The small GTPase KlRho5 responds to oxidative stress and affects cytokinesis

Small GTPases comprise a huge family of molecular switches, which cycle between a GDP-bound inactive and a GTP-bound active state and thereby regulate essential biological processes from signaling cascades and intracellular protein trafficking to cell motility (reviewed in Takai et al., 2001). Among them, the subfamily of Ras-homologous (Rho) GTPases is widespread within eukaryotes, with the model yeast Saccharomyces cerevisiae carrying six members, Rho1 to Rho5 and Cdc42 (Park and Bi, 2007). Rho5 was the last to be identified, then as a negative regulator of cell wall integrity (CWI) signaling (Schmitz et al., 2002a). It is now considered a homolog of human Rac1, which controls actin cytoskeleton dynamics and also affects energy metabolism in mammalian cells, with its malfunction associated with diverse diseases, including cancer development and neurodegenerative disorders (Eliáš and Klimeš, 2012; Marei and Malliri, 2017). Supporting its homologous function, we found that activation of Rho5 in S. cerevisiae depends on a dimeric guanine-nucleotide-exchange-factor (GEF), consisting of Dck1 and Lmo1, the yeast homologs of mammalian DOCK and ELMO proteins (Schmitz et al., 2015).

Further investigations in S. cerevisiae underlined the importance of Rho5 as a putative hub for various signal transduction processes (reviewed in Hühn et al., 2019). Thus, besides regulating the CWI pathway, it participates in the regulation of the high-osmolarity glycerol (HOG) pathway (Annan et al., 2008). Rho5 also reacts to glucose starvation by translocating rapidly from the plasma membrane to mitochondria (Schmitz et al., 2018, 2015). An implication in nutrient sensing was also deduced from the synthetic phenotypes of a rho5 deletion with mutants defective in nutrient signaling components, such as sch9 and gpr1 (Schmitz et al., 2018). Large-scale screens further supported such an association, as Rho5 was found to interact with a large number of glycolytic enzymes as well as with Ras1 (Singh et al., 2019a). Rho5 also translocates rapidly to the mitochondria under oxidative stress, depending on its interaction with and activation by both GEF subunits (Sterk et al., 2019). Recent studies further confirmed the involvement of Rho5 in mitochondrial dynamics, as they revealed interactions with Atg21, a central component of autophagy, and Msp1, an outer mitochondrial membrane protein (Singh et al., 2019a). This was in line with previous suggestions of a role for Rho5 in the regulation of yeast mitophagy and apoptosis (Schmitz et al., 2015; Singh et al., 2008). Finally, interactions with major eisosome components such as Pil1 and Sur7 provided additional evidence for its relation to cell surface and cell wall organization, as indicated by the initial genetic studies (Schmitz et al., 2002b; Singh et al., 2019a).

With regard to its domain structure, Rho5 in S. cerevisiae shares many features with other eukaryotic GTPases. Thus, it carries the two switch regions required for nucleotide binding and activation, as well as a C-terminal CAAX box as a target for lipid modification and association with the plasma membrane (Sterk et al., 2019). The latter also relies on a conserved polybasic region (PBR), which precedes the CAAX motif. A specific feature of Rho5 in S. cerevisiae, as opposed to its human Rac1 homolog, is an extension of 98 amino acid residues upstream of the PBR, which participates in its proper intracellular distribution.

The yeast Kluyveromyces lactis is amenable to both classical and molecular genetic techniques developed for S. cerevisiae, but relies on a predominantly respiratory rather than a fermentative metabolism (Rodicio and Heinisch, 2013). In broad terms, the physiology of K. lactis cells thus mimics that of healthy mammalian cells, whilst the fermentative S. cerevisiae would be more similar to cancer cells (Andrejc et al., 2017). The respiratory orientation of K. lactis is also reflected in the fact that in cells grown in glucose media, mitochondria are more abundant in images from transmission electron microscopy (TEM) than in S. cerevisiae, whilst other features such as the cell wall are alike in the two yeasts (Backhaus et al., 2010). In line with these observations, the p53-dependent generation of reactive oxygen species (ROS) and related apoptotic processes were shown to be independent of the carbon source in K. lactis, but only occurred under non-fermentative conditions in S. cerevisiae (Kumar et al., 2017).

Despite the similarities in molecular investigations to S. cerevisiae, very little information is available on GTPases in K. lactis. In fact, as far as Rho-type GTPases are concerned, only KlRho1 has been studied in some detail. It is an essential protein that interacts with protein kinase C (KlPkc1), and is activated by the GEF KlRom2, suggesting that it serves a similar function in CWI signaling as its homolog in S. cerevisiae (Lorberg et al., 2003; Rodicio et al., 2006). Non-Rho-type GTPases studied so far in K. lactis include Gpa1, the alpha-subunit of the trimeric complex in the mating pheromone pathway (reviewed in Coria et al., 2006), KlRas2, whose role so far has only been studied in relation to mating and mating-type switching (Barsoum et al., 2011), and the eukaryotic initiation factor eIF2 for its function in translation (Llacer et al., 2018). Additional GTPases in K. lactis have been described with five septins, which control cytokinesis and are functionally conserved to their counterparts in S. cerevisiae (Rippert and Heinisch, 2016). On the other hand, KlMyo1, a component of the constrictive actomyosin ring (AMR), seems to be less important for K. lactis than its homolog in S. cerevisiae, as Klmyo1 deletions are perfectly viable, although they are temperature sensitive and display morphological defects at 37°C (Rippert et al., 2014). Nevertheless, in wild-type K. lactis cells the dynamics of AMR constriction and the formation of primary and secondary septa follow those known for S. cerevisiae, and budding leaves relatively flat scars on the mother cells (Rippert et al., 2014).

Given the medical importance of Rac1 in humans, and the association of Rho5 with oxidative stress response and apoptosis in S. cerevisiae, in the present study we decided to study the role of Rho5, and that of its dimeric GEF, in the respiratory yeast K. lactis. We showed that, similarly to its homolog in S. cerevisiae, KlRho5 responds to both oxidative stress and nutrient limitation by translocating together with its GEF to mitochondria. Yet, genetic interactions with two key components of nutrient sensing, Sch9 and Gpr1, are not conserved in K. lactis, probably because of its predominantly respiratory metabolism. Importantly, a lack of KlRho5 activity results in significantly thicker cell walls and in distinct morphological defects caused by an aberrant budding process, which is different from S. cerevisiae and reminiscent of Rac1 functions in the organization of the mammalian cytoskeleton.

Rho5 together with its dimeric GEF, henceforth referred to as the DLR complex, mediates the response to oxidative stress and glucose starvation in the fermentative yeast S. cerevisiae (Schmitz et al., 2018, 2015). We wondered whether (1) this complex exists in the more respiratory-oriented yeast K. lactis and (2) if its intracellular distribution and physiological function is affected in a similar way under stress. To address the first question, a search for homologous genes with synteny in the yeast gene order browser (http://ygob.ucd.ie/), complemented by sequence alignments of the deduced proteins, revealed the presence of homologs of ScRHO5, ScDCK1 and ScLMO1 in the K. lactis genome, with overall amino acid identities of 47%, 38% and 27%, respectively (see Fig. S1 for alignments). As expected, regions covering putative functional domains of the two GEF subunits, such as the SH3 domain near the N-terminus of KlDck1 and putative sites of interactions, show higher degrees of conservation. The GTPase KlRho5 was also aligned with its human homolog HsRac1, showing 43% amino acid identity, which is significantly higher compared to only 32% between ScRho5 and HsRac1 (Fig. S1). The lower identities with ScRho5 can be attributed to its extension by 98 amino acids, which precedes the C-terminus and serves important functions in the S. cerevisiae GTPase (Sterk et al., 2019). This extension is much shorter or completely absent in KlRho5 and HsRac1, respectively (Fig. S1).

Next, the three identified genes were cloned with their flanking regions by PCR from the genome of a K. lactis wild-type strain to facilitate their molecular analysis, as outlined in the Materials and Methods section. To confirm that the Rho5 proteins of K. lactis and S. cerevisiae were functionally equivalent, we employed two independent tests for heterologous complementation. First, either the wild-type KlRHO5 gene or its activated mutant allele KlRHO5^Q69H was introduced into a recipient strain carrying an Scrho5 deletion, which confers hyper-resistance towards hydrogen peroxide (Schmitz et al., 2015). As shown in Fig. S2A, the KlRHO5 alleles both restored the growth sensitivity under oxidative stress in drop dilution assays, in contrast to the empty control vector. In fact, expression of the activated allele provoked a hyper-sensitivity, similar to published data on its S. cerevisiae homolog (Sterk et al., 2019). Complementation capacity was also validated for the synthetic lethality of rho5 null mutants with an sch9 deletion in S. cerevisiae (Schmitz et al., 2018). Introduction of the wild-type KlRHO5 gene or its activated mutant allele KlRHO5^Q69H into a heterozygous diploid strain (rho5/RHO5 sch9/SCH9), followed by sporulation and tetrad analysis, yielded only viable segregants with the double deletion (rho5 sch9) if they also carried a plasmid encoding a functional KlRho5 (Fig. S2B).

To assess the physiological role of the putative DLR complex in K. lactis (KlDLR), the three genes were individually deleted (Fig. 1A). All three deletions (Klrho5, Kldck1 and Kllmo1) produced viable segregants in tetrad analyses, with colony sizes similar to the wild-type segregants on YEPD plates, indicating that KlRho5 activity is not essential (Fig. 1B). Segregants from such analyses were selected for growth assays on rich medium containing hydrogen peroxide, since Scrho5 deletions are known to be hyper-resistant to this stressor. As evident from Fig. 1C, cells lacking any of the three KlDLR components indeed displayed an increased resistance towards oxidative stress as compared to the wild type.

Since the hyper-resistance of Klrho5 deletions towards hydrogen peroxide appears to be too moderate to serve as a suitable phenotype for genetic screens, we looked for stronger criteria to perform complementation studies. As stated above, Scrho5 deletions are synthetically lethal with an Scsch9 deletion and synthetically sick with Scgpr1 deletions, attributed to a connection to nutrient sensing (Schmitz et al., 2018). Therefore, we identified the homologs KlSCH9 and KlGPR1 by synteny analyses and cloned and deleted the respective genes from the K. lactis genome (Fig. 2A,B). Segregants carrying a Klsch9 deletion obtained from a heterozygous diploid strain produced drastically smaller colonies than those of the wild type upon germination on YEPD plates (Fig. 2C), whereas strains carrying the Klgpr1 deletion were not affected (Fig. 2D), both reminiscent of the respective deletion strains in S. cerevisiae (Schmitz et al., 2018). However, in contrast to S. cerevisiae, both the Klrho5 Klsch9 and the Klrho5 Klgpr1 double deletions were viable, and neither showed an additive phenotype as compared to the segregants with the single deletions (Fig. 2E and F, respectively).

A total of fifteen other gene deletions were tested in attempts to find a synthetic lethality with the Klrho5 deletion, either in analogy to reported genetic interactions in S. cerevisiae or because they seemed plausible in the light of the phenotypes observed herein. These included crosses with strains defective in carbohydrate signaling, the pentose phosphate pathway, oxidative stress response, cell wall synthesis, cytokinesis or DNA repair. As listed in Table S3, haploid segregants carrying the double deletions in combination with Klrho5 were obtained from tetrad analyses in all cases, with no obvious additive growth phenotypes, i.e. they were neither synthetically lethal nor obviously synthetically sick.

An interesting phenotype reported for an Scrho5 deletion was a reduced chronological life span in the genetic background of the BY-strain series in baker's yeast (Cao et al., 2016). We therefore assessed the loss of cell viability in stationary phase cells kept in synthetic minimal medium for three different haploid strains, each, carrying either a Klrho5 deletion or a wild-type KlRHO5 allele. All strains showed a similar loss of viability over a period of 3 weeks, with no significant differences between wild type and the deletions (Fig. S3), indicating that KlRho5 is not involved in the regulation of chronological life span in K. lactis.

Given the differences in primary structures of the three KlDLR proteins to their homologs in S. cerevisiae, and the lack of analogous synthetic interactions, we next addressed the distribution of the trimeric complex in K. lactis under different growth conditions to shed some light on their cellular functions. To this end, GFP fusions were constructed for each of the three genes at their original chromosomal loci, expressed from their native promoters. KlDck1-GFP and KlLmo1-GFP were obtained as C-terminal fusions, in analogy to their functional S. cerevisiae homologs. GFP-KlRho5 was tested as an N-terminal fusion, to prevent interference with the C-terminally located CAAX-box, which is required for proper localization (Fig. S1; Sterk et al., 2019). We noted that the genomic GFP-KlRHO5 fusion encoded a GTPase that was not fully functional, as judged by the increased appearance of morphologically aberrant cells (see next section). In order to avoid this phenotype, GFP-KlRHO5 was cloned into a non-replicating vector and integrated at the K. lactis leu2 locus, leaving the wild-type KlRHO5 allele at its native chromosomal position intact.

While the GFP-tagged variants of KlDck1 and KlLmo1 showed a diffused cytosolic distribution in cells grown under standard conditions, GFP-KlRho5 localized at the plasma membrane, as expected (Fig. 3A; note that KlDck1-GFP and KlLmo1-GFP occasionally seem to form patches at internal structures. Closer inspection showed that these signals neither colocalize with a nuclear marker nor with stained vacuoles, as shown in Fig. S4). However, upon addition of hydrogen peroxide, all three signals appeared at the mitochondria within the 2–3 min needed to prepare the samples for microscopy. Thus, all three K. lactis DLR components translocated like their S. cerevisiae homologs, indicating a conserved function in oxidative stress response (Schmitz et al., 2015).

Cdc42, another member of the Rho-GTPase family, which is involved in polarity establishment and morphogenesis in S. cerevisiae (Woods and Lew, 2019), has been shown to serve overlapping functions with Rho5 homologs in different filamentous fungi (reviewed in Hühn et al., 2019). Therefore, we assessed the intracellular localization of a GFP-KlCdc42 fusion. Under standard growth conditions in S. cerevisiae, this GTPase accumulates at the growing bud tip and is associated with the surface of vacuoles and other intracellular membranes (Woods and Lew, 2019). As shown in Fig. 3B, GFP-KlCdc42 also translocated efficiently to mitochondria when cells were exposed to hydrogen peroxide, indicating that it indeed could serve a complementary function to KlRho5. This notion was further supported by the suppression analyses of the morphological defects in a Klrho5 deletion as discussed in the next section. As a note of caution, it should be mentioned that the GFP-KlCdc42 fusion construct was not functional, since it did not rescue the lethality of a Klcdc42 deletion as deduced from tetrad analysis of a suitable heterozygous diploid strain (Klcdc42::kanMX/KlCDC42 GFP-KlCDC42::KlLEU2/Klleu2; data not shown).

In accordance with the observations in S. cerevisiae, the tagged DLR components of K. lactis also quantitatively associated with mitochondria upon glucose starvation (Fig. 4). Thus, despite the metabolic differences between the highly fermentative baker's and the respiratory-driven milk yeast, glucose starvation has a similar effect on the association of Rho5 with mitochondria in both.

While experimental evidence gathered so far indicates that lack of Rho5 in baker's yeast only moderately affects the number of actin patches per cell (Schmitz et al., 2018), we were intrigued to find a distinct morphological phenotype in cells with a Klrho5 deletion. As shown in Fig. 5A, most cells carrying the Klrho5 deletion, but not those of the wild type, had protruded bud scars (‘humps’) in addition to new middle- to large-sized buds. Similar protrusions were found in strains carrying Kldck1 and Kllmo1 deletions (Fig. 5A), indicating that lack of KlRho5 activity causes these morphological defects. To confirm that this phenotype was not caused by a second-site mutation in the genetic background of the strains employed, we first introduced the respective wild-type genes on K. lactis centromeric vectors. This only led to partial reversion of the phenotype, i.e. a lower percentage of cells with protrusions as compared to the deletion strain carrying an empty vector (data not shown). We assumed that the partial recovery was because of an insufficient plasmid stability in the respective transformants, previously deduced from the poor transmission of centromeric vectors through meiosis in tetrad analyses in several K. lactis strains (Rippert et al., 2014). Therefore, the wild-type KlRHO5, KlDCK1 and KlLMO1 genes were cloned into integrative vectors (pMMO5, pMMO7 and pMMO6, respectively) and inserted by homologous recombination at the Klleu2 locus into the corresponding deletion strains (KHO277-3D rho5; KHO254-1A dck1, KHO255-1B lmo1). Microscopic inspection revealed that 99% (n=209), 98% (n=103), and 94% (n=118) of the respective cells displayed wild-type morphologies. In contrast, more than 70% of cells from all recipient strains, or those carrying the empty vector integrated at the Klleu2 locus, had protrusions.

To assess possible overlapping functions with KlRho5 regarding the morphology, we constructed an activated allele of KlCDC42, encoding a G12V substitution in the GTPase on an integrative vector, and inserted it at the Klleu2 locus. The resulting strain was crossed with Klrho5 deletions and subjected to tetrad analyses. Out of 21 segregants with the double-mutant genotype Klrho5 KlCDC42^G12V examined under the microscope, 18 were found to grow like the wild type. This strongly indicated that KlCDC42^G12V can suppress the aberrant cell morphology associated with the Klrho5 deletion.

In an attempt to further substantiate these findings, we also constructed deletions in genes encoding the putative GEF KlCdc24 and the scaffold protein KlBem1 in a diploid K. lactis strain, since both proteins are related to regulation of morphogenesis mediated by Cdc42 in S. cerevisiae. As with the heterozygous Klcdc42 deletion, segregants from tetrad analyses carrying the respective deletion alleles did not produce colonies, indicating that both genes are essential (Fig. S5A–C). Similar results were also obtained with a heterozygous Kliqg1 deletion, whose homolog encodes a key regulator of cytokinesis in S. cerevisiae (Fig. S5D). Further analyses of gene functions in all these cases were impeded by the fact that we still lack a tightly regulated promoter for conditional gene expression in K. lactis (Rippert et al., 2014).

Two mechanisms may explain the formation of the protrusions in cells lacking active KlRho5: (1) abortive budding, i.e. cells trying to form a bud get stalled and initiate the formation of a new bud at another site, or (2) defective cytokinesis, i.e. the bud neck may be unusually elongated prior to cell separation and thus leave a ‘hump’. In order to distinguish between these possibilities, time-lapse images of growing cells from a Klrho5 deletion strain were produced and compared to wild-type cells (Fig. 5B; Movies 1 and 2), which showed that indeed the protrusions are remnants of previous budding events, rather than resulting from aborted ones. The same was true for Kldck1 and Kllmo1 deletions (Fig. 5B and time-lapse Movies 3 and 4).

To study these structural defects in more detail, we investigated the morphology of the aberrant protrusions by TEM. As summarized in Fig. 5C, Klrho5, Kldck1 and Kllmo1 deletions frequently produced a distorted primary septum, either bending into the developing bud or not being properly closed, while the wild-type controls displayed regular and fairly straight primary and secondary septa, as described previously (Rippert et al., 2014). In S. cerevisiae, secondary septa are deposited from both mother and daughter cells on to the primary septa. Assuming that K. lactis follows a similar strategy, a compensatory deposition of excessive glucan may explain the elongation of the bud necks in the mutants, which can ultimately be detected even by light microscopy (Schmidt et al., 2002). Moreover, the strengths of the lateral cell walls were determined from TEM images by measuring lateral cell wall thickness at ten points in each of 14 cells (Fig. 6). While the wild type yielded mean values of ∼85 nm, all three mutants lacking the KlDLR components increased in thickness to 98 nm, 105 nm and 112 nm for Kllmo1, Kldck1 and Klrho5 strains, respectively. This indicates that KlRho5 activity is not only required for proper synthesis of the primary septum during cytokinesis, but also participates in the general regulation of cell wall biosynthesis.

The observed protrusions of the bud scars strongly suggested that spatial and/or temporal control of cytokinesis is impaired in mutants in which KlRho5 is either absent or cannot be activated. Given the role of the human Rac1 homolog in the organization of the actin cytoskeleton, we therefore combined strains producing KlLifeAct-mRuby and KlMyo1-mCherry fusions, as markers for the contractile AMR, with the GFP-KlRho5 fusion and performed live-cell fluorescence microscopy. KlRho5 accumulated at the bud neck during the late stages of cytokinesis. Thus, an accumulation was observed at the bud neck of cells with large buds, colocalizing with KlLifeAct (Fig. 7A). On the other hand, the KlRho5 signal was distributed in patches throughout the plasma membrane of both mother and daughter in cells with small buds, thus colocalizing in this region with KlMyo1 upon its appearance (Fig. 7B; Fig. S6). It should be noted that at the neck of small-budded yeast cells KlRho5 was present, but only accumulated in cells with larger buds, i.e. during later stages of cytokinesis. Pearson coefficients were determined to substantiate these findings (see example images in Fig. 7 and Fig. S6). On average, values of 0.44±0.22 (mean±s.d.) were obtained in a total of 22 images examined for the colocalization of GFP-Rho5 with KlMyo1-mCherry, and 0.54±0.10 (n=8) for the colocalization of the GTPase with actin in large-budded cells. In contrast, actin patches in mother and daughter cells did not notably colocalize with GFP-KlRho5. These findings support a role of KlRho5 in the regulation of cytokinesis in K. lactis, which probably is exerted after constriction of the AMR. This function is either less pronounced or not shared by its baker's yeast homolog, ScRho5 (Schmitz et al., 2018).

In the present study, we provide evidence that Dck1, Lmo1 and Rho5 can form a functional unit in K. lactis, which corresponds to the S. cerevisiae DLR complex and to human Rac1 in its association with the dimeric GEF DOCK/ELMO. The identified coding sequences are most probably true homologs of the S. cerevisiae genes in K. lactis, since they are syntenic in their chromosomal context and alignments of the deduced amino acid sequences revealed a high conservation, especially when focused on known functional domains. This bioinformatic evidence indicating a functional conservation is strongly supported by our heterologous expression studies, which showed complementation of the rho5 deletion in S. cerevisiae by two variants of KlRHO5. With regard to the GTPase, the hypervariable region preceding the C-terminal end is of special interest. For ScRho5, a proper intracellular distribution and an adequate response to oxidative stress depends on the insertion of 98 amino acids, which is not present in the human homolog Rac1, upstream of the C-terminally located PBR sequence, where Rac1 cannot complement the yeast defects (Sterk et al., 2019). In contrast, the functional complementation by KlRho5 in S. cerevisiae, whose extension comprises only 44 amino acid residues, indicates that a few conserved residues upstream of the PBR may suffice to render the GTPase active and fully functional in yeasts. We speculate that the extension of the positively charged region retained in the K. lactis sequence, especially the KPQKS motif preceding the PBR, may contribute to the interaction with the negatively charged phospholipids and thereby to the membrane association of KlRho5, in analogy to other Ras- and Rho-type GTPases (Fig. S1; Lam and Hordijk, 2013; Williams, 2003). This notion is supported by the fact that human Rac1 can at least partially complement the Scrho5 deletion, if its C-terminus is replaced by the one of the extended ScRho5 (Sterk et al., 2019).

Another indication of functional conservation of the DLR complex in K. lactis is the rapid translocation of all three proteins to mitochondria after exposure to oxidative stress, and their similar physiological response when compared to S. cerevisiae (Schmitz et al., 2015). Thus, both Klrho5 and Scrho5 deletions, as well as those in the genes for the GEF subunits, provoked a hyper-resistance towards hydrogen peroxide. In general, K. lactis can withstand much higher concentrations of H₂O₂ than S. cerevisiae, probably because of its more respiratory lifestyle accompanied by higher production of ROS (Lamas-Maceiras et al., 2015; Santomartino et al., 2019). In S. cerevisiae, the increased resistance of mutants lacking DLR complex components was correlated to an impairment of mitophagy and a reduced apoptosis (Schmitz et al., 2015). In this context, it is important that we did not find an effect of a Klrho5 deletion on the chronological lifespan, in contrast to a report on Scrho5 (Cao et al., 2016). However, we were not able to reproduce the reduction in CLS for a Scrho5 deletion in the genetic background of the S. cerevisiae strain routinely used in our laboratory (unpublished results), indicating that it may be either strain dependent or very sensitive to minor changes in experimental conditions.

There are also pronounced similarities in the response of Rho5 to glucose starvation in the two yeast species. Thus, the DLR complex quantitatively translocates to mitochondria under such conditions both in S. cerevisiae (Schmitz et al., 2018), and in K. lactis, as shown herein. However, none of the synthetic phenotypes observed for Scrho5 in conjunction with deletions in genes involved in nutrient sensing, such as sch9 or gpr1, appears in K. lactis. This can be attributed to the fact that K. lactis is a Crabtree-negative yeast that does not switch to fermentative metabolism in the presence of high glucose concentrations (reviewed in Rodicio and Heinisch, 2013). Consequently, key regulators of central carbohydrate metabolism, such as the SNF1 complex and its downstream transcription factors KlCat8 and KlSip4, have been embedded into different regulatory networks as compared to S. cerevisiae (Mehlgarten et al., 2015; Mojardin et al., 2018). One of the physiological results of this rewiring is the repression of ethanol utilization by glycerol as a non-fermentable carbon source, which is not observed in S. cerevisiae (Rodicio et al., 2008). Thus, either KlRho5 is not involved with the regulatory circuits governed by KlSch9 and KlGpr1 in contrast to their S. cerevisiae counterparts, or, if the genetic interactions are retained, they produce less severe phenotypes not reflected in the growth characteristics tested herein. Addressing these questions would require detailed metabolic studies of the Klsch9 and Klgpr1 deletion mutants, which are beyond the scope of this work, but offer interesting perspectives for future investigations.

Most importantly, we found that cells lacking KlRho5 activity display morphological defects not observed in S. cerevisiae (Schmitz et al., 2018). The data presented herein indicate that KlRho5 is involved in the regulation of cytokinesis and accumulates at the bud neck after constriction of the AMR. We previously showed that cytokinesis in K. lactis basically relies on the same structural and regulatory components as that in S. cerevisiae, i.e. a constricting AMR controlled by Hof1 and Cyk3, as well as the septins that form the bud-neck compartment (Rippert and Heinisch, 2016; Rippert et al., 2014). The assembly of septins and AMR constriction are also central events in mammalian cytokinesis and as such essential for human health (Mangione and Gould, 2019). Interestingly, Rac1 in its association with DOCK/ELMO regulates the dynamics of the actin cytoskeleton in mammalian cells (reviewed in Singh et al., 2019b). While this function seems to be poorly conserved in S. cerevisiae (Schmitz et al., 2018), our data indicate that KlRho5 may be required for proper cytokinesis in K. lactis. As a consequence, closure of the primary septum and deposition of the secondary septa are deregulated in the deletion mutants, and aberrant cytokinesis leads to protruding bud scars retained on the mother cell surface. This is somewhat reminiscent of the filamentous fungus Eremothecium gossypii (formerly known as Ashbya gossypii), where rho5 mutants were also shown to produce aberrant septa (Nordmann et al., 2014). It is noteworthy that the protruded bud phenotype observed in the Klrho5 deletion is different from the multiple buds produced by a Klmyo1 deletion, which also interferes with proper cytokinesis (Rippert et al., 2014). This indicates that it is the actin moiety of the AMR that is affected by the lack of KlRho5. This view is also substantiated by the observation that accumulation of KlRho5 at the bud neck occurred after AMR constriction, when KlMyo1 was no longer detectable.

In S. cerevisiae, the homolog of mammalian IQGAP, Iqg1, is recruited together with actin to the bud neck after deposition of myosin II (Lippincott and Li, 1998). Interestingly, both activated Rac1 and Cdc42 can bind to IQGAPs in mammalian cells (Nouri et al., 2020; Ozdemir et al., 2018). This offers the intriguing possibility that KlRho5 and KlCdc42 may control the recruitment of actin to the bud neck in K. lactis during cytokinesis by interacting with KlIqg1. The GTPases could then be activated to participate in the control of cell wall synthesis, especially the primary septum. It will be interesting to follow up these associations in future research.

Consistent with a conserved role in the negative regulation of cell wall synthesis through CWI signaling (Schmitz et al., 2002a), Klrho5 deletions also displayed an increased thickness of the lateral cell walls. In conclusion, while the influence of Rho5 on the actin cytoskeleton could have degenerated in the fermentative baker's yeast as compared to Rac1, it seems to be retained from mammalian cells to fungi in general, and especially in the respiratory-oriented yeast K. lactis. This is a further example of why K. lactis in many cases may be a better eukaryotic model system than S. cerevisiae, as previously proposed for the study of tauopathies related to Alzheimer's disease (Heinisch and Brandt, 2016).

Yeast strains employed and their genotypes are listed in Table S1. All strains were derived from the congenic series described previously (Heinisch et al., 2010), which is based on the sequenced K. lactis type strain CBS2359. Yeast cell culture and genetic techniques followed standard procedures (Rose et al., 1990). Rich medium (YEPD) contained 1% yeast extract, 2% Bacto Peptone (Difco Laboratories Inc., Detroit, MI, USA) and 2% glucose. Synthetic media were prepared as described in Rose et al. (1990), with the omission of amino acids or bases as required for selection of plasmids or deletion markers and 2% glucose (SCD). For selection of the kanMX marker, 100 mg/l of G418 was added to the medium after sterilization.

For serial drop-dilution assays, cells were grown overnight at 30°C in SCD to late logarithmic phase, diluted in fresh medium to an OD₆₀₀ of 0.1, grown for another 3 h and subjected to 10-fold serial dilutions. A 3 µl aliquot of each dilution was spotted on to the indicated media and plates were incubated for 3 days at 30°C. Images were scanned and adjusted for brightness and contrast using Corel Photo Paint with the same settings for the entire plate, prior to compilation of lanes into the final figures. Growth curves were obtained in 100 µl cultures in 96-well plates, in YEPD with or without hydrogen peroxide as indicated, and recorded in a Varioskan LUX plate reader (Thermo Scientific, Bremen, Germany) as described previously (Sterk et al., 2019).

K. lactis strains were crossed on malt agar plates (5% malt, 3% agar), incubated for 2 days at 30°C and replica plated on to appropriate synthetic media for selection of diploids. For tetrad analyses, diploid strains were grown to late logarithmic phase in liquid YEPD, collected by centrifugation (10,000 g for 1 min at room temperature) and dropped on to 1% potassium acetate plates for sporulation at 30°C. After 1–2 days and microscopic inspection for ascus formation, a sample of each culture was resuspended in 100 µl of sterile water, and 4 µl of Zymolyase 100T (10 mg/ml) was added, followed by incubation for 7–10 min at room temperature. A 15 µl volume of the suspension was streaked out on to a YEPD plate and spores were segregated using a Singer MSM400 micromanipulator (Singer Instruments, Roadwater, Somerset, UK). Plates were incubated for 3 days at 30°C, and colony formation was documented by scanning and image processing as described for the serial drop-dilution assays.

For manipulations in E. coli, strain DH5α was employed with standard media as described previously (Straede and Heinisch, 2007).

Wild-type genes from K. lactis were obtained by PCR using appropriate oligonucleotides with restriction sites and genomic DNA of strain CBS2359 or its derivatives as templates. Clones were verified by sequencing to contain no errors leading to amino acid exchanges in the protein coding regions. All vectors and plasmids employed are listed in Table S2. In the case of KlDCK1, PCR cloning was combined with gap repair analogous to the procedure described for S. cerevisiae (Oldenburg et al., 1997), since because of the large open reading frame of KlDCK1 (5823 bp), all primary clones constructed contained PCR errors in the protein coding sequence. Therefore, the wild-type sequence was regenerated by gap repair. Briefly, one of the plasmids with an erroneous PCR insertion was used to excise an internal fragment, dephosphorylated and introduced into a K. lactis recipient strain with a wild-type genomic KlDCK1 allele. Sequencing of a recombinant clone isolated from such a transformant (pJJH1409) confirmed that it carried the correct KlDCK1 coding region.

Deletion strains were obtained by one-step gene replacement techniques described for S. cerevisiae (Rothstein, 1991), either by employing PCR products with 45–50 bp of flanking sequences homologous to the target gene, or by first creating longer flanking regions by in vivo recombination of the PCR products with the respective plasmids in S. cerevisiae, subcloning in pUK1921 (Heinisch, 1993) or its derivatives, and transformation of K. lactis after excision with restriction endonucleases or PCR amplification, as described previously (Heinisch et al., 2010). The recipient strain was either diploid and homozygous for the ku80 deletion or haploid and also carrying this deletion to facilitate homologous recombination (Zonneveld and Steensma, 2003). Target gene deletions were obtained with either the kanMX or the SkHIS3 cassette from the Longtine collection (Longtine et al., 1998), SpHIS5 from pUG27 (Gueldener et al., 2002), or ScLEU2 cassettes described previously (Heinisch et al., 2010). In the majority of cases, the ku80 deletion was later eliminated by crossing with suitable KU80 wild-type strains and tetrad analysis to avoid genomic instability. Fluorescence markers were derived from 3x mCherry and 3x myEGFP vectors described in Maeder et al. (2007), but carrying single fluorescence tags in the Longtine vector backbones (Table S2).

The mutant allele KlRHO5^Q69H, which corresponds to the activated allele of ScRHO5^Q91H described by Schmitz et al. (2002a), was obtained by string DNA synthesis from GeneArt (ThermoFisher Scientific) as a fragment containing the mutation and substituted for the respective wild-type sequence in a KlRHO5 clone via homologous recombination in yeast. Similarly, a DNA fragment carrying the activated KlCDC42^G12V allele was obtained by string DNA synthesis and cloned into a vector carrying the wild-type allele.

A LifeAct probe with GFP as a fluorescent marker to follow actin dynamics in vivo was previously constructed for use in K. lactis, given that neither rhodamine phalloidin nor the commonly used LifeAct probe for S. cerevisiae worked in this yeast, probably because of a slightly different actin composition (Rippert et al., 2014). For colocalizations with GFP-KlRho5, the fluorophore was replaced by mRuby using homologous recombination and combined in an integrative plasmid with the tagged RHO5 allele. KlMyo1-mCherry constructs have also been described previously (Rippert et al., 2014).

The setup used for fluorescence microscopy consisted of a Zeiss Axioplan 2 microscope (Carl Zeiss, Jena, Germany) equipped with a 100× alpha-Plan Fluor objective (numerical aperture 1.45) and differential interference contrast. Sample handling and image processing have been described in detail previously (Schmitz et al., 2015). Vacuoles were stained with 7-amino-4-chloromethylcoumarin (CMAC) as follows. A 1 ml sample of yeast cells grown to an OD₆₀₀ of 1.0 in synthetic medium was collected by centrifugation (10,000 g for 1 min at room temperature) and resuspended in 100 µl of fresh medium. An 8 µl volume of a 10 mM CMAC stock solution was added, and cells were incubated for 1 h, washed with 1 ml of fresh medium and used for microscopic inspection. Colocalization was quantified by determining the Pearson coefficient, using the colocalize function of the FIJI/ImageJ program.

Yeast cells were grown to logarithmic phase in YEPD with shaking at 30°C and subjected to TEM as described in detail previously (Backhaus et al., 2010). In brief, cells were fixed with glutaraldehyde and potassium permanganate. After dehydration with ethanol and acetone, cells were suspended in Epon 812 for polymerization. To obtain 70 nm thick sections, a Leica Ultracut UC7 ultramicrotome was employed, and, after staining with uranyl acetate and lead citrate, samples were examined using a Zeiss TEM902A electron microscope at 80 kV.

We are grateful to Hans-Peter Schmitz for invaluable advice on fluorescence microscopy and statistical analyses. We also thank Andrea Murra for excellent technical assistance and Rosaura Rodicio for critical reading of the manuscript.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258301.