Targeted Disruption of the CP2 Gene, a Member of the NTF Family of Transcription Factors*

The NTF-like family of transcription factors have been implicated in developmental regulation in organisms as diverse asDrosophila and man. The two mammalian members of this family, CP2 (LBP-1c/LSF) and LBP-1a (NF2d9), are highly related proteins sharing an overall amino acid identity of 72%. CP2, the best characterized of these factors, is a ubiquitously expressed 66-kDa protein that binds the regulatory regions of many diverse genes. Consequently, a role for CP2 has been proposed in globin gene expression, T-cell responses to mitogenic stimulation, and several other cellular processes. To elucidate the in vivo role of CP2, we have generated mice nullizygous for the CP2 allele. These animals were born in a normal Mendelian distribution and displayed no defects in growth, behavior, fertility, or development. Specifically, no perturbation of hematopoietic differentiation, globin gene expression, or immunological responses to T- and B-cell mitogenic stimulation was observed. RNA and protein analysis confirmed that the nullizygous mice expressed no full-length or truncated version of CP2. Electrophoretic mobility shift assays with nuclear extracts from multiple tissues demonstrated loss of CP2 DNA binding activity in the −/− lines. However, a slower migrating complex that was ablated with antiserum to NF2d9, the murine homologue of LBP-1a, was observed with these extracts. Furthermore, we demonstrate that recombinant LBP-1a can bind to known CP2 consensus sites and form protein complexes with previously defined heteromeric partners of CP2. These results suggest that LBP-1a/NF2d9 may compensate for loss of CP2 expression in vivo and that further analysis of the role of the NTF family of proteins requires the targeting of the NF2d9 gene.

Cellular diversity is generated by unique combinations of transcription factors interacting to specify patterns of gene expression in different cell types. This is particularly evident during development where biochemical and genetic studies have identified numerous proteins essential for the processes that govern embryogenesis. Many of these factors are highly conserved in evolution, playing critical roles in organisms as diverse as Drosophila and man. One family of transcription factors that typifies these principles is the NTF-like group of proteins. The founding member of this family is the developmentally programmed Drosophila factor, NTF-1 (neurogenic element binding transcription factor) (1). NTF-1 (also known as Grainyhead or Elf-1) was first identified through its ability to bind a cis element critical for expression of the Dopa decarboxylase gene (2,3). Subsequently, NTF-1 was shown to bind to promoters of other developmentally regulated genes including Ultrabithorax, fushi tarazu, and engrailed (2). NTF-1 has also been linked to dorsal/ventral and terminal patterning through the formation of multiprotein complexes that influence transcription from the decapentaplegic and tailless genes (4,5). More recently, tissue-specific isoforms of the protein have been described in Drosophila, and mutation of these isoforms or the ubiquitously expressed gene results in pupal lethality with gross developmental defects (1,6).
In mammals, two highly related NTF-like genes have been identified. In humans they are known as LBP-1a and CP2 (LBP-1c/LSF), whereas the mouse homologues are referred to as NF2d9 and CP2, respectively (7)(8)(9)(10). The human genes are 72% identical in overall amino acid sequence but share higher sequence identity (88%) in the N-terminal halves of the proteins than the C-terminal halves (52%). The homology with the NTF gene is also in the N-terminal region with three amino acid stretches, 148 to 159, 205 to 216, and 233 to 246 showing 66, 75, and 79% identity, respectively. The NTF-like gene family has been shown to have a variety of cellular and developmental functions in human and murine cells. The best characterized member of the family, CP2, was initially identified as a factor that binds to, and stimulates transcription from, the murine ␥-fibrinogen promoter and the viral SV40 major late promoter (11,12). Binding sites for CP2 have also been defined in regulatory regions of the human immunodeficiency virus (HIV) 1 where it acts in concert with YY1 to repress transcription (9,(13)(14)(15)(16). In the context of non-viral gene regulation, CP2 has been shown to bind homomerically to the human c-fos, ornithine decarboxylase, c-myc, and DNA polymerase promoters and the murine ␣-globin and fibrinogen promoters and activate transcription in vitro (17,18). Binding to the regulatory elements in the fos and ornithine decarboxylase promoters is modulated by cell growth signals. Mitogenic stimulation of resting T-cells is associated with rapid phosphorylation of CP2 by the mitogen-activated protein kinase pp44 (extracellular signal-regulated kinase 1) and a consequent increase in its DNA binding activity (17). This modulation suggests that CP2 contributes to the regulation of early response genes and therefore plays a role as a cell growth regulator.
A developmental role for CP2 has been identified in studies of globin gene regulation. In this context, CP2 binds to the stage selector element in the proximal ␥-gene promoter as a heteromeric complex with a recently cloned fetal/erythroidspecific partner protein, NF-E4 (19). 2 This complex, known as the stage selector protein (SSP), contributes to the preferential recruitment of the ␤-globin locus control region to the ␥-promoter during fetal erythropoiesis (20,21). SSP binding sites have also been defined in the ⑀-promoter and in the regions of DNase1 hypersensitivity that constitute the locus control region (19,22,23).
Despite the extensive literature examining CP2 function in vitro and in cell lines, the in vivo role of this factor remains unknown. To address this, we have generated a CP2 null mutation in mice by homologous recombination. Mice lacking CP2 expression were examined for defects in growth and development, with a particular emphasis on hematopoiesis, immune, and neural function. We observed no significant abnormality in CP2 Ϫ/Ϫ mice compared with wild type littermates. We have shown through DNA binding and protein-protein interaction studies that the lack of a discernible phenotype may be due to a complete rescue by NF2d9, the murine homologue of LBP1a.

EXPERIMENTAL PROCEDURES
Generation of CP2 Ϫ/Ϫ Mice-We isolated eight CP2 genomic clones by screening a 129-derived ES cell phage library with a full-length mouse CP2 cDNA probe. Duplicate lifts screened with a probe specific for exon two, containing the initiation ATG, identified one clone with a 12-kb insert that encodes the first two exons and the 5Ј untranslated region. Detailed restriction endonuclease mapping of this fragment confirmed the previously reported genomic structure of murine CP2 with the exception of an EcoRI polymorphism detected in the 5Ј untranslated region (see Fig. 1A) (24). Subsequently, a 6.6-kb NcoI fragment containing the 5Ј untranslated region and a portion of the first exon was subcloned into pSL301. A XhoI-SalI fragment containing a phosphoglycerate kinase promoter-regulated hygromycin resistance expression cassette was cloned into a downstream SalI site. Finally, a 3.4-kb SalI-NotI fragment containing 2.4 kb of the second intron of CP2 and a 1-kb HSV-TK expression cassette fragment (a kind gift of Dr. J. van-Deursen) was cloned downstream of this region to provide 3Ј homology and a negative selectable marker. This construct, pK01HygTK, was linearized with NotI and transfected by electroporation into RW8 embryonic stem cells (Genomic Systems Inc). The cells were cultured on primary irradiated embryonic STO feeder cells in the presence of 140 fg/ml hygromycin and 0.2 fM FIAU. Resistant clones were screened by Southern blot analysis using a unique 0.5-kb SalI-NcoI fragment located 5Ј to the targeted sequence. Four karyotypically normal CP2 ϩ/Ϫ clones were microinjected into C57BL/6 blastocysts, of which three clones were transmitted through the germ line.
In Situ Hybridization Studies-Embryo sections were prepared as described previously (25). Briefly, C57/BL6 mice, overdosed with ketamine and rhompin, were perfused intracardially with paraformaldehyde, and the embryos from timed matings were postfixed in a similar solution. Sections of 10 -14 M were cut with a cryostat, mounted on glass slides, and stored at Ϫ20°C. These slides were subsequently probed with sense and antisense riboprobes generated by [ 33 P]UTP labeling from Bluescript plasmids encoding the complete cDNAs of CP2 and NF2d9 (the latter were a kind gift of Dr. M. Negishi) (10). Specific signals were developed by dipping the slides in NTB2 emulsion (Kodak Scientific Imaging Systems) and exposed at 4°C for two weeks. The sections were counterstained using 0.1% toluidine blue in distilled water and analyzed by phase-contrast microscopy.
Phenotypic Analysis-Tissues from normal and age-matched het-erozygote and homozygote knockout mice were removed and fixed in formalin, and embedded paraffin sections were prepared. These sections were stained with hematoxylin and eosin and examined by light microscopy. Peripheral blood (150 l) was obtained by retro-orbital puncture and blood cell counts, and erythrocyte parameters were determined utilizing an automated analyzer (Coulter). In addition an aliquot was stained with Wright's Giemsa or methylene blue to study hematopoietic cell morphology and reticulocytes, respectively. Bone marrow hematopoietic progenitors were cultured in methylcellulose in the presence of IL3, erythropoeitin, and stem cell factor (Terry Fox Laboratories, Vancouver, Canada). For immunological studies, single cell suspensions were prepared from spleen, lymph node, thymus, and bone marrow and stained with cell type-specific markers for granulocytes (Gr1), T-cells (CD8 and CD4), B-cells (B220 and IA b ), and NK cells (NK1.1). Fluorescence analysis was performed utilizing a FACScan flow cytometer (Becton Dickinson, Mountain View, CA). Similar cellular suspensions were utilized to assess the proliferative potential of T-and B-cells, culturing 10 5 cells in the presence of anti-CD3⑀ ϩ/Ϫ phorbol 12-myristate 13-acetate, concanavalin A, lipopolysaccharide, phytohemagluttinin, ionomycin, or a combination of phorbol 12-myristate 13acetate, phytohemagluttinin, and ionomycin for 48 h as described previously (17). Analysis of Gene Expression by Ribonuclease Protection Assays (RPA)-RNA was prepared from various tissues and from peripheral blood and bone marrow cells using RNAzol B or RNASTAT 60. Murine CP2 and NF2d9 cDNA fragments spanning exons 2 and 3 and a portion of exon 4 were subcloned into pSP73. RPA studies were performed using the Ambion RPAII kit as per the manufacturer's instructions. For studies of murine and transgenic human globin gene expression RPA was performed utilizing probes specific for the murine , ␣, ⑀y, ␤h1, and ␤ major transcripts and the human ⑀-, G ␥-, and ␤-globin transcripts (a kind gift of Dr. K. Gaensler).
Nuclear Extract Preparation and Electrophoretic Mobility Shift Assays (EMSA)-Crude nuclear extracts were prepared from various primary tissues (26,27) and quantitated using the Bio-Rad protein assay system as per the manufacturer's instructions. EMSAs were performed by incubating varying amounts of nuclear extract with 10 5 cpm of [ 32 P]dCTP end-labeled double stranded oligonucleotides encoding the CCAAT box region of the murine ␣-globin promoter, the ␥-fibrinogen promoter, or the SV40 major late promoter in a 20-l reaction containing 500 ng of poly[d(I⅐C)], 6 mM MgCl 2 , 16.5 mM KCl, and 100 g of bovine serum albumin (21,27). For antibody studies, 3 l of preimmune serum or rabbit anti-mouse CP2 antibody were preincubated for 10 min with the binding reaction prior to addition of the probe. Polyclonal antiserum against NF2d9 was kindly provided by Dr. M. Negishi. After incubation at 4°C for 10 min and 25°C for 20 min, samples were electrophoresed on a 4% non-denaturing polyacrylamide gel in 0.5 ϫ Tris borate-EDTA buffer for 90 min at 10 V/cm. Recombinant CP2 and NF2d9 were prepared as glutathione S-transferase fusion proteins as described previously (19).
Yeast Two Hybrid Assays-cDNA sequences encoding the C-terminal 250 amino acids (amino acids 250 -500) of CP2 and the corresponding region of LBP-1a were inserted into the yeast expression vector pGBT9. The resultant plasmids encode hybrid proteins containing the DNA binding domain of GAL4 fused to CP2 or LBP-1a residues. The yeast reporter strain, HF7C, was transformed with this vector and a second plasmid encoding a hybrid protein of the GAL4 transactivation domain and either the NF-E4 cDNA or RING1B cDNA. The yeast were plated on leucine/tryptophan/histidine plates and incubated at 30°C for 3 days (28). Protein interactions are indicated by growth on these plates. Transfection efficiency was monitored by plating of an aliquot of the transformation on leucine/tryptophan plates (data not shown).
Transgenic Mice-CP2 Ϫ/Ϫ mice were bred with mice transgenic for a single copy of a yeast artificial chromosome (YAC) containing 250 kb of the human ␤-globin locus (␤-YAC) (a kind gift of Karin Gaensler, University of California, San Francisco, CA) (29). Timed pregnancies were set up utilizing CP2 ϩ/Ϫ females and CP2 ϩ/Ϫ YAC ϩ males, where the day of plug formation was taken as 0.5 days post-coitum (E0.5 dpc). Embryos were collected on days E9.5 dpc, E10.5 dpc, E11.5 dpc, and E14.5 dpc and genotyped by standard methodologies utilizing the CP2 probes described above and a probe specific for IVSII of the human A ␥-globin gene (a kind gift of Dr. Karin Gaensler). RNA from yolk sac and fetal liver was prepared, and RPA analysis was performed as described above. 2 J. M. C. and S. M. J., submitted for publication.

RESULTS
Targeting of the CP2 Genomic Locus-To disrupt the murine CP2 gene, a targeting vector was designed that replaced the first untranslated exon and the entire second exon containing the initiation codon and the trans-activation domain with a hygromycin expression cassette (Fig. 1A). In addition, this cassette introduced termination codons in all open reading frames. RW8 embryonic stem cells were electroporated with this construct and selected in hygromycin and FIAU. Southern analysis of resistant clones demonstrated a 9.0-kb EcoRI fragment in addition to the 10.5-kb wild type allele, at a mean frequency of one in 25 clones (data not shown). Four independently targeted clones, with normal karyotypes, were injected into C57BL/6J blastocysts, three of the clones being transmitted through the germ line. Interbreeding of mice heterozygous for the CP2 allele (CP2 ϩ/Ϫ ) resulted in litters of normal size with the expected Mendelian frequency of genotypes. Of 256 total offspring tested, 72 were CP2 ϩ/ϩ (28%), 125 were CP2 ϩ/Ϫ (49%), and 60 animals (23%) were nullizygous (CP2 Ϫ/Ϫ ) for the CP2 allele.
Expression of CP2 in Homozygous CP2 Ϫ/Ϫ Animals-To confirm the loss of CP2 gene expression in nullizygous animals, RNA was prepared from various tissues of both wild type and CP2 Ϫ/Ϫ mice and analyzed by RNase protection analysis. A specific band of 380 nucleotides was observed in all tissues derived from wild type animals utilizing a riboprobe that hybridizes to exons 2-4 ( Fig. 1B, lanes 1-5). In contrast, no signal was observed utilizing RNA derived from the brain, heart, kidney, lung, and spleen of CP2 Ϫ/Ϫ mice (Fig. 1B, lanes 7-11). An actin probe controlled for the integrity of the RNA (Fig. 1C).
To confirm the loss of CP2 expression, and to rule out a cryptic splicing event that might produce a functional CP2 transcript, RNA from wild type and CP2 Ϫ/Ϫ tissues was assayed by RT polymerase chain reaction utilizing primers specific to the 3Ј end of the mRNA transcript. Although a CP2-specific signal was observed in all wild type tissues tested, no signal was detected from RNA derived from CP2 Ϫ/Ϫ animals (data not shown).
Phenotypic Analysis of CP2 Ϫ/Ϫ Animals-Male and female knockout mice grew normally and were healthy up to 18 months of age. No abnormal behavioral patterns were observed. The fertility of CP2 Ϫ/Ϫ animals was normal, and no increase in morbidity was observed when compared with littermate controls. Careful histopathological examination of brain, spleen, kidney, liver, thymus, lymph nodes, heart, skin, muscle, and bone from CP2 Ϫ/Ϫ animals, performed at 3, 9, and 15 months, was identical to wild type littermate controls (data not shown).
Examination of Hematopoiesis in CP2 Ϫ/Ϫ Animals-CP2 has been implicated in the regulation of several hematopoietic genes, particularly those of the globin loci (7,19). To determine whether loss of CP2 expression resulted in changes in hematopoiesis, the hematological parameters of CP2 Ϫ/Ϫ mice were assayed and compared with those of wild type littermates. No significant difference in total cell counts, hematocrits, reticulocytes, differential white cell counts, or the ␣-/␤-globin ratio was observed (Table I). In addition, the numbers of bone marrow progenitors, as measured by colony-forming unit activity, were similar in CP2 ϩ/ϩ and CP2 Ϫ/Ϫ animals (data not shown). Similar studies of lymphopoiesis were stimulated by recent studies implicating CP2 in the modulation of T-cell proliferative responses (17). However, extensive analysis of T, B, and NK phenotypes, as well as functional assays of B-and T-cell Homologous recombinants were identified utilizing a unique 5Ј probe (hatched box). The sizes of the wild type and disrupted alleles detected by this probe are indicated. B, RNase protection analysis of total RNA isolated from multiple tissues of CP2 ϩ/ϩ and CP2 Ϫ/Ϫ mice. Total RNA was extracted from brain (B), heart (H), kidney (K), lung (L), and spleen (S) of CP2 ϩ/ϩ and CP2 Ϫ/Ϫ animals or from the murine erythroleukemia cell line (MEL). RNA was hybridized to a mixture of CP2and actin-specific probes adjusted to equal specific activities. Protected fragments are indicated at the right. This experiment is representative of 10 animals assayed. Dashes indicate empty lanes. function, failed to identify a difference between CP2 ϩ/ϩ and CP2 Ϫ/Ϫ cells (data not shown).
It is possible that despite normal adult erythropoiesis, the loss of CP2 expression may affect either ␣or ␤-globin gene expression during hematopoietic ontogeny. CP2 was initially identified as an ␣-globin CCAAT box binding activity, suggesting a possible role in ␣-globin gene expression. However, neithernor ␣-globin gene expression was perturbed in yolk sac or fetal liver cells ( Fig. 2A). We have demonstrated that CP2 is a component of the ␥-globin promoter-binding SSP complex and suggested that the ␥ to ␤ switch in the ␤-globin subtype may be perturbed in a CP2 null environment. To test this hypothesis, we bred CP2 Ϫ/Ϫ animals with mice transgenic for a 240-kb YAC containing the human ␤-globin locus (␤YAC). Subsequently, we bred male progeny transgenic for the ␤YAC with CP2 ϩ/Ϫ females and examined the expression of both human and mouse ␤-globin-like genes at several developmental stages. As shown in Fig. 2B, both human and murine ␤-globin-like gene expression in yolk sac, fetal liver, and bone marrow were identical in CP2 ϩ/Ϫ and CP2 Ϫ/Ϫ embryos and adult mice, respectively.
DNA Binding Activity in Extracts from CP2 Ϫ/Ϫ Mice-To examine CP2 DNA binding site occupancy in the null mice, we prepared crude nuclear extracts from lung, kidney, heart, and liver of wild type and CP2Ϫ/Ϫ animals and performed EMSA using a double stranded oligonucleotide probe containing the ␣-globin CCAAT box (7). Utilizing equal amounts of protein in each lane, a band of similar electrophoretic mobility was observed in all wild type tissues (Fig. 3A, compare lanes 1, 3, 5,  and 7). In contrast, extracts from CP2 Ϫ/Ϫ tissues failed to show the band seen with wild type extract and instead showed a DNA-protein complex with a slower migration pattern (Fig. 3A, compare lanes 2, 4, 6, and 8 with 1, 3, 5, and 7, respectively). This result was not dependent on the amount of protein added, as 2-to 8-fold more protein from CP2 Ϫ/Ϫ liver extract incubated with the probe generated an identical band shift (Fig. 3A,  compare lane 7 with lanes 8 -11).
The lack of an obvious phenotype in the CP2 Ϫ/Ϫ animals coupled with the persistent DNA site occupancy observed with extracts from nullizygous tissues suggested the presence of a ubiquitous CP2-like factor that could compensate for the lack of CP2. One candidate factor was Nf2d9, the murine homologue of the human NTF-like gene, LBP-1a (10). Support for this hypothesis was obtained by studying the relative electrophoretic mobilities of CP2 and NF2d9. Both molecules bound the ␣-globin CCAAT box and ␥-fibrogen probes, the NF2d9-DNA complex having a perceptibly slower mobility ( Fig. 3B and data not shown). To determine whether the protein-DNA complex generated with CP2 Ϫ/Ϫ extracts contained NF2d9, we performed competition experiments utilizing excess concentrations of unlabeled oligonucleotides that have been previously shown to bind CP2 and/or NF2d9(7, 10). These oligonucleotides were capable of ablating both wild type and mutant binding activity (data not shown). We also investigated the ability of monoclonal antisera generated against recombinant NF2d9 to dis-rupt binding activity. This antisera does not cross-react with CP2 as assessed by immunoblotting. 3 Addition of the antibody induced a partial supershift of wild type binding activity (Fig.  3C, compare lanes 1 and 3). In contrast, mutant activity was completely supershifted (Fig. 3C, compare lanes 2 and 4). These data suggest that NF2d9 can maintain DNA site occupancy at CP2 binding sites.
Expression of CP2 and NF2d9 during Mouse Development-To determine whether the distribution of expression of CP2 and NF2d9 was similar, we performed in situ hybridization on embryo sections using antisense and sense probes specific for each mRNA transcript. Normal embryos were examined at E9.5, E11.5, and E13.5 dpc. Probes were of similar specific activity, and sense probes produced little background signal from embryos probed at all developmental stages (Fig. 4,  A and C). However, utilizing an antisense probe, we observed CP2 expression in most tissues at similar levels at E13.5 dpc (Fig. 4B). In contrast, although expression of NF2d9 was observed in all tissues, it was markedly higher in the fetal liver (Fig. 4D, arrow). It is possible that loss of CP2 expression results in up-regulation of NF2d9 expression. To test this

FIG. 2. Effects of loss of CP2 expression on globin gene expression.
A, analysis of murine ␣-like globin gene expression. Total RNA was extracted from yolk sacs (E9.5 dpc) and fetal liver (E14.5 dpc) of wild type (CP2 ϩ/Ϫ ) or CP2 Ϫ/Ϫ animals and hybridized to a mixture of murine -globin, ␣-globin, and actin riboprobes. This experiment is representative of several littermates assayed. B, analysis of murine and human ␤-globin-like gene expression in mice transgenic for the ␤-globin locus YAC. Total RNA was extracted from yolk sacs (E9.5 dpc), fetal liver (E14.5 dpc), or adult bone marrow (Adult) of heterozygote (CP2 ϩ/Ϫ ) or CP2 Ϫ/Ϫ animals transgenic for the human ␤-globin YAC and hybridized to a mixture of human ␥-globin, murine ␤H1-globin, and actin riboprobes (d9.5) or human ␥-globin, human ␤-globin, murine ␤ majorglobin, and actin riboprobes (d14.5 and adult). Protected fragments are indicated. Littermates that do not contain the ␤-globin YAC are included for comparison. This experiment is representative of several litters assayed. hypothesis, we determined the expression of NF2d9 in the brain, kidney, and heart of CP2 ϩ/ϩ and CP2 null animals. As shown in Fig. 4E, no significant change in the relative expression of NF2d9 was observed in any of these tissues when compared with the actin control.
LBP-1a Can Functionally Replace CP2 Activity-The DNA binding, transactivation, and expression patterns of LBP-1a/ NF2d9 suggested that this protein could potentially compensate for the loss of CP2. To evaluate this further, we examined whether LBP-1a could fulfill the protein-protein interaction role of CP2. We have recently defined two transcription factors FIG. 3. A, analysis of protein/DNA binding activity in nuclear extracts from CP2 ϩ/ϩ and CP2 Ϫ/Ϫ tissues. Equal amounts of protein derived from nuclear extracts from lung, kidney, heart, and liver of CP2 ϩ/ϩ and CP2 Ϫ/Ϫ animals (lanes 1-8) were incubated with an ␣-globin CCAAT box CP2 binding probe. Both the CP2 complex (CP2) and a slower migrating complex (*) are indicated at right. In lanes 9 -11, 2-, 5-, and 10-fold more extract was incubated with the probe compared with lanes 7 and 8. Similar results were observed with a ␥-fibrinogen probe (data not shown). B, recombinant CP2 and NF2d9 have differing electrophoretic mobilities. Equal amounts of recombinant protein were incubated with an ␣-globin CCAAT box CP2 binding probe. Both the CP2 complex (CP2) and the NF2d9 complex are indicated. Similar results were observed with a ␥-fibrinogen probe (data not shown). C, effect of NF2d9-specific antisera on EMSA. Nuclear extracts derived from livers of CP2 ϩ/ϩ (lanes 1 and 3) or CP2 Ϫ/Ϫ (lanes 2 and 4) mice were incubated with an ␣-globin CCAAT box CP2 binding probe in the presence of 1 l of NF2d9-specific monoclonal antiserum (lanes 3 and 4). An approximately 5-fold excess of nuclear extract was used in lanes 2 and 4 in contrast with lanes 1 and 2.

FIG. 4. Expression pattern of CP2 and NF2d9 in wild type and mutant embryos.
A-D, sequential sagittal sections of wild type murine embryos E13.5 dpc were subjected to in situ hybridization with either CP2 (A) or NF2d9 (C) 33 P-labeled sense riboprobes or CP2 (B) or NF2d9 (D) antisense riboprobes. E, analysis of NF2d9 expression in CP2 ϩ/ϩ and CP2 null animals. Total RNA was extracted from brain, heart, and kidney of wild type (CP2 ϩ/ϩ ) or CP2 Ϫ/Ϫ animals and hybridized to a mixture of human NF2d9 and actin riboprobes (d9.5). Protected fragments are indicated. This experiment is representative of several litters assayed. that specifically interact with CP2. 4 The first, NF-E4, is the fetal/erythroid-specific component of the SSP. The second, RING1B, is a RING finger domain-containing protein involved in the regulation of CP2-dependent transcription. Utilizing the yeast two hybrid assay system we compared the ability of CP2 and LBP-1a to interact with these known CP2 partners. As shown in Fig. 5A, growth on leucine/tryptophan/histidine plates, which is indicative of a protein-protein interaction, was observed with both CP2 and LBP-1a and NF-E4. Protein interactions were also observed between CP2 and LBP-1a and RING1B (Fig. 5B). No growth was observed with controls lacking either of the interacting proteins. DISCUSSION Prompted by studies documenting the importance of the NTF-1 gene in Drosophila development, we have examined the effects of gene targeting of the mammalian NTF-like gene, CP2, in mice. These experiments assumed additional importance with the identification of CP2 as a major component of the SSP, a protein complex involved in the regulation of fetal hematopoiesis and with the identification of CP2 as a key factor in the T-cell proliferative response (17,19). To our surprise, no difference in hematopoiesis, globin chain synthesis, or immunological function between wild type and CP2 null animals was observed. Indeed, the general physiology, behavior, and repro-ductive capacity of CP2 Ϫ/Ϫ animals was identical to wild type littermates. Examination of the binding activities of nuclear extracts suggested that CP2 consensus binding sites are occupied in CP2 Ϫ/Ϫ animals by NF2d9, a protein highly related to CP2. In addition, we have shown that NF-E4 and RING1B, known heteromeric partners of CP2, also form protein complexes with LBP-1a/NF2d9. The similar patterns of expression of the two highly related genes coupled with the DNA and protein binding data suggests that the lack of a discernible phenotype in the CP2 nullizygous mice may be due to rescue by NF2d9.
Although our data is consistent with redundancy of function in the mammalian NTF-like gene family, it was essential to rule out the possibility that the knockout phenotype was masked by the production of a truncated or alternately spliced form of CP2. Several lines of evidence suggest that this did not occur. First, RNase protection and RT polymerase chain reaction analysis failed to show evidence of either the 5Ј exons 1 and 2 or the 3Ј end of the coding sequence. Second, DNAprotein interaction studies demonstrated loss of the CP2 homodimeric band with the appearance of a slower migrating complex, which we attributed to NF2d9. The ability of LBP-1a/ NF2d9 to bind to CP2 consensus sites is not surprising. Examination of the amino acid sequence of the respective proteins reveals striking homology in the region that we, and others, have identified as the DNA binding domain (18,30). 5 Between residues 150 and 291, the core of the binding domain, the two proteins share 90% identity and 96% similarity. Previous studies have demonstrated that LBP-1a and CP2 can bind the CP2 consensus sequence adjacent to the HIV initiation site (9). We have extended that observation, confirming that LBP-1a also binds to the CP2 sites in the SV40 major late promoter and the murine ␥-fibrinogen and ␣-globin promoters. These sequences are archetypal CP2 binding sites in that they consist of a pair of direct repeats (G/A)CTGG spaced by an intervening sequence of variable content, but set length, which restricts protein binding to a single face of the DNA helix (18,27). It is therefore likely that the other target sites for CP2 will also allow binding of LBP-1a.
The migration pattern we observed with CP2 Ϫ/Ϫ extracts in the EMSA was consistent with a protein-DNA complex containing NF2d9. The slightly slower migration reflects the larger size of the NF2d9 protein and is consistent with previous reports and our observations of the difference in the migration of recombinant CP2 and NF2d9 (Fig. 3B) (9). The ability of antiserum that recognizes only NF2d9 to displace the complex observed in CP2 Ϫ/Ϫ animals, coupled with the partial displacement observed with wild type extract, further supports this conclusion. Previous studies and the results reported here demonstrate the ability of LBP-1a/NF2d9 to functionally compensate for CP2 in its transcriptional roles (9). The activation domains of the two proteins have been mapped by our group to the N-terminal 40 amino acids. In this region, the two factors are 88% identical. We and others have demonstrated transcriptional activity of both CP2 and LBP-1a/NF2d9 in yeast and mammalian cells and in in vitro transcription assays (9,15,17,30). 6 The ability of LBP-1a/NF2d9 to fully compensate for CP2 loss in the context of heteromeric protein interactions was less assured. Sequence comparison of the previously characterized dimerization domain of CP2 reveals that it shares 52% identity and 75% similarity with LBP-1a at amino acid level. Recently, the dimerization domain of CP2 has been refined to a region 4  and NF-E4. The Saccharomyces cerevisiae reporter strain HF7C was transformed with the indicated plasmids. pGB-LBP-1a and pGB-CP2 contain the dimerization domains of LBP-1a and CP2, respectively, fused in frame with the DNA binding domain of GAL4 (amino acids 1-147). pACT-NF-E4 contains the entire coding sequence of the NF-E4 cDNA fused in frame with the activation domain of GAL4 (amino acids 768 -881). A specific interaction between pTD encoding the SV40 Tantigen and pVA3 encoding p53 has been reported previously. Yeast transformants were streaked onto synthetic media plates lacking leucine, tryptophan, and histidine (LTH Ϫ ) to assess potential protein interactions. B, yeast two hybrid assay of CP2/LBP-1a and Ring1B. The experiments were performed using the above methodology with the exception that pACT-Ring1B was substituted for pACT-NF-E4. pACT-Ring1B contains the entire coding region of the Ring1B cDNA fused in frame with the activation domain of GAL4 (amino acids 768 -881). between residues 266 and 403 (18). In this sequence, the two proteins share 63% amino acid homology and 85% amino acid similarity. Therefore it was not surprising that protein-protein interactions between LBP-1a and previously identified partners of CP2 were conserved.
Our studies of gene expression lend further support to our hypothesis that LBP-1a/NF2d9 rescues CP2 nullizygous animals. In situ hybridization analysis revealed that CP2 and NF2d9 are widely co-expressed, albeit at differing levels. One striking difference was observed in the pattern of expression between the two genes, with NF2d9 being present at significantly higher levels in the fetal liver. Two interpretations of this result are suggested by the known roles of CP2 and NF2d9. First, NF2d9 has been linked to gender-specific expression of the steroid 16␣-hydroxylase P450 (Cy2d-9) gene in the mouse liver (10). In this setting, NF2d9 forms a heteromeric complex with an unknown partner protein. Enrichment of NF2d9 in this organ may indicate that it is the preferred protein partner for this developmental process for reasons that are as yet unknown. Second, the fetal liver is a site of hematopoiesis in the developing mouse. Studies from mice transgenic for the human ␤-globin locus YAC demonstrate that a distinct fetal stage of human ␥-gene expression occurs between days 10.5 and 13.5 (29,31). As the SSP is involved in the preferential expression of this gene during fetal erythropoiesis it is conceivable that NF2d9 and not CP2 is the primary partner of NF-E4 in the formation of this complex (21). Studies of human globin chain synthesis in NF2d9 nullizygous mice will address this question.
The ability of one highly related gene to compensate for the loss of another in gene-targeting experiments is widely recognized. Redundancy may be observed for all functions of the protein or may be limited to a single organ system (32,33). Studies of the maf transcriptional regulators highlight the complexity of gene-targeting analysis in a multigene family. Although two of the family members, mafK and mafG, share overall 62% amino acid identity, with higher degrees of homology in known functional domains, the phenotype of nullizygous animals is widely divergent (34). Unlike the normal phenotype observed with mafK-deficient animals, mafG nullizygotes display impaired megakaryocytic development and behavioral defects. This divergence occurs despite largely identical patterns of expression.
Despite the evidence suggesting that NF2d9 can fully compensate for the loss of CP2 in vivo, it is possible that a subtle CP2-specific phenotype exists in the nullizygous mice. For example, although we have performed extensive phenotypic analysis of this strain, it is possible that we have not identified the tissues or tissues that require study. For example, we have closely evaluated the histological features of the central nervous system in the CP2 Ϫ/Ϫ animals, as well as their behavioral patterns in view of the variations in patterns of expression in various parts of the developing brain. Although we have observed no abnormality, it is possible that an abnormal phenotype may become evident with the establishment of the CP2 null genotype in different inbred strains. Our studies documenting the pattern of expression of CP2 and NF2d9 failed to provide clues as to possible organs in which CP2 may be non-redundant. However, they did suggest that NF2d9 may play a key role in fetal liver function. The generation of NF2d9-deficient animals and their interbreeding with the CP2 nullizygous mice will address this issue.