Alternative Splicing Regulates Prdm1/Blimp-1 DNA Binding Activities and Corepressor Interactions

Abstract

Prdm1/Blimp-1 is a master regulator of gene expression in diverse tissues of the developing embryo and adult organism. Its C-terminal zinc finger domain mediates nuclear import, DNA binding, and recruitment of the corepressors G9a and HDAC1/2. Alternatively spliced transcripts lacking exon 7 sequences encode a structurally divergent isoform (Blimp-1Δexon7) predicted to have distinct functions. Here we demonstrate that the short Blimp-1Δexon7 isoform lacks DNA binding activity and fails to bind G9a or HDAC1/2 but retains the ability to interact with PRMT5. To investigate functional roles of alternative splicing in vivo, we engineered novel mouse strains via embryonic stem (ES) cell technology. Like null mutants, embryos carrying a targeted deletion of exon 7 and exclusively expressing Blimp-1Δexon7 die at around embryonic day 10.5 (E10.5) due to placental defects. In heterozygous Δexon7 mice, there is no evidence of dominant-negative effects. Mice carrying a knock-in allele with an exon 6-exon 7 fusion express full-length Blimp-1 only, develop normally, are healthy and fertile as adults, and efficiently generate mature plasma cells. These findings strongly suggest that the short Blimp-1Δexon7 isoform is dispensable. We propose that developmentally regulated alternative splicing is influenced by chromatin structure at the locus and fine-tunes Blimp-1's functional capabilities.

INTRODUCTION

Originally cloned as a silencer of beta interferon gene expression (34) and subsequently identified as a master regulator of plasma cell terminal differentiation (68, 82), the zinc finger transcriptional repressor Prdm1/Blimp-1 governs a rapidly expanding repertoire of developmental processes (3). In the early embryo, Blimp-1 acts downstream of BMP/Smad signals to silence the default somatic pathway and specify primordial germ cells (58, 83). At later stages, Blimp-1 regulates development of the placenta, posterior forelimb, pharyngeal arches, secondary heart field, and sensory vibrissae (64). Blimp-1 has essential roles in adult tissue homeostasis. Besides plasma cells, Blimp-1 also controls gene expression in CD4 and CD8 T lymphocytes (30, 32, 48, 65, 70), the sebaceous gland (25), and skin epidermis (46). Additionally, Blimp-1 is required during osteoclast differentiation (53, 57) and postnatal reprogramming of intestinal enterocytes (22).

Blimp-1 has a modular structure comprising an amino-terminal PR/SET domain, a central proline- and serine-rich region, and 5 carboxy-terminal zinc fingers (81). The first two C₂H₂ fingers are sufficient for recognition of a Blimp-1 consensus motif, AGTGAAAGTG (35), that closely resembles the IRF interferon-stimulated regulatory element (ISRE) (GAAANNGAAANN) (34, 39). Blimp-1 binds competitively with IRF-1 and IRF-2 at these sequences (39, 80). An artificially engineered truncated Blimp-1/Prdm1 protein lacking the C-terminal zinc finger domain (ZFD) fails to display functional activity in vivo (31, 64). Blimp-1 collaboratively silences target gene expression in association with chromatin-modifying corepressors, including histone deacetylases 1 and 2 (HDAC1/2) (85), G9a lysine methyltransferase (19), protein arginine methyltransferase 5 (PRMT5) (1), and lysine-specific demethylase 1 (LSD1) (75). HDAC2 (85) and G9a (19) have been shown to interact with the zinc finger domain, whereas LSD1 (75), HDAC2 (85), and Groucho (62) associations were previously mapped to the proline- and serine-rich region.

In eukaryotes, the presence of structurally related but distinct protein isoforms produced by alternative mRNA processing greatly augments proteome diversity (47, 56). These variants can contribute shared, unique, or sometimes opposing functions. The activities of several well-known C₂H₂ zinc finger proteins are governed by this mechanism (50, 84). Early characterization of the Prdm1 gene structure revealed alternatively spliced transcripts lacking exon 7 sequences that encode a drastically truncated zinc finger domain (81). The short Blimp-1Δexon7 isoform failed to bind the Blimp-1 consensus site upstream of the c-myc promoter (81) and displayed a distinct localization pattern within the nuclei of transfected cells (66). This alternative isoform is conserved in humans and potentially plays a dominant-negative regulatory role (66). Recent experiments demonstrated expression of this splice variant by human natural killer (NK) T cells (72), whereas only the full-length product was detectable in mouse NK T cells (33).

To examine possible functional contributions made by this alternative isoform during early mouse development and/or adult tissue homeostasis, we engineered mouse strains that exclusively express full-length Blimp-1 or the short Blimp-1Δexon7 isoform only. Like loss-of-function mutants, Prdm1Δex7 embryos arrest at around embryonic day 10.5 (E10.5) due to placental defects, despite abundant expression of Blimp-1Δexon7 protein that efficiently localizes to the nucleus. Heterozygous animals coexpressing full-length Blimp-1 and Blimp-1Δexon7 develop normally, are healthy and fertile as adults, and contain B lymphocytes that efficiently undergo terminal differentiation to become plasma cells. Thus, the present experiments provide no evidence suggestive of dominant-negative effects in vivo. Mice carrying the ex6+7 knock-in allele that express full-length Blimp-1 only are also indistinguishable from wild-type mice. Complementary chromatin immunoprecipitation (ChIP), systematic evolution of ligands by exponential enrichment (SELEX), and coimmunoprecipitation experiments confirmed that the short Blimp-1Δexon7 isoform lacks detectable DNA binding activity, fails to interact with HDAC1 and G9a, and displays reduced binding to HDAC2. Surprisingly, Blimp-1Δexon7 efficiently binds cytoplasmic PRMT5, whereas full-length Blimp-1 preferentially binds nuclear PRMT5. Finally, we document robust expression of the alternatively spliced variant in immature dendritic cells (DCs). Collectively, our experiments demonstrate that developmentally regulated alternative splicing has a dramatic impact on the functional activities of Blimp-1/Prdm1.

MATERIALS AND METHODS

Gene targeting.

Targeting vectors were generated by subcloning a 9-kb fragment (Acc65I-XhoI fragment) of the bacterial artificial chromosome bMQ-126O12 (Gene Services) into a modified version of pBlueScript II (Stratagene). For the Δex7 construct, SacII and ApaI sites were used to introduce a mutagenized fragment containing a loxP site at the SacII site and a loxP-flanked pgk-neomycin (flox-Neo) gene (74) inserted 120 bp 3′ of exon 7 into exogenous BamHI and NheI sites. The ex6+7 targeting vector was generated by inserting a PmlI-ApaI fragment containing exon 6 fused to exon 7, with flox-Neo inserted into the downstream intron. For both constructs, an hsv-tk cassette was inserted outside the 3′ homology region. Gene targeting was performed in CCE mouse embryonic stem cells as described previously (55). Targeted clones were identified by Southern blotting and transiently transfected with pMC1-Cre to remove the drug resistance cassette.

PCR genotyping.

Genotyping was performed as described previously (55). All animal experiments were performed in accordance with Home Office regulations. The primers and cycling conditions were as follows: for Prdm1Δex7, primers Δex7 COM (GCAATGTCTGTGCCAAGACGTTC), Δex7 WT (GCTCAAGCTGGGTCTCCTATGG), and Δex7 MUT (GCTGGGCACACAGTACTCTGAGGTAC), with cycling at 94°C for 20 s, 58°C for 30 s, and 72°C for 30 s for 38 cycles; and for Prdm1 ex6-7, primers 67A (GCTGGGCACACAGTACTCTGAGGTAC), 67B (CAGGTCCACCTGAGAGTGCACAG), and 67C (TAGAATTCCTGCAGGTCGAG), with cycling at 94°C for 15 s, 57°C for 15 s, and 72°C for 30 s for 35 cycles.

RT-PCR.

Reverse transcription-PCR (RT-PCR) was performed as described previously (55, 64). Primer sequences were as follows: Ex4F, GTTATTGGCGTGGTAAGTAAGG; Ex5R, ATTTATCACTGTGAGCTCTCCAG; Ex6F, GGTTACAAGACTCTTCCTTAC; Ex8R, GCTCTTGTGACACTGGGCACA; HprtF, GCTGGTGAAAAGGACCTCT; and HprtR, CACAGGACTAGAACACCTGC.

Immunoprecipitation and Western blotting.

Cell lysates were prepared and Western blot analysis was performed as described previously (55). The following antibodies were used: Blimp-1 (1:500) (SC-130917; Santa Cruz), β-tubulin (1:1,000) (SC-9104; Santa Cruz), G9a (1:200) (D141-3; MBL), HDAC2 (1:5,000) (AB14169; Abcam), HDAC1 (1:500) (AB31263; Abcam), PRMT5 (1:2,000) (07-405; Millipore), anti-mouse immunoglobulin (Ig)–horseradish peroxidase (HRP) (1:500) (NA931V; GE Healthcare), anti-rat Ig–HRP (1:1,000) (NA935V; GE Healthcare), anti-rabbit Ig–HRP (1:2,000) (NA934V; GE Healthcare), and anti-Armenian hamster Ig–HRP (1:2,000) (C-2904; Santa Cruz) antibodies.

For coimmunoprecipitation analysis, nuclear and cytoplasmic extracts (67) were prepared from HeLa cells transfected with pCAGGS-Blimp-1, Blimp-1Δexon7, or a control empty vector (55). Blimp-1 was immunoprecipitated using mouse anti-Blimp-1 (NB600235; Novus) or rabbit anti-Blimp-1 (9115; Cell Signaling) antibody and protein G Dynabeads (Invitrogen).

Immunohistochemistry.

E9.5 placentae were processed and stained with rabbit anti-Blimp-1 (1:1,000; a kind gift of Reuben Tooze) as described previously (55).

Primary cell culture.

Lipopolysaccharide (LPS)-stimulated splenic B cells (55) and bone marrow dendritic cells (BMDCs) (26) were cultured as described previously.

Immunofluorescence analysis.

Flow cytometry was performed using a FACSCalibur flow cytometer (BD Biosciences), and data were analyzed with FlowJo software (Tree Star). For plasma cell staining, day 3 LPS-treated splenocytes were incubated with phycoerythrin (PE)-conjugated anti-mouse CD45 receptor (CD45R)/B220 (553090; BD Pharmingen) and allophycocyanin (APC)-conjugated rat anti-mouse CD138 (Syndecan-1) (558626; BD Pharmingen). 7-Amino-actinomycin D (7-AAD) (559925; BD Pharmingen) was added to exclude dead cells from the analysis. Additionally, forward and side scatter gating allowed us to enrich for plasma cells. Reported numbers refer to the percentages of cells within the indicated gates.

Purification of zinc finger domains.

DNA fragments encoding the segment from amino acid 468 to the carboxyl terminus of Blimp-1 and Blimp-1Δexon7 were PCR amplified using the primers ZF-For (GATAAGATCTGCCAGCATGAAGGACGAGAGTAG) and ZF-Rev (GATACTCGAGTTAAGGATCCATCGGTTCAACTG) (BglII and XhoI sites are underlined) and cloned into a modified version of pET28a (Novagen) containing a 6× histidine tag and a hemagglutinin (HA) tag. Proteins were expressed in T7 Express Escherichia coli (C3013H; New England BioLabs) induced with 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) at 37°C for 3 h in LB medium containing 50 μM zinc chloride and 25 μg/ml kanamycin. Cells were lysed in buffer L (50 mM Tris-HCl, pH 7.3, 300 mM potassium chloride, 20 mM imidazole, 10 μM zinc chloride, 0.01% Triton X-100, 10 mM 2-mercaptoethanol, 20 units/ml lysozyme, 10 μl/ml EDTA-free protease inhibitor [Sigma]) by incubation on ice for 40 min followed by 5 rounds of sonication. Clarified lysates were incubated with Ni-nitrilotriacetic acid (Ni-NTA) agarose beads (Qiagen) for 1 h at 4°C. Beads were washed 4 times with buffer W (20 mM Tris-HCl, pH 7.3, 100 mM potassium chloride, 20 mM imidazole, 10 μM zinc chloride, 100 μM dipotassium EDTA, 10 mM 2-mercaptoethanol, 20% glycerol). Proteins were eluted using buffer E (20 mM Tris, pH 7.3, 100 mM potassium chloride, 250 mM imidazole, 10 μM zinc chloride, 100 μM dipotassium EDTA, 10 mM 2-mercaptoethanol, 20% glycerol, 2 μl/ml EDTA-free protease inhibitor [Sigma]). The concentration of His-tagged protein was estimated by Coomassie blue staining and expressed relative to that of bovine serum albumin standards.

EMSA.

Oligonucleotides used to generate electrophoretic mobility shift assay (EMSA) probes were as follows: for PRDI, TAGGGATCAAGTGAAAGTGAAAGTGAAAGTGAGATC and TAGGGATCTCACTTTCACTTTCACTTTCACTTGATC (35); for 13× GT, TAGGACGGTGTGTGTGTGTGTGTGTGTGTGTGTGAGG and TAGGCCTCACACACACACACACACACACACACACCGT; and for 4× TTTGTG, TAGGACGGTGTTTGTGTTTGTGTTTGTGTTTGTGAGG and TAGGCCTCACAAACACAAACACAAACACAAACACCGT. Annealed oligonucleotides were 5′ end filled with [α-³²P]dCTP by use of the Klenow fragment of DNA polymerase. EMSA binding reaction mixtures were made up to a 20-μl volume at room temperature for 20 min and contained 20 mM HEPES, pH 7.9, 12 mM Tris-HCl, pH 7.3, 60 mM potassium chloride, 2.5 mM magnesium chloride, 120 μM dipotassium EDTA, 50 μM zinc chloride, 12% glycerol, 2.5 mM dithiothreitol, 6 mM 2-mercaptoethanol, 50 ng/μl poly(deoxyinosine-deoxycytosine), 100 ng/μl bovine serum albumin, 20,000 cpm of DNA probe, and 3, 10, 30, or 100 nM recombinant protein. Antibody supershifting was performed by addition of 30 ng/μl anti-HA antibody (clone 3F10; Roche). Reaction mixtures were electrophoresed through a gel containing 0.5× Tris-borate-EDTA (TBE), 5% acrylamide, and 1% glycerol at 150 V for 1.5 h.

SELEX.

A SELEX library was generated by performing a single round of PCR using the oligonucleotides GATAGGCGCGCCCTAACTAGCGACCTCG(N)₂₀GAGTACGGCACGCTATGGCGCGCCTATC and GATAGGCGCGCCATAGCGTGCCGTACTC. The library was digested with AscI (sites are underlined in the sequences) and end filled with [α-³²P]dCTP by use of the Klenow fragment. For DNA selection, EMSA binding reaction mixtures contained 24 ng of labeled probe and 100 nM recombinant protein. Gels were exposed to X-ray film, and developed films were used as a guide to excise supershifted complexes. For Blimp-1Δexon7, for which a shifted complex was not observed, a large section of the lane was excised to isolate low-abundance complexes that might be present. Purified DNA was used as the template for a PCR with the primers GATAGGCGCGCCCTAACTAGCGACCTCG and GATAGGCGCGCCATAGCGTGCCGTACTC. PCR products were extracted with phenol-chloroform, digested with AscI, labeled with Klenow fragment, and used in EMSA. This process was repeated 4 times, at which point PCR-amplified DNA was cloned into pCR-TOPO-XL (Invitrogen) and sequenced (GeneServices).

Chromatin immunoprecipitation.

ChIP was performed as described previously (8). Full-length Blimp-1 and Blimp-1Δexon7 coding sequences were PCR amplified with the primers GFP-For (GAGACTCGAGACCATGGGAATGGACATGGAGGACGCTGATATGAC) and GFP-Rev (GAGAGGTACCCAGGATCCATCGGTTCAACTGTTTCTTG) and cloned into pEGFP-N2 (XhoI and Acc65I sites are underlined) to generate carboxy-terminal enhanced green fluorescent protein (EGFP) fusions. Expression constructs were transfected into HeLa cells by use of Lipofectamine 2000 (Invitrogen). At 48 h posttransfection, cells were cross-linked with 1% formaldehyde for 20 min at room temperature and quenched with 125 mM glycine. Isolated chromatin was immunoprecipitated with mouse anti-GFP (clone 3E6; Invitrogen) and protein G Dynabeads (Invitrogen). DNAs in immunoprecipitated samples were then purified and analyzed by quantitative PCR (qPCR) using QuantiTek SYBR green PCR mix on a Rotorgene Q machine (Qiagen). Primers used for qPCR were as follows: for c-Myc, CAGTGCGTTCTCGGTGTG and CAGCCGAGCACTCTAGCTCT; for Tapbp, CCAGGCACCTTCACCTAACC and CAGCCATGAAGCCTCCTCTT; for Tln, GGGCACCTATCCACTGTCC and TCCCCTACATTTGCATAGCC; for Psmb8, GCTCGGACCCAGGACACTAC and TACTGCCCCGACCTGCAT; for Psmb10, GGGCACAGCAAGGGACAT and GTGGCGGTTTTCTGCATCTT; and for Krt10, TGGACACACCCTCTCAGTATATAAAGG and AGAGTAGTGCTTGCTTGAGCTGTATC (13).

RESULTS

Alternative splicing regulates Blimp-1 interactions with transcriptional corepressors.

Chromatin-modifying corepressor interactions have been mapped to the Blimp-1 proline-serine-rich and zinc finger domains (3). The histone H3 lysine 9 methyltransferase G9a/EHMT2 (76) associates with zinc fingers 1 and 2 to mediate repression at the beta interferon gene promoter (19). HDAC1 and -2 interactions with the proline-serine-rich and zinc finger domains are required for silencing of c-myc transcription (75, 85). In primordial germ cells, complexes with the arginine methyltransferase PRMT5 govern nuclear localization and probably contribute to epigenetic reprogramming (1). However, the fine specificity of Blimp-1 interactions with PRMT5 has yet to be characterized.

To investigate Blimp-1Δexon7 associations with transcriptional corepressors, we performed coimmunoprecipitation experiments using extracts prepared from transfected HeLa cells (Fig. 1). In contrast to full-length Blimp-1, Blimp-1Δexon7 failed to interact with HDAC1 or G9a and showed only residual binding to HDAC2 in nuclear extracts (Fig. 1A). Intriguingly, cytoplasmic extracts contained Blimp-1Δexon7–PRMT5 complexes (Fig. 1B), whereas full-length Blimp-1 selectively bound nuclear PRMT5 (Fig. 1A).

Fig 1.

Blimp-1Δexon7 fails to bind nuclear G9a, HDAC1/2, or PRMT5 but forms complexes with cytoplasmic PRMT5. (A) Immunoprecipitation (I.P.) and Western blot analysis of nuclear extracts from HeLa cells transfected with untagged constructs. Lysates were precipitated with either rabbit or mouse monoclonal Blimp-1 antibody. As expected, full-length Blimp-1 interacted with endogenous HDAC1, HDAC2, G9a, and Prmt5. In contrast, the Δexon7 isoform showed weak binding to HDAC2 and failed to bind the other corepressors in nuclear extracts. L, G9a long isoform; S, G9a short isoform. (B) Immunoprecipitation and Western blotting of cytoplasmic extracts. Lysates were precipitated with mouse monoclonal Blimp-1 antibody. PRMT5 associations with Blimp-1Δexon7 were readily detectable. All samples were normalized for the amount of immunoprecipitated Blimp-1 or Blimp-1Δexon7. (C) Selective expression of β-tubulin, Grb2 (cytoplasmic), and lamin A (nuclear) markers by cytoplasmic and nuclear extracts.

Blimp-1Δexon7 lacks DNA binding activity.

Blimp-1 represses c-myc expression to arrest cell division and promote plasma cell terminal differentiation (42, 43). The alternative Blimp-1Δexon7 isoform fails to bind the well-characterized consensus target site upstream of the c-myc promoter (66, 81). We also observed that Blimp-1Δexon7 failed to recognize the consensus site upstream of beta interferon (Fig. 2C). Additional candidate targets with conserved consensus binding motifs have been described extensively for a wide variety of settings (6, 12, 13, 46, 49, 65, 72). For instance, Blimp-1 controls the major histocompatibility complex (MHC) class I peptide loading pathway by direct repression of gamma interferon-induced genes, including the Tapbp (Tapasin) chaperone gene and the Psmb8 (Lmp7) and Psmb10 (Mecl1) proteasomal component genes (12). We performed ChIP experiments to examine Blimp-1Δexon7 occupancy at the following previously characterized target genes: c-myc, Tapbp, Tln, Psmb8, and Psmb10 (12, 13). As expected, full-length Blimp-1 efficiently bound these sequences but not the nontarget Krt10 gene (Fig. 2A). In contrast, Blimp-1Δexon7 gave no significant ChIP enrichment (Fig. 2A).

Fig 2.

The short Blimp-1Δexon7 isoform lacks DNA binding activity. (A) Anti-GFP ChIP of Blimp-1-GFP-, Blimp-1Δexon7-GFP-, and mock-transfected HeLa cells. Data represent mean % of input ± standard errors of the means (SEM) for triplicate transfections. (B) Western blot confirming comparable expression levels of Blimp-1 and Blimp-1Δexon7 in transfection reaction mixtures for ChIP. (C) EMSA using Blimp-1 and Δexon7 ZFDs and the PRDI sequence of the beta interferon promoter (35). ss, antibody supershift. (D) Fourth round of SELEX using Blimp-1 and Blimp-1Δexon7 ZFDs. ns, nonspecific band. (E) Alignment of selected Blimp-1 ZFD sequences from SELEX round 4. Perfect GAAAG Blimp-1 motifs (green) and single-mismatch motifs (blue) are indicated. (F) Alignment of Blimp-1Δexon7 ZFD sequences from SELEX round 4. TG and TTTGT motifs occasionally present in selected Blimp-1Δexon7 sequences are highlighted in pink. (G) Neither the Blimp-1Δexon7 nor Blimp-1 ZFD gave detectable binding to these sequences in EMSA.

To test whether Blimp-1Δexon7 may instead have the ability to recognize a diverse set of target sites with a distinct sequence motif(s), we exploited SELEX because this technique offers an unbiased approach for assessing DNA binding potential. As expected, the control Blimp-1 ZFD bound to representative sequences containing the GAAAG core motif (Fig. 2D and E). In contrast, even after multiple rounds of selection, there was no sequence enrichment, and we failed to detect any protein-DNA complex with the Blimp-1Δexon7 ZFD (Fig. 2D, F, And G). These results strongly argue that the short Blimp-1Δexon7 isoform lacks DNA binding activity.

Targeted deletion of exon 7 phenocopies the null allele.

Prdm1 is dynamically expressed at early developmental stages (64, 83). Null mutants die at around E10.5 due to placental defects (83). Prdm1 deficiency disrupts specification of primordial germ cells and severely compromises development of the pharyngeal arches (58, 83). Blimp-1/Prdm1 is also required at later stages in the secondary heart field, posterior forelimb, and sensory vibrissae (64). Importantly, genetic studies clearly demonstrate that the penetrance of these tissue disturbances is gene dosage dependent (55, 64).

To describe functional contributions made by the short Blimp-1Δexon7 isoform in vivo, we exploited a gene targeting strategy (Fig. 3). Two different mutant strains were engineered. The Prdm1Δex7 allele results in targeted deletion of exon 7 sequences and expression of the short isoform only (Fig. 3A and G). The Prdm1ex6+7 fusion allele eliminates the intronic region and thus permits expression of full-length Blimp-1 only (Fig. 3D and G). Heterozygous mice carrying these mutations are viable and fertile (Table 1). However, we failed to recover any live-born homozygous offspring from heterozygous Prdm1Δex7/+ intercross matings (Table 1).

Fig 3.

Generation of Prdm1Δex7 deletion and Prdm1-ex6+7 fusion alleles. (A) Prdm1Δex7 gene targeting strategy. LoxP sites (red arrows), Southern blot probes, restriction sites, and digest fragment sizes are indicated. E, EcoRI; S, SpeI. (B) Southern blot verification of Prdm1Δex7 targeting. Heterozygous targeted clones (+/Targ) were analyzed with 5′, 3′, internal (Int.), and Neo probes, using the indicated digests. The asterisk indicates a product from the neomycin-resistant embryonic fibroblast feeder cells. (C) Cre excision to generate the Prdm1Δex7 allele, verified using a SpeI digest and the internal probe. (D) Prdm1ex6+7 gene targeting strategy. E, EcoRI, B, BamHI. (E) Southern blot verification of Prdm1ex6+7-targeted clones, using an EcoRI digest and the 5′ probe. (F) Cre excision of the Neo cassette from the Prdm1ex6+7 allele, verified using a BamHI digest and the internal probe. (G) Transcript products produced by the wild-type (top), ex6+7 (middle), and Δex7 (bottom) alleles. (H) Sequence comparison of the wild-type (top) and Δex7 (bottom) zinc finger domains. Linker sequences (gray) and zinc binding residues (yellow) are indicated. Amino acids encoded by exon 7 are underlined in blue (top), and the exon 6-exon 8 splice junction is indicated with a black arrow (bottom). Alternative splicing disrupts zinc fingers 1 to 3 and replaces a zinc-binding histidine with phenylalanine (blue highlighting).

Table 1.

Genotypes of heterozygous intercross progeny

Prdm1 intercross	Age of progeny	No. (%) of progeny with indicated genotype			Total no. of progeny
+/Δex7 × +/Δex7		+/+	+/Δex7	Δex7/Δex7
	E9.5	18	35	17 (24.3)	70
	E10.5	16	21	9 (19.6)	46
	E12.5	9	18	0	27
	Weanlings	46	88	0	134
+/ex6+7 × +/ex6+7		+/+	+/ex6+7	ex6+7/ex6+7
	Weanlings	24	57	26 (24.3)	107

To evaluate the onset of lethality, we collected embryos at different stages. Like null mutants (83), homozygous Δex7 mutant embryos were recovered at Mendelian ratios at E9.5 but were absent after E10.5 (Table 1). Similarly, Δex7 mutant embryos and placentae displayed severe morphological disturbances identical to those previously reported for mice completely lacking the Blimp-1 protein (83) (Fig. 4A and E). In contrast, RT-PCR and Western blot experiments clearly demonstrated abundant expression of Δexon7 transcripts and protein (Fig. 4B to D). Thus, E9.5 homozygous Δex7 mutant embryos exclusively express Δexon7 transcripts at wild-type levels (Fig. 4B). Likewise, the Δexon7 protein is expressed robustly in mutant embryos (Fig. 4C) and placentae (Fig. 4D). Moreover, nuclear Blimp-1Δexon7 protein was clearly visible within the spongiotrophoblasts present in mutant E9.5 placentae (Fig. 4E). Nevertheless, the short Blimp-1Δexon7 isoform cannot functionally substitute for full-length Blimp-1 and fails to support embryonic development.

Fig 4.

Exon 7 deletion causes disturbances in embryonic and placental development. (A) Δex7 homozygous mutants are severely growth retarded compared to wild-type littermates at E10.5. (B) RT-PCR analysis using primers targeting Prdm1 exons 6 and 8 to detect alternative splicing and primers targeting exons 4 and 5 to detect total transcript levels. (C) Western blotting of E9.5 embryos shows that the Blimp-1Δexon7 protein in Δex7/Δex7 mutants is expressed at increased levels relative to those of full-length Blimp-1 (Blimp-1^FL) in wild-type littermates. (D) Western blotting of E9.5 placenta lysates shows that Blimp-1Δexon7 expression in Δex7 mutants is moderately increased in comparison to that of full-length Blimp-1 (Blimp-1^FL) in wild-type littermates. (E) Immunohistochemical staining of E9.5 wild-type and Δex7 placentae with rabbit polyclonal anti-Blimp-1 antibody. The higher magnification in the bottom panels reveals Blimp-1-positive nuclei (white arrows).

The short Blimp-1Δexon7 isoform is dispensable for embryonic development and plasma cell terminal differentiation.

Homozygous Prdm1ex6+7 mutants are born at Mendelian ratios (Table 1), are healthy and fertile as adults, and display no visible abnormalities or signs of disease when housed in a specific-pathogen-free environment. Blimp-1's functional requirements during plasma cell terminal differentiation have been analyzed extensively (29, 68, 82). To evaluate plasma cell maturation and confirm the mutation results for expression of full-length Blimp-1 only, we examined ex6+7 LPS-stimulated splenic B cells. RT-PCR and Western blot experiments demonstrated robust expression of full-length Blimp-1 transcripts and protein (Fig. 5A and B). Additionally, normal levels of Ig secretion were clearly observed (Fig. 5C). Flow cytometry analysis confirmed that full-length Blimp-1 expressed on its own efficiently promoted plasma cell differentiation (Fig. 5D and E). Thus, wild-type and homozygous ex6+7 mutant animals displayed indistinguishable numbers of plasma cells as judged by the appearance of a B220 and CD138 double-positive population (Fig. 5D and E). As a control, adult homozygotes that carried the Δex1A deletion encompassing the NF-κB binding sites (55) selectively lacked plasma cells (Fig. 5D and E).

Fig 5.

Targeted deletion of Δexon7 expression fails to perturb embryonic development or plasma cell terminal differentiation. (A) RT-PCR confirming the absence of exon 7 alternative splicing in LPS-stimulated splenocytes from ex6+7 homozygous mice. (B and C) Western blots of LPS-stimulated splenocytes from wild-type and ex6+7 mice showing normal levels of Blimp-1 (B) and immunoglobulin (C) expression. SNH fibroblasts and the J558L plasmacytoma cell line were included as negative and positive controls for Blimp-1 expression, respectively. μM, membrane IgM; μS, secreted IgM. (D and E) Flow cytometric analysis of LPS-stimulated splenocytes from wild-type, ex6+7, and Δex1A mice. B220⁺ CD138⁺ plasma cell numbers were comparable between wild-type, homozygous ex6+7, and heterozygous +/Δex7 mice but were dramatically reduced in Δex1A mice. (F) Western blot confirming coexpression of Blimp-1Δexon7 and the full-length protein in LPS-stimulated +/Δex7 heterozygous splenocytes.

Previous studies suggested that the Blimp-1Δexon7 isoform acts as a dominant-negative regulator of plasma cell differentiation (66). In contrast, here we observed efficient formation of plasma cells by Δex7 heterozygous mice coexpressing both the short isoform and full-length Blimp-1 under the control of endogenous regulatory elements (Fig. 5E and F). Thus, we concluded that the Blimp-1Δexon7 isoform lacks detectable feedback-suppressive activities during plasma cell differentiation.

Developmentally regulated alternative splicing in dendritic cells.

Blimp-1/Prdm1 protein expression levels increase dramatically during DC differentiation (6, 55, 73). Recent studies demonstrated that Blimp-1 directly targets cytokine gene expression (6) and binds competitively with the transcriptional activators IRF8 and PU.1 at the CIITA promoter (73). To examine alternative splicing during BMDC maturation and to confirm that Prdm1ex6+7 homozygous mutant BMDCs exclusively express full-length Blimp-1, we undertook RT-PCR, Western blot, and Northern blot analyses. As expected, the full-length Blimp-1 protein was strongly expressed by LPS-treated day 7 BMDCs from wild-type control and ex6+7 mice (Fig. 6A), whereas only weak expression was detectable in immature BMDCs (Fig. 6A). Similarly, Northern blots revealed a dramatic increase in mRNA expression levels in LPS-treated day 7 BMDC cultures (Fig. 6B). Mature plasma cells coexpressed roughly equal amounts of 5.7-, 4.3-, and 3.6-kb mRNA species (Fig. 6B), generated via usage of the 3 alternative polyadenylation [poly(A)] signals located in the 3′ untranslated region (3′-UTR) (81), whereas LPS-stimulated BMDCs seemed to express predominantly the 5.7-kb species due to usage of the most distal poly(A) site (Fig. 6B). RT-PCR experiments demonstrated that wild-type immature BMDCs strongly coexpressed both full-length and Δexon7 transcripts, whereas day 7 LPS-treated BMDCs exclusively expressed full-length transcripts (Fig. 6C). As expected, immature BMDCs from ex6+7 mice lacked expression of Δexon7 transcripts (Fig. 6C).

Fig 6.

Developmentally regulated exon 7 alternative splicing in BMDCs. (A) Western blot of wild-type and ex6+7 BMDCs. Blimp-1 protein was undetectable at day 4, weakly expressed in day 7 cultures, and strongly induced by LPS treatment. Wild-type and Δex7 embryo lysates were used to demonstrate the positions of the full-length (FL) and Δexon7 proteins. (B) Northern blot of RNAs from day 4 and day 7 untreated and day 7 LPS-treated BMDCs. Day 4 and day 7 untreated BMDCs expressed similar amounts of Prdm1 RNA. LPS treatment of BMDCs strongly induced expression of the largest (5.7 kb) Prdm1 transcript, whereas the shorter transcripts (4.3 and 3.6 kb) were upregulated to a lesser degree. In contrast to BMDCs, LPS-stimulated splenocytes expressed the shortest transcript most strongly, although they also contained substantial amounts of the longer transcripts. rRNA was used as a loading control (bottom). (C) RT-PCR analysis of wild-type and ex6+7 BMDCs. Primers targeting exons 6 and 7 detected Δex7 transcripts in wild-type BMDCs but not ex6+7 cultures. Expression of alternatively spliced transcripts was markedly downregulated upon LPS treatment.

DISCUSSION

Alternative splicing regulates Blimp-1 recruitment of corepressors and DNA binding activity.

The present ChIP experiments confirm that Blimp-1Δexon7 fails to bind the consensus motif upstream of the c-myc promoter (66, 81) and also demonstrate its inability to recognize a panel of target sites recently described in the literature (12, 13). As judged by EMSA and SELEX experiments, the recombinant Blimp-1Δexon7 zinc finger domain clearly lacks DNA binding activity. The exon 7 deletion selectively removes zinc fingers 1 through 3, whereas fingers 4 and 5 remain largely intact. Three C₂H₂ zinc fingers are often required for optimal DNA binding, but stable interactions with only a subset of zinc fingers have also been documented (40). Combinatorial functional activities are also a common feature of zinc finger proteins. For example, Friend of GATA (FOG) CCHC zinc fingers interact with the N-terminal CCCC finger of GATA-1 (15), whereas the general transcription factor TFIIIA binds to both 5S DNA and rRNA by using overlapping C₂H₂ fingers (37). The truncated Blimp-1Δexon7 zinc fingers lack DNA binding activity, but this variant structure could potentially mediate RNA binding or protein-protein interactions.

Blimp-1 lacks a classical nuclear localization signal (NLS), and similar to other zinc finger proteins (7, 23, 51), it depends on its zinc finger domain for nuclear import (19, 64). Like the case for Ikaros splice variants (7), nuclear import of Blimp-1Δexon7 is not strictly dependent on DNA binding. Unlike full-length Blimp-1, seen diffusely distributed throughout the nucleus, the short Δexon7 isoform becomes localized to discrete nuclear foci (66), and a significant proportion is often detectable in the cytoplasm. Efficient nuclear retention of Blimp-1 may therefore depend on its strong affinity for DNA target sites widely distributed throughout the genome. In contrast, Blimp-1Δexon7 fails to bind chromatin and may be sequestered within a distinct nuclear subcompartment.

G9a recruitment depends on the region spanning zinc fingers 1 and 2 (19). We found here that Blimp-1Δexon7 fails to interact with HDAC1 and G9a and only weakly binds to HDAC2. These results refine our understanding of the G9a associations and strongly imply that zinc finger 2 provides the crucial binding interface. HDAC2 was previously shown to associate bimodally with the zinc finger domain and the proline- and serine-rich region (85). Thus, residual HDAC2 binding to Blimp-1Δexon7 can be attributed to its interaction with the proline- and serine-rich region.

Considerable evidence suggests that Blimp-1/Prdm1 plays an active role as a scaffolding protein that organizes the structure and composition of gene silencing complexes. It is well known that G9a and HDAC1/2 interact with the C-terminal binding protein (CtBP) as a core component of higher-order structures (69). CtBP corepressors are essential for mouse embryonic and placental development (24). CtBP associations with numerous zinc finger proteins, including Prdm3 and Prdm16, have been characterized (27, 28). It will be interesting to learn whether Blimp-1 functions together with higher-order CtBP-HDAC-G9a complexes and possibly also recruits cell-type-specific partners to repress its transcriptional targets.

Transient Blimp-1 associations with PRMT5 direct histone arginine methylation and regulate target gene expression in primordial germ cells (PGCs) (1). However, fine-mapping studies of Blimp-1 associations with PRMT5 have not been reported. PRMT5 is predominantly cytoplasmic (63) but localizes to the nucleus in specific cell types, such as PGCs and trophoblasts (1, 14, 78). Surprisingly, Blimp-1Δexon7 fails to bind PRMT5 in nuclear extracts but shows a strong association in the cytoplasmic fraction. In contrast, full-length Blimp-1 exclusively binds nuclear PRMT5. One possibility is that the short Blimp-1Δexon7 isoform differentially associates with distinct PRMT5 complexes in the cytoplasm. Cytosolic Prmt5 is a core component of the methylosome complex, which is responsible for promoting assembly of Sm spliceosome proteins into snRNPs (16). In contrast, nuclear PRMT5 is found in association with Groucho (60) and HDAC2/mSin3A/Brg1 (59) repressor complexes that mediate histone methylation. PRMT5 methylation of the MBD2 repressor inhibits its DNA and HDAC binding (77), whereas methylation of nucleolin and Sm proteins regulates their nuclear localization (52, 79). Mutually exclusive recruitment of the cofactors pICln and RioK1 determines PRMT5 substrate specificity (17). Interestingly, RioK1 is strictly cytoplasmic, while pICln1 is predominantly nuclear (17). It will be important to learn more about PRMT5's regulatory functions in general, as well as structural features of Blimp-1Δexon7 that mediate selective interactions with cytoplasmic PRMT5 complexes.

Blimp-1Δexon7 behaves as a functional null and not a dominant-negative regulator.

The present experiments demonstrate that Δex7 homozygous mutant embryos die at about E10.5 due to placental defects. The short Blimp-1Δexon7 isoform expressed at physiological levels under the control of endogenous regulatory elements has no detectable feedback-suppressive activity, and in the context of embryonic development, it cannot substitute for full-length Blimp-1. Interestingly, in comparison to full-length Blimp-1 in wild-type embryos, expression of the short Blimp-1Δexon7 isoform is moderately increased in homozygous mutants. DNA binding activity is therefore required to mediate a previously described autoinhibitory feedback loop via recognition of the consensus binding site located in intron 2 at the Prdm1 locus (46). Like the case for an artificially engineered truncated Blimp-1 protein entirely lacking the C-terminal zinc finger domain, the Δexon7 isoform fails to mediate feedback suppression and shows increased levels of expression.

Alternative splicing may regulate Blimp-1 activity.

Alternative splicing provides an important regulatory mechanism governing the activities of numerous zinc finger transcription factors. For example, the members of the Ikaros family of zinc finger proteins contain both N-terminal DNA-binding zinc fingers and C-terminal zinc fingers that govern protein partnerships (21, 50, 54). Alternative splicing produces dominant-negative variants that lack DNA binding activity but retain the ability to form higher-order complexes (4, 36, 54). Similarly, zinc finger-deficient splice variants of KLF6 and ZIP possess dominant-negative activities (11, 86). The Wilms's tumor gene, WT-1, generates alternatively spliced zinc fingers including or excluding the lysine-threonine-serine (KTS) tripeptide (2, 20). The WT-1 (−KTS) isoform binds DNA and regulates transcription, whereas WT-1 (+KTS) binds RNA and associates with the splicing machinery (5, 41).

Conditional loss of Blimp-1 function in T cells causes an inflammatory bowel disease resulting in weight loss and death (30, 48). In contrast, in this study, Δex7 heterozygotes were indistinguishable from their wild-type littermates. The Δex7 heterozygotes developed normally, were healthy and fertile as adults, and efficiently generated mature plasma cells. The present experiments demonstrate that Blimp-1Δexon7 lacks DNA binding capability but also has no dominant-negative regulatory activity.

Recent evidence strongly suggests that transcription and splicing are mechanistically coupled (61). Thus, RNA polymerase II elongation rates influence alternative splicing events, and conversely, splicing factors such as SC35 regulate transcription elongation (10, 44). Moreover, local chromatin structure and histone modifications have been linked to alternative splicing patterns (38, 45). We detected Δexon7 transcripts in resting B cells (66) and immature BMDCs. Alternative splicing was dramatically reduced coincident with strongly upregulated expression upon maturation. Consistent with previous studies that demonstrated enhanced recruitment of splicing machinery to activated promoter regions (71), we also observed here that the weakly activated locus permits exon skipping, whereas increased levels of transcription favor exon inclusion. Thus, the pattern of Prdm1 alternative splicing is tightly coupled to the level of transcription at the locus.

The present experiments demonstrate that exon 7 alternative splicing is dispensable for embryonic development and plasma cell differentiation. Expression of the alternative Δexon7 isoform was recently documented for human natural killer cell subpopulations (72), but this product was undetectable in murine NK cells (33). These observations could potentially be explained due to species differences. Alternatively, culture conditions and experimental protocols for isolating NK cell subsets from human peripheral blood or mouse spleens could potentially have an impact on their maturation status. Additional evidence for evolutionary differences in key regulatory elements has been reported. For example, in humans, the BLIMP-1β isoform lacking the PR/SET domain is generated via an alternative transcriptional start site lying upstream of exon 4 (18). This truncated isoform retains G9a and DNA binding capabilities, has impaired repressor activity, and likely functions as a dominant-negative regulator in human NK and myeloma tumor cells (18, 72). The mouse locus lacks this alternative transcriptional initiation site (9). Studies of mice, zebrafish, and Xenopus demonstrate striking species differences in Prdm1/Blimp-1 developmental functions, expression patterns, and alternative promoter usage (3, 55). However, Prdm1 exon-intron boundaries are highly conserved between vertebrates, and the potential role of zinc finger alternative splicing in these diverse vertebrates remains to be explored. It is possible, of course, that exon 7 alternative splicing plays an important regulatory role and fine-tunes Blimp-1's functional capabilities during maturation of mammalian immune responses. Further examination of possible contributions to differentiation of plasma cell subpopulations, T cell homeostasis, and effector and memory functions in infection model systems awaits the development of backcrossed animals.