In Vivo Yeast Cell Morphogenesis Is Regulated by a p21-Activated Kinase in the Human Pathogen Penicillium marneffei

Pathogens have developed diverse strategies to infect their hosts and evade the host defense systems. Many pathogens reside within host phagocytic cells, thus evading much of the host immune system. For dimorphic fungal pathogens which grow in a multicellular hyphal form, a central attribute which facilitates growth inside host cells without rapid killing is the capacity to switch from the hyphal growth form to a unicellular yeast form. Blocking this transition abolishes or severely reduces pathogenicity. Host body temperature (37°C) is the most common inducer of the hyphal to yeast transition in vitro for many dimorphic fungi, and it is often assumed that this is the inducer in vivo. This work describes the identification and analysis of a new pathway involved in sensing the environment inside a host cell by a dimorphic fungal pathogen, Penicillium marneffei. The pakB gene, encoding a p21-activated kinase, defines this pathway and operates independently of known effectors in P. marneffei. Expression of pakB is upregulated in P. marneffei yeast cells isolated from macrophages but absent from in vitro cultured yeast cells produced at 37°C. Deletion of pakB leads to a failure to produce yeast cells inside macrophages but no effect in vitro at 37°C. Loss of pakB also leads to the inappropriate production of yeast cells at 25°C in vitro, and the mechanism underlying this requires the activity of the central regulator of asexual development. The data shows that this new pathway is central to eliciting the appropriate morphogenetic response by the pathogen to the host environment independently of the common temperature signal, thus clearly separating the temperature- and intracellular-dependent signaling systems.


Introduction
Host immune systems actively survey and attempt to kill invading pathogens, so for pathogens to successfully infect a host the pathogen must be able to evade or tolerate these systems. A number of pathogens enter phagocytic cells and primarily reside within these cells to avoid the host's immune defense system. To continually reside within phagocytic cells of the immune system without disrupting their integrity, pathogens such as fungi which can grow in a filamentous, multicellular hyphal form, must be able to produce a uninucleate yeast growth form. The ability to switch between the filamentous and yeast forms is a tightly regulated process known as dimorphic switching. Dimorphism has been shown to be a critical pathogenicity determinant.
Penicillium marneffei exhibits dimorphic switching and hence can grow in two distinct cellular forms; multicellular hyphae and unicellular yeast. P. marneffei is the only known Penicillium species which is dimorphic and the switch between growth forms is regulated by temperature [1]. At 25uC, in the saprophytic growth phase, P. marneffei grows as multinucleate, septate, branched hyphae. These hyphae produce conidia, the infectious agent, from specialized multicellular structures termed conidiophores. When switched to 37uC, P. marneffei undergoes a developmental process termed arthroconidiation. Cellular and nuclear division become coupled, double septa are laid down and hyphae fragment at these septation sites to liberate uninucleate yeast cells which subsequently divide by fission [1]. The yeast cells are the pathogenic form and multiple yeast cells are seen in the pulmonary alveolar macrophages and peripheral blood mononuclear cells of infected individuals [2]. P. marneffei infection is likely to occur through inhalation of the conidia produced by the filamentous saprophytic form [2]. It has been proposed that the conidia bind to laminin in the bronchoalveolar epithelia via a sialic acid-specific lectin [3,4]. The conidia are then ingested by host pulmonary alveolar macrophages where they germinate into unicellular yeast cells which divide by fission. Therefore the ability to produce infectious propagules such as asexual spores (conidiation) in the saprophytic growth state and the capacity upon infection to switch between a multicellular hyphal growth form and a unicellular yeast pathogenic form are both crucial for pathogenicity.
Polarity establishment is necessary for the differentiation of distinct cell types during development. The Rho GTPases Cdc42 and Rac act as molecular switches to localize or activate proteins associated with polarized growth. The CDC42 homologue in P. marneffei, cflA, is required for germination of conidia at both 25uC and 37uC, polarized growth and division of hyphae at 25uC and for polarized growth of yeast cells at 37uC [5]. The P. marneffei genome also encodes a second Rac-like Rho GTPase, cflB. Similar to cflA, cflB is required for the polarized growth and division of hyphae at 25uC [6]. However, unlike cflA, cflB plays a key role during asexual development (conidiation) at 25uC and is not required for the polarized growth of yeast cells at 37uC [6]. In Saccharomyces cerevisiae, the Rho GTPase Cdc42p activates the p21 activated kinases (PAKs) Ste20p and Cla4p [7][8][9][10][11]. P. marneffei possesses both STE20 and CLA4 homologues; pakA (STE20) and pakB (CLA4). Characterization of pakA in P. marneffei has shown that this gene is essential for conidial germination at 37uC and polarized growth of yeast cells, acting downstream of both a heterotrimeric G protein and Cdc42 pathway [12]. DpakA and pakA strains containing a mutation in the conserved Cdc42/Rac Interactive Binding (CRIB) domain (pakA H108G ) fail to germinate into pathogenic yeast cells in vivo [12].
This study describes the characterization of the second PAK in P. marneffei, PakB. The pakB gene is expressed during hyphal growth and asexual development at 25uC and is essential for the generation of these 25uC-specific cell types. Deletion of pakB results in yeast-like morphology and the inappropriate production of yeast cells at 25uC. Deletion of the primary regulator of asexual development, brlA, in the DpakB strain results in suppression of this inappropriate yeast cell production suggesting that these yeast cells are dependent on the conidiation program. PakB is also essential for yeast morphogenesis during infection but not in vitro. Macrophages infected with the DpakB strain exhibit highly branched, septate, hyphal cell growth but no yeast cells. These results suggest that the developmental pathways regulating conidiation at 25uC and yeast cell production at 37uC share a number of regulatory components including PakB and that the developmental outcome of each pathway is regulated in part by the mode of cellular division.

Results
Cloning the CLA4 p21-activated kinase orthologue from P. marneffei A previous low stringency hybridisation screen of a P. marneffei genomic library using an Aspergillus nidulans sequence with strong homology to S. cerevisiae Ste20p yielded five positive clones, which fell into two classes based on restriction enzyme digestion patterns [12]. Sequencing of a cloned fragment from one of these classes (pKB5751) revealed strong sequence homology to STE20-like PAKs and the gene within this clone was subsequently named pakA [12]. A fragment from the second class of clones was also subcloned (pKB4904) and sequencing revealed strong sequence homology to CLA4-like PAKs from Candida albicans (48% identity, 58% similarity), Ashbya gossypii (49% identity, 57% similarity), Ustilago maydis (53% identity, 61% similarity) and S. cerevisiae (48% identity, 57% similarity) ( Figure S1). The gene within this clone was named pakB. The homology is to a large extent restricted to the CRIB and kinase domains where, for example, PakB shows 74% and 76% identity, respectively, to the same domains in U. maydis Cla4. The predicted PakB protein exhibits 37% identity and 46% similarity to PakA.
The predicted PakB protein is 733 amino acids in length and contains a PH domain at 81-191, a Cdc42/Rac Interactive Binding (CRIB) domain (also called PBD for p21-Rho-binding domain) at positions 195-256 and a predicted kinase domain at 443-712 (http://pfam.sanger.ac.uk/). An 11 amino acid sequence in the non-catalytic C-terminal region of S. cerevisiae Ste20p has been shown to be required for interaction with Ste4p, the beta subunit of a heterotrimeric G protein, during pheromone signaling [13,14]. This region in S. cerevisiae Cla4p has also been shown to interact weakly with Ste4p [13]. The consensus sequence SSLwPLI/VXwwb (where X is any residue, w for A, I, L, S or T and b is for basic residues) is also found in P. marneffei PakA (619-629).
RNA was isolated from vegetative hyphae grown for 2 days in liquid medium at 25uC, asexual development (conidiation) cultures grown for 4 days on solid medium at 25uC and yeast cells grown for 6 days in liquid medium at 37uC. The level of pakB expression varied substantially depending on cell type. The level of pakB transcript was highest during vegetative hyphal growth at 25uC, lower during asexual development at 25uC and barely detectable during vegetative yeast growth at 37uC relative to the benA control ( Figure 1A).
P. marneffei infection is believed to occur by inhalation of conidia, which bind to the laminin in the bronchoalveolar epithelium [2][3][4]. Conidia are then ingested by pulmonary alveolar macrophages and germinate directly into uninucleate yeast cells which proliferate within the macrophage [1]. To examine if pakB is expressed during infection, RNA was isolated from yeast cells derived either from LPS activated J774 murine macrophages at 37uC infected with wildtype conidia 24 hours post-infection or from yeast cells incubated in macrophage growth media at 37uC for 24 hours (Materials and Methods). Substantial levels of pakB expression were detected in cells isolated from infected macrophages suggesting that pakB expression is induced during infection ( Figure 1B). Low levels of pakB expression was detected in yeast cells derived from the macrophage medium control ( Figure 1B).

PakB is localized to the hyphal apex, conidiophores and septation sites
To investigate the localization of PakB, a triple HA tag was inserted into a non-conserved region of PakB between the CRIB and kinase domains. The pakB + HA construct was co-transformed with the barA + gene into the P. marneffei strain G487 (niaD pyrG areA 2 ). Transformants were selected for glufosinate resistance and confirmed by Southern blot analysis of genomic DNA (Materials and Methods). Anti-HA immunostaining was performed on two of the pakB + HA strains after 4 days growth at 25uC. PakB was observed concentrated at the hyphal apex (Figure 2A-B) and also localized to all of the cell types of the conidiophore ( Figure 2C). PakB was particularly concentrated at the phialide to conidium interface and around the periphery of newly formed, but not old, conidia ( Figure 2D). In addition, PakB was co-localized at nascent septation sites presenting either as a single band colocalised with calcofluor stained septa ( Figure 2E), two bands on either side of the

Author Summary
Dimorphic fungal pathogens pose significant health and agricultural problems worldwide. These fungi have the capacity to switch between a multicellular hyphal growth form and a unicellular yeast growth form. Often one form is pathogenic, found in infected hosts, and the other is not. Many dimorphic fungal pathogens of humans produce the yeast form during infection and this form resides within host phagocytic immune cells, where it can tolerate killing by these cells and is not exposed to the acquired immune system. Inhibiting the pathogen's ability to switch growth forms has been shown to block pathogenicity. This study identifies a pathway used by the fungal pathogen to sense the host and switch to the appropriate growth form. This study provides new insights into the molecular mechanisms which are important for pathogenicity and may identify factors which can be targeted to block the ability of the pathogen to successfully reside within host cells.
calcofluor stained septa ( Figure 2F) or two spots on either side of the calcofluor stained septa ( Figure 2G). PakB was not observed at older septa (data not shown).
To investigate the localization of pakB during infection, LPS activated J774 murine macrophages were infected with the pakB + HA strains and anti-HA immunostaining and calcofluor staining were performed 24 hours post-infection (Materials and Methods). PakB was localized around the cell periphery ( Figure 2H-I). PakB was not localized either at nascent septation sites prior to, or immediately after, cell wall deposition (indicated by calcofluor staining) ( Figure 2H). Weak localization at septation sites could be observed prior to cell separation ( Figure 2I). PakB was localized at, and adjacent to, the division site during cell separation ( Figure 2I).

Deletion of pakB results in inappropriate yeast-like growth at 25uC
A construct in which a region encompassing the pakB coding sequence (+323 to +2618) was replaced with the pyrG selectable marker was used to create a pakB deletion strain. P. marneffei strain G487 (niaD pyrG areA 2 ) was transformed with this construct and pyrG + transformants selected. Genomic DNA from the PyrG + transformants was screened by Southern blotting to identify strains which possessed a restriction pattern consistent with replacement of pakB by pyrG at the genomic locus (data not shown). To generate a DpakB pyrG 2 strain, a DpakB::pyrG + deletion strain was plated on medium containing 5fluoroorotic acid (5-FOA) (Materials and Methods). This strain was cotransformed with plasmids containing pakB + and pyrG + genes and co-transformants confirmed by Southern blot analysis. The DpakB pakB + transformants contained 2-8 copies of pakB.
After 10 days at 25uC, wildtype P. marneffei grows as polarized vegetative hyphae which bear asexual structures (conidiophores). Colonies appear fuzzy and the surface is green due to the presence of pigmented asexual spores (conidia) on conidiophores ( Figure 3A-B). The DpakB strain produced compact, mucoid yeast-like colonies after 10 days growth at 25uC which resemble the yeast colonies produced by the wildtype at 37uC ( Figure 3A versus C). Despite this the DpakB strain conidiated upon longer incubation at 25uC (14 days). Conidiophore structures were visible under higher magnification, however, these were unevenly dispersed over the yeast-like colonies and not as profuse as in wildtype ( Figure 3B). Transformation of the DpakB strain with pakB + (DpakB pakB + ) completely restored the wildtype phenotype ( Figure 3).
Deletion of pakB results in inappropriate expression of hyphal specific and yeast specific genes at 25uC Deletion of pakB resulted in yeast-like growth at 25uC ( Figure 3A). To investigate the molecular basis underpinning this phenotype RT PCR analysis was performed to see if it correlates with a decrease in the expression of hyphal specific genes or an increase in the expression of yeast specific genes. RNA was isolated from both the wildtype and the DpakB strain grown as vegetative hyphae for 2 days in liquid medium at 25uC and as yeast cells for 6 days in liquid medium at 37uC and used for RT PCR with primers for a number of cell type specific genes (Canovas and Andrianopoulos, unpublished). In wildtype, the 2E11 probe was expressed specifically in hyphae at 25uC and not in yeast cells at 37uC ( Figure 1C). In contrast, the 2E11 transcript was not detectable in the DpakB strain at 25uC ( Figure 1C). In wildtype, the 2E4 transcript was expressed at a low level at 25uC and expression was greatly increased at 37uC ( Figure 1D) while expression of 5B10 in wildtype was not detectable at 25uC but was high at 37uC ( Figure 1E). Both 2E4 and 5B10 transcripts were highly expressed in the DpakB strain at 25uC ( Figure 1D and E). The amount of 2E4 and 5B10 transcript was also slightly increased in the DpakB strain at 37uC ( Figure 1D and E). This suggests that in the DpakB strain both a decrease in the expression of hyphal specific genes and an Figure 1. pakB is expressed predominately at 25uC and deletion results in inappropriate expression of hyphal specific and yeast specific genes at 25uC. (A) Expression of pakB shown by RT PCR on RNA isolated from vegetative hyphae grown for 2 days in liquid medium at 25uC (25uC veg.), asexual development cultures grown for 4 days on solid medium at 25uC (25uC dev.) and yeast cells grown for 6 days in liquid medium at 37uC (37uC). A benA loading control is shown. pakB is expressed at the highest level in vegetative hyphae at 25uC and to a lesser extent during asexual development at 25uC. Expression of pakB is barely detectable at 37uC. (B) RT PCR performed on RNA isolated from wildtype P. marneffei grown in the absence (2) or presence (+) of LPS activated murine macrophages after 24 hrs. In the infection process, macrophages were incubated for 2 hours with wildtype conidia to allow for phagocytosis before washing to remove non-phagocytosed conidia. Therefore, all P. marneffei cells are intracellular after 24 hrs. The benA loading control is shown. Expression of pakB is increased during macrophage infection. (C-E) RT PCR on RNA isolated from both the wildtype (+) and DpakB (D) strains grown as vegetative hyphae grown for 2 days in liquid medium at 25uC (25uC) and as yeast cells grown for 6 days in liquid medium at 37uC (37uC). The benA loading controls are shown. (C) In wildtype, the hyphal specific probe 2E11 is expressed specifically in hyphae at 25uC and not in yeast cells at 37uC. The 2E11 transcript is not detectable in the DpakB strain at 25uC. (D) In wildtype, 2E4 is expressed at a low level at 25uC and expression is greatly increased at 37uC. Expression of 2E4 is increased in the DpakB strain at 25uC. The amount of 2E4 expression is also slightly increased in the DpakB strain at 37uC. (E) The transcript of the yeast specific probe 5B10 is not detected in wildtype at 25uC, whereas, expression is high at 37uC. 5B10 is highly expressed in the DpakB strain at 25uC. The amount of 5B10 expression is also slightly increased in the DpakB strain at 37uC. doi:10.1371/journal.ppat.1000678.g001 increase in the expression of yeast specific genes may contribute to the yeast-like growth phenotype at 25uC.

The DpakB strain inappropriately produces yeast cells at 25uC
In order to characterize the cellular basis behind the yeast-like colonial morphology of the DpakB strain at 25uC, scanning electron microscopy (SEM) was performed on both wildtype and DpakB strains after 10 days growth at 25uC (Materials and Methods). Wildtype colonies appeared as a mass of entangled, branched hyphae which radiated from the centre of the colony in a polarized fashion ( Figure 4A). The edges of the colony were not clearly defined due to countless hyphae which extended great distances from the colony periphery ( Figure 4A). In contrast, the colonies of the DpakB strain were compact and therefore the colony edges were distinct ( Figure 4A). The hyphae of the DpakB strain were more branched than wildtype and often invaded the agar surface ( Figure 4B). Invasive growth is a common characteristic of wildtype arthroconidial growth at 37uC (not shown). In contrast to wildtype, in which yeast cells are never seen at 25uC, individual yeast cells were also observed around the DpakB colony periphery ( Figure 4C). The wildtype and DpakB strains were also grown on the standard 37uC media, brain heart infusion (BHI), for 10 days at 25uC and for 5 days at 37uC. SEM of the colony surface showed wildtype colonies at 25uC were comprised of large bundles of branched hyphae ( Figure 4D). At 37uC, wildtype undergoes arthroconidiation, a developmental process where hyphae fragment to liberate yeast cells which subsequently divide by fission. After 5 days at 37uC, the wildtype colony was comprised of a mass of fragmented hyphae and yeast cells ( Figure 4D). At 25uC, the colony surface of the DpakB strain resembled that of wildtype at 37uC ( Figure 4E). The DpakB strain also appeared to produce more yeast cells than wildtype at 37uC ( Figure 4E). Wildtype colonies appear fuzzy due to the presence of radial, polarized hyphae growing along the agar surface and aerial hyphae. The hyphae are covered with asexual structures (conidiophores). The DpakB strain produces compact, mucoid, yeast-like colonies at 25uC. Colonies of the DpakB strain are not covered with conidiophores after 10 days growth. The DpakB pakB + strain is indistinguishable from wildtype. (B) After 14 days, the colony surface of wildtype (pakB + ) shows many conidiophore structures (206 magnification). In contrast to wildtype, the DpakB strain produces fewer conidiophores and these are unevenly dispersed over the yeastlike colonial surface. The DpakB pakB + strain is indistinguishable from wildtype. (C) Colonies of the pakB strains at 37uC (326 magnification). Wildtype yeast colonies are compact and mucoid. The DpakB and DpakB pakB + strains are indistinguishable from wildtype at 37uC. doi:10.1371/journal.ppat.1000678.g003 Deletion of pakB affects nuclear index and branching but not septation or actin distribution To investigate further the molecular mechanisms underlying the deregulation of yeast cell morphogenesis in the DpakB strain at 25uC, the wildtype, DpakB and DpakB pakB + strains were grown for 4 days at 25uC, stained with calcofluor and Hoechst 33258 and observed microscopically ( Figure 5). Wildtype P. marneffei grows as highly polarized, branched, septate hyphae ( Figure 5). Apical cells are multinucleate whereas subapical cells are predominately uninucleate unless dividing ( Figure 5B). In all conditions examined, the DpakB pakB + strains were indistinguishable from wildtype (data not shown). The DpakB strain exhibited a compact colony morphology where hyphae were tightly packed and highly branched ( Figure 5A). Apical cell branching, which was not observed in wildtype, was frequently observed in the DpakB strain ( Figure 5A). Septa were present in the DpakB strain ( Figure 5A). In contrast to wildtype, the DpakB strain exhibited an increase in the number of nuclei per subapical cell compartment ( Figure 5B). To quantify this increase, the number of nuclei per cellular compartment was recorded for 100 cells on three separate occasions. Only 20.562.04% and 20.462.72% of wildtype and DpakB pakB + subapical cells contained more than one nucleus, respectively. In contrast, 64.261.18% of subapical cells of the DpakB strain contained more than one nucleus.
To examine if deletion of pakB affects actin distribution, immunostaining using mouse anti-actin was performed on the wildtype and DpakB strains. In the wildtype at 25uC, actin is localized as cortical actin spots along the hyphae and concentrated at nascent septation sites and the hyphal apex ( Figure 5C and D). Actin was normally distributed in the DpakB strain ( Figure 5C and E).
pakB is required for conidiation at 25uC and affects conidial germination kinetics Wildtype P. marneffei begins asexual development after 4 days growth at 25uC, with the production of a specialized aerial stalk from which differentiated cells are produced sequentially in a budding fashion: metulae bud from the stalk, phialides bud from metulae and uninucleate conidia bud from phialides. To investigate if the deletion of pakB results in aberrant asexual development at 25uC, wildtype, DpakB, DpakB pakB + and DcflB strains were grown for 14 days at 25uC and were examined by SEM ( Figure 6). Deletion of the P. marneffei RAC homologue, cflB, results in conidiation defects at 25uC [6]. Numerous conidiophores were observed on the surface of wildtype colonies ( Figure 6). The DpakB pakB + strain was indistinguishable from the wildtype (data not shown). The DpakB strain also produced conidiophores in which all conidiophore cell types were observed, however a large number of conidiophores in the DpakB strain had abnormally large conidia ( Figure 6). In addition, more than one conidium per phialide was rarely observed and the site of conidium to phialide attachment was not as constricted as in wildtype ( Figure 6). In contrast to the wildtype, DpakB and DpakB pakB + strains, conidiophores in the DcflB strain could not be readily distinguished but presumptive conidia of varied size were noted, similar to the DpakB strain although not as extreme ( Figure 6). A unique phenotype of the DcflB strain is that lysed conidiophore structures in which the metulae and phialides have ruptured or deflated were frequently observed ( Figure 6).
To investigate the cellular defects underlying the aberrant conidiophores of the DpakB strain, calcofluor staining was performed on wildtype, DpakB and DpakB pakB + strains after 14 days growth at 25uC (Materials and Methods). The DpakB conidiophores displayed septation defects in conidiophores ( Figure 7A-B). In contrast to wildtype conidiophores, in which two separate chitin disks can be observed at the phialide/conidium and conidium/conidium junctions, only one septum, no septa or incomplete septa were observed at the phialide to conidia cell boundaries in DpakB conidiophores ( Figure 7B).
To assess any defects in conidia that may be the result of the aberrant conidiophore morphogenesis in the mutant strains, calcofluor staining was performed on conidial suspensions of the wildtype, DpakB, DpakB pakB + and DcflB strains (Materials and Methods). The conidia produced by the wildtype and DpakB pakB + strains were homogeneous in size and showed uniform calcofluor staining around the cell periphery ( Figure 7C and data not shown). A proportion of conidia produced by the DcflB strain displayed a size increase, however, calcofluor staining remained uniform ( Figure 7C). In contrast, conidia from the DpakB strain differed greatly in size and showed uneven calcofluor staining ( Figure 7C). In addition, DpakB conidial preparations contained numerous yeast cells ( Figure 7C). Nuclear staining revealed that these yeast cells were uninucleate (data not shown). The germination and colony forming ability of conidia from the wildtype, DpakB, and DcflB strains was determined to investigate the potential consequences of the aberrant morphogenesis. The germination kinetics were measured by counting the number of ungerminated versus germinated conidia (conidia with a visible germ tube) in a population of 100 in three independent experiments after 15 hours in liquid medium at both 25uC and 37uC (Table 1). Despite the large increase in conidial size, conidia of the DpakB strain germinated well at both 25uC and 37uC and actually showed a slight increase in germination compared to the wildtype control (Table 1). In contrast, the conidia of the DcflB strain showed a reduction in germination compared to the other strains (Table 1). It was also evident that DpakB conidia prematurely extended secondary germ tubes. To quantify this, the number of conidia with 1, 2 or 3 or more germ tubes was counted in a population of 100 in three independent experiments ( Table 2). After 15 hrs at either 25uC or 37uC, the DpakB conidia showed a significant increase in germ tube emergence ( Table 2). The ability of single conidia to form colonies was also assessed after 5 days at both 25uC and 37uC ( Table 3). The colony forming units were measured by counting the number of colonies arising from 100 plated conidia in three independent experiments (Table 3). Consistent with the germination data, the DpakB strain showed wildtype viability ( Table 3). The DcflB strain showed a reduction in the ability to form colonies suggesting that the previously observed ungerminated conidia were inviable rather than delayed (Table 3).

Yeast cells produced by the DpakB strain at 25uC are derived from conidiophores
When examining asexual development in the DpakB strain, a number of conidiophores were observed in which the individual cell types had become detached (data not shown). As the DpakB strain inappropriately produced yeast cells at 25uC ( Figure 4C and 7C) we hypothesized that these yeast cells may arise from an inappropriate switch from conidiation to arthroconidiation. To investigate this possibility, a DpakB DbrlA double mutant was generated (Materials and Methods). brlA encodes the primary regulator of asexual development which is necessary and sufficient for asexual development. The DbrlA mutant produces aerial stalks but is unable to produce the various budded cell types of the conidiophore (Borneman and Andrianopoulos, unpublished). Wildtype, DpakB, DbrlA and DpakB DbrlA strains were grown on ANM + (NH 4 ) 2 SO 4 for 14 days at 25uC. Cell suspensions were made in 0.005% Tween 80 solution and filtered through (B) Co-staining with calcofluor and Hoechst 33258 reveals that wildtype (pakB + ) subapical cells are predominately uninucleate. Occasionally, dividing nuclei can be observed (white arrow). In contrast to wildtype, the subapical cells of the DpakB strain frequently possess 2 or more nuclei (white arrow). Cell ends indicated by white arrowheads and nuclei by double white arrowheads. (C) In wildtype (pakB + ) hyphae, actin is evident as cortical patches located along hyphae and is concentrated at the hyphal apex (white arrowhead). Both cortical actin patches and actin concentrated at the hyphal apex are observed in the DpakB strain (white arrowheads). (D & E) Actin is concentrated at nascent septation sites in wildtype (pakB + ) hyphae (white arrowhead). Like wildtype, actin in the DpakB strain is observed concentrated at nascent septation sites (white arrowheads). Images were captured using differential interference contrast (DIC) or with epifluorescence to observe calcofluor stained fungal cell walls (CAL), Hoechst 33258 stained nuclei (H) or immunofluorescently labeled actin (ACTIN). Scale bars, 20 mm. doi:10.1371/journal.ppat.1000678.g005 Miracloth to remove hyphal cells. As expected, conidia but no yeast cells were observed in wildtype suspensions while the DpakB strain produced conidia of varying size and numerous yeast cells. No conidia or yeast cells were observed in the DbrlA strain. Interestingly, no conidia or yeast cells were observed for the DpakB DbrlA strains suggesting that the yeast cells observed in the DpakB strain arise from the conidiation program (data not shown).
RNA was isolated from wildtype and DbrlA strains grown for 4 days on solid medium at 25uC to investigate if brlA is required for pakB expression during asexual development. Expression of pakB during asexual development was similar in both wildtype and the DbrlA strains showing that pakB expression during conidiation is independent of brlA (data not shown).

PakB is not required for yeast cell morphogenesis in vitro
Given the inappropriate production of yeast cell at 25uC, the effects of pakB deletion on in vitro yeast cell morphogenesis were assessed. Wildtype P. marneffei produced compact, mucoid, yeast colonies after 5 days growth at 37uC in vitro and both the DpakB and DpakB pakB + strains were indistinguishable from wildtype ( Figure 3C). To examine yeast cell morphogenesis in vitro, the wildtype (pakB + ), DpakB, DpakB pakB + and DcflB strains were inoculated on agar-coated BHI slides and incubated for 5 days at 37uC. Wildtype conidia germinate at 37uC to produce polarized arthroconidiating hyphae, an intermediary cell type which is specific to in vitro yeast morphogenesis and which is not manifested in macrophages. Nuclear division and septation become coupled in arthroconidiating hyphae, double septa are laid down and fragmentation occurs along this plane to liberate uninucleate yeast cells which consequently divide by fission. Numerous yeast cells were observed for the wildtype strain after 5 days at 37uC ( Figure 8A). Similarly, the DpakB, DpakB pakB + and DcflB strains produced abundant yeast cells comparable to wildtype and these yeast cells were also uninucleate ( Figure 8A). Thus PakB appears to play no role in vitro. This is consistent with the expression data under this condition in which the pakB transcript was barely detectable ( Figure 1A). The wildtype, DpakB and DpakB pakB + strains also produced numerous yeast cells after 6 days growth at 37uC on Sab and ME yeast medium, comparable to those produced on BHI (data not shown).

Expression of pakB mutant alleles cannot rescue the DpakB phenotype at 25uC
In order to examine the role of conserved domains of PakB in the phenotypes observed, mutant alleles were generated which altered the CRIB (H204G and D195-256) and putative GBB domain (D719-729). The equivalent CRIB mutations in S. cerevisiae Ste20p prevents the Cdc42p interaction (H345G) or generate a constitutively active protein which bypasses the requirement for Cdc42p (DCRIB) [8]. Deletion of the GBB domain prevents Ste20p interaction with Ste4p, therefore affecting activation of the MAPK cascade [13]. These alleles, in addition to the wildtype allele, were placed under the control of the inducible xylP promoter, which is only expressed in the presence of xylose. The DpakB pyrG 2 strain was transformed with these constructs and transformants were directly selected for pyrG + and confirmed by Southern blot analysis of genomic DNA (Materials and Methods). The xylP(p)pakB, xylP(p)pakB H204G , xylP(p)pakB DCRIB and xylP(p)-pakB DGBB strains were grown on media with or without 1% xylose for 10 days ( Figure 9A) or for 4 days on agar-coated slides for microscopic observation ( Figure 9B-D). As expected, all strains exhibited the DpakB phenotype at 25uC on non-inducing medium ( Figure 9A-B). On inducing medium, expression of the xylP(p)pakB construct completely restored the wildtype phenotype ( Figure 9A and C). Expression of the xylP(p)pakB H204G allele partially restored the wildtype hyphal phenotype such that colonies were more filamentous ( Figure 9A) and hyphae were less tightly packed with less apical branching ( Figure 9B-C). In contrast, strains expressing either the xylP(p)pakB DCRIB or xylP(p)pakB DGBB allele were indistinguishable at the colonial level on non-inducing and inducing medium ( Figure 9A). Both the xylP(p)pakB DCRIB and xylP(p)-pakB DGBB strains displayed compact, tightly packed hyphae exhibiting apical branching similar to DpakB, however, the xylP(p)pakB DCRIB strains were also hyperbranched and an increase in yeast cells around the colony periphery were observed ( Figure 9C-D). These yeast cells appeared to be dividing by fission ( Figure 9D).
To investigate the effect of the mutant alleles on asexual development at 25uC, these strains were grown for 14 days at 25uC and examined by SEM. All strains exhibited the DpakB phenotype on non-inducing medium. Unlike the DpakB pakB + Figure 6. Deletion of pakB results in conidiation defects at 25uC. Wildtype (pakB + cflB + ), DcflB and DpakB strains were grown on ANM + (NH 4 ) 2 SO 4 for 14 days at 25uC. SEM of the surface of a wildtype (pakB + cflB + ) colony at 25uC shows numerous asexual development structures (conidiophores). In contrast to wildtype in which conidia are uniform in size, a large number of conidiophores of the DpakB strain bear abnormally large conidia. Single conidiophores can be observed in which separate phialides produce conidia which differ dramatically in size. More than one conidium per phialide is rarely observed and the site of conidium attachment to the phialide is not as constricted as in wildtype (white arrowheads). Like the DpakB strain, conidia of the DcflB strain can vary in size. Lysed conidiophore structures are frequently observed in the DcflB strain, in which the metulae and phialides have ruptured (white arrow) or deflated (double white arrowhead). Scale bars, 5 mm. doi:10.1371/journal.ppat.1000678.g006 strain, whose conidiophores were indistinguishable from wildtype, the xylP(p)pakB strain did not show full complementation on inducing medium; conidiophores mainly consisted of a single phialide and conidiophores with more than one conidium per phialide was rarely observed (data not shown). This suggests that the formation of complex multicellular conidiophores is sensitive to pakB expression levels ( Figure 1A), either due to overexpression of PakB in conidiophore cell types or poor induction of the xylP promoter in conidiophore cell types. On inducing medium, the xylP(p)pakB H204G , xylP(p)pakB DCRIB and xylP(p)pakB DGBB strains were indistinguishable from the DpakB strain (data not shown). Unlike the chains of conidia observed in wildtype conidiophores, more than one conidium per phialide is rarely observed in the DpakB conidiophores. (B-E) Septa in conidiophores. In wildtype conidiophores (B), two separate chitin disks can be observed at phialide to conidium and conidium to conidium connections, indicating cellular division and cell separation. This is not observed in the DpakB conidiophores (C-E). Only one septum (C), no septa (D) or incomplete septa (E) are observed at the phialide to conidia cell boundaries. (F) Conidia were scraped off the colony surface, re-suspended in 0.005% Tween 80 solution, filtered through Miracloth and stained with calcofluor (CAL). Conidia of wildtype (pakB + cflB + ) are uniform in size. In contrast, conidia of the DcflB strain varied in size. Similar to the DcflB strain, conidia of the DpakB strain also varied in size. However, conidia of the DpakB strain were swollen to a greater extent than those from the DcflB strain and often exhibited uneven calcofluor staining. In addition, numerous yeast cells were observed in the DpakB strain. Images were captured using differential interference contrast (DIC) or with epifluorescence to observe calcofluor stained fungal cell walls (CAL). Scale bars, 20 mm (A and F) and 2.5 mm (B-E To assess whether the H204G, DCRIB or DGBB mutations affect PakB localization, HA-tagged pakB H204G , pakB DCRIB and pakB DGBB constructs were generated and co-transformed with the barA + gene into the P. marneffei strain G487 (niaD pyrG areA 2 ). Transformants were selected for glufosinate resistance and confirmed by Southern blot analysis of genomic DNA (Materials and Methods). Anti-HA immunostaining was performed on two strains of each genotype after 4 days growth at 25uC. Like wildtype, both the PakB H204G HA and PakB DGBB HA proteins were concentrated at the hyphal apex, however, the PakB DCRIB HA protein was not (data not shown). The PakB H204G HA, PakB DGBB HA and PakB DCRIB HA proteins showed wildtype localization in conidiophores and at nascent septation sites (data not shown).
Expression of pakB mutant alleles at 37uC in vitro results in yeast cells which resemble those produced in vivo pakB was not expressed in vitro on BHI at 37uC ( Figure 1A) but expression was induced during macrophage infection ( Figure 1B). Yeast cells produced in vivo are shorter and rounder than those produced in vitro and develop directly from conidia rather than via arthroconidiating hyphae. To examine if induced expression of pakB can effect the changes in yeast cell morphology observed in vivo, the wildtype, DpakB, xylP(p)pakB, xylP(p)pakB H204G , xylP(p)-pakB DCRIB and xylP(p)pakB DGBB strains were grown on BHI with or without 1% xylose for 5 days ( Figure 10). As expected, all of the strains were indistinguishable from wildtype and the DpakB strain on non-inducing medium ( Figure 10A). On inducing medium, the xylP(p)pakB strains (copy numbers 2-12) were indistinguishable from the wildtype and the DpakB strains ( Figure 10B). Therefore an increase in pakB expression does not explain the production of rounded yeast cells in vivo. On inducing medium, the xylP(p)-pakB H204G , xylP(p)pakB DCRIB and xylP(p)pakB DGBB strains produced yeast cells which were rounder and greatly reduced in length compared to both the DpakB and xylP(p)pakB strains and a number of yeast cells were produced which appeared to be dividing in a budding manner, similar to that observed during conidiation at 25uC ( Figure 10B-C). The yeast cells produced by the xylP(p)-pakB DCRIB strains were also often curled ( Figure 10B).

PakB is essential for yeast morphogenesis during infection
To investigate if pakB is required for yeast growth during infection, LPS activated J774 murine macrophages were infected with wildtype, DpakB and DpakB pakB + conidia and observed microscopically 24 hours post-infection (Materials and Methods). Calcofluor staining was performed to allow visualization of fungal cell walls. After 24 hours, numerous yeast cells dividing by fission were observed in macrophages infected with wildtype (pakB + ) ( Figure 8B) or DpakB pakB + conidia (data not shown). In contrast, the majority of macrophages infected with DpakB conidia contained highly branched, septate, hyphal cells but no yeast cells ( Figure 8B). A small proportion of macrophages infected with DpakB conidia contained large, swollen conidia ( Figure 8B). To quantify this difference, the number of cells containing at least one septum was recorded for approximately 100 cells on three separate occasions. In contrast to wildtype and DpakB pakB + , in which 17.263.70% and 15.361.74% of cells in infected macrophages contain septa, 71.162.24% of DpakB cells contain septa. This indicates that pakB is required for yeast cell division during infection. To determine if the role played by PakB in vivo is during the initiation or maintenance of yeast-like growth, macrophages were also infected with wildtype, DpakB and DpakB pakB + yeast cells (Materials and Methods). Similar to infection with conidia, after 24 hours numerous yeast cells were observed in macrophages infected with wildtype or DpakB pakB + yeast cells whereas macrophages infected with DpakB yeast cells contained only septate hyphal cells ( Figure S2).
To assess if cellular morphology affects the oxidative state of the host, superoxide production was detected by Nitrotetrazolium Blue Chloride (NBT) staining of macrophages 24 hours postinfection (Materials and Methods). No difference in superoxide production was observed in macrophages infected with either wildtype or the DpakB strain (data not shown). In addition, the presence or absence of superoxide did not affect the growth phenotype of the DpakB strain (data not shown).
To examine whether the presence of host extracts is sufficient to induce the morphological switch, lysed macrophage extracts were added to wildtype, DpakB and DpakB pakB + conidia (Materials and Methods). After 24 hrs at 37uC, all strains were growing as hyphae indicating that the addition of lysed macrophage extracts is insufficient to induce yeast-like growth (data not shown). To observe if being intracellular is sufficient to induce the morphological switch, macrophages infected with wildtype, DpakB and DpakB pakB + conidia were incubated at 25uC. Unlike at 37uC, after   24 hrs all strains grew as hyphae indicating that it is a combination of host and temperature signals which induces yeast-like growth (data not shown).
To investigate if the mutant alleles can complement the phenotype of the DpakB strain in vivo, pakB + , pakB H204G , pakB DCRIB and pakB DGBB constructs were co-transformed with the barA + gene into the DpakB strain and transformants were selected for glufosinate resistance and confirmed by Southern blot analysis of genomic DNA (Materials and Methods). These strains exhibited identical phenotypes to the equivalent xylP overexpression strains Yeast cells produced by the DpakB and DpakB xylP(p)pakB strains on BHI +1% xylose exhibit wildtype morphology. The DpakB xylP(p)pakB H204G , DpakB xylP(p)pakB DCRIB and DpakB xylP(p)pakB DGBB strains produce yeast cells which are rounder and greatly reduced in length compared to both the DpakB and DpakB xylP(p)pakB strains. The yeast cells produced by the DpakB xylP(p)pakB DCRIB strains can also be curled (white arrowheads). (C) On inducing medium, a number of yeast cells produced by the DpakB xylP(p)pakB H204G , DpakB xylP(p)pakB DCRIB and DpakB xylP(p)pakB DGBB strains appear to be dividing in a budding-like manner (white arrowheads). Images were captured using differential interference contrast (DIC) or with epifluorescence to observe calcofluor stained fungal cell walls (CAL). Scale bars, 20 mm. doi:10.1371/journal.ppat.1000678.g010 grown on inducing medium at 25uC (data not shown). LPS activated J774 murine macrophages were infected with conidia from these strains and examined microscopically after 24 hours (Materials and Methods). Calcofluor staining was performed to allow visualization of fungal cell walls. The number of cells containing a septum was recorded for approximately 100 cells, in four transformants of each genotype, in three independent experiments (Table S1) and significant differences assessed by two-level nested analysis of variance (ANOVA) ( Table S2). ANOVA showed there was no significant difference between the wildtype, DpakB pakB + and DpakB pakB H204G strains but there was a significant difference between these strains and the DpakB, DpakB pakB DCRIB and DpakB pakB DGBB strains. There was also a significant difference between the DpakB pakB DCRIB strains and either the DpakB pakB DGBB or DpakB strains. No significant differences were detected between transformants of the same genotype (Table S2). After 24 hours, numerous yeast cells dividing by fission were observed in macrophages infected with wildtype, DpakB pakB + or DpakB pakB H204G conidia ( Figure 8C) of which 17.263.70% (wildtype), 19.560.86% (DpakB pakB + ) and 22.261.3% (DpakB pakB H204G ) of cells contained at least one septum. In contrast, macrophages infected with either of the DpakB, DpakB pakB DCRIB or DpakB pakB DGBB strains contained both septate yeast and hyphal cells ( Figure 8C) and the number of septa was substantially higher than the wildtype (wildtype 17.263.70%; DpakB strain 71.162.24%; DpakB pakB DCRIB 42.661.9%; DpakB pakB DGBB 62.261.99%).
To assess whether the H204G, DCRIB or DGBB mutations affect PakB localization during macrophage infection, anti-HA immunostaining was performed on the PakB H204G HA, PakB DCRIB HA and PakB DGBB HA strains after 24 hours post-infection of LPS activated J774 murine macrophages (Materials and Methods). The PakB H204G HA, PakB DGBB HA and PakB DCRIB HA proteins showed wildtype localization during macrophage infection (data not shown).
The Rho type GTPases CDC42 and Rac are known regulators of PAKs and interact via the CRIB domain. As the DpakB pakB H204G and DpakB pakB DCRIB strains showed either no effect or only partial deregulation of yeast morphogenesis during infection, the in vivo phenotype of cflA G14V , cflA 120A and DcflB mutants was assessed [5,6] (Materials and Methods). After 24 hours, numerous yeast cells dividing by fission were observed in macrophages infected with wildtype (pakB + ) ( Figure 8B). In contrast, DcflB conidia in infected macrophages remained predominately ungerminated ( Figure 8B) (76.266.00% for DcflB compared to 13.364.86% for wildtype). Unexpectedly, numerous yeast cells with wildtype morphology were observed in macrophages infected with the cflA G14V and cflA 120A mutants ( Figure S3). These results suggest that any interaction between CflA and PakB is nonessential for in vivo morphogenesis and that the defects in yeast cell morphology observed in the cflA G14V and cflA 120A mutants in vitro [5] are circumvented when growing inside host cells.

Intracellular morphogenesis and growth
Many pathogens reside within host phagocytic cells where they evade much of the host immune system. For dimorphic fungal pathogens a central attribute which facilitates this immune system avoidance is the capacity to switch from a multicellular hyphal growth form to a unicellular yeast form and it has been demonstrated that blocking this transition abrogates pathogenicity [15,16]. Host body temperature (37uC) is the clearest ex vivo inducer of the hyphal to yeast transition in many dimorphic fungi and it is often assumed that this is the inducer in vivo. Here we show that pakB, which encodes the second p21 activated kinase in the dimorphic pathogen P. marneffei, is strongly upregulated upon phagocytosis by macrophages and is essential for yeast morphogenesis but not growth in macrophages. In contrast PakB plays no role in yeast morphogenesis at 37uC in vitro. This clearly places PakB in a new signalling pathway which responds to host cell inductive signals, not temperature, and is necessary for intracellular yeast morphogenesis and consequently pathogenicity.
The mechanism by which PakB controls yeast cell morphogenesis inside host cells is unique. The Rho type GTPases CDC42 and Rac are known regulators of PAKs in many eukaryotes, interacting via the CRIB domain of these kinases. Consistent with this PakB and CflB orthologues from a number of fungal pathogens have been shown to either physically or genetically interact, yet none of these systems represent intracellular pathogens. The P. marneffei DcflB (Rac) mutant strain fails to germinate in macrophages and mutation of the PakB CRIB domain (pakB H204G ) does not recapitulate this phenotype nor that of the DpakB mutant. In strains where the entire CRIB or predicted Gb binding domains are deleted, there is a partial deregulation of morphogenesis leading to the production of both yeast and hyphal cells in macrophages. Previous studies that have shown that CflA is required for correct yeast cell morphogenesis in P. marneffei during in vitro growth, potentially implicating CflA as a regulator for PakB. However the morphology of yeast cells in CRIB domain pakB mutants, both in vitro and inside macrophages, is essentially wildtype and cflA G14V and cflA 120A mutants produce wildtype yeast cells in vivo showing that any interaction between CflA and PakB is non-essential for in vivo morphogenesis and that the defects in yeast cell morphogenesis observed in the cflA G14V and cflA 120A mutants in vitro may be due to a general defects in morphogenesis.
In dimorphic fungi which have a predominant yeast phase such as C. albicans and Y. lipolytica, mutation of the pakB orthologue (CLA4) leads to defective hyphal formation and invasive growth, and consequent changes in pathogenicity for C. albicans [17,18]. For fungi whose predominant growth phase is hyphal and pathogenic phase is yeast, mutation of pakB has the opposite effect. The simplest explanation for these opposing effects is that pakB orthologues control a conserved fundamental cellular process which is recruited by the organism-specific regulatory systems for either the yeast-hyphal or hyphal-yeast dimorphic switch. Based on the results described here, and discussed below, this process is likely to be the regulation of both morphogenesis and septation during cellular division.

PakB is required for hyphal morphogenesis at 25uC
Dimorphic fungal pathogens must regulate both the transition and maintenance of two vegetative growth states, multicellular hyphal and unicellular yeast, as well as other morphogenetic programs such as asexual development [19]. This capacity is lost when the pakB gene is deleted from P. marneffei. The pakB gene is specifically expressed during hyphal growth and asexual development at 25uC and is required for the generation of these 25uC specific cell types. Loss of pakB leads to defects in polarised growth of hyphal cells and these defects partially overlap with those for cflA (CDC42) and cflB (Rac) mutants. In contrast, the second PAK in P. marneffei plays no role in hyphal cell morphogenesis [12]. Furthermore, loss of pakB leads to inappropriate production of yeast cells at 25uC and it might be argued that PakB is either required to promote hyphal growth or to negatively regulate yeast growth, and that yeast growth is the default state. In support of this hypothesis, a number of hyphal specific genes showed decreased expression while yeast specific genes showed increased expression in the DpakB strain. Studies of the P. marneffei transcriptional corepressor tupA lead to a similar conclusion about the default growth state [20].
Whilst the nature of the default state is likely to be correct, it is clear that pakB plays a more fundamental role beyond hyphal morphogenesis. The inappropriately produced yeast cells in the DpakB strain are not derived from vegetative hyphal cells but from differentiating conidiophore cells. Deletion of the primary regulator of asexual development which is not expressed in vegetative hyphae, the C 2 H 2 Zn finger transcription factor gene brlA, abolished production of yeast cells at 25uC in the DpakB strain. This suggests overlap in the morphogenetic mechanisms controlling conidiation at 25uC and yeast cell production at 37uC and supports previous studies in P. marneffei and other dimorphic fungi [16,20,21,22,23]. More importantly it points to the underlying mechanism being the control of the mode of cellular division. One of the major differences between vegetative hyphal cells and differentiated conidiophore cells is that the former divide by septation (analogous to fission but without cell separation) while the latter divide by budding, so a possible explanation is that PakB is required for the correct execution of budding division during conidiation in P. marneffei and loss of PakB defaults to a fission mode of division with cell separation in conidiophore cell types; as occurs in P. marneffei yeast cells at 37uC. In support of this hypothesis, deletion of the CLA4 homologue in U. maydis results in yeast cells which separate by fission, instead of the normal budding mode at the distal tips of yeast cells, with constriction occurring at a centrally located septum [24]. Although this situation differs from P. marneffei, as it is occurring in undifferentiated cell types, it suggests that fission is acting as the default mechanism for division when budding is aberrant. These results suggest that budding is a derived mode of division and this is consistent with our understanding of the molecular mechanisms which underlie budding, as septation (fission) follows the isotropic growth phase during the budding program.

pakB regulates septation during cellular division
Asexual development is a developmental program in which the hyphal-based apical growth and septation mode of growth switches to acropetal budding division without cell separation and finally basipetal budding division with cell separation, with the concomitant switch to uninucleate cells. Many dimorphic fungi produce yeast cells that divide by budding while P. marneffei produces yeast cells that divide by fission. In all of these instances, the coupling of septation to cell separation determines whether yeast or hyphal (or pseudohyphal) cells will be generated. The pakB deletion strain is able to undergo asexual development, albeit less profusely than wildtype, producing swollen conidia with an enlarged conidium attachment site, more akin to the septa of vegetative hyphal cells. The septa separating the other conidiophore cell types are also aberrant. This suggests that PakB activity regulates the constriction, and possibly formation, of septa during conidiogenesis and is supported by the localization of PakB to conidiophores where it is particularly concentrated at phialide to conidium boundaries. Despite the abnormal morphology of the DpakB conidia, they were still able to germinate normally at both 25uC and 37uC suggesting that it is only the final stages of conidial separation which are affected in the DpakB mutant. This is similar to the rice pathogen Magnaporthe grisea in which deletion of the CLA4 homologue, CHM1, results in irregularly shaped conidia with reduced constriction at the conidium attachment site but is unlike the rye pathogen Claviceps purpurea, in which the Dcla4 strain is completely unable to sporulate [25,26].
Loss of pakB completely blocks the link between septation and cell separation in P. marneffei during intracellular growth, leading to the formation of hyphal cells. This is consistent with the localization of PakB to septation sites in vivo where it is specifically localized to septa only after cell wall deposition had occurred suggesting that PakB is required for cell separation rather than septation. The DpakB phenotypes are consistent with the role of CLA4 homologues in regulating cytokinesis. S. cerevisiae, C. albicans and U. maydis CLA4 mutants are also unable to undergo normal cytokinesis during budding [18,24,27,28]. In S. cerevisiae, CLA4 is known to regulate the activity of Lte1p by localisation to the bud cortex, an important event in the Mitotic Exit Network (MEN), and loss of Lte1p leads to delayed cytokinesis [29,30]. P. marneffei and U. maydis do not have clear LTE1 orthologues and loss of pakB resulted in increased numbers of nuclei per cell compartment inconsistent with reduced or delayed mitotic exit. Furthermore, given the role of CLA4 orthologues in cytokinesis, it may have been expected that deletion of P. marneffei pakB would also result in septation defects in vegetative cells at 25uC. However, DpakB hyphae displayed normal calcofluor stained septa and actin ring formation. Normal calcofluor-white stained septa were also present in the CLA4 deletion strains of C. purpurea, M. grisea and A. gossypii [25,26,31]. Interestingly, the DAgcla4 mutation in A. gossypii results in the absence of almost all actin rings [31].

A role for PakB in the polarisome
The localization of wildtype PakB as a cap at the hyphal apex is very similar to that observed for the A. nidulans polarisome component SpaA [32]. The polarisome is a complex of polarity determining proteins originally identified in S. cerevisiae which include Bni1p (formin), Spa2p (scaffold), Bud6p and Pea2p. In S. cerevisiae Ste20p directly phosphorylates Bni1p [33]. Similar to the DpakB mutant, deletion of A. nidulans sepA (BNI1), spaA (SPA2) or budA (BUD6) results in apical branching [32,34]. In contrast, deletion of P. marneffei pakA (STE20) does not result in apical branching and PakA co-localizes with actin to discrete spots concentrated at the hyphal apex which is inconsistent with polarisome morphology [12]. Thus, in contrast to S. cerevisiae, it is the Cla4p homologue PakB and not the Ste20p homologue PakA which activates the polarisome during hyphal growth.
The CRIB domain is required for Ste20p localization via interaction with Cdc42p in S. cerevisiae. Both the STE20 DCRIB and STE20 H345G mutations result in reduced localization to sites of polarized growth [8] as does the equivalent mutation (pakA H108G ) in P. marneffei [12]. In contrast, the PakB H204G HA and PakB D-CRIB HA proteins exhibited wildtype localization, with the exception of reduced localization to the hyphal apex in the latter, suggesting that the CRIB domain is largely non-essential for localization. This may represent a new paradigm for PAK function in non-yeast fungi and the mechanism by which PakB is correctly localised remains to be determined. p21-activated kinases control cellular morphogenesis and pathogenicity P. marneffei, like many other dimorphic fungal pathogens, has two p21-activated kinases which are responsible for a variety of signaling and morphogenetic activities. The pakA gene encodes a Ste20p-like PAK which is essential for polarity establishment during conidial germination and polarised growth of the pathogenic yeast cells at 37uC such that conidia from DpakA strains fail to germinate upon phagocytosis by macrophages [12]. Based on genetic interaction studies it was shown that PakA lies downstream of CflA. However, pakA is not required for germination, polarised hyphal growth or asexual development at 25uC and it was postulated that this role may be filled by pakB [12]. Based on the data presented here, it is clear that pakB fulfils the predicted role of a polarity determinant during hyphal growth and asexual development but plays no role in conidial germination at 25uC, suggesting that this process, unlike its counterpart at 37uC and in vivo, may be PAK-independent. Formal proof of this hypothesis will require the generation of a DpakA DpakB double mutant. Unexpectedly, pakB plays a critical role in the formation of yeast cells in host cells, instead producing highly branched, septate, hyphal cells, but not in vitro. Upregulation of pakB expression in P. marneffei isolated from macrophages as opposed to in vitro cultured yeast cells shows that PakB activity is likely to be regulated at both the expression level as well as post-translationally. Identifying these host cell specific signals is the important next step in understanding how pathogens sense and respond to their hosts.

Molecular techniques
P. marneffei genomic DNA was isolated as previously described [21]. Southern and northern blotting was performed with Amersham Hybond N+ membrane according to the manufacturer's instructions. Filters were hybridized using [a-32 P]dATP labeled probes by standard methods [35].

Expression analysis
RNA was isolated from 2161 (wildtype) vegetative hyphal cells grown at 25uC for 2 days in liquid medium, from asexually developing cultures grown on solid medium at 25uC for 4 days and from yeast cells grown at 37uC for 6 days in liquid medium. RNA was isolated from yeast cells derived either from LPS activated J774 murine macrophages at 37uC infected with wildtype conidia 24 hours post-infection or from yeast cells incubated in macrophage growth medium (complete DMEM) at 37uC for 24 hours. Macrophages were infected as described below. RNA was also isolated from the DpakB strain after 2 days growth at 25uC in liquid medium and after 6 days at 37uC in liquid medium. RNA was extracted using TRIzol Reagent (Invitrogen) and a MP FastPrep-24 bead beater according to the manufacturer's instructions. RNA was DNaseI treated (Promega) prior to RT PCR analysis and a no cDNA synthesis control was performed to ensure no DNA contamination was present. Expression of pakB, brlA and the 25uC or 37uC specific probes 2E11, 2B10 and 2E4 was determined by RT PCR (Invitrogen Superscript III One-Step RT-PCR with Platinum Taq) using the primers: pakB-Y41

Cloning and plasmid construction
Previously, the A. nidulans STE20 homologous sequence was used to screen a P. marneffei genomic library (constructed in lGEM-11) at low stringency (50% formamide, 2xSSC, 37uC) [12]. A 4 kb SacII/XhoI hybridizing fragment from a second positively hybridizing clone was subcloned into SacII/XhoI digested pBlue-script II SK + (pKB4752). This clone did not contain the entire pakB ORF so a 5.8 kb PstI hybridizing fragment was also subcloned into PstI digested pBluescript II SK + (pKB5794). To generate a clone containing the entire pakB ORF, a 3.2 kb SacI fragment from pKB4752 was cloned into SacI digested pKB5794, this generated a 8.6 kb pakB clone (pKB4904). Double stranded sequencing was performed on 4.2 kb of clone pKB4904 and analyzed using Sequencher TM 3.1.1 (Gene Codes Corporation).
The HA tagged constructs were generated by replacing the 406 bp SacII/HindII fragment of pakB with the SacII/HindII fragment of pKB4693 (3x HA tag) in plasmids pKB6941, pKB7306, pKB7171 and pKB6932, generating pKB7116 (pakB + HA), pKB7118 (pakB H204G HA), pKB7172 (pakB DCRIB HA), and pKB7064 (pakB DGBB HA). This deletes from amino acid 328-343. This region is a poorly conserved region between the conserved CRIB and kinase domains and includes the first intron. To test for functionality, the pKB7116 (pakB + HA) plasmid was co-transformed with the barA + gene into the DpakB strain. Transformants were selected for glufosinate resistance and confirmed by Southern blot analysis of genomic DNA. Transformation of the DpakB strain with the pakB + HA plasmid (pKB7116) complemented the DpakB phenotype.

Fungal strains and media
Strains used in this study are shown in Table 4. P. marneffei FRR2161, cflA G14V , cflA 120A and DcflB have been previously described [5,6]. Transformation was performed using the previously described protoplast method [21]. The DpakB strain (DpakB::pyrG + ) was generated by transformation of strain G487 (niaD 2 pyrG 2 areA 2 ) with linearized pKB6019 and selecting for pyrG + . The DpakB pyrG 2 strain was isolated by plating the DpakB strain (DpakB::pyrG + ) on medium containing 1 mg mL 21 5fluoroorotic acid (5-FOA) supplemented with 10 mM c-amino butyric acid (GABA) and 5 mM uracil to select for the loss of the pyrG marker. The strain is unable to grow in the absence of 5 mM uracil.
The DbrlA DpakB double mutant was generated by transformation of strain G526 (DpkuA niaD 2 pyrG 2 areA 2 ) (K. Boyce and A. Andrianopoulos, unpublished) with linearized pAB5229 (A. Borneman and A. Andrianopoulos, unpublished) and selecting for pyrG + . Deletion of brlA was confirmed by Southern blot analysis. A DbrlA pyrG 2 strain was isolated by plating the DbrlA strain on medium containing 1 mg mL 21 5-FOA supplemented with 10 mM GABA and 5 mM uracil. The DbrlA pyrG 2 strain was transformed with linearized pKB6019 and DbrlA DpakB double mutants selected for pyrG + . Deletion of pakB was confirmed by Southern blot analysis.
In vivo macrophage assay J774 murine macrophages (1610 5 ) were seeded into each well of a 6 well microtitre tray containing one sterile coverslip and 2 mL of complete Dulbecco's Modified Eagle Medium (complete DMEM: DMEM, 10% fetal bovine serum, 8 mM L-glutamine and penicillin-streptomycin). Macrophages were incubated at 37uC for 24 hours before activation with 0.1 mg mL 21 lipopolysaccharide (LPS) from E. coli (Sigma). Macrophages were incubated a further 24 hours at 37uC, washed 3x in phosphate buffered saline (PBS) and 2 mL of complete DMEM medium containing 1610 6 conidia or yeast cells (grown at 37uC for 8 days in liquid BHI medium) was added. A control lacking conidia or yeast cells was also performed. Macrophages were incubated for 2 hours at 37uC (to allow conidia or yeast cells to be engulfed), washed once in PBS (to remove free conidia) and incubated a further 24 hours at 37uC. Macrophages were fixed in 4% paraformaldehyde and stained with 1 mg mL 21 fluorescent brightener 28 (calcofluor -CAL) to observe fungal cell walls. The numbers of germinated conidia was measured microscopically by counting the numbers of germinated conidia (conidia with a visible germ tube) or yeast cells in a population of approximately 100 fungal cells in macrophages. The numbers of cells with septa was measured microscopically by counting in a population of approximately 100 cells. Three independent experiments were performed. Mean and standard error of the mean values were calculated using GraphPad Prism3. To analyze the significance of the septation results, two-level nested analysis of variance (ANOVA) was performed on the data to test if the percentage of septate cells was significantly different among genotypes and also between transformants of the same genotype (http://udel.edu/ ,mcdonald/statnested.html) (Table S2).
Superoxide production was detected by addition of a 0.05% Nitrotetrazolium Blue Chloride (NBT) (Sigma) solution (in 0.05 M sodium phosphate buffer (pH 7.5)) to macrophages 24 hours postinfection. Macrophages were incubated for 1 hour and observed by differential interference contrast (DIC) microscopy.
To examine whether the presence of host extracts is sufficient to induce the morphological switch, LPS activated J774 murine macrophages in complete DMEM were lysed by freezing at 270uC for 20 minutes and slow thawing. The lysed extracts were added to wildtype, DpakB and DpakB pakB + conidia and incubated for 24 hrs at 37uC. Minus macrophage controls were also performed for comparison. Microscopy P. marneffei strains were grown on slides covered with a thin layer of solid medium, with one end resting in liquid medium [21]. Strains were grown on ANM medium supplemented with (NH 4 ) 2 SO 4 at 25uC for 2 or 4 days. To observe conidia and yeast cells, asexually developing plates were harvested into 0.005% Tween 80 solution and filtered through Miracloth to remove hyphae. To observe conidiophores, asexually developing plates were scraped onto a coverslip containing 5 mL 0.005% Tween 80 solution. At 37uC strains were grown on BHI medium for 5 days or Sab and ME medium for 6 days. xylP(p) strains were grown on ANM plus 10 mM (NH 4 ) 2 SO 4 61% xylose at 25uC and on BHI 61% xylose at 37uC. For germination experiments, approximately 10 6 spores were inoculated into 300 mL of SD plus 10 mM (NH 4 ) 2 SO 4 and incubated for 15 hours at 25uC or 37uC. The rates of germination were measured microscopically by counting the numbers of germinating conidia (conidia with a visible germ tube) in a population of 100 cells. The number of conidia with 1, 2 or 3 or more germ tubes was counted in a population of 100 cells. Three independent experiments were performed. Mean and standard error of the mean values were calculated using GraphPad Prism3. Immunofluorescence microscopy for examination of the actin cytoskeleton was performed with a mouse C4 monoclonal  anti-actin primary (Chemicon International, Inc.) and an ALEXA 488 rabbit anti-mouse secondary antibody (Molecular Probes). PakB HA immunofluorescence localization was performed with 3F10 rat monoclonal anti-HA primary (Boehringer Mannheim) and an ALEXA 488 goat anti-rat secondary antibody (Molecular Probes) using standard protocols [38]. Immunofluorescence microscopy controls using only primary or secondary antibodies were performed to confirm the specificity of the antibodies. Slides were examined using differential interference contrast (DIC) and epifluorescence optics for antibody fluorescence, cell wall staining with fluorescent brightener 28 (calcofluor -CAL) or nuclear staining with Hoechst 33258 and viewed on a Reichart Jung Polyvar II microscope. Images were captured using a SPOT CCD camera (Diagnostic Instruments Inc) and processed in Adobe Photoshop TM .
For scanning electron microscopy (SEM), strains were grown on ANM or BHI for 10 days at 25uC or BHI for 5 days at 37uC. xylP(p) strains were grown on carbon-free ANM plus 10 mM (NH 4 ) 2 SO 4 supplemented with either 1% glucose or 1% xylose at 25uC for 14 days. Agar cubes containing the fungal biomass were fixed with 2.5% glutaraldehyde in PBS buffer for 2 hours, washed 3 x in PBS and postfixed with 1% osmium tetroxide for 2 hours. Samples were then washed 3 x in PBS and ethanol dehydrated by washing in increasing concentrations of ethanol. Samples were dried in a Balzers CPD 030 Critical Point Dryer and gold coated in an Edwards S150B Gold Sputter Coater. Samples were examined with a Philips XL30 FEG Field Emission Scanning Electron Microscope. Figure S1 PakB is the P. marneffei Cla4 homologue. Unrooted phylogenic tree of Ste20p homologues from Saccharomyces cerevisiae (ScSte20), Ashbya gossypii (AgSte20), Candida albicans (CaCst20), Ustilago maydis (UmSmu1) and Penicillium marneffei (PmPakA) and Cla4p homologues from S. cerevisiae (ScCla4), A. gossypii (AgCla4), C. albicans (CaCla4), U. maydis (UmCla4) and Clavaceps purpurea (CpCla4). The third PAK in S. cerevisiae, ScSkm1, is also included. P. marneffei PakB (PmPakB) shows more sequence homology to Cla4p homologues than to Ste20p homologues. Found at: doi:10.1371/journal.ppat.1000678.s001 (0.56 MB PDF) Figure S2 Infection of macrophages with DpakB yeast cells results in morphological defects in vivo. LPS activated J774 murine macrophages infected with yeast suspensions of the wildtype (pakB + ), DpakB and DpakB pakB + strains. After 24 hours, numerous yeast cells dividing by fission were observed in macrophages infected with wildtype (pakB + ) or the DpakB pakB + strains. Macrophages infected with DpakB yeast cells contained highly branched, septate, hyphal cells. Images were captured using differential interference contrast (DIC) or with epifluorescence to observe calcofluor stained fungal cell walls (CAL). Scale bars, 20 mm. Found at: doi:10.1371/journal.ppat.1000678.s002 (4.99 MB PDF) Figure S3 cflA mutants produce yeast cells with wildtype morphology in vivo. LPS activated J774 murine macrophages infected with conidial suspensions of cflA + , cflA G14V and cflA D120A strains. After 24 hours, numerous yeast cells dividing by fission were observed in macrophages infected with all strains. Images were captured using differential interference contrast (DIC) or with epifluorescence to observe calcofluor stained fungal cell walls (CAL). Scale bars, 10 mm.