Genetic Variation Stimulated by Epigenetic Modification

Homologous recombination is essential for maintaining genomic integrity. A common repair mechanism, it uses a homologous or homeologous donor as a template for repair of a damaged target gene. Such repair must be regulated, both to identify appropriate donors for repair, and to avoid excess or inappropriate recombination. We show that modifications of donor chromatin structure can promote homology-directed repair. These experiments demonstrate that either the activator VP16 or the histone chaperone, HIRA, accelerated gene conversion approximately 10-fold when tethered within the donor array for Ig gene conversion in the chicken B cell line DT40. VP16 greatly increased levels of acetylated histones H3 and H4, while tethered HIRA did not affect histone acetylation, but caused an increase in local nucleosome density and levels of histone H3.3. Thus, epigenetic modification can stimulate genetic variation. The evidence that distinct activating modifications can promote similar functional outcomes suggests that a variety of chromatin changes may regulate homologous recombination, and that disregulation of epigenetic marks may have deleterious genetic consequences.


Introduction
Homologous recombination depends upon a DNA donor molecule to serve as a template for correction of a damaged recipient [1,2]. Homologous recombination is a critical pathway for error-free restoration of broken DNA, but it can also lead to mutagenesis and chromosomal rearrangements. templated by a homolog rather than a sister chromatid can lead to loss of heterozygosity associated with both cancer and aging; and homologous recombination between nonallelic repeated sequences can cause genomic instability and human genetic disease [3,4,5,6].
Chromatin status contributes to regulation of homologous recombination. Dynamic changes in chromatin structure occur at a site of DNA damage, and are important for maintaining broken DNA ends in close proximity [7,8,9,10,11]. In addition, we recently showed that local repressive modifications at donor chromatin can diminish homologous recombination [12]. This raised the possibility that activating modifications might stimulate recombination. We have now tested this, by asking if gene conversion can be promoted by local recruitment to the donors of factors associated with activation of chromatin. We assayed the effects of two distinct regulators, VP16 and HIRA. VP16 is a potent and well-characterized transactivator derived from herpesvirus, which has been associated with the relaxation of chromatin and and is known to interact with chromatin remodeling [13,14] and histone acetyltransferase complexes [15,16]. HIRA is a histone chaperone capable of nucleosome assembly and deposition outside of S-phase [17,18], with a role in deposition of the histone variant H3.3 [17,19].
Tethered VP16 or HIRA caused the level of gene conversion to increase by approximately an order of magnitude (8.4-fold and 11.0-fold, respectively). While these two tethered factors had similar functional outcomes, the localized changes in chromatin structure that they produced were quite distinct. Tethered VP16 greatly increased levels of acetylated histones, a mark associated with permissive DNA structure. Tethered HIRA did not alter local histone acetylation, but increased nucleosome deposition and caused local ordering of chromatin structure. Thus, different pathways of modification, and therefore different chromatin states, can achieve the same end: promotion of homologous recombination. Epigenetic regulation may be an important mechanism of both preserving and modifying genomic structure.

Tethered VP16 accelerates gene conversion
To determine the effects of chromatin structure on gene conversion, we have taken advantage of the powerful physiological model provided by the chicken B cell line, DT40. In the DT40 cell line, which derives from a B cell lymphoma, the Ig heavy (IgH) and light (Igl) chain variable (V) regions constitutively diversify by gene conversion. Gene conversion is templated by an array of homeologous upstream pseudo-V (yV) regions ( Figure 1A), which lack promoters and are nonfunctional. The yVl donors are enriched for acetylated histones H3 and H4 (AcH3 and AcH4), although they are not transcribed [12].
To ask if further enrichment of AcH3 and AcH4 could be achieved, or could affect recombination, we used the DT40 PolyLacO-l cell line constructed by our laboratory. In this cell line, polymerized lactose operator (PolyLacO) has been inserted into the yVl array between yVl17-yVl20, 17 kb upstream of the expressed Vl gene ( Figure 1A; [12]. This allows us to assay the effects of tethered regulatory factors expressed as fusions with lactose repressor. We generated DT40 PolyLacO-l transfectants stably expressing the activation domain of VP16 fused to GFP-LacI, GFP-LacI-VP16. A single green spot was readily imaged within the nuclei of the DT40 PolyLacO-l GFP-LacI-VP16 transfectants, providing evidence of GFP-LacI-VP16 expression and binding to PolyLacO ( Figure 1B). The cell cycle profile ( Figure 1C) and levels of l transcription ( Figure 1D) were unaltered relative to control DT40 PolyLacO-l GFP-LacI transfectants.
We assayed histone acetylation in PolyLacO-l GFP-LacI-VP16 and control DT40 PolyLacO-l GFP-LacI cells by chromatin immunoprecipitation (ChIP). There are two l light chain alleles in DT40 cells, one rearranged and actively expressing l light chain (l R ), and the other unrearranged and not transcribed (l U ). Allelespecific chromatin modifications were assayed at the rearranged allele of Vl (Vl R ); the unrearranged allele of Vl (Vl U ); and yVl17 (yVl17 R ), where sequence polymorphism was created by insertion of PolyLacO-l, providing the only site in the yVl array at which the rearranged and unrearranged alleles can be distinguished by use of specific PCR primers. Modifications on both alleles were assayed by amplification of five sites, including the region between yVl1 and the Vl gene (yV1-Vl); yVl1; yVl5; yVl13; yVl24. Strikingly, tethered VP16 increased levels of both AcH3 ( Figure 2A) and AcH4 ( Figure 2B); while tethered GFP-LacI did not have this effect. Enrichment of AcH3 was 40fold relative to input at the transcribed Vl R allele, and 10-fold at the allele-specific site, yVl17. There was essentially no enrichment at the nontranscribed Vl U allele (Figure 2A, left), indicating that the activity of VP16 was specifically targeted to the PolyLacOcontaining allele. AcH4 also displayed allele-specific enrichment, in the range of 45-to 50-fold at Vl R and yVl17; and 20-fold at most upstream sites in the yVl array. Very little enrichment was observed at the most downstream element, yVl1. This site similarly shows very little enrichment in AcH3 or AcH4 in the parental DT40 line [12], Comparison of enrichment of AcH3 and AcH4 in DT40 PolyLacO-l cells in which either GFP or VP16 was tethered to the yVl array showed that tethered VP16 caused a dramatic increase in local histone acetylation at almost all sites interrogated. An  exception was yVl5, where tethered VP16 diminished AcH3 levels. This site is notable for its naturally high levels of AcH3 and AcH4 in the parental DT40 cell line [12]. As with AcH3, we did not observe any enrichment of AcH4 at the nontranscribed Vl U allele control ( Figure 2B). The dramatic enrichment of both AcH3 and AcH4 at the rearranged and expressed allele Vl R is especially notable because tethered VP16 did not cause an increase in Vl transcript levels ( Figure 1D).
We used the sIgM loss variant assay to ask if tethered VP16 altered the clonal rate of Vl sequence diversification. This fluctuation assay measures the fraction of cells that no longer express structurally intact sIgM, and thus scores mutation events resulting from gene conversion, point mutation, insertion or deletion [12,20,21]. While gene conversion is the predominant pathway of Ig gene diversification in chicken B cells, if gene conversion is impaired (for example by the absence of essential recombination factors [20,22,23,24,25,26], other repair outcomes, particularly point mutations, will predominate. Thus diversification is monitored best by this loss of function assay, rather than an sIgM 2 to sIgM + gain of function assay, which scores only gene conversion [27] and is sensitive to variations in reversion frequencies at different sites within Vl [28]. Independent clonal derivatives of DT40 PolyLacO-l GFP-LacI and DT40 PolyLacO-l GFP-LacI-VP16 transfectants were established by limiting dilution cloning of sIgM + cells, and following 4 weeks of culture the fraction of sIgM 2 cells in each population was determined by flow cytometry of cells stained with anti-IgM antibody. A considerable fraction of sIgM 2 cells was evident in most clonal populations of DT40 PolyLacO-l GFP-LacI-VP16 transfectants relative to DT40 PolyLacO-l GFP-LacI transfectants (e.g. Figure 2C). Comparison of median percentages of sIgM 2 cells showed that DT40 PolyLacO-l GFP-LacI-VP16 transfectants exhibited an 8.4-fold increase in clonal diversification rate relative to DT40 PolyLacO-l GFP-LacI control cells ( Figure 2D).

Histone H3.3 is enriched at the yVl donor genes
One histone variant associated with chromatin activation and deposited independently of replication is H3.3, the principal H3like variant to contain activation-associated postranslational modifications [29,30]. As the presence of H3.3 could provide another mechanism to activate donor chromatin for recombination, we asked if H3.3 was enriched at the yVl array. To distinguish H3 from H3.3 by ChIP, we generated DT40 derivatives that stably express FLAG-tagged human H3.3 (H3.3-FLAG; [31] and characterized H3.3 deposition by ChIP with an anti-FLAG antibody. H3.3-FLAG was 4-fold enriched at the expressed Vl R gene relative to input DNA, and considerably enriched throughout the yVl array ( Figure 3). Because levels of H3.3-FLAG expression may vary among different transfectants, we normalized levels at the light chain locus relative to levels at the ovalbumin (Ova) gene, which is not expressed in B cells and was devoid of H3.3. Enrichment of H3.3 within the yVl array does not simply represent a graded spreading of this variant from the transcribed Vl R gene to sites upstream, as sites near Vl R did not consistently display higher levels of enrichment than more distant sites. This non-uniform chromatin structure suggests that ciselements, particularly in the yVl5 region, may be especially important for regulation of expression and/or diversification.

Tethered HIRA promotes gene conversion
Enrichment of H3.3 at the yVl donors raised the possibility that this variant histone may be important for efficient homologous recombination. The histone chaperone HIRA can promote assembly of nucleosomes containing histone H3.3 [17,19], suggesting that tethered HIRA might accelerate Ig gene diversification, analogous to tethered VP16. To test this, we generated stable DT40 PolyLacO-l HIRA-LacI transfectants, verified expression of the protein by Western blotting (Figure 4A), and showed that tethered HIRA did not affect cell cycle profile ( Figure 4B) or l transcript levels ( Figure 4C).
To ask if tethered HIRA affects the clonal rate of Ig gene diversification, we established independent clonal populations of DT40 PolyLacO-l HIRA-LacI transfectants by limiting dilution cloning of sIgM + cells, and following 4 weeks of culture the fraction of sIgM 2 cells in each population was determined. Flow cytometry showed a considerable fraction of sIgM 2 cells in most HIRA-LacI transfectants, but not in GFP-LacI control transfectants (e.g. Figure 4D). Comparison of median percentages of sIgM 2 cells showed that DT40 PolyLacO-l HIRA-LacI cells exhibited an 11.0-fold increase in diversification relative to DT40 PolyLacO-l GFP-LacI controls ( Figure 4E). This did not simply reflect overexpression of HIRA, as only a modest increase in the rate of clonal diversification (1.7-fold) was evident in transfectants of DT40 lacking a PolyLacO element but stably expressing HIRA-LacI ( Figure 4E).

Tethered VP16 and tethered HIRA promote gene conversion
Sequence analysis of Vl regions amplified by single cell PCR from cells expressing tethered VP16 or HIRA showed that most diversification was by gene conversion (Supplementary Figures S1 and S2). Table 1 summarizes mutational events. The great majority of sequence changes in both DT40 PolyLacO-l GFP-LacI-VP16 transfectants and DT40 PolyLacO-l R HIRA-LacI transfectants were due to gene conversion (82% and 87%, respectively), comparable to levels of gene conversion documented in DT40 PolyLacO-l GFP-LacI transfectants (80%; Cummings et al. 2007). However, the fraction of gene conversion events containing only a single substitution per tract was considerably higher in DT40 PolyLacO-l GFP-LacI-VP16 transfectants and DT40 PolyLacO-l R HIRA-LacI transfectants (37% and 24%, respectively) than in DT40 PolyLacO-l R GFP-LacI transfectants (3%). Thus, while essentially all gene conversion tracts in the GFP-LacI control cells contained at least two mutations, acceleration of diversification by either tethered VP16 or HIRA was accompanied by a diminished number of nucleotide substitutions per tract.  Tethered HIRA causes a local increase in histone deposition and nucleosome density ChIP analysis showed that, in contrast to VP16, tethered HIRA did not affect AcH3 and AcH4 levels, relative to a GFP-LacI control ( Figure 5A). To ask if tethered HIRA increased H3.3 levels, we generated DT40 PolyLacO-l FLAG-H3.3 HIRA-LacI transfectants, and then determined enrichment of FLAG-H3.3 at yVl17 R , where a polymorphism enables allele-specific analysis of chromatin. Those experiments showed a modest (1.4-fold) enrichment of FLAG-H3.3 relative to a control line ( Figure 5B), reproducible in an independent HIRA-LacI transfectant (not shown). Moreover, a comparable enrichment was evident when pan-H3 antibodies were used for ChIP (which detect H3 and other variants, including H3.3). Thus, tethered HIRA appeared to cause a net local enrichment of histones.
To probe nucleosome loading, nuclei were harvested from DT40 PolyLacO-l GFP-LacI, GFP-LacI-VP16 and HIRA-LacI transfectants and treated with varied amounts of micrococcal nuclease (MNase), which attacks the linker regions between nucleosomes to produce a characteristic ladder. Genomic DNA was then purified, resolved and blotted, and the PolyLacO region probed by hybridization. The MNase digestion patterns of DT40 PolyLacO-l GFP-LacI and DT40 PolyLacO-l GFP-LacI-VP16 cells were similar to each other ( Figure 5C, left, middle) and to the parental DT40 PolyLacO-l line (data not shown). Digestion produced a ladder of fragments (labeled 1-4 in the figure), which appear as blurred rather than sharp bands, reflecting heterogeneity in size of nucleosome multimers produced upon partial digestion. In contrast, MNase digestion of chromatin from DT40 PolyLacO-l HIRA-LacI cells revealed a distinctive laddering pattern, in which a sharply defined series of digestion products spanned from the 1-mers produced upon limit digestion up to 8-mers and beyond ( Figure 5C, right). This effect has been observed in two independent HIRA-LacI transfectants (not shown).
The diminished heterogeneity of products of partial digestion of chromatin from DT40 PolyLacO-l HIRA-LacI cells could reflect more ordered nucleosome phasing. Alternatively, increased nucleosome density may contribute to size homogeneity of partial digestion products. As diagrammed ( Figure 5D), a decrease in linker length would limit nuclease access, and cause products of partial digestion to be less heterogeneous in size, thus generating the homogeneous, sharply-defined fragments evident in HIRA-LacI transfectants. Increased loading of nucleosomes would be consistent with the known role of HIRA as a histone chaperone, as well as with the evidence that tethered HIRA increased local enrichment of H3.3 ( Figure 5B).

Discussion
We have shown that localized changes in donor chromatin structure can accelerate gene conversion. Our experiments showed two different chromatin modifiers, the activation domain of the transcription factor VP16, and the histone chaperone, HIRA, promoted gene conversion when tethered to the yVl donor array in DT40 B cells. These results identify epigenetic modification as a powerful mechanism for the regulation of homology-directed repair. Additionally, it provides both positive and negative regulation, as permissive donor chromatin structure can stimulate or repressive chromatin structure inhibit [12] homologous recombination.
The DT40 Ig yVl donor array normally carries marks typical of transcriptionally active chromatin, including enrichment of AcH3, AcH4, and H3.3. Tethered VP16 or HIRA increased levels of these marks, while activating recombination but not transcription. This suggests that marks that activate recombination and transcription may be overlapping, reflecting dependence of both processes on similarly relaxed chromatin structure. Consistent with this, marks characteristic of activated chromatin correlate with activation of Ig V, D and J segments for V(D)J recombination, a site-specific recombination process that is developmentally regulated, [32] and accompanied by sterile transcription of the target genes [33,34,35,36]. Moreover, transcriptional regulatory elements can contribute to activation of V(D)J recombination [32,37]; and retroviral integration favors transcribed regions [38].
Tethered VP16 increased levels of AcH3 and AcH4. That histone acetylation might promote gene conversion was suggested by evidence that Ig gene conversion was stimulated by culture of cells with the histone deacetylase inhibitor trichostatin A (TSA) [39,40]. However, because TSA functions throughout the genome, its effects could be mediated not only by local changes in chromatin structure but also by promoting expression of transacting factors which in turn stimulated gene conversion.
Tethered HIRA modulated nucleosome composition by depositing increased levels of H3.3, and appeared to increase local nucleosome density. How might this result in a more permissive chromatin state for gene conversion? Based upon the relative sensitivities of intact nucleosomes to solvent conditions, Jin and Felsenfeld [41] concluded that H3.3-containing nucleosomes were intrinsically less stable, and likely to be more sensitive to chromatin remodeling complexes. Thus, by altering the fraction of H3.3, HIRA might predispose nucleosomes toward displacement during critical steps of homologous recombination. Tethered HIRA had no effect on local histone acetylation, so deposition of H3.3 by HIRA was not sufficient for additional recruitment of HAT activity and local increase of histone acetylation, in contrast to tethered VP16. HIRA has been associated with senescence-associated heterochromatin foci (SAHFs) [42,43], but we did not find evidence that tethered HIRA induced a heterochromatic state within yVl array. We were unable to detect enrichment of diMe-, triMe-H3K9 or the repressive histone variant macroH2A (data not shown), all of which have been demonstrated to be components of SAHFs [42,43].
Tethered VP16 or HIRA accelerated clonal diversification rates by promoting gene conversion, but this was accompanied by an increase in the fraction of conversion tracts (37% and 24%, respectively) that carried only a single nucleotide substitution, a category essentially absent (3%) in GFP-LacI transfectants. This may reflect a decrease in the physical length of conversion tracts, which could in turn be due to limiting levels of factors necessary to promote gene conversion at the elevated rates dictated by tethered VP16 or HIRA. Alternatively, as gene conversion must involve some form of invasion of the donor by the recipient DNA, alterations in chromatin structure induced by tethered VP16 or HIRA may affect nucleosome position in a way that limits the physical extent of sequence accessible for transfer. For example, the increase in nucleosome density caused by tethered HIRA may be accompanied by a decrease in the average linker length; and, if linker regions are preferred templates for gene conversion, shortening of those regions may contribute to decreased gene conversion tract length in the HIRA transfectants. Alternatively, nucleosome phasing may be changed, so that even though chromatin is more accessible for the homology search, the regions within the yV segments that can participate in gene conversion are shorter. Finally, the tethered factors may create chromatin that is more accessible to mismatch repair, so a significant fraction of mutations could in principle be reverted, to create a large class of mutants with only single base changes.
Just as epigenetic regulation can activate chromatin for homologous recombination, the disregulation of chromatin structure may promote recombination leading to loss of heterozygosity characteristic of aging cells and of cancer [44,45,46,47]. These events have important implications for human disease, as recombination between nonallelic low copy repeats has been demonstrated to create DNA microdeletions that are associated with complex genetic traits including mental retardation [48], autism [49], and schizophrenia [50]. That activating marks at donor alleles promote gene conversion further suggests that chromatin structure should be considered in design of donors for in situ gene correction. The evidence that the same functional endpoint can be achieved by distinct chromatin marks shows that the cell has available multiple, redundant pathways for effecting local alterations of chromatin structure that regulate homologous recombination.

Materials and Methods
Cell culturing, imaging, RT-PCR, quantitation of sIgM loss variants and sequence analysis Cell culture and transfection were carried out as described [12,21]. Control experiments showed that the PolyLacO tag does not affect cell proliferation, cell cycle profile or Ig gene diversification. The p39ss-EGFP and p39ss-EGFP-VP16 plasmids [15,51] were provided by Dr. Andrew Belmont (University of Illinois); and pH3.3-FLAG [31] by Dr. Gary Felsenfeld (NIH).
For fluorescence imaging, cells (2610 5 ) were cytospun onto glass slides and fixed with 2% paraformaldehyde for 15 min. To visualize the nucleus, cells were stained with DAPI (Sigma, Saint Louis, MO). Fluorescent images were acquired using the DeltaVision microscopy system (Applied Precision) and processed with softWoRx software (Applied Precision).
To quantitate sIgM loss variants [12,20,21], in brief, sIgM + cells were isolated by flow cytometry followed by limiting dilution cloning, expanded for 4 weeks, then approximately 1610 6 cells were stained with anti-chicken IgM-RPE (Southern Biotechnology Associates, Birmingham, AL), and analyzed on a FACScan with CellQuest software (BD Biosciences). For sequence analysis, sIgM 2 cells were sorted, aliquoted to single wells, Vl regions amplified and sequenced, and their sequences compared to the yVl donors to determine if mutations were templated or nontemplated. Sequences were derived from at least two independently transfected lines. Only unique sequences were included for classification of the mutations.

Chromatin immunoprecipitation (ChIP)
ChIP was carried out as previously described [12]. For all experiments at least two chromatin preparations from at least two independent stably-transfected lines were analyzed. Figures present one representative experiment in which results from analysis of at least three separate amplifications were used to calculate a standard deviation. Separate amplifications of serial dilutions of template DNA were carried out to establish that the measured product intensities were within the linear range. Enrichment of the experimental amplicon was normalized to enrichment of an internal control amplicon from the ovalbumin (Ova) gene, which is not transcribed in chicken B cells. Ova was amplified in the same tube by duplex PCR; and enrichment upon ChIP with specific antibodies was normalized to parallel experiments in which ChIP was carried out with total input DNA controls. Inclusion of the Ova internal control amplicon permitted normalization for IP efficiency, background carryover, differences in effects due to variations in transgene expression, and error due to gel loading. The yVl:Ova enrichment ratio from immunoprecipitations was normalized to yVl:Ova ratio from total input DNA to determine fold enrichment = [(yVl/Ova)Ab]/ [(yVl/Ova)Input]. As an additional control, the ratio of the experimental and control amplicons in the total input control was compared to a control ChIP with polyspecific IgG; in all cases, enrichment in input and IgG controls were essentially equal.
PCR primers for ChIP were:

Micrococcal nuclease digestion and Southern blotting
Nuclei were prepared and digests were performed with 0, 3, 7, 15, 30, and 60 units of MNase as described [53]. Following MNase digestion, DNA was extracted three times with phenol:chloroform:isoamyl alcohol, once with chloroform, and precipitated with ethanol; and 20 mg resolved by agarose gel electrophoresis, and transferred for hybridization. The lacO probe was labeled by random priming, using as template a ,300 bp fragment containing LacO repeats.