Genetic evidence that formins function within the nucleus.

The murine limb deformity (ld) locus encodes a set of proteins, termed formins, that are required for embryonic limb and kidney development. Previous studies had indicated that these proteins are located in the nucleus and cytoplasm and have biochemical properties consistent with an action within the nucleus. To test the notion that nuclear localization is crucial for formin function, we carried out molecular and biochemical studies on three ld alleles. We show that two transgene-induced alleles, ldTgHd and ldTgBri, generate similar COOH-truncated formins that lack the terminal 110 amino acids, while a third allele, ldIn2, generates a less extensively truncated formin that lacks the terminal 42 amino acids. Using subcellular fractionation analysis, we find that wild-type formin is detected in both nuclear and cytosolic fractions; in contrast, the truncated formins encoded by ldTgHd and ldTgBri are strictly cytosolic. The less extensively truncated ldIn2 formin shows a similar, but less complete, localization defect. Consistent with this weaker cellular phenotype, hind limbs from ldIn2 mice have milder skeletal defects than those of ldTgBri mice. These observations define a small region in the carboxyl terminus that is required for nuclear localization and suggest that nuclear localization plays a role in formin action.

The murine limb deformity (ld) locus encodes a set of proteins, termed formins, that are required for embryonic limb and kidney development. Previous studies had indicated that these proteins are located in the nucleus and cytoplasm and have biochemical properties consistent with an action within the nucleus. To test the notion that nuclear localization is crucial for formin function, we carried out molecular and biochemical studies on three ld alleles. We show that two transgene-induced alleles, ld TgHd and ld TgBri , generate similar COOH-truncated formins that lack the terminal 110 amino acids, while a third allele, ld In2 , generates a less extensively truncated formin that lacks the terminal 42 amino acids. Using subcellular fractionation analysis, we find that wild-type formin is detected in both nuclear and cytosolic fractions; in contrast, the truncated formins encoded by ld TgHd and ld TgBri are strictly cytosolic. The less extensively truncated ld In2 formin shows a similar, but less complete, localization defect. Consistent with this weaker cellular phenotype, hind limbs from ld In2 mice have milder skeletal defects than those of ld TgBri mice. These observations define a small region in the carboxyl terminus that is required for nuclear localization and suggest that nuclear localization plays a role in formin action.
The murine limb deformity (ld) gene is required for normal limb and kidney development (1,2). Mice homozygous for ld mutations exhibit fusion of the long bones of the limbs and fusion and reduction of the metacarpals/metatarsals and phalanges. This limb defect is first manifested in the embryonic limb bud as a failure of proper apical ectodermal ridge development and a reduction in the width of the anteroposterior limb axis (3). These two defects are likely related to the loss of fgf-4 expression in the apical ectodermal ridge and the reduction of shh and HoxD expression in limb bud mesoderm (4,5). In addition, mutant mice can suffer unilateral or bilateral renal agenesis or hypoplasia, although this phenotype is of variable penetrance and expressivity. All five alleles of ld (ld TgHd , ld J , ld OR , ld TgBri , ld In2 ) display similar limb defects (1, 6 -8), while the kidney defects differ in penetrance and severity among the alleles (9) .
The recovery of a transgene insertion allele, ld TgHd , allowed the cloning of the ld locus by transgene tagging (1,2). The ld locus produces a complex set of transcripts encoding novel proteins termed formins (2). Three ld alleles, ld TgHd , ld TgBri , and ld In2 , contain structural alterations in the 3Ј end of formin transcripts (7,10). The molecular bases of two other ld alleles, ld J and ld OR , are unknown.
At least four major transcripts arise from the ld locus (Fig.  1A), and their embryonic expression patterns are consistent with the ld mutant phenotype. Isoforms I, II, and III share a common amino-terminal exon and are expressed coordinately (11). These three isoforms are expressed in the pronephros, mesonephros, and metanephros of midgestation mouse embryos but are absent from the embryonic limb buds (4). Like isoforms I, II, and III, isoform IV is present throughout renal development; in addition, it is expressed in early limb buds, with highest levels in the apical ectodermal ridge (4,11). These differential expression patterns suggest that all four isoforms may play important roles during renal development, while isoform IV may play a role in the ld limb defects. RNase protection assays indicate that other, less well characterized ld transcripts are also present in the ectoderm of limb buds (11).
The biochemical function of formins is unknown, although recent studies have provided a number of clues. First, a central portion of the proteins encoded by isoforms I, II, and IV contains a large number of proline residues, and some sequences in this region match consensus sequences for Src homology 3 ligands. Indeed, fusion proteins containing this region of formin have been demonstrated to bind in vitro to Src homology 3 domains (12,13). This proline-rich region also binds in vitro to a class of novel proteins containing WW domains (12). These findings suggest that this proline-rich region probably mediates physical interactions of formin with associated proteins. Second, formins share some sequence similarity with the Drosophila genes diaphanous (14) and cappuccino (15), which are required for cytokinesis and egg polarity, respectively. Third, immunohistochemical studies indicate that chicken formin is localized in a punctate nuclear pattern (16). Finally, formins harvested from transfected COS cells are phosphorylated mostly on serine and bind to DNA-cellulose columns (17), suggesting that formins may interact directly or indirectly with nucleic acids.
To learn more about the function of formins, we have compared the structural and biochemical properties of wild-type formin to mutant formins encoded by several ld alleles. We find that three ld alleles encode COOH-truncated but stable proteins. Formins from ld TgHd and ld TgBri cells have identical truncations and are strictly cytosolic, whereas a large proportion of formin from wild-type cells are stably associated with nuclei. Formin from ld In2 cells shows a less severe localization defect. In addition, morphological comparisons show that ld In2 mice have a less severe limb phenotype than ld TgBri mice.

EXPERIMENTAL PROCEDURES
3Ј-Rapid Amplification of cDNA Ends-3Ј-Rapid amplification of cDNA ends was performed as described previously (18) to isolate mutant cDNAs. One g of polyadenylated RNA (Pharmacia QuickPrep), * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U60966, U60967, U60968, and U60969. ‡ Supported in part by a National Institutes of Health Medical Scientist Training Program fellowship.
§ To whom correspondence should be addressed: Dept. of Genetics, isolated from wild-type or mutant fibroblast cell lines, was used for first strand cDNA synthesis reactions (Superscript, Life Technologies, Inc.) primed with 400 ng of the following oligonucleotide: 5Ј gtcgacatcgatctcgagtttttttttttttttttt 3Ј. First round polymerase chain reaction was done using an adapter-primer (5Ј aagtcgacatcgatctcgag 3Ј) and an ld-specific primer (5Ј gggatgaatctggtggactatg 3Ј; the first nucleotide corresponds to position 3247 of isoform IV). First round products were gel isolated and used as a template for a second round of polymerase chain reaction using the adapter-primer and a nested ld-specific primer (5Ј ggtaggatcccaggaagctggaacagacaag 3Ј; position 3298 of isoform IV). Products were gel isolated, restricted with BamHI and ClaI, and cloned into pBluescript (Stratagene). Subclones were sequenced on both DNA strands by the Sanger method using Sequenase (U. S. Biochemicals). Antibodies-Antigens for antibodies A and B were generated by constructing formin/glutathione S-transferase fusion proteins in the pGEX vector (19). Antigen A contains amino acids 519 to 641 of isoform IV, and antigen B contains amino acids 1-457. Fusion proteins were purified by affinity chromatography as described (19). Protein preparations were further purified by preparative SDS-polyacrylamide gel electrophoresis, and gel slices were injected into rabbits (Pocono Rabbit Farms). Rabbit bleeds were initially screened by an enzyme-linked immunoabsorbent assay, and bleeds with high titers were selected for further characterization. Antiserum A was purified on an affinity column (20) containing maltose-binding protein (New England BioLabs) fused to amino acids 519 -641 of isoform IV. Affinity-purified antibody A was used at a 1:1000 dilution for Western blots. Antiserum B was used as a crude serum at a 1:2000 dilution. Control Western blots on subcellular fractions were done using a 1:4 dilution of hybridoma supernatant containing anti-c-myc monoclonal antibody 3C7 (21), a generous gift from Lewis Chodosh.
Immunoprecipitations-35 S-labeled in vitro translation products for isoforms I and IV were generated with rabbit reticulocyte lysate (Promega) as described (17). Ten l of the lysate were incubated with 400 l of radioimmune precipitation buffer buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 8.0), and 2 l of either preimmune serum, affinity-purified antibody A, or antiserum B. After 1 h at 4°C, 20 l of protein A-Sepharose (Pharmacia Biotech) were added, and the mixture was rotated for another 30 min at 4°C. The immunoprecipitated material was washed four times with radioimmune precipitation buffer buffer, resolved by SDS-polyacrylamide gel electrophoresis, and visualized by autoradiography.
Subcellular Fractionation-For each fractionation experiment, five nearly confluent 15-cm plates of fibroblast monolayers (17) were washed twice with phosphate-buffered saline. Subsequently, all steps were performed on ice or at 4°C. Cells were scraped off with a rubber scraper and spun down at 1500 ϫ g for 5 min. The cell pellet was resuspended in a hypotonic buffer (20 mM Hepes pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 1 g/ml aprotonin, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin) and allowed to swell on ice for 10 min. Cells were then lysed by vigorous Dounce homogenization (100 strokes) using a tightfitting Dounce homogenizer. At this point and throughout the procedure, cell lysis and recovery of nuclei were monitored by phase microscopy. The homogenate was spun at 1500 ϫ g for 5 min to pellet nuclei and unlysed cells. This pellet was washed twice by resuspending it in hypotonic buffer containing 0.5% Nonidet P-40, Dounce homogenization an additional 10 times, and then spinning at 1500 ϫ g for 5 min. The resulting pellet (nuclear fraction) was resuspended in SDS-polyacrylamide gel electrophoresis sample buffer and sonicated prior to analysis by Western blotting. Meanwhile, the supernatant from the first spin was centrifuged at 100,000 ϫ g for 30 min. The supernatant (S100) is the cytosolic, or cytoplasmic fraction, while the pellet (P100) is a membrane fraction.
For the nuclear elution experiments, washed nuclei were resuspended in extraction buffer (20 mM Hepes pH 7.4, 0.5% Nonidet P-40, 1.5 mM MgCl 2 , and varying concentrations of NaCl) and rotated for 30 min at 4°C. The mixture was then spun at 1500 ϫ g for 5 min, and both the pellet and supernatant were analyzed by Western blotting. Western blots were performed using standard procedures (20) and detected using enhanced chemoluminescence (Amersham).
Skeletal Stains-Skeletal analysis was done on embryonic and adult limbs. To obtain embryos for staining, females were examined daily for vaginal plugs. Noon of the day of the vaginal plug was designated as embryonic day 0.5. Day 15.5 embryos were dissected and fixed overnight in 4% paraformaldehyde at 4°C. After a washing with phosphatebuffered saline (twice, 30 min each), embryos were treated with 5% trichloroacetic acid for 3 h. They were then washed with phosphatebuffered saline (twice, 5 min each), followed by 70% ethanol (5 min). Embryos were then stained overnight in 0.1% Alcian green in 70% ethanol. After washing (5 parts 70% ethanol: 1 part 5% trichloroacetic acid) for several hours and dehydrating in ethanol (2 times, 30 min), samples were cleared in methylsalicylate. Adult limbs were stained with Alcian blue and Alizarin Red S as described previously (22). After clearing, limbs were dissected to facilitate photography.

Three Limb Deformity Alleles Encode COOH-truncated
Formins-Previous studies have shown that three ld alleles, ld TgHd , ld TgBri , and ld In2 , are associated with structural alterations within the formin gene locus (6,7,10). The sites of these alterations within formin genomic DNA are schematically shown in Fig. 1A. In the two transgene-induced alleles, ld TgHd and ld TgBri , the disruptions occur in the intronic sequences 5Ј of exon A, located close to the 3Ј end of the formin gene (2, 7). In the ld In2 allele, caused by a chromosomal inversion event, the rearrangement breakpoint occurs in intronic sequences 3Ј of The genomic organization has been simplified for the sake of clarity. Arrows, the sites of structural alteration in the ld TgHd , ld TgBri , and ld In2 alleles. Exons A and B located near the 3Ј end were the first exons to be identified from the ld locus (2). The four major ld mRNA isoforms are diagrammed below, with the sites of the three mutations indicated by arrowheads. Note that the coding sequence of isoform III terminates early and is unaffected by the mutations. B, alignment of predicted polypeptide sequences from wild-type, ld TgBri , ld TgHd , and ld In2 formins, based on analysis of cDNAs obtained from 3Ј-rapid amplification of cDNA ends. The sequence shown begins just before the beginning of residues encoded by exon A. Residues that match the wild-type sequence are boxed. The first arrow is positioned immediately 5Ј of the first exon A-encoded residue. At this point, the sequences of ld TgBri and ld TgHd formins diverge from the wild-type sequence and terminate (*) shortly thereafter. The second arrow is positioned immediately 3Ј of the last exon B-encoded residue. At this point, the ld In2 sequence diverges from that of wild-type and then terminates. exon B (6). Based on these observations, processed transcripts produced from ld TgHd and ld TgBri are expected to deviate from wild-type transcripts immediately 5Ј of exon A, while transcripts from ld In2 should deviate immediately 3Ј of exon B (Fig.  1A, arrowheads). These predictions are consistent with RNase protection experiments using riboprobes spanning these putative divergence points (7,10).
To obtain precise information about the mutant transcripts and their products, we used 3Ј-rapid amplification of cDNA ends to clone cDNA sequences from the 3Ј end of the mutant transcripts (see "Experimental Procedures"). As expected, ld TgHd and ld TgBri cDNAs lack exons A and B and sequences further downstream. When conceptually translated, these cDNAs encode polypeptides that are identical with the wild type before the start of exon A, at which point they diverge and contain a few new amino acids before prematurely terminating (Fig. 1B). Thus, ld TgHd and ld TgBri formins lack the 110 COOHterminal amino acids of the wild-type product. In contrast, the ld In2 cDNA contains exons A and B but diverges thereafter. When translated, the ld In2 cDNA encodes a polypeptide that terminates prematurely after the exon B-encoded sequences, resulting in the loss of 42 COOH-terminal amino acids (Fig.  1B). In these three alleles, the COOH truncations affect the proteins encoded by isoforms I, II, and IV. Isoform III would not be affected since an alternative splice results in translational termination before these breakpoints (2).
Detection of Truncated Formins in Mutant Cells-Previously described anti-formin antibodies (17) were raised against synthetic peptides derived from the extreme COOH terminus of formin and therefore could not be used to detect the putative truncated mutant isoforms described above. To detect these mutant formins, we raised polyclonal antisera against two new formin fusion proteins. The first antiserum (A) was raised against the region (amino acids 519 -641 of formin isoform IV) immediately upstream of the proline-rich segment (Fig. 2A).
These epitopes are common to isoforms I, II, and IV. Isoform III does not encode these epitopes due to an alternative splice which results in early termination. A second polyclonal serum (B) was raised against amino acids 1-457 of formin isoform IV. This second antiserum recognizes epitopes unique to isoform IV and was used in this study to confirm results obtained using antiserum A. To test the isoform specificity of our antisera, we used these antisera to immunoprecipitate in vitro translated isoforms I and IV. As expected, antiserum A precipitated both isoforms I and IV, whereas antiserum B precipitated only isoform IV (Fig. 2B). No immunoreactivity was obtained from preimmune serum.
We used affinity-purified antibody A to probe Western blots containing total cell lysates from wild-type and ld fibroblasts (Fig. 2C). Consistent with previous immunoprecipitation experiments using anti-peptide antibodies (17), wild-type fibroblasts contained a predominant immunoreactive band migrating at ϳ165 kDa. This band has the same molecular weight as in vitro translated isoform IV, which is the major isoform in fibroblasts (17). This 165-kDa band is also detected using antiserum B (Fig. 3B), further supporting its identification as the formin gene product. Most definitively, mutant mice containing a targeted disruption of formin isoform IV lack this 165-kDa protein. 1 In Western blots probed with antibody A, there is also a prominent band at ϳ110 kDa; at present we do not know the identity of this polypeptide. This 110-kDa band is not recognized with antiserum B and is not altered in the ld mutants, suggesting that it is likely to be a nonspecific band. In the experiment shown in Fig. 2C, the 110-kDa band is apparently decreased in the ld In2 sample, but this result is not seen in other experiments and is unlikely to be significant. In both ld TgHd and ld TgBri cell lysates, the 165-kDa band is absent; instead there is a smaller band migrating at ϳ152 kDa (Fig.  2C). In ld In2 lysates, the 165-kDa band is replaced by one at ϳ160 kDa. Mutant tissue lysates from brain, kidney, and muscle also showed truncated bands when compared with wild-type tissues (data not shown). Taken together, these results identify formin isoform IV as a 165-kDa protein and demonstrate that the alleles ld TgHd , ld TgBri , and ld In2 produce stable, truncated formins consistent with their mutant cDNAs.
A Large Portion of Wild-type Formin Is Stably Associated with Nuclei in Fibroblasts-Since the truncated ld formins are stable and show normal tissue distribution, 2 we sought to uncover biochemical differences between wild-type and mutant formins. We used biochemical subcellular fractionation techniques (see "Materials and Methods") to localize wild-type formin in primary (Fig. 3) and established cell lines (data not shown). Briefly, fibroblast monolayers were swelled in a hypotonic solution and lysed by Dounce homogenization. Nuclei and unlysed cells were pelleted by a 1000 ϫ g spin. Purified nuclei were obtained from this crude nuclear pellet by two washings with a Nonidet P-40 containing buffer. The original supernatant (nonnuclear fraction) was spun at 100,000 ϫ g to obtain the soluble S100 fraction (cytosol) and the pelleted P100 fraction (large complexes and cellular membranes). When Western blots containing these fractions were probed with antiserum A or B (Fig. 3, A and B), we found that a substantial portion of the total formin protein is retained in the nuclear fraction under these separation techniques. At 100 mM NaCl most of the nuclear formin remains stably associated with nuclei, but at 250 mM NaCl the majority is eluted from the nuclei (Fig. 3C).
Our cell fractionation experiments also show that much of the nonnuclear formin sediments into the P100 fraction. This 1 A. Wynshaw-Boris, D. Chan, and P. Leder, manuscript in preparation.

FIG. 2. Detection of wild-type and truncated formins.
A, diagram of isoform IV, with lines indicating the residues present in the two immunogens. The epitopes in antigen A are common to isoforms I, II, and IV, whereas epitopes in antigen B are present only in isoform IV. B, immunoprecipitation of in vitro translated formins. Isoforms I and IV were in vitro translated with rabbit reticulocyte lysate and immunoprecipitated with preimmune serum, affinity-purified antibody A, or antiserum B. Molecular weight markers (in kilodaltons) are indicated. C, Western blot of total cellular lysates from wild-type and mutant cells. The Western blot was incubated with affinity-purified antibody A, followed by horseradish peroxidase-labeled goat anti-rabbit antibody, and detected with enhanced chemoluminescence. Molecular weight markers (in kilodaltons) are indicated. As described in the text, the upper bands are formin polypeptides. The 110-kDa band is likely a nonspecific background band.
result raises the possibility that the nonnuclear formin may be associated with a large macromolecular complex. We find that the amount of formin present in the S100 fraction can vary substantially in different experiments, suggesting that it is sensitive to experimental conditions. ld TgHd and ld TgBri Formins Are Cytosolic-We performed subcellular fractionation experiments on primary fibroblasts cultured from homozygous mutant animals. With both ld TgBri and ld TgHd fibroblasts, we obtained a fractionation profile strikingly distinct from that of wild-type fibroblasts (Fig. 4). There are two major differences in the localization of ld TgBri and ld TgHd formins compared to wild-type formins (Fig. 4, A-C). First, the mutant formins are absent from the nuclear fraction and instead are found almost entirely in the S100 fraction. Second, very little of the formin in the nonnuclear fraction sediments during the 100,000 ϫ g centrifugation (P100), suggesting that these truncated formins are not associated with a large macromolecular complex. These differences were reproducible on multiple experiments. ld In2 cells showed a similar but less dramatic localization defect. ld In2 formin is found predominantly in the S100 and P100 fractions, but a smaller portion is present in the nuclear fraction (Fig. 4D).
We checked our fractionation protocol by light microscopy at various stages of the procedure and found clean separations in each of the cell lines (data not shown). Furthermore, control experiments using the nuclear marker c-Myc showed that the nuclear localization defect that we observed in mutant cells is specific for formin. C-Myc is detected only in the nuclear fraction of wild-type and mutant cells (Fig. 4E).
To rule out the possibility that these strikingly different fractionation profiles were due to stochastic differences among primary cell lines or to experimental variability, primary cell lines from heterozygous ld TgBri /ϩ and ld In2 /ϩ mice were derived (we were unable to perform similar experiments with ld TgHd mice because this mutant is no longer available). Immunoblot analysis demonstrates that the heterozygous ld TgBri /ϩ cell line contains both the wild-type 165-kDa formin and the mutant 152-kDa formin (Fig. 5A). Thus, both the wild-type and mutant isoforms can be analyzed within the same cell line and within a single fractionation experiment. With the ld TgBri /ϩ cell line, a substantial portion of the wild-type formin is present in the nuclear fraction, while the truncated, mutant formin is found solely in the S100 fraction (Fig.  5A). Similarly, the truncated ld In2 formin in the ld In2 /ϩ cell line is found in proportionately larger quantities in the S100 fraction than wild-type formin, although the difference is not as great as with ld TgBri formin (Fig. 5B). These results agree well with the results obtained from homozygous cells.
ld In2 Mice Have Less Severe Hind Limb Defects Than ld TgBri Mice-As described above, ld In2 formin contains a less severe COOH truncation than ld TgBri and ld TgHd formins. In addition, FIG. 3. Subcellular fractionation of wild-type cells. A and B, wild-type fibroblasts were lysed by Dounce homogenization and separated into three fractions: nuclei; S100 supernatant; and P100 pellet. Western blots of these samples were incubated with affinity-purified antibody A (A) or antiserum B (B). Molecular weight markers (in kilodaltons) are indicated. C, extraction of formins from nuclei with increasing salt. The nuclear fraction was incubated for 30 min in the indicated concentration of NaCl, followed by a 1000 ϫ g spin. The supernatant (extract) and pellet were analyzed by a Western blot using affinitypurified antibody A.  5. Subcellular fractionation of heterozygous ld TgBri /؉ and ld In2 /؉ fibroblasts. A, heterozygous ld TgBri /ϩ fibroblasts were fractionated and analyzed by a Western blot using affinity-purified antibody A. The first two lanes contain total cell lysates from wild-type and homozygous ld TgBri fibroblasts and demonstrate the size difference between the wild-type and mutant formin. B, heterozygous ld In2 /ϩ fibroblasts were fractionated and analyzed by a Western blot using affinitypurified antibody A. The first lane has been intentionally underexposed to show clearly the two distinct bands (arrows) present in these cells.
ld In2 formin has a less dramatic localization defect than ld TgBri and ld TgHd formins. This correlation between degree of truncation and severity of localization defect raises the possibility that ld In2 may be a weaker allele than both ld TgBri and ld TgHd . Consistent with this view, we noticed that ld In2 mice could usually be distinguished from ld TgBri mice by inspection of their hind limbs. The distal portion of ld In2 hind limbs is typically thicker than that of ld TgBri hind limbs, which appear to contain fewer digits. To define this observation further, we performed skeletal stains on adult and embryonic limbs from these two mutants (Table I; Fig. 6). ld TgBri hind limbs typically (95%) contained only a single digit, resulting in a distal hind limb that is very narrow (Fig. 6C). In contrast, only 16% of ld In2 hind limbs displayed such a severe defect, with most hind limbs (84%) showing two or more distinct digits (Fig. 6B). No significant differences were observed in the forelimbs. DISCUSSION We have shown that three ld alleles result in COOH-truncated formins. Two such alleles, ld TgHd and ld TgBri , encode similarly truncated formins that show abnormal subcellular localization. Whereas a substantial proportion of wild-type formin is stably associated with nuclei, both these mutant formins are localized almost exclusively to the cytosol. Furthermore, the mutant formins are not sedimented by 100,000 ϫ g centrifugation, suggesting that they are not part of a large macromolecular complex. We conclude that the COOH residues deleted in ld TgHd and ld TgBri are essential for nuclear localization and possibly for interactions with other molecules. A third allele, ld In2 , leads to a smaller truncation, and the corresponding formin shows a less complete localization defect.
At present the precise role of the highly basic COOH region (pI ϭ 9.6 for the COOH-terminal 110 residues) in nuclear localization is unclear. Inspection of this sequence reveals a candidate nuclear localization sequence (KKAKKEHK) at the ld TgHd and ld TgBri breakpoint junction (Fig. 1B) similar to the motif defined for the SV40 large T antigen (23). This sequence is unaltered in ld In2 formin. However, we have been unable to demonstrate a nuclear localization function for this sequence. Fusion of the COOH-terminal 182 residues to the marker protein pyruvate kinase (normally a cytosolic protein) fails to direct the marker to the nucleus. 2 Further studies are required to elucidate the mechanism through which deletion of the COOH region of formin disrupts nuclear localization. An intriguing possibility is that the COOH-terminal residues are required for interactions with other proteins that are critical for transporting or tethering formin to the nucleus. The hypothesis that the COOH-terminal residues are essential for protein-protein interactions is supported by the observation that the ld Bri and ld TgHd formins apparently do not sediment when centrifuged at 100,000 ϫ g. In future experiments, it will be interesting to use the COOH-terminal residues as a probe to search for proteins that interact with formins. However, it is possible that it is not the COOH region per se that is directly involved in nuclear localization, but rather that other critical portions of the molecule suffer conformational changes due to the truncations.
Work on cappuccino, a Drosophila gene with sequence homology to formins, also suggests an important role for the carboxy region of formin. cappuccino belongs to a family of related genes containing two regions of homology (termed FH1 and FH2) to formins (15). One of these homology regions, FH2, is located near the COOH terminus of formin and ends slightly amino terminal to the breakpoint junction found in ld TgHd and ld TgBri . The homology between cappuccino and formins, however, is particularly high and extends beyond FH2 and into the extreme COOH terminus. Significantly, two cappuccino mutants, cap 3871 and cap 38 , contain mutations in the carboxy region (15), close to the corresponding region on formin where the breakpoint junctions of ld TgHd , ld TgBri , and ld In2 reside. These results reinforce our general conclusion that the COOH region of formin plays a critical function, and in future work it will be interesting to determine whether defective subcellular localization occurs in these two cappuccino mutants.
The observation that ld In2 shows both a less severe truncation and a less complete localization defect when compared with either ld TgHd or ld TgBri raises the issue of allelic strength. Since the site of formin action is thought to reside in the nucleus (16,17), our results suggest that ld TgHd and ld TgBri are severe loss-of-function alleles, possibly null alleles, and that ld In2 is a weaker allele. Consistent with this view, the hind limbs of ld In2 mice show less severe digit reductions than those of ld TgBri mice. Since these ld alleles are on mixed genetic backgrounds, it remains possible that these phenotypic differences reflect strain differences. Similar comparisons, unfortunately, could not be performed on ld TgHd hind limbs, since this line is no longer available.
The renal phenotype of ld In2 mice also appears to be signif-  6. Comparison of limb phenotypes between ld In2 and ld TgBri mice. Day 15.5 embryos were dissected and stained with Alcian green to reveal cartilage. The distal portion of a representative hind limb is shown for a wild-type embryo (A), an ld In2 homozygous embryo (B), and an ld TgBri homozygous embryo (C). The arrowheads in B and C indicate well-defined digits. The hind limbs in B and C are representative examples of the most common phenotypes (see Table I) seen in ld In2 and ld TgBri hind limbs, respectively. icantly less severe than that of ld TgHd or ld TgBri mice. First, the incidence of renal aplasia ranges from 57% with ld TgHd and 33% with ld TgBri to 2.5% with ld In2 mice (6,7,9,24). Second, unlike the other ld alleles, in which renal aplasia is the predominant phenotype, the most common renal phenotypes in ld In2 mutant mice are hydronephrosis and hydroureter. Renal agenesis in ld In2 mice is much rarer (6). The renal aplasia in ld animals is due to failure of proper ureteric bud outgrowth into the metanephric cap (9). Hydronephrosis and hydroureter, seen in ld In2 mice, may be less severe manifestations of the same underlying developmental defect. These observations are consistent with the model that ld TgBri and ld TgHd are more severe loss-of-function alleles than ld In2 . As with the discrepancies in limb phenotype, it is possible that these differences in penetrance and expressivity reflect the influence of modifying loci as well as the intrinsic strengths of the alleles. Mating experiments, however, have suggested that the differences in the renal phenotype are more likely a result of intrinsic allelic strength (9) .
In summary, our biochemical analyses of wild-type and mutant formins define a region required for the nuclear localization of formin and emphasize the nucleus as the site of action of formins. Our results provide genetic evidence that formins function within the nucleus. Consistent with these results, previous studies showed that chicken formin is present in a punctate nuclear pattern (16) and that murine formins are nuclear phosphoproteins that bind to DNA-cellulose (17). While the precise biochemical function of formins remains to be determined, the recent identification of two classes of proteins that potentially interact with formins should provide new avenues toward discovering their function within the nucleus (12).