The Tumor-sensitive Calmodulin-like Protein Is a Specific Light Chain of Human Unconventional Myosin X*

Human calmodulin-like protein (CLP) is an epithelial-specific Ca2+-binding protein whose expression is strongly down-regulated in cancers. Like calmodulin, CLP is thought to regulate cellular processes via Ca2+-dependent interactions with specific target proteins. Using gel overlays, we identified a ∼210-kDa protein binding specifically and in a Ca2+-dependent manner to CLP, but not to calmodulin. Yeast two-hybrid screening yielded a CLP-interacting clone encoding the three light chain binding IQ motifs of human “unconventional” myosin X. Pull-down experiments showed CLP binding to the IQ domain to be direct and Ca2+-dependent. CLP interacted strongly with IQ motif 3 (K d ∼0.5 nm) as determined by surface plasmon resonance. Epitope-tagged myosin X was localized preferentially at the cell periphery in MCF-7 cells, and CLP colocalized with myosin X in these cells. Myosin X was able to coprecipitate CLP and, to a lesser extent, calmodulin from transfected COS-1 cells, indicating that CLP is a specific light chain of myosin Xin vivo. Because unconventional myosins participate in cellular processes ranging from membrane trafficking to signaling and cell motility, myosin X is an attractive CLP target. Altered myosin X regulation in (tumor) cells lacking CLP may have as yet unknown consequences for cell growth and differentiation.

Human calmodulin-like protein (CLP) is an epithelialspecific Ca 2؉ -binding protein whose expression is strongly down-regulated in cancers. Like calmodulin, CLP is thought to regulate cellular processes via Ca 2؉dependent interactions with specific target proteins. Using gel overlays, we identified a ϳ210-kDa protein binding specifically and in a Ca 2؉ -dependent manner to CLP, but not to calmodulin. Yeast two-hybrid screening yielded a CLP-interacting clone encoding the three light chain binding IQ motifs of human "unconventional" myosin X. Pull-down experiments showed CLP binding to the IQ domain to be direct and Ca 2؉ -dependent. CLP interacted strongly with IQ motif 3 (K d ϳ0.5 nM) as determined by surface plasmon resonance. Epitope-tagged myosin X was localized preferentially at the cell periphery in MCF-7 cells, and CLP colocalized with myosin X in these cells. Myosin X was able to coprecipitate CLP and, to a lesser extent, calmodulin from transfected COS-1 cells, indicating that CLP is a specific light chain of myosin X in vivo. Because unconventional myosins participate in cellular processes ranging from membrane trafficking to signaling and cell motility, myosin X is an attractive CLP target. Altered myosin X regulation in (tumor) cells lacking CLP may have as yet unknown consequences for cell growth and differentiation.
Human calmodulin-like protein (CLP) 1 is encoded by an intronless gene localized on chromosome 10p13-ter (1), a chromo-somal region known to be lost in a number of breast and other cancer types (2)(3)(4)(5)(6). Initial studies failed to show expression of the CLP gene in a number of tissues and cell types (7); however, in an independent study aimed at identifying mRNAs coding for transformation-sensitive proteins, CLP was recloned as a protein named NB1 (for Normal-Breast-1) from a normal primary breast epithelial cell line that had been subtracted with RNA purified from the same cell line after chemical transformation (8). CLP was subsequently shown to be expressed in a highly tissue-specific manner in basal and suprabasal epithelial cells of the normal breast, cervix, prostate, and skin, and to be down-regulated in naturally occurring breast, cervix, and prostate tumors as well as in artificially immortalized and transformed breast epithelial cells in culture (9). Recently, CLP was found to be down-regulated in 80% of primary breast tumors of both noninvasive as well as invasive and metastatic stages (10). Given the apparent correlation between its downregulation and the tumorigenic state, the available data support the notion that CLP expression may be incompatible with the transformed state of a cell. The mechanism responsible for this incompatibility is presently unknown, as is the physiological function of CLP, though they are likely related.
CLP's deduced amino acid sequence of 148 residues is identical in length and shows 85% identity to calmodulin (CaM) (7,11). Accordingly, CLP shows a number of characteristics similar to those of CaM; however, it also displays unique features that suggest its functional divergence (11)(12)(13). For example, CLP displays a Ca 2ϩ -dependent electrophoretic mobility shift that is typical of CaM, and it binds four Ca 2ϩ as does CaM, albeit with an ϳ8-fold lower affinity (11,12). Purified CLP is able to fully activate several known CaM-regulated enzymes, including CaM kinase II and cyclic nucleotide phosphodiesterase (although the K m for CLP activation of the phosphodiesterase is higher than that for CaM). By contrast, CLP is ineffective at stimulating other CaM-dependent enzymes such as calcineurin, NO synthase, and smooth muscle myosin light chain kinase (11,13). At present there are no known CaM targets that show higher affinity for CLP than for CaM. On the other hand, the unique expression pattern of CLP and its possible involvement in specific cellular processes (e.g. cell differentiation) raise the possibility that it may have its own distinct targets. Knowing the identity of the proteins with which CLP normally interacts will greatly aid in understanding the normal physiological function(s) of CLP as well as the consequences of its down-regulation in transformed cells. Here, we therefore set out to identify such proteins using gel overlays and yeast two-hybrid interaction screening, and report the cloning and characterization of a human unconventional myosin (myosin X) as a specific target of CLP.

Expression and Purification of Recombinant Human CLP-Human
CLP was expressed in Escherichia coli JM109(DE3) from plasmid pGem-CLP and purified as described (14). Site-directed mutagenesis (15) was performed to generate an expression vector for CLP F99Y in which phenylalanine 99 was replaced by tyrosine. A 475-bp EcoRI/ BamHI cassette from pGem-CLP was first subcloned into M13mp18. Recombinant phage were propagated in E. coli RZ1032, and uracilcontaining single-stranded DNA was isolated for in vitro hybridization to the mutant oligonucleotide 5Ј-d(GCG CTG ACG TAG CCG TTG CC)-3Ј. The hybrid was elongated by T7 DNA polymerase (Sequenase 2.0, United States Biochemical Corp.) ligated by T4 ligase and transformed into E. coli JM101. The mutated cassette was then excised from double-stranded replicative form DNA of an appropriate M13 clone and religated into EcoRI/BamHI-digested pGem-CLP, yielding expression vector pGem-CLP F99Y . CLP F99Y was expressed as described (14).
CLP Overlays-CLP F99Y was radiolabeled using "Enzymobeads" according to the manufacturer's instructions (Bio-Rad, Brussels, Belgium) in 50 mM MOPS (pH 7.5), 100 mM NaCl, 2 mM CaCl 2 , 0.1% glucose (w/v), and 1 mCi of Na 125 I/100 g of recombinant protein. The specific activity was ϳ1.5 ϫ 10 6 cpm/g protein. For biotin labeling, purified recombinant CLP was dissolved at 12 mg/ml in 20 mM HEPES (pH 7.0), 5 mM ␤-mercaptoethanol, 5 mM CaCl 2 , and biotin X-NHS (Calbiochem, San Diego, CA) was dissolved at 20 mg/ml in Me 2 SO. Biotin solution was added to the CLP solution to 20% (v/v), and the mixture incubated for 1 h at room temperature. The reaction was then quenched by adding ammonium acetate to 1 M, and labeled CLP was separated from free label by gel filtration using Sephadex G-25 (Sigma). Breast and cervix tissues from biopsy specimens, and lung from autopsy were obtained from Drs. C. Moll and R. Caduff (Department of Pathology, University Hospital, Zurich, Switzerland). 1 g of tissue was minced with a razor blade and homogenized with a Polytron at 4°C in 4 ml of 10 mM Tris-HCl (pH 7.0), 150 mM KCl, 10 mM 1,4-dithio-DL-threitol, 2 mM EDTA, and 0.5 mM phenylmethylsulfonyl fluoride. The homogenate was passed through four layers of cheesecloth and centrifuged for 1 h at 24,000 ϫ g and 4°C. The pellet was resuspended in the same buffer and stored at Ϫ20°C until use. For overlays, protein samples were run in SDS-PAGE gels and blotted to PVDF membranes (Millipore, Bedford, MA) using standard Western blotting procedures (16). Blotted membranes were incubated for 5 min in CaTBST (2 mM CaCl 2 , 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.05% Tween 20) ϩ 1.5% H 2 O 2 , and blocked overnight in 5% nonfat dry milk in CaTBST. Membranes were probed in 10 5 cpm/ml (5-10 nM) of iodinated CLP F99Y or in 75 nM biotinylated CLP in CaTBST for 90 min, and then washed four times (5 min each time) in CaTBST, and exposed to Fuji x-ray film at Ϫ70°C if 125 I-labeled CLP F99Y was used. If biotin-labeled CLP was used, the blot was probed with biotin-avidin-horseradish peroxidase (1:3000, Amersham Pharmacia Biotech), washed as before, and developed using Renaissance TM chemiluminescent reagent (PerkinElmer Life Sciences).
Plasmid Construction-pAS2-CLP and pAS2-CaM were made by subcloning a NcoI/XbaI fragment carrying the entire coding sequence from pKK-CLP and pKK-CaM, respectively (11), into pAS2-1 (CLONTECH, Palo Alto, CA). pGADGH-IQ123K corresponds to the original plasmid isolated from the HeLa Matchmaker cDNA library (CLONTECH). It contains nt 2333-2674 of the human myosin X sequence (accession no. AF234532) fused with a linker of sequence 5Ј-d(GAA TTC GGC ACG AG)-3Ј at the 5Ј end and 5Ј-d(GAG CTC (T) 15 )-3Ј at the 3Ј end to the pGADGH vector at cloning sites EcoRI and XhoI, respectively. The resulting vector expresses amino acids 705-817 of myosin X as a Cterminal fusion protein with the yeast Gal4 activation domain. In addition, the myosin X sequence is followed by lysines resulting from the oligo(dT) primer used in library construction as well as a few extra C-terminal residues encoded by the vector. pGADGH-IQ0 was constructed by deletion of an ApaI fragment from this construct and expressed amino acids 705-751 of myosin X in frame with the Gal4 activation domain. pGADGH-IQ1, -IQ12, and -IQ123 were constructed by amplifying appropriate fragments from pGADGH-IQ123K using a pGADGH-specific upstream primer and downstream primers terminating at nt 2511 (IQ1), nt 2579 (IQ12), or nt 2674 (IQ123). The resulting PCR products were cut with BamHI (cuts within the upstream pGADGH sequence), and then cloned into pBluescript, which had been cut with EcoRV, T-tailed, and BamHI cut. The BamHI/XhoI inserts from this vector were then transferred to pGADGH. The resulting constructs expressed amino acids 705-763 (IQ1), 705-785 (IQ12), or 705-817 (IQ123) of myosin X in fusion with the Gal4 activation domain. pEGFP-HA-His and pEBFP-HA-His were constructed by inserting a double-stranded oligonucleotide with the (top strand) sequence 5Ј-d(p-GAT CCT ACC CCT ATG ATG TGC CTG ACT ATG CCC ACC ACC ACC ACC ACC ACC TGG TGC CCA GGG GCA GCA)-3Ј into the BspEI site of plasmids pEGFP and pEBFP (CLONTECH), thereby reconstituting the BspEI site at the 3Ј end only, and introducing the peptide sequence SYPYDVPDYAHHHHHHLVPRGSR, which contains an HA tag followed by a His tag, to the C terminus of the GFP/BFP tag. The full-length myosin X coding sequence from nucleotide position 233 to 6405 (accession no. AF234532) was then cloned as a KpnI/XbaI fragment from pBluescript-MyoX into pEBFP-HA-His, creating plasmid pEBFP-HA-His-MyoX, which expresses the human myosin X with an N-terminal BFP, HA, and His tag. pEYFP-CLP and pEYFP-CaM were constructed by PCR amplification using pKK-CLP and pKK-CaM (11) as templates together with primers pKK233-2K (5Ј-d(GTG AGC GGA TAA CAG GTA CCC ACA GGA AAC AG)-3Ј) and pKK233-2r (5Ј-d(CCG CCA AAA CAG CCA AGC TTG CAT GCC TG)-3Ј). The KpnI/XmaI fragment from each PCR product was then ligated into pEYFP (CLON-TECH). In-frame GST fusion constructs GST-IQ1, GST-IQ12, and GST-IQ123 were made by amplifying sequences corresponding to nt 2344 -2511 (IQ1), 2344 -2579 (IQ12), or 2344 -2674 (IQ123) of myosin X from plasmid pGADGH-IQ123K, TA-cloning the resulting PCR fragments into pCR2.1 (Invitrogen), and subcloning the EcoRI insert fragments from this vector into pGEX-2TK (Amersham Pharmacia Biotech). All vector inserts were completely sequenced to confirm their integrity.
Yeast Two-hybrid Interaction Analysis-Yeast two-hybrid screening was performed according to the Matchmaker Two-Hybrid System 2 instructions (CLONTECH). Yeast were transformed using the lithium acetate method (17) with sheared herring sperm DNA as carrier. Selection of initial positives was done after 14 days of growth on SD/ϪTrp/ ϪLeu/ϪHis agar plates. 5 mM or 40 mM 3-aminotriazole (Sigma) was added to all ϪHis media used for yeast strains CG1945 or Y190, respectively. Positive clones were propagated on SD/ϪTrp/ϪLeu media. Yeast media components were from CLONTECH and Difco. ␤-Galactosidase assays were performed by freeze-thaw of yeast colonies on filter lifts of 3-day-old streak plates. Bait plasmid was dropped out by cycloheximide counterselection, using concentrations of 1 and 10 g/ml for strains CG1945 and Y190, respectively. Yeast matings were performed in liquid culture using microtiter plate volumes and yeast strains Y187 (mating type ␣) and either Y190 or CG1945 (a-type). Plasmid inserts from positive yeast clones were directly amplified after picking 3-dayold yeast colonies into TE containing 0.25 unit/l lyticase (Sigma). Following 2 h at 37°C, the yeast were subjected to three rounds of freeze-thaw and heated to 95°C for 5 min. They were then chilled on ice, pelleted, and the supernatant used as template for PCR with pGADGH-specific primers. Library plasmids were prepared from counterselected yeast containing only "prey" plasmids by an alkaline lysis method (18) and transformed directly into E. coli strain DH5␣.
Cloning of Full-length Human Myosin X-5Ј-Rapid amplification of cDNA ends was performed on HeLa cell Marathon-Ready cDNA as recommended by the supplier (CLONTECH), using nested primers 5Ј-d(GTC TTC CCC AGC TGC CAC TCG CTG TTG G)-3Ј and 5Ј-d(AGG CAT CAT AGA GCT GCA GCA GGC TCG)-3Ј based on the sequence from the yeast two-hybrid clones. This resulted in a ϳ2-kb PCR product, which was sequenced directly using a primer walking strategy and extended through the initiator codon and ϳ30 bp of the 5Ј-untranslated region of the myosin X cDNA. The entire 6-kb human myosin X open reading frame (ORF) was then PCR-amplified using information from this sequence and from human ESTs homologous to bovine myosin X sequence. Because the 6-kb PCR product proved resistant to standard cloning techniques, we cloned the myosin X ORF cDNA in pieces, which were then reassembled. The 3Ј-most 2 kb of the ORF were amplified from Marathon-Ready cDNA using Taq DNA polymerase and cloned as an EcoRI/XbaI fragment into pBluescript. The 5Ј-terminal 4 kb of the ORF was amplified from Marathon-Ready cDNA using Pwo DNA polymerase (Roche Molecular Biochemicals) and cloned as a KpnI/EcoRI fragment into pBluescript. This fragment contained two mutations, one of which resulted in a frameshift and was corrected by site-directed mutagenesis using the GeneEditor kit (Promega, Madison, WI). The 3Ј fragment was then subcloned into the EcoRI and XbaI sites of this vector. The resulting full-length cDNA in pBluescript (pBluescript-MyoX) contains one discordance from the original sequence (accession no. AF234532), an A 3 G transition at base 2261, which results in an arginine for glutamine substitution at amino acid 679. During all PCR reactions, DNA polymerase was not added until the tubes reached 94°C (hot start PCR).
GST Pull-down Assays-GST and GST fusion proteins were expressed in E. coli BL21(DE3) from pGEX-2TK derived vectors upon isopropyl-1thio-␤-D-galactopyranoside induction as described (16). Cells were pelleted, resuspended in Tris-buffered saline (50 mM Tris-HCl (pH 7.4), 150 mM NaCl) plus protease inhibitors (0.2 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 2 g/ml leupeptin, 0.2 g/ml aprotinin, 10 mM EDTA) and 30 mM ␤-mercaptoethanol, and lysed by the addition of sarkosyl (Curtis-Matheson Scientific, Houston, TX) to a final concentration of 1.5%. After 15 min on ice, the lysate was cleared by centrifugation at 15,000 ϫ g and supplemented by the addition of Triton X-100 to 3%. This lysate was then bound to glutathione-Sepharose (Sigma) and washed with TBST and Tris-buffered saline. The quantity of bound fusion protein was estimated by Coomassie Blue staining of SDS-PAGE gels of known amounts of fusion protein-containing glutathione-Sepharose beads, and glutathione-Sepharose was added to equalize the amount of fusion protein/ml of beads among various fusions. Varying amounts of CLP or CaM were mixed with glutathione Sepharose beads containing equal quantities (ϳ5 g) of GST-IQ123, GST-IQ12, GST-IQ1, or GST alone in the presence of 5 mM CaCl 2 (Ca) or 5 mM EDTA in TBST and allowed to bind overnight at 4°C. The beads were washed four times with CaTBST or with 5 mM EDTA in TBST and solubilized in Laemmli buffer (16). These samples were then analyzed by Western blot for CLP using affinity-purified rabbit anti-human CLP antibody TG-7, or for CLP/CaM using unpurified TG-7 antiserum, which recognizes both CLP and CaM equally well (10).
Surface Plasmon Resonance Analysis-Surface plasmon resonance analysis was performed on a BIAcore 1000 instrument using Sensor Chip SA (streptavidin) chips (Biacore Inc, Piscataway, NJ). Affinity was measured for the N-terminally biotinylated IQ3 peptide (biotin-RFLHLKKAAIVFQKQLRGQIARRVYRQ-NH 2 ) synthesized in the Mayo peptide core facility. Approximately 1000 RU of peptide were bound to the chip, and varying CLP and CaM concentrations were passed over the chip. The concentration of purified CLP (14) was measured by UV absorbance at 276 nm using a molar extinction coefficient of 1500 M Ϫ1 (11). Bovine brain CaM was from Calbiochem (San Diego, CA) and quantitated by UV absorbance at 276 nm using a molar extinction coefficient of 3300 M Ϫ1 (19). CLP and CaM were diluted from 0.1 mM stocks immediately prior to sensorgram runs. All sensorgram runs were performed in a running buffer of 50 mM HEPES (pH 7.4), 150 mM NaCl, 1 mM ␤-mercaptoethanol, 0.02% w/v NaN 3 , 0.005% Surfactant P-20 (Biacore Inc.), and 1 mM CaCl 2 . Regeneration was performed using running buffer with 2 mM EDTA in place of CaCl 2 and was complete as determined by comparing blank buffer injections to 1 M injections of CLP or CaM. Sensorgrams were blank subtracted and baseline-adjusted, then RU at equilibrium were measured. For each of three experiments, a Scatchard plot was done and the x-intercept of a least squares fitted line was calculated. Each point in that experiment was then normalized to the x-intercept (i.e. divided by the value of the x-intercept and multiplied by 100), all data points were pooled, and a Scatchard plot representing all three experiments was plotted. Least squares fitting was then used to calculate a best fit line, and the affinity of CLP for the IQ3 peptide was estimated from the slope of that line. Kinetic rate constants were not estimated because analysis of flow rate dependence of the apparent rate constants demonstrated that CLP binding was mass transfer limited up to a flow rate of 100 l/min.
Immunofluorescence-MCF-7 cells were grown on glass coverslips and transfected with purified plasmid DNA as described above. Transfected cells were grown for 2 days, then rinsed in DPBS (Life Technologies, Inc.) and fixed for 20 min in 0.1 M PIPES (pH 6.95), 1 mM EGTA, 3 mM MgSO 4 , 3% formaldehyde. Coverslips were then rinsed three times (3 min each time) in DPBS, permeabilized for 2 min in 0.2% Triton X-100 (Curtis-Matheson Scientific), and blocked for 1 h in block-ing buffer (5% v/v normal goat serum and 2% w/v BSA in DPBS). Coverslips were then incubated for 1 h in primary antibody (polyclonal rabbit anti-HA IgG, Upstate Biotechnology, Lake Placid, NY) diluted 1:50 (10 g/ml) in blocking buffer, rinsed three times in DPBS, and incubated for 1 h in secondary antibody (5 g/ml Alexa-350 goat antirabbit IgG; Molecular Probes, Eugene, OR) and 25 ng/ml rhodaminephalloidin (Sigma) in blocking buffer. Following three rinses in DPBS and a final rinse in dH 2 O, coverslips were mounted with ProLong (Molecular Probes) on glass slides.

Identification of a ϳ210-kDa CLP-binding Protein by Gel
Overlay-To identify specific CLP targets in total tissue extracts, we used a gel overlay method proven to be successful in the identification of CaM target proteins in various tissues and subcellular fractions (20 -22). In previous studies with CaM, the best results were obtained when 125 I-labeled protein was used as probe. Unfortunately, in CLP a Phe residue replaces the Tyr 99 iodinated in CaM, leading to a dramatic reduction in the success of CLP iodination. We therefore engineered (by site-directed mutagenesis) a CLP F99Y expression construct. Purified CLP F99Y was then labeled with 125 I to a high specific activity and used as probe in overlay experiments, assuming that the single conservative amino acid replacement would not lead to a major change in target binding specificity. This approach identified a relatively small number (two to four major bands) of putative CLP-binding proteins in human breast, cervix, and lung (Fig. 1). The most prominent was a protein with an estimated molecular mass of 210 kDa. This protein (p210) was also detected by biotinylated CLP (data not shown), confirming that detection does not result from the F99Y substitution. p210 reacted only weakly or not at all with CaM in the overlays (Fig. 1).
CLP-binding proteins, including a band for p210, were also detected in the 100,000 ϫ g microsomal membrane fraction from cultured HeLa cells ( Fig. 2A), as well as from several other normal and transformed epithelial cell lines (data not shown). Importantly, the 210-kDa protein interacted with CLP in a specific, Ca 2ϩ -dependent manner; labeled CLP could be competed off the blot with excess unlabeled CLP, but not with unlabeled CaM (Fig. 2), and binding was detected in the presence of Ca 2ϩ , but not in the presence of EDTA (data not shown).
Identification of a Specific CLP-interacting Protein by Yeast Two-hybrid Screening-We turned to the yeast two-hybrid approach (23) to identify CLP interacting proteins and to potentially clone the p210 protein. We used the Matchmaker TM yeast two-hybrid system (CLONTECH) to screen an oligo(dT)primed, directionally cloned HeLa cell cDNA library with a CLP "bait" fused in frame to the Gal4 DNA binding domain in the plasmid pAS2-1 (CLONTECH). Screening 1.7 ϫ 10 7 initial transformants yielded 1070 colonies, of which 266 were ␤-ga- lactosidase positive in a colony lift assay. Following cycloheximide counterselection to cure strains of the bait plasmid, two colonies for each ␤-galactosidase-positive transformant were mated against Y187 yeast cells containing bait plasmids coding for CLP or a laminin fusion protein (negative control bait). Library plasmids remaining positive for interaction with CLP and testing negative for interaction with laminin were rescued, and their inserts were amplified by PCR and sequenced. 18 of the library plasmids positive for CLP interaction contained a ϳ350-bp insert whose sequence showed high similarity (86.3% identity) to a cDNA sequence for bovine myosin X (GenBank accession no. U55042). This insert obviously represented the cDNA of the corresponding human myosin X, which had not yet been cloned. Each of the insert sequences was primed from an internal A-rich stretch of the myosin X RNA and encoded a region mainly made up of the three IQ domains of the myosin X heavy chain. IQ domains are named for the isoleucine-glutamine (single-letter code IQ) pair, which is part of the signature sequence characterizing these domains (24). Importantly, IQ domains are known binding sites for calmodulin (and for other Ca 2ϩ binding light chains) in several "unconventional" myosins (24,25).
Using sequence information from the original yeast two hybrid clone, we performed a 5Ј-rapid amplification of cDNA ends using HeLa cell Marathon-Ready cDNA (CLONTECH). The resulting PCR product contained the entire N-terminal coding sequence of human myosin X including the initiator codon and ϳ30 bp of 5Ј-untranslated region. Using this sequence and sequences present in the human expressed sequence tag data base homologous to the bovine myosin X sequence, we successfully PCR-amplified the entire 6-kb human myosin X ORF (accession no. AF234532). Comparison of this cDNA with subsequently posted human genomic sequence revealed no discordances and allowed for the mapping of the intron-exon structure (Fig. 3). The human myosin X heavy chain contains 2058 amino acids with a calculated molecular mass of 238 kDa. Amino acid sequence comparison of the human with the murine (accession no. CAB56466) and bovine (accession no. T18519) myosin X heavy chain amino acid sequence revealed 96% and 93% identity, respectively.
Correspondence of p210 to Myosin X and Confirmation of Interaction with CLP-Identification of myosin X as a potential CLP interactor suggested that myosin X and p210 may be the same protein. Several lines of evidence support this notion. The calculated molecular mass of myosin X (238 kDa) matches well with that of p210, considering that the M r of p210 was estimated based on its migration in an SDS-polyacrylamide gel with only few marker proteins in the high molecular weight range. In addition, we tested extracts from yeast expressing the original Gal4AD-myosin X "prey" fusion protein for reactivity to biotinylated CLP. These extracts show a CLP reactive band of 36 kDa, which is the expected size for the Gal4AD-myosin X IQ-domain fusion protein. This band shows a similar affinity for CLP and CaM as does p210 as determined by competitive overlay (Fig. 2B). Taken together these data strongly suggest that myosin X is at least one component and most likely the only component of p210.
We next characterized CLP's interaction with myosin X in more detail. Coexpression of BFP-HA-His-tagged myosin X and YFP-tagged CLP in MCF-7 cells showed intense localization at the cell periphery for both proteins with some cytoplasmic localization also evident (Fig. 4B). By contrast, YFP-CaM demonstrated general cytoplasmic and nuclear localization and much poorer colocalization with myosin X (Fig. 4C). This mostly cytoplasmic and nuclear localization was also evident in cells that were singly transfected with YFP-CLP (Fig. 4A), suggesting that CLP localization to the plasma membrane was enhanced and perhaps "driven" by coexpression of myosin X.
CLP and, to a lesser extent, CaM coprecipitated with myosin X (Fig. 5). BFP-HA-His-tagged myosin X or control BFP-HA-His were coexpressed with YFP-CLP, YFP-CaM, or YFP in transiently transfected COS-1 cells. BFP-HA-His-myosin X or control BFP-HA-His were then precipitated with Ni 2ϩ -NTAagarose or with immobilized anti-HA antibody. Equivalent amounts of precipitated BFP-HA-His-myosin X and BFP-HA-His were loaded onto SDS-PAGE gels, and these precipitates were analyzed by Western blot using an anti-GFP antibody that recognizes both BFP and YFP. CLP and, to a lesser extent, CaM coprecipitated with myosin X (Fig. 5B, lanes 2 and 3) while no YFP was precipitated (Fig. 5B, lane 1) even though equivalent amounts of all three proteins were expressed in total cell lysates (Fig. 5A). On the other hand, when BFP-HA-His was precipitated, essentially no YFP-CLP or YFP-CaM were coprecipitated (Fig. 5B, lanes 4 and 5). An interesting observation was the detection of a second, more slowly migrating YFP-CLP band, which coprecipitated with myosin X (Fig.  5B, lane 3).
CLP Binds with High Affinity to IQ3 of Myosin X-We next wished to determine which region of myosin X binds specifically to CLP. The yeast two-hybrid screen strongly suggested that the IQ motifs are involved in CLP binding. This is because the original yeast two-hybrid clone (pGADGH-IQ123K) only encodes a small portion of the head domain, all three IQ motifs, and a few amino acid residues of the coiled coil, followed by an artificial polylysine stretch introduced by the oligo(dT) primer used in library construction. Yeast two-hybrid vectors were constructed containing about half of the first IQ motif (pGADGH-IQ0, amino acids 705-751 of hMyoX), the first IQ motif (pGADGH-IQ1, amino acids 705-763 of hMyoX), the first two IQ motifs (pGADGH-IQ12, amino acids 705-786 of hMyoX), and all three IQ motifs plus eight amino acids of the coiled coil domain (pGADGH-IQ123). Y190 or CG1945 cells were cotransformed with these vectors and either pAS2-CLP or pAS2-CaM baits and assayed for growth on SD/ϪTrp/ϪLeu/ ϪHis medium or for ␤-galactosidase expression in a filter lift assay (Fig. 6). In these experiments a strong association was only observed between CLP and IQ123 or IQ123K. In the less stringent Y190 yeast strain, we observed weak interaction between CaM and IQ12, IQ123, or IQ123K as well as between CLP and IQ12. Similar results were obtained when single pGADGH-IQ "prey" transformants were mated with Y187 cells containing either pAS2-CLP or pAS2-CaM.
These results suggested that the third IQ motif was critical for CLP binding. We used GST fusion proteins in a pull-down experiment to demonstrate direct interaction between CLP and the myosin X IQ motifs (Fig. 7). GST-IQ1, GST-IQ12, GST-IQ123, or GST alone as control was incubated with bacterial lysates containing various concentrations of CaM or CLP. At high CLP and CaM concentrations (5 M) strong CLP binding to GST-IQ123 was observed and weaker binding of CaM was also observed. Both of these binding events were Ca 2ϩ -dependent as complex formation was inhibited in the presence of EDTA (Fig. 7B). At high CLP concentrations (500 nM), CLP bound to all three fusion constructs, but as CLP concentration was reduced, binding was lost, first from GST-IQ1 and then from GST-IQ12. At the lowest CLP concentration tested (5 nM), only weak binding of CLP to GST-IQ12 was observed, whereas strong binding to GST-IQ123 was still observed (Fig. 7A). In no case did we observe CLP or CaM binding to GST alone (Fig.  7A). These results indicate that CLP, and to a lesser extent CaM, can bind directly to the IQ motifs of myosin X, that this binding is Ca 2ϩ -sensitive, and that the third IQ motif is required for the highest affinity CLP binding.
Finally, we used surface plasmon resonance to measure the affinity of CLP for a peptide corresponding to the third IQ motif of myosin X (Fig. 8). Using the known extinction coefficients for CLP (11) and CaM (19), dilutions of these proteins were made and their binding to the biotinylated IQ3 peptide immobilized on a streptavidin surface chip was measured in a BIAcore 1000 instrument (Fig. 8A). After blank subtraction, these values were used to calculate a binding affinity of CLP for the IQ3 peptide using a Scatchard plot (Fig. 8B). The resulting value of 0.5 nM for the K d gives an indication of the affinity that CLP likely has for the intact protein. In addition, we measured the affinity of CaM for the IQ3 peptide and found that CaM binds to the peptide-coated chip surface with an affinity of ϳ1 M.

DISCUSSION
To test the hypothesis that specific targets for CLP exist, we performed overlay experiments on different human tissue extracts including breast, lung, and cervix. These overlays demonstrated the existence of a prominent band at ϳ210 kDa (p210) that bound to labeled CLP, but not to labeled CaM. Competition experiments using an excess of unlabeled CLP or unlabeled CaM further demonstrated the specificity of the CLP-p210 interaction. Yeast two-hybrid screening identified human myosin X as a CLP interactor and suggested its identity to p210. In hindsight, this finding is perhaps not surprising given the known association of CaM with a number of unconventional myosins (28). Early overlay experiments had demonstrated weak interaction between CLP and the chicken skeletal muscle myosin used as a marker for gel electrophoresis. In addition, some binding of CLP to purified chicken brush border myosin I was also observed, though this binding could be effectively competed away by CaM. 2 Myosin X was originally identified by PCR screening of cDNAs using degenerate primers specific for the myosin head domain (29,30). The resulting cDNA fragments included one for a novel myosin named myosin X. This sequence was subsequently localized to chromosome 15 in mice and to chromosome 5 in humans (31). Very recently, the cloning and initial characterization of full-length myosin X from mice, cows, and humans has been reported (26,27). Although little is known about myosin X function, some possibilities have been inferred from its domain structure (see Refs. 26 and 27; Fig. 3) and its relationship to other myosins. Like other myosins, myosin X is a molecular motor able to translocate actin filaments in an in vitro motility assay (32) and it cosediments with F-actin in an ATP-dependent manner (27). Like myosins VII and XV (28), myosin X also contains a FERM domain (band 4.1/ezrin/radixin/moesin, also known as talin homology domain), which appears to be involved in interaction with transmembrane proteins linked to the cytoskeleton. In mammals, myosins VII and XV are implicated in deafness syndromes in which deafness is caused by a disturbance of the stereociliar cytoskeleton (33,34). An involvement of myosin X in actin cytoskeletal organization is therefore plausible and has gained recent support through the demonstration of colocalization of endogenous myosin X with actin bundles in lamellopodia and membrane ruffles in Madin-Darby bovine kidney cells as well as in several other cell lines (27). In addition, the data reported by Berg et al. (27) suggest that myosin X may relocate with dynamic actin redistribution. The localization we observed of recombinant myosin X (and of CLP) in transiently cotransfected MCF7 cells (Fig. 4) matches very well with that reported for the endogenous myosin X. In addition to the FERM domain, myosin X contains three pleckstrin homology domains (27). Pleckstrin homology domains are known to interact with phosphatidylinositol phosphates, and these lipids are concentrated at the ruffled edge of chemotactic cells, among other places (35). The notable biochemical property of p210 (alias myosin X) of poor solubility in aqueous buffers without detergents fits well with a possible membrane lipid interaction of this protein.
The interaction of myosin X with CLP was confirmed by a number of methods. In cells cotransfected with tagged myosin X and CLP, CLP precipitated along with myosin X when the latter was pulled down with either anti-HA beads or Ni 2ϩ -NTA beads. This indicates that the two full-length proteins can bind to each other in vivo. We consistently observed a second CLP band coprecipitating with myosin X (Fig. 5). Although Ca 2ϩ -CLP runs differently than apo-CLP in SDS-PAGE gels (11), the additional band is unlikely to represent a Ca 2ϩ -bound form of CLP because the sample loading buffer contained 5 mM EDTA, which is sufficient to cause CLP and CaM to run in their Ca 2ϩ -free form (11). Mobility shifts such as the one we observed are common among phosphorylated proteins, and it is therefore possible that a fraction of the myosin X-bound CLP is phosphorylated. Indeed, CLP shares with CaM several serine and threonine residues (Thr 79 , Ser 101 , Thr 117 ) as well as a tyrosine residue (Tyr 138 ) known to be targets for phosphorylation in CaM (36 -38). A substantial fraction of CaM tightly associated with the plasma membrane and underlying cytoskeleton has been shown to be phosphorylated by membrane-associated protein kinases (39). Assuming that CLP can be similarly phosphorylated, and given the cytoskeletal/membrane-associated location of myosin X, it may well be that a significant fraction of the myosin X-bound CLP in vivo is phosphorylated. Both yeast two-hybrid and GST pull-down experiments implicate IQ3 as a major CLP binding site. Although the two-hybrid screen indicated that IQ2 can also bind some CLP, the pulldown assay allowed us to estimate more closely the affinity of CLP for IQ1 and IQ2. In particular, we observed loss of CLP binding to GST-IQ12 as we reduced the CLP concentration from 50 to 5 nM, suggesting an affinity of CLP for IQ2 on the order of tens of nanomolar. Using surface plasmon resonance we found CLP's affinity for a synthetic IQ3 peptide to be 0.5 Ϯ 0.2 nM, which is in the range of high affinity CaM targets (40) and unequivocally confirms that myosin X is a bona fide CLP interacting protein. The affinity of CLP for myosin X as estimated by competitive filter overlay is somewhat lower than this (low nanomolar). The overlay technique, however, typically underestimates true affinities (20). In addition, affinities measured between peptides and CaM typically overestimate CaM affinities for the parent protein. For example, CaM binds to the intact plasma membrane calcium ATPase with 10-fold lower affinity than it binds to the CaM binding peptide from that protein (41,42). Thus, the actual affinity of CLP for intact myosin X will be less (i.e. the K d will be higher) than 0.5 nM, but still in the low nanomolar range.
It should be noted that full-length myosin X (and its IQ123 domain) can also bind CaM (as demonstrated by coprecipitation and yeast two-hybrid mating assays), albeit with reduced affinity when compared with CLP. The interaction of CLP with the myosin X IQ domains is Ca 2ϩ -dependent. This contrasts with the binding of CaM to some other myosins, which is often Ca 2ϩ -independent and frequently inhibited by Ca 2ϩ binding (25). Thus, the conditions used to analyze CLP binding may not be optimal for CaM interaction with myosin X. Because there are three IQ motifs, each potentially able to bind a CaM-like light chain, the "occupancy" state and actual light chain composition of myosin X may be complex and may change depending on the free Ca 2ϩ and relative concentration of potential light chains (e.g. CaM versus CLP). The distribution of myosin X has recently been investigated in murine, bovine, and human tissues (26,27), indicating that myosin X is fairly ubiquitous but generally expressed at low levels. Because CLP expression is confined to epithelial cells, it obviously cannot be the sole light chain of myosin X. In cells that express significant amounts of CLP (e.g. mammary epithelial cells, keratinocytes), CLP may well serve as one of the light chains of myosin X, whereas CaM (and/or additional, yet to be discovered light chains) will likely fulfill this purpose in other cell types. Indeed, several other CaM-related proteins have recently been discovered that serve as light chains of unconventional myosins. For example, the CaM-related EF-hand protein Mlc1p has been FIG. 6. Identification of CLP binding sites on myosin X by yeast twohybrid analysis. Top, Y190 and CG1945 cells were cotransformed with the indicated plasmids, and growth on SD/ϪTrp/ ϪLeu/ϪHis/ϩ3-aminotriazole medium was scored after 4 days as no colonies (Ϫ), very small/few colonies (ϩ/Ϫ), some colonies (ϩ), and many colonies (ϩϩϩ). Bottom, Y190 and CG1945 cells were cotransformed with the same plasmids as above, grown for 2 days on SD/ϪTrp/ϪLeu medium, and then assayed for ␤-galactosidase on filter lifts. FIG. 7. Demonstration of direct CLP-myosin X interaction by GST pull-downs. A, 5 g of purified GST alone or of different GST-IQ fusion proteins (indicated on top of the gel) were mixed with 1 g of CLP in a final volume of 12 ml of CaTBST for a final protein concentration of 5 nM CLP. After rocking overnight, the beads were collected and washed four times with CaTBST, and one-fifth of the total was analyzed by Western blot using anti-CLP antibodies. B, 5 g of GST-IQ123 were mixed with 50 g of CLP or CaM in a final volume of 600 l of CaTBST (Ca) or 5 mM EDTA in TBST (E) for a final CLP or CaM concentration of 5 M. After washing, one-fifth of the total was analyzed by Western blot using unpurified anti-CLP antiserum TG7, which recognizes CLP and CaM equally well.
FIG. 8. Surface plasmon resonance analysis of the binding of CLP and CaM to the IQ3 peptide of myosin X. Biacore analysis was performed on a BIAcore 1000 instrument using streptavidin chips coated with biotinylated IQ3 peptide. A, typical sensorgrams using 1000 and 100 nM CLP and CaM. B, Scatchard plot of CLP-IQ3 binding using the equilibrium phase RU for amount bound. For each of three experiments (diamond, triangle and square symbols), a Scatchard plot was done and the x-intercept of a least squares fitted line was calculated. Each point in that experiment was then normalized to that x-intercept, all data points were pooled, and a Scatchard plot containing all three experiments was plotted as shown. Least squares fitting was used to calculate the best fit lines shown.
shown to function as a light chain for Myo2p in Saccharomyces cerevisiae (43) and MICLC (another CaM-related "noncalmodulin" protein) has been identified as a light chain for Acanthamoeba myosin IC (44). In addition, chicken myosin V binds to at least two other noncalmodulin proteins that likely correspond to light chains (45). On the other hand, the abundance of CLP in several epithelial cells (e.g. skin or breast) seems to exceed the level required if CLP were to serve only as a light chain for the much less abundant myosin X. It is therefore possible that additional specific target proteins besides myosin X exist for CLP in these specialized tissues.
The identification of myosin X as a specific CLP target represents an important step toward the elucidation of CLP's biological function and potential involvement in tumor suppression. As mentioned above, the domain structure of myosin X suggests a role in actin organization at ruffling membrane edges. As these are frequently associated with chemotactic responses, it may well be that myosin X is involved in the organization of cellular membranes in response to specific cellular signals. In this context it is of interest that keratinocytes exhibit a rapid chemotactic response, which is decreased upon differentiation, a time when CLP is up-regulated in these cells. This, along with the observation that CLP is down-regulated in tumors, suggests that CLP may function to inhibit the membrane cytoskeletal reorganization required for some types of cell motility, perhaps via its interaction with myosin X. The identification of myosin X as a CLP target represents a crucial step toward elucidating the reason for CLP's consistent downregulation in human cancers and may point the way toward novel therapeutic strategies.