Interaction of Bnr1p with a Novel Src Homology 3 Domain-containing Hof1p

Proteins containing the formin homology (FH) domains FH1 and FH2 are involved in cytokinesis or establishment of cell polarity in a variety of organisms. We have shown that the FH proteins Bni1p and Bnr1p are potential targets of the Rho family small GTP-binding proteins and bind to an actin-binding protein, profilin, at their proline-rich FH1 domains to regulate reorganization of the actin cytoskeleton in the yeast Saccharomyces cerevisiae. We found here that a novel Src homology 3 (SH3) domain-containing protein, encoded by YMR032w, interacted with Bnr1p in a GTP-Rho4p-dependent manner through the FH1 domain of Bnr1p and the SH3 domain of Ymr032wp. Ymr032wp weakly bound to Bni1p. Ymr032wp was homologous to cdc15p, which is involved in cytokinesis inSchizosaccharomyces pombe, and we named this geneHOF1 (homolog of cdc 15). Both Bnr1p and Hof1p were localized at the bud neck, and both the bnr1 andhof1 mutations showed synthetic lethal interactions with the bni1 mutation. The hof1 mutant cells showed phenotypes similar to those of the septin mutants, indicating thatHOF1 is involved in cytokinesis. These results indicate that Bnr1p directly interacts with Hof1p as well as with profilin to regulate cytoskeletal functions in S. cerevisiae.

in the budding processes (for reviews, see Refs. 4 and 5). We have isolated BNI1 as a potential target of RHO1, which links RHO1 with the actin cytoskeleton (6). BNI1 has subsequently been shown to be a potential target of CDC42, RHO3, and RHO4 (7). BNR1 is a BNI1-related gene and is a potential target of RHO4 (8). Bni1p and Bnr1p are members of the FH 1 family of proteins, which are defined by the presence of two formin homology domains, the proline-rich FH1 domain and the FH2 domain. The FH proteins play an important role in the actin cytoskeleton-dependent processes, including cytokinesis and establishment of cell polarity (for reviews, see Refs. 9 and 10). We have recently shown that Bni1p interacts with elongation factor 1␣, which binds to and bundles actin filaments (11), and that Spa2p is required for the localization of Bni1p at the bud tip (12). Bni1p and Bnr1p, at their FH1 domains, bind to an actin monomer-binding protein, profilin, which is implicated in actin polymerization (7,8). A proline-rich sequence also interacts with an SH3 domain, which is found in a wide variety of proteins, ranging from cytoskeletal components to signal transducing enzymes (for a review, see Ref. 13). Actually, the FH1 domain of mouse formin binds to SH3 domain-containing proteins (14), although its physiological significance remains obscure.
We show here that a novel SH3 domain-containing protein, Hof1p, directly binds to the FH1 domain of Bnr1p through the SH3 domain in a GTP-Rho4p-dependent manner. The hof1 mutant shows a deficiency in cytokinesis. Both Hof1p and Bnr1p are localized at the bud neck, and both the hof1 and bnr1 mutations show synthetic lethal interactions with the bni1 mutation. Our results suggest that Bnr1p interacts with both Hof1p and profilin at its FH1 domain to regulate reorganization of the actin cytoskeleton.

EXPERIMENTAL PROCEDURES
Strains, Media, and Standard Methods-Wild type yeast strains OHNY1 (MATa ura3 leu2 trp1 his3 ade2) and OHNY3 (MATa/MAT␣ ura3/ura3 leu2/leu2 trp1/trp1 his3/his3 ade2/ade2) (15) were used for cytological and genetic studies. TAT7 (MATa trp1 leu2 his3 LYS2::lexA-HIS3 ura3::lexA-lacZ) was used for the two-hybrid studies. Yeast strains were usually grown in rich medium (yeast extract peptone dextrose adenine uracil, YPDAU), and yeast transformants were selected in dextrose-containing (synthetic dextrose) or galactose-containing (synthetic galactose) selection media (12). Yeast transformations were performed by the lithium acetate methods (16). Standard yeast genetic manipulations were performed as described (17). An Escherichia coli strain DH5␣ was used for construction and propagation of plasmids and purification of recombinant proteins. * This investigation was supported by grants-in-aid for Scientific Research and for Cancer Research from the Ministry of Education, Science, Sports, and Culture of Japan (1997), by grants-in-aid for Abnormalities in Hormone Receptor Mechanisms and for Aging and Health from the Ministry of Health and Welfare of Japan (1997), and by a grant from the Human Frontier Science Program (1997). 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. ‡ Present address: Tsukuba Research Laboratories, Eisai Co., Ltd., 5-1-3, Tokodai, Tsukuba, Ibaraki 300-2635, Japan. § Present address: Second Department of Surgery, Osaka University Medical School, Suita 565-0871, Osaka, Japan.
Yeast Two-hybrid Method-A plasmid containing a gene fused to DBD LexA was transformed into TAT7, and the resultant transformant was retransformed by a plasmid containing a gene fused to AD GAL4 . Cells of each transformant were cultured in synthetic dextrose medium lacking tryptophan and leucine, and ␤-galactosidase activity was measured quantitatively according to the O-nitrophenyl-␤-D-galactopyranoside assay method (21). The values (Miller units) are the averages of ␤-galactosidase activity for three transformants, and each measured value was within 50% of the average. For qualitative assay for ␤-galactosidase activity, cells of each transformant were placed on the nitrocellulose filter and stained with 5-bromo-4-chloro-3-indolyl ␤-D-galacto-pyranoside (22).
Disruption of HOF1-pUC19-hof1::LEU2 was cut with BamHI and SmaI, and the digested DNA was introduced into OHNY3. The genomic DNA was isolated from each transformant, and the proper disruption of HOF1 was verified by polymerase chain reaction (data not shown). A transformant in which one HOF1 allele was disrupted was subjected to tetrad analysis. All dissected asci (11 asci) showed a 2 Leu Ϫ :2 Leu ϩ segregation pattern, and all of the Leu ϩ clones showed the temperature-sensitive growth phenotype. One Leu ϩ strain, TKUK1, was used as a hof1 mutant in this study.
Other Procedures-SDS-polyacrylamide gel electrophoresis and determination of protein concentrations were performed as described (27,28).  (Fig. 1A). In contrast, Bni1p(1239 -1953), containing the FH1, FH2, and its C-terminal domains, weakly bound to Hof1p, but neither the FH1 domain itself nor full-length Bni1p did (Fig. 1B). Western blot analysis indicated that the levels of expression of Bni1ps were similar to those of Bnr1ps (data not shown). Deletion analysis of Ymr032wp indicated that Ymr032wp bound to Bnr1p at its SH3 domain (Fig. 1C). Ymr032wp is homologous to cdc15p, which is involved in cytokinesis in Schizosaccharomyces pombe (29). Both of the proteins have a region at the N terminus with potential to form coiled-coils, a C-terminal SH3 domain, and a PEST region, which is implicated in turnover of proteins (30). We named Ymr032wp Hof1p (homolog of cdc 15).
Rho4p-dependent Interaction of Bnr1p with Hof1p-We examined by the two-hybrid method whether full-length Bnr1p interacted with Hof1p in a Rho4p-dependent manner. Wild type or mutant Rho4p carrying an amino acid substitution that keeps Rho4p in the GTP-bound form (Rho4p(Q70L)) (31) or the GDP-bound or nucleotide-free form (Rho4p(T25N)) (32) was overexpressed in the two-hybrid reporter strain, TAT7, expressing full-length Bnr1p and full-length Hof1p. These Rho4ps did not possess the C-terminal lipid modification site to prevent association of Rho4ps with the membranes (8). Hof1p interacted with full-length Bnr1p when Rho4p(Q70L), but not wild type Rho4p or Rho4p(T25N), was overexpressed (Fig. 3A). We next examined whether full-length Bni1p interacted with Hof1p in a Rho family-dependent manner. However, Hof1p did not interact with full-length Bni1p even when Rho3p(Q74L) or Rho4p(Q70L) was overexpressed (Fig. 3B). Western blot analysis indicated that wild type and mutant Rho4ps and wild type and mutant Rho3ps were expressed at a similar level (data not shown). We could not examine the effect of Rho1p-(Q68L) or Cdc42p(G12V) on the Bni1p-Hof1p interaction because overex-pression of these proteins severely inhibited cell growth even with the C-terminal mutations. 2 These results indicate that full-length Bnr1p interacts with Hof1p in a GTP-Rho4p-dependent manner but that full-length Bni1p does not interact with Hof1p, irrespective of the presence or the absence of the GTP-Rho family.
Involvement of HOF1 in Cytokinesis-The HOF1 gene was disrupted with LEU2. The hof1 mutant showed the temperature-sensitive growth phenotype (data not shown). The growtharrested cells of the hof1 mutant were observed under a microscope and 70% of the hof1 mutant cells had a large and elongated bud (Fig. 4). Staining of DNA and chitin revealed that these hof1 mutant cells showed multinucleate and the deposition of chitin throughout the cells (Fig. 4A). Staining of actin revealed that the hof1 mutant cells possessed cortical actin patches at the bud tip but not at the mother-bud neck (Fig. 4B). These phenotypes are similar to those of the septin mutants (33). The septin family members, including CDC3, CDC10, CDC11, and CDC12, are involved in cytokinesis (for reviews, see Refs. 34 and 35). Our results indicate that the hof1 mutant is deficient in cytokinesis.
hof1 mutation was also synthetically lethal with the bni1 mutation (data not shown). In contrast, the hof1 mutation was not synthetically lethal with the bnr1 mutation (data not shown). These results suggest that HOF1 and BNR1 function in a similar signaling pathway.
Localization of Hof1p at the Bud Neck-The localization of Hof1p was examined by the immunofluorescence microscopic analysis. The HA-HOF1 gene was expressed under the control of the HOF1 promoter. However, a positive staining was not observed (data not shown). HA-HOF1 was therefore expressed under the control of the GAL1 promoter on a single copy plasmid. The expression of HA-Hof1p suppressed the temperaturesensitive growth phenotype of the hof1 mutant (data not shown), indicating that the addition of the HA tag does not impair the functions of HOF1. HA-Hof1p was localized at the bud neck in both small-and large-budded cells, as was Cdc11p (Fig. 5). Localization of HA-Bnr1p was similarly examined, and HA-Bnr1p was also localized at the bud neck (Fig. 5). DISCUSSION We have tested here the possible interactions of Bnr1p with four SH3 domain-containing proteins, Hof1p, Abp1p, Sla1p, and Rvs167p, and shown that Bnr1p interacts with only Hof1p. Abp1p, Sla1p, and Rvs167p are co-localized with cortical actin patches or actin cables (for a review, see Ref. 4), whereas Bnr1p and Hof1p are co-localized at the bud neck through the cell cycle, suggesting that Bnr1p specifically interacts with Hof1p among many SH3 domain-containing proteins.
We have shown that this interaction of Bnr1p with Hof1p is direct and GTP-Rho4p-dependent. GTP-Cdc42 stimulates an actin-depolymerizing activity of its target, neural Wiskott-Aldrich syndrome protein (36). This finding, together with the present results that although the FH1 domain of Bnr1p interacts with Hof1p even in the absence of GTP-Rho4p, full-length Bnr1p interacts with it only in a GTP-Rho4p-dependent manner, indicates that the binding of Rho family to each target molecule makes its conformation active as described for the Ras family-Raf interaction system (for a review, see Ref. 37).
We have shown that Hof1p also interacts with Bni1p, but this interaction is very weak. The FH1 domain of Bni1p was not sufficient for the interaction with Hof1p, and full-length Bni1p did not interact with Hof1p in a GTP-Rho family-dependent manner. Moreover, Bni1p is mainly localized at the bud tip (12). Therefore, the physiological significance of the Bni1p-Hof1p interaction remains obscure.
The PLP, and PPAPPLP. The SH3 domain of Hof1p interacted with a region of Bnr1p containing the former two (Bnr1p(765-806)) but not that containing the latter two (Bnr1p(818 -850)) in the two-hybrid method, although the interaction was relatively weak (data not shown). The FH1 domain of Bnr1p(765-806) also interacted with profilin in the two-hybrid method (data not shown). These results indicate that the FH1 domain of Bnr1p binds to both Hof1p and profilin and are consistent with an earlier finding that both the SH3 domain and profilin possess a similar ligand-binding surface, which consists of an elongated patch of aromatic residues (for a review, see Ref. 38). In contrast, the FH1 domain of Bni1p contains four proline-rich sequences, PPPPPPPPPPVP, PAPPPPPPPPPPP, PPPPPLP, and PPAPP, but the former two proline-rich sequences do not possess a leucine residue at the Ϫ1 position from the C terminus. The failure of the FH1 domain of Bni1p to interact with Hof1p may be due to this structural difference. In S. pombe, cdc15p (29), cdc3p (profilin) (39), and cdc12p (an FH protein) (40) are involved in cytokinesis. It has not yet been shown that cdc15p interacts with cdc12p, but cdc3p interacts with cdc12p (40). Interactions of an FH protein with both profilin and an SH3 domain-containing protein may be a general feature in the modes of action of FH proteins.
We have lastly shown that the disruption mutant of HOF1 is deficient in cytokinesis. This result is consistent with the facts that cdc15 is also involved in cytokinesis in S. pombe (29) and that Hof1p is localized at the bud neck, as is a septin, Cdc11p. Hof1p shares homologous regions, other than the SH3 domain, with cdc15p and uncharacterized proteins in tapeworm and mouse (29). These proteins may be generally involved in cytokinesis. A protein related to mammalian IQGAP, Iqg1p/Cyk1p, has recently been implicated in cytokinesis in S. cerevisiae (41,42). Iqg1p/Cyk1p is localized at the bud neck and associates with actin filaments. It would be interesting to examine whether Hof1p physically or functionally interacts with Iqg1p/Cyk1p.