A reassessment of histologic classification and an immunohistochemical study of 88 retinoblastomas. A special reference to the advent of bipolar-like cells

2901

A Reassessment of Histologic Classification and an lmmunohistochemical Study of 88 Retinoblastomas A Special Reference to the Advent of Bipolar-Like Cells

Wei He, M.D.,*,t Hiroshi Hashimoto, M.D.,* Masazumi Tsuneyoshi, M.D.,* Munetomo Enjoji, M.D.,* and Hajime Znomata, M.D.t

Background. Despite perpetual efforts of investigators, the histogenesis of retinoblastoma is still in dispute and histologic classification satisfactorily predictive of prognosis does not seem to be in use.

Methods. The authors studied 88 cases of retinoblastoma clinicopathologically and immunohistochemically, paying special attention to the presence of a “bipolar-like cell” element that would be used as one of the criteria for the diagnosis of differentiated retinoblastoma.

Results. Twelve cases of retinoblastoma with the bipolar-like cell element in the absence of rosettes and 41 cases of the tumor with rosettes were classified as differentiated retinoblastomas. The other 35 cases without rosettes or bipolar-like cells were classified as undifferentiated tumors. Tumor cells forming rosettes usually had positive results for synaptophysin and neuron-specific enolase (NSE) and negative results for glial fibrillary acidic protein (GFAP) and S-100 protein; however, undifferentiated cells had negative results for these four antibodies. The bipolar-like cells had positive results for synaptophysin and NSE but negative results for GFAP and S-100 protein. Twelve tumors with bipolar-like cells that lacked rosettes showed no optic nerve invasion, and the patients had a significantly better prognosis (100% 5-year survival rate) than 35 patients with undifferentiated tumors (71% 5-year survival rate) ( P x 0.01).

Conclusions. The findings support a neuronal origin of the tumor and indicate that, not only the rosettes sym- bolizing the photoreceptor differentiation, but also other neuronal elements, such as bipolar-like cells, can be used

From the *Second Department of Pathology and tDepartment of Ophthalmology, Faculty of Medicine, Kyushu University, Fu- kuoka, Japan.

Address for reprints: Masazumi Tsuneyoshi, M.D., Second De- partment of Pathology, Faculty of Medicine, Kyushu University 60, 3-1-1 Maidashi, Higashi-ku, Fukuoka 812, Japan.

Accepted for publication March 25, 1992.

as criteria for histologic classification of retinoblastoma. Cancer 1992; 70:2901-8.

Key words: retinoblastoma, clinicopathology, immunohistochemistry, bipolar-like cell, synaptophysin, neuron-specific enolase, glial fibrillary acidic protein, S-100 protein.

Despite the continuous efforts of investigators, the histogenesis of retinoblastoma is still in dispute. The theory that the tumor is derived from glial cells of the retina first was proposed by Virchow in 1864’ and was supported later by other authors, who referred to the tumor as glioma of the retina.’ However, the old view of glial origin has been superseded by significant data supporting a neuronal origin, which were obtained by means of immunohistochemica13-8 and electron microscopic studies.9-” The results indicated that the tumor originated from neuroblastic cells of the retina and could differentiate toward photoreceptors of the normal retina. Other investigators, who studied retinoblastomas electron microscopically,’2-’4 immunohistochem- i~ally,’~,’~ by cell culture,”-’9 or by means of molecular techniques,20,21 further suggested that the tumor arose from primitive stem cells capable of differentiation along both neuronal and glial cell lines.” However, although many studies of this tumor indicated that formation of rosettes by retinoblastoma cells was one of the favorable prognostic f a ~ t o r s , ~ , ~ ~ - ’ ~ there does not seem to be a histologic classification satisfactorily predictive of To reassess histologic classifications related to the prognosis, we studied an appre- ciable number of cases of retinoblastoma clinicopathologically and immunohistochemically, directing our attention to “bipolar-like cells,” a well-differentiated

2902 CANCER December 25,2992, Volume 70, No. 12

element of this tumor, to discuss the histogenesis of the tumor.

Materials and Methods

A total of 122 specimens from 98 patients with retinoblastoma were collected in the Department of Ophthal- mology, Faculty of Medicine, Kyushu University, dur- ing the period 1958-1989 and have been examined in the Second Department of Pathology clinicopathologically and immunohistochemically in recent years. Histo- pathologic reports for these patients were reviewed, and the prognosis of these patients also was investigated. Among these specimens, 88 samples of forma- lin-fixed, paraffin-embedded tumors from 88 patients were available for this study, including 66 unilateral and 22 bilateral cases.

All paraffin blocks were recut and stained with hematoxylin and eosin for another histopathologic observation. Five-micron sections were mounted on the poly(L-1isine)-treated slides for the immunohistochemical study. They were processed with the avidin- biotin peroxidase complex method. Primary antibodies against the following antigens were used: neuron-specific enolase (NSE) (polyclonal, 1 :300; Dako, Glostrup, Denmark), synaptophysin (monoclonal, 1: 100; Boeh- ringer Mannheim, Biochemica, Mannheim), glial fibrillary acidic protein (GFAP) (polyclonal, 1 :400; Dako), and S-100 protein (polyclonal, 1:400; Dakopatts, Co- penhagen, Denmark). The site of localization of the peroxidase label was identified by the formation of a brown reactive product with the use of hydrogen peroxidase and diaminobenzidine. Negative controls were provided by omission of primary antibodies and positive controls by the normal retina in the eyes enucleated because of orbital tumors or by the uninvolved retina in the specimens of retinoblastoma.

The Kaplan-Meier method was used to estimate the survival curves, and generalized Wilcoxon test was applied to assess the differences between individual survival curves made according to the classifications proposed by the current study and the World Health Orga- nization (WHO) in 1980.29

Results

General Information

In the current study, the average age of patients with retinoblastoma, excluding the two older ages of 14 and 17 years, was 18.4 months, with the peak incidence at 1 year old (29.9%). The male to female ratio was 1.5:l.O. Laterality and age distribution of the patients are shown in Figure 1, with 66 (66.5%) unilateral and 22 (33.5%)

bilateral cases. All bilateral cases were found in the younger age groups (0-3 years old).

Gross Appearance

The tumor growth was exophytic in 10 cases and endo- phytic in 33. In 44 cases, a mixed exophytic and endo- phytic growth pattern was found in the same eye. In a 17-year-old patient, the tumor grew in a diffuse infil- trating pattern.

Microscopic Findings and Histologic Classification

For the 88 cases, the histologic typing of the tumor by differentiation and optic nerve invasion by the tumor are shown in Table 1. Regardless of types, the cells in general appeared undifferentiated, with hyperchromatic nuclei and scanty cytoplasm, showing little evidence of photoreceptor differentiation. In many cases, these cells created one or more of the three forms of rosettes in different amounts. Flexner-Wintersteiner rosettes were found in 41 cases and composed of radially arranged cuboidal or columnar cells with basal and cy- lindric nuclei and fine cytoplasmic extensions through the limiting membrane into the lumen (Fig. 2, left). The rosettes of this form were accompanied by two other forms of rosettes in some of the cases. Homer-Wright rosettes seen in 40 cases were composed of the tumor cells lining up around a tangle of fibrils (Fig. 2, middle). Fleurettes encountered in 10 cases were made up of tumor cells with small hyperchromatic nuclei and abundant pale eosinophilic perikaryon (Fig. 2, right).

In addition to these conventional cell types, other peculiar tumor cells were observed in 27 of the 88 cases. They were a little larger than the others and had small, round, centrally located nuclei and ample clear or pale- staining cytoplasm, presenting a well-diff erentiated appearance (Fig. 3, top). We provisionally termed these cells bipolar-like cells because they histologically re- sembled the bipolar cells in the normal retina (Fig. 3,

0 : B i l a t e r a l (N = 2 2 ) 0 : U n i l a t e r a l (N = 66)

Figure 1. Laterality and age distribution of patients with retinoblastoma.

Retinoblastoma/He et al . 2903

Table 1. Histologic Features of Retinoblastomas No. of Patients with patients ON1 (%) No. died (YO)

Differentiated RB 53 4 (10) 3 (8) With rosettes 41 4 (10) 3 (8)

(no rosettes) 12 0 (0) 0 (0) Undifferentiated RB 35 12 (34) 0 (29)

With bipolar-like cells

Total 88 15 (17) 13 (16)

47 tumors without rosettes, which would be classified as undifferentiated retinoblastoma according to the WHO typing, 12 had various amounts of bipolar-like cells. The current study classifies these 12 tumors with bipolar-like cells as one particular type of differentiated retinoblastoma, in addition to 41 tumors of the rosette- forming type. The other 35 cases without rosettes or bipolar-like cells are classified as undifferentiated tumors.

ONI: optic nerve invasion; R B retinoblastorna.

Immunohistochemical Findings

bottom). They usually proliferated in clusters or vague nests and in various amounts and were surrounded by or continuous to the adjacent undifferentiated or rosette-forming retinoblastoma cells directly or with an intervention of transitional cells. In several cases, the bipolar-like cells occupied most of the tumors. It is inter- esting that they scarcely were found in the areas forming rosettes and occasionally were seen adjacent to ne- crotic areas. These cells obviously were never enclosed retinal neurons because they were simply in prolifera- tion. The bipolar-like cells seemed to be much more differentiated toward neuronal cells than the conventional undifferentiated retinoblastoma cells, although the former cells did not form rosettes or show any differentiation -ta%vard photoreceptor cells. Among 41 cases with rosettes and fleurettes commonly classified as differentiated tumors according to WHO typingz9 15 cases with bipolar-like cells were encountered. Of the

The immunohistochemical results are shown in Table 2. Neuron-specific enolase. In the normal human ret-

ina, ganglion cells and bipolar cells had positive results for NSE. Photoreceptor cells had only weakly positive results, whereas the rod and cone had strongly positive results. Both inner and outer plexiform layers stained positively, although the latter stained more strongly. Glial cells had negative results for NSE (Fig. 4, left). In retinoblastomas, 58 (66%) of the 88 cases displayed tumor cells reactive to NSE. All cases of well-differentiated retinoblastoma contained reactive cells forming Homer-Wright rosettes, Flexner-Wintersteiner rosettes (Fig. 4, middle), or fleurettes. The bipolar-like cells showed positive cytoplasmic staining, the reactivity usually being stronger than that of rosettes and fleurettes (Fig. 4, right). Undifferentiated retinoblastoma cells appearing in both differentiated and undifferentiated retinoblastoma had negative results for NSE.

Figure 2. Photomicrographs of differentiated retinoblastoma with (left) Flexner-Wintersteiner rosettes (H & E, original magnification X490), (middle) Homer-Wright rosettes (H & E, original magnification X490), and (right) fluerettes (H & E, original magnification X500).


Figure 3. (Top) Bipolar-like cells adjacent to undifferentiated retinoblastoma cells (H & E, X260). (Bottom left) Higher magnification of bipolar-like cells (H & E, X450). (Bottom right) Stratified structure of the normal retina with bipolar cells in the inner nuclear layer (H & E, X450).

Synaptophysin. In the retina, the three types of neurons had negative results for synaptophysin, and the two plexiform layers with ball-like or stalk-like pedi- cles had strongly positive results. The pedicle structures were not delineated in the specimens stained with NSE. In 55 (62%) of the 88 retinoblastoma cases, the tumor cells had positive results for synaptophysin. The rosettes had positive results for synaptophysin, although

the fleurettes were not reactive. The undifferentiated tumor cells had negative results for synaptophysin and NSE. Dense immunoreactive products for synaptophysin were demonstrated on ball-like or stalk-like processes in areas composed of bipolar-like cells (Fig. 5, left). The similar immunopositive structures also were found occasionally in areas of rosettes.

Glial fibrillary acidic protein. In the normal ret-

Retinoblastoma/He e t al. 2905

Table 2. Immunohistochemistry in the Normal Retina and in Retinoblastomas

NSE SNS GFAP s-100

Retina Photoreceptor Cone and rod Bipolar cell Ganglion cell Astrocyte Miiller cell

Retinoblastoma Rosette BLC UDC

+ ++ ++ ++

++ +++ -

+ ++ -

-

- +

++ +

NSE: neuron-specific enolase; SNS: synaptophysin; GFAP: glial fibrillary acidic protein; 5-100: 5-100 protein; BLC: bipolar-like cells; UDC: undifferentiated cells; +: weakly positive; ++: moderately positive; +++: strongly positve; -: negative.

ina, Muller cells with long fibers between photoreceptor cells and astrocytes or nerve fibers in the inner fiber layer stained positively for GFAP. No neuronal cells had positive results. In the retinoblastomas, no tumor cells had positive results, although more than half of the tumors contained nontumorous cells with positive results for GFAP. The GFAP-positive cells were situ- ated predominantly in the perivascular areas, probably as reactive astrocytes (Fig. 5, right). These cells were

recognized more often in well-diff erentiated tumors than in undifferentiated retinoblastomas, with 12 of the 35 undifferentiated cases demonstrating no reactive astrocytes. The areas composed of bipolar-like cells had negative results for GFAP.

S-100 protein. In the normal retina, the glial cells stained positively for S-100 protein and GFAP. Large ganglion cells in the ganglion cell layer, which were associated with nerve fibers, also stained positively. Two other kinds of neuronal cells, bipolar cells and photoreceptors, had negative results. In the retinoblastomas, cells with positive results for S-100 protein were seen frequently. Most of these cells seemed to be perivascular reactive astrocytes, as shown by the GFAP staining. Tumor cells forming rosettes also had negative results. In two cases, tumor cells had weakly positive results for S-100 protein; these cells had positive results for NSE and synaptophysin but negative results for GFAP. All undifferentiated retinoblastoma cells had negative results for S-100 protein.

Prognosis Wi th Statistical Analysis

The age and sex of patients and the laterality and gross appearance of tumors did not appear to exert any imme- diate or vital influence on prognosis.

All 12 patients with differentiated tumors contain- ing bipolar-like cells in the absence of rosettes survived

Figure 4. Immunoreactivity to NSE. (Left) The normal human retina. Ganglion cells and bipolar cells are positive for NSE. Photoreceptor cells are weakly positive, whereas the rod and cone are strongly positive. No glial cells stain positively (original magnification X280). (Middle) Well-differentiated retinoblastoma. Flexner-Wintersteiner rosettes stand out positively (original magnification XSOO). (Right) Bipolar-like cells are strongly positive (original magnification X330).


100

90-

80- x 7 0 ~ m 60- ' 50-

._ 40- 2

30-

20- 10-

- v

c

-

5 years, although the 5-year survival rate of 35 patients with undifferentiated retinoblastoma was 71 YO ( P < 0.01) (Fig. 6). This result is the basis on which we classified the above 12 cases as differentiated retinoblastoma, despite the absence of rosettes, in addition to the 41 differentiated tumors with rosettes or fleurettes. The survival curves by the trial in the current study were analyzed statistically and compared with the WHO classification of retinoblastoma. According to the current trial, the differentiated retinoblastoma had a significantly better prognosis (94%) than the undifferentiated retinoblastoma (71%) ( P < 0.05) (Fig. 7), whereas the

--A

B

Figure 5. Immunoreactivity to other markers in retinoblastoma. (Left) Ball- like or stalk-like processes in the areas composed of bipolar-like cells are strongly stained for synaptophysin (X400). (Right) Reactive proliferating astrocytes are positive for GFAP, but tumor cells are negative (X250).

80- X 70-

m 60- ' 50-

.$ 40-

,-. v

c

-

difference in the 5-year survival rate was not significant between patients with differentiated (92%) and undifferentiated (79%) retinoblastoma according to the classification by the WHO.

Discussion

The results obtained in this study may provide additional support for the neuronal theory regarding the origin of retinoblastoma. Certain histologic similarities between differentiated retinoblastoma cells and normal retinal neurons were indicated by the presence of rosettes and bipolar-like cells, whereas no similarity was

I .

'""I 90 h4-

L -B

(P < 0.01)

q , , , , , ,

0 1 2 3 4 5 6 Time (Years)

Figure 6. Survival curves of (A) 12 cases of retinoblastoma with bipolar-like cells that have statistically better prognosis (100% 5- year survival rate) and (B) 35 cases of undifferentiated retinoblastoma (71 Yo 5-year survival rate). The difference is significant ( P < 0.01).

(P < 0.05)

Retinoblastoma/He et al . 2907

observed between well-differentiated retinoblastoma cells and glial cells. Immunohistochemical profiles of some tumor cells in well-diff erentiated retinoblastomas were similar to those of the normal retina. As already shown by other investigator^,^-^*^*^ rosettes had positive results for neuronal markers, such as NSE and synaptophysin, but negative results for glial markers, such as GFAP and S-100 protein. The bipolar-like cells had strongly positive results for NSE and synaptophysin but negative results for GFAP and 5-100 protein, indi- cating that these cells were also neuronal. The positive staining of ball-like or stalk-like processes was considered evidence of synapse formation4 and an additional demonstration of the neuronal differentiation because such structures never were found in glial cells.30 No cases had tumor cells with positive results for GFAP.

The resemblance of the above bipolar-like cells microscopically and immunohistochemically to bipolar cells in the normal retina and the clinicopathologic im- plication for better prognosis were documented in this study. Embryologically, the retina is made up of neuro- blasts forming photoreceptors and various neuronal cells, including ganglion cells and bipolar cells. The photoreceptors mature later than the ganglion cells and even bipolar ~ e l l s . ~ * , ~ ’ In the development of embryonic sensory retina, the bipolar cells and photoreceptors orig- inate from the same outer neuroblastic layer. The ganglion cells and amacrine cells arise from the same inner neuroblastic layer.32 Retinoblastomas, which are composed of tumor cells comparable to primitive retino- blasts, are well known for the formation of rosettes with cells similar to embryonic photoreceptors and the formation of amacrine cells.33 However, questions arise about other neuronal cells, such as whether there is a possibility that the rosette formation is accompanied by simultaneous differentiation toward ganglion or bipolar cells. Craft et al. suggested that various modes of retinoblastoma cell differentiation to ganglion cells or bipolar cells would be further investigated.” Synapses formed by bipolar cells and photoreceptors, as seen in the inner plexiform layer of the normal retina, also have been detected in the r e t i n o b l a ~ t o m a . ~ ~ , ~ ~ To the best of our knowledge, there have been no well-documented articles on the differentiation of retinoblastoma cells toward neuronal cells other than photoreceptors. The advent of bipolar-like cells in the retinoblastoma, as con- firmed in the current study, indicates a possibility of bipolar cell differentiation.

Although many studies indicated that rosette formation in retinoblastoma was an important factor regarding favorable progn~sis,~,’~-’~ some investigators have insisted that the presence or absence of rosettes alone is not essential for prediction of prognosis.26-28 We found no significant difference in prognosis be-

tween differentiated and undifferentiated tumors according to the classification by the WHO based only on the presence or absence of rosettes. According to our classification proposed in this study, however, there was a significant difference in the 5-year survival rate between patients with differentiated and undifferentiated retinoblastomas. The difference in the 5-year survival rate with the WHO classification certainly is caused by using bipolar-like cells as a criterion in the current study, in addition to rosettes. We classified 12 of the cases with tumors showing no rosettes, but having bipolar-like cells, as well differentiated, although they originally would have been classified as undiff erentiated tumors according to the WHO classification. The current classification is considered to be reasonable because the above 12 cases lacked optic nerve invasion by the tumor. In contrast, optic nerve invasion was found in approximately one-third (12 of 35) of undifferentiated tumors in this study. It is well known that optic nerve invasion sharply worsens the progn~sis.’’-’~ However, other prognostic factors, such as patient age and sex and the laterality and gross appearance of tumors, were not correlated intimately with patient survival in the study.

In summary, the identification of bipolar-like cells in cases without rosettes and the histologic classification based on the differentiation toward such neuronal cells, in addition to the presence of cells forming rosettes, have been proven to be helpful in predicting the biologic behavior of a retinoblastoma.

References

1.

2.

3.

4.

5.

6.

7.

8.

9.

Virchow R. Die Krankhaften Geschwiilste. vol. 2. Berlin: August Hirschwald, 1864: 15 1. Parkhill EM, Benedict WL. Glioma of the retina: a histopathologic study. Am ] Ophthalmol 1941; 24:1354-73. Perentes E, Herbort CP, Rubinstein LJ, Herman MM, Uffer S, Donoso LA, et al. Immunohistochemical characterization of human retinoblastomas in situ with multiple markers. Am J Oph- thalmol 1987; 103:647-58. Kivela T, Tarkkanen A, Virtanen I. Synaptophysin in the human retina and retinoblastoma: an immunohistochemical and West- ern blotting study. Invest Ophthalmol Vis Sci 1989; 30212-9. Lane JC, Klintworth GK. A study of astrocytes in retinoblastomas using the immunoperoxidase technique and antibodies to glial fibrillary acidic protein. Am J Ophthalmol 1983; 95:197- 207. Albert DM, Rabson AS, Dalton AJ. Tissue culture study of human retinoblastoma. Invest Ophthalmol 1970; 9:64-72. Kivela T. Neuron-specific enolase in retinoblastoma: an immunohistochemical study. Acta Ophthalmol 1986; 64:19-25. Kivela T, Tarkkanen A. S-100 protein in retinoblastoma revi- sited: an immunohistochemical study. Acta Ophthalmol 1986;

Ts’o MOM, Zimmerman LE, Fine BS, Ellsworth RM. A cause of radioresistance in retinoblastoma: photoreceptor differentiation. Trans Am Acad Ophthalmol Otolaryngol 1970; 74:959-69.

64:664-73.


10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

Ts’o MOM, Fine BS, Zimmerman LE. The Flexner-Wintersteiner rosettes in retinoblastoma. Arch Pathol 1969; 88:664-71. Ts’o MOM, Zimmerman LE, Fine BS. The nature of retinoblastoma: I. Photoreceptor differentiation. A clinical and histopathologic study. A m J Ophthalmol 1970; 69:339-49. Craft JL, Sang DN, Dryja TI‘, Brockhurst RJ, Robinson NL, Al- bert DM. Glial cell component in retinoblastoma. E x p Eye Res

Ohnishi Y. The histogenesis of retinoblastoma: an electron-microscopic analysis of rosette. Ophthalmologica 1977; 174:129- 36. Radnot M. On the histogenesis of retinoblastoma. ] Pediatr Oph-

Terenghi G, Polak JM, Ballesta J, Cocchia D, Michetti F, Dahl D, et al. Immunocytochemistry of neuronal and glial markers in retinoblastoma. Virchows Arch [A] 1984; 404:61-73. Kyritsis AP, Tsokos M, Triche TJ, Chader GJ. Retinoblastoma: origin from a primitive neuroectodermal cell? Nature 1984;

Jiang Q, Lim R, Blodi FC. Dual properties of cultured retinoblastoma cells: immunohistochemical characterization of neuronal and glial markers. Exp Eye Res 1984; 39:207-15. Kyritsis AP, Tsokos M, Triche TJ, Chader GJ. Retinoblastoma: a primitive tumor with multipotential characteristics. Invest Oph- thalmol Vis Sci 1986; 27:1760-4. Tsokos M, Kyritsis AP, Chader GJ, Triche TJ. Differentiation of human retinoblastoma in vitro into cell types with characteristics observed in embryonal or mature retina. A m J Pathol 1986;

Herman MM, Perentes E, Katsetos CD, Darcel F, Frankfurter A, Collins VP, et al. Neuroblastic differentiation potential of the human retinoblastoma cell lines Y-79 and WERI-Rbl main- tained in an organ culture system: an immunohistochemical, electron microscopic, and biochemical study. A m ] Pathol 1989;

1985; 40~647-59.

thalmol 1975; 12:259-63.

307:471-3.

123~542-52.

1341115-32.

21.

22. 23.

24.

25.

26.

27.

28.

29.

30.

31.

32. 33.

34.

Van Veelen CWM, Staal GEJ. How did retinoblastoma arise? Nature 1984; 310:731. Garner A. Commentary. Histopathology 1989; 15:199-202. Tsukahara I. A histopathologic study on the prognosis and radiosensitivity of retinoblastoma. Arch Ophthalmol 1960;

Yanoff M, Fine BS. Retinoblastoma and pseudoglioma. In: Yan- off M, Fine BS, editors. Ocular pathology. 3rd ed. Philadelphia: JB Lippincott, 1989:684-94. Shields JA, Augsburger JJ. Current approaches to the diagnosis and management of retinoblastoma. Sum Ophthalmol 1981;

Sevel D, Rohm GF, Sealy R. Clinical significance of the fluerette in retinoblastoma. Br J Ophthalmol 1974; 58:687-93. Kobrin JL, Blodi FC. Prognosis in retinoblastoma: influence of histopathologic characteristics. J Pediatr Ophthalmol Strabismus

Mashiah M, Barishak YR. Photoreceptor differentiation in retinoblastoma and its significance in prognosis. B r J Ophthalmol 1977;

Zimmerman LE, Sobin LH. Histological typing of tumours of the eye and its adnexa. International histological classification of tumours. vol. 24. Geneva: World Health Organization,

Radnot M. Synaptic lamellae in retinoblastoma. Am J Ophthal-

Hogan MJ, Alvarado JA, Weddell J. Histology of the human eye. Philadelphia: WB Saunders, 1971:399-410. Ogden TE. Retina. vol. 1. St. Louis: CV Mosby, 1989:5-16. Sang DN, Albert DM. Retinoblastoma: clinical and histopathologic features. Hum Pathol 1982; 13:133-47. Dickson DH, Ramsey MS, Tonus JG, Synapse formation in retinoblastoma tumours. Br ] Ophthalmol 1976; 60:371-5.

63~131-4.

25:347-72.

1978; 151278-81.

61 ~417-22.

1980:64-5.

mol 1975; 79~393-404.

Documents

A reassessment of histologic classification and an immunohistochemical study of 88 retinoblastomas. A special reference to the advent of bipolar-like cells