As part of our ongoing ENU-mutagenesis screen, we identified the mutant mouse line droopy eye (drey). Mutant embryos exhibit a variety of incompletely penetrant defects including expansion of the retinal-pigmented epithelium over the dorsal half of the eye (present in 116/147 embryos analyzed; 79%) and neural tube closure defects consisting of spina bifida (21/147; 14%) and exencephaly (81/147; 55%; Figures 1A and 1B). A small percentage of embryos (4/147; 3%) show severe defects in mesoderm development including a malformed allantois, somite defects, and posterior mesoderm truncations (Figures 1C and 1D). Conversely, some drey mutant embryos (7/31; 23%) do not exhibit any detectible phenotypes, and in a few rare cases, drey mutant animals survive postnatally (4/92; 4%) and could reproduce (1/92; 1%).#

Using meiotic recombination mapping, the drey mutation mapped to a 1 Mb region on mouse chromosome 3, which contains six predicted transcripts including the D3Ertd300e transcript (Figure 1E; NCBI accession #BC052702). D3Ertd300e is annotated as a p38-interacting protein based on identification of the human homolog in a yeast two-hybrid screen using p38α as bait (Y.L. and J.H., unpublished data; NCBI accession #AF093250). Furthermore, we demonstrate that D3Ertd300e interacts with p38 MAPK and is required for its activity in vivo (see below and Figure 2), thus we rename D3Ertd300e: p38-interacting protein (p38IP). There is very little known about the p38IP gene and there are no apparent studies of p38IP function. Human p38IP/C13orf19 was found to have altered expression in prostate cancers relative to normal prostate tissue (Schmidt et al. 2001 2005) and to be expressed in human hematopoietic stem cells (Gomes et al., 2002). Since p38 MAP Kinase has been shown to regulate a variety of biological processes such as apoptosis, proliferation, and gene expression (reviewed in New and Han, 1998), p38IP was an interesting candidate to test as the responsible gene for a regulator of neural tube closure. Upon sequencing drey genomic DNA around the splice junctions of the predicted p38IP transcript, a T to C transition was identified in the splice donor consensus sequence at the end of exon 14 (Figure 1F). The p38IP protein contains a classical nuclear localization sequence, a PEST sequence, and a serine-rich region at the C terminus (Figure 1G).#

To determine if the mutation disrupts p38IP splicing, an RT-PCR assay was performed using primers flanking exon 14 (Figure 1H). Amplification of wild-type cDNA resulted in a single 500 bp PCR product, whereas mutant cDNA produced multiple products including a 500 bp product at low abundance, indicating that a small proportion of transcripts are spliced normally in mutant embryos. Sequencing of the aberrantly spliced transcripts indicates that they encode disrupted proteins with premature stop codons encountered at either 951 or 948 bp. The small proportion of normal transcripts produced likely accounts for the incomplete penetrance of phenotypes observed in drey mutant embryos.#

To confirm that mutation of p38IP is responsible for the developmental defects in p38IP^drey mutants, a second allele of p38IP was obtained from the BayGenomics genetrap resource and used in a complementation test cross. ES cell clone RRK304 (designated p38IP^RRK) contains an insertion in the seventh intron of the p38IP gene, which is predicted to fuse the first 133 amino acids of p38IP to a β-galactosidase and neomycin phosphotransferase cassette. To determine the fidelity of splicing of the genetrap cassette into the p38IP transcript, RT-PCR using primers that flank exons 7 and 8 was performed. The absence of any detectable normal transcript in p38IP^RRK/RRK mutant embryos suggests that this allele is likely a null or a severe hypomorph (data not shown). Similar to p38IP^drey/drey mutant embryos, transheterozygous p38IP^drey/RRK mutant embryos exhibit incompletely penetrant phenotypes that range from gastrulation and neural tube defects to morphologically normal embryos (Figure 1I and data not shown). This genetic experiment confirms that the mutation in p38IP is responsible for the drey phenotypes.#

To determine if p38IP is expressed in a pattern that is consistent with a role in regulation of gastrulation or neurulation, p38IP expression was examined by in situ hybridization using an antisense RNA probe against the p38IP transcript in wild-type embryos and LacZ in p38IP^RRK/+ embryos. These experiments reveal that the p38IP transcript is expressed ubiquitously at all stages examined (E7.5–E12.5; Figure 1J and data not shown).#

As mentioned above, we identified the human homolog of p38IP in a yeast two-hybrid screen using p38α as bait. Six colonies representing three different genes were identified out of 2 × 10⁷ colonies screened. Two of the proteins were known p38α interactors: the p38α substrate MK2 and Pax6 (Mikkola et al. 1999; Stokoe et al. 1992). One undescribed 353 amino acid peptide was identified, and because there was no information available about the gene, we named it p38-Interacting Protein (p38IP) (Y.L. and J.H., unpublished data; NCBI accession #AF093250). To confirm the interaction, we tested the ability of p38IP to interact with p38α in coimmunoprecipitation assays. When YFP-tagged p38IP and Flag-tagged p38α are coexpressed in 293T cells, p38IP associates with p38α (Figure 2A). Interaction between endogenous p38IP and p38α is also detected following immunoprecipitation of untransfected 293T cells (Figure 2B). To determine the specificity of p38IP binding to p38, p38 isoforms with the TGY dual phosphorylation site changed to AGF were used in a yeast two-hybrid interaction assay (Figures 2C–2F). Cells cotransfected with p38IP and p38α are able to grow on SD-LWH media and induce β-galactosidase activity, indicating an interaction, whereas p38IP does not interact by these assays in yeast with p38β, p38δ, p38γ, or vector control pGBT9 (Figures 2C–2E). Coexpression of p38IP and p38α is able to significantly induce β-galactosidase activity comparable to that of p53 and large T antigen (Figure 2F).#

In comparison to human p38IP, mouse p38IP ends at amino acid 530, resulting in a single N-terminal Serine-rich domain. To determine if the mouse p38IP also interacts with p38, we tested binding in a yeast two-hybrid assay. Identical to human p38IP, mouse p38IP only interacted with the p38α isoform when scored for growth on medium lacking histidine (SD-LWH) and induction of β-galactosidase (Figure 2G and data not shown). To determine if the truncated p38IP proteins that are potentially produced in p38IP^drey and p38IP^RRK mutant mice (Figure 1G) can bind to p38α, we mapped the binding domain in the p38IP protein. Yeast coexpressing N-terminal p38IP (1–383) and p38α are not able to grow on SD-LWH media or induce the β-galactosidase reporter, indicating that p38IP does not interact with this fragment. In contrast, cells cotransfected with C-terminal p38IP (380–733) and p38α are able to grow on SD-LWH media and induce β-galactosidase, indicating that the p38α interaction domain is in the C-terminal region of the protein. The interaction domain also was confirmed in a coimmunoprecipitation assay using in vitro translated proteins (data summarized in Figure 2G). These data reveal that the truncated proteins potentially produced by the two mutant alleles of p38IP do not bind p38α.#

To determine the in vivo relevance of the interaction between p38IP and p38, we examined whether p38 activity was misregulated in the drey mutant mouse. The activation of p38 and its downstream substrates CREB and ATF2 were examined using phospho-specific antibodies in wild-type and drey mutant eyes, an organ that is affected by p38IP truncation. In the wild-type E12.5 eye, p38, CREB, and ATF2 are phosphorylated in the retinal-pigmented epithelium (RPE; p38: 53%, CREB: 67%, ATF2: 89% of cells), neural retina (NR; p38: 69%, CREB: 57%, ATF2: 79% of cells), and lens (p38: 44%, CREB: 46%, ATF2: 74% of cells; Figures 2H, 2J, and 2L). In contrast, in the p38IP^drey/drey mutant eye, phosphorylated p38 is only detected at low levels in a few cells scattered throughout the eye (RPE: 1%; NR: 6%; lens: 7% of cells; Figure 2I). The phosphorylation of p38 substrates is not detected in specific cell layers in p38IP^drey/drey mutant eyes. Phosphorylated CREB is detected in the lens of p38IP^drey/drey eyes (53%) but is greatly reduced in the RPE (4%) and neural retina (4%; Figure 2K). While phosphorylated ATF2 is detected in the neural retina (88%) and lens (75%) of p38IP^drey/drey eyes, it is significantly reduced in the RPE (9%; Figure 2M). These data reveal that p38IP interacts with p38α, and in mutant embryonic tissues in which truncated p38IP cannot bind p38α, activation of p38 and downstream substrates is impaired in vivo.#

To investigate further the role of p38IP in embryogenesis, we examined mutant phenotypes in mouse embryos homozygous for the more severe p38IP^RRK allele (Figure 3). E9.5 and E10.5 mutant embryos exhibit multiple developmental defects consistent with abnormalities in development of mesoderm (Figures 3A–3D). Yolk sac membranes are wrinkled and poorly vascularized, and mutants develop a malformed allantois that fails to fuse to the chorion. Mutant embryos are necrotic, developmentally delayed, exhibit misshapen head folds and exencephaly, and fail to form somites or form only a few anterior somites.#

These widespread abnormalities in development of mesodermally derived tissues could originate during gastrulation when mesoderm is specified and migrates. At gastrulation, p38IP^RRK/RRK mutant embryos contain a mass of cells on the posterior side (Figures 3E and 3F). Histological analysis reveals that a significant proportion of the mesoderm failed to migrate away from the primitive streak (Figures 3G and 3H). Mesoderm migration defects were consistently observed although the extent varied between mutants.#

This phenotype is strikingly similar to that of Fgfr1 and Fgf8 mutant embryos (Deng et al. 1994; Sun et al. 1999; Yamaguchi et al. 1994). Fgfr1 is required for the expression of a series of mesoderm-specific markers such as Sprouty2, Tbx6, Brachyury, and Lim1; moreover, Fgfr1 is required for the expression of Snail, which mediates downregulation of the transcript for the E-cadherin cell adhesion protein to promote EMT (Ciruna and Rossant, 2001). However, in contrast to Fgf mutants, Sprouty2, Tbx6, and Brachyury are correctly expressed in p38IP^RRK/RRK mutant embryos (Figures 4A–4F). In wild-type embryos, Lim1 is expressed at higher levels in cells as they migrate away from the primitive streak (Barnes et al., 1994). In mutant embryos, Lim1 is expressed, although in a more limited range (Figures 4G and 4H), suggesting that cells are turning on the mesenchymal developmental program but nevertheless fail to properly migrate away from the streak.#

At E8.5, examination of molecular marker expression reveals that mesoderm is properly specified although disorganized in mutant embryos (Figures 4I–4T). Presomitic mesoderm is specified as evidenced by Fgf8 and Tbx6 expression (Figures 4I, 4J, 4M,and 4N). Mox1, a somite marker, is expressed in mutant embryos, although not organized into somitomeres (Figures 4K and 4L). Less severely affected mutant embryos do form some anterior somites as evident by expression of Twist (Figures 4S and 4T). Often the paraxial mesoderm is not divided into discrete left and right domains, likely due to defects in development of the midline. Examination of Brachyury, a midline axial mesoderm marker, shows that the midline is present in mutant embryos but is often severely disrupted and discontinuous (Figures 4Q and 4R). Less severely affected embryos, however, show division of Snail and Twist expression into discrete left and right domains, indicating a more normal midline (Figures 4O, 4P, 4S, and 4T). These results indicate that both axial and nonaxial mesoderm is specified in p38IP^RRK/RRK mutant embryos but disorganized.#

To determine if these defects in mesoderm migration are related to defects in p38 activation, p38 phosphorylation was determined in the primitive streak of E7.5 wild-type and mutant embryos. In wild-type embryos, phosphorylated p38 is detected ubiquitously in all germ layers (Figures 5A–5C). Despite the ubiquitous expression pattern of p38IP, in mutant embryos, phosphorylated p38 is not detected specifically in the primitive streak (between arrows) and in the mesoderm that fails to migrate away from the streak (arrowhead, Figures 5D–5F). Phosphorylated p38 is detected, however, in the rest of the embryo, including the few mesoderm cells that are able to migrate away from the primitive streak.#

Fgfr1 regulates the expression of Snail transcript and the Snail protein represses E-cadherin transcription to promote EMT (Ciruna and Rossant, 2001). To determine if this pathway is disrupted in p38IP^RRK/RRK mutant embryos, the expression of Snail and E-cadherin was examined during gastrulation. Snail is expressed at normal levels in the primitive streak of mutant embryos, indicating that p38IP is not required for regulation of Snail expression (Figures 5G and 5J). Consistent with the expression of Snail, which acts as a transcriptional repressor at the E-cadherin promoter, the E-cadherin transcript is downregulated in the mesoderm of mutant embryos, although these cells fail to migrate (Figures 5H and 5L).#

The E-cadherin protein is localized to the adherence junctions of cells in the epithelium, and the E-cadherin transcript gets rapidly downregulated as cells exit the epithelial cell layer (Figure 5I). Yet, in the cells that have just exited the epithelial layer and downregulated the E-cadherin transcript, E-cadherin protein is still present at the junctions and gets downregulated as cells migrate away from the streak (Figure 5M). This observation suggests that E-cadherin protein expression is also regulated posttranscriptionally. To determine if p38IP is required for posttranscriptional regulation of E-cadherin, E-cadherin protein expression was examined in mutant embryos (Figures 5N and 5O). Strikingly, in mutant embryos, the E-cadherin protein remains localized to the junction of the cells that fail to migrate away from the primitive streak. These results raise the intriguing possibility that two pathways act independently to regulate E-cadherin in the primitive streak to allow mesoderm migration: Fgf/Snail acts at the transcriptional level to downregulate the E-cadherin transcript and p38IP acts to downregulate or destabilize the E-cadherin protein.#

When cells undergo EMT, the expression levels of a number of cell adhesion markers become either up- or downregulated. To determine if p38IP is required for other aspects of EMT, additional EMT markers were examined in mutant embryos. Markers of mesoderm migration such as Integrin-α5 and Fibronectin are efficiently upregulated in mutant mesoderm that fails to migrate away from the primitive streak (Figures 5P–5S). In addition, as cells delaminate from the primitive streak, they change morphology as highlighted by staining with an anti-β-catenin antibody (Figures 5T and 5U). Taken together, these results suggest that while p38IP^RRK/RRK mutant mesoderm undergoes a partial EMT, it fails to complete EMT and downregulate the E-cadherin protein.#

To study further the requirement for p38IP in mesoderm migration, mesoderm from the primitive streak of E7.5 wild-type embryos or the cells that fail to migrate away from the primitive streak of mutant embryos were dissected and cultured on fibronectin-coated plates. Wild-type mesoderm migrates extensively away from the explant coincident with the downregulation of E-cadherin protein (Figures 6A and 6B and Burdsal et al., 1993). As seen in vivo, cells from the mutant primitive streak that failed to downregulate E-cadherin also failed to migrate when explanted (Figures 6C and 6D). To determine whether p38IP^RRK/RRK mutant cells can migrate if E-cadherin function is blocked, explants were incubated with a function-perturbing anti-E-cadherin antibody. This resulted in both the downregulation of E-cadherin and extensive migration of mutant cells away from the explant (Figures 6E and 6F). These results suggest that p38IP^RRK/RRK mutant cells likely remain in the primitive streak due to a deficiency in the downregulation of E-cadherin protein, but otherwise p38IP is not required for general cell migration.#

p38IP^RRK/RRK mutant embryos fail to activate p38 specifically in the primitive streak and the mesoderm that fails to downregulate E-cadherin (Figures 5F and 5O). To determine if the failure to downregulate E-cadherin in p38IP^RRK/RRK mutant embryos is due to a failure to activate p38 in the primitive streak, explants from wild-type embryos were cultured in the presence of either a specific p38 kinase inhibitor or vehicle control (DMSO). The chemical inhibitor SB203580 targets both p38α and p38β but not p38δ or p38γ (Cuenda et al., 1995). In control explants incubated with DMSO, cells are capable of downregulating E-cadherin and migrating away from the explant (Figures 6G and 6H). In contrast, explants treated with 20 μM SB203580 did not downregulate E-cadherin and failed to migrate away from the explant (Figures 6I and 6J). This result indicates that p38 activity is required for downregulation of E-cadherin and EMT. To confirm this and to test whether p38 is also required for mesoderm migration after EMT, explants were incubated in 20 μM SB203580 and the function blocking anti-E-cadherin antibody (Figures 6K and 6L). Similar to p38IP^RRK/RRK mutant explants, addition of anti- E-cadherin antibody rescued EMT and cell migration. These results indicate that both p38IP and p38 are required for downregulation of E-cadherin but not for mesoderm migration.#

To determine if Fgf signaling is required for p38 activation in the primitive streak, p38 phosphorylation was examined in Fgf8 mutant embryos (Figures 6M and 6N). In contrast to p38IP^RRK/RRK mutants (Figures 5D–5F), Fgf8 mutant embryos showed robust activation of p38 in the primitive streak and the cells that fail to migrate. The Drosophila homolog of NIK/Map4k4, Misshapen (Msn), requires p38 activity for some in vivo responses (Paricio et al., 1999), and the phenotype of NIK^−/− embryos (Xue et al., 2001) is similar to that of p38IP^RRK/RRK, suggesting that they may act in a similar pathway. Similar to p38IP mutant embryos, p38 is not activated in the primitive streak (between arrows) or the mesoderm that accumulates in the streak of NIK mutant embryos (arrowhead, Figure 6O). These data indicate that NIK may act upstream of p38 in a pathway to promote EMT and that this pathway is independent of Fgf signaling. This possibility is also supported by the similarities in expression of mesoderm markers in p38IP^RRK/RRK and NIK mutant embryos as compared to Fgf8 or Fgfr1 mutant embryos (Deng et al. 1994; Sun et al. 1999; Xue et al. 2001; Yamaguchi et al. 1994).#

We demonstrate that p38IP specifically binds p38 in both yeast two-hybrid and coimmunoprecipitation assays. We mapped the binding domain to the C-terminal region of p38IP. Significantly, the mutations in both the p38IP^RRK and p38IP^drey alleles result in deletion of the C-terminal domain of p38IP, and these mutant proteins cannot bind to p38. Furthermore, we demonstrate that p38 activation in vivo is compromised in both p38IP^RRK and p38IP^drey mutants. Future experiments are required to elucidate the biochemical mechanism(s) by which p38IP functions to regulate p38.#

We found that p38IP binds to only p38α in a yeast two-hybrid assay, yet previous studies of p38α mutant embryos demonstrated that p38α is essential for development of the placenta but not for gastrulation (Adams et al. 2000; Mudgett et al. 2000). This raises the possibility that p38IP may be required for activation of other p38 isoforms in vivo. To begin to investigate this possibility, we determined the expression pattern of p38β, p38γ, and p38δ to determine if the other p38 isoforms are expressed in the primitive streak. p38γ and p38δ are expressed in the extraembryonic regions of the gastrula while p38β is expressed in the epiblast (Figure S1). This data combined with studies in explanted mesoderm in which an inhibitor of p38α and p38β prevents EMT (Figure 6) suggest that p38IP may be required for activation of both p38α and p38β in vivo. An alternate explanation is that p38α may normally mediate EMT in the primitive streak but in the absence of p38α, p38β may be able to compensate either by altering its normal expression pattern or activity. In support of these ideas, we examined activation of p38 in p38α mutant gastrula stage embryos and found that p38 is phosphorylated normally throughout the epiblast and streak (I.E.Z. and L.N., unpublished data). In addition, we cannot exclude the possibility that p38IP is also required for the function of other, yet to be discovered, factors.#

Our data provide an intriguing link between p38 activation and the NIK pathway. This relationship to NIK is surprising as NIK activates JNK but not p38 in cultured cells (Su et al., 1997). Similarly, Misshapen (Msn), the Drosophila NIK homolog, activates JNK to regulate dorsal closure (Su et al., 1998), and Msn couples Frizzled and Disheveled to JNK activation to regulate planar cell polarity in the Drosophila wing and eye (Paricio et al., 1999). Interestingly in this context, JNK acts redundantly with Dp38a and Dp38b, suggesting that Msn is capable of regulating p38 activity in vivo. During mouse embryogenesis, a role for NIK in eye and neural tube morphogenesis has not been determined as NIK mutant embryos die around E9.5 (Xue et al., 2001). However, JNK mutant embryos (either single or double mutants) do not exhibit the severe gastrulation defects observed in NIK mutant embryos (Kuan et al. 1999; Sabapathy et al. 1999), suggesting that NIK may activate JNK-independent pathways in vivo. Indeed, our studies indicate that p38 activity is specifically disrupted in NIK mutant embryos, and mutations in either NIK or p38IP result in similar gastrulation defects, providing additional in vivo support for a link between NIK and p38 activation.#

Neural tube defects (NTDs) occur when the neural tube fails to close completely during embryogenesis leading to exencephaly/anencephaly and spina bifida. NTDs are one of the most common birth defects observed in humans and represent a complex genetic disease (reviewed in Copp et al., 2003). The mouse has provided over 100 potential candidate genes for NTDs, and in some instances mutations in these genes have been identified in human patients (reviewed in Copp et al. 2003; Zohn et al. 2005). Furthermore, identification of hypomorphic mutations associated with NTDs indicates that these types of mutations will reveal many more potential candidate genes. We show that a severe allele of p38IP results in gastrulation defects whereas the milder drey allele results in exencephaly and spina bifida. Future experiments will determine if the morphogenic defects in drey mutant mice such as the failure to close the neural tube and eye defects are due to specific defects in p38 activation.#

The EMT that occurs in the primitive streak of the gastrulating mouse embryo and during invasion and metastasis of tumor cells involve regulation of E-cadherin expression by Snail (reviewed in Barrallo-Gimeno and Nieto, 2005). Furthermore, the loss of E-cadherin expression is a central event in the transition of tumors from noninvasive to invasive carcinomas (Batlle et al. 2000; Cano et al. 2000). Here we present evidence of a Snail-independent pathway in which p38 is required in the primitive streak to downregulate E-cadherin expression at the posttranscriptional level. Whether this Snail-independent, p38-dependent pathway also functions in EMT during tumor invasion remains to be determined.#

drey was identified in a screen for recessive ENU-induced mutations that cause morphological abnormalities at E12.5 (Garcia-Garcia et al., 2005; Kasarskis et al., 1998; Zohn et al., 2005). The drey mutation was generated on a C57BL/6J genetic background and backcrossed to C3H. In a mapping cross of 480 opportunities for recombination, drey was mapped between Massachusetts Institute of Technology (MIT) simple sequence length polymorphism (SSLP) markers D3mit6 and D3mit137. For high-resolution mapping, additional polymorphic DNA markers were generated based on nucleotide repeat sequences (see http://mouse.ski.mskcc.org/ for primer sequences). A mouse embryonic stem (ES) cell line carrying an insertion in the p38IP gene (RRK304) was obtained from BayGenomics http://baygenomics.ucsf.edu/ database. Other mutant mouse lines used were Fgf8 and NIK (Meyers et al. 1998; Xue et al. 2001).#

cDNAs of the p38IP gene were amplified by RT-PCR (Superscript One-Step RT-PCR, Invitrogen) using RNA from E10.5 drey/drey and C57BL/6 control embryos. The splice site junctions around exon 14 were sequenced and a T to C transition discovered in the splice donor consensus sequence. Sequencing was confirmed using ten additional drey/drey mutant embryos.#

Whole-mount and section RNA in situs were performed as described (Holmes and Niswander 2001; Liu et al. 1998). p38IP expression pattern was determined using an antisense RNA probe synthesized from IMAGE clone: 2598858 or by expression of the lacZ transcript in p38IP^RRK/+ embryos. Immunofluorescence experiments were performed as described (Timmer et al., 2002) using anti-E-cadherin antibody (1:100 dilution of anti-uvomorulin antibody; Sigma F3648), Hoechst (10 μg/ml; clone DECMA-1, Sigma), anti-Phospho-p38 MAP Kinase (Thr180/Tyr182) antibody (1:100 dilution; Cell Signaling #9211), anti-Phospho-ATF2 antibody (1:100 dilution; Cell Signaling #9221), anti-Phospho-CREB antibody (1:100 dilution; Cell Signaling #9191), anti-Integrin-α5 (1:100; BD-Pharmingen #553319), anti-fibronectin (1:100; Sigma F3648), anti-β-catenin (1:1000; Sigma C7082).#

Primitive streak explants were dissected from E7.5 wild-type and mutant embryos and grown on fibronectin-coated plates for 3 days as described (Ciruna and Rossant, 2001). Media was supplemented with 20 μM SB203580 (Calbiochem) in DMSO and/or anti-E-cadherin (1:100 dilution of anti-uvomorulin antibody; Sigma).#

To identify p38-interacting proteins, we used a kinase-inactive mutant of p38α fused with the GAL4 DNA binding domain (pGBT9-p38(AF)) as bait and performed a yeast two-hybrid screen of a Hela cell cDNA library constructed in the activation domain plasmid pACT as described by the manufacturer (BD Clontech). Sequencing showed that the clone encodes amino acids 381–733 of p38IP (1–733). PCR and Southern blots were used to clone full-length human p38IP from a phage library. The specificity of the interaction was tested by immunoprecipitation as described (Ge et al., 2002).#