BCL6 – regulated by AhR/ARNT and wild-type MEF2B – drives expression of germinal center markers MYBL1 and LMO2

Abstract

Genetic heterogeneity is widespread in tumors, but poorly documented in cell lines. According to immunoglobulin hypermutation analysis, the diffuse large B-cell lymphoma cell line U-2932 comprises two subpopulations faithfully representing original tumor subclones. We set out to identify molecular causes underlying subclone-specific expression affecting 221 genes including surface markers and the germinal center oncogenes BCL6 and MYC. Genomic copy number variations explained 58/221 genes differentially expressed in the two U-2932 clones. Subclone-specific expression of the aryl-hydrocarbon receptor (AhR) and the resulting activity of the AhR/ARNT complex underlaid differential regulation of 11 genes including MEF2B. Knock-down and inhibitor experiments confirmed that AhR/ARNT regulates MEF2B, a key transcription factor for BCL6. AhR, MEF2B and BCL6 levels correlated not only in the U-2932 subclones but in the majority of 23 cell lines tested, indicting overexpression of AhR as a novel mechanism behind BCL6 diffuse large B-cell lymphoma. Enforced modulation of BCL6 affected 48/221 signature genes. Although BCL6 is known as a transcriptional repressor, 28 genes were up-regulated, including LMO2 and MYBL1 which, like BCL6, signify germinal center diffuse large B-cell lymphoma. Supporting the notion that BCL6 can induce gene expression, BCL6 and the majority of potential targets were co-regulated in a series of B-cell lines. In conclusion, genomic copy number aberrations, activation of AhR/ARNT, and overexpression of BCL6 are collectively responsible for differential expression of more than 100 genes in subclones of the U-2932 cell line. It is particularly interesting that BCL6 – regulated by AhR/ARNT and wild-type MEF2B – may drive expression of germinal center markers in diffuse large B-cell lymphoma.

Introduction

Diffuse large B-cell lymphoma (DLBCL) is an aggressive non-Hodgkin lymphoma. Expression array analyses have identified two molecularly distinct forms of the tumor, termed germinal center (GC) and activated B-cell (ABC) forms.1 Primary mediastinal B-cell lymphoma is a third variant of this tumor marked by its characteristic gene expression signature.2 Diverse molecular and cytogenetic alterations characterize the three subtypes of DLBCL.3 Translocations affecting BCL2 typically occur in the GC subtype while somatic mutations involving the NFκB pathway are more often found in ABC DLBCL.43

Cancers evolve under selective forces, including host immunity and drug inhibition, fuelled by mutational alterations explaining the clonal heterogeneity found in tumors including DLBCL.65 Clonal evolution may also be the reason for the development of molecularly related tumors in one patient. Thus, up to 30% of nodular lymphocyte-predominant Hodgkin lymphomas transform into DLBCL.7 Furthermore, clonal relationships have been reported for single patients with classical Hodgkin lymphoma and non-Hodgkin lymphoma.98

The U-2932 cell line was derived from a DLBCL patient with a history of Hodgkin lymphoma.10 According to the results of gene expression analysis, U-2932 meets the criteria for an ABC DLBCL cell line.11 We recently showed that U-2932 comprises two subclones (R1 and R2) derived from a presumptive mother clone with genomic BCL2 amplification which acquired distinct sets of secondary rearrangements leading to the alternative overexpression of BCL6 or MYC in the respective daughter clones.12 Immunoglobulin gene hypermutation analysis showed that R1 and R2 represent subclones of the original tumor.12 Thus, the two U-2932 subclones seemed to be ideal to study the cellular consequences of clonal evolution. More than 200 genes showed >10-fold expression differences between R1 and R2.12 Hence, the two U-2932 subclones sensitively model the cellular consequences of clonal evolution in vitro. Here, we set out to identify molecular causes for subclone-specific gene expression, a necessary first step in addressing the serious therapeutic challenges posed by tumor heterogeneity.

Methods

Human cell lines and treatments

All cell lines are held by the DSMZ cell line bank ( http://www.dsmz.de ) and were cultured as described previously.13

Knock-down experiments

Transfection with short interfering (si) RNA oligonucleotides was performed as described elsewhere.14 Gene-specific siRNA oligonucleotides and AllStars negative control siRNA were obtained from Qiagen (Hilden, Germany). The siRNA (80 pmol) were transfected into 1×10 cells by electroporation using an EPI-2500 impulse generator (Fischer, Heidelberg, Germany) at 350 V for 10 ms. Treated cells were harvested after 20 h.

Cell surface marker analysis and cell sorting

Cells were immunophenotyped and sorted on a FACSAriaIII (Becton Dickinson; Heidelberg, Germany). PerCP-Cy5.5-conjugated CD20 and APC-conjugated CD38 antibodies were purchased from Becton Dickinson, and APC-conjugated CD22, CD27 and CD59 antibodies were from Life Technologies (Darmstadt, Germany). The APC-conjugated CD24 antibody was obtained from Miltenyi Biotec (Bergisch Gladbach, Germany). U-2932 subclones R1 (CD20/CD38) and R2 (CD20/CD38) were phenotypically stable for more than 100 days in culture. During the period of this study, resorting was not necessary.

Cytogenetic analysis

CytoScan HD Array (Affymetrix; Santa Clara, CA, USA) hybridization analysis was performed to identify numerical aberrations. DNA was prepared using the Qiagen Gentra Puregene Kit (Qiagen; Hilden, Germany). Data were analyzed using the Chromosome Analysis Suite software version 2.0.1.2 (Affymetrix). Fluorescence in situ hybridization (FISH) was performed using bacterial artificial chromosome (BAC) clones (BACPAC Resources; Oakland, CA, USA) to analyze gene rearrangements at BCL6 using protocols described previously.1615

DNA microarray hybridization and quantitative genomic polymerase chain reaction analysis

Total RNA (500 ng) was used for biotin labeling according to the 3′ IVT Express Kit (Affymetrix). The biotinylated cRNA (7.5 μg) was fragmented and placed in a hybridization cocktail containing four biotinylated hybridization controls (BioB, BioC, BioD, and Cre). Samples were hybridized to an identical lot of Affymetrix GeneChip HG-U133 Plus 2.0 for 16 h at 45°C. Steps for washing and SA-PE staining were processed on the fluidics station 450 using the recommended FS450 protocol (Affymetrix). Images were analyzed on a GCS3000 Scanner with the GCOS1.2 Software Suite (Affymetrix). Data were analyzed using GeneSpring 11.5.1 (Agilent Technologies; Santa Clara, CA, USA). Signal intensities (raw data) were log₂-transformed and normalized using RMA.

Quantitative genomic polymerase chain reaction was performed on a 7500 Applied Biosystems real-time polymerase chain reaction system using the SYBR green assay (Applied Biosystems; Darmstadt, Germany) with ABL1 as the internal control and the diploid cell line NC-NC as a reference. Primers are shown in Online Supplementary Table S1.

Western blot analysis

Samples were prepared as described by Quentmeier et al.17 We used the nuclear extract kit (Active Motif; La Hulpe, Belgium) to separate cytoplasmic and nuclear proteins. Bands on nitrocellulose membranes were visualized with the biotin/streptavidin-horseradish peroxidase system (GE Healthcare; Little Chalfont, UK) in combination with Renaissance Western Blot Chemoluminescence Reagent (Perkin Elmer; Waltham, MA, USA).

Statistical analysis

Data were organized using commonly employed spreadsheet programs. Statistical tests were performed using the R software, as described in Online Supplementary Table S2.18 For analysis of transcription factor binding sites, we made use of a commercial bioinformatics service (Biobase; Wolfenbüttel, Germany).

Results

Numerical aberrations causing differential gene expression in subclones of the diffuse large B-cell lymphoma cell line U-2932

The DLBCL cell line U-2932 comprises two populations distinguishable by their expression of various B-cell markers including CD20 and CD38. According to the results of immunoglobulin hypermutation and cytogenetic analyses, the populations – named R1 and R2 - represent two subclones of the original tumor.12 Here, we set out to determine why more than 200 genes showed gross (>10-fold) expression differences between R1 and R2 (Online Supplementary Table S3).

We applied a whole genome array to investigate the extent to which subclone-specific gene expression might be attributable to numerical aberrations. Twenty-six percent (58/221) of the differentially expressed genes showed concordant numerical disparities between R1 and R2 (Online Supplementary Table S3). Statistical analysis of copy number aberrations and expression data revealed that numerical differences between the two subclones effectively predict differences in gene expression (sensitivity 0.64; specificity 0.94; accuracy 0.78). A McNemar chisquared test with continuity correction rejected non-correlation of the two parameters with a P-value of 0.0036 (P values <0.05 are considered statistically significant). In fact, for several genes, the correlation between ploidy status and expression level was so stringent that a causal relation could be directly inferred: CHMP2B and CGGBP1 on chr 3(p11.1-2) were highly amplified in R1 (13 n) and hemizygously lost in R2 (1 n) with corresponding differences in gene expression (Figure 1A,B, Online Supplementary Figure S1). Loss of expression in population R2 as a result of homozygous deletion was found for ITM2B and RB1 on chr 13(q14.2) (Figure 1A,B, Online Supplementary Figure S1).

Figure 1.Copy number aberrations of U-2932 subclones R1 and R2. (A) Genomic copy number aberrations according to CytoScan HD Array analysis. (B) Results of quantitative genomic polymerase chain reaction showing the ploidy status of genes in regions of interest. Pink: U-2932 R1; blue: U-2932 R2. Normal cell line NC-NC was used as a diploid reference and ABL1 as an endogenous control. Error bars indicate the standard deviation (SD) (n=3). Benjamini-Hochberg corrected t-test P-values: P<0.05 (*), <0.01 (**), <0.001 (***). In addition to CHMP2B and CGGBP1, C3orf38 and ZNF654 were also affected by chr 3(p11.1-2) amplification; further targets of chr 13(q14.2) deletion besides ITM2B and RB1 were CAB39L, CDADC1, and FNDC3A (Online Supplementary Table S3).

Thus, for a sizable minority of genes (≤26%) numerical aberrations provided a potential explanation, in some cases compelling, for the differences in gene expression between R1 and R2, leaving the question of what caused expression discrepancies for the remaining 163 genes unaffected by concordant copy number aberrations.

AhR/ARNT activity driving gene expression in U-2932 subclone R1

Constitutive activation of signaling chains can trigger aberrant expression of anti-apoptotic and proliferative genes, thereby promoting lymphomagenesis. Notable in this context was that the canonical AhR/ARNT target CYP1A1 was high in R1 and low in R2 (Figure 2A). AhR levels paralleled expression of CYP1A1, suggesting that R1-high genes might be controlled by AhR expression (Figure 2A).

Figure 2. AhR/ARNT and potential targets in subclones of the DLBCL cell line U-2932. (A) Transcript levels were assessed by quantitative real-time polymerase chain reaction. (B) Protein expression was determined by western blot analysis. Note the positive correlation between expression of BCL6 and its potential transcriptional regulators MEF2B and AhR in subclones of the DLBCL cell line U-2932. Error bars indicate the standard deviation (n=3). Benjamini-Hochberg corrected t-test P-values: P<0.05 (*), <0.01 (**), <0.001 (***).

It is widely believed that ligand-activated AhR translocates to the nucleus where it binds to ARNT. AhR/ARNT regulation in B cells may operate differently. Normal, resting B cells express little AhR. Upon stimulation (e.g. with CD40 ligand) AhR is up-regulated and translocates to the nucleus in the absence of exogenous AhR ligand, explaining how B-cell lymphoma cell lines show constitutive nuclear localization of the AhR/ARNT complex.2119 In accordance with this observation, AhR localized to the nucleus of U-2932 cells, independently of ligand stimulation in both subclones (Online Supplementary Figure S2). Thus, the major difference between subclones R1 and R2 concerned AhR expression level rather than subcellular localization (Figure 2A,B).

We, therefore, hypothesized that AhR regulation might underlie differential gene expression in the two U-2932 subclones. To find out how many of the R1-specific genes (126/221 genes with >10-fold expression difference were high in R1) were affected by AhR/ARNT activity, we treated U-2932 R1 cells with the AhR/ARNT inhibitor GNF351 (1 μM, 24 h). GNF351 treatment effected relocalization of AhR from the nucleus to the cytoplasm and led to down-regulation of the AhR/ARNT target CYP1A1 (Figure 3A,B). Eleven of the 126 R1-high genes (9%) were inhibited by the AhR/ARNT inhibitor, confirming that the difference in AhR levels contributed to non-genomic divergences in gene expression between the two subclones (partially shown in Figure 3B).

Figure 3.Effect of the AhR/ARNT inhibitor GNF351 on U-2932 R1 cells. (A) Translocation of AhR into the cytoplasm after stimulation with GNF351 (1 μM). C: cytoplasm; N: nucleus. Protein expression was determined by western blot analysis. (B) Inhibitory effect of GNF351 (1 μM) on expression of target genes including MEF2B. (C) Knockdown of ARNT led to reduced expression of MEF2B. (D) Knockdown of MEF2B led to reduced expression of BCL6. (B-D) RNA expression levels were determined by quantitative real-time polymerase chain reaction. Error bars indicate the standard deviation (n=3). Benjamini-Hochberg corrected t-test P-values: P<0.05 (*), <0.01 (**), <0.001 (***).

AhR has been described as an epigenetically regulated gene.2322 To test whether AhR expression differences in DLBCL cell lines were subject to epigenetic regulation we treated AhR-positive (SU-DHL6, U-2932 R1) and AhR–negative (OCI-LY19, U-2932 R2) cell lines with inhibitors of histone deacetylation and DNA methylation. Both types of inhibitor significantly induced AhR expression in the negative cell lines, suggesting that differences at the epigenetic level were responsible for the expression of AhR in DLBCL cell lines (Table 1). This view was supported by the finding that the AhR promoter of the cell line most sensitive to inhibition of DNA methylation (OCI-LY19) was highly methylated (Online Supplementary Figure S3).

Table 1.Effect of 5-Aza-2′ deoxycytidine and trichostatin A on AhR expression.

MEF2B: a transcriptional regulator of BCL6

One target of the AhR antagonist was MEF2B, member of the ‘myocyte enhancer-binding factor 2′ family of transcription factors (Figure 3B). Confirming that MEF2B lies downstream of AhR/ARNT, ARNT knockdown reduced MEF2B expression in U-2932 R1 cells (Figure 3C). MEF2B is a transcriptional regulator that cooperates with co-repressors and histone-modifying enzymes.2724 One target of MEF2B is BCL6, a proto-oncogene selectively expressed in GC B-cells.28 Thus, it was not surprising that AhR/ARNT inhibition not only affected MEF2B but also BCL6 (Figure 3B). Confirming previously published data, knockdown of MEF2B down-regulated BCL6 (Figure 3D).28 This, together with the U-2932 subclone-restricted expression pattern of AhR, MEF2B and BCL6, indicated that overexpression of AhR might well explain deregulation of the GC oncogene BCL6 (Figure 2A,B).

To test whether AhR, MEF2B and BCL6 showed coordinated regulation in DLBCL cell lines in general, we analyzed the expression of these genes and of ARNT in a panel of 23 DLBCL cell lines. ARNT was constitutively expressed in all cell lines. Expression levels of the other three genes varied, but in a highly correlated fashion (Figure 4, Online Supplementary Figure S4). A positive correlation between AhR and MEF2B (both positive with ΔΔCt values ≥0.15) could be shown (Pearson correlation coefficient 0.34). Values for sensitivity (0.61), specificity (1.0) and accuracy (0.78) were determined. With a P-value according to the Fisher exact test of 0.0026, independence of their expression levels could be rejected with a significance of 0.05 (Online Supplementary Table S2). Likewise, independence of MEF2B and BCL6 expression levels (both positive with ΔΔCt values ≥0.15) could be rejected with a P-value according to the Fisher exact test of 0.0001 against a level of significance of 0.05, together with a Pearson correlation coefficient of 0.69, a sensitivity of 0.92, a specificity of 0.9, and an accuracy of 0.91 (Online Supplementary Table S2).

Figure 4.Expression of AhR, ARNT, MEF2B and BCL6 in DLBCL cell lines. Transcript levels were assessed by quantitative real-time polymerase chain reaction and normalized against U-2932 subclone R1. TBP was used as the endogenous control. The mutational status of MEF2B shown had been assessed by Pasqualucci⁴ (red) or by us (black). We sequenced MEF2B from AA1 to AA141 including the mutational hot spot only; w, wild-type; m, missense mutation. BCL6 translocations were analyzed by three-color FISH using straddling (211G3) and flanking centromeric/telomeric (88P6/732M13) RP11-library BAC clones; w, wildtype, r; rearrangements involving the BCL6 site. Error bars indicate the standard deviation (n=3). The positive correlations between AhR-MEF2B (P=0.0026) and MEF2B-BCL6 (P=0.0001) according to the Fisher exact test are shown.

Until now, it has only been reported that mutant MEF2B enhances transcription of target genes, thereby contributing to the genesis of BCL6-positive DLBCL.28 Mutations in MEF2B are frequent in DLBCL and in follicular lymphoma.3029244 MEF2B point mutations are carried by 11% of GC and ABC DLBC lymphomas.3 However, unmutated MEF2B triggered BCL6 expression in U-2932 cells.28 Results of sequencing analyses showed that BCL6 levels in cell lines with MEF2B mutations were not generally high (Figure 4). We also found that cytogenetic translocations involving BCL6 were uncorrelated with BCL6 expression (Figure 4). Where BCL6 translocations were present in DLBCL cell lines, none involved canonical immunoglobulin gene partners reported in patients with BCL6 overexpression. Whole exome sequencing showed that the U-2932 subclones did not carry missense mutations in the open reading frame of IRF4 or STAT5, negative regulators of BCL6 (data not shown).3231 Exon 1 of BCL6 – important as a binding region for repressive BCL6 and STAT5 – was also unmutated (data not shown).3433 Thus, our results suggested that the AhR/ARNT-induced expression of wildtype MEF2B might be an independent regulator of BCL6 expression, besides canonical BCL6 translocations, BCL6 promoter hypermutation and MEF2B mutations.

BCL6: regulator of gene expression

More than 200 genes were differentially expressed in the two subclones of the U-2932 cell line (Online Supplementary Table S3). Having shown that copy number aberrations and AhR/ARNT pathway activation were responsible for a substantial proportion of these differences, for most genes the underlying molecular cause for subclone-specific expression remained to be elucidated. To determine the extent to which activation of the transcriptional regulator BCL6 in the R1 clone (Figure 2) contributed to this phenomenon, we ectopically expressed this oncogene in the BCL6-negative subclone R2. As previously reported, we infected R2 cells with a retroviral BCL6 construct (MSCV-BCL6-IRES-GFP)or with empty vector (MSCV-IRES-GFP).12 Treated cells were single-cell sorted to isolate clones with defined levels of BCL6. According to fluorescence microscopy, all MSCVBCL6-IRES-GFP clones carried the construct. However, only a minority (5/25) expressed BCL6 mRNA, of which only two expressed BCL6 protein (Online Supplementary Figure S5).

Ectopic expression of BCL6 in subclone R2 affected 48 of the 221 (22%) genes that were differentially expressed 10-fold or more (Figure 5A). Interestingly, 28 of these 48 genes were up-regulated by BCL6, although BCL6 is believed to act as a transcriptional repressor (Figure 5A,B).35 This also held true at the protein level. Under the influence of BCL6, the B-cell marker CD24 was repressed and, at the same time, CD20, CD22, CD27, CD38 and CD59 were significantly induced (Figure 5C). Confirming that BCL6 positively regulates the expression of target genes, the majority of the BCL6-stimulated genes of Figure 5 were positively correlated with BCL6 in a series of B-cell non-Hodgkin lymphoma cell lines (Online Supplementary Table S4).

Figure 5. BCL6-regulated genes. Ectopic expression of BCL6 in the U-2932 subclone R2 cells modulates diverse target genes. (A) Heat map according to expression array analysis. (B) Quantitative realtime polymerase chain reaction of selected BCL6 target genes. Error bars indicate the standard deviation (n=3). Benjamini-Hochberg corrected t-test P-values: P<0.05 (*), <0.01 (**), <0.001 (***). (C) Cell surface marker expression assessed by flow cytometry. Red: cells gated with forward scatter/side scatter (FSC/SSC) and singlet gates; green: FSC/SSC, singlet and GFP gates. Numbers in the boxes represent mean fluorescence intensity. Before flow cytometric analysis, the retrovirally transfected clone R2/MSCV BCL6 cl. 24 had been subjected to a second-round of single-cell cloning (R2/MSCV BCL6 cl. 24/13). Not all of the CD antigens tested belong to the group of 221 genes with >10-fold expression differences between R1 and R2.

Thus, our data demonstrated that differences in BCL6 expression resulted not only in down-regulation but also in up-regulation of a panel of genes, begging the question of how a transcriptional repressor might induce gene expression.

LMO2: a mediator for BCL6-initiated gene induction?

To assess whether the BCL6-triggered induction of gene expression was brought about by inhibition of a transcriptional repressor we analyzed which transcription factor binding sites (TFBS) were over-represented in BCL6-regulated genes. Putative TFBS were analyzed in the promoter region (±1000 bp relative to the transcriptional start site) of 160 genes equally up- and down-regulated by BCL6 (by at least a factor of 2). Confounding our original prediction, BCL6-up-regulated genes did not show enrichment for cognate repressors. It was notable that 68% of the BCL6-up-regulated genes contained binding sites for LMO2 complexes which were totally absent near down-regulated genes.

LMO2 forms complexes with distinct sets of partners in different types of cell. In contrast to hematopoietic stem cells and erythroid cells, the LMO2 complex in DLBCL cells excludes TAL1 and GATA proteins.36 Accordingly, neither TAL1 nor GATA1 nor GATA2 was expressed in the U-2932 subclones (Figure 6A). In contrast, other common LMO2 partners, such as E2A (TCF3) and SP1, but also DLBCL-specific partners such as ELK1, LEF1 and NFATc1 were expressed (Figure 6A). LMO2 was the only transcription factor complex gene expressed at different levels in the two subclones, suggesting that its presence might be crucial for the activity of the whole complex (Figure 6A). The expression of LMO2 in U-2932 cells was under the control of BCL6: cells expressing BCL6 were LMO2-positive, cells without BCL6 were LMO2-negative (Figures 5B and 6B). This and the fact that 68% of the BCL6-up-regulated genes contained TFBS for LMO2 complexes highlighted LMO2 as a plausible mediator for the inductive effects of BCL6. The observation that BCL6 stimulated expression of LMO2 was noteworthy in another context. On the basis of expression array data, U-2932 had been categorized as an ABC DLBCL cell line.11 However, LMO2 is one of the genes whose expression is indicative of the GC type of DLBCL.371 Applying a short list of five ABC marker genes and five GC markers, the U-2932 subclone R1 differed from R2 by elevated expression of four out of five GC markers (Figure 6C).37 Besides LMO2, BCL6, MYBL1 and NEK6 are also GC markers highly expressed in subclone R1 (Figure 6C). Ectopic expression of one of these genes (BCL6) led to the induction of two other GC markers (LMO2 and MYBL1) (Figure 5B).

Figure 6.Subclone-specific expression of BCL6 and LMO2. (A) Gene expression heat map of LMO2 complex genes. (B) Effect of BCL6 on LMO2 protein expression (C) Heat map of ABC and GC genes in U 2932 R1 and R2.

Discussion

Tumors often show heterogeneity at the chromosomal and genetic levels explaining why analysis of the clonal architecture of tumors is such an important topic in cancer research. A disadvantage of studies with primary tumor cells is that they rarely allow direct conclusions to be drawn on the cellular consequences of a specific mutation. Functional studies are, therefore, often conducted with cells that ectopically express the mutated gene, e.g. cytokine-dependent cell lines or knock-in mice. However, in most cases these systems do not truly represent the histological origin of the studied tumor. Immortalized cell lines have often proven suitable for elucidating the function of a mutated gene. However, originating from a single cell, a cell line usually does not represent more than one clone. Exceptionally, the DLBCL cell line U-2932 consists of two populations representing two tumor clones, as evidenced by immunoglobulin hypermutation analysis.12 The two U-2932 populations overexpress common (BCL2) and subclone-specific oncogenes (BCL6 versus MYC).12

Genomic amplification (BCL2; 6–7× in both clones) and a subclone-restricted translocation [t(8;14) affecting MYC in R2] explain overexpression of two key oncogenes. Here, we set out to determine why BCL6 was exclusively expressed in subclone R1. BCL6 was not subject to any of the canonical BCL6 rearrangements in U-2932. The BCL6 repressors STAT5 and IRF4 were unmutated, as was BCL6 exon 1 – important for negative autoregulation of BCL6. The transcriptional activator MEF2B, mutated in 11% of DLBCL, was also unmutated. However, the wild-type form of MEF2B can also regulate BCL6.28 MEF2B and BCL6 levels were highly correlated in the U-2932 subclones and in a large panel of DLBCL cell lines. Our inhibitor and knock-down experiments demonstrated that MEF2B is a downstream target of the AhR/ARNT pathway. Consequently AhR and MEF2B levels were highly correlated in a series of DLBCL cell lines.

AhR has been described as an epigenetically regulated gene.2322 Accordingly, inhibitors of histone deacetylation and DNA methylation triggered expression of this gene in AhR-negative cell lines. Thus, results with the U-2932 subclones and other DLBCL cell lines indicate that AhR-regulated high expression of wild-type MEF2B might be an additional, independent cause of the aberrant expression of BCL6 in DLBCL besides canonical BCL6 translocations, BCL6 hypermutation and MEF2B mutations.

That aberrant activation of signaling cascades can trigger the expression of a characteristic set of target genes is a well-described phenomenon. Inhibitor experiments showed that AhR/ARNT activation in subclone R1 was responsible for the expression of BCL6 and ten additional genes. However, gross (>10×), subclone-specific expression differences were observed in 221 genes, not just 11. Genomic copy number aberrations correlated with and explained subclone specific expression differences for 58 genes. BCL6 itself regulated 48 subclone-characterizing genes. Interestingly, 28/48 genes were induced when BCL6 was ectopically expressed, suggesting a novel - most likely indirect - transcriptional role rather than the well-described suppressive role of BCL6. It is also a novel finding that the GC marker BCL6 triggered expression of other GC markers, MYBL1 and LMO2. The transcriptional BCL6 target LMO2 forms part of a transcriptional complex with cognate TFBS in 68% of the BCL6-up-regulated genes, which are absent in BCL6-repressed genes, suggesting that LMO2 plays a role in the observed stimulatory effects of BCL6.

In summary, genomic copy number aberrations, activation of the AhR/ARNT complex, and overexpression of BCL6 act together to regulate over 100 genes in subclones of the DLBCL cell line U-2932. The levels of AhR, MEF2B and BCL6 were strongly correlated in a panel of DLBCL cell lines. Knock-down and inhibitor experiments confirmed that AhR/ARNT regulates MEF2B, a key transcription factor for the regulation of BCL6. Another new finding is that BCL6 not only acts as a repressor, but is also capable of inducing expression of genes including the GC markers LMO2 and MYBL1. Thus, BCL6 – regulated by AhR/ARNT and wild-type MEF2B – may control expression of GC markers in DLBCL (Figure 7). Collectively our findings document an in vitro model for genetic heterogeneity in DLBCL to serve as a novel tool for investigating the pathology and therapeutics of a clonally diverse cancer.

Figure 7. BCL6 – regulated by AhR/ARNT and wild-type MEF2B – controls target genes including SOCS2 and LMO2. Epigenetically regulated AhR controls expression of MEF2B, a transcription factor for BCL6. BCL6 represses (e.g. SOCS2) and induces (e.g. LMO2) target genes. LMO2 is involved in the positive regulatory process.

Acknowledgments

The authors would like to thank Wilhelm Sander-Stiftung for financial support.

Footnotes

• The online version of this article has a Supplementary Appendix.
• Authorship and Disclosures Information on authorship, contributions, and financial & other disclosures was provided by the authors and is available with the online version of this article at www.haematologica.org.

• Received November 13, 2014.
• Accepted March 4, 2015.