Investigating epigenetic effects of activation-induced deaminase in chronic lymphocytic leukemia

Activation induced deaminase (AID) has two distinct and well defined roles, both relying on its deoxycytidine (dC) deaminating function: one as a DNA mutator and another in DNA demethylation. In chronic lymphocytic leukemia (CLL), AID was previously shown to be an independent negative prognostic factor. While there is substantial impact on DNA mutations, effects of AID on gene expression by promoter demethylation of disease related target genes in leukemia has not been addressed. To shed light on this question, we aimed at determining genome wide methylation changes as well as gene expression changes in response to AID expression in CLL. Although we found minor differences in individual methylation variable positions following AID expression, we could not find recurrent methylation changes of specific target sites or changes in global methylation.


Introduction
Activation induced deaminase (AID, encoded by the AICDA gene) was identified as a key enzyme responsible for somatic hypermutation (SHM) and class switch recombination (CSR) of antibody genes in germinal center B cells [1]. These on-target activities of AID are achieved by deaminating cytosines to uracils at the genomic antibody locus. This initiates an error prone repair process resulting in localized hypermutation in the antigen binding variable region and altered antibody affinity during SHM. During CSR, altered antibody effector function is accomplished through generation of double strand breaks at antibody switch regions [2].
Aside of these on-target activities, AID was also shown to perform substantial off-target effects by mediating genome wide mutations (off-target hypermutation) and translocations (off-target CSR) [3]. These off-target effects significantly contribute to lymphomagenesis as well as to clonal evolution and treatment resistance [4,5]. Apart from these well described mutagenic effects, AID was also shown to contribute to DNA demethylation by deaminating methylated cytosines, thereby generating thymines, which are eventually replaced by PLOS  unmethylated dCs by the DNA mismatch repair machinery independent of proliferation or DNA replication [6]. These AID specific epigenetic effects were initially described during reprogramming of germ cells and induced pluripotent stem cells, but were also described for breast cancer, where AID was shown to regulate expression of specific genes important for epithelial-mesenchymal transition [7,8]. More recently, methylation dynamics during germinal center formation was also attributed to AID activity [9]. In chronic lymphocytic leukemia (CLL), a chronic B cell malignancy of the elderly, AID transcripts in leukemic cells from peripheral blood are detectable in about half of the patients [10]. Patients with AID-expressing cells have a markedly shorter time to treatment, a worse clinical prognosis and more adverse cytogenetic aberrations [11,12]. Surprisingly, AID expression also correlates with expression of a non-hypermutated B cell receptor (BCR) and is associated with active CSR and an increased proliferative and antiapoptotic potential in a subpopulation of leukemic cells, whereas CLL samples carrying a hypermutated BCR frequently lack AID transcripts [12,13]. However, although there is an inverse correlation between AID expression and mutated BCRs in CLL, BCR mutation status and AID expression remain both independent parameters for shorter time to treatment in multivariate analyses [12]. CLL samples were also shown to exhibit diverse methylation patterns with high interpatient heterogeneity and distinct methylation dynamics during disease progression [14,15]. While substantial impact of AID on mutations in CLL was suggested [16], up to now, there are no studies addressing a possible involvement of AID in particular epigenetic changes in CLL.

Patient samples and plasmids
Peripheral blood mononuclear cells from CLL patients (S1 Table) were obtained upon informed consent and ethical approval by the Ethics Committee of the Province of Salzburg (415-E/1287/4-2011, 415-E/1287/8-2011) by Ficoll density gradient centrifugation. The determination of prognostic markers and FISH analysis for trisomy 12, del11q, del13q, and del17p was performed routinely at our department as described previously [17]. Plamids pGFP and pAID-GFP were constructed as previously described [18].

Bioinformatics
Data of the Infinium MethylationEPIC BeadChip (Illumina) were analyzed with the Chip Analysis Methylation Pipeline (ChAMP), adapted for paired analyses using R/Bioconductor, including beta-mixture quantile normalization (BMIQ), linear models and empirical bayes methods (limma) for significant methylation variable positions (MVP) (p<0.05 or p<0.001 as stated) [19]. These MVPs were further grouped into differentially methylated regions (DMRs) defined as regions with 3 MVP in 1000 bp with p<0.05. Adjusted p-values were calculated based on the false discovery rate (FDR) according to the Benjamini-Hochberg method.

TaqMan gene expression assays (384-well microfluidic cards)
For Samples were analyzed on a ViiA 7 Real-Time PCR System (Thermo Fisher Scientific) as follows: 95˚C for 10 min, 50 cycles with 95˚C for 15 sec, 60˚C for 1 min. For transfected samples the transcript levels were normalized to GAPDH and the relative fold change was calculated with the 2 -ΔΔCT method using the pGFP transfected condition as the calibrator.
For untransfected samples the transcript levels were normalized to GAPDH and the relative expression was calculated with the 2 -ΔCT method. Samples with "Undetermined" Cm values for AICDA were considered negative for AID expression and grouped accordingly. Target genes below the detection limit with "Undetermined" Cm values were artificially set to 10− 11 (n.d. = not detected).

Bacterial mutator assay
AID-GFP fusion gene or GFP of our CLL transfection vectors were PCR-amplified and cloned into a pTrc99 backbone. pTrc99-AID was used as a positive control. Bacterial assays were performed as previously described [20]. The Uracil-DNA-Glycosylase (UNG)-deficient E.coli strain BW310 was transformed with pTrc99-AID, pTrc99-AID-GFP or pTrc99-GFP. Upon AID-induced mutations of the Rifampicin (Rif)-sensitive allele in BW310 bacteria, Rif-resistant clones appear in presence of Rif. The mutational frequency was determined by plating cultures of individual clones grown over night at 30˚C in rich medium supplemented with Carbenicillin (CB) (100 μg/ml) and IPTG (1mM) on LB agar supplemented with CB (100 μg/ ml) or CB (100 μg/ml) and Rif (100 μg/ml). AID activity was analyzed by calculating the mutational frequency as ratio of Rif-resistant clones to total clones in the starting culture.

Results
To draw light on the question of involvement of AID in epigenetic changes in CLL, we performed methylation array analyses using Illumina Infinium Methylation EPIC BeadChips to assess methylation changes in primary AID non-expressing CLL samples upon transfection with AID encoding constructs. Therefore, we first selected four AID non-expressing CLL samples (#1, #2, #3, and #4) and transfected cells individually with constructs encoding a functional AID-GFP fusion protein or a GFP control (activity of the AID-GFP fusion protein is determined by bacterial mutator assays, S1 Fig). In order to show expression of the AID-GFP fusion protein, we performed western blot experiments on cell lysates from transfected unsorted CLL samples (S2 Fig). For methylation array analyses we sorted cells according to GFP fluorescence and CD5+CD19+ expression by flow cytometry (Fig 1A and 1B) and performed individual EPIC array analysis on DNA from AID-GFP and GFP transfected samples ( Fig 1C). No significant global methylation changes were detectable in GFP versus AID-GFP transfected samples, reflected in similar methylation values within the respective CpG regions between GFP and AID-GFP expressing samples (S2 Table). The analysis of individual methylation variable positions (MVPs) between paired AID-GFP and GFP samples revealed only minor differences, with methylation changes at most +/-20%. In total, we detected 34,868 MVPs with p<0.05 (16,135 hypermethylated and 18,733 hypomethylated MVPs) within or in proximity to 12,727 genes and 293 MVPs with p<0.001 (151 hypermethylated and 142 hypomethylated MVPs) with 184 related genes. Five of these MVPs mapped to the gene body of AICDA, most likely coming from the transfected AID encoding plasmid (S3 Table). However, none of these MVPs (except one for AICDA) were significantly deregulated using adjusted pvalues (false detection rates <0.05, S3 Table).
To determine whether the minor methylation changes detected upon AID expression lead to an altered gene expression, we first ranked the most likely candidate genes with recurrently altered CpG methylation. Therefore, we grouped MVPs into differentially methylated regions (DMRs), defined as regions with at least 3 MVPs within a region of 1kb with p<0.05, resulting in a list of 125 DMRs containing 396 MVPs (S4 Table). Upon exclusion of pseudogenes and inclusion of DMRs harboring at least 2 hypomethylated MVPs located within 1500 bp upstream of the transcription start site (TSS), the gene body or 5 0 -UTR/3 0 -UTRs, we ended up with a list of 45 target genes (Fig 1D). For each patient, differences in DNA methylation upon AID expression (delta beta value) of the 145 MVPs grouped into 45 DMRs are shown in Fig  1D. To assess whether any of these 45 genes exhibit differential gene expression dependent on AID, we transfected 6 AID non-expressing CLL samples with constructs encoding AID-GFP or GFP and extracted total RNA from GFP positively sorted cells. Reversely transcribed complementary DNA of these samples were subjected to TaqMan 384-well microfluidic cards designed to detect specific transcription levels of our 45 candidate genes in addition to AICDA. In accordance with the overall low methylation changes, our results did not reveal a significant impact of AID on differential gene expression and only minor differences in transcription levels could be detected. Of note, only 26 of our set of 45 genes showed detectable expression. Differential expression of AICDA served as an internal control (Fig 2A). Using the Taqman microfluidic cards, we also analyzed a cohort of 32 unselected untreated CLL patients (patient information in S1 Table) for expression of AICDA and our set of candidate genes. As shown in Fig 2B, AICDA expression in 21 of 32 patients (65.6%) was below detection limit and again no differential expression of any of these genes could be noticed in dependence of AID expression.
To address differential gene expression in a larger cohort, we finally reanalyzed a published gene expression dataset (GSE39671) [21]. These microarray expression data from 130 CLL samples revealed bimodal gene expression of AICDA, dividing samples into 91 AICDA non-(or low)-expressing samples and 39 AICDA-expressing samples (S5 Table). In this large cohort, we found five of our 45 target genes to be differentially expressed (FDR<0.05; GABRB3, GFPT1, UNKL, GPT, DAPK1) in AICDA-high vs low groups, albeit at very low fold change (S5 Table, S3 Fig) and partially with inverse fold change compared to our results from the TaqMan 384-well microfluidic cards (Fig 2B). In addition, differential expression of these candidates could not be validated using another dataset (GSE22762) from CLL samples [22]. In this dataset, AICDA expression was not bimodal but rather exhibited a range from low to high (S6 Table) with no apparent correlation (R 2 range: 0,0001-0,055) with any of our candidate genes (S7 Table).

Conclusion
We conclude that unlike to data from breast cancer, where AID was recently shown to robustly induce gene expression of genes important for epithelial-mesenchymal transition [8], we could not find comparably convincing and robust AID dependent induction of target gene demethylation and concurrent gene expression in CLL. While we defined a small set of genes with AID-dependent methylation differences, we were unable to detect substantial influence on gene expression differences of these candidate genes, neither in our own cohort of CLL samples nor in data from published datasets [21,22]. Hence, we assume that possibly either AID does not have a particular set of specific target genes for demethylation in CLL or that there is a high interpatient target heterogeneity, implying that AID-if at all-rather unspecifically induces genome wide 'off-target' methylation changes. Alternatively, AID could induce methylation changes in CLL, which were not covered by the EPIC beadchip array or transcriptome analysis, which could affect genome instability rather than gene expression changes. Furthermore, the possibility that we missed single CpGs influencing gene expression, especially in promotor region, by grouping at least 3 MVPs within a region of 1kb to DMRs cannot be excluded.
Summarizing, our results suggest that the well described mutagenic effect and not targeted epigenetic activity of AID likely accounts for the observed worse prognosis of patients with AID positive CLL.   Table. Global methylation values in GFP and AID-GFP transfected samples using Illumina Infinium Methylation EPIC BeadChips. The methylation rate (0 = all demethylated; 1 = all methylated) for CpG sites is depicted with minimum, first quartile, median, mean, third quartile, maximum and sum values grouped according to CpG regions (ALL = all CpGs; TSS1500 = within 1500 bp upstream of the TSS; TSS200 = within 200 bp upstream of the TSS; 1stExon = within the first exon; 5UTR = within the 5 0 -UTR; Body = within the gene body; 3UTR = within the 3 0 -UTR; N Shelf = within 2-4 kb upstream of CpG islands; N Shore = within 0-2 kb upstream of CpG islands; CpG islands = within CpG islands; S Shore = within 0-2 kb downstream of CpG islands; S Shelf = within 2-4 kb downstream of CpG islands; Open Sea = isolated CpGs in the genome). (XLSX) S3