Loss of the B-lineage–specific gene expression program in Hodgkin and Reed-Sternberg cells of Hodgkin lymphoma

Classical Hodgkin lymphoma (cHL) is characterized by the presence of rare typical Hodgkin and Reed-Sternberg (HRS) cells, which represent the malignant cells of the tumor. The origin of these cells was controversially discussed until molecular studies showed their clonality and lymphoid derivation. In most cases, the malignant cells are derived from germinal center (GC) B cells, whereas in few cases the HRS cells show a T-cell origin.1 In B-cell–derived cases, the occurrence of somatic deleterious (“crippling”) mutations in the immunoglobulin genes of HRS cells in about 25% of the cHL cases has led to the idea that these cells are derived from preapoptotic GC B cells.1 2 

One reason for the longstanding controversy about the cellular origin of HRS cells was their unusual immunophenotype, which cannot be related to any normal cell of the hematopoietic lineage. Typical B-lymphocyte markers such as CD19, CD20, CD22, and the common leukocyte marker CD45 are either not expressed or by only a small proportion of the malignant cells.3-5 Additionally, B-cell receptor (BCR) itself as well as transcription factors important for proper B-cell function such as PU.1, Oct-2, and BOB-1 have been shown to be down-regulated in HRS cells.6-8 Surprisingly, Pax-5, a transcription factor essential for B-cell commitment in early B cells and also maintenance of B-cell identity in mature cells, is still expressed in HRS cells in the majority of cases.9 On the other hand, inappropriate lineage marker expression of dendritic cells, monocytes, or macrophages can be found in HRS cells.1 

Unlike in cHL, the tumor cells in B-cell non-Hodgkin lymphoma (NHL) usually retain expression of typical B-lineage markers. In addition, the malignant cells show a differentiation-linked phenotype, preserving the differentiation state of their normal counterparts during their oncogenic transformation.10,11 This is not only the case for the expression of differentiation markers such as CD77 in Burkitt lymphoma or Bcl-6 in follicular lymphomas, which are normally expressed by GC B cells, but also for complex biologic features as the ongoing hypermutation process and the retained follicular growth pattern in follicular lymphoma.12-14 

To comprehensively analyze to what extent the B-lineage identity is affected in HRS cells, we compared large-scale gene expression profiles from HL cell lines and normal B-cell subsets regarding the expression of B-lineage–associated genes. This analysis was performed using the serial analysis of gene expression (SAGE) technique and oligonucleotide DNA chip microarrays.

The HL cell lines used originate from patients with HL of the nodular sclerosis (NS; L428 and HDLM2) or mixed cellularity (MC) subtype (KMH2 and L1236). HDLM2 is of T-cell origin; the 3 other HL cell lines are of B-cell origin.15-17 All cell lines were grown in RPMI 1640 with Glutamax-1 (Invitrogen, Karlsruhe, Germany), supplemented with 10% fetal calf serum (FCS) and 100 U/mL penicillin/streptomycin at 37°C in an atmosphere containing 5% CO₂.

Tonsils were taken from patients during routine tonsillectomy and mononucleated cells were purified by Ficoll density centrifugation. Centroblasts were stained consecutively with rat anti-CD77 (Immunotech, Hamburg, Germany), mouse antirat IgM (Serotec, Eching, Germany) and antimouse IgG1 microbeads (Miltenyi Biotec, Bergisch-Gladbach, Germany) and enriched by magnetic cell separation (Miltenyi Biotec). The purity of CD77⁺ cells used for this study was higher than 95%. In total, 2.5 × 10⁸ centroblasts pooled from 8 donors were used for SAGE.

SAGE, developed by Velculescu et al, generates a quantitative gene expression profile by high-throughput sequencing.18The method is based on the principle that a nucleotide sequence of 10 bases length (a gene tag) can mostly be assigned to a unique transcript, if its position in the mRNA molecule is known (downstream of the NlaIII restriction site nearest to the poly A sequence). Total RNA from centroblasts and the L1236 cell line was extracted with Trizol (Gibco). Poly-A⁺ RNA was selected using the Oligotex mRNA kit (Qiagen, Hilden, Germany) and 5 μg was used to synthesize ds-cDNA with the cDNA synthesis system (Gibco). SAGE was performed for both cell populations following the protocol version 1.0 d, available fromhttp://www.sagenet.org/sage.htm, with the following slight modifications. As described by Powell,19 biotinylated primers were introduced in the ditag-polymerase chain reaction (ditag-PCR), and after the BsmF1 reaction magnetic beads coupled to streptavidin (Dynal, Osly, Germany) were used to purify the resulting ditags. Additionally, a heating step was included before gel purification of the concatemers.20 SAGE profiles were generated by sequencing about 30 000 tags of both cell populations. The tag numbers given in the text and tables are normalized to 30 000 and rounded up or down. The data were analyzed using SAGE Software 2000, Version 4.12.

RNA isolation and cDNA as well as cRNA synthesis were performed as described.21 The cRNA was fragmented according to the Affymetrix (High Wycombe, United Kingdom) protocol, and 15 μg biotinylated cRNA was hybridized to U95A microarrays (Affymetrix), on which approximately 12 000 cDNAs are represented by oligonucleotide probes. In a supervised analysis, the 4 HL cell lines, as well as 20 samples corresponding to the normal B-cell subsets (5 samples each naive B cells, GC centrocytes, GC centroblasts, and memory B cells, which were purified by magnetic cell separation) were defined as separate clusters, and the differentially expressed genes were identified by the Genes@Work software platform (IBM, White Plains, NY), which is a gene expression analysis tool based on the pattern discovery algorithm called structural pattern localization analysis by sequential histograms (SPLASH).22 

The SPLASH software used for data evaluation is designed to identify genes consistently differentially expressed between groups of samples, so that we needed to use at least 4 HL cell lines and chose the ones most widely accepted as being derived from HRS cells. One of the 4 HL cell lines included in the microarray analysis (HDLM2) is of T-cell origin.17 The inclusion of this line does not hamper the identification of down-regulated B-cell–specific genes in the 3 B-cell–derived HL lines, because HDLM2 cells, as expected, lack B-cell marker expression and the software identifies only genes consistently down-regulated in all 4 cell lines. Moreover, inclusion of the HDLM2 cell line lends additional weight to the identification of lymphocyte- and hematopoietic-specific genes showing reduced expression in HRS cells.

The isolation of the normal B-cell subsets and the computational analysis is described elsewhere.21 Normal B-cell subsets were isolated according to the following marker expression: naive B cells (CD27⁻, CD10⁻, CD3⁻, CD14⁻, and IgD⁺), centroblasts (CD77⁺), centrocytes (CD77⁻, CD39⁻, CD3⁻, and CD10⁺), and memory B cells (CD10⁻, CD3⁻, CD38^low, CD14⁻, and CD27⁺).

To search for genes specifically expressed in B cells we screened the PubMed database with search terms such as “B cell-specific expression,” “B cell-restricted expression,” “B-cell marker human staining.” The vast majority of these genes was identified as being down-regulated in HL cell lines (Table1; Figure1).

Total RNA of HL cell lines was isolated with Trizol (Gibco) followed by cDNA synthesis using the Omniscript reverse transcription (RT) kit (Qiagen). RT-PCR for the complete coding sequence ofPax-5 was performed using aliquots corresponding to 30 ng total RNA in a final volume of 30 μL containing 100 μM of each deoxyribonucleoside triphosphate (dNTP), 1 × High Fidelity PCR buffer (Roche, Mannheim, Germany), 2 mM MgCl₂, 2 U High Fidelity DNA polymerase mix (Roche), and 125 nM of each primer (Pax-5 forward: 5′-AAA AAG GCA CAA AAA AGT GGA AAC-3′ and Pax-5 reverse: 5′-ACC CTC AAT AGG TGC CAT CAG T-3′). The amplification program consisted of 60 seconds at 95°C, followed by 30 cycles of 50 seconds at 95°C, 30 seconds at 59°C, 60 seconds at 72°C, and finally 5 minutes at 72°C. RT-PCR products (1287 bp) were gel-purified and directly sequenced with the amplification primers, as described. To obtain sequence information about the complete coding sequence, PCR products were additionally sequenced with 4 internal oligonucleotides (5′-TGA CAC CGT GCC TAG CGT CAG-3′, 5′-GAC CGC GTG TTT GAG AGG CAG-3′, 5′-GTG GTC TGC TCG GGC TTG ATG G-3′, 5′-CAC TAT GCT GTG ACT GGA AGC TG-3′).

Immunohistochemical stainings were performed on sections of tonsils obtained by routine tonsillectomy and in sections of frozen (RP105, BCMA, BLNK, CD52, Blk, SIT, and Lck) and paraffin-embedded (Syk and SWAP-70) lymph node biopsies from patients with cHL. Cases 7, 8, 12, and 13 were of the NS subtype of cHL and the remaining cases were of the MC subtype of cHL. The sections were stained with anti-RP105 (CD180; 1:100; Bioscience, San Diego, CA), anti-Lck (1:100; Santa Cruz Biotechnologies, Santa Cruz, CA), anti-Syk (1:500; Santa Cruz Biotechnologies), anti-BCMA (1:100; Santa Cruz Biotechnologies), anti-BLNK (1:20, Santa Cruz Biotechnologies), anti-Blk (1:25, Serotec), anti-SIT (1:50; Abcam, Cambridge, United Kingdom), and anti-CD52 (1:80, Biosource, Solingen, Germany) followed by a standard ABC staining and developed with Fast Red or diaminobenzedine tetrahydrochloride (DAB) substrate (Dako, Hamburg, Germany). Epstein-Barr virus (EBV) status of patients with cHL was investigated by EB-encoded small RNA (EBER) in situ hybridization.23 

SAGE profiles consisting of about 30 000 tags were generated from tonsillar centroblasts, the putative nonmalignant counterpart of HRS cells, and the L1236 cell line with confirmed origin from the HRS cells of a patient with cHL.15 16 SAGE software was used to identify genes that were deregulated by at least a factor of 5.0, and 280 distinct transcripts were found to be down-regulated in the HRS cells according to this criterion. The analysis of genes up-regulated in HRS cells will be published elsewhere. Because we were interested in the down-regulation of genes normally expressed in B cells, we analyzed the expression profiles for such genes, dividing them into 3 classes according to their expression pattern: B-cell–, lymphoid-, and hematopoietic-specific (defined here collectively as B-lineage–specific genes). Intriguingly, 31 of the down-regulated transcripts represented B-lineage–specific genes (Table 1). Among the genes showing a 2-fold to less than 5-fold difference in gene expression level, down-regulation of 5 additional B-lineage markers was identified (Table 1).

Most of the B-cell markers known to be down-regulated in HRS cells (eg, CD19, CD20, CD45, CD79a/b, OBF-1, Oct-2)4,5,7 8 were also identified in this analysis. However, the majority of B-lineage genes identified as down-regulated in this study have not been described in cHL so far. In particular, we note lack of or reduced expression of many molecules important in BCR signaling, for example, multiple protein tyrosine kinases (Syk, Lyn, Blk) or the BLNK protein, but also of important transcription factors such as Spi-B, Lyl-1, and A-myb (Figure 1; Table 1).

The extensive down-regulation of B-lineage genes prompted us to verify and expand this finding by investigating gene expression profiles generated with microarrays (Affymetrix). In this analysis, 4 HL cell lines, L1236, L428, KMH2, and HDLM2, were defined as a cluster (see “Materials and methods”), and were compared with the main normal B-cell populations (defined as the second cluster), namely, 5 samples each of naive B cells, CD77⁺ GC cells (centroblasts), CD77⁻ GC cells (centrocytes), and memory B cells.21 This comparison allowed us to investigate whether the absence of the B-lineage–specific transcription program is a feature of all HL cell lines analyzed. The set of genes found to be significantly down-regulated in all HL cell lines revealed an extensive defect in B-lineage gene expression that resembled the results of the SAGE study (Figure 1). In total, 374 entries showed a reduced expression in HRS cell lines, from which 46 genes known as B-lineage–specific genes could be identified.

Most B-lineage genes found to be down-regulated in the SAGE profile were also identified as being down-regulated in the microarray analysis (30 genes marked with ‘S’ in Figure 1), providing a cross-validation for both methods. However, within both analyses, some genes were identified that were not present in the other. Detection of transcripts in the SAGE profile may be restricted by the size of the profile (refer to the footnote of Table 1), and polymorphisms or lack of NlaIII restriction sites in certain genes. On the oligonucleotide microarray, information of about 12 000 cDNAs is represented, which do not cover all transcripts expressed in a cell, and for some transcripts the oligonucleotides may be suboptimal. Differences between the 2 data sets may additionally be due to the different B-cell populations with which the HL cell lines were compared. For example, A-myb as a GC-specific transcription factor was not identified in the microarray analysis, because other normal B-cell populations besides GCB cells do not express A-myb, so that it is not consistently expressed in normal B cells.

Down-regulated protein expression of several B-cell markers has already been published (CD19, CD20, Oct-2, OBF-1).5,7 8For 9 additional markers we investigated protein expression by immunohistochemical staining, performed on frozen or paraffin-embedded sections of reactive tonsils (Table 2) and of lymph node biopsies of cHL (Table 2; Figure2). RP105 is a Toll-like receptor with reported expression on most mature B cells and importance in lipopolysaccharide (LPS)–induced B-cell activation. Stainings were also performed for Syk, Lck, and Blk, 3 tyrosine kinases. Syk is essential for BCR and general immunoreceptor signaling. Lck is reported to be expressed weakly in resting B cells but up-regulated on B-cell stimulation. Blk is also thought to play a role in BCR signal transduction. BCMA is an integral membrane protein in the Golgi apparatus and is expressed primarily in mature B cells. BLNK, a protein with scaffold function, is critically involved in BCR signaling. CD52 is a glycopeptide antigen highly expressed on T and B cells. SWAP-70, originally suggested to be involved in switch recombination, is found in the cytoplasm of resting B cells and translocates to the nucleus on cellular activation. SIT is a SHP2-interacting transmembrane adaptor protein that is thought to be selectively expressed in lymphocytes.

Regarding the lymph node biopsies of cHL, no protein expression of Syk, BLNK, and CD52 was detected in HRS cells in any of the cases investigated (Table 2; Figure 2). Lck, RP105, SIT, and SWAP-70 expression was found in one case each, with only some HRS cells being positive in these cases. For BCMA we found a more diverse pattern because 3 cases of 12 showed protein expression in some HRS cells. Overall, the results of the immunohistochemical stainings of lymph node sections of cHL are thus largely consistent with the mRNA data and confirm the down-regulated expression of these 9 B-cell–lineage proteins in HRS cells in the majority of cases analyzed.

The observed extensive defect in the B-lineage–specific gene expression program in HRS cells prompted us to search for retained expression of B-cell markers in the HL cell lines. Because non–B-cell profiles were not available to identify genes specifically expressed in B cells, we decided to manually screen the expression data for most of the established B-cell–specific genes (see “Materials and methods”). Only very few additional B-cell markers were identified as being consistently expressed in HRS cells. In addition to the expression of Pax-5 (shown by RT-PCR; see “Analysis of Pax-5”), only the mRNA molecules for B7.1/B7.2, CD40, and several major histocompatibility complex (MHC) class II alleles were found in HRS cells at a similar level as or an even higher level than that in GC and memory B cells (data not shown), as already known from previous studies (for reviews, see Küppers1 and Poppema24). The expression of these molecules may point to an antigen-presenting function of HRS cells.24 25 

Pax-5 is a key transcription factor involved in the establishment and maintenance of B-cell identity. Paradoxically, Pax-5 is expressed by most HRS cells, but many genes known to be positively regulated by Pax-5 (eg, CD19, CD21, CD22,CD72, BLNK, CD79a; Table 1; Figure1),26,27 are down-regulated in HRS cells.9 We therefore analyzed the 4 HL cell lines by RT-PCR and found Pax-5 expression in the 3 cell lines of B-cell origin, in line with previous data.9 The sequence analysis showed no mutations in the coding sequence in any cell line. One conservative base substitution was detected in 2 of the cell lines, presumably representing a genetic polymorphism. The lack of mutations argues against an intrinsic defect in Pax-5 protein function as a reason for the lost B-lineage phenotype of HRS cells.

In this study, comprehensive gene expression profiles from HL cell lines and normal B-cell subsets were analyzed regarding the B-cell phenotype of HRS cells. It is already known from previous studies that HRS cells often lack expression of several B-lineage markers (ie, B-cell–, lymphoid-, hematopoietic-specifically expressed genes), such as CD19, CD20, CD45, CD79a/b, OBF-1, and PU.1.3-8 Reduced expression of these molecules in HL cell lines was also observed in this study. We show here that there is not only down-regulation of single B-lineage markers, but rather a defect in the B-lineage gene expression program as such. RNA expression of nearly all known genes specific for B cells, lymphocytes, and hematopoietic cells is either lacking or greatly reduced in HRS cells (Figure3). For 9 of these 45 down-regulated genes, down-regulation of protein expression was confirmed for HRS cells on tissue sections. In contrast to these data, a large-scale expressed sequence tag (EST) analysis reported a gene expression pattern of HRS cells supporting a B-cell origin.28 However, as the cDNA data represent a mixture of sequences from HRS cells of cHL and Hodgkin cells of lymphocyte predominant HL (L&H cells), the latter of which being known to phenotypically resemble B cells,29 the observed B-cell phenotype likely reflects the cDNA population of the L&H cells.

Among the genes that were previously unknown to be down-regulated, we identified genes coding for molecules involved in LPS-induced B-cell activation (RP105, MD-1), surface markers (CD37, CD53), the tumor necrosis factor (TNF) receptor BCMA, and the transcription factors Spi-B and Lyl-1. Numerous other molecules acting in signal transduction showed reduced or undetectable expression. In particular, various components of the BCR signaling pathway seem to be severely affected. This involves (1) the tyrosine kinases Syk, Lyn, and Blk, normally activated by BCR crosslinking30; (2) the scaffold protein BLNK, which is phosphorylated by Syk and acts by recruiting tyrosine kinases and downstream signaling molecules to the BCR31; (3) PLC-γ and Vav, signal intermediates that control Ca²⁺flux and protein kinase C (PKC) and mitogen-activated protein (MAP) kinase activation; and (4) the immunoreceptor tyrosine-based inhibition motif (ITIM)–containing molecule CD72 and the phosphatase SHP-1, which are involved in negative regulation of BCR signaling.32,33 Thus, HRS cells have not only lost BCR expression but also the corresponding signal transduction machinery. It is therefore unlikely that in HRS cells another factor, such as the EBV-encoded LMP2A protein,34 replaces the function of the BCR by using its classical signaling pathway.

Some of the molecules discussed play a role also in pathways other than BCR signaling. The down-regulation of these factors may consequently affect the activity of those pathways in HRS cells. For example, as SHIP negatively regulates signaling through the chemokine receptor CXCR4,35 its down-regulation could result in enhanced activity of this receptor. SHP-1 normally acts in negative regulation of the JAK/STAT pathway36 which is activated, for example, by cytokine signaling. Interestingly, STAT3 has been shown to be constitutively active in HRS cells.37 38 The down-regulation of SHP-1 may contribute to this constitutive activation.

On the basis of the data presented, we propose that the lack of expression of B-lineage–specific genes in HRS cells reflects a fundamental defect in maintaining the gene expression program of B cells in HRS cells. Several explanations for this phenomenon can be envisioned.

The decreased mRNA levels of several genes identified in this analysis (eg, CD20, CD19, CD22, Blk, and EBF),39 and also expression of Syndecan-1 or MUM-1 reported for cHL suggest a phenotypical similarity of HRS cells and plasma cells.40-42 However, down-regulated expression of Ig genes, BCMA, and Oct-243,44 but retained expression of Pax-5 are inconsistent with terminal differentiation of B cells.45 46 It can, however, not be excluded that B cells of a particular differentiation stage (preplasma cells?) exist, which may partly resemble HRS cells.

Because the Pax-5 protein, which is important for establishment and maintenance of B-cell identity,26 27 is expressed in most cases of cHL, we searched for mutations interfering with protein function. However, the lack of mutations in the Pax-5 coding sequence in the 3 HL cell lines analyzed here argues against a functional defect in this protein. This may point to a defect of a (yet unknown) factor essential for B-lineage identity.

Another explanation for loss of the B-lineage gene expression program could be its suppression by a dominant factor. A protein known to play a role in a variety of cell fate decisions is the Notch receptor, which is also involved in the decision between T- and B-cell fate. Overexpression of a constitutively active Notch 1 transgene results in a block in B-cell development and an increased T-cell number in the bone marrow.47 Interestingly, overexpression of Notch 1 (and Notch 2, whose function in hematopoietic cells is, however, unclear) has been observed in primary HRS cells and HL cell lines.48,49 However, it is questionable whether Notch contributes to the lost B-lineage identity, because the Notch pathway seems not to be intrinsically activated in HL cell lines49and its activity in primary HRS cells remains to be clarified.

The loss of tissue-specific markers is a well-known feature of many (mainly progressed) tumors, a process also termed dedifferentiation.50,51 However, malignant B cells normally express markers and retain complex biologic features attributable to specific B-cell differentiation stages (see “Introduction”). It has been suggested that this retained differentiation pattern in B-cell NHL relies on an intimate linkage of the transforming event(s) and the differentiation program of the progenitor cell (“differentiation-linked lymphomagenesis”).10 11 This hypothesis is supported mainly by the fact that most B-cell NHLs carry translocations of oncogenes into the Ig locus and that the function of these oncogenes is dependent on proper transcriptional activity of the Ig loci. However, this idea does not hold true for HRS cells because the B-cell phenotype appears not to be essential for, or even incompatible with, the pathogenesis of cHL. Hence, one may speculate that the nature of at least one transforming event in cHL must differ fundamentally from those causing other B-cell lymphomas.

The present results offer a straightforward explanation for the puzzling finding that the B-lineage–derived HRS cells persist without a functional BCR. In mature B cells BCR signaling plays a vital role52 and GC B cells that have lost BCR expression because of somatic mutations rapidly undergo apoptosis.53,54 It is likely that this dependence on BCR expression is based on a B-lineage–specific gene expression program, which is defective or simply lacking in HRS cells. Regarding the proposed origin of HRS cells from preapoptotic GC B cells,1 2 events leading to the loss of the B-lineage–specific transcription program may thus prevent the induction of apoptosis in these cells.

In conclusion, loss of B-lineage identity may allow HRS cells to survive in the absence of BCR signaling. The fundamental defect in maintaining the B-cell differentiation program is a hallmark of cHL, stands in contrast to other B-cell NHL, and may be caused by a novel, as yet undefined, transforming event in the pathogenesis of cHL.

We are grateful to Yvonne Blum, Michaela Fahrig, Christine Gerhardt, Julia Jesdinsky, Tanja Schaffer, and Vladan Milijkovic for excellent technical assistance and to Julia Kurth and Ulla Strobl for critically reading this manuscript.