Distinctive patterns of BCL6 molecular alterations and their functional consequences in different subgroups of diffuse large B-cell lymphoma

Patient population

We studied previously 240 cases of DLBCL profiled with gene expression by complementary DNA (cDNA) microarray technology.⁴ A panel of hematopathologists confirmed the diagnosis of DLBCL in all patients. The cases were classified into different molecularly defined subgroups using the Bayesian method.²⁴ Of these 240 cases, 92 cases were classified as GCB-DLBCL, 82 cases ABC-DLBCL, 20 cases PM-DLBCL and 46 cases unclassifiable DLBCL. All patients included in this series had received an anthracycline-containing chemotherapy regimen. The Institutional Review Board of the University of Nebraska Medical center approved this study.

Immunohistochemical staining of tissue sections

Immunohistochemical (IHC) staining for CD20 and BCL6 was performed on 138 cases with available archival paraffin blocks as described previously.²⁵ A prior comparative study of IHC staining in three different laboratories validated the reproducibility of staining and evaluation.²⁶ Briefly, tissue microarrays (TMAs) were prepared from cases with adequate archival paraffin-embedded tissue. Sections of 5 µm were cut from each TMA and stained with an antibody to BCL6 (Polyclonal, Santa Cruz Biotechnology, Santa Cruz, CA, USA), as described previously.²⁵ CD20 stains were performed to evaluate each core for involvement by tumor. The percentage of tumor cells stained with each antibody was recorded in 10% increments and for each case, the core with the highest percentage of tumor cells stained was used for analysis. Cases were considered positive if 30% or more of the tumor cells stained for BCL6.²⁵

Detection of the 3q27 translocation by fluorescence in situ hybridization

Interphase fluorescence in situ hybridization (FISH) analysis for chromosome 3q27 (BCL6) translocations was performed on 133 cases using formalin-fixed, paraffin-embedded tissue sections, as described previously with minor modification.⁸ Briefly, a BCL6 break-apart probe (BAP; Abbott-Vysis, Downers Grove, IL, USA) was used to detect BCL6 translocations at major breakpoint region (MBR) and a BCL6 home-brew BAP (Clone RP11-1144D2 and RP11-67E18) at the alternative breakpoint region (ABR) (78/133 cases).^27,28 Nuclei were counterstained with 4,6-diamidino-2-phenylindole (DAPI) in Antifade solution and the slides were visualized using an Olympus BX51 fluorescence microscope. Images were captured and archived using Cyto Vision software (Applied Imaging, Santa Clara, CA, USA). To analyze the hybridization, a total of 50–100 nuclei per case were scored for the presence of the BCL6 translocation. The normal cutoff for this FISH assay in paraffin tissue sections has been established to be 20% by prior studies. The gain/amplification data on the 3q27 region of these 133 cases was obtained from a previous study,² using comparative genomic hybridization (CGH) and compared with BCL6 mRNA and protein expression. The FISH data for t(14;18) were obtained from our previous study.⁸ To examine the correlation between BCL6 translocation and p53 genomic abnormalities, 66 cases of the cohort with p53 gene mutation as well as deletion data were obtained from a previous study for analysis.²⁹

Mutational analysis of BCL6 intron-1 and exon-1

Genomic DNA extracted from 128 frozen tissue samples used for GEP was available for this analysis. The major mutation cluster (MMC) in BCL6 intron-1 (~ 700 bp) was divided into three overlapping fragments for PCR amplification (accession no.: AC072022.19; 37735–38012 (F1); 37975–38220(F2); 38200–38391(F3)) using the following PCR primer sets (F1′/F1″: 5′-(TTTTCCGCTCTTGCCAAATGCTTT/5′-GGAGGGGAATTAGGGG; F2′/F2″: 5′-GGTTTTTGGAAAGGAGGT/5′-CTGCTTTCCTTGCTCCGTC and F3′/F3″: 5′-TGCGGCTGTGTTTTTT/5′-ATTCCCTGCCTTCGAGCCG, respectively), see Figure 3 also. Mutational analysis of exon-1 (E1; 37375–37636) was performed using a previously described primer set (E1/E1″: 5′-ACGCTCTGCTTATGAGGA/5′-CGGCAGCAACAGCAATAA, respectively).²¹ The PCR mixtures contained 500 ng of genomic DNA, 200 µm dNTP, 1.5 mm MgCl₂, 50 µm primer and 1.5 U Amplitaq polymerase in a reaction volume of 100 µl. Forty-five cycles of PCR were performed after an initial denaturation (94°C, 9 min) followed by denaturation at 94°C for 75 s., annealing at 60–68°C for 75 s (60°C for E1, 68°C for F1, 60°C for F2 and 64°C for F3), extension at 72°C for 40 s, and final extension at 72°C for 7 min. Mutants were then identified using denaturing high-performance liquid chromatography in a WAVE DNA Fragment Analysis System (Transgenomic, Omaha, NE, USA).³⁰ Mutant fractions were collected on the Wave system on DNA samples that exhibited shifted peaks compared with wild-type DNA. These fractions were reamplified and then sequenced. The detailed analysis of the mutations was performed with the Vector NTI software program (Invitrogen Inc., CA, USA), and compared with germ line BCL6.

Analysis of BCL6 and target gene expression

We evaluated BCL6 mRNA levels measured by the Lymphochip microarray in all DLBCL cases from our previous study.⁴ Expression levels were log-transformed and mean-centered across samples. The value was expressed as the number of fold changes over the reference pool mRNA.⁵ The median of hybridization values from four BCL6 cDNA clones, immobilized on the Lymphochip, was used as BCL6 mRNA expression values. We also used BCL6 transcript level, BCL6 protein expression, and BCL6 translocation and mutation data to supervise the analysis of the expression of known BCL6 targets^31–34 (see Supplementary Reference list for known target genes) present on the Lymphochip for each subgroup of DLBCL. When the GCB cases were divided into quartiles according to their BCL6 mRNA expression levels, we observed repression of most of these targets and there was a general trend toward increased suppression with higher transcript levels. We selected a subgroup of these genes that showed a consistent pattern of repression and considered this subgroup as the most reliable BCL6 targets in the GCB environment. This set of target genes was then used to assess BCL6 activity in other subgroups. For target genes represented by three or more cDNA clones, the median hybridization values were used for analysis. If two clones were present, the mean value was used. The three clones of nuclear factor-κB1 (NF-κB1) on the Lymphochip showed inconsistent expression levels and, therefore we examined the cases that were also profiled with Affymetrix chips. The clone on the Lymphochip (UniqID: 31267), which showed maximum correlation (r=0.75) with expression levels measured with Affymetrix chips was selected for analysis.

Statistical analysis

The Kaplan–Meier method was used to estimate OS of the patients, and the log-rank test was used to compare the survival distributions between subgroups with and without BCL6 mutation/translocation. OS was defined as the time from diagnosis to death resulting from any cause or, for patients remaining alive, the time from diagnosis to last contact. Event-free survival was defined as the time from diagnosis to the first occurrence of relapse or death from any cause or, for patients remaining alive and relapse-free, the time from diagnosis to last contact. Fisher's exact test was used to analyze categorical data and the Wilcoxon's rank-sum test was used to analyze continuous data between groups. P-values for the Kaplan–Meier curves were not adjusted for subgroup analysis. For other comparisons, if the overall P-value comparing subgroups and BCL6 variable was significant, then pair-wise comparisons between the subgroups were performed. These post hoc tests were adjusted for multiple comparisons using the Bonferroni method. SAS software was used for the data analysis (SAS Institute Inc., Cary, NC, USA).

Case characteristics

The 240 DLBCL cases profiled with the Lymphochip⁴ were divided into GCB (92 cases; 38.7%), ABC (82 cases; 34.2%), PM (20 cases; 7.9%) and unclassifiable (46 cases; 19.2%) subgroups. BCL6 mRNA expression data were available in all of the above cases. Of the 240 cases, BCL6 IHC data were available in 138 cases (57 GCB, 44 ABC, 10 PM and 27 unclassifiable). In the same series of patients, FISH data on BCL6 MBR translocation and gain/amplification were obtained in 133 cases (51 GCB, 46 ABC, 12 PM and 24 unclassifiable) and ABR translocation on 78 cases. BCL6 mutational data was obtained in 128 cases (50 GCB, 43 ABC, 12 PM and 23 unclassifiable). All datasets were available in 106 cases. Clinical data were available on 236 cases (91 GCB, 82 ABC, 19 PM and 44 unclassifiable). No significant differences were observed for any of the clinical parameters in the subgroups with IHC data or FISH data when compared with the entire group of DLBCL.

Prevalence of BCL6 translocations in DLBCL subgroups

The BCL6 gene translocation at the MBR was analyzed in 133 cases by FISH and observed in 18.8% (25/133) of DLBCL cases. The incidence of translocation was higher in the ABC (24%) and PM (33%) subgroups, as compared with the GCB subgroup (10%) (Table 1). We also compared the t(14;18) data obtained previously on these cases⁸ and observed that only two of 23 BCL6 translocated cases were also positive for the t(14;18), one a GCB- and the other a PM-DLBCL. There does not seem to be any association between these two translocations (Supplementary Table ST1).

We also evaluated BCL6 rearrangement at ABR in 78 informative DLBCL cases. Of these, five cases (6.4%) were positive using a BAP for this region. Of the five positive cases, three cases were in the ABC and one case each in the GCB and unclassifiable subgroup. As we expected that BCL6 translocation at MBR would be associated with deregulated BCL6 expression, we compared BCL6 mRNA and protein expression in cases with and without BCL6 translocation. When all the DLBCL cases were analyzed as a group, we observed no significant difference in either BCL6 mRNA expression (P=0.58) or protein (P=0.56) in cases with or without translocation. When the subgroups were analyzed separately, the only significant difference was observed in the ABC subgroup (Figure 1), where BCL6 mRNA expression was increased by 1.5-fold in the translocated cases (P=0.045). Similar results were obtained when the cases with translocation at the ABR were added to the analysis. Actually, cases (n=5) with translocation at the ABR were associated with low BCL6 protein (P<0.025, χ² test) and mRNA expression (threefold difference, P=0.05) compared with cases without translocation.

Of the 133 cases studied for BCL6 translocation, 66 cases were analyzed for p53 mutation and deletion²⁹ (Supplementary Table ST2). There was no significant difference (P=0.80 by the χ² test) in the frequency of p53 abnormalities in cases with and without BCL6 translocation. Similar findings were obtained when cases with translocation at ABR were added (P=0.2 by χ² test).

Among the 25 BCL6 translocated cases, data on IgH BAP were available in 16 cases, of which six were negative and therefore had a non-IgH translocation partner. Of the remaining 10 cases positive for IgH BAP, FISH using IgH/BCL6 fusion probes (Cancer Genetics Inc., River Vale, NJ, USA) were performed and seven cases had IgH/BCL6 translocation, two cases had variant BCL6 translocation and one was inconclusive. Of the seven cases with IgH/BCL6 translocation, four cases were in the ABC, two in the GCB and one in the unclassifiable subgroups and all the cases with variant translocations were in the ABC subgroup. Therefore, the incidence of IgH/BCL6 vs non-IgH/BCL6 translocation in DLBCL was almost same the (8:7). No differences in BCL6 mRNA or protein expression were observed between cases with or without t(3;14).

BCL6 gene mutations in DLBCL subgroups and correlation with mRNA and protein expression

We defined the mutational spectrum in our cases, including the frequently mutated 5′ region in intron-1 (MMC),²² and a newly defined region containing the negative regulatory elements for BCL6 in exon-1.²¹ Of the 128 analyzed cases, intron-1 mutations were detected in 78 cases (61%), with higher frequencies in the GCB (74%) and PM (75%) subgroups than in the ABC subgroup (44%) (Table 2). There were 302 different mutations consisting mostly of single-base substitutions (n=251), followed by single-base insertions (n=38) and single-base deletions (n=13). The number of mutations per case ranged from 1 to 24 with an average of 3.9 mutations per case (Table 2). G was the most frequently mutated nucleotide, and transversions were more frequent than transitions (140 vs 110) (see Supplementary Figure 1). The average number of mutations was highest in PM-DLBCL subgroup (Table 2). Cases of DLBCL with mutations had higher BCL6 mRNA expression than cases with wild-type BCL6 (P=0.0013); but this was not statistically significant in individual DLBCL subgroups (Figure 2). The GCB subgroup had high BCL6 protein and there was a trend toward higher BCL6 protein level in mutated cases (Figure 2).

Previous studies have reported clustering of mutations in the MMC^35,36 and postulated the presence of two³⁷ or three³⁸ major clusters. We also observed a similar clustering pattern and defined three MMCs that had a mutational frequency ≥3-fold higher than the entire MMC region (1–1.5 × 10⁻²/bp vs 5.4 × 10⁻³/bp). These clusters were located at positions 80–123, 277–316 and 470–513 in the MMC region (Figure 3). The first mutation cluster (80–123) has been reported by others.^37,38 In subgroup analysis, the presence of two major clusters was observed in the GCB subgroup (277–316 and 470–513), but clustering was not discernable in the other subgroups probably because of the lower number of cases with mutations. In agreement with previous studies, three single-nucleotide polymorphisms (SNPs) in the MMC were observed at positions 397 (C → G; 20%), 502 (A → G; 10%) and 520 (del T; 10%). The SNPs at 397 and 520 were observed with similar frequencies in the GCB (6 of 14) and ABC (4 of 14) subgroups; however, the SNP at 502 was observed more frequently in the GCB subgroup (6 of 8) than in the other subgroups (PM and ABC with one case each). Exon-1 mutations were detected in 10% (13 of 128) of the cases and the majority of these cases were in the GCB subgroup (9 of 13). These mutations were generally clustered in the DNAse-I hypersensitive region, particularly within the BCL6, STAT1 and predicted CEBP binding sites. In general, mutations in the BCL6-binding sites were associated with increased BCL6 mRNA and protein expression (see Supplementary Figure 2).

Relationship between BCL6 translocation and 3q27gains with BCL6 mRNA and protein expression

Using a 30% cutoff for positive cellular staining, the expression of BCL6 protein was detected in 78 of 138 DLBCL cases (55%) with the highest expression in the GCB subgroup (Table 3). In the ABC subgroup, there was a statistically significant association between BCL6 mRNA and protein expression (P=0.0025), but not in other subgroups (Figure 4). BCL6 mRNA and protein expression was not significantly associated with BCL6 translocation status at the MBR, but significantly lower expression was observed in cases with ABR (mRNA: P=0.05 and protein P=0.025).

We also correlated BCL6 protein and mRNA expression with 3q27 gains obtained from CGH performed separately.² Of the 123 DLBCL cases evaluated, 3q27 gain was observed in 22 cases (17.8 %) cases and the majority of 3q27 gains were observed in the ABC subgroup (17/42; 40% vs 3/45; 6% in GCB; P<0.001); however, these gains were not associated with increased BCL6 mRNA or protein expression.

Association of BCL6 gene abnormalities and BCL6 target gene repression

We used BCL6 mRNA and protein expression, and mutation and translocation data, to perform supervised analyses of the expression of BCL6 target genes^31,39,40 in the different subgroups of DLBCL. Target gene selection was described in the Materials and methods section and repression of the majority of the target genes was observed in all three subsets when the level of BCL6 expression was sufficiently high at both protein or mRNA level. For example, repression of the target genes was observed in the GCB subgroup from the second to fourth quartile of BCL6 transcript levels, but in the ABC and PM subgroup significant repression was observed only in the fourth quartile (Figure 5a). However, the pattern of repression differed among the subsets and some of the target genes, for example, CD80, STAT3, NFκB1 and BCL-xl, were not consistently repressed in the ABC or PM subgroups. P53 expression levels did not show any significant inverse correlation with BCL6 expression in any subgroup (Figures 5a and b).

We did not observe any significant correlation between the mutational status in intron-1 and the transcriptional levels of target genes; however, in the GCB subgroup, the majority of target genes showed lower expression in cases carrying exon-1 mutations (see Supplementary Figure 3).

There was no evidence of repression of known target genes in cases with BCL6 translocation in any of the subgroups of DLBCL. We also examined the expression of eight known translocation partner genes (IL-21R, MBNL, Pim-1 kinase, JAW1, CD71, eIFF4G1, CIITA-8 and MBNL)^41,42 in BCL6 translocated cases. Cases with non-IgH partners showed higher (≥1.5-fold) expression of five (out of eight) known translocation partners genes (IL-21R, MBNL, Pim-1 kinase, JAW1, CD71 and eIFF4G1) compared with cases with IgH/BCL6 translocation (Figure 6).

Clinical characteristics and OS

The major clinical characteristics of patients with or without BCL6 mutations or translocations and with different levels of BCL6 mRNA and protein expression are given in the supplementary tables (see Supplementary Tables ST2–5). Patients with BCL6 mutations were more likely to be less than 60 years old than the wild-type cases (P=0.01), but were similar in the other clinical characteristics examined. GCB-DLBCL cases with lower BCL6 protein expression were more likely to be males (P=0.005). There were no other major clinical differences.

BCL6 protein expression (≥30%) was predictive of favorable OS (P=0.001) in the entire group of DLBCL (Figure 7a), and this prediction was also observed with other cutoffs (≥10%, P=0.004; ≥50%, P<0.001). Higher mRNA levels were also associated with a more favorable outcome (P<0.001) when the cases were divided into quartiles or halves according to their mRNA expression levels (Figure 7b). In subgroup analysis, high levels of BCL6 protein expression were associated with better OS in the ABC subgroup (BCL6 ≥50%; P=0.026). BCL6 protein expression was also a significant predictor of survival in the low International Prognostic Index (IPI) group of patients (P=0.001). This association was also observed for mRNA expression (P=0.009 for quartiles and 0.003 for halves) (Figure 7 c–d). In the high IPI risk group, high BCL6 protein expression was significantly associated with a favorable outcome (P=0.005); however, this association was marginally significant in BCL6 mRNA expression (P=0.06) (see Supplementary Figure 4). BCL6 translocation showed no association with OS in DLBCL as a single entity or in subgroup analysis.

The presence of BCL6 mutations was a significant predictor of favorable OS (P<0.001) in DLBCL and in the ABC subgroup (P=0.04), but was of marginal significance in the GCB subgroup (P=0.08). The association of mutational status with favorable outcome was also observed in both the low IPI (P=0.02) and high IPI risk groups (P=0.008) (see Supplementary Figure 5).