement for arginine by tumour cells to promote NO-mediated inflammation (i.e. tumour genomic instability and angiogenesis), with evidence for increased iNOS levels in ovarian cancer being associated with a poor prognosis.26–28 Alternatively, ovarian tumoral AS may channel arginine into metabolic pathways that promote tumour growth via the synthesis of polyamines, pyrimidines and proline, which is involved in the synthesis of collagen and the extracellular matrix. The polyamines putrescine, spermidine and spermine are increasingly being linked to various aspects of tumorigenesis, including the transformation of cells, tumour growth, invasion and metastasis.7, 9 Of note may be that urinary polyamine levels have been linked to chemosensitivity in patients with ovarian cancer,29, 30 with high levels predictive of a good outcome. The term ‘arginine switch’ has recently been coined emphasising that a balance exists between the processes of inflammation and repair, which are dysregulated in cancer.20 Both inflammatory cytokines and metabolites, such as NO and agmatine, regulate the arginine switching mechanism. For example, NO has been shown to negatively regulate AS activity in intestinal epithelial cells following IL-1β treatment, possibly via post-translational modification in the form of S-nitrosylation at the Cys-132 residue.31, 32 Intracellular arginine levels may also fluctuate in response to arginase and the arginine transporter, both of which are TNF-α-inducible.33, 34 Further work is therefore required to assess not only AS activity following TNF-α stimulation, but also subsequent arginine utilization in normal and malignant ovarian epithelial cells.

In addition to our analysis of dysregulated expression of AS in ovarian cancer, the cancer profiling array highlights several other areas of interest. AS mRNA expression also was increased in non-small cell lung and stomach tumours compared with matched normal samples, suggesting that this enzyme may have a role in other types of epithelial neoplasia. In contrast, AS mRNA expression was down-regulated in 2 non-epithelial tumour types compared with corresponding normal tissues, namely melanoma and renal cancer. This supports recent published data on loss of AS in these and other tumours known for their aggressive clinical behaviour, reinforcing that arginine utilization in cancer is complex and poorly understood.17, 18 Importantly, manipulation of plasma arginine may be of therapeutic value in tumours with loss of AS expression, and currently is under clinical investigation.35–38 Thus, tumoral AS overexpression as well as repression, the latter due to epigenetic silencing of the AS promoter,38 appear to impact on arginine handling by the tumour, with resultant phenotypic and therapeutic differences.

In conclusion, we have identified that AS, a key enzyme in arginine biosynthesis, is TNF-α-inducible and dysregulated in ovarian cancer. Further work addressing the role of AS overexpression in this and other cancers may lead to novel therapeutic approaches that target the link between tumour inflammation and metabolism.

Acknowledgements

We are very grateful to Prof. Ian Hart (Tumour Biology, Cancer Research UK) and Dr. Tim Crook for helpful discussion of our manuscript and Mr. George Elia (Histopathology, The London Research Institute) for technical assistance with the immunohistochemistry. Dr. PW Szlosarek is a recipient of a Cancer Research UK Clinical Research Training Fellowship.

Ancillary