Acetyltransferases GCN5 and PCAF are required for B cell mat- uration in mice

B lymphocyte development includes two DNA recombination processes, the V(D)J recombination of immunoglobulin (Igh) gene variable region and class switching when the Igh constant regions are changed from IgM to IgG, IgA, or IgE. The V(D)J recombination is required for successful maturation of B cells from pro-B to pre-B to immature-B and then to mature B cells in the bone marrow. The CSR occurs outside the bone marrow when mature B cells migrate to peripheral lymphoid organs, such as spleen and lymph nodes. Both V(D)J recombination and CSR depend on an “open chromatin” state that makes DNA accessible to specific enzymes, recombination activating gene (RAG), and activation-induced cytidine deaminase (AID). Acetyltransferases GCN5 and PCAF possess redundant functions acetylating histone H3 lysine 9 (H3K9). Here, we generated by complex breeding a mouse model with B cells lacking both GCN5 and PCAF. We found that double-deficient mice possess low levels of mature B cells in the bone marrow and peripheral organs, accumulation of pro-B cells in bone marrow, and reduced CSR levels. We concluded that both GCN5 and PCAF are required for B cell development in vivo.


Introduction
Development of B lymphocytes starts in the bone marrow where progenitor (pro)-B cells using recombination-activating genes (RAG) generate DNA double-strand breaks (DSBs) and initiate the V(D)J recombination. In maturating B cells, the V(D)J recombination process is genetic recombination of variable (V), diversity (D), and joining (J) gene segments arranging into a newly formed VDJ part of immunoglobulin gene (Ig) (1-4). Following the V(D)J recombination, B cells develop from pro-B cells expressing specific markers cluster of differentiation 19 (CD19), B220/CD45, and CD43 (CD19+B220+CD43+) to pre-B cells (CD19+B220+CD43-), immature B (CD19+B220+IgM+, low immunoglobulin M, IgM) and mature B (CD19+B220+IgM+, high IgM) cells in bone marrow (1). Mature B lymphocytes leave the bone marrow and through the blood migrate to periphery populating spleen and lymph nodes.
Then, mature B cells initiate another DNA recombination process to change the constant regions of immunoglobulin genes referred to as class switch recombination (CSR). During the CSR in mice, IgM is replaced by IgG3, IgG1, IgG2a, IgG2b, IgE, and IgA (1). The CSR is initiated by non-productive transcription known as a germ-line transcription (GLT) which is needed to separate two DNA strands. Single-stranded DNA is then targeted by activation-induced cytidine deaminase (AID), a B lymphocyte-specific enzyme deaminating cytosine to uracil (C to U). Then uracil DNA N-glycosylase (UNG) removes the uracil from DNA leading to the single-strand break formation (SSB) (5). Two SSBs facing each other then form a DSB and allow recombination (1-4). Both V(D)J and CSR can be regarded as processes following fundamentally similar strategies of genomic recombination (6,7).
The DSBs formed during the V(D)J recombination and CSR are recognized, processed, and repaired by the non-homologous end-joining pathway (NHEJ) and initiate more complex signaling and chromatin modification known as DNA damage response (DDR) (1-4). The NHEJ is initiated when Ku70 and Ku80 heterodimer (Ku) recognizes and binds the DSBs. Ku serves as a platform for downstream factors including DNA-dependent protein kinase catalytic subunit (DNA-PKcs), X-ray repair cross-complementing protein 4 (XRCC4), XRCC4-like factor (XLF), Paralogue of XRCC4 and XLF (PAXX), a modulator of retrovirus infection (MRI), XRCC4 and DNA ligase 4 (LIG4) (2)(3)(4). There are additional factors that are sometimes optional for NHEJ, including Artemis with nuclease activity required for processing of hairpin-sealed DNA ends and overhangs.
One type of DDR pathway acts downstream of ataxia telangiectasia mutated (ATM) protein kinase, which is activated by DSBs and then phosphorylates multiple substrates, including NHEJ and DDR factors. ATM phosphorylates histone H2AX, which in turn recruits mediator of DNA damage checkpoint 1 (MDC1) and facilitates the accumulation of really interesting new gene (RING) finger motif (RNF) 8 and RNF168 ubiquitin-ligases, and then the p53-binding protein (53BP1). Phosphorylation of H2AX is related to the acetylation of histones, including histone H3K9 (8). Acetylation of histone H3K9 is mediated by GCN5 and PCAF (8,9).
General control non-depressible 5 (GCN5) acetyltransferase is also known as lysine acetyltransferase (KAT) 2A. Germline inactivation of GCN5 in mice resulted in early embryonic lethality due to the role of the protein in neurogenesis (25). GCN5 is functionally redundant with another acetyltransferase, KAT2B, also referred to as p300/CBP-associated factor (PCAF). While inactivation of Pcaf gene in mice has no detectable phenotype, double knockout of Gcn5/Pcaf genes resulted in even earlier embryonic lethality (25). Because histone H3K9 acetylation is related to DDR (8), one could propose that GCN5, PCAF, or both enzymes are required for lymphocyte development in vivo. However, embryonic lethality of Gcn5 -/-and Gcn5 -/-Pcaf -/-mice (25) challenged the studies. To overcome the obstacle, we developed a complex mouse model, when Pcaf gene was germline-inactivated (25), while floxed Gcn5 gene was conditionally-inactivated in B cell lineages by CRE recombinase expressed under Cd19 promoter (26). To sort out CRE-positive and CRE-negative cells, we used Rosa26-stop-YFP knockin which only expressed YFP following the CRE activation (27).
Here, we found that GCN5 and PCAF acetyltransferases are functionally redundant during early B cell maturation, while GCN5 is required for robust CSR.

Generation of mice lacking GCN5 and PCAF in B cells
Combined inactivation of Gcn5 and Pcaf genes in mice results in embryonic lethality (25). To overcome this challenge, we designed a complex genetic model when floxed Gcn5 gene is conditionally inactivated in B cell lineages by CRE enzyme under the Cd19 promoter (Cd19 Cre+ ) (26). To sort out the cells with activated CRE, we used a model with knocked-in yellow fluorescent protein gene (YFP) into ROSA-26 locus. The YFP is inactive until CRE removes "STOP" signal (Rosa-26-YFP+) (27). Thus, we obtained Gcn5 f/f Pcaf -/-Cd19 +/Cre YFP + mice and simpler controls. Further in the text, we will skip Cd19 +/Cre and YFP+ for simplicity in most of the cases, and will refer to the mice based on the status of Gcn5 and Pcaf genes, i.e. as Gcn5 f/f Pcaf -/-, Gcn5 f/f , Pcaf -/-and WT. When the CRE is active and describing sorted B cells, we indicate Gcn5 -/-, a knockout status of the gene. Lack of GCN5 and PCAF was also validated by western blot ( Figure S1) and quantitative PCR.

Mice lacking GCN5 and PCAF in B cells are alive and possess small spleens
We obtained alive mice with germline inactivation of Pcaf gene and conditional inactivation of Gcn5 in B cells under the Cd19 promoter ( Figure 1). We found that germline inactivation of Pcaf gene alone has no detectable effect on mouse development, in line with the previous observation (25). Conditional inactivation of Gcn5 gene in B cells had no visible effect on sizes of WT and Pcaf-deficient mice, which were 15 to 19 g on average (p>0.1433) ( Figure 1A). However, inactivation of Gcn5 resulted in smaller spleens in mice (Gcn5 -/-, 54 mg), when compared to WT (69 mg) and Pcaf -/-(72 mg) mice. Combined inactivation of Pcaf and Gcn5 in B cells resulted in even smaller spleens (Gcn5 -/-Pcaf -/-, 29mg, p<0.0001). Spleens of mice without CRE activity with Gcn5 gene being floxed and functional (Gcn5 f/f Pcaf -/-, 67 mg), were comparable in size to the ones of WT (Figure 1 B, C).

Mice lacking GCN5 and PCAF in B cells are alive and possess small spleens
To detect mature Gcn5 -/-Pcaf -/-B cells, we identified B220+IgM+ cells in the spleen using flow cytometry (Figure 2 A, B). Inactivation of Pcaf gene alone did not affect B cell proportions in the spleen (58%) when compared to WT mice (52%, p=0.4777) (Figure 2A

Inactivation of Gcn5 and Pcaf results in a reduced proportion of B cells in the blood
To detect mature B cells in the blood, we used B220 markers ( Figure 2C, D). Inactivation of Pcaf alone resulted in 62% of B cells after red blood cells were lysed, which was comparable to WT mice with 65% of B cells in the blood (p=0.92). Inactivation of Gcn5 gene alone resulted in a modest reduction of B cell proportion to 56% (p=0.12), while combined inactivation of Gcn5 and Pcaf led to very low B cell levels in blood (22%, p<0.0001). Levels of B cells in blood of control mice without CRE recombinase expression, when Gcn5 gene was functional (Gcn5 f/f Pcaf -/-) were comparable to WT mice (78%). We concluded that GCN5 and PCAF are both required and functionally redundant for B cell development in mice.
One reason of low B cell count in spleen and blood of Gcn5 -/-Pcaf -/-mice could be cell death following normal development of B cells in bone marrow and migration to periphery. Another option could be blocked or delayed maturation of B cells in bone marrow during the earlier stages. To test the latter possibility, we analyzed B cell development in bone marrow of the mice (Figure 3).

Inactivation of Gcn5 and Pcaf results in accumulation of pro-B cells in bone marrow
To characterize B cell maturation in bone marrow, we followed the expression of B220 (B220+IgM-) and IgM (B220+IgM+) on the lymphocyte surface. Inactivation of Gcn5 or Pcaf resulted in an insignificant decline of B220+IgM+ population (26%, 5-6 million) when compared to WT (32%, 8 million, p=0.44) (Figure 3 A, B). Combined inactivation of Gcn5 and Pcaf resulted in an additional reduction of mature B cells in bone marrow (3 million, 16%) (Figure 3 A, B).
We concluded that GCN5 and PCAF are required for the maturation of B cells from the pro-B cell stage to pre-B and later to mature B cells.

GCN5 and PCAF are required for robust class switch recombination
The CSR depends on the ATM-dependent DDR (1, 3,4,20). Because H3K9 acetylation works downstream of H2AX phosphorylation, and GCN5/PCAF might work downstream of ATM/ATR/DNA-PKcs, we tested if the CSR depends on GCN5 and PCAF (Figure 4). We purified B splenocytes from 8 to 12 weeks old mice and stimulated the CSR from IgM to IgG3 using established protocols (32,34). We focused on matched pairs of Gcn5 f/f (the functional equivalent of WT cells) and Gcn5 -/-, as well as Gcn5 f/f Pcaf -/-(the functional equivalent of Pcaf -/-) and Gcn5 -/-Pcaf -/-cells. We used Aid -/-cells as a CSR-deficient control to detect an experimental background (Figure 4). Inactivation of Pcaf alone had no effect on CSR levels (WT vs Gcn5 f/f Pcaf -/-, p>0.96). Contrary, the inactivation of Gcn5 resulted in a reduction of CSR levels from about 14% in WT and Gcn5 f/f cells to 6% in Gcn5 -/-cells, *p<0.0008 (Figure 4). Combined deletion of Pcaf and Gcn5 resulted in a similar reduction from 12% in Gcn5 f/f Pcaf -/-cells to 6% in Gcn5 -/-Pcaf -/-cells, **p=0.0080. We concluded that GCN5 is required for robust CSR to IgG3, because additional inactivation of Pcaf did not affect CSR levels when compared Gcn5 -/-and Gcn5 -/-Pcaf -/-B cells, p>0.9999 (Figure 4).

Discussion
Both GCN5 and PCAF are involved in chromatin modification and DDR response, which made them relevant candidates to facilitate lymphocyte development (8,9). One challenge was the lack of a relevant in vivo model because GCN5 and PCAF have certain redundant functions in acetylating H3K9, and because of the germline inactivation of Gcn5 results in early embryonic lethality in mice (25). Here, we generated and analyzed a complex mouse model which allowed studying of double-deficient Gcn5 -/-Pcaf -/-B cells development in vivo and ex vivo. We used a germline knockout of Pcaf (25), a conditional knockout of Gcn5 f/f , a knockin of CRE recombinase expressed under the B cell-specific Cd19 promoter (26), and a knockin of YFP to track the activity of CRE recombinase (27).
For such a complex mouse model (Gcn5 f/f Pcaf -/-Cd19 +/cre Rosa-26-YFP + ), multiple controls were used. In one line of the controls, the mice lacking PCAF, having floxed Gcn5 gene but expressing no CRE recombinase were used (Gcn5 f/f Pcaf -/-Rosa-26-YFP + , Figure S2). The GCN5-deficient and GCN5/PCAF double-deficient B cells possessed developmental delay with lower levels of mature B cells in spleen and blood, and accumulation of progenitor B cells in bone marrow (Figures 1-4). Contrary, the control mice without CRE expression demonstrated WT levels of B cell development in all the groups, i.e. WT levels of B220+IgM+ mature B cells in the spleen ( Figure S2 A  The mice lacking PCAF and with conditional knockout of Gcn5 in B cells were of normal size and visually not different from the WT and PCAF-deficient littermates ( Figure  1A). One clear feature the Gcn5 f/f Pcaf -/-Cd19 +/cre Rosa-26-YFP + mice had was a small spleen (Figure 1 B and C) which was also the case in mice lacking Gcn5 in B cells. The small spleen could indicate a defect in B cell development, and we indeed found low numbers of mature B cells in the spleen, blood, and bone marrow. One could propose that mature B cells lacking GCN5 or both GCN5 and PCAF possess low proliferation speed or tend to trigger apoptosis. Alternatively, GCN5 and PCAF might be required for the V(D)J recombination. This option could be tested by, for example, using vAbl pre-B cell lines as we and others did before, e.g. (3,10,11,16,(20)(21)(22)(23)(24). Another intriguing question is whether the physical presence or enzymatic activity of GCN5 and PCAF are required for the observed phenotypes, i.e. abrogated B cell maturation and reduced levels of CSR. To investigate this question one could use specific inhibitors of GCN5 and PCAF enzymes, or enzyme-dead mutations introduced to the Gcn5 and Pcaf genes.
The CSR levels were reduced in B cells lacking GCN5 (Figure 4). The challenge in this set of experiments was that the mice of Gcn5 f/f Pcaf -/-Cd19 +/cre Rosa-26-YFP + genotype were rather rare and possessed a very low number of suitable B splenocytes (Figures 1  and 2). Although our data on IgG3 is sufficient, one could extend the study in the future by generating knockout cell lines lacking GCN5 and PCAF and suitable for CSR. One possible model system is CH12F3 cells capable to support CSR to IgA (36), which were used in the past for this kind of experiment (17,29,34). The CSR itself is a complex multistage process. Generating relevant cell lines will also provide tools to determine specific stages of CSR affected in GCN5-deficient mice, i.e. germline transcription, AID recruitment, generation of DSBs, or DNA repair.

Conclusions
Acetyltransferases GCN5 and PCAF possess redundant functions in B cell maturation. GCN5 is required for class switch recombination in vivo.