JOURNAL OF PATHOLOGY, VOL.
164: 107- 1 17 (1 99 1)
GASTROINTESTINAL STROMAL TUMOURS: CORRELATION OF IMMUNOPHENOTYPE WITH CLINICOPATHOLOGICAL FEATURES P. L. NEWMAN*, C. WADDEN AND
c. D.
M. FLETCHER
Department of Histopathology, St Thomas's Hospital (U.M.D.S.), London, U.K.; *Department of Histopathology, University Hospital of Wales, Cardiff, U.K. Received 25 October 1990 Accepted23 November 1990
SUMMARY Stromal tumours of the gastrointestinal tract remain a persistent source of controversy with regard to both their proposed lines of differentiation and the difficulty in predicting their biological behaviour. We have examined 60 cases with a panel of seven antibodies directed at the identification of smooth muscular or neural differentiation. In our hands, 36 per cent of cases showed neural differentiation (although only 6.6 per cent expressed S-100 protein); 31 per cent appeared smooth muscular; 20 per cent manifested bidirectional differentiation; and 13 per cent were negative for all the markers used. Histological appearances do not reliably reflect immunophenotype. We have attempted to correlate immunophenotype with the site of the lesions and with their clinical behaviour. Mean follow-up of 5 years was obtained in 42 cases. Turnours with a neural phenotype have the best prognosis. Gastric tumours expressing both desmin and smooth muscle actin (in the absence of other markers) with up to 4 mitoses per 30 HPF behave in a benign fashion. Larger studies are required to substantiate the value of immunophenotyping this complex group of turnours. KEY WORDS-stomach,
intestine, leiomyoma, leiomyosarcoma, schwannorna, immunohistochemistry.
those with diploid lesions, but even so, some with diploid lesions died of their disease. Since 1941 when Golden a n d Stout' published In addition to the problem of biological behavthe first large series of gastrointestinal (GI) iour, the histogenesis of these tumours has also been smooth muscle tumours indicating that histological the subject of several studies, with widely differing appearance did not correlate well with biological results. Ail gastrointestinal stromal tumours were behaviour, several studies have addressed them- originally considered to be of smooth muscle origin, selves to the problem of assessing prognosis in this but while electron microscopy has been reported to group of turnours.*-" Initially, these concentrated show evidence of smooth muscle differentiation in on histological features, particularly mitotic counts, some series,"." it purports to show nerve sheath cellularity, and anaplasia, as well as size and differentiation in others." However, in many location. Some degree of correlation was found, but cases it appears that only scattered cells show differsome of the features assessed are very subjective and entiated features, the bulk of the tumour being subject to considerable inter-observer variation.'* undifferentiated, a n d the possibility of entrapment Later, flow c y t ~ m e t r i c " ~ 'and ~ m o r p h o m e t r i ~ ' ~ of residual normal tissue components within the techniques were applied in a n attempt to establish tumour cannot be excluded. a n objective measure of prognosis. These showed, Immunohistochemical studies have also produced not surprisingly, that patients with aneuploid diametrically opposed views as to the nature of these tumours had a shorter average survival time than tumours. Largely based on S- 100 protein a n d desAddressee for correspondence: Dr C. D. M. Fletcher, min immunoreactivity, some groups have favoured smooth muscle differentiation in the majority of Department of Histopathology, St Thomas's Hospital, London case^,^^.*^ some neura1,'8.2' and Pike et a1.22found SEI 7EH, U.K. INTRODUCTION
0022-3417/91/050107-11 $05.50 0 1991 by John Wiley & Sons, Ltd
108
P. L. NEWMAN ET AL.
Table I-Antibodies
used in this study of gastrointestinal stromal tumours Monoclonal polyclonal
Antibody Desmin Smooth muscle actin (IA4) (SMA)
M M
PGP9.5 Neuron-specificenolase (NSE) Glial fibrillary acidic protein (GFAP) Leu I
P P P P M
s-100
Source Dako Sigma Dako Ultraclone Dako Dako Becton Dickinson
Dilution 1 /SO
1/16 000 1j2000 lj800 1/SO0 1 /SO0
1/10
Table 11-Histological criteria for grading gastrointestinal stromal tumours Benign
or Borderline or or
0-2 mitoses per 30 HPF 0 mitoses per 30 HPF
Spindle cell lesion, no atypia Epithelioid lesion
2-3 mitoses per 30 HPF
Spindle cell lesion, mild pleomorphism/ hyperchromasia Spindle cell lesion, no atypia Epithelioid lesion
3 4 mitoses per 30 HPF 1 mitosis per 30 HPF
or
> 5 mitoses per 30 HPF > 3 mitoses per 30 HPF
or
> 2 mitoses per 30 HPF
Malignant
Spindle cell lesion, no atypia Spindle cell lesion, frank pleomorphism/ hyperchromasia Epithelioid lesion
I HPF = 0.1 59 mm2on the microscopes used
different immunophenotypes in different regions of the bowel. It is therefore clear that gastrointestinal stromal tumours are poorly categorized. Possibly they represent a heterogeneous group and this may explain why it is so difficult to adopt a single set of morphological criteria of malignancy with any degree of confidence. In this retrospective study we have therefore attempted to correlate immunophenotype, as defined by a panel of antibodies, with prognosis, in particular trying to identify tumours with features of neural differentiation, and possibly identify good or bad prognostic groups. MATERIALS AND METHODS All cases indexed as leiomyoma, leiomyosarcoma, leiomyoblastoma, schwannoma, or neurofibroma within the gastrointestinal tract were retrieved from the files of the histopathology department of St Thomas’s Hospital, producing a total of 66 cases.
Lesions in the oesophagus were excluded as the pathology of these is well described.23 Fresh 4 pm sections were cut of all the formalinfixed, paraffin-embedded blocks available, and a representative block was selected from each case for immunohistochemical staining. In two cases, two blocks were selected as the tumours showed a range of histological appearances. Immunohistochemistry was performed using the avidinbiotin-peroxidase (ABC) technique, employing the antibodies listed in Table I. Appropriate controls were used throughout. The tumours were assessed on the basis of their haematoxylin and eosin appearance, and categorized as benign, malignant, or borderline using the criteria set out in Table 11, and without knowledge of the clinical outcome. Mitoses were counted in 30 random high-power fields (where 1 HPF=0.159mm2) taken from the block which subjectively appeared the most cellular and mitotic. No attempt was made to differentiate between smooth muscle and neural tumours at this stage. The immunohistochemical results were
GI STROMAL TUMOURS
assessed by each author independently and without knowledge of the other results. There was good concordance between the observers in the assessment of positive or negative staining. Clinical follow-up was obtained by examination of patients’ hospital notes and by contacting general practitioners. Five cases were excluded on the grounds of supplementary immunohistochemistry and additional clinical information (four were metastases and one was a mesenteric tumour) and a further case from the 1950s was deleted owing to inadequate tissue preservation. This left 60 cases for study.
109
although basically of spindle cell morphology, the nuclei were rather plumper than the majority of spindle cell lesions, and there was focal nuclear palisading, with a rather syncytial, plexiform appearance in areas, and with some cytoplasmic vacuolation and degeneration of the ground substance. All nine of these lesions were located in the stomach, and none appeared histologically malignant. Of the 11 classified as malignant, 6 died of tumour; 4 were lost to follow-up; and one died 5 years post-operatively of an unknown cause. Immunohistochemistry
RESULTS Clinical Of the 60 cases included in the study, adequate follow-up was available on 42, ranging from an early tumour related death at 8 months to 19 years’ disease-free survival (mean follow-up 5 years). Thirty-one patients were female, 27 male, and in two the gender was not recorded. The ages ranged from 35 to 85 years (mean 62 years). Three patients suffered from von Recklinghausen’s disease. In 45 cases (75 per cent), the tumours were located in the stomach; 1 1 (1 8 per cent) were in the small bowel (4 duodenum, 2 jejunum, 5 unspecified); and 4 were from the rectum. Of the 42 cases with follow-up, 6 (14.25 per cent) died of their disease (4 small bowel tumours, 2 stomach). Histology On the basis of the criteria in Table 11, 11 cases were classified as malignant, 4 as borderline, and 45 as benign. In practice, on the basis of size, 24 of the ‘benign’ lesions would have had a rider on the histology report expressing uncertainty as to the likely behaviour of GI stromal tumours. Eight tumours were epithelioid; four showed an even mixture of epithelioid and spindle cell areas; and the remaining 48 cases had a predominantly spindle cell morphology. When subsequently correlating morphology with immunohistochemistry, it was abundantly clear that no histological features, such as nuclear morphology, palisading, or cytoplasmic eosinophilia, could distinguish spindle cell lesions with a neural or smooth muscle phenotype. There were nine tumours that were hard to classify into spindle or epithelioid types, in that
The immunophenotypes of the tumours are recorded in Table 111. Nineteen (3 1 per cent) cases showed evidence of smooth muscle differentiation as defined by staining solely with desmin or smooth muscle actin (SMA) antibodies (muscle tumours) (Fig. I). Twenty-two (36 per cent) showed neural differentiation (neural tumours) (Fig. 2) and eight (13 per cent) were negative with all antibodies. Twelve tumours (20 per cent) reacted with both smooth muscle and neural markers other than S- 100 (bidirectional tumours) (Fig. 3). No tumour expressed a combination of S-100 and muscle markers. All nine of the syncytial plexiform, palisaded lesions stained with neural markers (NSE and/or PGP) only (Fig. 4). Only a single case each stained with Leu-7 (also desmin/SMA positive) and GFAP (also S-lOO/NSE/PGP 9.5 positive). Of the eight cases which were entirely negative, despite good internal controls, four were spindle-celled and two each were epithelioid or of mixed type. Many tumours showed scattered cells reactive with S-100 or SMA, which tended to be located at the periphery of the lesion and were interpreted as entrapped normal structures (see Discussion). Correlation of immunophenotype and site
Of the tumours showing smooth muscle differentiation, the distribution within the gut was similar to the overall distribution of lesions, with 14 cases (73.6 per cent) located in the stomach and five cases (26.3 per cent) in the small bowel. More striking was the distribution of neural tumours. A11 tumours in which S-100 was identified were situated outside the stomach, while 17 of the 18 (94 per cent) tumours staining for NSE and/or PGP 9.5 in the absence of S-100 were situated within the stomach. The totally immunonegative and bidirectional differentiation
110
P. L. NEWMAN ET AL.
Table 111-Immunophenotyping
Neural tumours (22 cases)
Muscle tumours ( I 9 cases) DES and SMA DES only SMA only
of 60 gastrointestinal stromal tumours
S-lOO,”SE/PGP 9.5 S-lOO/NSE PGP9.5/NSE PGP 9.5 only NSE only
14 3
2
Bidirectional tumours ( 1 2 cases) DESjSMAjPGP 9.5jNSE DESjSMAjPGP 9.5 DESjSMAjNSE DESjPGP 9.5 SMAjPGP 9.5/NSE SMAjPGP 9.5 SMAjNSE
3 1
I 4
I
Immunonegative tumours (8 cases) 2 1 2 2 3 1 1
One case contained a quite different nodule which expressed DESjSMA (see Discussion) and is also listed under muscle tumours
DES = Desmin; SMA = smooth muscle actin; NSE = neuron-specific enolase
Fig. I-A
spindle cell tumonr of the stomach (a) which proved to be muscular and here shows strong SMA positivity (b)
GI STROMAL TUMOURS
111
Fig. 2-A small bowel spindle cell lesion (a), in this case showing positivity for both S-100 (b) and NSE (c)
112
P. L. NEWMAN ET A L .
Fig. 3-Another gastric spindle cell tumour (a), on this occasion expressing both desmin (b) and PGP 9.5 (c). Note the morphological similarities between the tumours illustrated in Figs 1-3
113
GI STROMAL TUMOURS
Fig. &An example of the distinctive syncytial, palisaded tumours which were found only in the stomach (a) and which invariably expressed NSE (b) and/or PGP 9.5
cases did not differ significantly in distribution from the series as a whole. Correlation of immunophenotype and outcome Six out of 42 cases (14.25 per cent) are known to have died of their tumours. Of these, two were immunonegative; two showed mixed muscle/neural differentiation; one showed muscle differentiation; and one was NSE positive. Four arose in the small bowel and 2 in the stomach. The additional four cases diagnosed as malignant on histological criteria but on which no follow-up was available all arose in the stomach. Two had a mixed immunophenotype, one stained with desmin only, and one with SMA only. All the malignant gastric tumours showed some areas with epithelioid features. None of the gastric tumours, including those classified as borderline, which stained for muscle markers and had a purely spindle cell morphology led to the death of the patient. None of the tumours expressing S-100 positivity led to the death of the patient, and indeed only one of the 22 cases (4.5 per cent) showing purely neural markers proved fatal.
DISCUSSION In this study we have found evidence of both neural and smooth muscle differentiation in GI stromal tumours in roughly equal proportions (36 and 3 1 per cent of cases respectively), while 20 per cent showed bidirectional differentiation and 13 per cent were negative for all the markers used. Of the 22 neural cases, only four stained with anti-S-100 protein antibodies (18 per cent of neural cases or 6.6 per cent of total). This figure is roughly in agreement with studies in which reactivity with S-I00 was the principal criterion used for neural differentiation, and which tended to suggest a much lower percentage of neural cases than we have found (Hjermstad et a1.-5 per cent,” Marco et al.-11.9 per cent,24Daimaru et d-7.8 per cent25).The four S-100 positive cases were all located outside the stomach, which would be consistent with the findings of Pike et ~ 1 who . found ~ ~ only 1 out of 20 gastric tumours to be S-100 positive, while 18 out of 20 intestinal cases gave a positive reaction. Pike et al.’s paper also comments on occasional S-100
114
P. L. NEWMAN ET A L .
Fig. 5-111 the majority of cases, nerves and/or tiny axons were demonstrated within the tumour by S-100 staining. Tumour cells are uniformly negative
positive cells in 15 further gastric tumours, which they interpreted as possibly representing entrapped Schwann cells. We also encountered scattered positive cells and small aggregates in many tumours, which we too considered were in keeping with entrapped neural elements and not due to tumour cell staining (Fig. 5). Mazur and Clarke2’apparently included this type of staining in their positive results, leading to S-100 positivity in 28.5 per cent, whereas if only their diffusely positive case is included, a rate of 3-5 per cent S-100 positive tumours is found. The assumption that scattered positive cells are not tumour cells is further explained by their finding of only one tumour which showed ultrastructural evidence of Schwann cell differentiation-this was the case with diffuse S- 100 positivity. It is interesting that while tumours staining with S-100 were located outside the stomach, 17 out of 18 tumours staining with other neural markers were gastric. The exception was a duodenal lesion. Within this group, only one patient is known to have died of this tumour, a high-grade epithelioid
neoplasm in which there was no difficulty in making a diagnosis of malignancy histologically and which caused death 8 months after diagnosis. This is a mortality rate of 6.6 per cent in those cases with adequate follow-up, compared with 18.5 per cent of the non-neural (muscle, bidirectional, or negative) cases. Several of the S-100-negative neural cases had a quite distinctive histological pattern, which we came to recognize while examining this series. These tumours were composed predominantly of spindle cells that were somewhat plumper than those of the other spindle cell tumours, but not as round as the epithelioid lesions. The nuclei were vesicular and often small prominent basophilic nucleoli were present. The cytoplasm was eosinophilic with a delicate and rather fibrillary appearance, and the cell membranes were frequently indistinct, giving a rather syncytial appearance. Focally, the nuclei were palisaded and in some areas this led to a plexiform pattern. None of these lesions showed any mitotic activity. These lesions bore some morphological similarity to the tumours labelled ‘plexosarcomas’ by Herrera et a1.,26but differed in several respects. Our tumours of this type were all GFAP and S-100 negative, whereas Herrera et al. describe GFAP in all cases and S-100 in 3 out of 4. Also, and more importantly, our cases have all proved to be biologically benign, whereas the ‘plexosarcomas’ have run a fulminant malignant course. In this series, three patients with von Recklinghausen’s disease are included. However, only one of the neural tumours came from a von Recklinghausen patient (S-100, NGE, and PGP positive). The other two patients’ tumours showed strong reactions with muscle markers (desmin and SMA in one, SMA alone in the other). One of these patients also had two incidental microscopic lesions in the gastric wall which were totally immunonegative. This lack of neural differentiation in the GI stromal tumours of von Recklinghausen patients has been previously noted ultrastructurally by Schaldenbrand and Appelman*’ and immunohistochemically by Daimaru et al.25 With regard to other putative markers of neural differentiation, we also employed Leu 7, which is said to recognize myelin-associated g l y c o p r ~ t e i n . ~ ~ ~ ~ ~ However, only one of the 60 lesions reacted with this antibody and this case also expressed desmin and SMA. Given that Leu 7 has been shown to stain undoubted smooth muscle t u m o ~ r as s ~well ~ ~as~ ~ neural tumours, we preferred to classify this case as
GI STROMAL TUMOURS
muscular rather than bidirectional. We additionally employed an antibody to GFAP, which is said to stain a high proportion of peripheral nerve sheath turn our^,^',^^ although this has been disputed.29 While Daimaru et al.25 have found all gastrointestinal schwannomas to be GFAP positive, positive staining was identified in only one of our neural cases, which also expressed S-100, NSE, and PGP. This result is more in line with that of Saul et al." and would suggest that the value of this antibody in the recognition of nerve sheath tumours remains controversial. Thirty-one per cent of our series expressed smooth muscle markers only. Of these, 14/19 (73.6 per cent) expressed both desmin and SMA. This is roughly similar to the results achieved with desmin by Thompson and Evans'3 (20 per cent desmin positive), but much more frequent than in the series of Pike et ~ 1 (0/40), . ~ Hjermstad ~ et (1/45), Ricci et al." (OjS), and Yagihashi et al." (0/12), and less than in the series by Saul et (42/79) and Norton et a1.34(6/6). The only other studies that have stained for smooth muscle actin also give results at variance with ours. Ricci et al." report 5/8 (62 per cent) of cases positive, while Pike et ~ 1 . found ~' 34/41 (83 per cent) to be positive. If we include our bidirectional cases, we have a total of 26 cases staining for SMA (43 per cent), still considerably less than these other surveys. The reasons for these major differences are uncertain. One factor which has definitely varied in the above series is the origin of the antibodies used, and whether they are mono- or polyclonal. The tissue fixative and length of fixation may also vary, but these factors are usually not stated. As one found using the S-100 antibody, so with SMA there is frequently a scattering of positive cells within the tumour, the majority of which appear related to blood vessels and were interpreted by us as pericytes. It is possible that these have been considered occasional true positive tumour cells in some previous reports. Indeed, in a recent review of GI stromal tumours, Appelman states that SMA positivity may be found in almost all such turn our^.^^ There is also the problem of entrapment of bundles of normal smooth muscle from the gut wall which have been broken up by tumour. This is particularly prominent at the periphery of small gastric tumours. The easiest way to avoid this pitfall is to take particular note of the low power appearance of the tumour, at which the general outline of background muscle bundles can be best appreciated (Fig. 6 ) . As regards
115
Fig. &Low power examination of this gastric lesion stained for SMA highlights the outline of normal muscle bundles which have been disrupted by tumour
clinical outcome, the expression of desmin and SMA in gastric tumours appears to correlate with good prognosis. None of the ten cases with this phenotype in this location is known to have led to the patient's death. We have insufficient follow-up data on the intestinal cases to make meaningful comment. The lesions with bidirectional differentiation as a group showed no specific combination of immunoreactivity, although none stained with S-100. There was no characteristic location for these lesions, although two of the four rectal lesions were in this category. Their main interest lies in the concept of divergent differentiation. Before accepting this concept, we re-examined these cases to ensure that we were not falling into any of the diagnostic pitfalls outlined above, and confirmed our results. The same cells appeared to stain for both neural and muscle markers. Are the antibodies that we used specific? While NSE is sometimes disparaged as 'non-specific enolase' it is frequently used as a neural marker, and in this study it was frequently co-expressed with PGP 9.5, which is thought to
1 I6
P. L. N E W M A N ET AL.
reliably reflect neural crest differentiati~n~~.” and is found to be diagnostically useful by us. Likewise, SMA is considered a sensitive smooth muscle marker,38although it also reacts with pericytes and myofibroblasts, but in the context of these tumours it should be specific. We therefore appear to have good evidence of bidirectional differentiation, a phenomenon which is being increasingly documented in a wide range of neoplasms including relatively common lesions such as carcinomata of the bronchus and cervix. Although it would be simplest to attribute this phenomenon to origin from an undifferentiated mesenchymal cell, it most probably reflects aberrant expression (or at least ‘desuppression’) of the genetic information carried by almost any gastrointestinal stromal cell. It also further underlines the fact that immunophenotype has little or nothing to do with ‘histogenesis’. There is no doubt, however, that patterns of differentiation determined by immunohistochemistry are of considerable value both in classification and in clinicopathological studies of neoplasia. Eight tumours failed to react with any of our seven antibodies. Two of these were malignant both histologically and clinically, suggesting ‘dedifferentiation’. One of these malignant tumours contained an area of histologically quite different benign ‘leiomyoma’ which reacted with both SMA and desmin. In this case, there was circumstantial evidence that the tumour had been present for 21 years, making it tempting to surmise that loss of immunoreactivity had indeed accompanied malignant transformation. However, four of the inert tumours were unequivocally benign ‘microleiomyomas’ in which the concept of malignant dedifferentiation is untenable. ACKNOWLEDGEMENTS
This study was performed while Dr was a Visiting Fellow in the Department pathology at St Thomas’s. C.D.M.F. is Research Campaign Senior Clinical Fellow.
Newman of Histoa Cancer Research
REFERENCES I . Golden T , Stout AP. Smooth muscle tumor? of the gastrointestinal tract and retroperitoneal tissues. Surg Gynecol Ohstet 1941, 13: 7 8 6 8 10. 2. Berg J. McNeer G. Leiomyosarcoma of the stomach. A clinical and pathological study. Currcer 1960; 13: 25-33.
3. Smithwick W, Biesecker JL. Leand PM. Leiomyobkdstoma: behaviour and prognosis. Cancer 1969; 2 4 996-1003. 4. Lavin P, Hajdu SI, Foote FW. Gastric and extragastric leiomyoblastomas. Clinicopathologic study of 44 cases. Cancer 1972; 2 9 305-3 1 1. 5. Appelman HD, Helwig EB. Gastric epithelioid leiomyoma and leiomyosarcoma (leiomyoblastoma). Cuncer 1976; 3 8 708-728. 6. Appelman HD, Helwig EB. Sarcomas of the stomach. Am J Clin Parhol 1977; 67: 2-10. 7. Ranchod M, Kempson RL. Smooth muscle tumors of the gdstrointestinal tract and retroperitoneum. A pathologic analysis of 100 cases. Cuncer 1977; 3 9 255-262. 8. Akwari OE, Dozois RR, Weiland LH, Beahrs OH. Leiomyosarcoma of the small and large bowel. Cancer 1978; 42: 1375-1384. 9. Shiu MH, Farr G H , Papdchristou D N , Hajdu SI. MyosdrComdS of the stomach: natural history, prognostic factors and management. Cuncer 1982;49 177-187. 10. Evans HL. Smooth muscle tumors of the gastrointestinal tract. A studyoE56casesfollowedforaminimumof 10 years. Cuncer 1985; 5 6 2242-2250. I I . Ricci A Jr, Ciccarelli 0,Cartun RW, Newcomb P. A chicopathologic and immunohistochemical study of 16 patients with small intestinal leiomyosarcoma. Limited utility of immunophenotyping. Cancer 1987;60: 179CL-1799. I ? . Silverberg SG. Reproducibility of the mitosis count in the histologic diagnosis of smooth muscle tumors of the uterus. Hum P a r i d 1976; 7: 451454. 13. Tsushima K, Rainwater LM, Goellner J R , Van Heerden JA, Leiber MM. Leiomyosarcomas and benign tumors of the stomach: nuclear DNA patterns studied by flow cytometry. Mayo C/in Proc 1987; 62: 275-280. 14. Kiyabu MT, Bishop PC, Parker JW, Turner RR, Fitzgibbons PL. Smooth muscle tumors of the gastrointestinal tract. Flow cytometric quantitation of DNA and nuclear antigen content and correlation with histologic grade. Am J Surg Parhol 1988; 12: 95&960. 15. Federspiel BH, Sobin LH, Helwig EB, Mike1 UV, Bahr CF. Morphometry and cytophotoinetric assessment of DNA in smooth muscle tumors (leiomyomas and leiomyosarcomas) of the gastrointestinal tract. A n d Quum Gyro/ Hi.vo/ 1987; 9 105-1 14. 16. Weiss RA, Mackay B. Malignant smooth muscle tumors of the gastrointestinal tract: an ultrastructural study of 20 cases. Ultrastruct Prirho/ 1 9 8 l ; Z 231-240. 17. Knapp RH, Wick MR, Goellner JR. Leiomyoblastomas and their relationship to other smooth muscle tumors of the gastrointestinal tract. An electron microscopic study. Am J Surg Pafhol 1984; 8: 449461. 18. Yagihashi S. Kimura M, Kurotaki H. ct a/.Gastric SUbmUCosdl tumours of neurogenic origin with neuroaxonal and Schwann cell elements. J Pathol 1987; 153: 41-50, 19. Hjermstad BM, Sobin LH, Helwig EB. Stromal tumors of the gastrointestinal tract: myogenic or neurogenic? Am J Surg Purhol 1987; 11: 383-386. 20. Saul SA, Rast ML, Brooks JJ. The immunohistochemistry of gastrointestinal stromal tumors. Evidence supporting an origin from smooth muscle. Am JSurg Parhol 1987; 11: 464473. 21. Mazur MT, Clark HB. Gastric stromal tumors. Reappraisal of histogenesis. Am JSurg Parhol1983; 7: 507-519. 22. Pike AM. Lloyd RV, Appelman HD. Cell markers in gastrointestinal stromal tumors. Hum Parhol 1988; 1 9 830-834. 23. Seremetis M G , Lyons WS, DeGuzman VC, Peabody JW Leiomyomata of the esophagus. An analysis of 838 cases. Cancer 1976; 38: 2 166-2 177. 24. Marco U, Muntal BL, Targarona E, Marco C. Gastrointestinal stromal tumors. Immunohistochemical study. Lab h e s f 1989; 60: 57A (Abstract). 25. Daimaru Y, Kido H, Hashimoto H, Enjoji M . Benign schwannoma of the gastrointestinal tract. a clinicopathologic and immunohistochemical study. H u m Pathol 1988; 1 9 257-264. 26. Herrera GA, Cerezo L, Jones JE, er ( I / . Gastrointestinal autonomic Arch Parhol Lab Med 1989; 113: nerve tumors-’plexosarcomas’. 84C853.
GI STROMAL T U M O U R S 27. Schaldenbrand JD, Appelman HD. Solitary solid stromal gastrointestinal tumors in von Recklinghausen’s disease with minimal smooth muscle differentiation. Hum Purho/1984; 15: 229-232. 28. Swanson PE, Manivel JC, Wick M R . Immunoreactivity for Leu-7 in neurofibrosarcoma and other spindlecell sarcomas ofsoft tissue. Am J Pallrol1987; 1 2 6 546-560. 29. Johnson M D , Click AD, Davis BW. lmmunohistochemical evaluation of Leu-7, myelin-basic protein, S-I00 protein, glial fibrillary acidic protein and LN 3 immunoreactivity in nerve sheath tumors and sarcomas. Arch Prrfhol Lah Mrd 1988; 1 1 2 155-160. 30. Swanson PE, Stanley MW, Scheithauer BW, Wick M R . Primary cutaneous leiom yosarcoma. A hiStolOgicd1and immunohistochemical study of 9 cases with ultrastructural correlation. J Curan Parho/ 1988; 15: 129-141. 31. Kawahara E, Oda Y, Ooi A, Katsuda S , Nakanishi I, Umeda S. Expression ofglial fibrillaryacidic protein (GFAP) in peripheral nerve sheath tumors. A comparative study of immunoreactivity of GFAP, vimentin, S- 100 protein and neurofilament in 38 schwannomas and 18 neurofibromas. Am J S u r g Puthol 1988; 1 2 115-120. 32. Gray MH, Rosenberg AE, Dickersin G R , Bhan AK. GIial fibrillary acidic protein and keratin expression by benign and malignant nerve sheath tumors. Hum Pufhol 1989; 20: 1089-1096.
117
33. Thompson EM, Evans DJ. The significance of PGP 9.5 in tumorsan immunohistochemical study of gastrointestinal stromal tumors. Histopathologj 1990; 17: 175-177. 34. Norton AJ, Thomas JA, Isaacson PG. Cytokeratin-specific monoclonal antibodies are reactive with turnours of smooth muscle derivation. An immunocytochemical and biochemical study using antibodies to intermediate filament cytoskektdl proteins. Histoparhology 1987; 11: 487499. 35. Appelman HD. Mesenchymal tumors of the gut: historical perspectives, new approaches, new results. and does it make any difference? Monogr Patho/ 1990; 31: 220-246. 36. Thompson RJ, Doran J F , Jackson P, Dhillon AP. RodeJ. PGP9.5-a new marker for vertebrate neurons and neuroendocrine cells. Brain Res 1983; 278: 224228. 37. Rode J, Dhillon AP, Doran J F , Jackson P, Thompson RJ. PGP9.5-a new marker for human neuroendocrine tumours. Histopothologj. 1985;9 147-158. 38. Jones H, Steart PV, Du Boulay CEH, Roche WR. Alpha-smooth muscle actin as a marker for soft tissue tumours. A comparison with desmin. J Parliol1990: 162: 29-33.
164: 107- 1 17 (1 99 1)
GASTROINTESTINAL STROMAL TUMOURS: CORRELATION OF IMMUNOPHENOTYPE WITH CLINICOPATHOLOGICAL FEATURES P. L. NEWMAN*, C. WADDEN AND
c. D.
M. FLETCHER
Department of Histopathology, St Thomas's Hospital (U.M.D.S.), London, U.K.; *Department of Histopathology, University Hospital of Wales, Cardiff, U.K. Received 25 October 1990 Accepted23 November 1990
SUMMARY Stromal tumours of the gastrointestinal tract remain a persistent source of controversy with regard to both their proposed lines of differentiation and the difficulty in predicting their biological behaviour. We have examined 60 cases with a panel of seven antibodies directed at the identification of smooth muscular or neural differentiation. In our hands, 36 per cent of cases showed neural differentiation (although only 6.6 per cent expressed S-100 protein); 31 per cent appeared smooth muscular; 20 per cent manifested bidirectional differentiation; and 13 per cent were negative for all the markers used. Histological appearances do not reliably reflect immunophenotype. We have attempted to correlate immunophenotype with the site of the lesions and with their clinical behaviour. Mean follow-up of 5 years was obtained in 42 cases. Turnours with a neural phenotype have the best prognosis. Gastric tumours expressing both desmin and smooth muscle actin (in the absence of other markers) with up to 4 mitoses per 30 HPF behave in a benign fashion. Larger studies are required to substantiate the value of immunophenotyping this complex group of turnours. KEY WORDS-stomach,
intestine, leiomyoma, leiomyosarcoma, schwannorna, immunohistochemistry.
those with diploid lesions, but even so, some with diploid lesions died of their disease. Since 1941 when Golden a n d Stout' published In addition to the problem of biological behavthe first large series of gastrointestinal (GI) iour, the histogenesis of these tumours has also been smooth muscle tumours indicating that histological the subject of several studies, with widely differing appearance did not correlate well with biological results. Ail gastrointestinal stromal tumours were behaviour, several studies have addressed them- originally considered to be of smooth muscle origin, selves to the problem of assessing prognosis in this but while electron microscopy has been reported to group of turnours.*-" Initially, these concentrated show evidence of smooth muscle differentiation in on histological features, particularly mitotic counts, some series,"." it purports to show nerve sheath cellularity, and anaplasia, as well as size and differentiation in others." However, in many location. Some degree of correlation was found, but cases it appears that only scattered cells show differsome of the features assessed are very subjective and entiated features, the bulk of the tumour being subject to considerable inter-observer variation.'* undifferentiated, a n d the possibility of entrapment Later, flow c y t ~ m e t r i c " ~ 'and ~ m o r p h o m e t r i ~ ' ~ of residual normal tissue components within the techniques were applied in a n attempt to establish tumour cannot be excluded. a n objective measure of prognosis. These showed, Immunohistochemical studies have also produced not surprisingly, that patients with aneuploid diametrically opposed views as to the nature of these tumours had a shorter average survival time than tumours. Largely based on S- 100 protein a n d desAddressee for correspondence: Dr C. D. M. Fletcher, min immunoreactivity, some groups have favoured smooth muscle differentiation in the majority of Department of Histopathology, St Thomas's Hospital, London case^,^^.*^ some neura1,'8.2' and Pike et a1.22found SEI 7EH, U.K. INTRODUCTION
0022-3417/91/050107-11 $05.50 0 1991 by John Wiley & Sons, Ltd
108
P. L. NEWMAN ET AL.
Table I-Antibodies
used in this study of gastrointestinal stromal tumours Monoclonal polyclonal
Antibody Desmin Smooth muscle actin (IA4) (SMA)
M M
PGP9.5 Neuron-specificenolase (NSE) Glial fibrillary acidic protein (GFAP) Leu I
P P P P M
s-100
Source Dako Sigma Dako Ultraclone Dako Dako Becton Dickinson
Dilution 1 /SO
1/16 000 1j2000 lj800 1/SO0 1 /SO0
1/10
Table 11-Histological criteria for grading gastrointestinal stromal tumours Benign
or Borderline or or
0-2 mitoses per 30 HPF 0 mitoses per 30 HPF
Spindle cell lesion, no atypia Epithelioid lesion
2-3 mitoses per 30 HPF
Spindle cell lesion, mild pleomorphism/ hyperchromasia Spindle cell lesion, no atypia Epithelioid lesion
3 4 mitoses per 30 HPF 1 mitosis per 30 HPF
or
> 5 mitoses per 30 HPF > 3 mitoses per 30 HPF
or
> 2 mitoses per 30 HPF
Malignant
Spindle cell lesion, no atypia Spindle cell lesion, frank pleomorphism/ hyperchromasia Epithelioid lesion
I HPF = 0.1 59 mm2on the microscopes used
different immunophenotypes in different regions of the bowel. It is therefore clear that gastrointestinal stromal tumours are poorly categorized. Possibly they represent a heterogeneous group and this may explain why it is so difficult to adopt a single set of morphological criteria of malignancy with any degree of confidence. In this retrospective study we have therefore attempted to correlate immunophenotype, as defined by a panel of antibodies, with prognosis, in particular trying to identify tumours with features of neural differentiation, and possibly identify good or bad prognostic groups. MATERIALS AND METHODS All cases indexed as leiomyoma, leiomyosarcoma, leiomyoblastoma, schwannoma, or neurofibroma within the gastrointestinal tract were retrieved from the files of the histopathology department of St Thomas’s Hospital, producing a total of 66 cases.
Lesions in the oesophagus were excluded as the pathology of these is well described.23 Fresh 4 pm sections were cut of all the formalinfixed, paraffin-embedded blocks available, and a representative block was selected from each case for immunohistochemical staining. In two cases, two blocks were selected as the tumours showed a range of histological appearances. Immunohistochemistry was performed using the avidinbiotin-peroxidase (ABC) technique, employing the antibodies listed in Table I. Appropriate controls were used throughout. The tumours were assessed on the basis of their haematoxylin and eosin appearance, and categorized as benign, malignant, or borderline using the criteria set out in Table 11, and without knowledge of the clinical outcome. Mitoses were counted in 30 random high-power fields (where 1 HPF=0.159mm2) taken from the block which subjectively appeared the most cellular and mitotic. No attempt was made to differentiate between smooth muscle and neural tumours at this stage. The immunohistochemical results were
GI STROMAL TUMOURS
assessed by each author independently and without knowledge of the other results. There was good concordance between the observers in the assessment of positive or negative staining. Clinical follow-up was obtained by examination of patients’ hospital notes and by contacting general practitioners. Five cases were excluded on the grounds of supplementary immunohistochemistry and additional clinical information (four were metastases and one was a mesenteric tumour) and a further case from the 1950s was deleted owing to inadequate tissue preservation. This left 60 cases for study.
109
although basically of spindle cell morphology, the nuclei were rather plumper than the majority of spindle cell lesions, and there was focal nuclear palisading, with a rather syncytial, plexiform appearance in areas, and with some cytoplasmic vacuolation and degeneration of the ground substance. All nine of these lesions were located in the stomach, and none appeared histologically malignant. Of the 11 classified as malignant, 6 died of tumour; 4 were lost to follow-up; and one died 5 years post-operatively of an unknown cause. Immunohistochemistry
RESULTS Clinical Of the 60 cases included in the study, adequate follow-up was available on 42, ranging from an early tumour related death at 8 months to 19 years’ disease-free survival (mean follow-up 5 years). Thirty-one patients were female, 27 male, and in two the gender was not recorded. The ages ranged from 35 to 85 years (mean 62 years). Three patients suffered from von Recklinghausen’s disease. In 45 cases (75 per cent), the tumours were located in the stomach; 1 1 (1 8 per cent) were in the small bowel (4 duodenum, 2 jejunum, 5 unspecified); and 4 were from the rectum. Of the 42 cases with follow-up, 6 (14.25 per cent) died of their disease (4 small bowel tumours, 2 stomach). Histology On the basis of the criteria in Table 11, 11 cases were classified as malignant, 4 as borderline, and 45 as benign. In practice, on the basis of size, 24 of the ‘benign’ lesions would have had a rider on the histology report expressing uncertainty as to the likely behaviour of GI stromal tumours. Eight tumours were epithelioid; four showed an even mixture of epithelioid and spindle cell areas; and the remaining 48 cases had a predominantly spindle cell morphology. When subsequently correlating morphology with immunohistochemistry, it was abundantly clear that no histological features, such as nuclear morphology, palisading, or cytoplasmic eosinophilia, could distinguish spindle cell lesions with a neural or smooth muscle phenotype. There were nine tumours that were hard to classify into spindle or epithelioid types, in that
The immunophenotypes of the tumours are recorded in Table 111. Nineteen (3 1 per cent) cases showed evidence of smooth muscle differentiation as defined by staining solely with desmin or smooth muscle actin (SMA) antibodies (muscle tumours) (Fig. I). Twenty-two (36 per cent) showed neural differentiation (neural tumours) (Fig. 2) and eight (13 per cent) were negative with all antibodies. Twelve tumours (20 per cent) reacted with both smooth muscle and neural markers other than S- 100 (bidirectional tumours) (Fig. 3). No tumour expressed a combination of S-100 and muscle markers. All nine of the syncytial plexiform, palisaded lesions stained with neural markers (NSE and/or PGP) only (Fig. 4). Only a single case each stained with Leu-7 (also desmin/SMA positive) and GFAP (also S-lOO/NSE/PGP 9.5 positive). Of the eight cases which were entirely negative, despite good internal controls, four were spindle-celled and two each were epithelioid or of mixed type. Many tumours showed scattered cells reactive with S-100 or SMA, which tended to be located at the periphery of the lesion and were interpreted as entrapped normal structures (see Discussion). Correlation of immunophenotype and site
Of the tumours showing smooth muscle differentiation, the distribution within the gut was similar to the overall distribution of lesions, with 14 cases (73.6 per cent) located in the stomach and five cases (26.3 per cent) in the small bowel. More striking was the distribution of neural tumours. A11 tumours in which S-100 was identified were situated outside the stomach, while 17 of the 18 (94 per cent) tumours staining for NSE and/or PGP 9.5 in the absence of S-100 were situated within the stomach. The totally immunonegative and bidirectional differentiation
110
P. L. NEWMAN ET AL.
Table 111-Immunophenotyping
Neural tumours (22 cases)
Muscle tumours ( I 9 cases) DES and SMA DES only SMA only
of 60 gastrointestinal stromal tumours
S-lOO,”SE/PGP 9.5 S-lOO/NSE PGP9.5/NSE PGP 9.5 only NSE only
14 3
2
Bidirectional tumours ( 1 2 cases) DESjSMAjPGP 9.5jNSE DESjSMAjPGP 9.5 DESjSMAjNSE DESjPGP 9.5 SMAjPGP 9.5/NSE SMAjPGP 9.5 SMAjNSE
3 1
I 4
I
Immunonegative tumours (8 cases) 2 1 2 2 3 1 1
One case contained a quite different nodule which expressed DESjSMA (see Discussion) and is also listed under muscle tumours
DES = Desmin; SMA = smooth muscle actin; NSE = neuron-specific enolase
Fig. I-A
spindle cell tumonr of the stomach (a) which proved to be muscular and here shows strong SMA positivity (b)
GI STROMAL TUMOURS
111
Fig. 2-A small bowel spindle cell lesion (a), in this case showing positivity for both S-100 (b) and NSE (c)
112
P. L. NEWMAN ET A L .
Fig. 3-Another gastric spindle cell tumour (a), on this occasion expressing both desmin (b) and PGP 9.5 (c). Note the morphological similarities between the tumours illustrated in Figs 1-3
113
GI STROMAL TUMOURS
Fig. &An example of the distinctive syncytial, palisaded tumours which were found only in the stomach (a) and which invariably expressed NSE (b) and/or PGP 9.5
cases did not differ significantly in distribution from the series as a whole. Correlation of immunophenotype and outcome Six out of 42 cases (14.25 per cent) are known to have died of their tumours. Of these, two were immunonegative; two showed mixed muscle/neural differentiation; one showed muscle differentiation; and one was NSE positive. Four arose in the small bowel and 2 in the stomach. The additional four cases diagnosed as malignant on histological criteria but on which no follow-up was available all arose in the stomach. Two had a mixed immunophenotype, one stained with desmin only, and one with SMA only. All the malignant gastric tumours showed some areas with epithelioid features. None of the gastric tumours, including those classified as borderline, which stained for muscle markers and had a purely spindle cell morphology led to the death of the patient. None of the tumours expressing S-100 positivity led to the death of the patient, and indeed only one of the 22 cases (4.5 per cent) showing purely neural markers proved fatal.
DISCUSSION In this study we have found evidence of both neural and smooth muscle differentiation in GI stromal tumours in roughly equal proportions (36 and 3 1 per cent of cases respectively), while 20 per cent showed bidirectional differentiation and 13 per cent were negative for all the markers used. Of the 22 neural cases, only four stained with anti-S-100 protein antibodies (18 per cent of neural cases or 6.6 per cent of total). This figure is roughly in agreement with studies in which reactivity with S-I00 was the principal criterion used for neural differentiation, and which tended to suggest a much lower percentage of neural cases than we have found (Hjermstad et a1.-5 per cent,” Marco et al.-11.9 per cent,24Daimaru et d-7.8 per cent25).The four S-100 positive cases were all located outside the stomach, which would be consistent with the findings of Pike et ~ 1 who . found ~ ~ only 1 out of 20 gastric tumours to be S-100 positive, while 18 out of 20 intestinal cases gave a positive reaction. Pike et al.’s paper also comments on occasional S-100
114
P. L. NEWMAN ET A L .
Fig. 5-111 the majority of cases, nerves and/or tiny axons were demonstrated within the tumour by S-100 staining. Tumour cells are uniformly negative
positive cells in 15 further gastric tumours, which they interpreted as possibly representing entrapped Schwann cells. We also encountered scattered positive cells and small aggregates in many tumours, which we too considered were in keeping with entrapped neural elements and not due to tumour cell staining (Fig. 5). Mazur and Clarke2’apparently included this type of staining in their positive results, leading to S-100 positivity in 28.5 per cent, whereas if only their diffusely positive case is included, a rate of 3-5 per cent S-100 positive tumours is found. The assumption that scattered positive cells are not tumour cells is further explained by their finding of only one tumour which showed ultrastructural evidence of Schwann cell differentiation-this was the case with diffuse S- 100 positivity. It is interesting that while tumours staining with S-100 were located outside the stomach, 17 out of 18 tumours staining with other neural markers were gastric. The exception was a duodenal lesion. Within this group, only one patient is known to have died of this tumour, a high-grade epithelioid
neoplasm in which there was no difficulty in making a diagnosis of malignancy histologically and which caused death 8 months after diagnosis. This is a mortality rate of 6.6 per cent in those cases with adequate follow-up, compared with 18.5 per cent of the non-neural (muscle, bidirectional, or negative) cases. Several of the S-100-negative neural cases had a quite distinctive histological pattern, which we came to recognize while examining this series. These tumours were composed predominantly of spindle cells that were somewhat plumper than those of the other spindle cell tumours, but not as round as the epithelioid lesions. The nuclei were vesicular and often small prominent basophilic nucleoli were present. The cytoplasm was eosinophilic with a delicate and rather fibrillary appearance, and the cell membranes were frequently indistinct, giving a rather syncytial appearance. Focally, the nuclei were palisaded and in some areas this led to a plexiform pattern. None of these lesions showed any mitotic activity. These lesions bore some morphological similarity to the tumours labelled ‘plexosarcomas’ by Herrera et a1.,26but differed in several respects. Our tumours of this type were all GFAP and S-100 negative, whereas Herrera et al. describe GFAP in all cases and S-100 in 3 out of 4. Also, and more importantly, our cases have all proved to be biologically benign, whereas the ‘plexosarcomas’ have run a fulminant malignant course. In this series, three patients with von Recklinghausen’s disease are included. However, only one of the neural tumours came from a von Recklinghausen patient (S-100, NGE, and PGP positive). The other two patients’ tumours showed strong reactions with muscle markers (desmin and SMA in one, SMA alone in the other). One of these patients also had two incidental microscopic lesions in the gastric wall which were totally immunonegative. This lack of neural differentiation in the GI stromal tumours of von Recklinghausen patients has been previously noted ultrastructurally by Schaldenbrand and Appelman*’ and immunohistochemically by Daimaru et al.25 With regard to other putative markers of neural differentiation, we also employed Leu 7, which is said to recognize myelin-associated g l y c o p r ~ t e i n . ~ ~ ~ ~ ~ However, only one of the 60 lesions reacted with this antibody and this case also expressed desmin and SMA. Given that Leu 7 has been shown to stain undoubted smooth muscle t u m o ~ r as s ~well ~ ~as~ ~ neural tumours, we preferred to classify this case as
GI STROMAL TUMOURS
muscular rather than bidirectional. We additionally employed an antibody to GFAP, which is said to stain a high proportion of peripheral nerve sheath turn our^,^',^^ although this has been disputed.29 While Daimaru et al.25 have found all gastrointestinal schwannomas to be GFAP positive, positive staining was identified in only one of our neural cases, which also expressed S-100, NSE, and PGP. This result is more in line with that of Saul et al." and would suggest that the value of this antibody in the recognition of nerve sheath tumours remains controversial. Thirty-one per cent of our series expressed smooth muscle markers only. Of these, 14/19 (73.6 per cent) expressed both desmin and SMA. This is roughly similar to the results achieved with desmin by Thompson and Evans'3 (20 per cent desmin positive), but much more frequent than in the series of Pike et ~ 1 (0/40), . ~ Hjermstad ~ et (1/45), Ricci et al." (OjS), and Yagihashi et al." (0/12), and less than in the series by Saul et (42/79) and Norton et a1.34(6/6). The only other studies that have stained for smooth muscle actin also give results at variance with ours. Ricci et al." report 5/8 (62 per cent) of cases positive, while Pike et ~ 1 . found ~' 34/41 (83 per cent) to be positive. If we include our bidirectional cases, we have a total of 26 cases staining for SMA (43 per cent), still considerably less than these other surveys. The reasons for these major differences are uncertain. One factor which has definitely varied in the above series is the origin of the antibodies used, and whether they are mono- or polyclonal. The tissue fixative and length of fixation may also vary, but these factors are usually not stated. As one found using the S-100 antibody, so with SMA there is frequently a scattering of positive cells within the tumour, the majority of which appear related to blood vessels and were interpreted by us as pericytes. It is possible that these have been considered occasional true positive tumour cells in some previous reports. Indeed, in a recent review of GI stromal tumours, Appelman states that SMA positivity may be found in almost all such turn our^.^^ There is also the problem of entrapment of bundles of normal smooth muscle from the gut wall which have been broken up by tumour. This is particularly prominent at the periphery of small gastric tumours. The easiest way to avoid this pitfall is to take particular note of the low power appearance of the tumour, at which the general outline of background muscle bundles can be best appreciated (Fig. 6 ) . As regards
115
Fig. &Low power examination of this gastric lesion stained for SMA highlights the outline of normal muscle bundles which have been disrupted by tumour
clinical outcome, the expression of desmin and SMA in gastric tumours appears to correlate with good prognosis. None of the ten cases with this phenotype in this location is known to have led to the patient's death. We have insufficient follow-up data on the intestinal cases to make meaningful comment. The lesions with bidirectional differentiation as a group showed no specific combination of immunoreactivity, although none stained with S-100. There was no characteristic location for these lesions, although two of the four rectal lesions were in this category. Their main interest lies in the concept of divergent differentiation. Before accepting this concept, we re-examined these cases to ensure that we were not falling into any of the diagnostic pitfalls outlined above, and confirmed our results. The same cells appeared to stain for both neural and muscle markers. Are the antibodies that we used specific? While NSE is sometimes disparaged as 'non-specific enolase' it is frequently used as a neural marker, and in this study it was frequently co-expressed with PGP 9.5, which is thought to
1 I6
P. L. N E W M A N ET AL.
reliably reflect neural crest differentiati~n~~.” and is found to be diagnostically useful by us. Likewise, SMA is considered a sensitive smooth muscle marker,38although it also reacts with pericytes and myofibroblasts, but in the context of these tumours it should be specific. We therefore appear to have good evidence of bidirectional differentiation, a phenomenon which is being increasingly documented in a wide range of neoplasms including relatively common lesions such as carcinomata of the bronchus and cervix. Although it would be simplest to attribute this phenomenon to origin from an undifferentiated mesenchymal cell, it most probably reflects aberrant expression (or at least ‘desuppression’) of the genetic information carried by almost any gastrointestinal stromal cell. It also further underlines the fact that immunophenotype has little or nothing to do with ‘histogenesis’. There is no doubt, however, that patterns of differentiation determined by immunohistochemistry are of considerable value both in classification and in clinicopathological studies of neoplasia. Eight tumours failed to react with any of our seven antibodies. Two of these were malignant both histologically and clinically, suggesting ‘dedifferentiation’. One of these malignant tumours contained an area of histologically quite different benign ‘leiomyoma’ which reacted with both SMA and desmin. In this case, there was circumstantial evidence that the tumour had been present for 21 years, making it tempting to surmise that loss of immunoreactivity had indeed accompanied malignant transformation. However, four of the inert tumours were unequivocally benign ‘microleiomyomas’ in which the concept of malignant dedifferentiation is untenable. ACKNOWLEDGEMENTS
This study was performed while Dr was a Visiting Fellow in the Department pathology at St Thomas’s. C.D.M.F. is Research Campaign Senior Clinical Fellow.
Newman of Histoa Cancer Research
REFERENCES I . Golden T , Stout AP. Smooth muscle tumor? of the gastrointestinal tract and retroperitoneal tissues. Surg Gynecol Ohstet 1941, 13: 7 8 6 8 10. 2. Berg J. McNeer G. Leiomyosarcoma of the stomach. A clinical and pathological study. Currcer 1960; 13: 25-33.
3. Smithwick W, Biesecker JL. Leand PM. Leiomyobkdstoma: behaviour and prognosis. Cancer 1969; 2 4 996-1003. 4. Lavin P, Hajdu SI, Foote FW. Gastric and extragastric leiomyoblastomas. Clinicopathologic study of 44 cases. Cancer 1972; 2 9 305-3 1 1. 5. Appelman HD, Helwig EB. Gastric epithelioid leiomyoma and leiomyosarcoma (leiomyoblastoma). Cuncer 1976; 3 8 708-728. 6. Appelman HD, Helwig EB. Sarcomas of the stomach. Am J Clin Parhol 1977; 67: 2-10. 7. Ranchod M, Kempson RL. Smooth muscle tumors of the gdstrointestinal tract and retroperitoneum. A pathologic analysis of 100 cases. Cuncer 1977; 3 9 255-262. 8. Akwari OE, Dozois RR, Weiland LH, Beahrs OH. Leiomyosarcoma of the small and large bowel. Cancer 1978; 42: 1375-1384. 9. Shiu MH, Farr G H , Papdchristou D N , Hajdu SI. MyosdrComdS of the stomach: natural history, prognostic factors and management. Cuncer 1982;49 177-187. 10. Evans HL. Smooth muscle tumors of the gastrointestinal tract. A studyoE56casesfollowedforaminimumof 10 years. Cuncer 1985; 5 6 2242-2250. I I . Ricci A Jr, Ciccarelli 0,Cartun RW, Newcomb P. A chicopathologic and immunohistochemical study of 16 patients with small intestinal leiomyosarcoma. Limited utility of immunophenotyping. Cancer 1987;60: 179CL-1799. I ? . Silverberg SG. Reproducibility of the mitosis count in the histologic diagnosis of smooth muscle tumors of the uterus. Hum P a r i d 1976; 7: 451454. 13. Tsushima K, Rainwater LM, Goellner J R , Van Heerden JA, Leiber MM. Leiomyosarcomas and benign tumors of the stomach: nuclear DNA patterns studied by flow cytometry. Mayo C/in Proc 1987; 62: 275-280. 14. Kiyabu MT, Bishop PC, Parker JW, Turner RR, Fitzgibbons PL. Smooth muscle tumors of the gastrointestinal tract. Flow cytometric quantitation of DNA and nuclear antigen content and correlation with histologic grade. Am J Surg Parhol 1988; 12: 95&960. 15. Federspiel BH, Sobin LH, Helwig EB, Mike1 UV, Bahr CF. Morphometry and cytophotoinetric assessment of DNA in smooth muscle tumors (leiomyomas and leiomyosarcomas) of the gastrointestinal tract. A n d Quum Gyro/ Hi.vo/ 1987; 9 105-1 14. 16. Weiss RA, Mackay B. Malignant smooth muscle tumors of the gastrointestinal tract: an ultrastructural study of 20 cases. Ultrastruct Prirho/ 1 9 8 l ; Z 231-240. 17. Knapp RH, Wick MR, Goellner JR. Leiomyoblastomas and their relationship to other smooth muscle tumors of the gastrointestinal tract. An electron microscopic study. Am J Surg Pafhol 1984; 8: 449461. 18. Yagihashi S. Kimura M, Kurotaki H. ct a/.Gastric SUbmUCosdl tumours of neurogenic origin with neuroaxonal and Schwann cell elements. J Pathol 1987; 153: 41-50, 19. Hjermstad BM, Sobin LH, Helwig EB. Stromal tumors of the gastrointestinal tract: myogenic or neurogenic? Am J Surg Purhol 1987; 11: 383-386. 20. Saul SA, Rast ML, Brooks JJ. The immunohistochemistry of gastrointestinal stromal tumors. Evidence supporting an origin from smooth muscle. Am JSurg Parhol 1987; 11: 464473. 21. Mazur MT, Clark HB. Gastric stromal tumors. Reappraisal of histogenesis. Am JSurg Parhol1983; 7: 507-519. 22. Pike AM. Lloyd RV, Appelman HD. Cell markers in gastrointestinal stromal tumors. Hum Parhol 1988; 1 9 830-834. 23. Seremetis M G , Lyons WS, DeGuzman VC, Peabody JW Leiomyomata of the esophagus. An analysis of 838 cases. Cancer 1976; 38: 2 166-2 177. 24. Marco U, Muntal BL, Targarona E, Marco C. Gastrointestinal stromal tumors. Immunohistochemical study. Lab h e s f 1989; 60: 57A (Abstract). 25. Daimaru Y, Kido H, Hashimoto H, Enjoji M . Benign schwannoma of the gastrointestinal tract. a clinicopathologic and immunohistochemical study. H u m Pathol 1988; 1 9 257-264. 26. Herrera GA, Cerezo L, Jones JE, er ( I / . Gastrointestinal autonomic Arch Parhol Lab Med 1989; 113: nerve tumors-’plexosarcomas’. 84C853.
GI STROMAL T U M O U R S 27. Schaldenbrand JD, Appelman HD. Solitary solid stromal gastrointestinal tumors in von Recklinghausen’s disease with minimal smooth muscle differentiation. Hum Purho/1984; 15: 229-232. 28. Swanson PE, Manivel JC, Wick M R . Immunoreactivity for Leu-7 in neurofibrosarcoma and other spindlecell sarcomas ofsoft tissue. Am J Pallrol1987; 1 2 6 546-560. 29. Johnson M D , Click AD, Davis BW. lmmunohistochemical evaluation of Leu-7, myelin-basic protein, S-I00 protein, glial fibrillary acidic protein and LN 3 immunoreactivity in nerve sheath tumors and sarcomas. Arch Prrfhol Lah Mrd 1988; 1 1 2 155-160. 30. Swanson PE, Stanley MW, Scheithauer BW, Wick M R . Primary cutaneous leiom yosarcoma. A hiStolOgicd1and immunohistochemical study of 9 cases with ultrastructural correlation. J Curan Parho/ 1988; 15: 129-141. 31. Kawahara E, Oda Y, Ooi A, Katsuda S , Nakanishi I, Umeda S. Expression ofglial fibrillaryacidic protein (GFAP) in peripheral nerve sheath tumors. A comparative study of immunoreactivity of GFAP, vimentin, S- 100 protein and neurofilament in 38 schwannomas and 18 neurofibromas. Am J S u r g Puthol 1988; 1 2 115-120. 32. Gray MH, Rosenberg AE, Dickersin G R , Bhan AK. GIial fibrillary acidic protein and keratin expression by benign and malignant nerve sheath tumors. Hum Pufhol 1989; 20: 1089-1096.
117
33. Thompson EM, Evans DJ. The significance of PGP 9.5 in tumorsan immunohistochemical study of gastrointestinal stromal tumors. Histopathologj 1990; 17: 175-177. 34. Norton AJ, Thomas JA, Isaacson PG. Cytokeratin-specific monoclonal antibodies are reactive with turnours of smooth muscle derivation. An immunocytochemical and biochemical study using antibodies to intermediate filament cytoskektdl proteins. Histoparhology 1987; 11: 487499. 35. Appelman HD. Mesenchymal tumors of the gut: historical perspectives, new approaches, new results. and does it make any difference? Monogr Patho/ 1990; 31: 220-246. 36. Thompson RJ, Doran J F , Jackson P, Dhillon AP. RodeJ. PGP9.5-a new marker for vertebrate neurons and neuroendocrine cells. Brain Res 1983; 278: 224228. 37. Rode J, Dhillon AP, Doran J F , Jackson P, Thompson RJ. PGP9.5-a new marker for human neuroendocrine tumours. Histopothologj. 1985;9 147-158. 38. Jones H, Steart PV, Du Boulay CEH, Roche WR. Alpha-smooth muscle actin as a marker for soft tissue tumours. A comparison with desmin. J Parliol1990: 162: 29-33.
