|Home | About | Journals | Submit | Contact Us | Français|
The zinc-finger PR domain transcriptional repressor Blimp-1/Prdm1 plays essential roles in primordial germ cell specification, placental, heart, and forelimb development, plasma cell differentiation, and T-cell homeostasis. The present experiments demonstrate that the mouse Prdm1 gene has three alternative promoter regions. All three alternative first exons splice directly to exon 3, containing the translational start codon. To examine possible cell-type-specific functional activities in vivo, we generated targeted deletions that selectively eliminate two of these transcriptional start sites. Remarkably, mice lacking the previously described first exon develop normally and are fertile. However, this region contains NF-κB binding sites, and as shown here, NF-κB signaling is required for Prdm1 induction. Thus, mutant B cells fail to express Prdm1 in response to lipopolysaccharide stimulation and lack the ability to become antibody-secreting cells. An alternative distal promoter located ~70 kb upstream, giving rise to transcripts strongly expressed in the yolk sac, is dispensable. Thus, the deletion of exon 1B has no noticeable effect on expression levels in the embryo or adult tissues. Collectively, these experiments provide insight into the organization of the Prdm1 gene and demonstrate that NF-κB is a key mediator of Prdm1 expression.
The PR/SET domain zinc-finger transcriptional repressor Blimp-1/Prdm1 was initially cloned as a negative regulator of IFNB1 (beta interferon) expression (30) and later identified as a factor both necessary and sufficient for B-cell terminal differentiation and antibody secretion (74, 79). Blimp-1, the protein encoded by Prdm1, silences expression of key transcription factors, such as c-Myc, required for cell cycle progression (43), as well as Pax5, Id3, and Spi-B, which maintain mature B-cell identity (41, 71). Prdm1 inactivation in the T-cell lineage results in fatal inflammatory bowel disease associated with reduced interleukin 10 and upregulated expression of interleukin 2 and gamma interferon (28, 49). In the skin, Prdm1 is required for sebaceous gland homeostasis (22) and epidermal terminal differentiation (48). Prdm1 has a dynamic pattern of expression in the developing mouse embryo (10, 60, 67, 81). Loss-of-function mutant embryos fail to specify primordial germ cells, display pharyngeal arch defects, and die around embryonic day 10.5 (E10.5) due to placental insufficiency (60, 81). Conditional rescue experiments have revealed additional roles in multipotent progenitor cell populations in the forelimb, secondary heart field, and sensory vibrissae (67). Thus, Prdm1 regulates cell fate decisions in diverse contexts in the embryo and governs tissue homeostasis in multiple cell types in the adult organism.
The cis-acting regulatory elements that direct tissue-specific Prdm1 expression in these specialized cell types are largely unknown. A Venus fluorescent reporter transgene embedded within a 230-kb bacterial artificial chromosome (kb −140 to +90 relative to the transcription start site) faithfully drives temporally and spatially restricted expression at numerous sites in the embryo, including primordial germ cells, anterior definitive endoderm, somites, pharyngeal arches, limb buds, and dermal papillae (60, 61). In contrast, an enhanced green fluorescent protein reporter construct containing 4.4 kb upstream of the Prdm1 transcription start site is sufficient for expression in adult hematopoietic tissues and mediates lipopolysaccharide (LPS) responsiveness of splenic B cells (83). However, this construct also leads to ectopic expression at numerous tissue sites. The cis-acting regulatory elements controlling dynamic patterns of Prdm1 expression in vivo thus potentially span a large genomic region. Dose-dependent BMP-Smad signals activate Prdm1 expression in committed primordial germ cells when they initially appear at the base of the allantois (60). However, it remains unknown whether Prdm1 is a direct Smad target. A recent study identified a Gli3 binding site ~27 kb downstream of the Prdm1 coding region that drives expression in the developing limb (82). Similarly, studies of zebra fish have shown that Sonic Hedgehog controls Prdm1 expression during pectoral fin and muscle development (5, 40). However, multipotent progenitor cell populations allocated at numerous tissue sites express Prdm1 only transiently (67). Multiple, as yet uncharacterized enhancer and repressor elements are almost certainly required to regulate graded Prdm1 activities throughout development.
Alternative promoter usage offers an attractive mechanism for regulating Prdm1 gene expression. Two alternative promoters control spatially and temporally distinct blimp1/krox expression patterns during sea urchin development (44, 45). These alternative transcripts encode nearly identical proteins except that the 1b isoform contains 50 additional residues at its amino terminus. Specific morpholino knockdown of the 1a and 1b transcripts results in tissue-specific disturbances in the gut endoderm and vegetal plate, respectively (44). The activity of an alternative promoter region located 5′ of PRDM1 exon 4 that generates a protein lacking the PR/SET domain with reduced repressive activity on multiple target genes is elevated in human myeloma lines relative to levels in primary B cells (21).
The Prdm1 basal promoter and multiple transcriptional start sites were previously mapped immediately upstream of exon 1 (78). To learn more about developmentally regulated expression, we characterized the 5′ ends of Prdm1 transcripts in the developing embryo. We identified two novel alternative first exons that both splice directly to exon 3, containing the translational start site. Exon 1B, located 70 kb upstream of exon 1A, is strongly expressed in the yolk sac. An additional first exon (exon 1C) is located in the intron downstream of exon 1A. To evaluate the possibly distinct functional activities contributed by alternative promoters, we generated targeted alleles that selectively eliminate either exon 1A (Δex1A) or exon 1B (Δex1B) transcripts. The exon 1B deletion slightly decreases expression in the yolk sac but otherwise has no noticeable effect in the embryo or adult tissues. Surprisingly, the exon 1A deletion encompassing NF-κB sites upstream of the promoter eliminates Prdm1 expression in LPS-stimulated B cells and blocks plasma cell differentiation but fails to disrupt embryonic development. Consistent with this, we observe only modestly reduced Prdm1 expression levels in the embryo. However, compound heterozygotes also carrying the null allele display partially penetrant developmental defects. The novel alternative promoters described in this report are likely to play important roles in generating regulatory diversity and controlling gene dosage effects.
The Δex1A targeting vector was generated by ligating a 2.9-kb 5′ homology region (StuI-PstI), a loxP-flanked pgk-neomycin cassette from PGKneolox2DTA (76), and a 4.7-kb 3′ homology region (AfeI-EcoRV) into a modified version of pBSII-KS(−) (Stratagene). An hsv-thymidine kinase (hsv-tk) cassette was inserted outside the 3′ homology region. The Δex1B targeting vector was generated by ligating a 12.2-kb Acc65I-NdeI fragment from the bMQ-381N6 bacterial artificial chromosome (Geneservice, Cambridge, United Kingdom) into a modified version of pBSIIKS(−) (Stratagene). The loxP-flanked pgk-neomycin cassette was introduced at XhoI and SpeI sites, and the hsv-tk cassette was ligated outside of the 3′ homology region. Gene targeting was carried out in CCE embryonic stem (ES) cells. A linearized targeting vector (15 μg) was introduced by electroporation. Homologous recombinant clones were selected in the presence of G418 (200 μg/ml) and 1-[2′-deoxy-2′-fluoro-β-d-arabinofuranosyl]-5-iodouracil (0.1 μg/ml). Drug-resistant colonies were screened by Southern blot analysis using the restriction enzyme and probe combinations shown in Fig. Fig.2.2. For the Δex1A allele, we recovered 28 correctly targeted clones out of ~860 drug-resistant colonies, and in the case of the Δex1B allele, 16 correctly targeted clones out of ~1,150 colonies. Targeted clones were transiently transfected with pMC1Cre and screened for excision of the loxP-flanked pgk-neomycin cassette by Southern blotting.
DNA was prepared as described previously (53). The following primers and conditions were used for the Δex1A allele: common primer, GCCAGACCCTGAGATGACTACATTG; wild-type primer, CACAGCAAAACAAAAGCCCAAC; mutant primer, CGAAGCGGACAAGAACCACTACTG; 54°C annealing temperature, 40 cycles; for Δex1B, wild-type primer 1, TTGAGGTTCACGCACGAATG; wild-type primer 2, GACTTTTGCTTGCTATGCCCTG; mutant primer 1, CCTAAAAAGGTGCGAGTAAGGTGAG; mutant primer 2, TACATCCCCAGCCCAGAGGTTG; 58°C annealing temperature, 40 cycles. Prdm1BEH (81), Prdm1null (74), and Prdm1gfp (27) mice were genotyped as described previously.
5′ random amplification of cDNA ends (RACE) cloning was performed using the second-generation 5′/3′ RACE kit (03 353 621 001; Roche) with slight modifications. Two micrograms of total RNA was reverse transcribed using a primer annealing in exon 4 (Ex4Rev: CTCCTTACTTACCACGCCAA). First-strand cDNA was purified (11 732 668 001; High Pure PCR product purification kit; Roche), dA-tailed, and PCR amplified (Platinum Pfx polymerase; 11708; Invitrogen) using a nested primer in exon 3 (Ex3Rev1, GTGCTCGAGCGTCAGCGCCGGAATCCCAGG) and the oligo(dT)-anchor primer (GACCACGCGTATCGATGTCGACTTTTTTTTTTTTTTTTV). The XhoI and SalI cloning sites are underlined. An annealing temperature of 55°C was used for Pfx amplification. Subsequently, 0.5 μl of the reaction mixture was used as a template for a second round of amplification (ReddyMix PCR Mastermix; AB-0575/LD; Thermo Scientific) using Ex3Rev1 and the oligo(dT)-anchor primer (60°C annealing temperature). Products were gel purified (Qiaex II kit; 20051; Qiagen), digested with XhoI and SalI, cloned into pBSII-KS(−) (Stratagene), and sequenced (Geneservice). In some cases, where the second round of amplification failed to yield discrete products, the Pfx reaction mixture was instead amplified using ReddyMix PCR Mastermix with a nested primer (Ex3Rev2: CTGCCAGTCCTTGAAACTTC) in combination with the oligo(dT)-anchor primer. In this case, PCR products were gel purified and cloned directly into pCR-XL-TOPO (Invitrogen). Sequences were aligned to the mouse genome using the BLAT function of the University of California, Santa Cruz (UCSC), genome browser (31, 37).
The pCAGGS-Blimp-1 expression vector (67) was transiently transfected into COS-7 and HEK293 cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Total RNA was isolated using Trizol (Invitrogen), and reverse transcription-PCR (RT-PCR) was performed using the OneStep RT-PCR kit (Qiagen). Primers were as follows: exon 1A forward (Ex1AFor), CGTAGAAAAGGAGGGACCGCC; exon 1B forward (Ex1BFor), GTTTGCATTCACCGAAGTTGC; exon 1C forward (Ex1CFor), CCGGGACACAGGACGCAG; exon 3 reverse no. 1 (Ex3Rev1), CGTCAGCGCCGGAATCCCAGG; exon 3 reverse no. 2 (Ex3Rev2), CTGCCAGTCCTTGAAACTTC; Hprt forward (HprtFor), GCTGGTGAAAAGGACCTCT; Hprt reverse (HprtRev), CACAGGACTAGAACACCTGC.
An RNase protection assay (RPA) was carried out using the RPAIII kit (AM1415; Ambion) as described previously (2). Sizes of the probes and protected fragments are summarized in Table Table1.1. Band intensities were normalized to the Sp1 signal. Each percentage represents the average for two independent normalized mutant samples in comparison with the average for two independent normalized wild-type control littermate samples.
Age-matched and whenever possible sex-matched homozygous mutant and wild-type control littermates derived from intercross matings were sacrificed at 6 to 8 weeks of age. Spleen cell suspensions were depleted of erythrocytes by ammonium chloride-Tris treatment. To induce plasma cell differentiation, splenocytes (2.5 × 106 cells/ml) were cultured for 3 days in the presence of LPS (50 μg/ml) (Escherichia coli 055:B5; Difco Laboratories). Alternatively, for T-cell activation, the B cells were depleted using anti-CD45R (B220) magnetic microbeads (495-01; Miltenyi Biotec) according to the manufacturer's instructions. The nonadherent cells were cultured at a density of 5 × 105 cells/ml for 3 days on anti-mouse T-cell receptor β-chain (553166; BD Biosciences)-coated dishes.
BMDCs were isolated as described previously (24). Cultures were initially plated at 4 × 105 cells/ml in the presence of granulocyte/macrophage-colony stimulating factor (25 ng/ml) (415-ML; R&D Systems), fed on day 3 and day 6, and harvested on day 7. Where indicated, dendritic cell maturation was induced during the last 20 h of culture by the addition of LPS (20 μg/ml).
Cells were lysed in radioimmunoprecipitation assay buffer plus protease inhibitors and extracts prepared as described previously (67). Following sodium dodecyl sulfate-polyacrylamide gel electrophoresis, proteins were transferred onto a polyvinylidene difluoride membrane (Millipore) at 300 V for 2 h. Membranes were blocked for 1 h at room temperature in Tris-buffered saline-Tween containing 7% nonfat dry milk and incubated in primary anti-Blimp-1 (1:500, rat monoclonal 5E7) (27) antibody overnight at 4°C or in primary anti-ß-tubulin (1:1,000, rabbit polyclonal, sc-9104; Santa Cruz) or anti-mouse immunoglobulin (Ig) (H+L)-horseradish peroxidase (HRP) (1:500; NA931V; GE Healthcare) for 1 h at room temperature. Anti-rat Ig-HRP (1:1,000; NA935V; GE Healthcare) and anti-rabbit Ig (1:2,000; NA934V; GE Healthcare) secondary antibody incubations were for 1 h at room temperature. Bands were quantified using a ChemiDoc XRS imager (Bio-Rad) and the QuantityOne software program (Bio-Rad) and normalized to ß-tubulin.
E9.5 placentae were fixed overnight in 4% paraformaldehyde, dehydrated in ethanol, embedded, and sectioned at 6 μm. Sections were boiled for 20 min in antigen retrieval solution (Dako), blocked for 5 min in peroxidase quenching buffer (K4011; Dako), incubated in anti-Blimp-1 (1:1,000, rabbit polyclonal) (22) at 4°C overnight, washed in phosphate-buffered saline, developed using 3,3′-diaminobenzidine and the Dako peroxidase-labeled polymer kit, and then counterstained with hematoxylin. Visualization of primordial germ cells by staining for alkaline phosphatase activity was performed as described previously (39). For IgA staining, sections of intestine were submerged in optimal-cutting-temperature freezing compound and frozen in a dry-ice isopentane bath. Blocks were cryosectioned at 6 μm and stained with goat anti-mouse IgA-HRP (1:200; 1040-05; Southern Biotech).
Testes and ovaries were fixed overnight in Bouin's fixative, dehydrated in ethanol, embedded in paraffin wax, sectioned at 8 μm, and stained with hematoxylin and eosin. For skeletal staining, limbs were skinned, fixed in 95% ethanol, stained with alcian blue, cleared with 1% potassium hydroxide, stained with alizarin red, cleared again in 1% potassium hydroxide, and equilibrated in 100% glycerol as described previously (53).
Wild-type and p50/p65 doubly deficient 3T3 fibroblasts (63) were maintained in Dulbecco's modified Eagle medium with 10% fetal bovine serum and gentamicin (10 μg/ml). Cells were split the day before infection, cultured overnight to achieve 90% confluence, and rinsed twice with phosphate-buffered saline before addition of serum-free Dulbecco's modified Eagle medium. Following a 2-h incubation with Sendai virus (ATCC, Rockville, MD) at a multiplicity of infection of 2, medium was aspirated and cells were cultured with complete medium and at the appropriate times postinfection directly lysed in Trizol (Invitrogen, Carlsbad, CA). M12 B cells (32) were cultured in RPMI 1640 with 10% fetal bovine serum, gentamicin (10 μg/ml), and where appropriate LPS (2.5 μg/ml). The NF-κB inhibitor helenalin (10 μM; Biomol, Plymouth Meeting, PA) or BMS341380 (30 uM) (Bristol-Myers Squibb, Princeton, NJ) was added for 1 h. RNA was isolated using TRIzol reagent (Life Technologies), and cDNA was prepared using SuperScript III reverse transcriptase (Invitrogen) according to the manufacturer's protocol.
Quantitative PCR was performed using an ABI 7700 instrument (Applied Biosystems). Concentrations for stock reagents are as follows: 1× PCR buffer, 200 mM deoxynucleoside triphosphate, 0.4× SYBR green (Sigma), 150 nM 6-carboxy-x-rhodamine (Sigma), 1% dimethyl sulfoxide, and 1.25 U Taq polymerase. Conditions and primer concentrations used were as follows: mouse Prdm1, 500 nM 5′-GACGGGGGTACTTCTGTTCA-3′ and 50 nM 5′-GGCATTCTTGGGAACTGTGT-3′; 2.5 mM MgCl2; mouse beta-2-microglobulin, 300 nM 5′-AGACTGATACATACGCCTGCAG-3′ and 50 nM 5′-GCAGGTTCAAATGAATCTTCAG-3′. The amplification program for beta-2-microglobulin was as follows: 95°C for 5 min, 95°C for 20 s, 59°C for 1 min, and 82 to 84°C for 20s (collect data) for 40 cycles; melting curve, 95°C for 20 s, 59°C for 15 s, and up to 95°C for 20 s with a 19-min ramping time. For mouse Prdm1, the amplification step was as follows: 95°C for 15 s, 60°C for 30 s, and 72°C for 30 s (collect data) for 40 cycles.
Chromatin immunoprecipitation (ChIP) quantified by semiquantitative PCR and Southern blotting was performed as described previously (25) using anti-p65 (sc-372; SantaCruz Biotechnology, Santa Cruz, CA). The PCR primers were designed to amplify the following regions: the potential NF-κB binding site located at −94 to −85 relative to the transcriptional start site of Prdm1 exon 1a promoter. The Igκ intronic enhancer encompasses the previously described NF-κB binding site (72). The Bcl-6 binding site within intron 5 of mouse Prdm1 (77) was amplified as a negative control. Quantitative PCR was performed as previously described (48) with FastStart SYBR green master mix (Roche) on a Stratagene MX3000 real-time PCR system.
The TATA-less GC-rich promoter region upstream of exon 1A contains multiple transcription initiation sites (51, 78). Early experiments suggested that the in-frame translational start site present in mouse exon 1A together with exon 2 encodes an N-terminal extension (78, 79). However, exon 2 sequences are not found in human PRDM1 transcripts (23, 30, 78). We also noticed that exon 2 is not present in several mouse GenBank clones that splice directly from exon 1A to exon 3. To further characterize Prdm1 transcripts expressed during mouse development, we performed 5′ RACE and sequenced the products recovered from E9.5 embryos and yolk sacs. Interestingly, the majority of clones (seven of nine) isolated from yolk sacs contain an alternative first exon (exon 1B) located approximately 70 kb upstream of exon 1A (Fig. 1A to C). Consistent with this observation, RPA experiments demonstrate that exon 1B transcripts are selectively expressed in the yolk sac and barely detectable in the embryo (Fig. (Fig.1E).1E). We found a single GenBank Prdm1 clone (accession no. AK077622) derived from E8.0-stage embryos that contains exon 1B spliced to exon 3 (29). Moreover, cap analysis gene expression (CAGE) tags indicative of transcription start sites (35, 75) have been mapped to the region immediately upstream of exon 1B (Fig. (Fig.1D).1D). Additionally, recent reports describe bivalent chromatin domains containing both transcriptionally active (H3K4 trimethyl) and repressed (H3K27 trimethyl) histone modifications at promoters of developmentally regulated genes (3, 6). We used the UCSC genome browser (37) together with ChIP-sequencing data (36, 50) (http://www.broad.mit.edu/node/681) to visualize histone modifications across the Prdm1 locus. Interestingly, exon 1B is bivalent in both human and mouse ES cells (Fig. (Fig.1F1F).
Neither exon 1A nor exon 1B 5′ RACE clones contain exon 2 sequences. Rather, both exons splice directly to exon 3. To further investigate exon 2 expression, we performed RT-PCR analysis using primers anchored in exon 1A and exon 3. We screened a panel of tissues and cell lines including the BCL1 cell line originally used for cloning Prdm1 (79). As expected, exon 2 expression is detectable in cells transfected with the full-length cDNA containing exon 2 (79). However, endogenous transcripts exclusively contain exon 1A sequences joined directly to exon 3 (Fig. (Fig.1G).1G). Four GenBank expressed sequence tag (EST) clones contain exon 1A spliced to exon 3 (accession no. AK133503, AK149344, BC129801, and CX733088), whereas exon 2 is present only in the original Prdm1 clone (accession no. U08185) (79).
One possibility is that alternative promoters govern cell-type-specific patterns of Prdm1 expression. Selective loss of alternative transcripts could potentially cause tissue-specific developmental defects. To test this possibility, we engineered targeted deletions designed to specifically eliminate either exon 1A or 1B transcripts (Fig. (Fig.2).2). The exon 1A deletion (Δex1A) removes 2.18 kb (UCSC genome browser coordinates, chromosome 10: 44,178,130 to 44,180,309; July 2007 assembly, mm9) encompassing roughly −1.8 kb to +380 bp, including the basal promoter and transcriptional start sites (78) (Fig. (Fig.2A).2A). To ensure efficient splicing of exon 1B to exon 3, sequences 3′ to exon 1A were left largely intact (~150 bp was removed). The exon 1B deletion (Δex1B) removes 3.9 kb (chromosome 10: 44,247,021 to 44,250,926), approximately −2.6 kb to +1.3 kb relative to the transcriptional start site on clone AK077622 (Fig. (Fig.2D2D).
Because Blimp-1 requirements in the embryo are exquisitely dose dependent (60, 67, 81), we expected that the Δex1A or Δex1B deletion alleles would cause germ cell defects or embryonic lethality or perturb other Blimp-1-dependent developmental processes, such as forelimb and heart development. Remarkably, both Prdm1Δex1A/Δex1A and Prdm1Δex1B/Δex1B homozygous mutants (hereafter referred to as Δex1A and Δex1B mice) were recovered at Mendelian ratios from heterozygous intercross matings (Table (Table2).2). Adult homozygous mutants failed to display any gross abnormalities. Moreover, these mice are fertile and can be maintained as homozygous mutant breeding pairs.
To confirm the Δex1A and Δex1B targeted deletions selectively eliminate mRNA expression of the alternative transcripts as predicted, we analyzed E9.5 embryo and yolk sac mRNA by RT-PCR (Fig. (Fig.3A).3A). As expected, the Δex1A mutants entirely lack exon 1A but express exon 1B transcripts. Conversely, Δex1B mutants express exon 1A transcripts but lack exon 1B mRNA. Thus, our targeting strategies selectively eliminate expression of the alternative transcripts as intended.
As shown in Fig. Fig.3,3, RPA experiments demonstrated that the Δex1A and Δex1B deletion alleles modestly reduce total Prdm1 expression levels in the embryo and yolk sac, respectively. Thus, in Δex1A embryos, total Prdm1 mRNA levels detectable with the exon 4-5 probe are reduced by approximately 50% (Fig. (Fig.3B).3B). Similarly, in Δex1B yolk sac samples, alternative exon 1A transcripts are slightly upregulated, but total Prdm1 levels are reduced by approximately 60% (Fig. (Fig.3C3C).
Next, we examined Blimp-1 protein expression via Western blot analysis. At E9.5, Blimp-1 expression by Δex1B mice is indistinguishable from wild-type expression whereas Δex1A embryos contain slightly reduced levels of total Blimp-1 protein (Fig. (Fig.3D).3D). However, alkaline phosphatase staining revealed only a slight decrease in the numbers of primordial germ cells (Fig. (Fig.3F).3F). Similarly, we observed normal patterns of Blimp-1 expression in Δex1A E9.5 placentae (Fig. (Fig.3E).3E). In comparison, Prdm1+/− heterozygous embryos express reduced levels of the Blimp-1 protein (Fig. (Fig.3D)3D) and have roughly half the normal number of migrating primordial germ cells at the early headfold stage but otherwise fail to exhibit any detectable developmental abnormalities (60, 81).
To explore possible gene dosage effects, we crossed Δex1A and Δex1B homozygous mutants with Prdm1+/null mice (74). We reasoned that further reducing Blimp-1 expression levels in compound heterozygotes could potentially reveal developmental defects. As expected, null/Δex1B compound heterozygotes were born at Mendelian ratios (Table (Table2),2), exhibited no visible abnormalities, and were fertile. In contrast, null/Δex1A animals were underrepresented at weaning. Homozygous null embryos fail to survive beyond E10.5 (60, 81). The null/Δex1A embryos were present at Mendelian ratios at E10.5 but underrepresented beginning at E14.5 (Table (Table3).3). We observed a spectrum of developmental abnormalities similar to those described previously for Sox2-Cre rescued and Blimp-1gfp/gfp embryos (67). However, in contrast, here we recovered substantial numbers of live-born null/Δex1A animals. Interestingly, many of the surviving compound heterozygotes displayed partially penetrant phenotypic disturbances, including germ cell defects (12 of 13 mice) in both males and females (Fig. 4C to E) and a rudimentary or missing fifth digit of the forelimb (6 of 13 mice) (Fig. 4A and B).
Mice lacking Prdm1 expression in B cells display defects in antibody production, but these mice are otherwise healthy and fertile (26, 27, 74). In contrast, conditional loss in T cells results in diarrhea, weight loss, and fatal colitis (28, 49). Our Δex1A and Δex1B mice, maintained under specific-pathogen-free conditions, fail to display any overt signs of disease.
To evaluate the possibility that alternative promoter usage regulates Prdm1 activities in B cells, we examined Δex1A and Δex1B spleen cells treated ex vivo under conditions that promote terminal B-cell differentiation. Strikingly, RPA experiments demonstrate that LPS-treated Δex1B splenocytes are indistinguishable from wild-type controls whereas, in contrast, Δex1A mutants show a greater than 95% reduction in total mRNA expression levels (Fig. (Fig.5A).5A). The mutation results in a loss of exon 1A transcripts and also eliminates alternative transcripts detectable with downstream probes spanning exons 3, 4-5, and 6 (Fig. (Fig.5A).5A). Blimp-1 protein levels were also dramatically reduced (Fig. (Fig.5B).5B). As predicted for loss of Prdm1 function in B cells (26, 27, 74), Δex1A LPS-stimulated splenocytes also display defective (~90% reduced) secreted IgM production (Fig. (Fig.5C).5C). The Δex1A deletion also results in markedly reduced levels of mucosal IgA expression on intestinal epithelial cells (Fig. (Fig.5D)5D) (26). Thus, we conclude that the Δex1A deletion eliminates Prdm1 function in the B-cell lineage.
Next, we tested Prdm1 expression in T lymphocytes and dendritic cells. RPA experiments demonstrate that Δex1A mutant T cells lack exon 1A transcripts and express marginally reduced levels of total Prdm1 mRNA (Fig. (Fig.6A).6A). As shown in Fig. Fig.6B,6B, Western blots similarly demonstrate protein expression is downregulated in Δex1A mutant T cells. We also observe reduced mRNA (Fig. (Fig.6C)6C) and protein (Fig. (Fig.6D)6D) expression by Δex1A BMDCs. In contrast, Δex1B mutant B-, T-, and dendritic-cell populations lack exon 1B transcripts, but total Prdm1 expression levels are indistinguishable from wild-type levels (Fig. (Fig.55 and and66).
Results above demonstrate upregulated expression of alternative transcripts in Δex1A embryos and Δex1B yolk sacs (Fig. 3B and C). Similarly, Δex1A BMDCs also display elevated expression of alternative exon 1B transcripts (Fig. 7G and H). However, exon 3 expression detectable with an exon 1B-3 probe was only marginally reduced in Δex1A embryos, suggesting that the Prdm1 gene has additional alternative transcriptional start sites (Fig. (Fig.7A).7A). To examine this possibility, we performed 5′ RACE on E9.5 Δex1A mutant embryos. A novel 5′ exon (exon 1C), located in the intron between exon 1A and exon 3 approximately 1.2 kb downstream of exon 1A, was cloned (Fig. (Fig.7B,7B, E, and F). RT-PCR and RPA experiments confirmed that exon 1C transcripts are spliced to exon 3 (Fig. 7C and G to K). We also detected 1C transcripts that retain intronic sequences (Fig. (Fig.7D).7D). Exon 1C expression is not simply due to aberrant activation of a cryptic promoter caused by the Δex1A deletion, because exon 1C transcripts are present in wild-type LPS-treated spleen cells, activated T cells, and embryonic tissues (Fig. 7C and D). Moreover, the region upstream of exon 1C contains numerous CAGE tags (35, 75) and has been identified as an ancient noncoding element conserved between human and elephant shark (80) (Fig. (Fig.7E7E).
To learn more about the possible underlying mechanism(s) responsible for selective loss of Prdm1 expression caused by the exon 1A deletion in plasma cells, we searched for candidate transcription factor binding sites mapped within this 2.1-kb genomic region. Additionally, we compared conserved binding motifs located near the alternative upstream exon 1B, as well as those located close to the alternative exon 1C transcriptional start site. As shown in Fig. 8A and B, the region surrounding exon 1B displays greater diversity than relatively well-conserved sequences located adjacent to the proximal exons 1A and 1C. Indeed, we found multiple transcription factor binding sites tightly clustered together immediately upstream of exon 1A. Besides previously described c-Fos/AP-1 (62) and Pax 5 (41) binding sites, strikingly conserved NF-κB binding sites are present within the exon 1A deletion. NF-κB transcription factors are known to function downstream of a wide variety of inflammatory stimuli (54, 64, 65). Blimp-1/Prdm1 was initially cloned as a negative regulator of IFNB1 (beta interferon) expression in virally infected cells (30).
To explore this possible relationship and directly test whether NF-κB is required for activation of Prdm1 expression, we examined wild-type and p50/p65 doubly deficient 3T3 fibroblasts (63) infected with Sendai virus. Results shown in Fig. Fig.9A9A demonstrate that fibroblasts lacking the NF-κB subunits p50 and p65 fail to express Prdm1 in response to Sendai virus infection. As in the case of endoplasmic reticulum stress (16), we also found here that Prdm1 induction in response to Sendai virus infection is insensitive to cycloheximide treatment, demonstrating a direct requirement for NF-κB independent of new protein synthesis (data not shown). Similarly, as shown in Fig. Fig.9B,9B, treatment with NF-κB inhibitors prevents Prdm1 induction in response to LPS in B-cell lines. Helenalin, which selectively alkylates p65/RelA (46), or BMS341380, which selectively inhibits IKKβ phosphorylation, both gave indistinguishable results. Thus, we conclude that NF-κB signaling is required for induction of Prdm1 in response to LPS stimulation. Finally, to demonstrate that NF-κB binds to the region immediately upstream of exon 1A, we performed ChIP experiments. As shown in Fig. 9C and D, significant levels of p65/RelA binding were detectable in LPS-treated M12 B-lymphoma cells. These results strongly suggest that occupancy of the NF-κB binding sites upstream of the exon 1A promoter is required for LPS-inducible Prdm1 expression.
Alternative promoter usage is a prevalent feature of mammalian genome architecture and evolution (4, 8, 34). Alternative promoters located in different genomic regions are often responsible for governing tissue-specific patterns of expression. Alternative first exons may also introduce sequence substitutions that change protein structure and/or influence mRNA stability (13, 38, 69). As many as 58% of mouse genes have alternative promoters (8, 34), but only a handful have been functionally characterized by targeted mutagenesis. Here we demonstrate for the first time that the murine Prdm1 gene uses alternative promoters that share overlapping activities during early development, whereas the previously characterized promoter region selectively functions to drive expression in plasma cells.
Our 5′ RACE experiments have identified an alternative promoter located 70 kb upstream of exon 1A. Thus, the Prdm1 transcription unit spans a much greater genomic distance than previously realized (~91 kb versus ~23 kb). Exon 1B transcripts are strongly expressed in the yolk sac. This alternative promoter also bears bivalent chromatin modifications in embryonic stem cells. However, our gene targeting experiments unequivocally demonstrate that this alternative first exon is dispensable for normal development. Thus, mice carrying the Δex1B targeted deletion express wild-type levels of Prdm1. Even in the context of a compound heterozygote, we fail to detect any evidence for developmental defects. The Prdm1 locus is adjacent to a gene desert (57) devoid of known protein-coding information, with the nearest annotated gene, Prep, located approximately 600 kb upstream. Interestingly, the region downstream of exon 1B is more highly conserved than upstream sequences and in all likelihood contains tissue-specific regulatory elements that act at a distance to govern Prdm1 gene expression.
We have demonstrated that exon 2, present in the original Prdm1 clone isolated by Turner and coworkers (79), is not normally expressed. Similarly, GenBank EST clones contain exon 1A spliced directly to exon 3. Our analysis of Δex1A deletion mice establishes that the AUG translational start codon present in exon 1A is not required for protein expression. The cluster of in-frame AUG codons in exon 3 is conserved across vertebrates (23, 78). During the course of this study, we discovered an additional first exon (exon 1C) located in the intron between exon 1A and exon 3. Besides exon 1C transcripts spliced to exon 3, additionally we observe exon 1C transcripts that contain intervening intronic sequences. These alternative transcripts are normally expressed at low levels in diverse cell types, including T cells, B cells, and dendritic cells. Given its evolutionary conservation, it is tempting to speculate that exon 1C may represent an ancient promoter region that has been superseded by the more distal exon 1A promoter. Interestingly, exon 1C transcripts contain two additional upstream in-frame AUG codons. Future experiments will evaluate whether these alternate translational start sites may contribute functional diversity.
The revised picture of the Prdm1 gene structure should prove useful for designing experiments aimed at mapping cis-acting regulatory elements and comparative studies of the 16 Prdm family members encoded in the mouse genome (17). A closely related family member, Prdm14, is also activated by BMP-Smad signals in prospective primordial germ cells and plays an essential role in establishment of the germ cell lineage (84). It will be interesting to compare Prdm1 and Prdm14 cis-regulatory sequences controlling gene expression and to engineer knock-in alleles that swap coding information to test for functional redundancy.
The Δex1A deletion eliminates Prdm1 expression, required for terminal B-cell differentiation into antibody-secreting cells. The simplest explanation is that this discrete 2.18-kb genomic region contains essential sequences normally bound by key transcription factors upstream of Prdm1. Consistent with this notion, c-Fos/AP-1 binding sites have been mapped to a fragment containing approximately kb −1.3 to −1.0 relative to the exon 1A transcription start site (62). Moreover, c-Fos is known to induce Prdm1 expression in B cells (62). However, c-Fos-deficient B cells still undergo plasmacytic differentiation in response to LPS (62). Thus, c-Fos binding is not essential for Prdm1 expression in B cells.
Similarly, previous studies demonstrated that Prdm1 expression is dramatically upregulated in response to viral infection, endoplasmic reticulum stress, cytokine signaling, and inflammatory stimuli (54, 64, 65) and the NF-κB signaling pathway plays a key role in B-cell development (11, 19, 64). The present experiments demonstrate for the first time that NF-κB signaling plays a key role in Prdm1 induction. Thus, NF-κB inhibitors block Prdm1 induction in B cells, and fibroblasts lacking the NF-κB subunits p50 and p65 fail to express Prdm1 in response to Sendai virus infection. ChIP experiments demonstrate LPS-inducible occupancy of a conserved NF-κB site upstream of the proximal exon 1A promoter in B cells. The exon 1A deletion encompasses these c-Fos/AP-1 and NF-κB binding sites as well as conserved Stat sites (15) and may therefore eliminate essential signals required for activation of Prdm1 expression.
The exon 1A deletion also removes two highly conserved consensus CTCF binding motifs. The CTCF protein has 11 zinc fingers that display nearly 100% amino acid sequence identity shared among vertebrates (66). Recent genome-wide mapping studies demonstrate that roughly half of the CTCF-binding sites mapped far away from genes, consistent with a potential role for these sequences as insulators (33). Only about 20% of CTCF sites are located near transcriptional start sites, and interestingly, as for Prdm1, a common characteristic shared by many of these genes is alternative promoter usage. The CTCF sites upstream of exon 1A are probably involved in organizing higher-order chromatin structure that controls developmentally regulated Prdm1 gene expression.
Another well-known Blimp-1 direct target is Pax5. Pax5 is required to establish and maintain B-cell lineage identity (12, 58), and elimination of Pax5 expression is an essential prerequisite for terminal differentiation (41). Mutually exclusive Prdm1 and Pax5 expression involves an autoregulatory feedback loop (41, 52, 71). The conserved Pax5 binding site is located inside the Δex1A deletion (52). Repression of Prdm1 expression also depends on a downstream Bcl-6 binding site (73, 77) and a Bach-2 binding site located near the exon 1A promoter (59). Thus, another, not mutually exclusive interpretation is that the Δex1A deletion causes a change in chromatin architecture and shifts the dynamic positioning of nucleosomes along the locus, which, selectively in the B-cell lineage, leads to sustained silencing and long-term occupancy by these Prdm1 transcriptional repressors.
In striking contrast to B cells, Δex1A T cells and BMDCs retain moderate levels of expression. Perhaps assembly of silent chromatin is less tightly controlled in these cell lineages. Alternatively, Prdm1 expression in these cell types may be governed by nonoverlapping, as yet ill-defined cis-acting regulatory elements. Consistent with this suggestion, we note with interest the two conserved NFAT binding motifs located immediately upstream of exon 1C. Members of the NFAT family of transcriptional factors associate with different partners to regulate key aspects of T-helper-cell differentiation (47). In T lymphocytes, Prdm1 expression is induced by inflammatory cytokines and in response to receptor signaling (20, 28, 49, 70). Prdm1 regulates T-cell proliferation, survival, homeostasis, and terminal differentiation (28, 49). However, expression does not seem to be restricted to a discrete subset. It will be important to learn how these NFAT binding sites mapped upstream of exon 1C may regulate Prdm1 expression levels and influence T-cell development and function. Prdm1 is induced during macrophage differentiation (9). The present experiments demonstrate for the first time that Prdm1 expression is dramatically upregulated during dendritic-cell maturation. It will be interesting to characterize functional activities of T-lymphocyte and dendritic-cell subsets carrying the Δex1A hypomorphic allele.
The Δex1A deletion selectively eliminates expression in plasma cells but only slightly decreases expression in the embryo. In the absence of the basal promoter, alternative promoters compensate and rescue all aspects of embryonic development. Compound heterozygotes also carrying the null allele with further reduced expression levels display a broad spectrum of developmental defects reflecting a generic loss of Prdm1 activities rather than inactivation within any particular expression domain. Thus, in the embryo, Prdm1 alternative promoters seem to regulate overall expression levels as opposed to governing tissue-specific expression patterns. Developmentally regulated Prdm1 expression in the embryo is known to be governed by BMP-Smad signaling cues (60), whereas in contrast, Prdm1 expression in different B-cell subpopulations responds to Toll-like receptor-NF-κB pathways (18, 42). Quite different regulatory cues are likely to control Prdm1 functional activities in diverse cell types.
Interestingly, in the skin, Prdm1 mRNA expression was increased in the absence of functional Blimp-1 (48). These results provide evidence for repression via an autoregulatory mechanism. Consistent with this idea, conserved Blimp-1 binding sites were characterized within intron 2 (48). Decreased Blimp-1 expression levels may therefore inactivate this negative autoregulatory feedback circuit and allow normally silent alternative promoters to become activated. Our Δex1A and Δex1B mutant strains may be valuable tools for studying the structural basis of these divergent regulatory inputs.
Prdm1 homologs have been found in many metazoans (17) and play important roles in zebra fish (5, 68), Xenopus (14), Drosophila (1, 55), and sea urchin development (44, 45). Even between Caenorhabditis elegans and humans, the coding sequence of the PR/SET domain and zinc fingers is relatively well conserved (78). However, Prdm1 expression patterns and functional activities show striking species differences. For instance, in mice Prdm1 is essential for placental development and germ cell specification (60, 81). In contrast, in zebra fish Prdm1 controls muscle cell fate (5, 68). Evidence to date strongly argues that these roles are not conserved across vertebrates. Thus, Prdm1 represents an early metazoan gene, having gained novel expression domains and diverse functions through the course of animal evolution.
Interestingly, Prdm1 first exon usage varies widely among vertebrates. A survey of Xenopus tropicalis GenBank ESTs using the UCSC genome browser shows a single clone initiated at the genomic region homologous to mouse exon 1A (accession no. CR414136). There are four Xenopus clones (accession no. CF784176, CR432771, CR417246, and AL649398) that initiate at an alternative first exon approximately 4 kb upstream of Xenopus exon 1, corresponding to a region approximately 8.5 kb upstream of exon 1A in mouse. The zebra fish ESTs all start at the same first exon. However, this promoter region shares no homology with sequences in the mouse genome. Similarly, alternative promoter regions found in sea urchin (44, 45) share no homology with mouse alternative exon 1A, 1B, or 1C. Considerable evidence demonstrates that motif-specific enhancer-promoter interactions regulate gene expression patterns (7). It seems likely that evolutionary changes in Prdm1 expression patterns involve not only changes in enhancer sequences but also changes in the promoter motifs with which they interact. Thus, the emergence of alternative promoters may mark the beginnings of new functions for this old gene.
We thank Ayesha Islam and Stephane Vincent for initial characterization of exon 1B transcripts, Chad Koonce for assistance with gene targeting in ES cells, and Carol Paterson and Emily Lejsek for blastocyst injections and genotyping assistance. We thank Reuben Tooze and Lynn Corcoran for generously providing rabbit polyclonal and rat monoclonal Blimp-1 antibodies, respectively.
This work was supported by a program grant from the Wellcome Trust.
Published ahead of print on 8 September 2009.