JOURNAL OF PATHOLOGY, VOL.153:233-244 (1987)

CELLULAR COMPOSITION OF FOLLICLESOF FOLLICLE CENTRE CELL LYMPHOMASIN RELATION TO GERMINAL CENTRES OF REACTIVE LYMPH NODES. A MORPHOMETRICAL ELECTROMICROSCOPICAL STUDY

J. P. J. PETERS*, L. H. P. M. RADEMAKERS*, R. J. DE BOERt AND J. A. M. VANUNNIK* * PathologischInstituut, University of Utrecht, Pasteurstraat 2, 3511 HX Utrecht, The Netherlands. tBioinformatics Group, University of Utrecht, Padualaan 8, 3584 CH Utrecht, The Netherlands.

Received 20 November 1986 Accepted 24 June 1987 SUMMARY The cell spectrum of neoplastic and benign reactive germinal centres wasdetermined ultrastructurally. The degree in which the cell composition found in reactive germinal centres is maintainedin analogous structures of follicular was investigated by pattern recognition methods and discriminantanalysis based on the frequencies of the various lymphoid and non-lymphoid cell types. The follicular lymphomasincluded lymphomas with predom- inantly (FCCL, Cb-cc) and lymphomas with predominantlycentroblasts (FCCL, Cb). Pattern analysis of FCCL Cb, FCCL Cb-cc and reactive germinal centresindicates that FCCL Cb follicles resemble reactive germinal centres in more aspects than follicles of FCCL Cb-cc. Clear statistical differences were encountered between the frequencies of thelymphoid cell types and of the follicular dendritic and histiocytic reticulum cells in follicles of FCCL Cb-cc and FCCL Cband reactive germinal centres. Application of a discriminant analysis using a combination of the frequency of centrocytesand follicular dendritic cells demonstrated that both types of neoplastic follicles and reactivegerminal centres were correctly classified on the basis of their cell spectrum. For the three groups the most potentdiscriminator was the , whereas the small and large centroblast were of less value. For discriminationbetween Cb-cc follicles and reactive germinal centres again the centrocyte was the most potent discriminator.Discrimination between FCCL Cb follicles and reac- tive germinal centres of FCCL Cb-cc follicles can be easily achievedusing the frequencies of small or centrocytes on their own. These findings indicate that (1) follicles of both FCCL Cb and FCCI Cb-ccdiffer greatly in the cellular composition and not only with respect to their content of centroblasts but also in their contentof follicular dendritic cells and (2) they may be considered as neoplasms representing different developmentalphases of germinal centres. KEYWORDSGerminal centre, follicle centre cell lymphomas, ultrastructure, morphometry

INTRODUCTION Rappaport et al.,' Lennert2 and Nathwani et al.3 Most of these criteria refer to changes in lymph The differentiation between follicular hyperplasianode architecture, though the absence of prominent and follicular is often a difficult problemphagocytosis and the uniformity of the population in the histopathology of lymphoid tissue. Histologicof the follicle centre cell population have to be taken criteria for differentiation have been defined byinto consideration. Apart from these morphological criteria the immuno-histochemical demonstration of immunoglobulin light chain restriction is prob- Addressee for correspondence: L. H. P. M. Rademakers, Pathologisch Instituut, University of Utrecht, Pasteurstraat 2, ably the most reliable criterion for differentiating 3511 HX Utrecht, TheNetherlands. between reactive and neoplastic follicles. 0022-3417/87/110233-12$06.00 ©1987 by John Wiley & Sons, Ltd. 234 J. P. J. PETERS ET AL.

The relationship of follicular lymphomas with Data analysis normal germinal centres has been stressed by histo- logica1,4 andimmunohistologicalstudies... The data were analysed by pattern recognition Ultrastructurally a remarkably wide spectrum ofmethods and several statistical procedures. lymphoid cells and of non-lymphoid cells is encoun- The data of the frequencies of the various cell tered in normal germinal centres.8 Follicular lym-types were subjected to a clusteranalysis:a pattern phomas can be classified according to the relative recognition method which aims to reveal the most proportion of germinal centre cells (centrocytes or striking pattern in a set of data (in this investigation centroblasts) and the size of the proliferatingthe distinct biopsies characterized by a number of cells.',2,9,1°In most follicular lymphomas theparameters). Cluster analysis is a non-supervised lymphoid cells in the follicular structures can bemethod :i.e. grouping of data is achieved by the related to normal germinal centre cells" ,1 2 but datamethod without intervention of the investigator. referring to the degree in which this spectrum isThe method, using Biopat," thus classifies the data maintained in follicular lymphomas containing into the most relevant groups. The outcome of the varying numbers of centroblasts are lacking. In thiscluster analysis is depicted in a dendrogram. A den- study we compare the cell composition of reactivedrogram shows the similarity level at which the and neoplasticfolliclesusing morphometricalrespective objects (here biopsies) and groups of electron microscopy in combination with patternobjects are combined into new groups (i.e. clusters). recognition methods. Cluster analysis is explained in more detail by Sneath and Soka1.16 The statistical procedures attempt to find dif- MATERIAL AND METHODS ferences in individual parameters between groups Lymph nodes which are predefined (i.e. supervised) by the investi- gator. The differences in frequency of the various Fresh biopsies were processed forcell types between reactive and lymphomatous folli- multidisciplinary studies according to the protocolcular structures were assayed by the Wilcoxon test. described by Van der Putte et al." This protocolThis procedure was applied at a probability level included routine histology, cytology, immunohisto-of 5 per cent. chemistry and enzyme histochemistry on frozen sec- Although the data of an individual parameter tions and enzymatically prepared cell suspensions,(cell type) may differ significantly between the and electron microscopy. 14 histopathological groups, this does not imply that Fifteen lymph nodes with follicular lymphomathe biopsies can be classified correctly on the basis and eleven lymph nodes with follicular hyperplasiaof this statistical difference. This problem can be were selected for our biopsy files on the basis ofresolved by subjecting the data of the various histopathologicaldiagnosis and immunohisto-parameters to a discriminantanalysis;a method chemistry. Histologically twelve cases were cate-that analysesthe data of several parameters gorisedas folliclecentre cell lymphomaconcomitantly. centroblastic centrocytic, follicular (FCCL Cb-cc) andthreeasfolliclecentrecelllymphoma centroblastic, follicular (FCCL Cb). RESULTS Morphometry In reactive germinal centres a large number of cell The lymphoid and non-lymphoid cell types weretypes can be identified at the ultrastructural level. identified in reactive germinal centres on the basis A number of these lymphoid and non-lymphoid cell of their ultrastructure. The relative frequency oftypes are illustrated in Fig. I. these cell types was determined on low power The following celltypes were encountered : electron micrographs. For this purpose three folli-, large ( > 12 pm in diameter) and cular structures in each biopsy were selected in 1 pm small centroblast ( < 12 pm in diameter), centro- sections. In ultrathin sections a set of six electroncyte, multilobated cell, centroplasmacytoid cell, micrographs (magnification 3700 x ) were lymphocyte, , follicular dendritic cell, randomly taken from each follicular structure.histiocytic reticulum cell, fibroblastic reticulum cell These micrographs represented a total area ofand dark reticulum cell. 14 000 pm2 per biopsy. The ultrastructural distinction of the above men- CELLULAR COMPOSITION OF FOLLICULAR LYMPHOMAS 235

Fig. I -Electron micrograph of a reactive germinal centre (biopsy R5) showing the extensive web of cytoplasmic extensions of folliculardendritic cells (FDC) aroundthe lymphoid cells. The lymphoid cell population isheterogenous; large centroblasts (CB), centrocytes (CC) and centroplasmacytoid cells (CPC) are indicated. Note the similar chromatin pattern of the latter two cell types.Asterisk lymphocyte;arrowsnuclei of FDC. x 4000 tioned cell types was based for most cell types on Lymphocytes are defined as round small cells. descriptions by Lennert andMiiller-Hermelink8 Their nucleus contains heavy marginal and central and Lennert.2 chromatin condensations. Cell organelles are sparse Multilobatedcells' 3have marked nuclear lobula-in thecytoplasm;only occasionally clusters of tions, a heterochromatic chromatin pattern and aelectron dense granules occur. cytoplasmic organization of organelles comparable Dark reticulumcells' 8are cells of various shapes with those of centrocytes. with an electron dense cytoplasm. In our material Centroplasmacytoid cells also resemble centro-the majority of these cells represent degenerating cytes, but have a well developed rough endoplasmicfollicular dendritic cells, in few cases histiocytic and reticulum and Golgi complex. These cells probably fibroblastic reticulum cells. represent the Ig-producing centrocytes previously In neoplastic follicles most of the above men- reported by Lennert and Stein." tioned cell types could be recognized on the basis 236 J. P. J. PETERS ET AL.

31.11, tr. - rC s c.%

:1' e

-06 =

`1:43 -N.

W , °

1 :

_

.x

.4 A

"44 -VI -1()IN.1-!`, . ,; . salt gra. -6- rj .44 le; .f a ,

. .0.41-;1'1 -1.7 , , !hull Ey . . 4-- ' Fig. 2Centroblastic-centrocytic (biopsy CB-CC 6). Neoplastic follicle with predominantly centrocytes (asterisk); CB: centroblast; HRC: histiocytic reticulum cell; LC: Lymphocyte; Arrows--broad cytoplasmic extensions of FDC. x 4000 of their ultrastructural characteristics (Fig. 2). In In FCCL Cb the centrocytes were larger and also FCCL Cb-cc the malignant centroblasts andthe centroblasts showed pronounced nuclear inden- centrocytes often had a strong resemblance to theirtations. However, such cells could be identified benign counterparts, whereas in other cases thesewithout difficulty (Fig. 3). cells were more irregular and had more pronounced chromatin condensations. In two cases nuclearUltrastructural analysis of the cell population in pockets were frequent in the nuclei of centrocytes.follicular structures In all these cases the centrocytes were smaller com- No clear polar distribution of germinal centre pared with their analogues in reactive germinalcells with respect to the light and dark zones as des- centres. The histiocytic reticulum cells and follicularcribedbyLennert2was observedbylight dendritic cells were morphologically different frommicroscopy in most of the reactive germinal centres. those in reactive germinal centres as describedOnly in a few cases was a slight peripheral previously.' 4 accumulation of centroblasts observed. A polar dis- CELLULAR COMPOSITION OF FOLLICULAR LYMPHOMAS 237

t

er'14- 41,,_ 2. ''.1r-ls C,L- '-.. 4C 'i '1 r-t .7i HI ie - 4 ,,, ..,, e .% ; 4,4--.zovii ,. , ,. ,( II, ._ 1 r=

1,1 AD 4 . 4 fr411.--a- 8.2211.aatt' Fig.3 Centrohlasticfollicular lymphoma (biopsy t_ B 14); neoplastic follicle. Small (asterisk) and large centroblasts (CB) are abundant. Centrocytes (CC) are of larger size than usual. Note FDC with broad cytoplasmicextensions. x 4000 tribution with respect to histiocytic reticulum cellsagglomerativeclusteranalysiswithWard's was also not evident in semithin sections. For thecriterion."' Note that the histopathological class of analysis of the cell population therefore the samplesthe biopsies is not taken into account in this (non- represented randomly intermingled types of germi-supervised) method : the biopsies are grouped on the nal centre cells. basis of their cell counts only. This analysis yields The data obtained after morphometrical analysisthe dendrogram depicted in Fig. 4. The dendrogram of reactive germinal centres (R) of each biopsy areclearly shows 3 groups (clusters) which largely cor- given in Table I a. The data on neoplastic germinalrespond to the 3 histopathological groups (reactive, centres are presented in Table 1 b. FCCL Cb and FCCL Cb-cc). Only germinal centres of the reactive lymph nodes RI and R6 are placed Cluster analysis 'incorrectly', in the FCCL Cb-cc cluster. These two The data, i.e. the 26 biopsies characterized bybiopsies have a relatively high centrocyte count. their respective cell counts, were subjected toThe fact that this non-supervised method yields the 00lJ Table laRelative frequency (%) of distinct cell types in normal germinal centres of reactive (R) lymph nodes R I R 2 R 3 R 4 R 5 R 6 R 7 R 8 R 9 RIO RII SmallLargeImmunoblast centroblast 70 730.6 631614 1 6013 3 1 5917 70.7 930.1 81 60.50.2 521510 4 784 21 53 14 81 20 60.3 MultilobatedCentroplasmacytoidCentrocyte cell 5 1 1 0.56 0.73 4414 3 0.51 0.24 60 3 1 46 7 1 66 0.62 1 44 0.90 PlasmaLymphocyte cell 06 0 1 50-51 0.24 14 80.5 0 40.2 7 1 30.54 0 15 0 HistiocyticFibroblasticFollicular reticulum dendritic reticulum cell cell cell 022.8 0/ 07 04 03 047 40.26 041 0.37 402 047 TotalMitosisDark14,000pm2 reticulumnumber of lymphoid of cell cells/ cells 503 2 1 450 0 1 441 30 426 20 406 0 1 399 0 1 397 0.40 391 1 365 0.30 333 0.90.3 338 0 1 Table 1 b Relative frequency (%) of distinct cell types in malignant follicles of follicular lymphomas of the centroblastic-centrocytic (Cb-cc) and centroblastic type (Cb) CB-CC CB-CC CB-CC CB-CC1 CB-CC CB-CC CB-CC CB-CC CB-CC CB-CC CB-CC CB-CC CB CB 5 11 12 13 14 CB 15 Immunoblast 1 2 2 1 3 1 4 0 3 6 0.8 7 0.2 8 1 9 0-810 0.8 1 I 0 0 0.5 CentroplasmacytoidCentrocyteSmallLarge centroblast 87 0.91 76 45 92 32 92 30 94 0.21-2 77 23 92 30.5 85 42 91 20.9 92 50.8 83 63 83 3 2257 0 234513 3430 9 cellMultilobated cell 08 720-4 0 11.5 20 00.2 36 1 0 1 0 1 00-70.7 00.81 20.3 30 1 72 15 06 608 PlasmaFollicularLymphocyte cell dendritic 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 cellHistiocytic reticul um 0 1 0.83 0 0 3 0.3 0.2 2 1 22 35 2 61.5 42 Fibroblasticcell reticulum 0 0.5 02 01.2 0.61 00.3 02 0.90 0 0 0 0 0 0.5 cellDark reticulum 0-30.7 0 02 0 0 0.6 1 6 0 0.8 0 1 0 0 0 TotalcellsMitosis 14,000pm2number of lymphoid of cells/ 647 0 524 0 514 0 492 0 488 0.2 468 0-2 468 0 468 1 462 0 400 0 369 0.6 367 1 384 200 0 0.5 207 0 240 J.P. J.PETERS ET AL.

_Ltr__' 1 RO1 13CBCCO2 17CBCCO6 6 606 22COCCI! 1E23 CBCC12 19CBCCO8 12CBCCO1 14 CBCCO3 20 CBCCO9 15CBCCO4 Lh 18 CBCCO7 21CBCCIO 16 CBCCO5 2 802 10 R10 3 003 4 004 1 7 007 8 ROB 5 005 9 RO9 11 All 24 CB13 25 C814 I E7 26 C815

48.00 42.00 36.00 30.00 I 24.00 18.00 12.00 6.00 0.00 45 00 39 00 33.00 27 00 21 00 15 00 9.00 3 00 Fig. 4Dendogram of an agglomerative cluster analysis of reactive germinal centres and neoplastic follicles. Vertical axis: germinal centres of the 26 lymph node biopsies characterized by 13 parameters and grouped according the similarity of their data. Horizontalaxis:dissimilarity level at which the objects are grouped. This level is determined by Wards criterion° and based on the mean city block distance.23 The 3 histopathological classes of follicles can be recognized in this diagram.

3 groups distinguished beforehand leads us to con-showedsignificantlyhighervaluesforsmall clude that the 3 groups differ clearly from each othercentroblasts, multilobated cells, lymphocytes and in cell composition. These differences between thefollicular dendritic cells. The data for centrocytes groups were investigated by statistical (supervised) were significantly lower. methods. In addition, significant differences were observed in the total number of cells/area between the three Statistical analysis histopathological groups indicating differences in cell size, since no gaps were present between the The mean values, range and the results of the cells. statistical evaluation of the differences between the data of each group are listed in Table II. Statistical analysis of the data with the Wilcoxon Discriminant analysis test of the distinct cell types revealed differences Discriminant analysis, using a subset of char- between the groups. Highly significant differencesacters (cell types), of the biopsies divided into the in the frequencies of the various parameters were3 histopathological groups, allows a 100 per cent observed between the reactive germinal centres andsuccessful classification of all biopsies. We first FCCL Cb-ccfollicles:the frequency of centrocyteschecked the importance of each parameter individu- in the FCCL Cb-cc follicles were significantly higherally. It appeared that if we only consider the centro- compared with the reactive-group, whereas thecyte count of each biopsy 25 (out of 26, i.e. 96 per numbers of centroblasts, centroplasmacytoid cells,cent successful classification) biopsies are classified follicular dendritic cells and histiocytic reticulum correctly (the reactive node R6 is again misclassified cells were significantly lower. This was also appar-to belong to the FCCL Cb-cc group). ent for the data on mitotic figures. It thus appears that knowledge of the centrocyte In follicles of FCCL Cb the frequencies ofcount only is almost (ie. for 96 per cent) sufficient , centrocytes, histiocytic reticulum for classifying any biopsy into its correct group. The cells and mitotic figures were significantly lower, insecond best is the small centroblast which classified comparison with the reactive germinal centres23 (89 per cent) of the biopsies correctly (R 1, R6 whilst the figures obtained for multilobated cellsand R8 are misclassified). All other cell types are and small centroblasts were higher. Compared withof no value for classification of the biopsies. Next the FCCL Cb-cc follicles the FCCL Cb folliclesa combination of parameters was tested :i.e. we Tablelymphomas of the centroblastic-centrocytic type (Cb-cc) and centroblasticII type (Cb). Mean values and range of the frequency of cell types occurring in follicular structures of hyperplastic lymph nodes (R) and follicular Group 1 n = 11 'CB-CC'Group 2 n = 12 Group n'CB' = 3 3 I vs 2 Wilcoxon test 1 vs 3 2 vs 3 Immunoblast i1.4 Range 0- 4 R 1.01.5 Range 1- 3 i0.2 Range 0- 0.5 CentroplasmacytoidCentrocyteSmallLargeMultilobated centroblast cell cell 5913-2 0.94.26.1 0.5-0.5-14 344-81 0-146-21 87 0-80.83.2 76-940- 0-3 1-56 6 451814 4.62-3 34-57 2-09-230-30 76 +*- -++* * +*-++ HistiocyticFollicularPlasmaLymphocyte cell dendritic reticulum cell cell 4-14-50-35-1 0-0-15 1-1 8 02-11-61.1 0- 38 - 10 4.001.2 0- 2-7-152 6 - +*-- +- * --+ TotalMitosisDarkFibroblastic reticulumnumber of lymphoid reticulum of cells cells/14,000 cells cell pm2 404 0.30-11.2 333-5140.3- 3 0-2- 0-372 476 0.20.80-4 367-647 0-0 2136 263 0-20 200-384 0 0.50- 0.5 - +*-+- -++ -+ : mean value; n : number of biopsies; + : significant difference (P<0.05);*P<0.01. 242 J. P. J. PETERS ET AL. tested the centrocyte count in combination withparameters centrocyte and follicular dendritic cell. every other parameter. It turned out that only theAfter checking up the clinical background,it combination of centrocyte and follicular dendriticappeared that R6 was the initial lymph node biopsy cell, although the latter cell type has no discriminat-of a patient who developed in the next 5 years ing value of its own, gives a 100 per cent correctpersistentgeneralizedlymphadenopathyand classification. symptoms of AIDS. Discriminant analysis of the reactive germinal Statistical analysis of each parameter separately centres versus FCCL Cb-cc follicles also indicatesshows that FCCL Cb-cc are a rather homogenous that the centrocyte and small centroblasts are thegroup in which centrocytes predominate. In con- best discriminators giving a correct classification oftrast the FCCL Cb appear to have a more hetero- 22 (96 per cent) and 20 (80 per cent) out of 23 bi-genous cellular composition. opsies respectively ; however the same mistakes The cellular composition of reactive germinal mentioned above are made. The centrocytes andcentres may range between a relative high number small centroblasts are followed in order of import-of centroblasts and a predominance of centrocytes ance by histiocytic reticulum cells, mitoses, folli-(Table la). Lennert2 described four phases of germi- cular dendritic cells, centroplasmacytoid cells andnal centre development :(i) a phase lasting about large centroblasts. Only with a combination of24hourswithpredominantly mediumsized centrocytes and follicular dendritic cells is a 100 percentroblasts; (ii) a phase, lasting one to three weeks, cent successful classification obtained. with predominantly centroblasts and 'starry sky' Analysis of the reactive germinal centres versusmacrophages; (iii) a phase lasting one to three FCCL Cbfolliclesshowsthatbothsmallmonths with an occasional zonal distribution and centroblasts and centrocytes on their own allowthe occurrence of immunoblasts, centroblasts, 1 successful (100 per cent) classification of the bi-centrocytes, plasma cells, follicular dendritic cells 1 opsies. Also FCCL Cb-cc follicles versus FCCL Cband histiocytic reticulum cells, and (iv) a phase in follicles can be reliably classified (100 per cent cor-which only centrocytes and follicular dendritic cells rect) by the small centroblast or centrocyte countsremain. This phasic development accounts for the on their own. large variance in cell frequencies observed in reac- tive germinal centres. Most of the reactive germinal centres studied can be considered as third phase DISCUSSION germinal centres of variousstages;the germinal centres (biopsies R 1 and R6) with high centrocyte In the present study the cellular composition ofcounts as fourth phase germinal centres. According the follicles from FCCL Cb-cc and FCCL Cb lym-to the cluster analysis these reactive germinal phomas was compared with that of reactive germi-centres have a cell spectrum which is comparable nal centres. The cluster analysis separates the datawith that of the FCCL Cb-cc. However, the folli- into 3 groups (FCCL Cb, reactive and FCCL Cb-cccular dendritic cells, present within the germinal respectively). The dendrogram (Fig. 4) shows thatcentres of these two biopsies, have the ultrastruc- follicles of FCCL Cb and reactive germinal centres tural characteristics which are specific for reactive are combined before (i.e. at a lower dissimilaritygerminal centres.' 4 Moreover, discriminant level) they combine with the Cb-cc group. This sug- analysis with a subset of parameters, classifies these gests that follicles of FCCL Cb are more closelygerminal centres correctly. related to reactive germinal centres than follicles of Consequently, the different types of follicular FCCL Cb-cc. lymphomas might be regarded as a malignant devi- Using discriminant analysis we were able to esti-ation of different developmental stages of germinal mate the importance of each parameter for thecentres. FCCL Cb-cc might represent the malignant separation of follicles of FCCL Cb, FCCL Cb-cccounterpart of an 'end stage' (of fourth phase) and reative germinal centres. For discriminationgerminal centre whereas FCCL Cb may represent between these three types of follicles most of the a malignant analogue of an early phase, presumably parameters had some discriminating capacity, The a second or third phase, germinal centre. centrocyte and centroblast were most important. As discussed before., we regard the non- However, biopsy R6 was incorrectly placed among lymphoid follicular dendritic cells as non-malignant the Cb-cc follicular lymphomas. Correct classifica- cells which develop within the neoplastic germinal tion of R6 was only achieved using a subset of the centre as a result of factors mediated by the lym- CELLULAR COMPOSITION OF FOLLICULAR LYMPHOMAS 243

phoid tumour cells. In the three cases examined 3.Nathwani BN, Windberg CD, Diamond LW, Bear- follicular dendritic cells were present in higher man RM, Kim H. Morphologic criteria for the dif- numbers in the follicles of FCCL Cb than in the ferentiation of follicular lymphoma from florid follicles of FCCL Cb-cc. This suggests that the dif- reactive : a study of 80 cases. ferentation of follicular dendritic cells from precur- Cancer 1981; 48: 1794-1806. 4.Lennert K. Follicular lymphoma :a tumor of the sor cells might be facilitated in FCCL Cb in germinal centers. Gann Mon Cancer Res 1973; 15: comparison with FCCL Cb-cc. From studies on 217-225. rodent systems it is known that follicular dendritic 5.Lukes RJ, Collins RD. New observations on folli- cells differentiate during germinal centre develop- cular lymphoma. Gann Mon Cancer Res 1973; 15: ment.22 Studies in our laboratory (Rademakers et 209-215. al., to be published) suggest that dark zones of 6.Stein H, Gerdes J, Mason DY. The normal and germinal centres, which are composed mainly of malignant germinal centre. In : Janossy G (ed) Clinics small to medium sized centroblasts, are important inHaematologyvol. 11,3.Thelymphocytes. locations for differentiation of follicular dendritic London:WB Saunders Company Ltd. 1982; 11,3: cells during the expansion of germinal centres. 531-599. 7.Harris NL, Nadler LM, Bhan AK. Immunohistolo- Ultrastructural analysis of the cell spectrum of gic characterization of two malignant lymphomas of follicular structures combined with powerful pat- type (centroblastic/centrocytic and terns recognition methods demonstrated that the centrocytic) with monoclonalantibodies:follicular classical FCCL Cb-cc and FCCL Cb are different and diffuse lymphomas of small-cleaved-cell type are entities and may be considered as neoplasms related but distinct entities. Am J Pathol1984;117: reflectingdifferentphases of germinalcentre 262-272. development. The follicles of FCCL Cb have a 8.Lennert K, Mfiller-Hermelink HK. Lymphocytes structural organization which shows more resem- und ihre Funktions-formen.Morphologic,Organisa- blance to benign reactive germinal centres than to tion and immunologische Bedeutung. Verh Anat Ges the follicles of FCCL Cb-cc. 1975; 69: 19-62. 9.Lukes RJ, Collins RD. A functional approach to the classification of malignant lymphoma.In:Recent results in Cancer Research 1974; Berlin, Heidelberg, ACKNOWLEDGEMENTS New York ; Springer Verlag46:18-30. 10.Lukes RJ, Collins RD. New approaches in the classi- We wish to thank Professor Dr. W. Den Otter fication of lymphomata.Brit./Cancer 1975; 31 Suppl and Dr. M. I. Cleton-Soetman for their critical II: 1-28. advice during preparation of the manuscript, Dr. 11.Levine GD, Dorfman RF. Nodular lymphoma : An Ph.M. Kluin for evaluation of the histopathology ultrastructural study of its relationship to germinal centers and a correlation of light and electron micro- of lymph nodes, Dr. J. J. M. Roelofs and Mrs. I. scopic findings. Cancer 1975; 35: 148-164. van der Tweel for statistical assistance, Mr. E. W. 12.Kaiserling E. Non-Hodgkin lymphomas: Ultrastruc- Dumernit for preparing the micrographs, Mrs. ture and cytogenesis. Progress in Pathology 1977; C. T. P. Oprel and E. C. Verbaan for typing the 105; Stuttgart : Fischer Verlag 30-58. manuscript. 13.Putte SCJ Van der, Schuurman HJ, Rademakers This study was supported by a grant from the LHPM, Kluin Ph, Unnik JAM Van. Malignant lym- Netherlands Cancer Society (Koningin Wilhelmina phoma of follicle centre cells with marked nuclear Fonds, K.W.F.). lobation. Virchows Arch (Cell Pathol) 1984; 46: 93-107. 14.Rademakers LHPM, Peters JPJ, Unnik JAM Van. Histiocytic and dendritic reticulum cells in follicular REFERENCES structures of follicular lymphoma and reactive hyperplasia : a quantitative electron microscopical I. Rappaport H, WinterWJ,Hicks EB. Follicular lym- analysis. Virchows Arch (Cell Pathol) 1983; 44: phoma : a reevaluation of its position in the scheme 85-98. of malignant lymphoma, based on a survey of 253 15.Hogeweg P, Hesper B. Oligothetic characterization cases. Cancer 1956; 9: 793-821. of clusters. Pattern Recognitition 1981; 14: 131-136. 2. Lennert K. Malignant lymphomas other than Hodg- 16.Sneath PHA, Sokal RR. Numerical taxonomy. The kin's Disease. In : Uehlinger E (ed). Hanbuch der principles and practice of numerical classification. speziellen pathologischenAnatomicund Histologie. Freeman 1973; San Francisco. Berlin:Springer Verlag, 1978; 3B: 38-40. 17.Lennert K, Stein H. The germinal centre: mor- 244 J. P. J. PETERS ET AL.

phology, histochemistry and .In:M. J De, Unnik JAM Van. Distribution of dendritic reti- Goes and E. Christophers eds. Lymphoproliferative culum cells in follicular lymphoma and reactive Diseases of the Skin. Springer Verlag1982;Berlin, hyperplasia: light microscopic identification and Heidelberg, New York, Tokyo, 3-15. general morphology. Virchows Arch (Cell Pathol) 18. Mollo F, Monga G, Stramignoni A. Dark reticular 1984;46: 215-228. cells in human lymphadenitis and lymphomas. Virch 22.Heusermann U, Zurborn K-H, Schreuder L, Stuttle Arch B (Cell Pathol)1969 ;3: 117-126. HJ. The origin of the dendritic reticulumcell:an 19. Ward JH. Hierarchical groupings to optimize an experimental enzyme-histochemical and electron objective function. J Am Stat Assoc 1963; 58: microscopic study on the rabbit spleen. Cell Tissue 236-244. Res 1980; 209: 279-294. 20. Hogeweg P. Iterative character weighing in numeri- 23.Bock HH. AutomatisChe Klassifikation. Göttingen : cal taxonomy. Comput Biol Med 1976 ; 6: 199-211. Vandenhoeck and Ruprecht 1974. 21. Peters JPJ. Rademakers LHPM, Roelofs JMM, Jong