- Email: [email protected]
Oncofetal fibronectin and oral squamous cell carcinoma
BJOM-172.QXD
11/24/01 9:51 AM
Page 471
BRITISH
British Journal of Oral and Maxillofacial Surgery (2001) 39, 471–477 © 2001 The British Association of Oral and Maxillofacial Surgeons doi: 10.1054/bjom.2001.0702, available online at http://www.idealibrary.com on
Journal of Oral and Maxillofacial Surgery
Oncofetal fibronectin and oral squamous cell carcinoma A. J. Lyons,* A. C. Bateman,† A. Spedding,‡ J. N. Primrose,§ U. Mandel¶ *Consultant Oral and Maxillofacial Surgeon, King’s College Hospital, London, UK; †Consultant Histopathologist, Southampton University Hospitals Trust, Southampton, UK; ‡Consultant Histopathologist, Queen Alexandra Hospital, Portsmouth, UK; §Professor, Academic Surgical Unit, Southampton University Hospitals Trust, Southampton, UK; ¶Associate Professor, Department of Oral Diagnostics, University of Copenhagen, Copenhagen, Denmark SUMMARY. Fibronectin is a cell matrix glycoprotein, which exists as a number of isoforms that are often found within the cell matrix that surrounds tumours. Collectively these tumour-associated isomers of fibronectin have been termed oncofetal fibronectin (OFFN). We looked for expression of OFFN within oral squamous cell carcinomas (SCC) and related its presence to prognosis. The investigation used a monoclonal antibody (MoAb 5C10) to the glycosylated variant of OFFN, and 100 archival specimens of oral SSC. Immunostaining for OFFN was intense in the adjacent stroma of 43 squamous carcinomas, weak in 27 and absent in 30. Cervical metastases were found in 17/27 (63%) specimens that stained intensely, 6/17 (35%) that stained weakly and 3/13 (23%) that did not stain. Of the 21 cases which had extracapsular lymph node spread, 81% were from those that stained intensely, 19% from those that stained weakly and none from those that did not stain for OFFN expression. Also, 21/44 patients (49%) died in group with intense OFFN staining, 6/26 (23%) in the group with weak staining and 3/30 (10%) in the group that did not stain. The presence of OFFN glycoprotein in oral SCC as evaluated by immunostaining with MoAb 5C10 correlates strongly with the presence of metastatic lymph node involvement, particularly extracapsular involvement, and mortality. We therefore suggest that the degree of expression of OFFN in tumours is a valuable prognostic indicator. © 2001 The British Association of Oral and Maxillofacial Surgeons
and ED-B). In view of these attributes, FN that contains ED-A and or ED-B has been named OFFN. Over the past two decades more has become known about the structure and function of fibronectin and OFFN. As the extra segments that constituted OFFN were only defined in the mid-1980s, it is possible that many of the properties ascribed to fibronectin in papers published before this date may not apply to OFFN or conversely may apply only to OFFN. Certainly, OFFN contains all the segments of the originally described fibronectin molecule, and so can be expected to share many of its attributes (Fig. 1). OFFN has also been detected in healing human cornea5 and related lesions, including fibrotic liver cirrhosis, dilated cardiomyopathy, proliferative glomerulonephritis and fibromatosis.6,7 There is some evidence that the regeneration of other tissues such as peripheral nerves is accompanied by a re-expression of ED-B OFFN.8 There is much evidence that the ED-B OFFN isoform is involved in angiogenesis. ED-B OFFN is never seen in mature vessels, but newly formed vessels in normal and tumour tissues are characterized by an abundant
INTRODUCTION Fibronectin was first extracted from human plasma by Morrison et al.1 in 1948 and was named ‘cold insoluble globulin’ as it precipitated in the presence of fibrinogen at low temperature. The molecule is made up of three types of peptide sequence homologues (I, II and III) and distinct domains that have the ability to bind fibrin, heparin, collagen, gelatin, factor XIIIa transglutaminase, Staphylococcus aureus and cells.2 It is present in the extracellular matrix and body fluids, and seems to have important roles in intercellular adhesion and in binding cells to the extracellular matrix.3 In 1985 Matsuura and Hakomori,4 reported the existence of a second type of fibronectin with a molecular weight of 310 000–335 000 Da, in which there was an extra glycosylated side chain, attached to the IIICS region of the molecule. They called this variant oncofetal fibronectin (OFFN), as its expression seemed to be confined to tumour and fetal tissues. Subsequently, further parts of the molecule that accounted for the higher molecular weight of this variant were defined (ED-A 471
BJOM-172.QXD
11/24/01 9:51 AM
472
Page 472
British Journal of Oral and Maxillofacial Surgery
Fig. 1 Oncofetal fibronectin molecule and interactions.
ED-B OFFN matrix.9 So, quantification of ED-B expression in tumour-associated vessels may predict the ability of these tumours to induce angiogenesis, and may be of similar prognostic significance to the evaluation of microvessel density.10 Deposition of OFFN has been observed in about 30 tumour types, including breast tumours meningiomas and pituitary adenomas.11 The intensity of OFFN staining in colorectal carcinoma has been directly correlated with advanced local stage, liver metastases and general prognosis.12 The expression of OFFN in head and neck squamous carcinomas has been evaluated in two published studies. In a Polish study by Jenek et al.,13 a number of extracellular matrix proteins (ECM) were examined using immunohistochemistry in specimens of laryngeal squamous carcinomas from 20 patients. Although the stage of the tumours was not described, all samples were taken from laryngectomy specimens, which may imply that these were large tumours, as in Northern Europe and the USA smaller lesions are usually treated by radiotherapy. OFFN staining was present and variable to some degree in well over half of them. These results are in contrast to those reported by Mandel et al.,11 who took fresh frozen tissue from 38 cases of oral squamous carcinoma. Radiotherapy had previously been given to 25 of the 38 patients, implying that these tumours were likely to be large or recurrent. All 38 cases stained strongly for both ED-B and glyocosylated OFFN with monoclonal antibodies. Their results concurred with those of other studies,7,14–16 all of which have shown that expression of ED-B invariably coexists with glycosylated OFFN. None of 15 premalignant tissue samples
taken stained for OFFN. This uniformity of staining for OFFN in oral SCC, as compared with the variability seen in laryngeal squamous cell carcinoma, might reflect the biological differences between the two types of tumour. For example, whereas the proportion of oral tumours that present with metastases in lymph nodes has been reported to be as high as 60%,17 that for laryngeal tumours has been placed at only 18%.18 In view of this controversy, an immunohistochemical study was done on 100 cases of oral SCC to find out if the degree of OFFN could predict the presence of metastases and outcome, as seemed to be the case in colorectal cancer.12 The expression of OFFN in the paraffin sections of 100 cases of oropharyngeal carcinoma was examined. All patients with this diagnosis were identified at two centres over 9-year period from their histopathological records, and the time from treatment to analysis of the data varied from 2 to 11 years. Only those patients with adequate clinical records and whose outcome was known (survived or died) were included in the study. To find out if the expression of OFFN by oropharyngeal carcinoma could be related to tumour behaviour and outcome, this expression was corelated with the following clinical and pathological variables: lymph node spread and extralymphatic spread; patient’s age, tobacco and alcohol consumption, and mortality; and the size,19 grade and site of the tumour. Both the previous studies in aerodigestive SCC had used fresh frozen material for immunohistochemical analysis, but valid results has been obtained by Infusa et al.12 with trypsin-treated archival material from colorectal tumours, so archival material was used in this study.
BJOM-172.QXD
11/24/01 9:51 AM
Page 473
Oncofetal fibronectin and oral squamous cell carcinoma
MATERIAL AND METHODS Light microscopy of sections stained with haematoxylin and eosin was used to select the most representative section from each of the cases used in the study. For most cases the main resection specimen was examined together with any metastases to lymph nodes that were resected at this operation or subsequent operations. For patients who only had radiotherapy, an incisional biopsy specimen was all that was examined. The formalin-fixed paraffin blocks sections were prepared, usually in batches of 10, by the following method. A microtome was used to cut 4-m sections off the paraffin blocks, which were placed on to 3⬘aminopropyl-triethoxy-silane (APES)-coated glass slides (Vector Laboratories) and dried at 60⬚C for 30 minutes. Endogenous peroxidase within the sections was blocked by the use of 5% hydrogen peroxide solution in alcohol. The sections were washed again with fresh alcohol and then rehydrated with water. Sections were then transferred to boiling citrate buffer at pH 6.0 in a pressure cooker and treated at high pressure for 2 minutes. After the pressure was released the sections were rapidly cooled in tap water, ensuring that they did not dry out at any time. After this, normal goat serum (10%) in phosphate-buffered saline (PBS) containing avidin blocking reagent at a
Fig. 2 Intensity of expression of oncofetal fibronectin and mortality.
473
dilution of one drop per ml was applied to the sections, which were then incubated at room temperature for 60 minutes. Excess serum was shaken off the sections, and primary antibody (MoAb 5C10) with one drop per ml of avidin blocking reagent was added. Negative controls were incubated with PBS, to which one drop per ml of avidin blocking agent had been added, and were then processed in the same way as the other samples. Monoclonal antibody (MoAb 5C10) was diluted in PBS in concentrations varying from neat to 1 in 1000. A dilution of 1 in 10 was found to produce optimal staining in a series of sections of oral squamous cell carcinoma. Sections were again washed three times with PBS as before and then with water, before incubation with biotinylated secondary antibody (Vector Laboratories) for 30 minutes at room temperature. Streptococcal ABC/peroxidase complex (Vector Laboratories) was then applied and incubated with the sections at room temperature for 30 minutes, after which they were rinsed as before with PBS and then with water. Labelling was visualized with diamino benzadine (Vector Laboratories) and counterstained with haematoxylin; the slides were then dehydrated, cleared and mounted. All applications and incubations were done at room temperature. Sections were examined by light microscopy to assess intensity of staining for OFFN, and graded as absent, weak, patchy, diffuse or strong (Fig. 2).
BJOM-172.QXD
11/24/01 9:52 AM
474
Page 474
British Journal of Oral and Maxillofacial Surgery
This was assessed subjectively by at least two histopathologists. Despite the apparent arbitrary nature of this grading, agreement between the two assessors was 98%. The two disputed sections were finally graded after a second look. Expression of OFFN progressed from strong to weak to no expression, and was hence ordered. In nearly all cases the data was judged to be parametric and compared against other categorical data (for example, presence or absence of lymph node metastases). Where OFFN was compared against age groups, the numbers in each group were too small to judge as parametric and the non-parametric Mann–Whitney U test was used. As mortality was a censored observation, being assessed at different times after treatment for each patient and the influence of OFFN was expected to be unaffected by the passage of time, the log rank test was used. All statistical analysis employed an SPSS version 9 statistics programme.
Table 2 Staining and extracapsular spread from lymph nodes Staining
Number of necks dissected
Number with extracapsular spread
Intense Weak None Total
27 17 13 57
16 4 0 20
2 for trend:18.4, P:0.001.
Table 3 Mortality Staining
Number of patients
Deaths (%)
Intense Weak None Total
21/44 6/26 3/30 20/30
49 23 10 …
Log rank analysis: strongly positive versus weakly positive, P:0.06; strongly positive versus negative, P:0.02; weakly positive versus negative, P:0.5.
RESULTS In all sections in which glycosylated OFFN was found, it was present in the stroma rather than within the epethelial SCC. The staining was not confined to, nor was it more intense, around blood vessels. In involved lymph nodes tumour stroma was spares in most sections, making identification of staining difficult. Staining for glycosylated OFFN was found to be intense in the stroma of 43 SCC, diffuse or weak in 27 and absent in 30. In the cases with intense staining of glycosylated OFFN, cervical metastases were found in 17 of 27 patients who had undergone neck dissection (63%), compared with 6 of 17 (35%) with weak of diffuse staining and 3 of 13 (23%) necks in cases with no staining (P90.04) (Table 1). The 20 cases in which there was extracapsular spread in the lymph node specimen are detailed in Table 2. Table 3 and Fig. 3 show the number of patients who died according to the intensity of staining; intense staining was associated with significantly more deaths than no staining (P:0.02). In the weakly positive group there were six deaths (23%), this did not differ significantly from the mortality in the the strongly staining Table 1 Neck dissection and nodal status Staining
Number of necks dissected
Intense Weak None Total
27 17 13 57
2 for trend:6.3, P:0.01.
Number (%) with invaded nodes 17 (63) 6 (35) 3 (23) 26 (46) Fig. 3 High power photomicrograph of intense OFFN staining.
BJOM-172.QXD
11/24/01 9:52 AM
Page 475
Oncofetal fibronectin and oral squamous cell carcinoma
group (P:0.06) or that with no staining (P:0.4). There was no correlation between glycosylated OFFN staining and the size, site or grade of the primary tumour, and OFFN staining did not correlate with age, or consumption of tobacco or alcohol.
DISCUSSION Immunohistochemical analysis of previously histologically fixed tissue showed distinct variation in staining for OFFN; it was absent in normal tissue and variable in tumour tissue, so that 30 oral SCCs did not stain at all for glycosylated OFFN, 26 stained weakly and 43 intensely. This was in contrast to the study by Mandel et al.,11 which used fresh tissue specimens, and all 38 fresh samples of SCC were found to stain intensely for glycosylated OFFN and ED-B fibronectin. However, in the study by Jenek et al.,13 only 15% of the laryngeal squamous carcinomas stained intensely for ED-B OFFN as assessed by MoAb to fresh frozen tissue, although 53% were deemed weak. There is therefore considerable disparity in the observed expression of fibronectin isoforms as judged by immunohistochemistry in the two previous studies of oral and laryngeal SCC. Although these studies used fresh unfixed tissue, the apparent OFFN expression seen in the present study using previously fixed tissue is similar to the expression of ED-B isoforms found by Jenek et al.13 Moreover, the study on colorectal carcinoma that used archival material gave significant results, and archival material from paraffin blocks is perhaps more likely to permit analysis of representative sections rather than those which may be necrotic or not representative of the bulk of the tumour. It has been suggested that ED-B OFFN is a marker of angiogenesis and that immunohistochemical staining of this moiety in tumours is often found in the endothelial cell cytoplasm.17 There was definitely no evidence of any focusing of staining around blood vessels in the sections examined in this study for glycosylated OFFN expression with MoAb 5C10. The pattern of staining of this antibody was compared with that of antibody BC1 against ED-B fibronectin in the study on oral SCC carried out by Mandel et al.11 and was described as ‘quite similar’, and neither of these monoclonal antibodies showed focal staining around blood vessels. This study is therefore in accord with Mandel et al.11 with respect to OFFN deposition not being focused around blood vessels. It does not support the conjecture of Castellani et al.20 that ED-B OFFN is found in the endothelial cell cytoplasm rather than other parts of the tumour stroma. The three markers of tumour behaviour analysed in this study showed a clear positive correlation with glycosylated OFFN expression. The incidence of histologically confirmed lymph node metastases in the
475
13 patients who had undergone neck dissection in the group where there was no glycosylated OFFN expression was only 23% (n:3). It was higher in the 17 patients who underwent neck dissection, where there was weak glycosylated OFFN expression at 35% (n:6), but highest at 63% (n:17) in the 27 patients who underwent neck dissection where glycosylated OFFN expression was judged to be intense. These results were significant (P:0.001). The effect on extracapsular tumour spread was most dramatic in that this occurred exclusively in samples positive for glycosylated OFFN and occurred in 16 of 17 cases that stained intensely with lymph node metastases, as opposed to weak staining where only 4 of 13 cases with lymph node metastases showed extracapsular spread (P:0.0002). This alteration in matrix glycoproteins may therefore be a prerequisite for breaches of the lymph node capsule. Both lymph node metastases and extracapsular spread are known to affect prognosis adversely. Therefore it is not surprising that mortality until the end of the study period, when follow-up was from 2 to 11 years, was lowest in patients whose tumours did not express glycosylated OFFN. The difference in metastases and outcome across the three categories of glycosylated OFFN staining could not be ascribed to other factors; in all the categories the size of the tumours was predominantly 2–4 cm (T2). There was no significant difference between the proportions of smaller T1 tumours in the three categories or the proportion of larger T3 and T4 tumours (P:0.0156 Mann–Whitney U test). It might be expected that tumours that expressed glycosylated OFFN would grow faster, and so be staged larger on presentation. However, there are many factors that drive patients to present with oral cancer, and although the length of time an ulcer or white patch is present is a significant factor, the size of the lesion is also likely to be an important trigger to seek medical advice. Therefore Oral carcinomas with varying glycosylated OFFN expression must be expected to present at the same size. It is noteworthy that the study by Mandel et al.11 on oral squamous carcinoma included only three tumours in each of the T1 and T2 stages. In addition, 25 of the 38 patients had undergone radiotherapy, which suggests that radiotherapy as a combined or sole treatment had already failed. This, together with the relative paucity of small tumours, suggests that a group of highly aggressive tumours had been selected. The histology of the tumour is recognized as a useful prognostic indicator in oral SCC,21 in that poorly differentiated tumours do have a greater propensity to metastasize to cervical lymph nodes. The majority of tumours were moderately differentiated in all three categories, and the proportions of well and poorly differentiated tumours showed no significant variation across the categories of
BJOM-172.QXD
11/24/01 9:52 AM
476
Page 476
British Journal of Oral and Maxillofacial Surgery
glycosylated OFFN expression. This was in contrast to the results of the study on colorectal carcinomas by Infusa et al.12 and that on 79 cases of breast cancer by Rosa et al.,22 where there was a correlation between histologic grade and glycosylated OFFN expression. The fact that glycosylated OFFN expression did not correlated with grade of tumour does not detract from this study. On the contrary, the fact that glycosylated OFFN expression seems to affect prognosis independently of grade of tumour increases the value of glycosylated OFFN expression as an independent prognostic indicator. Tobacco and alcohol consumption did not influence glycosylated OFFN expression. OF the 87 patients for whom data were available about tobacco consumption, 57 were smokers and 30 non-smokers, and this did not correlate with OFFN expression (P:0.46). Nearly all the patients in the sample population consumed more than 20 units of alcohol a week (62/75) and this did not influence glycosylated OFFN expression (P:0.61). However, to obtain meaningful data on the interrelation of these factors with OFFN expression would require a larger sample of patients. Surprisingly the age distribution differed significantly different when the age distribution across the three groups of glycosylated OFFN expression was examined by the Kruskal Wallis nonparametric test (P:0.05). However, further analysis when each one the three groups of glycosylated OFFN expression was compared with the other two groups separately by the Mann–Whitney U test showed that the age distribution of those that strongly expressed glycosylated OFFN deposition did not differ from that of the other two. However, the age distribution between the group that weakly expressed glycosylated OFFN and the group that did, was significantly different (P:0.02). This was almost certainly because 8 of the 23 patients who weakly expressed glycosylated OFFN were in the 40–50-year age group, and 9 of the 28 patients who did not express glycosylated OFFN were in the 70–80-year age group. It is difficult to explain this difference, and in view of the fact that there was no difference between these two groups and the group that strongly expressed glycosylated OFFN, it does seem reasonable to suggest that this was the result of a sampling error. Clearly, comparisons between different studies using MoAbs for OFFN are prone to considerable errors, as some use fresh frozen tissue and others use paraffinfixed tissue. Some use MoAbs to glycosylated OFFN while others use MoAbs to ED-A and/or ED-B OFFN. Infusa et al.12 used paraffin-fixed tissue and although a MoAb against glycosylated OFFN was used, it was FD-6 rather than the 5C10 monoclonal antibody as used in this study. However, there is much overlap between and within studies that may justify comparison, particularly as the fresh tissue exposed to three different MoAbs
5C10 and FD-6 to glycosylated OFFN as well as ED-B FN, stained almost identically.11 It is tempting to explain the anomaly between the study by Mandel et al.11 and the others as resulting from error, with OFFN reactivity being lost with some fixed specimens. If this is the reason it is perhaps surprising that this loss of reactivity confers a better prognosis for patients in whose tumours OFFN was expressed, which was shown to be highly significant. The expression of OFFN is probably not an all or nothing effect, but a progressive process by which the more aggressive a tumour becomes, the greater the OFFN production in the stroma, as illustrated by the intermediate position in outcome exhibited in cases that weakly expressed OFFN. The production of glycosylated OFFN is not suddenly switched on; it merely increases as the tissue transforms from normal to malignant. How the substrate specificity of the ␣-N-acetyl galactosaminyl transferase is altered to produce glycosylated OFFN is unclear, but this alteration seems more likely to occur in those oral SCC that are most likely to invade widely and metastasise. The results of this study on OFFN expression and SCC show OFFN expression to be a powerful prognostic indicators. Staining of biopsy specimens for OFFN before definitive treatment for patients with no sign of lymph node involvement in the neck may be useful in planning treatment. However, it is unfortunate that the expression of this moiety occurs in normal healing tissues and during other processes, which makes it unsuitable either as a marker for oral cancer or a target for therapeutic agents. ACKNOWLEDGEMENTS We thank The British Association of Oral and Maxillofacial Surgeons, and the Research and Development Unit at Portsmouth Hospitals NHS Trusts for financial support, and Mr B. T. Evans and Mr G. Zaki for allowing their patients to be enrolled into this study.
REFERENCES 1. Morrison PR, Edsall JT, Miller SG. Preparation and properties of serum and plasma proteins: the separation of purified fibrinogen from a fraction of human plasma. J Am Chem Soc 1948; 70: 3103–3108. 2. Schwarzbauer JE. Fibronectin: from gene to protein. Curr Opin Cell Biol 1991; 3: 786–791. 3. Hynes RO. Fibronectins. Sci Am 1986; 254: 32–41. 4. Matsuura H, Hakomori S. The oncofetal domain of fibronectin defined by monoclonal antibody FDC-6: its presence in fibronectins from fetal and tumour tissues and its absence from normal adult tissues and plasma. Proc Natl Acad Sci USA 1985; 82: 6517–6521. 5. Nickeleit V, Kaufman AH, Zagachin AU, Dutt JE, Foster RB. Healing corneas express embryonic fibronectin isoforms in the epethelium, subepethelial stroma and endothelium. Am J Pathol 1996; 149: 549–555. 6. Berndt A, Hosmehl H, Mandel U et al. TGF and bGF synthesis and localization in Dupuytren’s disease (nodular palmar fibromatosis) relative to cellular activity, myofibroblast phenotype and oncofetal variants of fibronectin. Histochem J 1995; 27: 1014–1020.
BJOM-172.QXD
11/24/01 9:52 AM
Page 477
Oncofetal fibronectin and oral squamous cell carcinoma 7. Gabler U, Berndt A, Kosmehl HA et al. Matrix remodelling in dilated cardiomyopathy entails the occurrence of oncofetal fibronectin molecular variants. Heart 1995; 75: 358–362. 8. Mathews GA, fFrench-Constant C. Embryonic fibronectins are upregulated following peripheral nerve injury. J Neurobiol 1995; 26: 171–188. 9. Carnomella B, Balza E, Siri A et al. A tumour-associated fibronectin isoform generated by alternate splicing of messengerRNA precursors. J Cell Biol 1989; 108: 1139–1148. 10. Castellani P, Viale G, Dorcaratto A et al. The fibronectin isoform containing the ED-B oncofetal domain: a marker of angiogenesis. Int J Cancer 1994; 59: 612–618. 11. Mandel U, Gaggero B, Reibel J, Hamilton Therkildsen M, Dabelsteen E, Clausen H. Oncofetal fibronectins in oral carcinomas: correlation of two different types. APMIS 1994; 102: 695–702. 12. Infusa H, Nakamura M, Adachi T et al. Localization of oncofetal and normal fibronectin in colorectal cancer. Correlation with histologic grade, liver metastasis and prognosis. Cancer 1995; 75: 2802–2808. 13. Jenek R, Kruk~Zagajewska A, Zeromski J. Extracellular matrix proteins expression and incidence of tumor infiltrating cells in laryngeal carcinoma. Patol Pol 1994; 45: 179–186. 14. Farhadian F, Contard F, Corbier A, Barrieux A, Rappaport L, Samuel J-L. Fibronectin expression during physiological and pathological cardiac growth. J Mol Cell Cardiol 1995; 27: 981–990. 15. Farhadian F, Barrieux A, Lortet S et al. Differential splicing of fibronectin pre-messenger ribonucleic acid during cardiac ontogeny and development of hypertrophy in the rat. Lab Invest 1996; 71: 552–559. 16. Kosmehl H, Berndt A, Katenkamp D. Molecular variants of fibronectin and laminin; structure, physiological occurrence and histopathological aspects. Virchows Arch 1996; 429: 311–322. 17. Southwick HW. Elective neck dissection for intraoral cancer. JAMA 1971; 217: 454–455. 18. Hibbert J. Scott-Browne’s Otolaryngology (V), 6th edn. Oxford: Butterworth, 1997: 11–21. 19. American Joint Committee on Cancer. Manual for Staging of Cancer, 4th edn. Philadelphia: Lippincott, 1992: 28–38.
477
20. Castellani P, Viale G, Dorcaratto A et al. The fibronectin isoform containing the ED-B oncofetal domain: a maker of angiogenesis. Int J Cancer 1994; 59: 612–618. 21. Henk JM, Langdon JD. Malignant Tumours of the Oral Cavity. London: Edward Arnold, 1985: 64. 22. Rosa BL, Vielh P, Matsuura H et al. Distribution of oncofetal fibronectin in human mammary tumours: immuno-fluorescence study on histological sections. Cancer Res 1990; 50: 1608–1612.
The Authors A. J. Lyons MS FDS FRCS Consultant Oral and Maxillofacial Surgeon, King’s College Hospital, London SE5 9RW, UK A. C. Bateman MD MRCPath Consultant Histopathologist, Southampton University Hospitals Trust, Southampton SO9 1XY, UK A. V. Spedding MRCPath Consultant Histopathologist, Queen Alexandra Hospital, Portsmouth, PO4 3LY, UK J. N. Primrose MD FRCS Professor, Academic Surgical Unit, Southampton University Hospitals Trust, Southampton SO9 1XY, UK U. Mandel Associate Professor, Department of Oral Diagnostics, University of Copenhagen, Copenhagen, Denmark Correspondence and requests for offprints to: Mr A. J. Lyons, Department of Oral and Maxillofacial Surgery, King’s College Hospital, Caldecot Road, London SE5 9RW, UK. Tel:;44 (0)207 346 3474; Fax:;44 (0)207 346 3574; E-mail: [email protected] Accepted 6 June 2001
11/24/01 9:51 AM
Page 471
BRITISH
British Journal of Oral and Maxillofacial Surgery (2001) 39, 471–477 © 2001 The British Association of Oral and Maxillofacial Surgeons doi: 10.1054/bjom.2001.0702, available online at http://www.idealibrary.com on
Journal of Oral and Maxillofacial Surgery
Oncofetal fibronectin and oral squamous cell carcinoma A. J. Lyons,* A. C. Bateman,† A. Spedding,‡ J. N. Primrose,§ U. Mandel¶ *Consultant Oral and Maxillofacial Surgeon, King’s College Hospital, London, UK; †Consultant Histopathologist, Southampton University Hospitals Trust, Southampton, UK; ‡Consultant Histopathologist, Queen Alexandra Hospital, Portsmouth, UK; §Professor, Academic Surgical Unit, Southampton University Hospitals Trust, Southampton, UK; ¶Associate Professor, Department of Oral Diagnostics, University of Copenhagen, Copenhagen, Denmark SUMMARY. Fibronectin is a cell matrix glycoprotein, which exists as a number of isoforms that are often found within the cell matrix that surrounds tumours. Collectively these tumour-associated isomers of fibronectin have been termed oncofetal fibronectin (OFFN). We looked for expression of OFFN within oral squamous cell carcinomas (SCC) and related its presence to prognosis. The investigation used a monoclonal antibody (MoAb 5C10) to the glycosylated variant of OFFN, and 100 archival specimens of oral SSC. Immunostaining for OFFN was intense in the adjacent stroma of 43 squamous carcinomas, weak in 27 and absent in 30. Cervical metastases were found in 17/27 (63%) specimens that stained intensely, 6/17 (35%) that stained weakly and 3/13 (23%) that did not stain. Of the 21 cases which had extracapsular lymph node spread, 81% were from those that stained intensely, 19% from those that stained weakly and none from those that did not stain for OFFN expression. Also, 21/44 patients (49%) died in group with intense OFFN staining, 6/26 (23%) in the group with weak staining and 3/30 (10%) in the group that did not stain. The presence of OFFN glycoprotein in oral SCC as evaluated by immunostaining with MoAb 5C10 correlates strongly with the presence of metastatic lymph node involvement, particularly extracapsular involvement, and mortality. We therefore suggest that the degree of expression of OFFN in tumours is a valuable prognostic indicator. © 2001 The British Association of Oral and Maxillofacial Surgeons
and ED-B). In view of these attributes, FN that contains ED-A and or ED-B has been named OFFN. Over the past two decades more has become known about the structure and function of fibronectin and OFFN. As the extra segments that constituted OFFN were only defined in the mid-1980s, it is possible that many of the properties ascribed to fibronectin in papers published before this date may not apply to OFFN or conversely may apply only to OFFN. Certainly, OFFN contains all the segments of the originally described fibronectin molecule, and so can be expected to share many of its attributes (Fig. 1). OFFN has also been detected in healing human cornea5 and related lesions, including fibrotic liver cirrhosis, dilated cardiomyopathy, proliferative glomerulonephritis and fibromatosis.6,7 There is some evidence that the regeneration of other tissues such as peripheral nerves is accompanied by a re-expression of ED-B OFFN.8 There is much evidence that the ED-B OFFN isoform is involved in angiogenesis. ED-B OFFN is never seen in mature vessels, but newly formed vessels in normal and tumour tissues are characterized by an abundant
INTRODUCTION Fibronectin was first extracted from human plasma by Morrison et al.1 in 1948 and was named ‘cold insoluble globulin’ as it precipitated in the presence of fibrinogen at low temperature. The molecule is made up of three types of peptide sequence homologues (I, II and III) and distinct domains that have the ability to bind fibrin, heparin, collagen, gelatin, factor XIIIa transglutaminase, Staphylococcus aureus and cells.2 It is present in the extracellular matrix and body fluids, and seems to have important roles in intercellular adhesion and in binding cells to the extracellular matrix.3 In 1985 Matsuura and Hakomori,4 reported the existence of a second type of fibronectin with a molecular weight of 310 000–335 000 Da, in which there was an extra glycosylated side chain, attached to the IIICS region of the molecule. They called this variant oncofetal fibronectin (OFFN), as its expression seemed to be confined to tumour and fetal tissues. Subsequently, further parts of the molecule that accounted for the higher molecular weight of this variant were defined (ED-A 471
BJOM-172.QXD
11/24/01 9:51 AM
472
Page 472
British Journal of Oral and Maxillofacial Surgery
Fig. 1 Oncofetal fibronectin molecule and interactions.
ED-B OFFN matrix.9 So, quantification of ED-B expression in tumour-associated vessels may predict the ability of these tumours to induce angiogenesis, and may be of similar prognostic significance to the evaluation of microvessel density.10 Deposition of OFFN has been observed in about 30 tumour types, including breast tumours meningiomas and pituitary adenomas.11 The intensity of OFFN staining in colorectal carcinoma has been directly correlated with advanced local stage, liver metastases and general prognosis.12 The expression of OFFN in head and neck squamous carcinomas has been evaluated in two published studies. In a Polish study by Jenek et al.,13 a number of extracellular matrix proteins (ECM) were examined using immunohistochemistry in specimens of laryngeal squamous carcinomas from 20 patients. Although the stage of the tumours was not described, all samples were taken from laryngectomy specimens, which may imply that these were large tumours, as in Northern Europe and the USA smaller lesions are usually treated by radiotherapy. OFFN staining was present and variable to some degree in well over half of them. These results are in contrast to those reported by Mandel et al.,11 who took fresh frozen tissue from 38 cases of oral squamous carcinoma. Radiotherapy had previously been given to 25 of the 38 patients, implying that these tumours were likely to be large or recurrent. All 38 cases stained strongly for both ED-B and glyocosylated OFFN with monoclonal antibodies. Their results concurred with those of other studies,7,14–16 all of which have shown that expression of ED-B invariably coexists with glycosylated OFFN. None of 15 premalignant tissue samples
taken stained for OFFN. This uniformity of staining for OFFN in oral SCC, as compared with the variability seen in laryngeal squamous cell carcinoma, might reflect the biological differences between the two types of tumour. For example, whereas the proportion of oral tumours that present with metastases in lymph nodes has been reported to be as high as 60%,17 that for laryngeal tumours has been placed at only 18%.18 In view of this controversy, an immunohistochemical study was done on 100 cases of oral SCC to find out if the degree of OFFN could predict the presence of metastases and outcome, as seemed to be the case in colorectal cancer.12 The expression of OFFN in the paraffin sections of 100 cases of oropharyngeal carcinoma was examined. All patients with this diagnosis were identified at two centres over 9-year period from their histopathological records, and the time from treatment to analysis of the data varied from 2 to 11 years. Only those patients with adequate clinical records and whose outcome was known (survived or died) were included in the study. To find out if the expression of OFFN by oropharyngeal carcinoma could be related to tumour behaviour and outcome, this expression was corelated with the following clinical and pathological variables: lymph node spread and extralymphatic spread; patient’s age, tobacco and alcohol consumption, and mortality; and the size,19 grade and site of the tumour. Both the previous studies in aerodigestive SCC had used fresh frozen material for immunohistochemical analysis, but valid results has been obtained by Infusa et al.12 with trypsin-treated archival material from colorectal tumours, so archival material was used in this study.
BJOM-172.QXD
11/24/01 9:51 AM
Page 473
Oncofetal fibronectin and oral squamous cell carcinoma
MATERIAL AND METHODS Light microscopy of sections stained with haematoxylin and eosin was used to select the most representative section from each of the cases used in the study. For most cases the main resection specimen was examined together with any metastases to lymph nodes that were resected at this operation or subsequent operations. For patients who only had radiotherapy, an incisional biopsy specimen was all that was examined. The formalin-fixed paraffin blocks sections were prepared, usually in batches of 10, by the following method. A microtome was used to cut 4-m sections off the paraffin blocks, which were placed on to 3⬘aminopropyl-triethoxy-silane (APES)-coated glass slides (Vector Laboratories) and dried at 60⬚C for 30 minutes. Endogenous peroxidase within the sections was blocked by the use of 5% hydrogen peroxide solution in alcohol. The sections were washed again with fresh alcohol and then rehydrated with water. Sections were then transferred to boiling citrate buffer at pH 6.0 in a pressure cooker and treated at high pressure for 2 minutes. After the pressure was released the sections were rapidly cooled in tap water, ensuring that they did not dry out at any time. After this, normal goat serum (10%) in phosphate-buffered saline (PBS) containing avidin blocking reagent at a
Fig. 2 Intensity of expression of oncofetal fibronectin and mortality.
473
dilution of one drop per ml was applied to the sections, which were then incubated at room temperature for 60 minutes. Excess serum was shaken off the sections, and primary antibody (MoAb 5C10) with one drop per ml of avidin blocking reagent was added. Negative controls were incubated with PBS, to which one drop per ml of avidin blocking agent had been added, and were then processed in the same way as the other samples. Monoclonal antibody (MoAb 5C10) was diluted in PBS in concentrations varying from neat to 1 in 1000. A dilution of 1 in 10 was found to produce optimal staining in a series of sections of oral squamous cell carcinoma. Sections were again washed three times with PBS as before and then with water, before incubation with biotinylated secondary antibody (Vector Laboratories) for 30 minutes at room temperature. Streptococcal ABC/peroxidase complex (Vector Laboratories) was then applied and incubated with the sections at room temperature for 30 minutes, after which they were rinsed as before with PBS and then with water. Labelling was visualized with diamino benzadine (Vector Laboratories) and counterstained with haematoxylin; the slides were then dehydrated, cleared and mounted. All applications and incubations were done at room temperature. Sections were examined by light microscopy to assess intensity of staining for OFFN, and graded as absent, weak, patchy, diffuse or strong (Fig. 2).
BJOM-172.QXD
11/24/01 9:52 AM
474
Page 474
British Journal of Oral and Maxillofacial Surgery
This was assessed subjectively by at least two histopathologists. Despite the apparent arbitrary nature of this grading, agreement between the two assessors was 98%. The two disputed sections were finally graded after a second look. Expression of OFFN progressed from strong to weak to no expression, and was hence ordered. In nearly all cases the data was judged to be parametric and compared against other categorical data (for example, presence or absence of lymph node metastases). Where OFFN was compared against age groups, the numbers in each group were too small to judge as parametric and the non-parametric Mann–Whitney U test was used. As mortality was a censored observation, being assessed at different times after treatment for each patient and the influence of OFFN was expected to be unaffected by the passage of time, the log rank test was used. All statistical analysis employed an SPSS version 9 statistics programme.
Table 2 Staining and extracapsular spread from lymph nodes Staining
Number of necks dissected
Number with extracapsular spread
Intense Weak None Total
27 17 13 57
16 4 0 20
2 for trend:18.4, P:0.001.
Table 3 Mortality Staining
Number of patients
Deaths (%)
Intense Weak None Total
21/44 6/26 3/30 20/30
49 23 10 …
Log rank analysis: strongly positive versus weakly positive, P:0.06; strongly positive versus negative, P:0.02; weakly positive versus negative, P:0.5.
RESULTS In all sections in which glycosylated OFFN was found, it was present in the stroma rather than within the epethelial SCC. The staining was not confined to, nor was it more intense, around blood vessels. In involved lymph nodes tumour stroma was spares in most sections, making identification of staining difficult. Staining for glycosylated OFFN was found to be intense in the stroma of 43 SCC, diffuse or weak in 27 and absent in 30. In the cases with intense staining of glycosylated OFFN, cervical metastases were found in 17 of 27 patients who had undergone neck dissection (63%), compared with 6 of 17 (35%) with weak of diffuse staining and 3 of 13 (23%) necks in cases with no staining (P90.04) (Table 1). The 20 cases in which there was extracapsular spread in the lymph node specimen are detailed in Table 2. Table 3 and Fig. 3 show the number of patients who died according to the intensity of staining; intense staining was associated with significantly more deaths than no staining (P:0.02). In the weakly positive group there were six deaths (23%), this did not differ significantly from the mortality in the the strongly staining Table 1 Neck dissection and nodal status Staining
Number of necks dissected
Intense Weak None Total
27 17 13 57
2 for trend:6.3, P:0.01.
Number (%) with invaded nodes 17 (63) 6 (35) 3 (23) 26 (46) Fig. 3 High power photomicrograph of intense OFFN staining.
BJOM-172.QXD
11/24/01 9:52 AM
Page 475
Oncofetal fibronectin and oral squamous cell carcinoma
group (P:0.06) or that with no staining (P:0.4). There was no correlation between glycosylated OFFN staining and the size, site or grade of the primary tumour, and OFFN staining did not correlate with age, or consumption of tobacco or alcohol.
DISCUSSION Immunohistochemical analysis of previously histologically fixed tissue showed distinct variation in staining for OFFN; it was absent in normal tissue and variable in tumour tissue, so that 30 oral SCCs did not stain at all for glycosylated OFFN, 26 stained weakly and 43 intensely. This was in contrast to the study by Mandel et al.,11 which used fresh tissue specimens, and all 38 fresh samples of SCC were found to stain intensely for glycosylated OFFN and ED-B fibronectin. However, in the study by Jenek et al.,13 only 15% of the laryngeal squamous carcinomas stained intensely for ED-B OFFN as assessed by MoAb to fresh frozen tissue, although 53% were deemed weak. There is therefore considerable disparity in the observed expression of fibronectin isoforms as judged by immunohistochemistry in the two previous studies of oral and laryngeal SCC. Although these studies used fresh unfixed tissue, the apparent OFFN expression seen in the present study using previously fixed tissue is similar to the expression of ED-B isoforms found by Jenek et al.13 Moreover, the study on colorectal carcinoma that used archival material gave significant results, and archival material from paraffin blocks is perhaps more likely to permit analysis of representative sections rather than those which may be necrotic or not representative of the bulk of the tumour. It has been suggested that ED-B OFFN is a marker of angiogenesis and that immunohistochemical staining of this moiety in tumours is often found in the endothelial cell cytoplasm.17 There was definitely no evidence of any focusing of staining around blood vessels in the sections examined in this study for glycosylated OFFN expression with MoAb 5C10. The pattern of staining of this antibody was compared with that of antibody BC1 against ED-B fibronectin in the study on oral SCC carried out by Mandel et al.11 and was described as ‘quite similar’, and neither of these monoclonal antibodies showed focal staining around blood vessels. This study is therefore in accord with Mandel et al.11 with respect to OFFN deposition not being focused around blood vessels. It does not support the conjecture of Castellani et al.20 that ED-B OFFN is found in the endothelial cell cytoplasm rather than other parts of the tumour stroma. The three markers of tumour behaviour analysed in this study showed a clear positive correlation with glycosylated OFFN expression. The incidence of histologically confirmed lymph node metastases in the
475
13 patients who had undergone neck dissection in the group where there was no glycosylated OFFN expression was only 23% (n:3). It was higher in the 17 patients who underwent neck dissection, where there was weak glycosylated OFFN expression at 35% (n:6), but highest at 63% (n:17) in the 27 patients who underwent neck dissection where glycosylated OFFN expression was judged to be intense. These results were significant (P:0.001). The effect on extracapsular tumour spread was most dramatic in that this occurred exclusively in samples positive for glycosylated OFFN and occurred in 16 of 17 cases that stained intensely with lymph node metastases, as opposed to weak staining where only 4 of 13 cases with lymph node metastases showed extracapsular spread (P:0.0002). This alteration in matrix glycoproteins may therefore be a prerequisite for breaches of the lymph node capsule. Both lymph node metastases and extracapsular spread are known to affect prognosis adversely. Therefore it is not surprising that mortality until the end of the study period, when follow-up was from 2 to 11 years, was lowest in patients whose tumours did not express glycosylated OFFN. The difference in metastases and outcome across the three categories of glycosylated OFFN staining could not be ascribed to other factors; in all the categories the size of the tumours was predominantly 2–4 cm (T2). There was no significant difference between the proportions of smaller T1 tumours in the three categories or the proportion of larger T3 and T4 tumours (P:0.0156 Mann–Whitney U test). It might be expected that tumours that expressed glycosylated OFFN would grow faster, and so be staged larger on presentation. However, there are many factors that drive patients to present with oral cancer, and although the length of time an ulcer or white patch is present is a significant factor, the size of the lesion is also likely to be an important trigger to seek medical advice. Therefore Oral carcinomas with varying glycosylated OFFN expression must be expected to present at the same size. It is noteworthy that the study by Mandel et al.11 on oral squamous carcinoma included only three tumours in each of the T1 and T2 stages. In addition, 25 of the 38 patients had undergone radiotherapy, which suggests that radiotherapy as a combined or sole treatment had already failed. This, together with the relative paucity of small tumours, suggests that a group of highly aggressive tumours had been selected. The histology of the tumour is recognized as a useful prognostic indicator in oral SCC,21 in that poorly differentiated tumours do have a greater propensity to metastasize to cervical lymph nodes. The majority of tumours were moderately differentiated in all three categories, and the proportions of well and poorly differentiated tumours showed no significant variation across the categories of
BJOM-172.QXD
11/24/01 9:52 AM
476
Page 476
British Journal of Oral and Maxillofacial Surgery
glycosylated OFFN expression. This was in contrast to the results of the study on colorectal carcinomas by Infusa et al.12 and that on 79 cases of breast cancer by Rosa et al.,22 where there was a correlation between histologic grade and glycosylated OFFN expression. The fact that glycosylated OFFN expression did not correlated with grade of tumour does not detract from this study. On the contrary, the fact that glycosylated OFFN expression seems to affect prognosis independently of grade of tumour increases the value of glycosylated OFFN expression as an independent prognostic indicator. Tobacco and alcohol consumption did not influence glycosylated OFFN expression. OF the 87 patients for whom data were available about tobacco consumption, 57 were smokers and 30 non-smokers, and this did not correlate with OFFN expression (P:0.46). Nearly all the patients in the sample population consumed more than 20 units of alcohol a week (62/75) and this did not influence glycosylated OFFN expression (P:0.61). However, to obtain meaningful data on the interrelation of these factors with OFFN expression would require a larger sample of patients. Surprisingly the age distribution differed significantly different when the age distribution across the three groups of glycosylated OFFN expression was examined by the Kruskal Wallis nonparametric test (P:0.05). However, further analysis when each one the three groups of glycosylated OFFN expression was compared with the other two groups separately by the Mann–Whitney U test showed that the age distribution of those that strongly expressed glycosylated OFFN deposition did not differ from that of the other two. However, the age distribution between the group that weakly expressed glycosylated OFFN and the group that did, was significantly different (P:0.02). This was almost certainly because 8 of the 23 patients who weakly expressed glycosylated OFFN were in the 40–50-year age group, and 9 of the 28 patients who did not express glycosylated OFFN were in the 70–80-year age group. It is difficult to explain this difference, and in view of the fact that there was no difference between these two groups and the group that strongly expressed glycosylated OFFN, it does seem reasonable to suggest that this was the result of a sampling error. Clearly, comparisons between different studies using MoAbs for OFFN are prone to considerable errors, as some use fresh frozen tissue and others use paraffinfixed tissue. Some use MoAbs to glycosylated OFFN while others use MoAbs to ED-A and/or ED-B OFFN. Infusa et al.12 used paraffin-fixed tissue and although a MoAb against glycosylated OFFN was used, it was FD-6 rather than the 5C10 monoclonal antibody as used in this study. However, there is much overlap between and within studies that may justify comparison, particularly as the fresh tissue exposed to three different MoAbs
5C10 and FD-6 to glycosylated OFFN as well as ED-B FN, stained almost identically.11 It is tempting to explain the anomaly between the study by Mandel et al.11 and the others as resulting from error, with OFFN reactivity being lost with some fixed specimens. If this is the reason it is perhaps surprising that this loss of reactivity confers a better prognosis for patients in whose tumours OFFN was expressed, which was shown to be highly significant. The expression of OFFN is probably not an all or nothing effect, but a progressive process by which the more aggressive a tumour becomes, the greater the OFFN production in the stroma, as illustrated by the intermediate position in outcome exhibited in cases that weakly expressed OFFN. The production of glycosylated OFFN is not suddenly switched on; it merely increases as the tissue transforms from normal to malignant. How the substrate specificity of the ␣-N-acetyl galactosaminyl transferase is altered to produce glycosylated OFFN is unclear, but this alteration seems more likely to occur in those oral SCC that are most likely to invade widely and metastasise. The results of this study on OFFN expression and SCC show OFFN expression to be a powerful prognostic indicators. Staining of biopsy specimens for OFFN before definitive treatment for patients with no sign of lymph node involvement in the neck may be useful in planning treatment. However, it is unfortunate that the expression of this moiety occurs in normal healing tissues and during other processes, which makes it unsuitable either as a marker for oral cancer or a target for therapeutic agents. ACKNOWLEDGEMENTS We thank The British Association of Oral and Maxillofacial Surgeons, and the Research and Development Unit at Portsmouth Hospitals NHS Trusts for financial support, and Mr B. T. Evans and Mr G. Zaki for allowing their patients to be enrolled into this study.
REFERENCES 1. Morrison PR, Edsall JT, Miller SG. Preparation and properties of serum and plasma proteins: the separation of purified fibrinogen from a fraction of human plasma. J Am Chem Soc 1948; 70: 3103–3108. 2. Schwarzbauer JE. Fibronectin: from gene to protein. Curr Opin Cell Biol 1991; 3: 786–791. 3. Hynes RO. Fibronectins. Sci Am 1986; 254: 32–41. 4. Matsuura H, Hakomori S. The oncofetal domain of fibronectin defined by monoclonal antibody FDC-6: its presence in fibronectins from fetal and tumour tissues and its absence from normal adult tissues and plasma. Proc Natl Acad Sci USA 1985; 82: 6517–6521. 5. Nickeleit V, Kaufman AH, Zagachin AU, Dutt JE, Foster RB. Healing corneas express embryonic fibronectin isoforms in the epethelium, subepethelial stroma and endothelium. Am J Pathol 1996; 149: 549–555. 6. Berndt A, Hosmehl H, Mandel U et al. TGF and bGF synthesis and localization in Dupuytren’s disease (nodular palmar fibromatosis) relative to cellular activity, myofibroblast phenotype and oncofetal variants of fibronectin. Histochem J 1995; 27: 1014–1020.
BJOM-172.QXD
11/24/01 9:52 AM
Page 477
Oncofetal fibronectin and oral squamous cell carcinoma 7. Gabler U, Berndt A, Kosmehl HA et al. Matrix remodelling in dilated cardiomyopathy entails the occurrence of oncofetal fibronectin molecular variants. Heart 1995; 75: 358–362. 8. Mathews GA, fFrench-Constant C. Embryonic fibronectins are upregulated following peripheral nerve injury. J Neurobiol 1995; 26: 171–188. 9. Carnomella B, Balza E, Siri A et al. A tumour-associated fibronectin isoform generated by alternate splicing of messengerRNA precursors. J Cell Biol 1989; 108: 1139–1148. 10. Castellani P, Viale G, Dorcaratto A et al. The fibronectin isoform containing the ED-B oncofetal domain: a marker of angiogenesis. Int J Cancer 1994; 59: 612–618. 11. Mandel U, Gaggero B, Reibel J, Hamilton Therkildsen M, Dabelsteen E, Clausen H. Oncofetal fibronectins in oral carcinomas: correlation of two different types. APMIS 1994; 102: 695–702. 12. Infusa H, Nakamura M, Adachi T et al. Localization of oncofetal and normal fibronectin in colorectal cancer. Correlation with histologic grade, liver metastasis and prognosis. Cancer 1995; 75: 2802–2808. 13. Jenek R, Kruk~Zagajewska A, Zeromski J. Extracellular matrix proteins expression and incidence of tumor infiltrating cells in laryngeal carcinoma. Patol Pol 1994; 45: 179–186. 14. Farhadian F, Contard F, Corbier A, Barrieux A, Rappaport L, Samuel J-L. Fibronectin expression during physiological and pathological cardiac growth. J Mol Cell Cardiol 1995; 27: 981–990. 15. Farhadian F, Barrieux A, Lortet S et al. Differential splicing of fibronectin pre-messenger ribonucleic acid during cardiac ontogeny and development of hypertrophy in the rat. Lab Invest 1996; 71: 552–559. 16. Kosmehl H, Berndt A, Katenkamp D. Molecular variants of fibronectin and laminin; structure, physiological occurrence and histopathological aspects. Virchows Arch 1996; 429: 311–322. 17. Southwick HW. Elective neck dissection for intraoral cancer. JAMA 1971; 217: 454–455. 18. Hibbert J. Scott-Browne’s Otolaryngology (V), 6th edn. Oxford: Butterworth, 1997: 11–21. 19. American Joint Committee on Cancer. Manual for Staging of Cancer, 4th edn. Philadelphia: Lippincott, 1992: 28–38.
477
20. Castellani P, Viale G, Dorcaratto A et al. The fibronectin isoform containing the ED-B oncofetal domain: a maker of angiogenesis. Int J Cancer 1994; 59: 612–618. 21. Henk JM, Langdon JD. Malignant Tumours of the Oral Cavity. London: Edward Arnold, 1985: 64. 22. Rosa BL, Vielh P, Matsuura H et al. Distribution of oncofetal fibronectin in human mammary tumours: immuno-fluorescence study on histological sections. Cancer Res 1990; 50: 1608–1612.
The Authors A. J. Lyons MS FDS FRCS Consultant Oral and Maxillofacial Surgeon, King’s College Hospital, London SE5 9RW, UK A. C. Bateman MD MRCPath Consultant Histopathologist, Southampton University Hospitals Trust, Southampton SO9 1XY, UK A. V. Spedding MRCPath Consultant Histopathologist, Queen Alexandra Hospital, Portsmouth, PO4 3LY, UK J. N. Primrose MD FRCS Professor, Academic Surgical Unit, Southampton University Hospitals Trust, Southampton SO9 1XY, UK U. Mandel Associate Professor, Department of Oral Diagnostics, University of Copenhagen, Copenhagen, Denmark Correspondence and requests for offprints to: Mr A. J. Lyons, Department of Oral and Maxillofacial Surgery, King’s College Hospital, Caldecot Road, London SE5 9RW, UK. Tel:;44 (0)207 346 3474; Fax:;44 (0)207 346 3574; E-mail: [email protected] Accepted 6 June 2001