Int. J . Cancer: 47, 853-857 (1991) 0 1991 Wiley-Liss, Inc.
Publication of the International Union Against Cancer Publication de I’Union Internationale Contre le Cancer
PROGNOSTIC SIGNIFICANCE OF MONOCLONAL ANTIBODY 3H- 1 REACTIVITY WITH SQUAMOUS-CELL HEAD-AND-NECK CANCERS John H. KEARSLEY~, J. Helen LEONARD, Hiroyulu TAKAHASHI, Diana BATTISTUTTA, Peter G . PARSONS and Denis J. MOSS Queensland Institute of Medical Research, Bramston Terrace, Herston QLB 4006, Australia. We report the reactivity of a novel murine IgM monoclonal antibody (MAb), 3H-I, in formalin-fixed tissue from a series of 44 patients with squamous-cell cancers (SCCs) of the upper aerodigestive tract. The antigen detected by 3H-I MAb is not expressed in simple squamous epithelia and only rarely in non-squamous malignancies. The following staining patterns were found to be associated with a progressively unfavourable prognosis: membranous, diffuse cytoplasmic and focal cytoplasmidnil staining. There were no statistically-significant associations between focal cytoplasmidnil staining and traditionally used prognostic parameters such as tumour size, nodal involvement or pathologic grade. However, multivariate analyses demonstrated that focal or nil staining was a significant independent prognostic factor for survival (p < 0.001) and was the only significant prognostic factor for relapse (p < 0.001). The converse applied to tumours with predominantly membranous staining ( p = 0.004). 3H-I MAb, with the advantage of relying on the pattern rather than the intensity of staining, makes it possible to identify different cellular phenotypic sub-populations more accurately than by conventional means, and may ultimately improve our understanding of the mechanisms underlying transformation from a benign or proliferative state to a malignant one.
Squamous-cell cancers (SCCs) of the upper aerodigestive tract represent 5-10% of newly diagnosed cancers in the Westem world (Pindborg, 1965) although their incidence is highest among individuals who maintain a heavy consumption of alcohol and cigarettes (Cann et al., 1985). In some countries, head-and-neck SCCs represent the most common form of malignant disease (Desai, 1983). Traditional therapy with surgery and/or radiotherapy can be curative in selected cases. However, early and late morbidity from the functional, cosmetic and socio-economic impact of treatment can significantly reduce a patient’s quality of life. Furthermore, incurable disease following treatment almost invariably remains localized in the head and neck region, so that terminal suffering is protracted and distressing. Prognostic information is essential for the evaluation, judgement and optimal treatment of patients with SCC of the upper aerodigestive tract. Although routine histomorphologic grading systems and light microscopic appearances are of some benefit in predicting clinical outcome (Zarbo and Crissman, 1988), such criteria remain subjective and poorly standardized, and must be regarded as relatively crude approximations of biologic behaviour at the all-important cellular level. The development of hybridoma technology (Kohler and Milstein, 1975) has provided a tool for a more accurate assessment of the structural and functional changes which occur in malignant disease. We have described, in a companion report (Parsons et al., 1991), the derivation and characterization of a novel IgM MAb 3H-1 which has a high specificity for squamous-cell disorders and which reacts with formalin-fixed tissues. MAb 3H-1 appears to produce at least 3 well-defined patterns of reactivity in SCCs and in squamo-proliferative lesions over a range of anatomic sites. The present report is an extension of our previous work and aims to determine whether the pattern of MAb 3H-1 reactivity carries prognostic significance in a series of formalin-fixed tissue sections of headand-neck SCCs.
MATERIAL AND METHODS
Tissue specimens Specimens were obtained from 44 untreated patients with SCCs from a range of anatomic sites within the upper aerodigestive tract (oral cavity 28, pharynx 10, larynx 6 ) . Normal squamous epithelium was present in the same section in many cases. There were 31 males and 13 females with a mean age of 59.2 years (range 41-79 years). All specimens had been fixed in buffered formaldehyde solution and routinely embedded in paraffin wax. Serial 4-pm-thick sections were cut from each block, and one slide was stained with haematoxylin and eosin for routine histopathologic assessment and grading. When keratinization with squamous epithelial pearls was predominant, the tumour was graded as well-differentiated. When keratinization was only present in some areas of the tumour, it was graded as poorly differentiated. When there appeared to be a fairly even mixture of keratinizing and non-keratinizing areas, the tumour was graded as moderately well differentiated. Some keratinization was present in all of the tumours studied. Monoclonal antibody MAb 3H-1 was derived as described by Parsons et al. (1991) from a fusion of BALB/c mouse splenocytes immunized with UV-irradiated keratinocytes which had previously been transformed with SV40 virus (KJD-USV40 cell line) (Brown and Gallimore, 1987). 3H-1 is an IgM MAb which recognizes an antigen with a molecular weight of approximately 55 kDa. Immunohistochemistry Immunohistochemical staining was performed using the avidin-biotin-peroxidase (ABC) technique (PK-4005; Vector, Burlingame, CA). Briefly, tissue sections were deparaffinized, dehydrated in 100% alcohol and then rehydrated through decreasing alcohol concentrations to water. Sections were treated with 0.3% (w/v) hydrogen peroxide and 20% (w/v) methanol in phosphate-buffered saline (5 m~ Na,HPO,, 3 m~ KH,PO,, 145 m~ NaCl pH7.2; PBS) for 10 min to eliminate endogenous peroxidase activity. Sections were then incubated with diluted horse serum for 20 min to eliminate non-specific binding of immunoglobulins. After removing excess horse serum, 3H- 1 MAb was added to each section as a supernatant and the slides were incubated for 30 min at room temperature. After several washes in PBS, sections were incubated with diluted biotinylated horse anti-mouse IgM immunoglobulin for 30 min at room temperature. The slides were then washed in PBS and incubated for 45 min with the Vectastain reagent, according to the manufacturer’s directions. After a further wash in PBS, sections were incubated for 3-10 min with 3amino-8-ethylcarbazole to which 10 drops of 30% H2,02had been added. The sections were finally counterstained with haemalum, rinsed and mounted under a cover-slip with Glycergel (DAKO, Carpinteria, CA). ‘To whom correspondence and reprint requests should be addressed. Received: November 5 , 1990.
854
KEARSLEY ET AL.
All immunoperoxidase procedures included both positive and negative control slides. The positive control was a formalin-fixed section from an SCC of the tongue which consistently showed strong immunoperoxidase staining with 3H- 1 antibody. Negative controls were incubated with PBS or an irrelevant murine IgM antibody (reactive with formalin-fixed material) instead of 3H-1 antibody in the first step and were then treated as described above. Immunohistochemical evaluation
Initial assessment was made at low microscopic power to examine the distribution of squamous cells and to ascertain whether there were any obvious gross variations in staining. Multiple sites from each SCC were studied and the distribution pattern of reaction product within positive cells was recorded as purely membranous, diffuse cytoplasmic, focal cytoplasmic or nil staining. The percentage of squamous cancer cells showing each type of immunoreactivity or no immunoreactivity was estimated as a percentage of the total number of malignant squamous cells seen in each section (Fig. 1). Statistical methods
Associations between categorical variables (T, N stage; pathologic grade; patterns of 3H- 1 staining; relapse-free and overall survivals) were investigated using chi-squared tests with Yates’ correction where appropriate (Armitage, 1971). Univariate differences in relapse and death rates over time according to these variables were considered using Kaplan and Meier (1958) survival analyses. Cox (1972) proportional haz-
ards regression was used in multivariate analyses to investigate the independent contributions of the various prognostic factors. All analyses were performed using the EGRET (Statistics and Epidemiology Research Corporation, Seattle, WA) and SPSSPC+ (SPSS, Chicago, IL) package on a microcomputer. In exploring for associations between the variables and with overall or recurrence-free survival, due to a lack of numbers in the extreme categories of the prognostic variables, tumour size was grouped as less than 4cm vs. greater than 4cm; nodal involvement was grouped as NO vs. N1-4, and pathologic grade grouped as well differentiated vs. moderately/poorly differentiated. Patients were grouped according to the percentage of tumour cells with focal cytoplasmic or nil staining. The quartiles of the distribution were W O % , 41-70%, 71-90% and 91-100%. RESULTS
Of the 44 SCC samples, 10 (23%) were graded as welldifferentiated, 27 (61%) were moderately differentiated, and 7 (16%) were poorly differentiated. Seventy-three percent of patients (32/44) had tumours measuring 2-6 cm. More than half (24 patients) had no nodal involvement. Three distinct patterns of MAb 3H- 1 reactivity were seen in this series and most tissue sections demonstrated varying proportions of cells with each pattern. Multiple sections from each SCC were screened and the results were consistently reproducible. The first pattern (membranous) was strictly limited to the cell membrane, (Fig. l a ) ; in the second (diffuse cytoplasmic)
FIGURE1 - Formalin-fixed SCC samples demonstrating the 3 patterns of 3H-1 reactivity (a) membranous, bar = 100 pm; (b) diffuse cytoplasmic, bar = 50 pm; (c) focal cytoplasmic, bar = 50 pm; ( d ) focal cytoplasmic (high-power), bar = 50 pm.
855
PROGNOSTIC SIGNIFICANCE OF MAb 3H-1
the membranous component was slightly less prominent but there was an additional diffusely granular cytoplasmic staining reaction, (Fig. lb). The third pattern of 3H-1 reactivity (focal cytoplasmic) consisted of focal, granular cytoplasmic deposits often situated in a paranuclear location, (Fig. lc,d). Some SCC cells did not stain at all (nil staining). The pattern of 3H-1 reactivity bore a close relationship to the architecture of individual tumours. In the tumours which invaded as cellular islands there was usually no reactivity in the peripheral cells, Adjacent, more centrally located cells in these islands usually demonstrated a focal cytoplasmic reactivity and the most central cells usually exhibited a membranous pattern (Fig. l a ) . However, the width of each reactive “layer” of cells could vary widely between individual tumours, from cellular islands composed totally of membranous reactivity to multicellular layers of nil or focal cytoplasmic reactivity (Fig. l a ) . There was no staining of keratin whorls when these structures were located in the centre of invading cellular islands. When squamous cancer cells were seen to invade in a diffuse fashion (in either small clumps or as linear streams of malignant cells), the staining pattern was usually focal cytoplasmic, although occasional examples of intense membranous and diffuse cytoplasmic staining were seen. Except for intense staining of the secretory epithelium of salivary glands, virtually no immunoperoxidase staining was present in cells of the surrounding tissue (blood vessels, stroma, skeletal muscle or lymphocytes). Statistical analyses The 44 patients were monitored to determine relapse and death rates. Seven patients were not included in the calculation of relapse rates because their tumours did not respond to initial therapy and consequently could not relapse (by definition). In total there were 28 patients who relapsed in 833 months of follow-up, and 30 patients who died in 1,096 months of follow-up. The median follow-up time for all the patients was 15.5 months, ranging from 3 to 120 months. Among the survivors, the shortest follow-up time was 16 months. Crude associations between variables The presence of nodal involvement was significantly associated with tumour sizes of more than 4 cm (p = 0.04). There was no statistical evidence to suggest that either nodal involvement or tumour size were associated with pathologic grade (p = 0.67 and 0.45, respectively). There were no statistically significant associations between quartiles of focal cytoplasmic or nil staining and tumour size (p = 0.76), nodal involvement (p = 0.98) or pathologic grade (p = 0.58). These data are presented in Table I. Kaplan-Meier survival rates Table 11 demonstrates 5-year disease-free and overall survival rates according to tumour size, degree of nodal involvement and histologic grade. Figure 2 demonstrates KaplanMeier survival curves according to percentage focal cytoplasmic or nil staining pattern in each sample. Multivariate analyses All variables were considered in their original ungrouped
TABLE I1 - FIVE-YEAR RECURRENCE-FREE AND OVERALL SURVIVAL F‘IGURES ACCORDING TO TUMOUR SIZE, NODAL INVOLVEMENT AND PATHOLOGIC GRADE
Parameter
Size 4cm Nodes: nil
involved Grade: well differentiated moderately differentiated poorly
differentiated
5-year disease-free survival
5-year swival
25% (10,45)’ 15% (3,39) 26% (10,45) 13% (2,34) 50% (16,77)
32% (14,52) 15% (13,34) 38% (19,58) 21% (6,42) 60% (20,85)
23% (8,42)
29% (13,47)
20% (6,61)
14% (0,47)
‘Parentheses indicate 95% confidence intervals.
‘
3 605040-
g 3 M
3020-
0 ’
7
14
21
1
35
2a
MONTHS
- t - t-I
-
10 0
I
9%(11-1)90-100
i
1$
$1
2a
35
MONTHS
FIGURE 2 - Kaplan-Meier survival curves for (a) recurrence-free and (b) overall survival according to percentage focal or nil staining pattern. form and in grouped form as above to allow for the testing of trends. Cox multiple regression modelling of these variables supported the descriptive observations about survival made from the Kaplan-Meier analyses. There were significant or nearly significant linear trends of increasing deaths with increasing tumour size (p = 0.07),
TABLE I - NUMBER OF SCC PATIENTS CROSS-CLASSIFIED BY QUARTILES OF FOCAWNIL STAINING AND GRADE, T AND N STAGES AND SURVIVAL
3y-1,
BF /nil
Grade
TS Stage
Status
N6 Stage
pattern
W*
M3
pl
TI2
T3.4
No
Ni4
Dead
Alive
040
4 3 2 1
7 7 6 7
1 2
8 7 4 7
4 5 5 4
6 7 5 6
6 5 4 5
4 9 7 10
8 3 2
41-70 7 1-90 91-100
1 3
‘Focal .-2WeIl differentiated.-3ModerateIy differentiated.-4Pmrly differentiated. -5Tumour stage.-6Nodal involvement
1
856
KEARSLEY ET A L .
nodal involvement (p = 0.05), and pathologic grade (p = 0.05). A multivariate model was defined which partitioned the components of variation due to each of these variables. To this was added the percentage of focal cytoplasmic or nil staining variable which was considered for any further independent contribution to the variation in death rates. A significant trend of increasing death rate with increasing percentage of focal cytoplasmic or nil staining was still demonstrated (p < 0.001). This trend can be readily observed in Table 111, where the adjusted relative death rates for the quartiles of the staining pattern are presented. Conversely, there was a significant trend of increasing survival with increasing percentage of membranous staining pattern (p = 0.004). The time taken for treated patients to relapse was analysed as an alternative end-point to death. Excluding the subjects for whom relapse data were not applicable, there were no significant associations between relapse rate and tumour size (p = 0.66) or nodal involvement (p = 0.73), and only a suggestion of a positive association with pathologic grade (p = 0.07). However, these variables were included in the model to ensure that the least-biased estimates of relapse rates were estimated for focal cytoplasmic or nil staining pattern. There was, after considering the contribution of these variables, a statistically significant trend of increasing relapse rate with increasing percentage of focal cytoplasmic or nil staining pattern (p < .001) (vide supra). There was also a significant converse trend of decreasing relapse rates with increasing of membranous staining pattern (p < 0.001). Patients with moderately differentiated SCCs (n = 27) were analysed separately by percentage focalhi1 staining pattern. After adjusting for the potential effects of tumour size and nodal involvement, there remained a statistically significant linear trend of increasing death rates with increasing focalhi1 staining (x' = 7.61, p = 0.006). Patients with moderately differentiated SCCs which had 91-100% of cells with focal or nil staining had an 8-fold greater mortality than those with 0 4 0 % focal or nil staining.
tive, poorly standardized and remains a relatively crude approximation of tumour kinetics at the all-important cellular level. Not only are the vast majority of SCCs graded as moderately differentiated by pathologists, but even some welldifferentiated SCCs (as in our study) may behave in an unexpectedly aggressive fashion. It was reassuring that increasing tumour size, nodal involvement and high pathologic grade are all strong predictors of disease-related death in our study. These clinico-pathologic features have previously been demonstrated to portend a gloomy clinical outlook (Davis, 1985), though correlation with disease relapse is often poor. The significant trends of increasing death and relapse rates with increasing percentage of focal or nil staining pattern is highly relevant to both pathologists and clinicians. For example, several of our well-differentiated SCCs demonstrated pronounced keratin whorling, but had focal cytoplasmic or non-reactive staining with 3H-1 MAb. All such patients either developed tumour recurrence soon after definitive surgery, or died from advanced local disease. In addition, many of our moderately differentiated SCCs could be further subdivided according to a predominant 3H-1 staining pattern. We suggest that the progression of 3H-1 staining reactivity from purely membranous through diffuse cytoplasmic, and focal cytoplasmic to complete absence of staining may be a better prognostic guideline for histopathologists in that the 3H- 1 staining pattern identifies sub-populations of cancer cells with unique biologic and growth characteristics, not currently detected by routine haematoxylin and eosin staining. Although the vast majority of our SCCs were derived from the upper aerodigestive tract, SCC samples examined from other anatomic sites, such as lung and cervix, followed exactly the same pattern of 3H-1 MAb reactivity. There have been a number of successful attempts at producing MAbs to SCC tissue (Brenner et al., 1982; Carey et al., 1983; Bernal and Speak, 1984; Boeheim et al., 1985; Kyoizumi et al., 1985; Stahel et a l . , 1985; Fernstein et al., 1986; Hanai et al., 1986; Kimmel and Carey, 1986; Gioanni et al., 1987; Myoken et al., 1987; Ranken et al., 1987). That few of these Mabs have found an important role to play in immunoDISCUSSION histochemistry reflects several shortcomings. Firstly, because Our findings confirm that MAb 3H-1 has a high specificity many are reactive only on frozen tissue sections, they are unfor SCCs and that a number of well-defined staining patterns likely to be used routinely in busy pathology departments and result when the antibody is applied to formalin-fixed tissue they are not applicable to large retrospective studies. Secondly, sections. It is now clear from our work with SCCs and squamo- a number react only with cytoplasmic epitopes, resulting in proliferative lesions that the staining pattern closely reflects varying degrees of cytoplasmic staining which must then be some functional activity of disordered squamous-cell behav- interpreted on a subjective scale of 0 to 3 + intensity to enable iour. Abnormal glycosylation of surface membrane proteins a discrimination to be made between tumours. Thirdly, most and lipids is considered to play an important role in aberrant react with normal epithelial cells (including normal squamous epithelial proliferation (Vigneswaran et a l . , 1990), and epithelia) in addition to reacting with other epithelial maligwhether MAb 3H- 1 recognizes abnormally glycosylated pro- nancies. Finally, few have demonstrated prognostic signifiteins or lipids, possibly of desmosomal origin (Vilela et d., cance in their ability to influence clinical decision-making. On the other hand, MAb 3H-1 is reactive on formalin-fixed sec1987), is the subject of further work in this laboratory. The results of our multivariate analyses have considerable tions and interpretation is not subjective as it is the pattern of practical importance in the clinical management of patients. It reactivity rather than the degree of staining intensity which is is already recognized that SCCs contain heterogeneous cell important. Moreover, only a very limited range of normal tispopulations which are not readily identified by standard histo- sues ( e . g . , upper gastrointestinal mucosa, salivary-gland epilogic criteria and that histomorphic grading of SCCs is subjec- thelium) and a very limited range of non-squamous malignancies produce consistent reactivity (Parsons et al., 1991). In conclusion, we describe a potentially important, novel TAR1,E Ill - RELATIVE -. RELAPSE AND DEATH RATES FOR ALL PATIENTS IgM MAb, 3H-1, which reacts with formalin-fixed tissues, is ACCORDING TO PERCENTAGE FOCAL CYTOPLASMIC OR NIL STAINING PARERN-ADJUSTED FOR TUMOUR SIZE, NODAL INVOLVEMENT AND highly specific for squamo-proliferative cells and whose patPATHOLOGIC GRADE tern of reactivity yields important prognostic information with Percentage which to plan optimal treatment for individual patients with Relative Relative focallnil death rate relapse rate SCC. staining ~~~~
0-40
41-70 7 1-90 91-100
~~
~
~~~~
1.0
2.3 (0.7, 7.9)' 3.5 (1.0, 12.6) 5.7 (1.6, 20.3)
'Figures in parentheses indicate 95% confidence intervals.
1.0
2.1 (0.6, 7.6) 6.1 (1.6, 23.3) 14.3 (3.6, 56.5)
ACKNOWLEDGEMENTS
We thank Mrs. J. Middleton for typing the manuscript. This work was supported by the Queensland Cancer Fund and J.H.K. was in receipt of a Research Fellowship from the Queensland Radium Institute.
PROGNOSTIC SIGNIFICANCE OF MAb 3H-1
857
REFERENCES ARMITAGE, P., Statistical methods in medical research. Blackwell, Oxford (1971). BERNAL, D. and SPEAK,J.A., Membrane antigen in small-cell carcinoma of the lung defined by monoclonal antibody SMI. Cancer Res., 44,256270 (1984). BOEHEIM, K., SPEAK,J.A., FREI,E. and BERNAL,S.D., SQMl antibody defines a surface membrane antigen in squamous carcinoma of the head and neck. Int. J. Cancer, 36, 137-142 (1985). BRENNER, B.G., JOTHY, S., SHUSTER,J. and FUKS, A., Monoclonal antibodies to human lung-tumour antigens demonstrated by immunofluorescence and immunoprecipitation. Cancer Res., 42, 3 187-3 192 (1982). BROWN, K.W. and GALLIMORE, P.H., Malignant progression of an SV40transformed epidermal keratinocyte cell line. Brit. J. Cancer, 56, 545-554 (1987). CANN,C.I., FRIED,M.P. and ROTHMAN, K.J., Epidemiology of squamow cell cancer of the head and neck. Otolaryng. Clin. N . Amer., 18, 367-388 (1985). CAREY, T.E., KIMMEL, K.A., SCHWARTZ, D.R., RICHTER, D.E., BAKER, S.R. and KRAUSE,C.J., Antibodies to human squamous-cell carcinoma. Otolaryng. Head Neck Surg., 91, 482491 (1983). Cox, D.R., Regression models and life tables. J . roy. stat. Soc., 34, 187-202 ( 1972). DAVIS,R.K., Prognostic variable in head and neck cancer. Otolaryng. Clin. N. Amer., 18, 411419 (1985). DESAI,P.B., Clinical cancer research in India: present status and future prospects. Ind. J. med. Res., 78, (Suppl.) 8 (1983). FERNSTEIN, P.D., PEKNY,K.W., REISFELD,R.A. and WALKER,L.E., Antigens associated with human squamous-cell lung carcinoma defined by murine monoclonal antibodies. Cancer, 46, 297&2977 (1986). GIOANNI, J., SAMSON, M., ZANGHELLINI, E., MAZEAU, C., ETTORE,F., DEMARD,F., CHAUVEL,P., DUPLAY,H., SCHNEIDER, M., LAURENT, J.C. and LALANNE, C.M., Characterization of a new surface epitope specific for human epithelial cells defined by a monoclonal antibody and application to tumour diagnosis. Cancer Res., 47, 44174424 (1987). HANAI,N., SHITARA,K. and YOSHIDA,H., Generation of monoclonal antibodies against human lung squamous-cell carcinoma and adenocarcinoma using mice rendered tolerant to normal human lung. Cancer Res., 46, 44384443 (1986).
KAPLAN,E.L. and MEIER,P., Non-parametric estimation from incomplete observations. J. Amer. stat. Ass., 53, 457481 (1958). KIMMEL,K.A. and CAREY,T.E., Altered expression in squamous carcinoma cells of an orientation-restricted epithelial antigen detected by monoclonal antibody A9. Cancer Res., 46,36163623 (1986). KOHLER,G. and MILSTEIN,C., Continuous cultures of fused cells secreting antibody of predefined specificity. Nature (Land.), 256, 495497 (1975). KYOIZUMI, S., AKIYAMA, M., KOUNO,N., KOBUKE,K., HAKODA, M., JONES,S.L. and YAMAKIDO, M., Monoclonal antibodies to human squamous-cell carcinoma of the lung and their application to tumour diagnosis. Cancer Res., 45, 3274-3281 (1985). MYOKEN,Y., MOROYAMA, T., MIYAUCHI,S., TAKADA, K. and NAMBA, M., Monoclonal antibodies against human oral squamous-cell carcinoma reacting with keratin proteins. Cancer, 60, 2927-2937 (1987). PARSONS, P.G., LEONARD,J.H., KEARSLEY, J.H., TAKAHASHI, H. and Moss, D.J., Characterisation of a novel monoclonal antibody, 3H-1, reactive with squamoproliferative lesions and squamous cell cancers. Int. J. Cancer, 47, This issue (1991). PINDBORG, J.J., Oral cancer from an international point of view. J. Cancer dent. Ass., 31, 219-226 (1965). RANKEN,R., WHITE, C.F., GOTTFRIED,T.G., YONKOVICH,S.J., BLAZEK,B.E., Moss, M.S., FEE, W.E., JR. and VICTORLIU, Y.S., Reactivity of monoclonal antibody 17.13 with human squamous-cell carcinoma and its application to tumour diagnosis. Cancer Res., 47,5684-5690 (1987). STAHEL,R.A., SPEAK,J.A. and BERNAL,S.D., Murine monoclonal antibody LAM2 defines cell membrane determinant with preferential expression on human lung small-cell carcinoma and squamous-cell carcinomas. Int. J. Cancer, 35, 11-17 (1985). VILELA,M.J., PARRISH,E.P., WRIGHT,D.H. and GARROD,D.R., Monoclonal antibody to desmosomal glycoprotein 1-a new epithelial marker for diagnostic pathology. J. Path., 153, 365-375 (1987). VIGNESWARAN, N., PETERS,K.P., HORNSTEIN, O.P. and DIEPGEN,T.L., Alteration of cell surface carbohydrates associated with ordered and disordered proliferation of oral epithelia: a lectin histochemical study in oral leukoplakias, papillomas and carcinomas. Cell Tiss. Kinet., 23, 41-55 (1990). ZARBO,R.J. and CRISSMAN, J.D., The surgical pathology of head and neck cancer. Semin. Oncol., 15, 10-19 (1988).
Publication of the International Union Against Cancer Publication de I’Union Internationale Contre le Cancer
PROGNOSTIC SIGNIFICANCE OF MONOCLONAL ANTIBODY 3H- 1 REACTIVITY WITH SQUAMOUS-CELL HEAD-AND-NECK CANCERS John H. KEARSLEY~, J. Helen LEONARD, Hiroyulu TAKAHASHI, Diana BATTISTUTTA, Peter G . PARSONS and Denis J. MOSS Queensland Institute of Medical Research, Bramston Terrace, Herston QLB 4006, Australia. We report the reactivity of a novel murine IgM monoclonal antibody (MAb), 3H-I, in formalin-fixed tissue from a series of 44 patients with squamous-cell cancers (SCCs) of the upper aerodigestive tract. The antigen detected by 3H-I MAb is not expressed in simple squamous epithelia and only rarely in non-squamous malignancies. The following staining patterns were found to be associated with a progressively unfavourable prognosis: membranous, diffuse cytoplasmic and focal cytoplasmidnil staining. There were no statistically-significant associations between focal cytoplasmidnil staining and traditionally used prognostic parameters such as tumour size, nodal involvement or pathologic grade. However, multivariate analyses demonstrated that focal or nil staining was a significant independent prognostic factor for survival (p < 0.001) and was the only significant prognostic factor for relapse (p < 0.001). The converse applied to tumours with predominantly membranous staining ( p = 0.004). 3H-I MAb, with the advantage of relying on the pattern rather than the intensity of staining, makes it possible to identify different cellular phenotypic sub-populations more accurately than by conventional means, and may ultimately improve our understanding of the mechanisms underlying transformation from a benign or proliferative state to a malignant one.
Squamous-cell cancers (SCCs) of the upper aerodigestive tract represent 5-10% of newly diagnosed cancers in the Westem world (Pindborg, 1965) although their incidence is highest among individuals who maintain a heavy consumption of alcohol and cigarettes (Cann et al., 1985). In some countries, head-and-neck SCCs represent the most common form of malignant disease (Desai, 1983). Traditional therapy with surgery and/or radiotherapy can be curative in selected cases. However, early and late morbidity from the functional, cosmetic and socio-economic impact of treatment can significantly reduce a patient’s quality of life. Furthermore, incurable disease following treatment almost invariably remains localized in the head and neck region, so that terminal suffering is protracted and distressing. Prognostic information is essential for the evaluation, judgement and optimal treatment of patients with SCC of the upper aerodigestive tract. Although routine histomorphologic grading systems and light microscopic appearances are of some benefit in predicting clinical outcome (Zarbo and Crissman, 1988), such criteria remain subjective and poorly standardized, and must be regarded as relatively crude approximations of biologic behaviour at the all-important cellular level. The development of hybridoma technology (Kohler and Milstein, 1975) has provided a tool for a more accurate assessment of the structural and functional changes which occur in malignant disease. We have described, in a companion report (Parsons et al., 1991), the derivation and characterization of a novel IgM MAb 3H-1 which has a high specificity for squamous-cell disorders and which reacts with formalin-fixed tissues. MAb 3H-1 appears to produce at least 3 well-defined patterns of reactivity in SCCs and in squamo-proliferative lesions over a range of anatomic sites. The present report is an extension of our previous work and aims to determine whether the pattern of MAb 3H-1 reactivity carries prognostic significance in a series of formalin-fixed tissue sections of headand-neck SCCs.
MATERIAL AND METHODS
Tissue specimens Specimens were obtained from 44 untreated patients with SCCs from a range of anatomic sites within the upper aerodigestive tract (oral cavity 28, pharynx 10, larynx 6 ) . Normal squamous epithelium was present in the same section in many cases. There were 31 males and 13 females with a mean age of 59.2 years (range 41-79 years). All specimens had been fixed in buffered formaldehyde solution and routinely embedded in paraffin wax. Serial 4-pm-thick sections were cut from each block, and one slide was stained with haematoxylin and eosin for routine histopathologic assessment and grading. When keratinization with squamous epithelial pearls was predominant, the tumour was graded as well-differentiated. When keratinization was only present in some areas of the tumour, it was graded as poorly differentiated. When there appeared to be a fairly even mixture of keratinizing and non-keratinizing areas, the tumour was graded as moderately well differentiated. Some keratinization was present in all of the tumours studied. Monoclonal antibody MAb 3H-1 was derived as described by Parsons et al. (1991) from a fusion of BALB/c mouse splenocytes immunized with UV-irradiated keratinocytes which had previously been transformed with SV40 virus (KJD-USV40 cell line) (Brown and Gallimore, 1987). 3H-1 is an IgM MAb which recognizes an antigen with a molecular weight of approximately 55 kDa. Immunohistochemistry Immunohistochemical staining was performed using the avidin-biotin-peroxidase (ABC) technique (PK-4005; Vector, Burlingame, CA). Briefly, tissue sections were deparaffinized, dehydrated in 100% alcohol and then rehydrated through decreasing alcohol concentrations to water. Sections were treated with 0.3% (w/v) hydrogen peroxide and 20% (w/v) methanol in phosphate-buffered saline (5 m~ Na,HPO,, 3 m~ KH,PO,, 145 m~ NaCl pH7.2; PBS) for 10 min to eliminate endogenous peroxidase activity. Sections were then incubated with diluted horse serum for 20 min to eliminate non-specific binding of immunoglobulins. After removing excess horse serum, 3H- 1 MAb was added to each section as a supernatant and the slides were incubated for 30 min at room temperature. After several washes in PBS, sections were incubated with diluted biotinylated horse anti-mouse IgM immunoglobulin for 30 min at room temperature. The slides were then washed in PBS and incubated for 45 min with the Vectastain reagent, according to the manufacturer’s directions. After a further wash in PBS, sections were incubated for 3-10 min with 3amino-8-ethylcarbazole to which 10 drops of 30% H2,02had been added. The sections were finally counterstained with haemalum, rinsed and mounted under a cover-slip with Glycergel (DAKO, Carpinteria, CA). ‘To whom correspondence and reprint requests should be addressed. Received: November 5 , 1990.
854
KEARSLEY ET AL.
All immunoperoxidase procedures included both positive and negative control slides. The positive control was a formalin-fixed section from an SCC of the tongue which consistently showed strong immunoperoxidase staining with 3H- 1 antibody. Negative controls were incubated with PBS or an irrelevant murine IgM antibody (reactive with formalin-fixed material) instead of 3H-1 antibody in the first step and were then treated as described above. Immunohistochemical evaluation
Initial assessment was made at low microscopic power to examine the distribution of squamous cells and to ascertain whether there were any obvious gross variations in staining. Multiple sites from each SCC were studied and the distribution pattern of reaction product within positive cells was recorded as purely membranous, diffuse cytoplasmic, focal cytoplasmic or nil staining. The percentage of squamous cancer cells showing each type of immunoreactivity or no immunoreactivity was estimated as a percentage of the total number of malignant squamous cells seen in each section (Fig. 1). Statistical methods
Associations between categorical variables (T, N stage; pathologic grade; patterns of 3H- 1 staining; relapse-free and overall survivals) were investigated using chi-squared tests with Yates’ correction where appropriate (Armitage, 1971). Univariate differences in relapse and death rates over time according to these variables were considered using Kaplan and Meier (1958) survival analyses. Cox (1972) proportional haz-
ards regression was used in multivariate analyses to investigate the independent contributions of the various prognostic factors. All analyses were performed using the EGRET (Statistics and Epidemiology Research Corporation, Seattle, WA) and SPSSPC+ (SPSS, Chicago, IL) package on a microcomputer. In exploring for associations between the variables and with overall or recurrence-free survival, due to a lack of numbers in the extreme categories of the prognostic variables, tumour size was grouped as less than 4cm vs. greater than 4cm; nodal involvement was grouped as NO vs. N1-4, and pathologic grade grouped as well differentiated vs. moderately/poorly differentiated. Patients were grouped according to the percentage of tumour cells with focal cytoplasmic or nil staining. The quartiles of the distribution were W O % , 41-70%, 71-90% and 91-100%. RESULTS
Of the 44 SCC samples, 10 (23%) were graded as welldifferentiated, 27 (61%) were moderately differentiated, and 7 (16%) were poorly differentiated. Seventy-three percent of patients (32/44) had tumours measuring 2-6 cm. More than half (24 patients) had no nodal involvement. Three distinct patterns of MAb 3H- 1 reactivity were seen in this series and most tissue sections demonstrated varying proportions of cells with each pattern. Multiple sections from each SCC were screened and the results were consistently reproducible. The first pattern (membranous) was strictly limited to the cell membrane, (Fig. l a ) ; in the second (diffuse cytoplasmic)
FIGURE1 - Formalin-fixed SCC samples demonstrating the 3 patterns of 3H-1 reactivity (a) membranous, bar = 100 pm; (b) diffuse cytoplasmic, bar = 50 pm; (c) focal cytoplasmic, bar = 50 pm; ( d ) focal cytoplasmic (high-power), bar = 50 pm.
855
PROGNOSTIC SIGNIFICANCE OF MAb 3H-1
the membranous component was slightly less prominent but there was an additional diffusely granular cytoplasmic staining reaction, (Fig. lb). The third pattern of 3H-1 reactivity (focal cytoplasmic) consisted of focal, granular cytoplasmic deposits often situated in a paranuclear location, (Fig. lc,d). Some SCC cells did not stain at all (nil staining). The pattern of 3H-1 reactivity bore a close relationship to the architecture of individual tumours. In the tumours which invaded as cellular islands there was usually no reactivity in the peripheral cells, Adjacent, more centrally located cells in these islands usually demonstrated a focal cytoplasmic reactivity and the most central cells usually exhibited a membranous pattern (Fig. l a ) . However, the width of each reactive “layer” of cells could vary widely between individual tumours, from cellular islands composed totally of membranous reactivity to multicellular layers of nil or focal cytoplasmic reactivity (Fig. l a ) . There was no staining of keratin whorls when these structures were located in the centre of invading cellular islands. When squamous cancer cells were seen to invade in a diffuse fashion (in either small clumps or as linear streams of malignant cells), the staining pattern was usually focal cytoplasmic, although occasional examples of intense membranous and diffuse cytoplasmic staining were seen. Except for intense staining of the secretory epithelium of salivary glands, virtually no immunoperoxidase staining was present in cells of the surrounding tissue (blood vessels, stroma, skeletal muscle or lymphocytes). Statistical analyses The 44 patients were monitored to determine relapse and death rates. Seven patients were not included in the calculation of relapse rates because their tumours did not respond to initial therapy and consequently could not relapse (by definition). In total there were 28 patients who relapsed in 833 months of follow-up, and 30 patients who died in 1,096 months of follow-up. The median follow-up time for all the patients was 15.5 months, ranging from 3 to 120 months. Among the survivors, the shortest follow-up time was 16 months. Crude associations between variables The presence of nodal involvement was significantly associated with tumour sizes of more than 4 cm (p = 0.04). There was no statistical evidence to suggest that either nodal involvement or tumour size were associated with pathologic grade (p = 0.67 and 0.45, respectively). There were no statistically significant associations between quartiles of focal cytoplasmic or nil staining and tumour size (p = 0.76), nodal involvement (p = 0.98) or pathologic grade (p = 0.58). These data are presented in Table I. Kaplan-Meier survival rates Table 11 demonstrates 5-year disease-free and overall survival rates according to tumour size, degree of nodal involvement and histologic grade. Figure 2 demonstrates KaplanMeier survival curves according to percentage focal cytoplasmic or nil staining pattern in each sample. Multivariate analyses All variables were considered in their original ungrouped
TABLE I1 - FIVE-YEAR RECURRENCE-FREE AND OVERALL SURVIVAL F‘IGURES ACCORDING TO TUMOUR SIZE, NODAL INVOLVEMENT AND PATHOLOGIC GRADE
Parameter
Size 4cm Nodes: nil
involved Grade: well differentiated moderately differentiated poorly
differentiated
5-year disease-free survival
5-year swival
25% (10,45)’ 15% (3,39) 26% (10,45) 13% (2,34) 50% (16,77)
32% (14,52) 15% (13,34) 38% (19,58) 21% (6,42) 60% (20,85)
23% (8,42)
29% (13,47)
20% (6,61)
14% (0,47)
‘Parentheses indicate 95% confidence intervals.
‘
3 605040-
g 3 M
3020-
0 ’
7
14
21
1
35
2a
MONTHS
- t - t-I
-
10 0
I
9%(11-1)90-100
i
1$
$1
2a
35
MONTHS
FIGURE 2 - Kaplan-Meier survival curves for (a) recurrence-free and (b) overall survival according to percentage focal or nil staining pattern. form and in grouped form as above to allow for the testing of trends. Cox multiple regression modelling of these variables supported the descriptive observations about survival made from the Kaplan-Meier analyses. There were significant or nearly significant linear trends of increasing deaths with increasing tumour size (p = 0.07),
TABLE I - NUMBER OF SCC PATIENTS CROSS-CLASSIFIED BY QUARTILES OF FOCAWNIL STAINING AND GRADE, T AND N STAGES AND SURVIVAL
3y-1,
BF /nil
Grade
TS Stage
Status
N6 Stage
pattern
W*
M3
pl
TI2
T3.4
No
Ni4
Dead
Alive
040
4 3 2 1
7 7 6 7
1 2
8 7 4 7
4 5 5 4
6 7 5 6
6 5 4 5
4 9 7 10
8 3 2
41-70 7 1-90 91-100
1 3
‘Focal .-2WeIl differentiated.-3ModerateIy differentiated.-4Pmrly differentiated. -5Tumour stage.-6Nodal involvement
1
856
KEARSLEY ET A L .
nodal involvement (p = 0.05), and pathologic grade (p = 0.05). A multivariate model was defined which partitioned the components of variation due to each of these variables. To this was added the percentage of focal cytoplasmic or nil staining variable which was considered for any further independent contribution to the variation in death rates. A significant trend of increasing death rate with increasing percentage of focal cytoplasmic or nil staining was still demonstrated (p < 0.001). This trend can be readily observed in Table 111, where the adjusted relative death rates for the quartiles of the staining pattern are presented. Conversely, there was a significant trend of increasing survival with increasing percentage of membranous staining pattern (p = 0.004). The time taken for treated patients to relapse was analysed as an alternative end-point to death. Excluding the subjects for whom relapse data were not applicable, there were no significant associations between relapse rate and tumour size (p = 0.66) or nodal involvement (p = 0.73), and only a suggestion of a positive association with pathologic grade (p = 0.07). However, these variables were included in the model to ensure that the least-biased estimates of relapse rates were estimated for focal cytoplasmic or nil staining pattern. There was, after considering the contribution of these variables, a statistically significant trend of increasing relapse rate with increasing percentage of focal cytoplasmic or nil staining pattern (p < .001) (vide supra). There was also a significant converse trend of decreasing relapse rates with increasing of membranous staining pattern (p < 0.001). Patients with moderately differentiated SCCs (n = 27) were analysed separately by percentage focalhi1 staining pattern. After adjusting for the potential effects of tumour size and nodal involvement, there remained a statistically significant linear trend of increasing death rates with increasing focalhi1 staining (x' = 7.61, p = 0.006). Patients with moderately differentiated SCCs which had 91-100% of cells with focal or nil staining had an 8-fold greater mortality than those with 0 4 0 % focal or nil staining.
tive, poorly standardized and remains a relatively crude approximation of tumour kinetics at the all-important cellular level. Not only are the vast majority of SCCs graded as moderately differentiated by pathologists, but even some welldifferentiated SCCs (as in our study) may behave in an unexpectedly aggressive fashion. It was reassuring that increasing tumour size, nodal involvement and high pathologic grade are all strong predictors of disease-related death in our study. These clinico-pathologic features have previously been demonstrated to portend a gloomy clinical outlook (Davis, 1985), though correlation with disease relapse is often poor. The significant trends of increasing death and relapse rates with increasing percentage of focal or nil staining pattern is highly relevant to both pathologists and clinicians. For example, several of our well-differentiated SCCs demonstrated pronounced keratin whorling, but had focal cytoplasmic or non-reactive staining with 3H-1 MAb. All such patients either developed tumour recurrence soon after definitive surgery, or died from advanced local disease. In addition, many of our moderately differentiated SCCs could be further subdivided according to a predominant 3H-1 staining pattern. We suggest that the progression of 3H-1 staining reactivity from purely membranous through diffuse cytoplasmic, and focal cytoplasmic to complete absence of staining may be a better prognostic guideline for histopathologists in that the 3H- 1 staining pattern identifies sub-populations of cancer cells with unique biologic and growth characteristics, not currently detected by routine haematoxylin and eosin staining. Although the vast majority of our SCCs were derived from the upper aerodigestive tract, SCC samples examined from other anatomic sites, such as lung and cervix, followed exactly the same pattern of 3H-1 MAb reactivity. There have been a number of successful attempts at producing MAbs to SCC tissue (Brenner et al., 1982; Carey et al., 1983; Bernal and Speak, 1984; Boeheim et al., 1985; Kyoizumi et al., 1985; Stahel et a l . , 1985; Fernstein et al., 1986; Hanai et al., 1986; Kimmel and Carey, 1986; Gioanni et al., 1987; Myoken et al., 1987; Ranken et al., 1987). That few of these Mabs have found an important role to play in immunoDISCUSSION histochemistry reflects several shortcomings. Firstly, because Our findings confirm that MAb 3H-1 has a high specificity many are reactive only on frozen tissue sections, they are unfor SCCs and that a number of well-defined staining patterns likely to be used routinely in busy pathology departments and result when the antibody is applied to formalin-fixed tissue they are not applicable to large retrospective studies. Secondly, sections. It is now clear from our work with SCCs and squamo- a number react only with cytoplasmic epitopes, resulting in proliferative lesions that the staining pattern closely reflects varying degrees of cytoplasmic staining which must then be some functional activity of disordered squamous-cell behav- interpreted on a subjective scale of 0 to 3 + intensity to enable iour. Abnormal glycosylation of surface membrane proteins a discrimination to be made between tumours. Thirdly, most and lipids is considered to play an important role in aberrant react with normal epithelial cells (including normal squamous epithelial proliferation (Vigneswaran et a l . , 1990), and epithelia) in addition to reacting with other epithelial maligwhether MAb 3H- 1 recognizes abnormally glycosylated pro- nancies. Finally, few have demonstrated prognostic signifiteins or lipids, possibly of desmosomal origin (Vilela et d., cance in their ability to influence clinical decision-making. On the other hand, MAb 3H-1 is reactive on formalin-fixed sec1987), is the subject of further work in this laboratory. The results of our multivariate analyses have considerable tions and interpretation is not subjective as it is the pattern of practical importance in the clinical management of patients. It reactivity rather than the degree of staining intensity which is is already recognized that SCCs contain heterogeneous cell important. Moreover, only a very limited range of normal tispopulations which are not readily identified by standard histo- sues ( e . g . , upper gastrointestinal mucosa, salivary-gland epilogic criteria and that histomorphic grading of SCCs is subjec- thelium) and a very limited range of non-squamous malignancies produce consistent reactivity (Parsons et al., 1991). In conclusion, we describe a potentially important, novel TAR1,E Ill - RELATIVE -. RELAPSE AND DEATH RATES FOR ALL PATIENTS IgM MAb, 3H-1, which reacts with formalin-fixed tissues, is ACCORDING TO PERCENTAGE FOCAL CYTOPLASMIC OR NIL STAINING PARERN-ADJUSTED FOR TUMOUR SIZE, NODAL INVOLVEMENT AND highly specific for squamo-proliferative cells and whose patPATHOLOGIC GRADE tern of reactivity yields important prognostic information with Percentage which to plan optimal treatment for individual patients with Relative Relative focallnil death rate relapse rate SCC. staining ~~~~
0-40
41-70 7 1-90 91-100
~~
~
~~~~
1.0
2.3 (0.7, 7.9)' 3.5 (1.0, 12.6) 5.7 (1.6, 20.3)
'Figures in parentheses indicate 95% confidence intervals.
1.0
2.1 (0.6, 7.6) 6.1 (1.6, 23.3) 14.3 (3.6, 56.5)
ACKNOWLEDGEMENTS
We thank Mrs. J. Middleton for typing the manuscript. This work was supported by the Queensland Cancer Fund and J.H.K. was in receipt of a Research Fellowship from the Queensland Radium Institute.
PROGNOSTIC SIGNIFICANCE OF MAb 3H-1
857
REFERENCES ARMITAGE, P., Statistical methods in medical research. Blackwell, Oxford (1971). BERNAL, D. and SPEAK,J.A., Membrane antigen in small-cell carcinoma of the lung defined by monoclonal antibody SMI. Cancer Res., 44,256270 (1984). BOEHEIM, K., SPEAK,J.A., FREI,E. and BERNAL,S.D., SQMl antibody defines a surface membrane antigen in squamous carcinoma of the head and neck. Int. J. Cancer, 36, 137-142 (1985). BRENNER, B.G., JOTHY, S., SHUSTER,J. and FUKS, A., Monoclonal antibodies to human lung-tumour antigens demonstrated by immunofluorescence and immunoprecipitation. Cancer Res., 42, 3 187-3 192 (1982). BROWN, K.W. and GALLIMORE, P.H., Malignant progression of an SV40transformed epidermal keratinocyte cell line. Brit. J. Cancer, 56, 545-554 (1987). CANN,C.I., FRIED,M.P. and ROTHMAN, K.J., Epidemiology of squamow cell cancer of the head and neck. Otolaryng. Clin. N . Amer., 18, 367-388 (1985). CAREY, T.E., KIMMEL, K.A., SCHWARTZ, D.R., RICHTER, D.E., BAKER, S.R. and KRAUSE,C.J., Antibodies to human squamous-cell carcinoma. Otolaryng. Head Neck Surg., 91, 482491 (1983). Cox, D.R., Regression models and life tables. J . roy. stat. Soc., 34, 187-202 ( 1972). DAVIS,R.K., Prognostic variable in head and neck cancer. Otolaryng. Clin. N. Amer., 18, 411419 (1985). DESAI,P.B., Clinical cancer research in India: present status and future prospects. Ind. J. med. Res., 78, (Suppl.) 8 (1983). FERNSTEIN, P.D., PEKNY,K.W., REISFELD,R.A. and WALKER,L.E., Antigens associated with human squamous-cell lung carcinoma defined by murine monoclonal antibodies. Cancer, 46, 297&2977 (1986). GIOANNI, J., SAMSON, M., ZANGHELLINI, E., MAZEAU, C., ETTORE,F., DEMARD,F., CHAUVEL,P., DUPLAY,H., SCHNEIDER, M., LAURENT, J.C. and LALANNE, C.M., Characterization of a new surface epitope specific for human epithelial cells defined by a monoclonal antibody and application to tumour diagnosis. Cancer Res., 47, 44174424 (1987). HANAI,N., SHITARA,K. and YOSHIDA,H., Generation of monoclonal antibodies against human lung squamous-cell carcinoma and adenocarcinoma using mice rendered tolerant to normal human lung. Cancer Res., 46, 44384443 (1986).
KAPLAN,E.L. and MEIER,P., Non-parametric estimation from incomplete observations. J. Amer. stat. Ass., 53, 457481 (1958). KIMMEL,K.A. and CAREY,T.E., Altered expression in squamous carcinoma cells of an orientation-restricted epithelial antigen detected by monoclonal antibody A9. Cancer Res., 46,36163623 (1986). KOHLER,G. and MILSTEIN,C., Continuous cultures of fused cells secreting antibody of predefined specificity. Nature (Land.), 256, 495497 (1975). KYOIZUMI, S., AKIYAMA, M., KOUNO,N., KOBUKE,K., HAKODA, M., JONES,S.L. and YAMAKIDO, M., Monoclonal antibodies to human squamous-cell carcinoma of the lung and their application to tumour diagnosis. Cancer Res., 45, 3274-3281 (1985). MYOKEN,Y., MOROYAMA, T., MIYAUCHI,S., TAKADA, K. and NAMBA, M., Monoclonal antibodies against human oral squamous-cell carcinoma reacting with keratin proteins. Cancer, 60, 2927-2937 (1987). PARSONS, P.G., LEONARD,J.H., KEARSLEY, J.H., TAKAHASHI, H. and Moss, D.J., Characterisation of a novel monoclonal antibody, 3H-1, reactive with squamoproliferative lesions and squamous cell cancers. Int. J. Cancer, 47, This issue (1991). PINDBORG, J.J., Oral cancer from an international point of view. J. Cancer dent. Ass., 31, 219-226 (1965). RANKEN,R., WHITE, C.F., GOTTFRIED,T.G., YONKOVICH,S.J., BLAZEK,B.E., Moss, M.S., FEE, W.E., JR. and VICTORLIU, Y.S., Reactivity of monoclonal antibody 17.13 with human squamous-cell carcinoma and its application to tumour diagnosis. Cancer Res., 47,5684-5690 (1987). STAHEL,R.A., SPEAK,J.A. and BERNAL,S.D., Murine monoclonal antibody LAM2 defines cell membrane determinant with preferential expression on human lung small-cell carcinoma and squamous-cell carcinomas. Int. J. Cancer, 35, 11-17 (1985). VILELA,M.J., PARRISH,E.P., WRIGHT,D.H. and GARROD,D.R., Monoclonal antibody to desmosomal glycoprotein 1-a new epithelial marker for diagnostic pathology. J. Path., 153, 365-375 (1987). VIGNESWARAN, N., PETERS,K.P., HORNSTEIN, O.P. and DIEPGEN,T.L., Alteration of cell surface carbohydrates associated with ordered and disordered proliferation of oral epithelia: a lectin histochemical study in oral leukoplakias, papillomas and carcinomas. Cell Tiss. Kinet., 23, 41-55 (1990). ZARBO,R.J. and CRISSMAN, J.D., The surgical pathology of head and neck cancer. Semin. Oncol., 15, 10-19 (1988).
