Paxillin is a protein containing four LIM domains, and functions in integrin signaling. We report here that two transcripts are generated from the paxillin gene locus in Drosophila; one encodes a protein homolog of the vertebrate Paxillin (DPxn37), and the other a protein with only three LIM domains, partly encoded by its own specific exon (PDLP). At the myotendinous junctions of Drosophila embryos where integrins play important roles, both DPxn37 and PDLP are highly expressed with different patterns; DPxn37 is predominantly concentrated at the center of the junctions, whereas PDLP is highly enriched at neighboring sides of the junction centers, primarily expressed in the mesodermal myotubes. Northern blot analysis revealed that DPxn37 is ubiquitously expressed throughout the life cycle, whereas PDLP expression exhibits a biphasic pattern during development, largely concomitant with muscle generation and remodeling. Our results collectively reveal that a unique system exists in Drosophila for the generation of a novel type of LIM-only protein, highly expressed in the embryonic musclature, largely utilizing the Paxillin LIM domains.
The LIM domain is a cysteine-rich sequence functioning as a protein-binding interface (Schmeickel and Beckerle, 1994), and is present in single or multiple copies in a variety of proteins that are generally involved in transcriptional regulation, cell signaling and control of cell fate (for review see Dawid et al., 1998). LIM domains were grouped by sequence similarity into four classes (A–D), with some outliers (Dawid et al., 1995), and LIM-containing proteins have been classified into three groups, based on sequence relationships among LIM domains and on the overall structure of the proteins (Dawid et al., 1998). Group 1 includes proteins bearing A and B class LIM domains in tandem, such as LHX (LIM homeodomain) proteins, LMOs (LIM-only proteins) and LIM kinases. Group 2 includes proteins composed largely of LIM domains of class C, such as CRIPs (cysteine-rich intestinal proteins) and CRPs (cysteine-rich proteins). Most of the proteins in group 3 contain class D LIM domains, and are more heterogeneous, containing different numbers of LIM domains located at the C-terminus, such as Paxillin, Zyxin and PINCH.
To investigate the possible role of paxillin during development as well as cell migration, we have initiated a reverse genetic approach in Drosophila melanogaster, and report here a unique use of paxillin LIM domains in Drosophila.
Results and discussion
To identify a gene encoding the Drosophila Paxillin homolog, we used TBLASTN to search the expressed-sequence tag (EST) database (Berkley Drosophila Genome Project, BDGP), and detected Drosophila EST CK00121 by homology to the LIM domains of human paxillin. By using the EST cDNA as a probe, a cDNA 2 kb long was isolated from a λgt10 cDNA library. The corresponding mRNA consists of an open reading frame 1743 nt long, and the predicted 581 amino acid protein is most similar to human Paxillin (Figure 1A and E). During our analysis, BDGP mapped the gene to polytene band 37E, and it has hence been designated DPxn37. DPxn37 protein shows 57% overall identity to human Paxillin and retains most of the protein interaction motifs, including five LD motifs and four LIM domains (LIM1–LIM4) (Figure 1E). Three out of the four major tyrosine phosphorylation sites of human Paxillin (Nakamura et al., 2000) are conserved at analogous positions; only the tyrosine residue that corresponds to Tyr31 is missing (Figure 1E). When expressed in mammalian cells, DPxn37 localizes to focal adhesions and is tyrosine-phosphorylated during cell adhesion (Figure 4; data not shown). Thus, DPxn37 is a Drosophila homolog of human Paxillin. Although human Paxillin consists of three isoforms generated by alternative splicing (Mazaki et al., 1997), the submitted genomic sequence reveals that Drosophila only encodes a Paxillin homolog to the α isoform (http://www.fruitfly.org). There are several members of the Paxillin super-family in human, including Hic-5 (Zhang et al., 2000) and Leupaxin (Lipsky et al., 1998), which have several LD motifs and four LIM domains with the same cysteine-rich consensus sequence. DPxn37 is also highly homologous to these proteins (Figure 1E), but distinct homologs of Hic-5 and Leupaxin are not found in the euchromatic genomic sequences of D. melanogaster.
When developmental northern blotting analysis was performed with poly(A)+ RNA preparations from the embryonic, larval, pupal or adult stages using the entire coding region of DPxn37 cDNA as a probe, two different size transcripts were detected. One is ∼2.4 kb long, corresponding to the full-length transcript for DPxn37, and expressed ubiquitously throughout the life cycle (Figure 1B). The other is only ∼1.4 kb long and too short to encode the entire coding region of DPxn37. Moreover, unlike the long transcript, the short version shows a biphasic pattern of expression during development (Figure 1B). We were therefore interested in identifying the nature of this small sized transcript. To this end, we divided DPxn37 cDNA into four pieces (#1, #2, #3 and #4 in Figure 1A), again performed northern blotting analysis, and found that probes #3 and #4 hybridized with both transcripts while probes #1 and #2 hybridized only with the long one (data not shown). Therefore, the short transcript appeared to contain only some of the 3′-portion of the DPxn37 gene, and the corresponding cDNA was then identified as described in Methods. Sequencing analysis of the cDNA thus obtained revealed that the short transcript encodes a protein with three LIM domains and 197 amino acids long (Figure 1A). The first and second LIM domains are the same as those in DPxn37. The third LIM domain, however, is generated by alternative splicing and contains the NH2-half of the DPxn37 LIM3 domain and a novel amino acid sequence at its C-terminal end, which is encoded by an extra exon located 0.3 kb 3′-downstream of the eighth exon of DPxn37 (Figure 1A). Despite this alternative splicing, a cysteine-rich sequence of the third LIM domain is conserved in the newly formed LIM domain, except that the last histidine residue is replaced by cysteine (Figure 1C). We named this protein paxillin-derived LIM-only protein (PDLP). The mRNA for PDLP seems to start at ∼0.1 kb 5′-upstream of exon 8 of DPxn37, and contains its own polyadenylation site located 0.64 kb 3′-downstream of DPxn37 exon 8 (data not shown; also see Figure 1A). Such a transcript corresponding to Drosophila PDLP and the exon specific to PDLP could not be found in mice and humans (data not shown).
The expression of DPxn37 and PDLP transcripts was then analyzed during embryogenesis (Figure 2). DPxn37 transcript is detected at a very early stage, perhaps partly representing their maternal expression; and is found more or less throughout the body at later stages. From stage 13, up-regulated expression of DPxn37 transcript was detected as a segmentary pattern along the somatic musculature. On the other hand, PDLP transcript is almost undetectable at the early stages, becoming detectable after stage 13. They are then also expressed in a segmentary pattern predominantly along the somatic musculature, although their pattern appears to be different from that of DPxn37 (also see below). PDLP mRNA is also expressed in cephalic muscles.
We next generated two different preparations of polyclonal antibodies, each specifically recognizing DPxn37 or PDLP (Figure 1D). Using these antibodies, the somatic musculature at stage 16 of embryogenesis was investigated (Figure 3). Phosphotyrosine is known to be highly enriched at the myotendinous junctions (Harden et al., 1996), at least partly representing integrin-mediated focal adhesion events. We found that the DPxn37 protein is abundant at the junctions, and colocalized well with phosphotyrosine. These areas highly positive for DPxn37 may include the epidermal tendon cells, which can be identified by staining with tendon cell markers such as Kakapo and α tubulin (Gregory and Brown, 1998; Strumpf and Volk, 1998; data not shown). The possible expression, however, of DPxn37 at the very ends of the myotube edges cannot be precluded. In sharp contrast, the PDLP protein overlapped only slightly with phosphotyrosine, and was almost undetectable at the center of the junctions, but was highly expressed on both sides of the junctions. These areas appeared to correspond primarily to the mesodermal myotubes, as assessed by staining of the myotube edges with αPS2 integrins (Brower et al., 1984; Bogaert et al., 1987; Figure 3). Both DPxn37 and PDLP signals were also detected at low levels in the cell body of the myotubes. With the late stage embryos, we also examined several other tissues and found that the PDLP protein is highly expressed in the cephalic muscles (data not shown), consistent with the mRNA expression.
We next examined the possible subcellular localization of DPxn37 and PDLP (Figure 4). DPxn37, when expressed in vertebrate fibroblasts, localized to focal adhesions, and colocalized well with endogenous Paxillin. On the other hand, PDLP localized along actin bundles, and only marginally overlapped with the DPxn37-containing focal adhesions, formed at the edges of actin bundles. Similar colocalization with actin bundles has been observed with two Drosophila LIM-only proteins, Mlp60A and Mlp84B, composed of one and five LIM domains, respectively (Stronach et al., 1996, 1999). These proteins are abundant in several musculatures at the late stages of embryogenesis and implicated in myogenesis. In vertebrates, there is also a CRP family of LIM-only proteins with the ability to localize to actin bundles, which are implicated in muscle differentiation (Sadler et al., 1992; Arber et al., 1994; Crawford et al., 1994). Therefore, unlike DPxn37, PDLP exhibits a property similar to those shared among muscle LIM proteins, although its biological significance remains to be elucidated (Arber et al., 1994; Stronach et al., 1996). It should be noted, however, that the cysteine-rich consensus sequence of these muscle LIM proteins and CRPs, which is CX2CX17HX2CX2CX2CX17/18CX2C (Stronach et al., 1996), is different from that of the LIM1 and LIM2 domains, which is CX2CX19HX2CX2CX17CX2D, and the LIM3 domain (see Figure 1C) of PDLP. Moreover, the LIM3 domain of vertebrate Paxillin has been implicated in the focal adhesion localization (Brown et al., 1996), and identification of sequences involved in the different subcellular localizations may be important for understanding possible functional differences between DPxn37 and PDLP. A fraction of PDLP, but not DPxn37, was also found in cell nuclei (Figure 4), again as observed with other Drosophila muscle LIM proteins as well as CRPs (Arber et al., 1994; Stronach et al., 1996). However, we did not observe PDLP to be concentrated in cell nuclei in the embryos (data not shown). The biological relevance and significance of the nuclear localization of these LIM-only proteins also remains to be elucidated, as discussed previously (Arber et al., 1994; Stronach et al., 1996).
A cysteine-rich consensus sequence of PDLP is highly homologous to that of Lepidopteran DALP (death-associated LIM-only protein), which also consists of only three LIM domains and is implicated in the programmed cell death of intersegmental muscles of adult moth Manduca sexta (Hu et al., 1999; see Figure 1F). We have not yet described the function of PDLP in adult insects. Overexpression of DALP has been shown to inhibit in vitro myotube formation of mouse C2C12 myoblasts, and hence DALP was proposed to function as a homolog of one of the Paxillin-superfamily members, Hic-5 (Hu et al., 1999). However, overexpression of PDLP did not cause such an inhibitory effect under conditions where overexpressed Hic-5 acted as an inhibitor (data not shown). Rather than being involved in the determination of the final fate of mesodermal cells in Drosophila embryogenesis (Bate, 1993), the timing of PDLP expression appears to be concomitant with late events in the myogenesis, cell migration and attachment, and cytoskeletal rearrangements, where integrins play important roles, again similar to those assigned for Mlp60A and Mlp84B (Stronach et al., 1996).
The myotendinous junctions of the Drosophila embryo are an ideal system for studies of primitive yet fundamental functions and diversities of integrins, where the tendon cells express αPS1βPS integrin and the muscle cells express αPS2βPS integrin (for review see Brown et al., 2000). They also provide model systems for the analysis of the inductive interactions between cells of distinct fates, as well as the musculature development in higher organisms (for review see Volk, 1999). The study described here revealed that Drosophila has a unique system expressing two types of LIM proteins from the paxillin gene locus, which show different patterns of expression at the myotendinous junctions during embryogenesis; the biological significance of which is thus worth analyzing further.
During our analysis, a paper describing the identification and properties of the Drosophila paxillin gene and its protein product has appeared (Wheeler and Hynes, 2001), in which PDLP was not identified.
Cloning of DPxn37 and PDLP cDNAs.
Using a 0.3 kb XhoI fragment of CK00121 (DDBJ/EMBL/GenBank accession No. AA142016) as a probe, an eye imaginal disc cDNA library in λgt10 (a gift from A. Cowman) was screened using standard protocols. Among 16 independent clones thus obtained from approximately one million screened, the longest clone being 2.0 kb in size, DPAX10 (DDBJ/EMBL/GenBank accession No. AB048194) was analyzed for nucleotide sequence of the DPxn37 transcript.
cDNA for PDLP (DDBJ/EMBL/GenBank accession No. AB056465) was isolated as follows using poly(A)+ RNA of the late pupae, by a method of 5′- and 3′-rapid amplification of cDNA ends (RACE), which was carried out according to the manufacturer's instructions (SMART RACE™ cDNA Amplification Kit, Clontech). 5′-RACE was first carried out using the primer 5′-GTTCAGCGCCGAGATGTAGTTCTC, synthesized based on the nucleotide sequence of DPxn37 exon 8 (nt 1590–1567 of DPxn37). Two different sized cDNAs, 1.5 and 0.7 kb long, were then obtained and cloned into pCR2.1-TOPO (Invitrogen). Sequence analysis revealed that the long cDNA corresponded to the DPxn37 transcript, and the short one contained an exon at its 5′-end that was not present in the DPxn37 transcript. 3′-RACE was next carried out using the primer 5′-TTCAGTCGAGGCAAGCACATCGAC, which corresponded to the 5′-end nucleotide sequence of the short cDNA, and cDNA 1.4 kb in size was obtained, which corresponds to the PDLP transcript, and subjected to sequence analysis after being cloned into pCR2.1-TOPO. The PDLP cDNA thus obtained contains a 207-bp 5′-untranslated region, and a 593-bp 3′-untranslated region.
Poly(A)+ RNAs from flies in various stages of development were prepared by Micro Poly(A) Pure™ mRNA purification kit (Ambion), and 4 μg were used per lane. Northern blotting was performed by standard methods, using 32P-labeled DPAX10 cDNA as a probe.
In situ hybridization using digoxygenin-labeled antisense RNA probes and alkaline phosphatase substrate was carried out as described (Ishimaru et al., 1999). The RNA probes correspond to nt 1–876 of the DPAX10 cDNA, and nt 709–1012 of the PDLP cDNA, which corresponds to the COOH-half of the LIM3 domain of PDLP and the 3′-untranslated region.
Antibodies and protein expression.
DNA fragments encoding the N-terminal region of DPxn37 (amino acids 2–346) and the PDLP-specific region (amino acids 168–197) were cloned into pGEX-2T and pGEX-2TK (Amersham Pharmacia), respectively, using PCR-coupled engineered restriction sites, and expressed in bacteria as fusion proteins with glutathione S-transferase (GST) sequence. Each fusion protein was purified to near homogeneity using glutathione–Sepharose beads (Amersham Pharmacia), and was immunized into rabbits. Resulting sera were first subjected to ammonium sulfate precipitation (30–50%). After being resolved in phosphate-buffered saline, each preparation of serum was pre-absorbed using GST protein coupled to NHS-activated Sepharose beads (Amersham Pharmacia Biotech) and then affinity-purified using the GST fusion protein used for the immunizations, coupled to NHS-activated Sepharose beads.
S2 cells, maintained at 27°C with 10% fetal calf serum (FCS)/HyQ CCM3 medium (Hyclone), were transfected by Cellfectin (Gibco-BRL) with pUAST-(myc)6-DPxn37 or pUAST-PDLP plasmids, constructed by using PCR-coupled engineered restriction sites, together with pWA-GAL4. At 48 h after transfection, cell proteins were subjected to immunoblotting analysis, coupled with SDS–PAGE, as described (Mazaki et al., 1997).
Whole mount staining of embryos was performed by standard procedures using affinity-purified rabbit polyclonal anti-DPxn37 or anti-PDLP antibodies, anti-α PS2 integrin antibody (CF.2C7, a gift from D.L. Brower) or mouse monoclonal anti-phosphotyrosine antibody 4G10 (UBI). In brief, embryos were collected, dechorinated, fixed with a mixture of 4% paraformaldehyde solution and heptane, and devitellinized with a methanol–heptane mixture. Embryos were then washed and permeabilized with PBS-S (PBS containing 0.2% saponin), and blocked with 5% FCS in PBS-S for 30 min before incubation with antibodies. Secondary antibodies used were Cy3-conjugated donkey anti-rabbit IgG and Cy5-conjugated donkey anti-mouse IgG (Jackson ImmunoResearch). Stained embryos were imaged by using a confocal laser scanning microscope (model 510; Carl Zeiss).
Protein subcellular localization.
NIH 3T3 cells transfected with EGFP- or DsRed-tagged DPxn37 and EGFP-tagged PDLP plasmids were fixed and subjected to immunofluorescence characterization, as described previously (Nakamura et al., 2000). These tagged constructs were made by ligating each entire region of DPxn37 and PDLP cDNA in-frame with the C-terminal end of each tag protein cDNA and cloned into pBabePuro vectors using PCR-coupled engineered restriction sites, similar to the case of human paxillin (Nakamura et al., 2000). Endogenous Paxillin was detected using polyclonal rabbit Ab199-217 (Mazaki et al., 1997), which does not cross-react with DPxn37 or PDLP, coupled with Cy3-conjugated donkey anti-rabbit IgG. F-actin was visualized by biotinylated-phalloidin and Cy5-conjugated streptavidin (Molecular Probes). Confocal images were acquired using a confocal laser scanning microscope (model 510; Carl Zeiss).
We are grateful to Talila Volk at the Weizmann Institute and Danny L. Brower at the University of Arizona for antibodies for marker proteins of tendon cells and myotubes, Yoshiaki Ito at Kyoto University for C2C12 cells, and Kiyoshi Nose at Showa University for Hic-5 cDNAs. We also thank Yumiko Shibata, Mihoko Sato, Manami Hiraishi and Rosemary Williams for their technical assistance, Mayumi Yoneda for her secretarial work, and Helena Akiko Popiel for her critical reading of the manuscript. This work was supported in part by Grants-in-aid from the Ministry of Education, Science, Sports and Culture of Japan; and Grants from Takeda Chemical Co. The Osaka Bioscience Institute was founded in commemoration of the one hundredth anniversary of the municipal government of Osaka City, and is supported by Osaka City.