Antonino Carbone, MD, Division of Pathology, Centro di Riferimento Oncologico, Istituto Nazionale Tumori, IRCCS, via Pedemontana Occidentale, Aviano I-33081, Italy. E-mail: email@example.com
Summary. Biological and clinical studies have shown that Hodgkin's disease (HD) can be divided into two major categories, termed nodular lymphocyte predominance HD (NLP HD) and classic HD (CHD). Within CHD four subtypes have been distinguished: nodular sclerosis, mixed cellularity, lymphocyte rich and lymphocyte depletion. To refine the histogenesis of the pathological spectrum of HD, 75 CHD and 13 NLP HD were analysed for the expression pattern of MUM1/IRF4 (Multiple Myeloma-1/Interferon Regulatory Factor-4), a lymphocyte-specific member of the IRF family, that is expressed by late centrocytes and post-germinal centre (GC) B cells. MUM1 reacted with Hodgkin's and Reed–Sternberg (HRS) cells of all CHD cases (75/75 cases), with a moderate to strong staining intensity. Conversely, lymphocyte and histiocyte (L & H) cells, the putative tumour cells of NLP HD, were negative for MUM-1 expression (9/13 cases) or displayed a weak reactivity for the antigen in < 10% neoplastic cells (4/13 cases). With respect to HD microenvironment, NLP HD displayed numerous MUM1-positive T lymphocytes located in close proximity to L & H cells whereas, in CHD, MUM1-positive T lymphocytes appeared to be distributed randomly with no specific relationship with HRS cells. Overall, this study shows that MUM1 expression differs in L & H cells versus HRS cells, corroborating the notion that NLP HD and CHD represent different stages of B-cell differentiation. As MUM1-positive T lymphocytes form rosettes around tumour cells of NLP HD, but not of CHD, these data point also to differences in the microenvironment of NLP HD and CHD, and postulate an interactive role of MUM1-positive T lymphocytes with L & H cells.
Several molecular studies have indicated that L & H cells are germinal centre (GC)-related B cells which have been stimulated and selected by antigen (Kanzler et al, 1996; Marafioti et al, 1997), whereas HRS cells may represent a B cell-derived monoclonal population related to GC/post-GC memory B cells that lost their dependence on antigenic stimulation (Kanzler et al, 1996; Küppers et al, 1999). The histogenesis of HD has been further refined by studies of biological markers identifying distinct subsets of mature B cells. In fact, L & H cells express BCL-6, a transcription factor expressed in GC B cells, whereas HRS cells express CD138/syndecan-1, a proteoglycan associated with post-GC terminal B-cell differentiation (Carbone et al, 1998).
Recently, MUM1/IRF4 (Multiple Myeloma-1/Interferon Regulatory Factor-4) has been added to the panel of phenotypic markers available for the characterization of lymphoid disorders (Falini et al, 2000). MUM1 was discovered because of its involvement in the t(6;14)(p25;q32) translocation of multiple myeloma, which causes the juxtaposition of the MUM1 gene, mapping at 6p25, to the IgH locus on 14q32 (Iida et al, 1997). MUM1 is a lymphocyte-specific member of the interferon regulatory factor (IRF) family of transcription factors which is also known as ICSAT (Interferon Consensus Sequence binding protein for Activated T cells) and Pip (PU.1 Interaction Partner) (Nguyen et al, 1997). Recent studies have shown that MUM1 is expressed in the final step of intra-GC B-cell differentiation, in subsequent steps of B-cell maturation towards plasma cells, and in lymphoid neoplasms thought to be derived from these cells (Falini et al, 2000; Gaidano & Carbone, 2000; Tsuboi et al, 2000). In the T-cell compartment, MUM1 expression is restricted to activated T cells (Falini et al, 2000).
Although MUM1 has been extensively studied on lymphoid neoplasms including multiple myeloma, diffuse large cell lymphomas, T/null anaplastic large cell lymphomas (Falini et al, 2000; Tsuboi et al, 2000) and acquired immunodeficiency syndrome (AIDS)-related non-Hodgkin's lymphomas (Carbone et al, 2000, 2001), little is known about its expression in HD. Because MUM1 has proved useful to refine the histogenesis of lymphoproliferative disorders, we analysed the expression pattern of MUM1 throughout the pathological spectrum of HD. Also, because MUM1 is expressed by activated T cells, we investigated MUM1 expression in the lymphocytes of HD microenvironment that are postulated to play a significant role in HD pathogenesis.
Materials and methods
Neoplastic samples. This study was conducted on 75 cases of CHD (55 NS, 17 MC, 3 LD) and 13 cases of NLP HD. HRS cells of 14 CHD (8 NS, 5 MC, 1 LD) expressed a B-cell phenotype (i.e. CD20+ and/or CD79a+), HRS cells of 2 CHD (2 NS) expressed a T-cell phenotype (i.e. CD3+), whereas HRS cells of 59 CHD (45 NS, 12 MC, 2 LD) failed to express B-cell or T-cell antigens. L & H cells of all NLP HD cases expressed a B-cell phenotype (i.e. CD20+).
NLP HD was diagnosed according to morphological and immunophenotypic criteria (Harris et al, 1994; Mason et al, 1994). In all 13 cases, several neoplastic cells exhibited the characteristic morphology of L & H cells with folded and lobated nuclei and inconspicuous nucleoli (‘popcorn’ cells). L & H cells consistently expressed CD20 and CD45 antigens (13/13), but lacked CD15 and CD30. All NLP HD cases displayed a nodular growth; a prominent meshwork of follicular dendritic cells was present within the nodules. The CD30+, CD45–, CD15+, EMA– diagnostic profile was required for the diagnosis of CHD (Harris et al, 1994). The Rye modification of the Lukes and Butler classification was used to classify the histological subtypes of CHD (Lukes et al, 1966b).
Tissues were fixed in Bouin solution or neutral buffered formalin. In most cases, a portion of unfixed tissue was snap frozen in liquid nitrogen and stored at −80°C.
Non-neoplastic samples. Ten lymph node samples with non-neoplastic lymphoid proliferations were also included in the study. The histopathological pattern was predominantly represented by hyperplastic changes of the lymphoid follicles.
Immunohistochemical studies. Immunohistochemistry was performed using the avidin–biotin–peroxidase complex (ABC-px) or alkaline phosphatase anti-alkaline phosphatase (APAAP) methods (Hsu et al, 1981; Cordell et al, 1984).
The expression of MUM1 was investigated using an affinity-purified polyclonal goat antibody (ICSAT/M-17) specific for the MUM1 protein (Iida et al, 1997) (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA). The M-17 antibody reacts with MUM1 of mouse, rat and human origins, although it does not cross-react with other members of the IRF family proteins. For MUM1 assessment, freshly cut paraffin-embedded sections were treated in a microwave oven at 250 W for 30 min in EGTA solution (1 mmol/l pH 8) or TEC (2 mmol/l Tris(hydroxymethyl)-aminomethan; 1·3 mmol/l Ethylenediamine-Tetraacetic acid; 1·1 mmol/l tri-sodium citrate dihydrate) solution (pH 7·8); immunostaining was performed on an automated immunostainer (Nexes, Ventana Medical Systems Inc., Tucson, AZ, USA) according to a modified version of the company's protocols.
Negative control experiments were performed by preabsorbing the anti-MUM1 antibody with a fivefold weight excess of blocking peptide (Santa Cruz Biotechnology Inc.) (overnight at 4°C) and then carrying out immunostaining as described above. None of the negative control sections was immunostained.
In selected cases, including three NLP HD and four CHD cases, the reactivity pattern of the ICSAT/M-17 polyclonal antibody was compared with that of MUM1p (Falini et al, 2000), a monoclonal antibody (mAb) raised against the human MUM1/IRF4 protein (kindly provided by Professor B. Falini, Institute of Haematology and Internal Medicine, University of Perugia, Italy). The reactivity pattern of both antibodies recognizing MUM1/IRF4 was generally superimposable.
Immunostaining for CD20 (L26; Dakopatts A/S, Glostrup, Denmark), CD3 (polyclonal antibody; Dakopatts) and CD57 (Leu7; Becton & Dickinson, San Jose, CA, USA) was performed on paraffin-embedded sections on an automated immunostainer (Nexes); deparaffinized tissue section were first pretreated in a microwave oven (Jet 900 W, Philiphs) for 30 min at 250 W in citrate buffer pH 6 and then immunostained using a diaminobenzidine (DAB) detection kit (Ventana Medical System).
Deparaffinized and cryostat sections were used for further immunophenotyping and lineage assignment of HD cases. Source and specificities of antibodies utilized in this study have been reported in detail previously (Carbone et al, 1995b, 1997, 1998).
Assessment of MUM1 staining in HD samples. At least 100 neoplastic cells per section, as defined by histological and immunohistological criteria (CD30 positivity for CHD and CD20 positivity for NLP HD), were independently counted by two of the authors. The percentage of MUM1-positive neoplastic cells was assigned to one of the following categories: 0, < 10%, 10% to 25%, 25% to 50%, 50% to 75%, and > 75%. Only definite and unambiguous staining on unequivocal malignant cells was accepted as positive. The intensity of staining (weak, moderate or strong) was also recorded.
A semiquantitative analysis was performed on paraffin-embedded tissue sections for microenvironmental lymphocytes expressing MUM1. In each case, the mean number of MUM1-expressing lymphocytes was derived from the number of MUM1-expressing lymphocytes directly surrounding each RS cell (on a total of 100 RS cells counted).
Two-colour staining. Multiple colour immunohistochemical stainings were performed to detect MUM1 plus CD3 or CD20 or CD57 in selected lymph node samples involved by CHD and NLP HD, as well as in non-neoplastic samples. Formalin-fixed paraffin-embedded tissue sections were first immunostained with anti-MUM1 antibody using a DAB detection kit (Ventana Medical System); subsequently, sections were treated twice for 5 min in citrate buffer (pH 6) in a microwave oven to denature bound antibody molecules and to inactivate the alkaline phosphatase. Then, sections were immunostained with anti-CD3 or -CD20 or -CD57 using the alkaline phosphatase anti-alkaline phosphatase (APAAP) method or using a Fast Red detection kit (Ventana Medical System).
Purification of peripheral CD4+ T cells and in vitro activation. Peripheral blood mononuclear cells (PBMNC) from normal donors were purified by Ficoll-Hypaque (Pharmacia, Uppsala, Sweden) density gradient separation and subsequently incubated with ammonium chloride for 5 min at 4°C to minimize the contamination with red blood cells. Monocyte depletion was carried out by plastic adherence. The CD4+ T cells were isolated from the mononuclear cell fraction using anti-CD4-conjugated immunomagnetic microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's instructions. The efficiency of separation, as evaluated by flow cytometry, was in all cases more than 95% of CD4+ cells (data not shown). The purified CD4+ T-cell fraction was activated by exposure to 12-O-tetradecanoylphorbol-13-acetate (TPA; Sigma Chemical Co, St Louis, MO, USA) (10 ng/ml) and ionomycin (Sigma) (1 µg/ml) as previously described (Carbone et al, 1995c). Basal and stimulated CD4+ T cells were tested for MUM1 expression using flow cytometry. Basal and stimulated CD4+ T cells were also tested for the CD57 and CD40L antigens that are usually expressed by microenvironmental T lymphocytes in NLP HD (Poppema, 1989) and CHD (Carbone et al, 1995a).
Flow cytometry analysis. To detect MUM1 expression, purified CD4+ T cells were fixed in 2% paraformaldehyde in phosphate-buffered saline (PBS) for 15 min at 4°C and then permeabilized in 1% Tween-20 for 30 min at 4°C. Cells were then incubated with anti-MUM1 (MUM1p) mAb for 30 min at 4°C, washed twice and finally incubated with a phycoerythrin (PE)-conjugated F(ab′)2 fragment of goat anti-mouse IgG secondary antibody (Coulter-Immunotech; Becton & Dickinson). Non-specific binding of antibodies was assessed by labelling cells with isotype-matched irrelevant mouse Igs (Becton & Dickinson). Viable antibody-labelled cells were identified according to their forward and right-angle scattering, electronically gated and analysed on a FACScalibur flow cytometer (Becton & Dickinson) using CellQuest software (Becton & Dickinson). For surface CD40L detection, cells were sequentially incubated with anti-CD40L mAb M90 (Immunex) and PE-conjugated F(ab′)2 fragments of goat anti-mouse IgG secondary antibody (Becton & Dickinson). Surface CD57 expression was evaluated by direct immunofluorescence using fluorescein isothiocyanate (FITC)-conjugated anti-CD57 mAb (Becton & Dickinson). For dual-colour immunofluorescence, cells were first incubated with anti-MUM1 and a PE-conjugated F(ab′)2 fragment of goat anti-mouse IgG followed by FITC-conjugated anti-CD57 mAb. Quadrant markers were set, following electronic compensation, to account for the highest levels of non-specific fluorescence displayed by cells sequentially labelled with isotype-matched control Igs, PE-conjugated F(ab′)2 goat anti-mouse Igs (red fluorescence) and FITC-conjugated isotype-matched control Igs (green fluorescence).
Expression of the MUM1 antigen in HD
NLP HD. In 9/13 (69%) NLP HD cases, L & H cells scored negative for MUM-1 expression. In the four remaining NLP HD, a weak or moderate expression of MUM-1 was detected on occasional cells accounting for < 10% of L & H cells (Table I; Fig 1A and B).
Table I. MUM1 antigen expression in L & H cells of nodular lymphocyte predominance Hodgkin's disease (NLP HD) and HRS cells of classic Hodgkin's disease (CHD).
CHD.Table I summarizes data on the expression of the MUM1 antigen by HRS cells of 75 cases of CHD representative of the different histological subtypes. A positive MUM1 staining was observed in the majority of HRS cells (> 75%) of all CHD cases examined (Fig 2A and B). MUM1 positivity was detected in all pathological subtypes of CHD, including NS, MC and LD. HRS cells showed moderate to strong MUM1 positivity (Fig 2). In several positive tumour cells, nuclear positivity was associated with diffuse cytoplasmic staining of weak intensity. There was no evidence of cell membrane staining. Among CHD cases, there was no apparent correlation between MUM1 expression and the antigenic phenotypes of HRS cells (B, T, undetermined) (Table I).
Taken together, these data demonstrated that, throughout the histological spectrum of HD, a high reactivity and staining intensity for MUM1 was preferentially expressed by tumour cells of CHD.
In cases in which residual normal lymphoid tissue was present, plasma cells together with small lymphoid cells and rare blastic cells were stained with the anti-MUM1 antibody.
Expression of MUM1 in microenvironmental reactive lymphocytes of HD
In 11 out of 13 NLP HD cases, MUM1-expressing lymphocytes were numerous within the HD nodules and were mainly located in close proximity to the L & H cells as rosettes (Figs 1 and 3A). In the remaining two NLP HD cases, MUM1-expressing lymphocytes were scanty. Two-colour immunohistochemistry and serial section analysis showed that MUM1-expressing T lymphocytes of NLP HD were CD3+ and CD20–, and usually CD57– (Fig 3A–D). As summarized in Table II, a mean of 2·5–6·5 MUM1+ T cells surrounded a single L & H cell in the NLP HD cases. In cases of CHD, numerous MUM1+ lymphoid cells were found in the involved areas (Table II), but they appeared to be distributed randomly and usually did not form rosettes around HRS cells (Fig 2).
Table II. A semi-quantitative evaluation of MUM1-expressing microenvironmental lymphocytes in Hodgkin's disease (HD).
Mean number of MUM1+ lymphocytes surrounding each tumour cell (range)*
In each case the mean number of MUM1-expressing lymphocytes was derived from the number of MUM1-expressing lymphocytes directly surrounding each tumour cell in a total of 100 tumour cells evaluated.
Expression of MUM1, CD57 and CD40L during activation of normal CD4+ T cells
Figure 4 shows a representative experiment illustrating the kinetics of MUM1 and CD57 expression during in vitro activation of normal peripheral blood CD4+ T cells by TPA and ionomycin.
A subset (10%) of basal CD4+ T cells expressed CD57, and its expression did not change during activation. In contrast, basal CD4+ T cells did not express MUM1. MUM1 expression was observed in purified CD4+ T cells after 24 h of activation and reached a maximum at 48 h (Fig 4), when 51% of CD4+ T cells showed MUM1 expression. Dual-colour immunofluorescence experiments (CD57/MUM1), performed at 24 h, disclosed that in purified CD4+ T cells the dominant population expressed only MUM1, whereas a small subset co-expressed MUM1 and CD57 (not shown). Furthermore, the kinetics of MUM1 expression were different to that of CD40L. An early peak of CD40L expression was observed in purified CD4+ T cells after 6–12 h of activation, followed by a return to basal levels (< 5% of positive cells) within 24–48 h, whereas MUM1 expression was observed at a later stage (24–48 h) and persisted stably (not shown).
Expression of MUM1 in non-neoplastic lymphoid tissue
As expected, a reactivity for MUM1 was detectable in a fraction of GC cells and in all plasma cells. Mantle zone lymphocytes were usually MUM1 negative. The paracortical/interfollicular zones were mostly MUM1 negative, with the exception of some small lymphoid cells and rare isolated large cells (with blastic morphology), which were represented by B cells and T cells (not shown). MUM1-positive GC cells were preferentially localized in the light zone of the GC, exhibited a moderate to strong nuclear staining and displayed a small cleaved cell/centrocyte morphology. MUM1-positive GC cells, which were predominantly represented by B cells (MUM1+, CD20+), scored consistently negative for CD57 (not shown).
To further refine the histogenesis of the pathological spectrum of HD, we have investigated the expression of MUM1 in a large series of HD samples representative of different histological subtypes and immunophenotypes. Our results indicate that the profile of MUM1 expression of CHD and NLP HD differs both in the neoplastic component and in the reactive microenvironment.
With respect to HD neoplastic cells, we have shown that HRS cells of CHD expressed MUM1 in all cases, irrespective of the HD histological subtype (NS, MC, LD) and the antigenic phenotype. Conversely, among NLP HD cases, the percentage of L & H cells expressing MUM1 ranged from 0 to 10%, and the intensity of staining was generally weak. Thus, high reactivity and strong intensity of MUM1 expression in the context of HD appears to be a specific feature of CHD as opposed to NLP HD. As, in physiological B-cell differentiation, MUM1 expression identifies the transition from GC B cells to immunoblasts and plasma cells, MUM1 positivity in CHD corroborates the notion that HRS cells reflect a post-GC phenotype. Indeed, HRS cells generally fail to express BCL-6, that is restricted to GC B cells, and stain positive for CD138, which clusters with late stages of B-cell maturation (Kadin, 2001). The observation that L & H cells of NLP HD are generally negative for MUM1 is in agreement with our previous observation that tumour cells of NLP HD consistently display a BCL-6+/CD138– phenotype, consistent with derivation from GC B cells (Falini et al, 1996; Carbone et al, 1997, 1998).
The profile of MUM1 expression in the pathological spectrum of HD also reveals differences in the microenvironment of CHD and NLP HD. Microenvironmental lymphocytes in CHD are thought to affect the tumour behaviour and phenotype, as exemplified by the case of T-cell/HRS-cell interactions through the CD40/CD40L system, which are thought to contribute growth signals to the neoplastic population of CHD (Gruss et al, 1994; Carbone et al, 1995a,b; Kadin, 2001). The small lymphocytes surrounding L & H cells in the nodules of NLP HD were substantially different from those of CHD and were represented by a mixture of polyclonal B cells and numerous T cells, which preferentially express either CD57 (Timmens et al, 1986) or MUM1 (this study). The almost exclusive presence of MUM1-expressing lymphocytes around tumour cells in NLP HD may provide a further support for the separation of NLP HD within the pathological spectrum of HD in terms of microenvironmental T-cell phenotype.
The functional significance of the differential distribution of the MUM1-positive (this study), CD57-positive (Timmens et al, 1986; Poppema, 1989) and CD40L-positive (Carbone et al, 1995a, 1998) T lymphocytes within the pathological spectrum of HD remains to be clarified. Based on kinetic data obtained in vitro in normal T lymphocytes, it may be hypothesized that T lymphocytes of CHD and of NLP HD represent different activation stages of CD4+ T cells. According to this model, HRS cells would interact predominantly with T cells expressing the CD4+/CD40L+/MUM1– phenotype and representing an early phase of T-cell activation. Conversely, L & H cells of NLP HD would preferentially interact with T cells expressing the CD4+/CD40L–/MUM1+/CD57+/– phenotype and representing a late phase of T-cell activation. Future studies performed on T lymphocytes obtained ex vivo from tissues of NLP HD and CHD are required to formally verify this hypothesis.
This work has been supported in part by the Associazione Italiana per la Ricerca sul Cancro, Milan, Italy; the Ministero della Sanità, Ricerca Finalizzata I.R.C.C.S., Rome, Italy; National Institutes of Health grant CA-37295; and MURST-Cofin 2000, Rome, Italy.