The function of two closely related Rho proteins is determined by an atypical switch I region Free

Small GTP-binding proteins are a ubiquitous protein class in eukaryotic cells. They are involved in diverse processes such as cancer development, cell morphology or vesicle trafficking (Takai et al., 2001). The structure and function of this protein class are highly conserved and well understood. Activation of small GTP-binding proteins is characterized by a GDP-GTP exchange involving regulators called guanine nucleotide exchange factors (GEFs). The active, GTP-bound form of the GTP-binding protein binds and activates the effector proteins. Inactivation is mediated by the binding of a GTPase-activating protein (GAP) that increases the intrinsically low GTPase activity of the GTP-binding protein. The large variety of Rho proteins has evolved by mutations that occurred after gene duplication (Boureux et al., 2007). Although the GTPase-switch mechanism remained highly conserved, the interaction interface varies between the different Rho proteins and thereby creates novel signaling junctions (Wherlock and Mellor, 2002). In this study, we have used a comparative analysis of duplicated and non-duplicated Rho proteins to understand how Rho protein function may have diverged after gene duplication.

Yeasts provide an excellent tool for studying gene duplication (Wolfe, 2006). Comparison of the well-annotated genome of Saccharomyces cerevisiae with several complete genomes of other yeasts allows the reliable identification of duplicated genes. The filamentous growing Ashbya gossypii is of special interest for studies on evolution of gene function in combination with the unicellular budding yeast S. cerevisiae (Philippsen et al., 2005). The different growth styles of the two organisms increase the possibility of observing evolutionary changes in function of gene-products that are involved in elementary processes such as polar growth, cell division and cell cycle control.

A typical member of the Rho GTP-binding protein family is the Rho1 protein from S. cerevisiae. Its function and its regulation are complex but well understood (Levin, 2005). Deletion of the RHO1 gene is lethal (Madaule et al., 1987), which reflects the multitude of tasks that have been attributed to the encoded protein. Two major functions of RHO1, regulation of cell wall biosynthesis and regulation of the actin cytoskeleton, are reflected by two different complementation groups of rho1 mutants (Saka et al., 2001). Cell wall biosynthesis is regulated by interaction of Rho1 with the glucan-synthase complex (Drgonova et al., 1996; Mazur and Baginsky, 1996; Qadota et al., 1996), which produces β-(1,3)-glucan, the major component of the yeast cell wall. Initially, Rho1 was isolated as the regulatory subunit of this complex. Regulation of the actin cytoskeleton is achieved by interaction with two different effector-proteins: the formin Bni1 (Kohno et al., 1996) and the single S. cerevisiae protein kinase C homolog Pkc1 (Nonaka et al., 1995). Although the connection to actin regulation via formins is direct as proteins of this class catalyze polymerization of actin cables, actin regulation via Pkc1 is not well understood and the targets of Pkc1 in actin regulation are unknown. In addition to these two main functions of Rho1, there is evidence for a variety of other functions (reviewed by Levin, 2005).

A first homolog of the S. cerevisiae RHO1 was cloned from A. gossypii before the second copy was discovered during the genome-sequencing project (Wendland and Philippsen, 2001). Deletion mutants of AgRHO1 (open reading frame ABR183W, AgRHO1b) grew slowly, suffered from significant cell lysis and actin defects, and died within 4 days. This suggested functions similar to the S. cerevisiae Rho1. A second RHO1-homolog (ABR182W, AgRHO1a), directly upstream of AgRHO1, was identified after the genome sequence became available (Dietrich et al., 2004). The genomic organization of the AgRHO1 genes together with the neighboring genes in A. gossypii and S. cerevisiae is shown in Fig. 1A. The two copies very probably originate from a tandem-duplication event. Little is known about the function of ABR182W. Deletion mutants have been reported to display a weak lysis phenotype and a weak sensitivity to the chitin-binding dye Calcofluor White (Walther and Wendland, 2005).

Here, we studied the differences in function of a pair of duplicated RHO-type GTP-binding proteins in A. gossypii and compared them with their single homolog in budding yeast. Surprisingly, we found that a single residue in AgRho1a is altered in the highly conserved switch I region. This residue contributes to divergence of protein function and alters GAP specificity without impairing effector interaction.

We first characterized Rho1 protein function in A. gossypii to test whether protein function diverged. An alignment of the protein sequences shows that AgRho1a is 73.9% identical to AgRho1b and 75.6% identical to ScRho1, whereas the latter two proteins share the highest number of identical residues (83.3%) (Fig. 1B). As a first step towards defining the functional difference between AgRHO1a and AgRHO1b, we compared the phenotypes of deletion mutants. Although Agrho1aΔ formed a mature mycelium that grew slightly slower than the wild-type control, deletion of AgRHO1b resulted in cells that never reached the state of a mature mycelium and died because of cell lysis (Fig. 1C). Agrho1aΔ cells also lysed but only few hyphae were affected. Fig. 1D shows an example of lysing hyphae taken from Movie 1 (supplementary material). We also monitored the strain deleted for AgRHO1a at different growth temperatures. Lysis of adult mycelia was quantified by an assay that visualizes the release of cytoplasmic alkaline phosphatase upon lysis by a color reaction (Fig. 1E, for details, refer to Materials and Methods). An Agslt2Δ strain was used as a lysis control. AgSLT2 is the homolog of the S. cerevisiae MAP kinase responsible for maintenance of cellular integrity (Torres et al., 1991). In contrast to Agrho1b deletion mutants, Agslt2 deletions can form a mature mycelium despite extensive lysis. Hyphae at the colony border grow normally and cannot be distinguished microscopically from wild type but the inner, older parts of the mycelium show extensive lysis. At 25°C and 30°C, Agrho1a deletion mutants released about half as much alkaline phosphatase as the Agslt2 deletion mutants. Surprisingly, at 37°C, lysis of Agrho1aΔ decreased to about 4% of that of Agslt2Δ cells. As Rho-type GTPases are often regulators of the actin cytoskeleton (Boureux et al., 2007), we speculated that lysis of the Agrho1aΔ might be due to a defect in actin regulation. However, we did not observe significant differences between the actin cytoskeleton of wild-type and Agrho1aΔ cells stained with Alexa-phalloidin (Fig. 2A). As visualization of actin requires fixation of the cells, differences in actin dynamics might have been missed. Actin cables can be visualized in living cells by the Abp140-GFP fusion protein (Yang and Pon, 2002). Interestingly, an Agrho1aΔ strain that carried a plasmid encoding AgAbp140-GFP grew poorly compared with the wild type with the same plasmid (Fig. 2B). Microscopy of the actin cytoskeleton was impossible because the small sick cells had high background fluorescence. However, the phenotype associated with expression of AgABP140-GFP suggests that the actin cytoskeleton of the Agrho1Δ strain has a weak defect. Therefore, we tested wild type and Agrho1aΔ for sensitivity to the actin polymerization inhibitor Latrunculin A. Fig. 2C,D shows that Agrho1aΔ cells are indeed slightly more sensitive to actin perturbation than wild-type cells, indicated by ∼1.5-fold increase of the inhibition zone by 10 nmol Latrunculin A. This suggests an involvement of AgRho1a in actin regulation. The observed lysis of hyphae could therefore be an indirect consequence of the actin defects, whereas the lethal lysis phenotype of the Agrho1b deletion could be the direct consequence of defects in cell wall biosynthesis.

A difference in protein function may be reflected by a different localization pattern. Therefore, we localized fusions of both proteins to YFP (AgRho1a) or CFP (AgRho1b) simultaneously by fluorescence microscopy. As we observed temperature-dependent lysis for the Agrho1aΔ mutants, we speculated that temperature might be an important factor for the function of AgRho1a. To test this, cells were grown either at 25°C or at 37°C. Fig. 3A shows that, at 25°C, AgRho1b localizes predominantly to the cortex and the cytoplasm at the tip of the growing hyphae, whereas AgRho1a localizes to the entire cytoplasm. Interestingly, a 1-hour heat shock at 37°C led to a change of localization pattern for both proteins. Now they were found at the cortex of the entire hyphae. In contrast to the localization of AgRho1b at 25°C, membrane association is not restricted to the tip region of the hyphae. This implies that at higher temperatures both proteins might share more functions than at lower temperatures, which might explain the reduced lysis at higher temperature observed for Agrho1aΔ mutants.

Whereas localization of the two AgRho1-proteins differs at the tip region of hyphae, we found both at developing (Fig. 3B, first row) and mature septa (Fig. 3B, second row), suggesting that both function in septum formation.

To test whether the area of AgRho1b localization at the tip at 25°C overlaps with the area of cell wall biosynthesis, we observed a GFP-fusion to AgRho1b in Aniline Blue stained hyphae. Aniline Blue is incorporated with 1,3-β-glucan into the forming cell wall (Kippert and Lloyd, 1995). Therefore, short incubation times with Aniline Blue preferentially stain the sites of cell wall biosynthesis. As shown in Fig. 3C, an overlap of Aniline Blue staining and GFP-AgRho1b is observed in growing tips and newly emerging branches. Even though this is not proof, a role of AgRho1b in cell wall synthesis is possible.

We performed complementation assays of yeast mutants to get further insight into the specific functions of the two AgRho1-proteins. S. cerevisiae is a close relative of A. gossypii but has only a single RHO1 gene. Therefore, complementation of S. cerevisiae rho1 mutants can be used to assay how the function of duplicated genes from A. gossypii may have evolved. To avoid artifacts from heterologous promoters, we constructed S. cerevisiae strains with an ORF replacement of the RHO1 gene by each of the A. gossypii homologs, resulting in a fusion of the A. gossypii genes to the ScRHO1 promoter. Fig. 4A shows representative examples of a tetrad analysis of heterozygous strains carrying ScRHO1 and either AgRHO1a (left side) or AgRHO1b (right side). Out of 50 tetrads analyzed for AgRHO1a, all showed a 2:2 segregation for viability and none of the viable spores carried the marker for AgRHO1a, proving that AgRHO1a is unable to complement the loss of ScRHO1. By contrast, segregants of a strain with AgRHO1b replacing ScRHO1 were viable, even though a 2:2 segregation with larger and smaller colonies was observed. The small colonies carried the marker of AgRHO1b in all cases tested. In addition to the tetrad analysis, we tested whether haploid cells, in which the genomic copy of ScRHO1 was replaced by either one of the AgRHO1 alleles, can be forced to loose a plasmid with an URA3 marker carrying the ScRHO1 gene on medium with 5-Fluoroorotic Acid (Fig. 4B). Consistent with the results from the tetrad analysis, plasmid loss was possible for Scrho1Δ::AgRHO1b but not for Scrho1Δ::AgRHO1a. However, in contrast to the tetrad analysis, where Scrho1Δ::AgRHO1b segregants were smaller, the Scrho1Δ::AgRHO1b strain grew like the wild type suggesting that the reason for the small size of the segregants is delayed germination and not a growth defect. To verify this, we tested the growth of the Scrho1Δ::AgRHO1b strain at high temperature (Fig. 4C), where a potential growth defect should become apparent. However, the segregants behaved like the wild type, even at temperatures as high as 39°C, suggesting that there is no growth defect associated with the Scrho1::AgRHO1b mutant.

This complementation of a Scrho1 deletion suggests that AgRHO1b and ScRHO1 functions are conserved, whereas the functions of AgRHO1a and ScRHO1 have diverged. The functions of AgRHO1b and ScRHO1 probably reflect the functions of the common ancestor RHO1 gene. This raises the question, has AgRHO1a lost the capability to act in both or only in one branch of ancestral Rho1-signaling? We performed complementation tests with temperature-sensitive alleles from both ScRHO1 complementation groups (Saka et al., 2001) to answer this question. Again, we used ORF-replacements of ScRHO1 with the A. gossypii counterpart but in addition we integrated temperature-sensitive ScRHO1 alleles at the LEU2 locus. For this purpose, we used the Scrho1-2 mutant that has a defect in the actin branch of Rho1 signaling and the Scrho1-4 mutant that has a defect in the glucan biosynthesis branch of Rho1 signaling (Saka et al., 2001). We had to increase the restrictive temperature to 39°C as the published 37°C (Saka et al., 2001) did not lead to the described phenotypes in our strain background. As shown in Fig. 4D (first row), cells with the Scrho1 mutants from both complementation groups can grow at the permissive temperature of 25°C but die at the restrictive temperature of 39°C. AgRHO1a only partially complements the mutant with a defect in actin regulation (Fig. 4D, second row). Together with the Latrunculin A sensitivity of the Agrho1a deletion, this supports the hypothesis of AgRHO1a being important for actin regulation. This experiment further suggests a loss of function of AgRHO1a in glucan biosynthesis as emphasized by the failure of AgRHO1a to complement the Scrho1-4 mutant, which has a defect in glucan biosynthesis. As expected from the ability of AgRHO1b to complement the loss of ScRHO1, AgRHO1b is able to complement the defects of both Scrho1-2 and Scrho1-4 (Fig. 4D, third row).

We wanted to know whether the observed changes in AgRho1a function are reflected by changes in its protein sequence compared with AgRho1b. Therefore, we compared the sequences of the homologs with each other and with more than 300 Rho-type GTP-binding proteins from different organisms. We identified a single histidine residue at position 39 in the switch I region of AgRho1a that is a conserved tyrosine in all 323 Rho-type GTP-binding proteins we examined. This atypical switch I region is shown in an alignment with examples from Rho-proteins of other model organisms in Fig. 5A. To exclude that this atypical residue was due only to a sequencing error or a mutation in our sample, we verified the histidine residue by resequencing the coding DNA region (not shown). We speculated that the atypical histidine might contribute to the difference in AgRHO1a and AgRHO1b function. If so, a mutant Agrho1a strain with a histidine-to-tyrosine exchange at position 39 should be able to take over the function of AgRho1b and consequently should be viable even if AgRHO1b is deleted, which is lethal in the wild-type background. To test this hypothesis, we mutated the cytosine at position 115 of AgRHO1a to a thymine, resulting in a change from histidine to tyrosine, analogous to the tyrosine of AgRho1b, and evaluated whether the mutated AgRho1a gains AgRho1b function. For this purpose, we integrated the mutated allele into an A. gossypii wild-type strain at the original AgRHO1a locus. The resulting strain was viable (Fig. 5B) although its radial growth speed was a bit slower than wild type (Fig. 5C). We then combined this mutation with a deletion of AgRHO1b. A heterokaryotic Agrho1bΔ strain was used as a control. Although spores from the strain only deleted for AgRHO1b never gave rise to a mature mycelium, the presence of Agrho1aH39Y restored growth and cells formed a mature mycelium (Fig. 5B). As shown in Fig. 5C, the growth speed of this strain was slower than wild type, and the difference in growth speed was less pronounced at 37°C. We hypothesized that, unlike the wild-type AgRHO1a, the mutant Agrho1a would be also capable of complementing the lethal ScRHO1 deletion. Therefore, we constructed an ORF-replacement of ScRHO1 with Agrho1aH39Y (Fig. 5D). In contrast to the wild-type AgRHO1a allele (Fig. 4A,B), Agrho1aH39Y fully restores growth in a background deleted for ScRHO1, even under heat stress at 39°C (Fig. 5D). Furthermore, GFP-AgRho1aH39Y localizes in a cortical pattern that is distinct from wild-type GFP-AgRho1a and similar to CFP-AgRho1b (Fig. 5E; Fig. 3A). Together with the complementation of the growth defects of an Agrho1b and an Scrho1 deletion, this is a strong indicator for the importance of the histidine 39 of AgRho1a for divergence of protein function between the two AgRho1 proteins.

We wondered how a single residue could influence the function of AgRho1a in such a dramatic way. The residue that is altered in AgRho1a is located in the switch I region of the small GTP-binding protein. This is one of the regions that undergoes a nucleotide-dependent conformational change (Milburn et al., 1990). A survey of protein structures shows that the tyrosine residue, which corresponds to the atypical histidine in AgRho1a, is directly involved in interaction of RhoA, Rac1 and Cdc42 with other proteins (Dvorsky et al., 2004). Interestingly, none of the 13 different structures of protein complexes where the conserved tyrosine in the switch I region participated in the interaction interface involved an effector molecule. Instead, all complexes were formed with regulators of small GTP-binding proteins like GEFs, GDP-dissociation inhibitors (GDIs) and mainly GAPs. This suggests that the different functions of the two AgRho1-proteins are not caused by different effector interactions. Consistently, the two AgRho1-proteins do not show differences in interactions with A. gossypii homologs of known S. cerevisiae Rho1 effector proteins in a two-hybrid assay. The only difference we observed was a weak interaction between AgRho1b and AgSec3 that was not found with AgRho1a (Table 1). However, Agrho1aH39Y or Agrho1bY40H mutations did not affect this interaction. Together with the results shown in Fig. 5D, this indicates that interaction with AgSec3 does not significantly contribute to the functional differences observed for the AgRho1 proteins.

Instead, we speculated that the difference in function of the AgRho1-proteins might come from a different affinity of AgRho1a and AgRho1b to a regulatory protein, probably belonging to the class of GAPs. As two important regulators of the different branches of ScRho1 signaling in S. cerevisiae are the two GAPs ScLrg1 and ScSac7 (Lorberg et al., 2001; Schmidt et al., 2002; Watanabe et al., 2001), we set up a two-hybrid assay to test for interaction of the homologous proteins from A. gossypii with the AgRho1-proteins. We have chosen the two-hybrid system for two reasons. First, owing to the low abundance of the GAP proteins, a successful co-immunoprecipitation of the GAPs together with the AgRho1-proteins from A. gossypii cell extracts was not possible. Second, Jaitner et al. (Jaitner et al., 1997) showed for a small GTP-binding protein and its binding partner that the interaction strength measured by two-hybrid experiments correlates in a semilogarithmic fashion with the dissociation constants. As only the GTP-bound forms of the GTP-binding proteins interact with the GAP proteins, we used both the wild-type and variants that carry a glutamine-to-histidine mutation in the switch II region, which results in GTP-locked variants of the proteins (Q68H, Q69H). We preferred this allele to other GTP-locked alleles because we already had experience from two-hybrid experiments with the homologous mutant from S. cerevisiae (Schmitz et al., 2001). As shown in Fig. 6A, GTP-locked AgRho1a interacts with AgLrg1, but interaction with AgSac7 was barely detectable. This raises the question, is this interaction completely absent or is it too weak to be detected by a two-hybrid assay? We addressed these possibilities by performing a co-precipitation of the two proteins in presence of aluminium fluoride. Aluminium fluoride forms a stable complex with GTP-binding proteins in the GDP-bound form and with their corresponding GAP proteins (Ahmadian et al., 1997). If there is any interaction between AgRho1a and AgSac7, the complex between GAP, GTP-binding protein and aluminium fluoride can be pulled down, even if the interaction is transient and weak. We expressed GST-tagged AgRho1a and AgRho1b and the hexahistidine-tagged GAP-domain of AgSac7 in E. coli. The GAP domain of AgSac7 can be pulled down with both AgRho1a and AgRho1b (Fig. 6B). Attempts to perform a similar experiment with AgLrg1 were unsuccessful because we were unable to obtain expression of soluble protein, even from constructs with fragments of different length and under different conditions. However, together with the two-hybrid data, the results for AgSac7 show that AgRho1a and AgSac7 do interact and can be co-purified if the complex is stabilized, but that this interaction is probably so weak that it is close to the detection limit of the two-hybrid assay.

The situation for AgRho1b is nearly the opposite to AgRho1a. GTP-AgRho1b shows only a weak interaction with AgLrg1 but a strong interaction with AgSac7 (Fig. 6A). This interaction is stronger than the background detected with a vector control, but clearly weaker than the interaction between AgSac7 and AgRho1b.

We also tested AgRho1aH39Y to see whether the mutation influences binding to a GAP protein. Indeed, the switch I mutation resulted in a stronger interaction between AgRho1a and both GAPs. Although this is obvious for AgSac7, where interaction of the protein with a wild-type switch I is barely detectable, we used 3-aminotriazole to obtain an estimate of interaction strength with AgLrg1 (Fig. 6C). In this assay, yeast cells can grow on higher concentrations of 3-aminotriazole if they produce more of the His3 reporter. Therefore, stronger interaction between bait and prey of a two-hybrid assay leads to higher resistance against 3-aminotriazole. Although yeast cells with the AgRho1a wild-type protein as a bait and AgLrg1 as a prey do not grow on 2 mM 3-aminotriazole, a strain with AgRho1aH39Y can grow even on 5 mM 3-aminotriazole, indicating a much stronger interaction. This change in GAP-interaction strength is probably the reason for the capability of an AgRho1aH39Y mutant to take over AgRho1b function.

As the position 39 (taking AgRho1a numbers as a reference) is not the only residue that is different between the AgRho1 proteins, we asked whether there are other residues involved in the different GAP interactions. We looked at the 27 residues that are known to form the interaction interface with GAP proteins from structural studies of homologous proteins (Dvorsky et al., 2004). In addition to position 39, only three of these (positions 45, 98 and 139) differ between AgRho1a and AgRho1b. Only one of these (position 45) is located in the switch I region (Fig. 5A). Interestingly, the glutamic acid at that position in AgRho1a is also found in S. cerevisiae Rho1, whereas the aspartic acid of AgRho1b is conserved in Rho1 of Kluyveromyces lactis. Therefore, we tested these proteins for interaction with the A. gossypii GAPs (Fig. 6D). Although there is hardly any interaction with both GAPs and KlRho1, ScRho1 interacts with AgLrg1 but not with AgSac7. In addition, we mutated tyrosine 39 of ScRho1 to histidine to mimic the switch I region of AgRho1a (Y39H). In agreement with our findings from above the Y39H mutation weakens the interaction with AgLrg1. These results suggest that in addition to the residue at position 39, other residues might contribute to the difference in GAP interaction of AgRho1a and AgRho1b. However, this may not be the residue at position 45 because this residue is identical in AgRho1b and KlRho1 but the latter does not interact with any of the A. gossypii GAP-proteins. Instead it might be that the interaction surface of the AgRho1 proteins and their GAPs might include different residues, which also requires adaptations in the sequence of the GAPs in addition to differences in the sequence of the Rho proteins. To determine whether there is a different regulation of the two AgRho1 proteins by the GAPs, as suggested by the interaction studies, we constructed strains carrying a mutation that encodes dominant active forms of AgRho1a and AgRho1b. We then compared the phenotypes of these strains with phenotypes of strains deleted for either Agsac7 or Aglrg1. Removal of a GAP from the regulatory circuit of a small GTP-binding protein leads to an accumulation of a GTP-bound Rho protein and therefore to a phenotype that is similar to the GTPase-deficient form of the small GTP-binding protein. A deletion of AgLRG1 or a GTPase-deficient allele of AgRHO1a resulted in slow growth compared with wild type. However, deletion of AgSAC7 or introduction of a GTPase-deficient allele of AgRHO1b was lethal (Fig. 7A). Together with the interaction data, these phenotypes support the idea that AgRho1a is the preferred target of AgLrg1, whereas AgRho1b might be the preferred target of AgSac7.

Our results suggest that GAP-mediated GTP hydrolysis of AgRho1a should be lower than GTP hydrolysis of AgRho1b in presence of AgSac7. To test this, we expressed the GTP-binding proteins and the GAP domains of these A. gossypii proteins in E. coli and purified them for an in vitro GAP assay (Fig. 7B). The assay we used detects the release of phosphate from GTP hydrolysis by a color reaction that can be monitored by an increase in absorbance at a wavelength of 650 nm. GTP hydrolysis of all control reactions, each GTP-binding protein and the GAP on its own was below OD 0.1. We used the GAP-GTP-binding protein pair human p50RhoGAP and RhoA that was supplied with the assay as a positive control. RhoA had the highest rate of GTP hydrolysis with both human p50RhoGAP and the AgSac7 GAP domain. With both of these GAP proteins, GTP hydrolysis of AgRho1b was significantly higher than GTP hydrolysis of AgRho1a. The interaction data, together with genetics and the GAP assay, suggest that both GAPs still are able to act on both GTP-binding proteins. However, this raises the question, is the GAP with the higher or the GAP with the lower affinity the regulator of a Rho protein in vivo? Both situations seem to be possible. A higher affinity would result in a greater specificity of the reaction, whereas a lower affinity would result in an increase of the GTP-bound form of the Rho protein. An answer to this question might come from the localization of the GAP proteins. To investigate this, we fused GFP to AgLrg1 and AgSac7. AgLrg1-GFP localized to the cortex of the hyphal tip, resembling the localization pattern of AgRho1b. AgSac7-GFP was found in the cytoplasm similar to AgRho1a (Fig. 7C). This favors the second situation described above. AgLrg1 is probably part of the regulatory circuit of AgRho1b. The purpose of its high affinity towards AgRho1a might be to inactivate AgRho1a efficiently at the tip region. The same may be true for AgSac7 and AgRho1b. AgSac7 probably regulates AgRho1a and might prevent activity of AgRho1b when AgRho1b is not at the tip.

In summary, our results show that AgRho1a function has evolved by mutation of a single residue in the switch I region, leading to a situation where AgLrg1 regulates AgRho1a and AgSac7 regulates AgRho1b, which in turn results in divergence of protein localization and function.

The initial question we addressed in this study was how the function of a Rho-type GTP-binding protein diverges after gene duplication. We found that during evolution, alteration of a single tyrosine to a histidine in the switch I region of one of the two Rho1 proteins encoded by the filamentous fungus A. gossypii can influence interaction with GTPase-activating proteins. This has an effect on protein function without altering the specificity for effector proteins. The finding that GTP-binding protein function evolves by mutation of the switch region is remarkable because this region, and especially the tyrosine at position 39 (AgRho1b numbering), is highly conserved. By contrast the large variety of Rho-GTP binding proteins has probably evolved by mutation of the more variable parts of the interaction interface (Wherlock and Mellor, 2002). There are only a few reports about small GTP-binding proteins where this is not the case: Heo and Meyer (Heo and Meyer, 2003) developed an algorithm to generate mutations that lead to a switch of Rho-protein function. They found that mutations outside the effector interaction interface can induce a switch of function. However, these switch of function mutations were also outside the conserved switch regions of the GTP-binding proteins. The latter is also true for a group of three closely related Rac proteins that have different functions (Haeusler et al., 2003). In this study, the residues that are thought to be responsible for the different functionality have an effect on the nucleotide binding capability and a different activation by a GEF protein but they are also located outside the switch region.

What is the effect of the differences in AgRho1a and AgRho1b sequence? We found that especially the histidine at position 39 in the switch I region of AgRho1a represents a mutation that occurred during evolution and alters interaction of AgRho1a with regulatory proteins, namely with GAPs. The homologous residue in the 323 other Rho-proteins we checked (including AgRho1b and ScRho1) was without exception a conserved tyrosine. Even though there are other differences between AgRho1a and AgRho1b, it was possible to revert AgRho1a function to AgRho1b function by simply changing histidine 39 of AgRho1a to a tyrosine. The resulting AgRho1aH39Y was able to compensate for the loss of AgRHO1b, whereas an AgRHO1b deletion alone was lethal (Fig. 5B,C). The localization pattern of AgRho1aH39Y mimicked that of AgRho1b (Fig. 5E). The atypical histidine 39 of AgRho1a is found in the switch I region, which is known to interact with GAPs in various small GTP-binding proteins (Dvorsky et al., 2004). For this reason, we investigated the interactions of AgRho1a and AgRho1b with the two Rho1-GAPs AgLrg1 and AgSac7. We found that Rho1a interacts preferentially with AgLrg1 and weakly with AgSac7, whereas AgRho1b shows strong interaction with both GAPs (Fig. 6). The difference in two-hybrid interaction strength is reflected by the ability of AgSac7 to activate GTP-hydrolysis more efficiently for AgRho1b than for AgRho1a as shown in an in vitro GTPase activity assay (Fig. 7). Interestingly, this strong interaction with the Rho GAPs seems to be crucial for AgRho1b function, as AgRho1aH39Y can provide cell viability in absence of AgRho1b and shows GAP interaction similar to AgRho1b. This finding implies that precisely controlled GTP hydrolysis is crucial for the different AgRho1 functions. Small GTP-binding proteins do not only work as on/off switches. It has been shown that the cycling between the GDP- and the GTP-bound form is important for their functionality (Gladfelter et al., 2002). Evidence for this theory comes from findings that cycling of ScCdc42 is important for its function (Gladfelter et al., 2002). Interestingly, an Sccdc42 mutant that was derived from an in vitro mutagenesis has an exchange of tyrosine for histidine in the switch I region like it is present in AgRho1a (Gladfelter et al., 2002). In agreement with our results, this Sccdc42 mutant could not be activated to the same level of GTPase activity by its GAP compared with wild-type ScCdc42 with a tyrosine in switch I. Therefore, AgRho1aH39Y might be able to compensate for AgRHO1b loss by having similar GDP/GTP cycling properties as AgRho1b and thus mimic AgRho1b action. Such a scenario would suggest that if both AgRho1 proteins are present at the same time together with one of the GAPs, the Rho protein with the lower affinity towards the GAP would be able to cycle between GTP- and GDP-bound state, and the majority of the other Rho proteins would be kept inactive. In such a situation, the GAP would colocalize with the Rho protein that the GAP binds with a lower affinity. In agreement with this theory, AgLrg1 and AgRho1b localize to the tips of hyphae, whereas AgSac7 and AgRho1a both are found in the cytoplasm (Fig. 7C). At a first glance, this seems to be contradictory to the phenotypes of the activated AgRho1 alleles and the GAP deletions (Fig. 7A). However, in an Agsac7Δ strain, AgRho1b would be active at a location, where it usually is completely inactive. Therefore, this phenotype correlates with the Agrho1bQ69H mutant. The same would be the case for the Aglrg1Δ strain and Agrho1aQ68H. This would also mean that the respective roles of the two GAPs are conserved between A. gossypii and S. cerevisiae, because ScLrg1 primarily regulates biosynthesis of 1,3-β-glucan and ScSac7 primarily regulates the Pkc1 branch of Rho1 signaling (Watanabe et al., 2001). Such a hypothesis is supported by the data we obtained from deletion analysis, localization and complementation studies (Figs 1, 2, 3, 4). All these indicate that AgRho1a is involved in actin regulation but has lost the glucan biosynthesis function that is preserved in ScRho1 and AgRho1b, as shown by complementation of S. cerevisiae mutants (Fig. 4). In addition AgRho1a gained a novel function that prevents cells from lysis at low temperature.

The fact that both Rho1 proteins have different functions in A. gossypii raises the question, why could this be an advantage for the cell? Different aspects have to be taken into account when answering this question. First, there are more Rho1 regulators in S. cerevisiae than in A. gossypii. Fig. 8 compares the core machinery of the Rho1 regulatory network of both organisms. The GEFs ScRom1 and ScRom2 (Ozaki et al., 1996), as well as the GAPs ScSac7 and ScBag7 (Schmidt et al., 2002), are so-called `Twin-ORFs', remnants of a whole genome duplication that occurred in the S. cerevisiae lineage (Dietrich et al., 2004). These proteins add a level of complexity to ScRho1 regulation and thus might allow the adaptation of activity of ScRho1 to a variety of intra- and extracellular signals. This observation leads to a simple explanation for the AgRHO1 duplication: it might have been maintained in the genomes of various species because it simply expands the organisms' capacity to fine-tune Rho1 activity to a similar level as with duplication of several GAP or GEF proteins. In good agreement with such a hypothesis is the fact that a tandem RHO1 duplication is also found in Kluyveromyces waltii, Saccharomyces kluyveri and Holleya sinecauda (Cliften et al., 2003; Kellis et al., 2004; Walther and Wendland, 2005). Like A. gossypii, K. waltii and S. kluyveri have not undergone the whole genome duplication that is found with S. cerevisiae, and therefore also have fewer GAP and GEF proteins. Even though there is no genome sequence available for H. sinecauda, the close evolutionary relationship to A. gossypii suggests a non-duplicated genome.

Another reason why it might be advantageous for A. gossypii to have two Rho1 proteins might be due to the filamentous growth of A. gossypii. In S. cerevisiae, cell cycle and cell division are tightly coupled (Philippsen et al., 2005). The major part of cell wall expansion and thus 1,3-β-glucan biosynthesis occurs during bud growth. The duration of this phase is regulated by the cell-cycle machinery, which means that full Rho1 activity in 1,3-β-glucan biosynthesis is required only for a specific period during the cell cycle. Therefore, the different branches of ScRho1 activity might be temporally restricted. This temporal restriction of polar growth is not present in A. gossypii, rather polar growth is active in the filamentous fungus all the time (Philippsen et al., 2005). The mycelium even maintains many growing tips at the same time and tip growth is independent from the asynchronous nuclear cycles (Gladfelter et al., 2006). This tremendous polarized growth requires a highly active 1,3-β-glucan biosynthesis rate. As a consequence, this may saturate most of the Rho1 activity. The presence of a second independently regulated Rho1 protein would allow the cell to uncouple the different branches of the Rho1-signaling pathway without the need to stop polar growth and glucan biosynthesis. Therefore, the different localization of the two Rho1-proteins that we observed for AgRho1a and AgRho1b (Fig. 3A), might help to completely uncouple the different branches of Rho1-signaling. This hypothesis suggests that the duplicated RHO1 gene might be common to all filamentous fungi. In contrast to this, only a single RHO1 homolog was identified in the genomes of fungi such as Neurospora crassa (Galagan et al., 2003) or Aspergillus fumigatus (Nierman et al., 2005). However, these species have another small GTP-binding protein involved in control of the actin cytoskeleton (Borkovich et al., 2004): a Rac1 homolog that is missing in all species closely related to S. cerevisiae and therefore also in A. gossypii. This suggests that some filamentous fungi might use different strategies to achieve fast filamentous growth.

In summary, we show here that a simple mutation in the switch I region of a GTP-binding protein can change its affinity towards its GAPs, which probably leads to slight variations of the GDP-GTP cycling and finally leads to a decoupling of still very similar protein function without impairing effector interaction.

According to the guidelines used for the A. gossypii gene annotation, we refer to ABR183W as AgRHO1b and to ABR182W as AgRHO1a, whereas in an initial characterization the authors referred to ABR183W as RHO1 and ABR182W as RHOH (Wendland and Philippsen, 2001; Walther and Wendland, 2005).

All strains are listed in supplementary material Table S1. Deletion strains were constructed by PCR-based gene targeting, according to the method described by Wendland et al. (Wendland et al., 2000). All oligonucleotides are listed in supplementary material Table S3. Strains were grown at the indicated temperatures either in AFM (Ashbya Full Medium) with or without Geneticin or CloNAT, or in synthetic medium when selecting for auxotrophic markers (Knechtle et al., 2003).

The genotypes of all strains used in this study are listed in supplementary material Table S2. S. cerevisiae strains were grown at 30°C if not indicated otherwise. Media were used according to Amberg et al. (Amberg et al., 2005).

All DNA manipulations were carried out according to Sambrook et al. (Sambrook et al., 2001) with DH5αF′ as host (Hanahan, 1983). Sequencing was carried out using an ABI DNA Sequencer according to the instructions of the manufacturer.

All plasmids and constructs are listed in supplementary material Table S4 together with construction details. Plasmid sequences are available from the authors upon request.

The overlay assay for detection of cellular lysis was adapted from Saka et al. (Saka et al., 2001). A mycelia suspension (4 μl) was spotted on AFM plates for every strain and incubated at 30°C for 36 hours. The plates were then shifted to 25°C, 30°C or 37°C for 14 hours and overlaid with 10 ml lysis indicator agarose (1% agarose in 0.05 M glycine-HCl buffer pH 9.5) containing 10 mM BCIP (5-bromo-4-chloro-3-indonyl phosphate disodium salt). The plates were scanned after 90 minutes of incubation at room temperature using an EPSON Perfection 1200 PHOTO scanner (Long Beach, USA). The average signal intensity of the blue channel of the RGB image was determined on an area of 2960 pixels using MetaMorph v6.2 software (Universal Imaging Corporation, Downingtown, USA). Lysis of deletion strains was calculated relative to negative (wild type) and positive control strains (Agslt2Δ).

Dense spore preparations (10 μl) were mixed with 4 ml warm AFM containing 1% molten agar and 0.03% Triton X-100 and immediately poured on AFM agar plates. Filter paper disks (5 mm in diameter) were placed on the spore agar surface. The indicated amounts of latrunculin A were pipetted on the disks in a total volume of 5 μl DMSO. The growth inhibition halos were measured after 36 hours of incubation at 25°C.

Cells were grown overnight in liquid medium. Prior to microscopy, they were placed on slides with small cavities that were filled with half concentrated AFM with 2% glucose and 0.5% agarose and allowed to recover for 2 hours at the indicated temperatures. The setup used for in vivo fluorescence microscopy consisted of a Zeiss Axioplan2e (Carl Zeiss, Jena, Germany) equipped with 100× alpha-Plan Fluar objective (NA 1.45) and differential-interference-contrast (DIC). Images were acquired using a Photometrics CoolSNAP HQ Camera (Roper Scientific, Tucson, USA). Fluorescence was excited with a xenon lamp and appropriate filtersets were used to adjust excitation and emission wavelengths. The setup was controlled by MetaMorph v6.2 (Universal Imaging Corporation). DIC images were acquired as single planes. Images were scaled using MetaMorph scale image command and Huygens essential software was used for image deconvolution. Image overlays were carried out using the MetaMorph overlay images command.

For two-hybrid experiments, prey and bait plasmids were transformed into a and α derivatives of S. cerevisiae strain PJ69-4a. Strains were crossed and selected on minimal medium without leucine and tryptophan containing 80 mg/l adenine. Transformants were grown to an OD₆₀₀ of 1.5 μl of these cultures were spotted on synthetic medium agar plates lacking leucine, tryptophan and histidine for selection on interaction. To test the strength of interactions, 5-aminotriazole was added to the media at the indicated concentrations.

Proteins were expressed using E. coli BL21(DE3) as a host (Stratagene, La Jolla, USA). Induction, expression and GST-purification were carried out as described in detail by Smith and Rittinger (Smith and Rittinger, 2002). Purification of the hexahistidine-tagged AgSac7p was performed using an ÄKTA FPLC system (GE Healthcare Life Sciences, Pittsburgh, USA) with 1 ml HisTrap FF column according to the instructions of the manufacturer. A Coomassie stained SDS-page of the purified proteins is shown in supplementary material Fig. S1.

GST pulldown of GAP and GTP-binding proteins was performed in the presence of aluminium fluoride by addition of 10 mM NaF and 450 μM AlCl₃ to lysis and washing buffers for stabilization of the complex. Apart from the addition of aluminium fluoride, buffers and conditions used were similar to the purification of GST-tagged proteins described above. Western blots were probed using anti-GST from rabbit (G7781 Sigma, St Louis, USA) and anti-Penta-HIS from mouse (34660 Qiagen, Hilden, Germany) as primary and anti-rabbit coupled to IRdye700 (611-730-127, Rockland Immunochemicals, Gilbertsville, USA) and anti-mouse coupled to IRdye800 (926-32210, Li-Cor, Lincoln, USA) as secondary antibodies. Signals were detected with an Odyssey infrared imaging system (Li-Cor, Lincoln, USA).

Western blots were probed with a 1:1000 dilution of antibodies directed against GFP (11 814 460 001 Roche, Mannheim, Germany).

The in vitro GAP assay was performed using the RhoGAP assay Biochem Kit (BK105, Cytoskeleton, Denver, CO). RhoA and p50RhoGAP were supplied with the kit.

We thank Peter Philippsen and Jürgen J. Heinisch for their constant support; Peter Philippsen, Amy S. Gladfelter and Anja Lorberg for discussions and critical reading of the manuscript; Mohammad Reza Ahmadian for discussions; Peter Uetz, Mike Hall, Yoshikazu Ohya and Dominic Hoepfner for strains and plasmids; and Lars Molzahn, Lynn Ullmann, Stéphanie Hauselmann and Déborah Ley for their work during their laboratory internship.