Normal Development of Mice and Unimpaired Cell Adhesion/Cell Motility/Actin-based Cytoskeleton without Compensatory Up-regulation of Ezrin or Radixin in Moesin Gene Knockout*

Ezrin/radixin/moesin (ERM) proteins are general cross-linkers between the plasma membrane and actin filaments. Because their expression is regulated in a tissue-specific manner, each ERM protein has been proposed to have unique functions. On the other hand, experiments at the cellular level and in vitro have suggested their functional redundancy. To assess the possible unique functions of ERM proteins in vivo , the moesin gene located on the X chromosome was disrupted by gene targeting in embryonic stem cells. Male mice hemizygous for the mutation as well as homozygous females were completely devoid of moesin but developed normally and were fertile, with no obvious histological abnormalities in any of the tissues examined. In the tissues of the mutant mice, moesin completely disappeared without affecting the expression levels or subcellular distribution of ezrin and radixin. Also, in platelets, fibroblasts, and mast cells isolated from moe-sin-deficient mice, targeted disruption of the moesin gene

Three closely related proteins, ezrin, radixin, and moesin constitute a gene family called the ERM 1 family. Ezrin, radixin, and moesin were identified in different directions (1-7), but isolation and sequencing of their cDNAs revealed that they were closely related (amino acid sequence identity of 70 -80% in the mouse) (8 -12). Independently of the lines of study on ERM proteins, another ERM-like protein was identified as a tumor suppressor or hereditary neurofibromatosis type 2 and named merlin (moesin/ezrin/radixin-like protein) or schwannomin (13,14).
One of the important questions regarding ERM proteins that have not yet been addressed is the extent to which ezrin, radixin, and moesin are functionally redundant. Targeted disruption of ERM protein genes would be one of the most direct ways to approach this issue. Among the ERM proteins, we expected moesin to be rather functionally unique, partly because this molecule lacks the polyproline stretch found in both ezrin and radixin and partly because only moesin is not tyrosine phosphorylated by the epidermal growth factor receptor (39). Therefore, to address the redundancy problem in ERM proteins we generated mice with a targeted null mutation of the moesin gene located on the X chromosome.
Construction of Targeting Vector-Mouse moesin genomic clones were isolated from a 129/Sv mouse genomic library using mouse moesin cDNA (nucleotides 1-315) (11) as a probe. The targeting vector (see Fig.  1A) was constructed by standard recombinant DNA techniques; the 0.7-kb PstI-KpnI fragment containing the 3Ј part of exon 3 was deleted and replaced by a loxP-neo cassette in which loxP sequences (42,43) flank phosphoglycerate kinase-neo cassette (44) oriented in the opposite orientation to moesin transcription. In addition, a splicing acceptor (SA) sequence (45,46) and a polyadenylation signal (PA) [XhoI-PstI fragment of pCAGGS (47)] were ligated 3Ј of this cassette. The coding region of the diphtheria toxin A gene driven by the MCl promoter (DT-A) was then ligated downstream from the vector construct for negative selection against random integration of the vector (48). As shown in Fig. 1, the final vector construct consists of a 7.7-kb 5Ј-homologous region, loxP/phosphoglycerate kinase-neo/loxP/SA/PA, a 1.5-kb 3Ј-homologous region, and DT-A.
Disruption of Moesin Gene in Embryonic Stem Cells and Generation of Moesin-deficient Mice-Embryonic stem (ES) cells were electroporated with 20 g of linearized targeting vector DNA using a Bio-Rad Gene Pulser at 0.25 V and 960 microfarad. Cells were plated on feeder cells in normal growth medium for 36 -48 h, followed by selection with 175 g/ml G418. After 8 -10 days, G418-resistant colonies were picked up. The colonies were screened individually by cleaving genomic DNA (7 g) with BamHI and by probing the Southern blots with the 230-bp cDNA sequence in exons 4 and 5 downstream from the 3Ј homologous region (3Ј probe; see Fig. 1A). Correct targeting was confirmed by Southern blotting of BstXI-digested genomic DNA with the 50-bp cDNA in exon 2 upstream from the 5Ј homologous region (5Ј probe; see Fig.  1A). Targeted clones were also checked for single integration by hybridization with a neo probe. Two correctly targeted ES cell clones (clones 145 and 199) were expanded and injected into the blastocysts from C57BL/6 mice, which were then transferred into the uteri of pseudopregnant ICR recipients. Male chimeras with extensive ES cell contributions to their coats were bred with C57BL/6 female mice. Tail DNA from agouti F1 offspring was genotyped by Southern blotting analysis. F1 heterozygous females and F1 wild-type males were interbred and the littermates were genotyped.
Immunoblotting-Protein extracts from ES cells (ϩ/Y, Ϫ/Y) and various tissues of wild-type and moesin-deficient mice were separated by SDS-polyacrylamide gel electrophoresis (10%) and then electrophoretically transferred from gels onto nitrocellulose membranes, followed by incubation with antibodies. For antibody detection, a blotting detection kit with biotinylated immunoglobulin and streptavidine-conjugated alkaline phosphatase (Amersham Pharmacia Biotech) was used.
Immunofluorescence Microscopy-Immunofluorescence microscopy of frozen tissue sections (10 m) and embryonic fibroblasts were performed as described previously (49). They were examined using confocal imaging system (Bio-Rad) equipped with a Zeiss Axiophot II photomicroscope (Carl Zeiss, Oberkochen, Germany).
Platelet Aggregation Assay-Blood was collected by cardiac puncture on sodium citrate (0.38%), mixed with an equal volume of suspension buffer (20 mM HEPES, 140 mM NaCl, 5 mM KCl, 5 mM MgCl 2 , 0.38% sodium citrate, pH 7.4) and then centrifuged at 120 ϫ g for 10 min at room temperature. The supernatant was collected as platelet-rich plasma, which was diluted to 3 ϫ 10 5 platelets/l with plasma. Platelet aggregation was assayed as described previously (50).
Mast Cell Culture-Mast cells were isolated according to the technique described previously (51). Briefly, bone marrow was carefully taken out from the femurs of wild-type or moesin-deficient mice and suspended in ␣-minimum essential medium containing 10% fetal calf serum and 40 units/ml recombinant murine interleukin-3. The cell suspensions were placed on culture dishes for 7 days, and nonadherent cells were transferred into fresh medium in new culture dishes every 4 days. After 6 -7 weeks of culture, homogeneous populations (Ͼ90%) of bone marrow-derived mast cells were obtained. They were placed on poly-L-lysine coated coverslips and then fixed with 2.5% glutaraldehyde and 2% formaldehyde in 0.1 M cacodylate buffer (pH 7.2), and processed for the scanning electron microscopic observation.

Targeted Disruption of Moesin Gene in Embryonic Stem
Cells-The mouse genomic clone used for construction of the targeting vector is shown in Fig. 1A. Judging from the structure of the human moesin gene consisting of 12 exons (52), this mouse clone was thought to contain exons 2-5. The targeting vector was designed with the expectation that homologous re-combination between the vector and the moesin gene would result in deletion of a 3Ј part of exon 3. Because the numbers of nucleotides in exon 1 (12 bp from ATG), exon 2 (84 bp) and exon 3 (96 bp) are multiples of 3, it was possible that an aberrant transcript would be produced from the targeted allele by skipping the disrupted exon 3. To minimize this possibility, a SA and PA were placed at the 3Ј end of LoxP-neo cassette.
The linearized targeting vector was introduced into J1 ES cells by electroporation, and cells were selected with G418. To screen for homologous recombination events, DNA from resistant clones were subjected to Southern blotting analysis with a probe corresponding to a sequence 3Ј of the recombination site (3Ј probe). The wild-type moesin allele displayed a 9.2-kb band on Southern blotting of BamHI-digested DNA, whereas the disrupted locus showed a 5.0-kb band (Fig. 1B). Correct targeting was confirmed by Southern blotting with a 5Ј probe, and targeted clones were also checked for single integration by hybridization with a neo probe. Of 430 G418-resistant clones examined, two had undergone a single homologous recombination event (clones 145 and 199) (Fig. 1B). As the moesin gene is located on the X chromosome (52,53) and ES cells were established from male mice, the moesin single knockout cells were expected to be deficient in moesin. Immunoblotting analysis was then performed using two different pAbs, TK88 and TK89, which recognized NH 2 -and COOH-terminal halves of all ERM proteins, respectively (Fig. 1C). In wild-type ES cells, both pAbs recognized three bands around 80 kDa, which corresponded to ezrin, radixin, and moesin from the top, whereas both clones 145 and 199 lacked the moesin band in TK88 as well as TK89 immunoblotting. In these moesin single knockout cells, additional bands did not appear at the smaller molecular mass region in TK88 or TK89 immunoblotting, confirming that the moesin single knockout cells are deficient in moesin. This was further confirmed by the immunoblotting with ezrin-, radixin-, and moesin-specific mAbs (M22 and CR22 recognized distinct epitopes of the COOH-terminal half of moesin) (Fig.  1C). Theoretically, of course, the possibility cannot be completely excluded that only the short fragment of moesin molecule, which cannot be detected with pAbs (TK88 or TK89) or mAbs (M22 or CR22), is expressed in small amounts in the moesin single knockout cells. It is not, however, likely that these small fragments of moesin, if any, can work as functional ERM proteins.
Generation of Moesin Null Mice-Clones 145 and 199 were injected into C57BL/6 recipient blastocysts, and male chimeric mice were crossed with C57BL/6 females. F1 female agouti pups, which were expected to be heterozygous for the mutant moesin allele, were then crossed with wild-type males and littermates were genotyped using the 3Ј probe ( Fig. 2A). This interbreeding yielded offspring at the expected Mendelian segregation ratio, 1:1:1:1 of wild-type male (ϩ/Y), wild-type female (ϩ/ϩ), heterozygous female (ϩ/Ϫ), and hemizygous male (Ϫ/Y). Moesin-deficient male mice (Ϫ/Y) developed and grew normally in the laboratory environment and showed no differences in weight, size, or reproductive ability from wild-type mice at least up to 12 months old. Furthermore, interbreeding between F2 hemizygous males (Ϫ/Y) and F2 heterozygous females (ϩ/Ϫ) yielded F3 homozygous females (Ϫ/Ϫ), which also showed normal development, growth, and reproductive capacity, at least up to 9 months old (data not shown). Various tissues were examined histologically in hematoxylin-eosin-stained sections of moesin-deficient mice (Ϫ/Y and Ϫ/Ϫ), and no significant abnormalities were detected (date not shown). Most of the following analyses were performed using hemizygous male mice (Ϫ/Y).

Expression Levels and Subcellular Distributions of Ezrin and
Radixin in Moesin-deficient Mice-We first examined whether the expression levels of ezrin and radixin were elevated in moesin-deficient mice in a compensatory manner by immunoblotting with anti-ERM pAb TK89 that recognized COOHterminal halves of all ERM family members (Fig. 2B). As reported previously, in wild-type mice most tissues co-expressed ezrin, radixin, and moesin in various expression ratios, but the liver and intestine lacked ezrin and radixin expression, respectively. Unexpectedly, in moesin-deficient mice, neither ezrin nor radixin expression was affected; no compensatory up-regulation of ezrin or radixin was detected in any of the tissues examined. Immunoblotting with ezrin-, radixin-, and moesinspecific mAbs also confirmed this conclusion, and anti-ERM pAb TK88 that recognized NH 2 -terminal halves of all ERM proteins gave the same results without additional bands at the lower molecular mass region (date not shown). We next examined the subcellular distributions of ezrin and radixin in various tissues of moesin-deficient mice. When frozen sections of the kidney from wild-type mice were doubly stained with anti-ezrin pAb (TK90)/anti-moesin mAb (M22) (Fig. 3A) or anti-radixin pAb (I1)/anti-moesin mAb (M22) (data not shown), intense ezrin, radixin, and moesin signals were detected from apical surfaces of proximal tubules and glomeruli. As reported previously, in endothelial cells of blood vessels, moesin was abundantly detected, whereas the expression levels of ezrin and radixin were rather low. In the kidney of moesin-deficient mice, in which the moesin signal was not detected by immunofluorescence microscopy, the subcellular distributions of ezrin (Fig. 3A) and radixin (date not shown) did not appear to be affected. In the liver of wild-type mice, the only ERM proteins expressed were radixin and moesin. In this tissue, moesin expression was mostly restricted to sinusoidal endothelial cells (Fig. 3B, panel b), whereas radixin was detected abundantly in bile canaliculi and in small amounts in sinusoidal endothelial cells (Fig. 3B, panel a). In the liver of moesin-deficient mice, the subcellular distribution of radixin did not appear to be affected (Fig. 3B, panels c and d), and no induction of ezrin was detected (data not shown). Further intensive immunofluorescence microscopic analyses led us to conclude that targeted disruption of the moesin gene did not affect

FIG. 1. Targeted disruption of the mouse moesin gene in ES cells. Panel
A, schematic representation of wild-type allele, targeting vector, and targeted allele of the moesin gene. The targeting vector contained the loxP/phosphoglycerate kinase-neo/loxP cassette and SA/PA in its middle portion to delete a 3Ј part of exon 3 (Ex3). DT-A, diphtheria toxin A gene; bars, 5Ј and 3Ј probes for Southern blotting. The BamHI fragments detected by the 3Ј probe (BamHI digest) and BstXI fragments detected by the 5Ј probe (BstXI digest) from wild-type and targeted alleles are also shown. B, BamHI; Bs, BstXI; P, PstI; K, KpnI. Panel B, Southern blot analysis of genomic DNA from wild-type cells (wild (ϩ/Y)) and two independent clones of targeted ES cells (145 (Ϫ/Y) and 199 (Ϫ/Y)). Using the 3Ј probe, BamHIdigested fragments from wild-type and targeted alleles yielded 9.2-and 5.0-kb bands, respectively (3Ј probe). The correct integration of the vector was confirmed by Southern blotting of BstXI digest with 5Ј probe (5Ј probe). Targeted clones were checked for single integration by hybridization with a neo probe (neo probe). Panel C, immunoblotting analysis of the expression of ERM proteins in wild-type clones (wild (ϩ/Y)) and targeted ES cell clones (145 (-/Y) and 199 (-/Y)). Anti-ERM pAb TK89, TK88 recognized COOH-and NH 2terminal halves of all ERM family members, whereas mAb M11, R21, and M22 were specific for ezrin, radixin, and moesin, respectively. Anti-mouse ERM mAb CR22 reacted strongly with moesin. Targeted ES cells completely lacked moesin expression. Arrows indicate ezrin (E), radixin (R), and moesin (M). either the expression levels or subcellular distribution of ezrin or radixin in any of the tissues examined. Because we detected no abnormalities in moesin-deficient mice at the tissue level, we next examined several types of cells isolated from moesindeficient mice, in which moesin was reported to be important for physiological functions.
Aggregation of Moesin-deficient Platelets-Among the ERM proteins, human platelets were reported to predominantly express moesin, and moesin was suggested to be involved in their aggregation process (54). Therefore, we first examined the aggregation activity of platelets isolated from moesin-deficient mice. Upon immunoblotting of wild-type mouse platelets, in addition to large amounts of moesin, significant amounts of ezrin and radixin were also detected (Fig. 4A). In platelets collected from moesin-deficient mice, the moesin band disappeared completely, leaving relatively small amounts of ezrin and radixin without compensatory up-regulation.
We then compared the aggregation ability of platelets between wild-type and moesin-deficient mice (Fig. 4B). Wildtype and moesin-deficient platelets were collected, and they were subjected to the aggregation assay; aggregation was initiated by addition of various concentrations of thrombin, ADP, or collagen, and the extent of aggregation was measured as the turbidity change with a platelet aggregometer. Unexpectedly, as shown in Fig. 4B, no differences were detected between wild-type and moesin-deficient platelets.
These observations led us to conclude that in platelets targeted disruption of the moesin gene did not affect their aggregation activity without any compensatory up-regulation of ezrin or radixin.
Stress Fiber/Focal Contact Formation and Cell Motility of Moesin-deficient Fibroblasts-Moesin was reported to be required for the Rho-dependent formation of actin stress fibers and focal contacts in fibroblasts (36). We then isolated and cultured embryonic fibroblasts from wild-type as well as moesin-deficient mice and stained them with rhodamine- phalloidin   FIG. 2. Moesin-deficient mice. A, genotypes of F2 mice. Genomic DNA was isolated from tails, and BamHI digests were hybridized with the 3Ј probe. In addition to wild-type male (ϩ/Y) and female (ϩ/ϩ) mice, males (Ϫ/Y), and females (-/ϩ), hemi-and heterozygous, respectively, for null mutation of the moesin gene were identified. B, expression of ERM proteins in various tissues of wild-type mice (ϩ/Y) and moesin knockout (Ϫ/Y) male mice. Tissues were homogenized in SDS sample buffer, and each lysate was applied onto SDS-polyacrylamide gel electrophoresis at the same amount of total protein (15 g/lane), followed by immunoblotting with anti-ERM pAb T89. T89 pAb recognized all ERM family members (see Fig. 1C). Arrows indicate ezrin (E), radixin (R), and moesin (M).

FIG. 3. Subcellular distribution of ERM proteins in kidney and liver of wild-type and moesin knockout male mice.
A, frozen sections of kidney from wild-type mice (a and b) and moesin knockout mice (c and d) were doubly stained with ezrin-specific pAb TK90 (a and c) and moesin-specific mAb M22 (b and d). Arrowheads, apical surfaces of proximal tubules; asterisks, glomeruli; arrows, endothelial cells. B, frozen sections of liver from wild-type mice (a and b) and moesin knockout mice (c and d) were doubly stained with radixin-specific pAb I1 (a and c)/moesin-specific mAb M22 (b and d). Arrows, sinusoidal endothelial cells; arrowheads, bile canaliculi; bar, 10 m. or anti-vinculin mAb. As shown in Fig. 5 (A-D), even in moesindeficient fibroblasts, well developed stress fibers with vinculinpositive focal contacts were observed, indicating that targeted disruption of the moesin gene did not affect stress fiber/focal contact formation in fibroblasts. We next compared the migration rate of moesin-deficient fibroblasts on coverslips with that of wild-type fibroblasts using a time lapse video system. Again, however, no statistically significant difference was detected by Student's t test (p Ͻ 0.005) (Fig. 5E). The mean migration speeds of moesin-deficient and wild-type fibroblasts on coverslips were 31.45 Ϯ 1.44 and 29.05 Ϯ 1.52 m/h, respectively (n ϭ 100).
Microvillar Formation in Moesin-deficient Mast Cells-Finally, we examined microvillar formation in moesin-deficient cells, in which ERM proteins were reported to be directly involved through experiments with antisense oligonucleotides (40). To obtain a homogeneous population of mast cells bearing well developed microvilli, we cultured bone marrow isolated from wild-type as well as moesin-deficient mice in medium containing interleukin-3 for 6 -7 weeks (51). Immunoblotting with anti-ERM pAb TK89 revealed that wild-type mast cells expressed large amounts of moesin, small amounts of radixin, and only trace amounts of ezrin and that targeted disruption of the moesin gene did not elevate the expression levels of ezrin or radixin (Fig. 6A). However, as shown in Fig. 6B, no significant differences were detected in the length or number of microvilli between the wild-type and moesin-deficient mast cells on scanning electron microscopy. DISCUSSION ERM (ezrin/radixin/moesin) proteins have been implicated as general cross-linkers between the plasma membrane and actin filaments (for reviews see Refs. [15][16][17][18][19]. In this study, we generated male and female mice hemi-and homozygous, respectively, for a null mutation in the moesin gene located on the X chromosome. Surprisingly, the mutant mice exhibited no obvious abnormalities in appearance or fertility, and a systemic histological scan of mutant tissues revealed no abnormalities. Our results clearly demonstrated that moesin is not required for normal mouse development or for survival in the laboratory environment. This is surprising in view of the degree of conservation of the moesin gene, for example, the occurrence of a moesin gene in Drosophila (55,56) and the tissue-specific regulated expression of this gene (57)(58)(59).
To date immunofluorescence microscopy and immunoblotting analyses have revealed that the ratio of the levels of ezrin, radixin, and moesin expression in individual cells varies between different tissues in wild-type mice. Furthermore, ezrin, radixin, and moesin are not always colocalized. From these in situ observations in wild-type mice, ezrin, radixin, and moesin have been suggested to have specific functions. However, experiments in vitro or at the cellular level have not clearly identified differences in function between these molecules. When the expression of any one or two ERM proteins was selectively suppressed by antisense oligonucleotides, no phenotypic changes were detected, and cell-cell/cell-matrix adhesion and microvillar formation were affected only when expression FIG. 4. Moesin-deficient platelets. A, immunoblotting of platelets isolated from wild-type mice (ϩ/Y) and moesin knockout (Ϫ/Y) male mice as well as human platelets with anti-ERM pAb TK89. In human platelets, moesin was abundantly detected in addition to small amounts of radixin. In wild-type mouse platelets, moesin was predominant, but ezrin as well as radixin were detected also at significant levels. B, platelet aggregation assay. Aggregation of platelets was initiated by the addition of thrombin (0.0125, 0.025, and 0.1 units/ml), ADP (0.5 M), or collagen (2 and 5 g/ml), and the extent of aggregation was measured as the turbidity change with a platelet aggregometer. of all family members was suppressed (40). Furthermore, the NH 2 -terminal halves of all ERM proteins directly bound to the cytoplasmic domains of CD44, Rho-GDP dissociation inhibitor, and PIP 2 (23,35) and formed both homo-and heterodimers (29 -31).
The present results favor the notion that ERM proteins are functionally redundant also at the whole body level. Interestingly, in moesin-deficient mice as well as isolated moesindeficient cells, targeted disruption of the moesin gene did not induce compensatory up-regulation of the other members of the ERM family. In any of the tissues examined in mutant mice, no changes were detected in the expression levels or subcellular distributions of ezrin or radixin. More importantly, in isolated moesin-deficient cells such as platelets and mast cells, in which moesin was predominantly expressed in wild-type cells, moesin completely disappeared, leaving relatively small amounts of ezrin and radixin. The aggregation activity of moesin-deficient platelets and the microvillar formation of moesin-deficient mast cells were not affected. Considering that ERM proteins were thought to be directly involved in platelet aggregation (54) and microvillar formation (4,31,40,60,61), we concluded that only a small fraction of total ERM proteins in wild-type platelets and mast cells are sufficient for their physiological functions. Similar observations, i.e. no phenotypic changes in knockout mice without up-regulation of other family members, has been reported in various systems. For example, mice devoid of components of intermediate-sized filaments such as vimentin and glial fibrillary acidic protein developed normally without compensatory expression of other intermediate filament components (62,63).
Another issue that we should discuss here is the function of moesin in intracellular signaling. We proposed that the crosslinking activities of ERM proteins were regulated by the Rhodependent signaling pathway through direct binding to Rho-GDP dissociation inhibitor and/or through Rho-dependent posphorylation (23,(35)(36)(37)(38). Ezrin and radixin bound to Rho-GDP dissociation inhibitor with similar binding constants to moesin and were phosphorylated by Rho kinase with similar efficiency to moesin in vitro. Mackay et al. (36) found that moesin was required for the Rho-dependent formation of stress fibers and focal contacts in permeabilized Swiss 3T3 cells. Also in this case, moesin was able to be replaced by ezrin and radixin. From the present results, we concluded that the functions of ERM proteins in the Rho-dependent signaling pathway are redundant also at the whole body level.
In conclusion, the present observations were consistent with the notion that ERM proteins are functionally redundant. The possibility cannot be excluded that a more distantly related protein to moesin in the band 4.1 superfamily functionally compensates for the lack of moesin. Further generation of mutant mice lacking ERM proteins, both singly and in combination, will lead to a better understanding of the physiological relevance of the occurrence of these three closely related proteins.
FIG. 5. Moesin-deficient fibroblasts. Fibroblasts were cultured from wild-type embryos (A and C) and moesin knockout embryos (B and D) and stained with rhodamine-phalloidin (A and B) or anti-vinculin mAb (C and D). Targeted disruption of the moesin gene did not affect the morphology, number of stress fibers, or focal contacts of cultured embryonic fibroblasts. The migration rates of these cells were examined on coverslips using a time lapse video system (E). There were no significant differences in migration rate between wild-type mice (ϩ/Y and ϩ/ϩ) and moesin-deficient mice (Ϫ/Y and Ϫ/Ϫ) (p Ͻ 0.005). Bar, 50 m.