Abstract
purpose. Several B-cell-associated transcription factors and their coactivators, including BCL-6, BSAP/PAX5, BOB.1/OBF.1, Oct.2, MUM1/IRF4, and PU.1, have been detected in peripheral B-cell lymphomas. There are limited data on their expression in centrally located lymphoid neoplasms, such as primary intraocular lymphoma (PIOL) or primary central nervous system lymphoma (PCNSL). PIOL is a rare non-Hodgkin lymphoma, considered a subtype of PCNSL. Both are usually diffuse, large B-cell lymphoma (DLBCL), rarely manifest outside the CNS, and carry a poor prognosis.
methods. Tissue biopsy specimens were examined from eight cases of PIOL and 42 cases of HIV-negative PCNSL, as well as 50 cases of peripheral DLBCL, for the above-mentioned transcription factors and for immunoglobulin heavy and light chains, using immunohistochemistry.
results. Immunoglobulin expression was demonstrated in 46 (92%) of 50 cases of PIOL/PCNSL but in only 27 (54%) of 50 cases of peripheral DLBCL. Positivity for BOB.1/OBF.1 and Oct.2 was observed in all immunoglobulin-expressing PIOL and PCNSL. BSAP/PAX5 expression occurred in 98% of PIOL/PCNSL, and MUM1/IRF4 immunoreactivity in 45 (90%) of 50 of these cases. PU.1 expression was observed in only 10% of the PIOL/PCNSL group in contrast to 23 (46%) of 50 peripheral DLBCLs. Aberrant coexpression of MUM1/IRF4, PAX5, MUM1/IRF4, and BCL-6 was observed in most PIOLs/PCNSLs.
conclusions. These data provide further support to the notion that peripheral and centrally located DLBCLs differ in clinical, immunophenotypic, and genotypic features, despite their similar morphologic characteristics. PIOL and PCNSL tumor cells are most likely to be derived from mature B-cells that have undergone the germinal center reaction.
Similar to other developmental processes, the differentiation of hematopoietic stem cells into B cells involves the sequential expression of structural and regulatory genes. After its commitment to B-cell lineage, the lymphocyte migrates from the bone marrow into the peripheral lymphoid tissue, passing through various stages of differentiation that can be defined by genotypic and phenotype changes. These include the rearrangement of immunoglobulin heavy (IgH) and light (IgL) chains, the expression of various antigens on the cell surface, as well as affinity maturation in the germinal centers.
1 It has been demonstrated that different transcription factors are expressed at various time points during B-cell maturation, and indeed are essential for the regulation of gene expression during both early and late B-cell development.
2 3 4 5 The physiological and clinical relevance of these transcription factors becomes apparent on disruption of or alterations of the genes that encode them.
To date, several B-cell-associated transcription factors and their coactivators have been detected in paraffin-embedded sections of human tissues. These include BCL-6,
6 BOB.1/OBF.1 (also known as OCA-B),
7 Oct.2,
7 MUM1/IRF4,
8 BSAP/PAX5,
9 10 and PU.1 (also known as Spi-1).
11 12 The relationship of these transcription factors to B-cell differentiation and the germinal center is demonstrated in
Figure 1 . Most studies investigating these molecules have been performed on specimens taken from patients with peripheral B-cell lymphomas
7 12 13 14 15 16 17 18 and, recently, in some lymphoid tumors of non-B lineage.
19 The literature, however, contains little/limited data on the expression of these transcription factors in extranodal “central” lymphoid neoplasms, defined herein as primary intraocular lymphoma (PIOL)
20 or primary central nervous system lymphoma (PCNSL).
21
PIOL is a rare, high-grade, malignant non-Hodgkin B-cell lymphoma, involving the retina, the vitreous, and occasionally the optic nerve.
22 23 24 25 26 It is considered a subtype of PCNSL, which is defined as a lymphoma limited to the craniospinal axis without evidence of systemic disease.
27 Most PIOL and PCNSL are diffuse, large B-cell lymphomas (DLBCLs), according to the updated World Health Organization (WHO) lymphoma classification,
28 and are associated with a poor prognosis.
29 30 Although there has been progress in the comprehension of the molecular pathology of PIOL,
23 31 32 the understanding of the histogenesis of PIOL remains limited, in part because of the paucity of tumor specimens available for examination. Because the expression of the B-lymphocyte-associated transcription factors could help define the cellular origin of PIOL, we used standard immunohistochemical techniques to investigate the expression of Ig and the transcription factors BSAP/PAX5, BOB.1/OBF.1, Oct.2, MUM1/IRF4, and PU.1 in PIOL and PCNSL and compared the results with those obtained in peripheral DLBCL.
A series of eight cases of PIOL were collected, and slides were retrospectively reviewed from the consultation files of the Reference Center for Hematopathology, Pathology Department of Charité University Medicine, Campus Benjamin Franklin (Charité CBF) and of the Department of Pathology, Pitie-Salpetriere, Paris. The specimens consisted of enucleated eyes (
n = 2) and chorioretinal biopsies (
n = 6). In addition, stereotactic biopsies of 34 (Charité CBF) and 8 (Casey Eye Institute and Department of Pathology) cases of PCNSL were reviewed. The results were compared with those obtained from 50 cases of peripheral (nonmediastinal) DLBCL (Charite CBF). The experiments were performed concurrently under the same experimental conditions; the results of 36 of 50 cases of peripheral DLBCL has been published.
14 All patients were HIV negative. All lymphomas were diagnosed according to the WHO Classification of Tumors.
28
The tissue biopsy specimens had been fixed in 4% buffered formalin and embedded in paraffin. In addition, hyperplastic tonsil, reactive lymphoplasmacellular infiltrates in the lacrimal gland, eye enucleated for malignant choroidal melanoma, and cerebral tissue obtained from autopsies (average postmortem period, 15 hours) were evaluated.
For immunostaining of paraffin-embedded tissues, tissue sections (4 μm) were cut from the paraffin blocks, dewaxed, rehydrated, and subjected to heat-induced epitope retrieval methods before incubation with the appropriate antibodies. Sections were immersed either in sodium citrate buffer solution at pH 6.0 or in EDTA solution (pH 8.0), and were subsequently heated in a pressure cooker, according to standard practices.
33 After they were rinsed in running water and Tris-buffered saline, the sections were incubated with the primary antibodies listed in
Table 1 . For detection of PAX5, BOB.1/OBF.1, Oct.2, MUM1/IRF4, and CD10, the alkaline-phosphatase anti-alkaline phosphatase complex (APAAP) was used as a stain,
34 whereas the En Vision (DakoCytomation, Glostrup, Denmark) method
35 was used for PU.1 detection. CD10 was examined in these experiments to help enable interpretation of the results with respect to B-cell differentiation and the germinal center. To demonstrate Ig expression, the streptavidin-biotin-peroxidase method was used.
36 Alkaline phosphatase was revealed by fast red as the chromogen, whereas peroxidase was developed in diaminobenzidine, giving a brown reaction product. Frozen sections for the validation of Ig expression in reactive lymphoid tissue were air dried overnight and fixed in acetone for 10 minutes, and the binding of the polyclonal antibodies (DakoCytomation) to IgH and IgL chains γ, μ, δ, κ, and λ was evaluated. Appropriate negative and positive tissue control experiments were performed with each investigation. The research was approved by the respective institutional review boards. All human tissue was managed in accordance with the guidelines set forth in the Declaration of Helsinki.
In agreement with results in previous studies,
7 14 18 19 the expression of transcription factors in reactive lymphoid tissue reflected the maturation stage of the B-cells in the different germinal center areas
(Fig. 1) .
Immunoglobulin Expression.
The palatine tonsils and the interstitial tissue of the lacrimal glands contained an infiltrate of small, bland, mixed T and B lymphocytes with reactive lymphoid follicles. The small B cells of the mantle of the secondary follicles were strongly immunoreactive for IgD and IgM (
Figs. 2A 2B ). The marginal zone B cells, in contrast, showed strong positivity for IgM, but only weak expression of IgD, consistent with previous studies.
14 Most germinal center B cells of the dark zone of the follicles were negative or only weakly positive for Ig protein, with the exception of B cells showing plasmacellular differentiation.
BSAP/PAX5.
In the lymphoid tissues of the tonsils and the lacrimal glands, strong BSAP/PAX5 expression was observed in the lymphocytes within the germinal centers and the mantle zone, as well as in intraepithelial lymphocytes
(Fig. 2C) . Only occasional cells were immunoreactive in the interfollicular zone. In the eyes enucleated for choroidal melanoma, occasional lymphocytes in the choroid and iris were positive for BSAP/PAX5. No BSAP/PAX5 immunoreactivity was observed in normal cerebral tissue.
BOB.1/OBF.1.
Strong immunoreactivity for BOB.1/OBF.1 was observed easily in all B-cell populations encountered in reactive lymphoid tissues. The strongest expression occurred in the nuclei of the germinal center cells
(Fig. 2D) .
In the eyes enucleated for malignant choroidal melanoma, BOB.1/OBF.1 nuclear staining was present in the basal epithelial layer of the cornea; the equatorial cells of the lens; the pigment epithelium of the ciliary body; occasional B-cells within the uvea; and the ganglion cell, inner plexiform, and outer plexiform layers of the retina
(Fig. 3A) . In normal cerebral tissue, no BOB.1/OBF.1 immunoreactivity was observed.
Oct.2.
The nuclei of all lymphoid cells in the dark zone of the germinal centers in the tonsils and reactive lymphoid infiltrates were strongly positive for Oct.2
(Fig. 2E) . In the light zone, most cells stained with a similar intensity. Follicle mantle lymphocytes displayed weak to moderate Oct.2 positivity. In the interfollicular zone, nuclear Oct.2 expression was present in scattered blasts and occasional small lymphocytes, as well as in plasma cells, which displayed a polytypical expression of IgL κ and λ
(Fig. 1) . In the eyes enucleated for malignant choroidal melanoma, Oct.2 expression was limited to occasional lymphocytes in the choroid and iris
(Fig. 3B) . In the nonneoplastic brain, Oct.2 expression was absent.
MUM1/IRF4.
Some marginal-zone B cells showed weak nuclear immunoreactivity for MUM1/IRF4
(Fig. 2F) . Most germinal center cells were negative for MUM1/IRF4, although several germinal center cells in the light zone displayed strong nuclear and moderate cytoplasmic positivity for MUM1/IRF4. Admixed MUM1/IRF4-positive plasma cells were present in the subepithelial areas in the tonsils, in the interstitial tissue of the lacrimal glands, and occasionally in the reactive germinal center. In the eyes enucleated for malignant choroidal melanoma, MUM1/IRF4 was observed in occasional melanocytes in the choroid, and demonstrated variable immunoreactivity in the malignant uveal melanoma cells
(Fig. 3C) . In normal cerebral tissue, no MUM1/IRF4 immunoreactivity was observed.
BCL-6.
Strong BCL6 expression was observed in the germinal center cells
(Fig. 2G) . Occasional extrafollicular blasts were immunoreactive for BCL6. In ocular tissues, some basal epithelial cells of the conjunctival and corneal epithelium, demonstrated moderate nuclear expression of BCL-6. Normal cerebral tissue was negative for BCL-6.
PU.1.
PU.1 expression was observed in the nuclei of centroblasts and centrocytes with a moderate intensity
(Fig. 2H) . Admixed macrophages demonstrated a strong staining for PU.1, in contrast to the reactive plasma cells, which were negative for this transcription factor. In normal cerebral and ocular tissues, no PU.1 immunoreactivity was observed.
The findings of the expression patterns of transcription factors in PIOL and PCNSL are summarized in
Table 2 . First, there were no differences in the staining patterns or staining intensity of the transcription factors in PIOL tumor cells in the vitreous when compared with those in the chorioretinal biopsy specimens or in the enucleated eye. Secondly, there were no fundamental differences in the staining patterns between PIOL and PCNSL; consequently, they are grouped together in
Table 2 , and the results compared with peripheral, diffuse, large B-cell lymphomas. For the sake of completeness, however, the expression patterns of the Ig transcription factors in PIOL and PCNSL are demonstrated separately in
Figures 4 and 5 , respectively.
In nearly all (92%) of PIOL and PCNSL cases, monoclonal expression of an IgL and/or IgH was demonstrated
(Figs. 4A 4B 5A 5B) . All PIOLs were Ig positive. Moderate to strong positivity for BOB.1/OBF.1 and Oct.2 was observed in all Ig-positive and Ig-negative PIOLs and PCNSLs
(Figs. 4D 4E 5D 5E) . Both BSAP/PAX5 and MUM1/IRF4 immunoreactivity occurred in 49 (98%) of 50 PIOL/PCNSL specimens, whereas BCL-6 expression was observed in 43 (86%) of 50 of the cases
(Figs. 4C 4F 5C 5F) . Weak PU.1 immunoreactivity was present in only 5 (10%) of the PIOL/PCNSL group; it is noteworthy that all PIOLs were PU.1 negative
(Figs. 4H 5H) . CD10 expression was observed in only 6 (12%) of 50 PIOL/PCNSL (
Table 2 ;
Figs. 4I 5I ).
A moderate to strong intensity of BOB.1/OBF.1 and a strong intensity of Oct.2 was observed in all DLBCLs (
Figs. 6D 6E ;
Table 2 ). Moderate to strong immunoreactivity for MUM1/IRF4 was seen in 45 (90%) of 50 of peripheral DLBCLs, BCL-6 positivity in 35 (70%) of 50 cases
(Fig. 6G) , and BSAP/PAX5 expression in 42 (84%) of 50 of cases
(Fig. 6C) . In contrast, weak immunoreactivity for PU.1 was observed in 23 (46%) of 50 peripheral DLBCLs
(Fig. 6I) . Immunoglobulin expression at the protein level was observed in only 27 (54%) of 50 cases of peripheral DLBCL
(Figs. 6A 6B) , consistent with previous findings.
14 Finally, CD10 expression was observed in 12 (24%) of 50 peripheral DLBCLs
(Fig. 6H) .
To gain new insights into the cellular origin and the maturation stage of PIOL and PCNSL, we investigated the expression of the Ig transcription factors PAX5, BOB.1/OBF.1, Oct.2, PU.1, BCL-6, and MUM1/IRF4, were investigated and compared with those patterns in peripheral DLBCLs. These transcription factors were chosen because of their roles in the regulation of gene expression during B-cell development. This is the first study in which the Ig transcription factor expression in PIOL and PCNSL has been investigated in detail. In agreement with previous findings,
14 37 we observed simultaneous expression of BOB.1/OBF.1 and Oct.2 in all reactive and normal B-cells that expressed Ig. Recently published data demonstrate that Oct.2 and BOB.1/OBF.1 transcription factors are coexpressed in almost all categories of peripheral B-cell neoplasms, with the exception of the classic Hodgkin lymphoma.
7 11 14 15 38 In the present study, we demonstrated a consistent expression of BOB.1/OBF.1 and Oct.2 in all Ig-expressing PIOLs, PCNSLs, and peripheral DLBCLs. This observation adds support to the notion that Oct.2 and BOB.1/OBF.1 are imperative for Ig expression, not only in normal, but also in neoplastic B cells.
39
One of the most interesting observations in this study was the finding that Ig expression was absent in neoplastic B cells, despite the presence of Oct.2 and BOB.1/OBF.1. This pattern of expression was noted in 46% of peripheral DLBCLs compared with only 8% of PIOL/PCNSLs. The difference between these two groups of essentially morphologically similar lymphomas is striking and suggests that, in contrast to PIOL/PCNSL tumor cells, a significant proportion of peripheral DLBCLs arise from germinal-center B cells that have either lost the ability to express Ig or have downregulated Ig expression. The consistent and strong expression of BOB.1/OBF.1 and Oct.2 in the Ig-negative PIOL/PCNSL and peripheral DLBCL indicates, however, that other mechanisms, possibly “crippling mutations” within the transcription control regions and/or coding sequences,
40 are involved to account for the lack of Ig transcription in these tumors. An alternative could be epigenetic silencing of the Ig heavy chain gene, as recently described in classic Hodgkin lymphoma.
41 To clarify this, further investigations are needed.
Another finding of interest of the present study was the absence of immunostaining of PU.1 in the majority (90%) of Ig-positive PIOL/PCNSL. This contrasted with a weak immunoreactivity of this transcription factor in 23 (46%) of 50 peripheral DLBCL cases and a corresponding lack of expression in 54% of cases. PU.1 is a hematopoietic transcription factor belonging to the Ets family. It is identical with the Spi-1 oncogene, which is implicated in spleen focus-forming, virus-induced murine erythroleukemias. PU.1 regulates the transcription of genes relevant to the development of B-cells, macrophages, and myeloid cells.
4 42 43 Moreover, it binds to several promoters, such as MUM1/IRF4, and regulates the expression of genes required for terminal B-cell differentiation. It is reported to be expressed in immature and mature B lymphocytes, being lost only in plasma cells and their precursors.
44 The physiological absence or downregulation of PU.1 expression has been reported in some plasma-cell–derived lymphomas,
45 whereas most other B-cell neoplasms are PU.1 positive. Most of the PIOLs and PCNSLs investigated in the present study demonstrated some degree of plasmacellular differentiation, possibly explaining their PU.1 negativity. Of note, there was no correlation between PU.1 negativity and Ig expression in either the PIOL/PCNSL or peripheral DLBCL specimens. Therefore, it is unlikely that PU.1 is responsible for the lack of Ig expression in Ig negative PIOLs/PCNSLs or DLBCLs.
Finally, the majority of the cases examined expressed MUM1/IRF4. Positivity was observed in 49 (98%) of 50 PIOL/PCNSL and in 47 (94%) of 50 DLBCL. In keeping with our preliminary findings in PIOL,
20 with data examining PCNSL
21 and with previous data on peripheral DLBCL,
8 46 MUM1/IRF4 was often coexpressed with BCL-6 in both central and peripheral DLBCL by the same neoplastic cells, an aberrant feature that does not correspond to the mutually exclusive pattern found in normal germinal centers.
8 The dysregulation of these transcription factors in PIOL/PCNSL is further undermined by the coexpression of BSAP/PAX5 and MUM1/IRF4 by the neoplastic B-cells, again an abnormal phenotype not found in the normal counterpart.
47 These findings, together with the detection of numerous IgH gene somatic mutations in PIOL
32 48 and PCNSL,
49 50 51 52 provide further evidence for the hypothesis that PIOL and PCNSL are derived from mature B cells that have undergone a prolonged interaction in the microenvironment of the germinal center or are at the late germinal center stage of differentiation. This may explain the extensive loss of CD10 (only 12% PIOL and PCNSL cases positive versus 24% of peripheral DLBCL cases) and the variable expression of both BCL-6 and MUM1/IRF4 in PIOL.
20
Germinal centers arise in peripheral lymphatic tissue (e.g., tonsil, mucosa-associated lymphatic tissue, and lymph nodes) as a result of antigen stimulation, and they are responsible for the production of B cells with high-affinity antigen receptors. The memory B cells and long-lived plasma cells, which emerge from the germinal centers, are attracted by homing receptors, including chemokine receptor ligands, in peripheral lymphatic tissue (e.g., the gastrointestinal tract) and the bone marrow, causing their migration to the respective tissues. Because the eye and the CNS in humans is devoid of a germinal center structure, our and others’ findings imply that PIOL/PCNSL are tumors that arise initially in an extraneural germinal center environment. Subsequent localization to the retina and/or to the CNS may involve the development of a “neurotropic” cellular phenotype, the influence of chemokine receptors and their ligands,
53 54 and/or the incomplete destruction of the neoplastic cells by the host immune system.
In summary, our data indicate that (1) PIOLs and PCNSLs demonstrated Ig expression more frequently than did peripheral DLBCL (92% vs. 54%); (2) similar to peripheral DLBCLs, all PIOLs and PCNSLs expressed BOB.1/OBF1 and Oct.2; (3) an aberrant coexpression of BCL-6 and MUM1/IRF4 and of MUM1 and PAX5 was present in most PIOLs, PCNSLs, and DLBCLs; (4) nearly all PIOLs and PCNSLs were immunonegative for PU.1, differing from peripheral DLBCLs, 46% of which were positive for this transcription factor. The latter finding may be explained by the greater degree of plasmacellular differentiation, possibly suggesting a more mature phenotype of PIOL and PCNSL, when compared with peripheral DLBCL. Our data provide further support for the notion that peripheral and “centrally” located DLBCLs differ in clinical and genotypic features, despite their similar morphologic characteristics. Further studies are needed to determine the exact cell of origin of PIOLs and PCNSLs and to establish whether these neoplastic precursors indeed arise in an extraneural location initially before relocalization to cerebral and ocular tissues; and to determine whether the ability of tumor cells to express Ig and its various transcription factors is of prognostic value in PIOL, PCNSL, and peripheral B-cell non-Hodgkin lymphoma.