Two GDP-mannose transporters contribute to hyphal form and cell wall integrity in Aspergillus nidulans

Filamentous fungi play important roles in nutrient cycling and industry, and several cause diseases of plants and animals. Increasingly, fungi are seen as major threats to immunosuppressed individuals, such as organ recipients and patients with cancer or HIV (Turner et al., 2006). One of the structures that differentiates a fungus from its host is the chitinous cell wall, which mediates many of the organism's interactions with its environment and regulates cell shape at all stages of development. An improved understanding of the fungal wall and of the processes that determine its integrity would add to our ability to control activities of fungal friends and foes (Maertens & Boogaerts, 2000). In addition to chitin and other polysaccharides, which play principally structural roles, a variety of proteins are also localized within cell walls (De Groot et al., 2005; Bowman & Free, 2006; Lesage & Bussey, 2006; Richard & Plaine, 2007). These play a variety of roles, including cell adhesion (Verstrepen et al., 2004), mediation of iron uptake (Protchenko et al., 2001), cell wall remodelling (reviewed by Lesage & Bussey, 2006), regulation of cell wall porosity (De Nobel et al., 1990), and maintenance of cell wall integrity (van der Vaart et al., 1995).

All cell wall proteins characterized to date are glycoproteins, which have been modified by post-translational addition of mannose and other sugars (De Groot et al., 2005; Bowman & Free, 2006). All evidence indicates that this process takes place within an endomembrane system (Bourett et al., 2007) that is functionally equivalent to that of higher plants and animals, though a significant structural difference exists in the fungal Golgi apparatus, which consists of unstacked, dispersed cisternae termed Golgi equivalents (Howard, 1981). (For the sake of brevity, this structure will be referred to simply as the Golgi in this paper.) Secreted fungal proteins typically contain evolutionarily conserved N-linked oligosaccharides similar to those found in animal glycoproteins (Dean, 1999), and many are also heavily O-glycosylated in domains rich in serine and threonine residues (Willer et al., 2005). In fungi, in contrast to mammals, mannosyl residues are the dominant structural component of both classes of oligosaccharide (Mansour & Levitz, 2003). In both classes, glycosylation is initiated in the endoplasmic reticulum (ER), where either a simple (O-linked) or complex pre-assembled (N-linked) precursor is transferred from a lipid donor to the recipient protein. After transfer of the glycoproteins to the Golgi via regulated vesicular traffic, oligosaccharide modification continues in a multistep process, dependent upon glycosidases, glycosyltransferases and membrane transporters (Gemmill & Trimble, 1999). The sugar donors in glycosyltransferase reactions are nucleotide sugars. These solutes must be transported from the cytosol into the Golgi lumen by specific nucleotide sugar transporters (NSTs), which are type III transmembrane proteins (Berninsone & Hirschberg, 2000). The best studied of these is the Saccharomyces cerevisiae GDP-mannose transporter (GMT) VRG4, which forms autodimers in the functional state (Gao & Dean, 2000).

Because cell wall glycoproteins must transit through and be modified within the endomembrane system, a wide range of intracellular proteins must be expected to play roles in determining wall structure and function (Goto, 2007). Accordingly, critical wall-related and morphogenetic roles have been demonstrated for several proteins involved in vesicular traffic (e.g. Whittaker et al., 1999; Shaw et al., 2002; Shi et al., 2004; Yang et al., 2008) and in the pathways of protein mannosylation (e.g. Shaw & Momany, 2002; Willer et al., 2005; Upadhyay & Shaw, 2006).

Our laboratory is interested in identifying new genes whose activity is important to the integrity of cell walls in filamentous fungi. Mutants with cell wall defects can be identified in a variety of ways; many, for instance, are highly sensitive to the chitin-binding agent Calcofluor White (CFW) (e.g. de Groot et al., 2001; Hill et al., 2006). In this study we report the isolation of one such mutation, calI11 (for Calcofluor hypersensitivity), whose phenotype also includes shortened hyphal compartments with increased branch density in the absence of CFW stress, as well as reduced wall mannosylation. We demonstrate that the calI11 mutation lies in an Aspergillus nidulans orthologue of the S. cerevisiae gene encoding the GMT VRG4 (Gao & Dean, 2000), which we designate gmtA. The distribution of the GmtA–GFP chimera is consistent with that of Golgi proteins. We also describe a paralogue of gmtA, designated gmtB, which can act as an extra-copy suppressor of the calI11 phenotype. To our knowledge, this is the first study of a GMT in a filamentous fungus.

Strains, media and basic culture methods.
Strains used in this study are listed in Table 1. Complete medium (CM) consisted of 1 % glucose, 0.2 % peptone, 0.1 % yeast extract, 0.1 % casamino acids, 5 % nitrate salts, 1 % trace elements, 0.1 % vitamin mix, 1.2 mM L-arginine, 10 mM uracil and 5 mM uridine. Vitamin mix and nitrate salts are described in the appendix of Kafer (1977). Trace element solution is described in Hill & Kafer (2001). Minimal medium (MM) consisted of 1 % glucose, 5 % nitrate salts, 1 % trace elements, 0.001 % thiamine hydrochloride and 25 ng biotin ml^–1. Solid media contained 1.5 % agar and 50 mg ampicillin ml^–1. All cultures were incubated at 30 °C.

Table 1. A. nidulans strains used in this study

Mutagenesis, screening, and Mendelian analyses.
Conidia of A. nidulans strain A28 were mutagenized to 50 % mortality with 4-nitroquinoline-1-oxide (NQO) as described by Harris et al. (1994). Survivors were screened for sensitivity to 10 µg CFW ml^–1 (Blankophor BBH, a gift from Bayer Corporation) as described in Hill et al. (2006). Strain RCH-30, which exhibited CFW hypersensitivity along with compact colony diameter, was selected for further study and crossed with strain GR5, which has wild-type branching and CFW sensitivity, using standard genetic methods (Kafer, 1977; Kaminskyj, 2000). Analysis of the resulting progeny showed that hypersensitivity and small colony diameter co-segregated in a 1 : 1 ratio, and both traits were recessive in the diploid state. Strain R205 (pyrG89; wA3; calI11; pyroA4) used elsewhere in this research was selected from this cross. Diploids constructed between RCH-30 and strains bearing mutations in cal loci described previously (calA–H, allele numbers 1–10; Hill et al., 2006) confirmed that the RCH-30 mutation was in an independent locus. For this reason we have designated the allele calI11, following the nomenclature recommendations of Clutterbuck & Arst (1995). Following the identification of a base substitution in the mutant strain allele of AN8848.3 (gmtA), a sexual cross was performed between calI11 strain R205 and phenA2 strain A498 to test for linkage between the two loci. Only 9.5 kb separated phenA from AN8848 (according to The Aspergillus nidulans Linkage Map; ).

Library transformations and identification of complementing genes.
Strain R205 was transformed according to Yelton et al. (1984) using the AMA-NotI genomic library (Osherov et al., 2000). Transformants were selected for restoration of pyrimidine prototrophy on minimal medium containing 1.1 M sorbitol. Conidia from transformed colonies were tested for restoration of wild-type colony morphology and for resistance to CFW, and two complemented strains (R205-XF1 and R205-XF2) were selected for further study. Genomic DNA was isolated from each strain, and plasmids were recovered by transforming competent Escherichia coli. Two recovered plasmids (designated pR205-XF1 and pR205-XF2) complemented the calI11 phenotype upon retransformation of strain R205. The genomic inserts of both plasmids were end-sequenced using vector-specific primers, and the resulting end sequences were compared with the Broad Institute Aspergillus nidulans sequenced DNA database (Aspergillus Sequencing Project, Broad Institute of MIT and Harvard; Galagan et al., 2005) to determine the intervening wild-type genomic sequences. One autocalled gene (AN8848.3) was represented in the genomic insert of pR205-XF1, and two (AN9298.3 and AN9299.3) in the insert of pR205-XF2. tBLASTN () was used to search NCBI databases for sequences showing homology to AN8848.3 (gmtA) and AN9298.3 (gmtB), and alignments with selected sequences were performed using CLUSTAL W (European Bioinformatics Institute; ).

Cloning of AN8848.3 (gmtA), AN9298.3 (gmtB) and AN9299.3.
Candidate genes were PCR-amplified from A28 genomic DNA with an additional 600–1000 bp of upstream sequence and 100 bp of downstream sequence, using Pfu Turbo polymerase (Stratagene) and gene-specific primers (Table 2). PCR products were purified using the Qiagen PCR purification kit, and gel pieces were purified using the Qiagen QIAquick Gel Extraction kit. The cleaned PCR products were ligated into the pRG3-AMA1-NotI plasmid (Fungal Genetics Stock Center, University of Missouri, Kansas City, MO), using the Quick Ligation kit (New England Biolabs). The constructs containing wild-type gmtA, gmtB and AN9299.3 were designated pLJH131, pLJH133 and pLJH134, respectively. Protoplasts from strain R205 were transformed separately with the different constructs, and transformants were tested for complementation of the calI11 phenotype, confirming that the complementing sequences are wild-type AN8848.3 (gmtA) and AN9298.3 (gmtB).

Table 2. Primers used in this study

Sequencing the mutant allele.
The gmtA and gmtB alleles of the calI11 mutant strain were amplified from R205 genomic DNA using the same primers used in the preceding section and Pfu Turbo. Two separate PCR reactions were run for each gene, and products from each reaction were cloned into pGEM-5Zf(+) (Promega) before transformation of competent E. coli. Two clones from each PCR reaction were selected for sequencing using overlapping primer sets (not listed), giving at least twofold coverage of each sequence. The respective sequences were aligned to the wild-type (strain A4) genomic DNA sequence in the Broad Institute database and to our own sequences of the identical region of strain GR5 (parent strain of strain R205) with CLUSTAL W in order to identify the genetic lesion in gmtA.

Expression of GFP and red fluorescent protein (RFP) chimeras.
We have not been successful in generating a stable transformant that expresses a GmtB–GFP chimera. A gmtA–GFP expression construct was generated by fusion PCR (Szewczyk et al., 2006). An initial round of PCR reactions was performed to amplify the 1 kb 5'- and 3'-flanking regions of gmtA using Pfu Turbo and genomic DNA of strain A1145 (Nayak et al., 2006) as a template. Primers used in amplifying the 5'-flanking regions were designed to omit the stop codons from the expression constructs (Table 2). Also included in the initial round of PCR reactions was a reaction amplifying the GA5-GFP-Af-pyrG cassette using plasmid pFNO3 (Yang et al., 2004; Fungal Genetics Stock Center) as a template. Each of the three PCR products was run on a 0.65 % agarose gel, and the appropriately sized fragments were excised and gel-purified as described above. Fusion PCR reactions to generate the linear gmtA-GFP-Af-pyrG expression construct contained 100 ng each of 5'- and 3'-flanking regions and GA5-GFP-Af-pyrG cassette. The fusion products were run on 0.65 % agarose gels, and the linear 4.6 kb gmtA-GFP-Af-pyrG construct was excised and gel-purified using the QIAquick Gel Extraction kit. The nkuA deletion strain A1145 was transformed according to Osmani et al. (2006) to create strain R633, and transformants were selected for conversion to pyrimidine prototrophy on MM plates supplemented with 1.1 M sorbitol, 0.05 µg pyridoxine ml^–1 and 0.1 µg riboflavin ml^–1.

In order to create strains that expressed GmtA–GFP along with CopA–monomeric RFP (mRFP), a GA5–mRFP cassette containing the riboB gene from Aspergillus fumigatus was constructed. The riboB gene (Afu1g13300), along with 500 bp of 5'-flanking and 100 bp of 3'-flanking sequence, was PCR-amplified using primers that added XbaI sites at the 5' and 3' ends (Table 2) using A. fumigatus (strain A1100) genomic DNA as a template. Plasmid pXDRFP4 (Yang et al., 2004; Fungal Genetics Stock Center) was digested with XbaI (which digests within the sequence between the mRFP sequence and the pyrG sequence, and in the PCR-BluntII-TOPO multi-cloning site) to remove pyrG, and XbaI-treated riboB was cloned into XbaI-digested pXDRFP4. The expression construct (CopA–mRFP) was generated in a manner similar to that utilized in creating the gmtA-GFP-Af-pyrG expression constructs. The initial round of PCR amplified 500 bp of the 5'- and 3'-flanking regions of CopA (AN3026.3) and the GA5-mRFP-Af-riboB cassette. Each of the three products was gel-purified, and 100 ng of each was used in a fusion PCR reaction that produced a linear 3.6 kb CopA-mRFP-Af-riboB construct. Gel-purified CopA-mRFP-Af-riboB was used to transform strain R633 (A1145/gmtA-GFP) according to Osmani et al. (2006), and transformants were selected for conversion to riboflavin prototrophy on MM plates supplemented with 1.1 M sorbitol and 0.05 µg pyridoxine ml^–1.

Microscopic methods and morphological characterization of the calI11 phenotype.
All observations were made using an Olympus BX51 epifluorescence microscope, in either fluorescence or transmitted-light mode, equipped with a SPOT RT-SEM digital camera (Diagnostic Instruments). Fluorescence observations employed a x100 1.35 numerical aperture objective.

Basic morphological observations were made by applying 5 µl drops containing between 500 and 2000 conidia onto agar media and incubating for 15–24 h, after which germlings were covered with liquid medium of matching solute composition and observed under a coverslip. Fluorescence observations were made using liquid-grown coverslip cultures (Harris et al., 1994). For quantitative assessment of septal distance, hyphal width, nuclear numbers, etc., coverslip cultures were fixed, stained with Hoechst 33258 dye (Sigma) and CFW (to stain nuclei and cell walls, respectively), mounted as described by Harris et al. (1994) and observed using the Olympus U-MWU2 filter set. Apical compartments were defined as those distal to the first septum, and apical branches were defined as any branch occurring in this compartment. Intercalary compartments are those other than the terminal compartment. Branch density (intercalary branch number per 10 µm) was calculated by counting the total number of branches in randomly selected five-compartment hyphal segments (intercalary). Spore body diameters were measured at right angles to the longitudinal axis of the germling. Measurements were made from photographs taken with a x40 objective and displayed at x1167 magnification. Statistical differences between populations (mutant vs wild-type vs complemented) were determined using two-tailed t tests at the 0.05 level of significance, based on the observation of at least 30 randomly selected germlings for each condition.

For staining with FITC–Concanavalin A (FITC–ConA; Sigma), coverslip cultures were washed for 1 min in deionized water, then incubated in 200 µl drops of 250 µg aqueous FITC–ConA ml^–1 for 30 min at room temperature, followed by a brief water rinse and immediate observation using the Olympus U-WMB2 filter set. Mixed cultures (mutant plus wild-type) were employed to ensure that identical conditions of illumination and photographic exposure were used for comparisons of staining intensity. For observation of cells expressing GFP or mRFP chimeras, coverslips with attached germlings were either transferred directly to slides and observed without fixation, or alternatively they were first fixed for 30 min at room temperature in PIPES-buffered formaldehyde (Kaminskyj, 2000). The U-MWB2 filter set was used for observation of the GFP signal, and the U-MWG2 filter set was used for mRFP. For simultaneous imaging of ER and Gmt–GFP, coverslip-grown germlings of strain R633 were incubated in 200 µl drops of 10 µM ER-Tracker Blue-White DPX (Molecular Probes) in growth medium for 30 min, followed by 30 min in dye-free medium and immediate observation using the U-MWU2 filter set. To assess potential colocalization of differently coloured signals (GFP vs ER-Tracker or GFP vs mRFP), separate channels were first imaged in grayscale, then converted to complementary pseudocolours, and finally merged in RGB mode using Adobe PhotoShop software.

Characterization of calI11 phenotype
The calI11 mutant strain was selected principally on the basis of its hypersensitivity to CFW. The strain's complete inability to tolerate exposure to CFW at tested levels (Fig. 1) was comparable to the most extreme sensitivity shown by cal mutant strains reported in previous studies (Hill et al., 2006). In addition, calI11 mutants exhibited severe morphological defects, demonstrated in Fig. 2(a–e) and Table 3, in which intercalary compartments were greatly shortened, with a comparable increase in branch density. Septum-to-septum distance was reduced to about 44 % of that of the wild-type. The length of apical compartments was also greatly reduced (to ∼22 % of the wild-type length), and all apical compartments contained hyphal branches, compared with just 5 % apical branching frequency in the wild-type. Hyphal width was irregular compared with the wild-type and was significantly increased (∼1.3-fold that of wild-type), as was the degree of swelling of spores during germination. (The ∼1.8-fold increase in diameter over wild-type equals an approximately sixfold increase in volume.) Many cells bore prominent vacuoles (Fig. 2b). These cellular defects were reflected in a greatly reduced colony size (Fig. 2f). These defects were not remedied by growth with high osmoticum or by media containing elevated levels of mannose.

(125K): Fig. 1. Sensitivities of mutant, wild-type and transformed strains to CFW. Colonies were grown for 2 days at 30 °C on CM plus pyrimidines, containing either 10 or 0 µg CFW ml–1. Each vertical column is a twofold dilution series, beginning with (top) 10 000 spores in 5 µl. R205+gmtA has been transformed with plasmid pLJH131. R205+gmtB has been transformed with plasmid pLJH133. R205+pRG3 has been transformed with the empty pRG3 vector (control).

(71K): Fig. 2. Effects of calI11 mutation on cell and colony morphology. (a, b) Transmitted light images of germlings of calI11 mutant strain R205 at two different magnifications (∼24 h old and ∼16 h old, respectively) showing hyperbranched, contorted growth form and excessive swelling of the spore body. (c) Transmitted light image of germling of strain GR5 (wild-type morphology) at ∼10 h. (d, e) Fluorescence images of strains R205 and GR5, respectively, stained to show nuclei and septa (examples indicated by arrows) using Hoechst 33258 and CFW. (f) Comparison of the tight colony morphology of 2-day-old spore-grown colonies of mutant strain R205 with that of wild-type strain GR5. Bars, 25 µm, except (a),100 µm.

Table 3. Quantitative morphological measurements of calI11 mutant, wild-type and complemented strains Only population means are reported (omitting measures of dispersion) for clarity of presentation. In each column, values with the same superscript letter are not significantly different.

Complementation of calI11
Transformation of the calI11 mutant strain R205 with an A. nidulans high copy plasmid genomic library yielded two strains showing wild-type colony morphology (see Methods), from which two complementing plasmids were recovered: one containing the predicted gene AN8848.3 (1312 bp), and a second containing predicted genes AN9298.3 (927 bp) and AN9299.3 (837 bp). Retransformation with each of the three separately cloned ORFs (including ∼600–1000 bp upstream promoter sequence) demonstrated that the complementing genes were AN8848.3 and AN9298.3 (Fig. 1, Table 3). All morphological aspects of the complemented phenotype were restored to essentially wild-type levels, with the exception of the frequency of branching in apical compartments, which was partially, but still significantly, restored (Table 3). Resistance to CFW was restored to nearly, though not fully, wild-type levels (Fig. 1).

A translated BLAST search revealed that the closest homologues of AN8848.3 and AN9298.3 were yeast GMTs (Supplementary Fig. S1; Fig. 3). To reflect this homology, we have assigned the functional gene names gmtA and gmtB, respectively. AN8848.3 is predicted to encode a 380 aa protein with ∼50 % identity and 55 % similarity to S. cerevisiae VRG4, while AN9298.3 encodes a predicted 271 aa product with ∼43 % identity and 50 % similarity to VRG4 (see Supplementary Fig. S1). GmtA and GmtB share 41 % identity and 58 % similarity. Each of the two gene products is predicted to be a multipass transmembrane protein (SOSUI Engine ver. 1.11; ).

(39K): Fig. 3. Location of amino acid substitution in calI11. CLUSTAL W alignment of the GALNK motifs of GMTs from several fungal species plus the protozoan parasite Leishmania donovani. These are NSTs for which experimental evidence exists to support substrate specificity for GDP-mannose. Conserved amino acids are indicated by shaded characters. The alanine to proline substitution in the calI11 mutant is indicated by an arrow.

Identification of the calI11 mutation
Sequence analysis of AN9298.3 (gmtB) copied from R205 genomic DNA revealed no genetic lesion. However, AN8848.3 (gmtA) from the same source contained a G to C mutation at base 943, which was predicted to cause an A315P amino acid substitution in predicted exon 4 (Fig. 3). This region is highly conserved in yeasts and filamentous fungi (Fig. 3; Gao & Dean, 2000). A cloned gmtA allele containing this same G to C substitution failed to complement the calI11 phenotype when transformed into strain R205 (data not shown). Further confirmation that calI is gmtA was obtained from the results of a sexual cross between strains R205 (calI11) and strain A498, containing the phenA2 allele, which is separated from the AN8848.3 locus on chromosome III by ∼9.5 kb. Out of 117 progeny tested, only one showed a recombinant phenotype, indicating an exceptionally tight degree of linkage (less than 1 cM map distance).

Mannose deficiency in calI11 cell walls
In light of the fact that the closest homologues of gmtA are transporters of GDP-mannose, we explored the possibility that the calI11 mutation might cause a detectable reduction in mannosylation of cell wall constituents. Fig. 4 shows a co-culture of strains GR5 and R205 treated with the mannose-binding lectin FITC–ConA, and demonstrates a marked reduction in staining of the mutant cell walls compared with those of wild-type (longer, thinner hyphae in Fig. 4). Mutants complemented with constructs containing either gmtA or gmtB were restored to wild-type levels of ConA staining (data not shown).

(89K): Fig. 4. Differential staining of cell walls of calI11 mutant and wild-type hyphae by FITC–ConA. (a) Fluorescence image of a mixed culture containing cells of calI11 mutant strain R205 and wild-type strain GR5. (b) Transmitted light image of the same field. A single wild-type hypha crosses the field horizontally, intersecting with a weakly stained mutant germling. Bar, 25 µm.

Cellular localization of GmtA
Transformation of mutant strain R205 with a plasmid expressing a C-terminal GFP-labelled wild-type copy of AN8848.3 complemented the calI11 phenotype, demonstrating the functionality of the GmtA–GFP hybrid protein (data not shown). GmtA–GFP displayed a punctate pattern of fluorescence (Fig. 5a), consistent with those observed in other studies using putative Golgi markers, and markedly different from that shown by the ER-specific dye ER-Tracker (Fig. 5b). Occasional ring-shaped profiles were also observed (Fig. 5b). We studied GmtA localization further by comparing the fluorescence pattern with that of a putative Golgi-localized protein, CopA (Breakspear et al., 2007), which displayed a punctate pattern of fluorescence similar to that observed with GmtA (Fig. 5c, d). Initial observations made with live cells indicated that the two proteins did not colocalize. However, the mobility of both signals was high enough to allow for some organelles to shift position while switching filters, so the observations were repeated using fixed cells (Fig. 5c, d). These observations confirmed that the two signals were in different compartments, with little indication of overlap. CopA distribution was highest within ∼20 µm of the hyphal apex, while GmtA distribution was higher in more basal regions (Fig. 5d).

(65K): Fig. 5. Subcellular localization of GmtA, ER and CopA. (a) Distribution of GmtA–GFP in living hyphal apices of strain R633. (b) Merged image of strain R633 stained with the ER-specific vital dye ER-Tracker (pseudo-coloured in red), demonstrating the separate localization of ER and GmtA–GFP. Arrows indicate ring-shaped localizations of GmtA–GFP. (c) Merged image of a fixed hypha of strain R559, expressing CopA–mRFP (red signal) and GmtA–GFP (green signal). (d) Merged image of a fixed hypha of strain R559, demonstrating the tip-high distribution of CopA–mRFP (red signal) and the less tip-oriented distribution of GmtA–GFP (green signal). Bars, 5 µm.

The most distinctive aspects of the calI11 phenotype are its extreme sensitivity to CFW, its widened stubby hyphal segments, and the increased frequency of branching, including branching in apical compartments. CFW hypersensitivity is well established as an indicator of defects in cell wall integrity (de Groot et al., 2001; Hill et al., 2006), and studies with cell wall mutants and gene deletions have demonstrated an increase in branch density as a result of some wall defects (e.g. Ichinomiya et al., 2002; Mellado et al., 2003). Wild-type A. nidulans hyphae generally have few or no branches in the apical compartment and one or fewer branches per intercalary compartment (Turner & Harris, 1999). Hyperbranching mutants tend to have closely spaced septa and wider hyphae, as seen in calI11, which leads to the proposal that the number of branches per unit of volume remains about the same in both wild-type and hyperbranching mutants (Turner & Harris, 1999). Since the widths of hyphal compartments in calI11 are highly irregular, no volumetric calculations were attempted; however, the combination of increased width and decreased length is consistent with the view that compartment volumes are not greatly changed. If, in fact, compartment length changes to keep pace with changes in diameter, then some aspects of hyperbranching may result from changes in determinants of cell polarity at the hyphal tip (Harris et al., 2005), which are ultimately responsible for establishing cell diameter.

The results of this investigation strongly support the conclusion that the defective gene in the calI11 mutation is a GMT located in a compartment of the Golgi which functions in mannosylation of hyphal wall constituents. All aspects of the calI11 phenotype are complemented by the wild-type allele of gene AN8848.3, whose closest homologues are fungal GMTs. Meiotic mapping demonstrates that the calI11 mutant allele resides in the region of chromosome III where AN8848.3 is located, and the AN8848.3 allele of the mutant strain contains a point mutation which renders that allele unable to complement calI11 when introduced into the pRG3 plasmid. Although substrate specificity among NSTs cannot be predicted from sequence similarity (Berninsone & Hirschberg, 2000), the greatly reduced staining of calI11-strain cell walls by the lectin ConA, which has strong binding affinity for mannosyl residues (Baenziger & Fiete, 1979), indirectly supports a defect in mannosylation.

We conclude that AN9298.3 (gmtB) represents a second GMT in the A. nidulans genome, based upon its sequence homology and its capacity to complement the defect in gmtA when overexpressed. The complementing ability of introduced NSTs can be used as a functional demonstration of transporter substrate specificity (Berninsone & Hirschberg, 2000; Cottrell et al., 2007). Whether gmtB functions in concert with gmtA, or instead performs independent functions in different compartments or at different developmental stages, will be the object of further studies.

A role for GMTs in hyphal morphogenesis and wall metabolism, as indicated in this study, is consistent with the results of many other studies in both yeasts and filamentous fungi. Defects in cell wall integrity (indicated by hypersensitivity to wall-compromising agents such as CFW and Congo red) and/or morphological defects, consisting usually of impaired establishment or maintenance of polarity, have been tied to defects in all major stages of mannosylation. These include defects in the cytosolic reactions that precede establishment of mannosyl linkages (Smith & Payton, 1994; Upadhyay & Shaw, 2006), defects in ER protein mannosyltransferases (PMTs) (Timpel et al., 1998; Momany et al., 1999; Shaw & Momany, 2002; Oka et al., 2004; Willer et al., 2005), and defects in the early Golgi mannosyltransferase Och1p (Bates et al., 2006). Furthermore, a screen for S. cerevisiae strains showing defects in polarity detected mutants in the GMT VRG4 (Mondésert et al., 1997).

In yeast, defects in protein glycosylation activate major components of the cell wall integrity pathway, including protein kinase C and MAP kinase pathways (Cullen et al., 2006). Mannosylation defects could lead to loss of wall integrity (with potential connections to polarity) through a variety of routes. In S. cerevisiae, mutations in VRG4 result in defects in both N-linked and O-linked protein glycosylation, as well as in synthesis of sphingolipids (Dean, 1999). Some 20–30 % of the mass of filamentous fungal walls consists of proteins, virtually all of which possess mannose-rich oligosaccharides (Bowman & Free, 2006). In Saccharomyces, deletion of some of these, such as Cwp2p (van der Vaart et al., 1995), results in increased sensitivity to wall-compromising agents such as CFW, Congo red and zymolyase. In addition, two plasma membrane proteins playing sensor roles in the yeast cell wall integrity pathway, Mid2p and Wsc1p, are heavily O-glycosylated (Philip & Levin, 2001). The reduction in production of sphingolipids (Dean, 1999) provides a potential direct link between glycosylation and cell polarity, since sphingolipids have been shown to play roles in polarity establishment and maintenance (e.g. Cheng et al., 2001), most likely through their structural roles as components of lipid rafts. Some of the hyphal growth forms observed by Cheng et al. (2001) upon repression of sphingolipid synthesis resemble the hyphal growth forms of the calI11 mutation.

The mutation identified in the calI11 allele is predicted to cause an amino acid substitution within a region that is highly conserved in fungi. This region, which contains the GALNK motif, has been shown to be required for nucleotide sugar binding (Gao et al., 2001). In S. cerevisiae the consensus GALNK motif spans residues 271–292, and the mutation that we identified affects an amino acid lying one residue outside the consensus region at the equivalent of VRG4 amino acid 293. The alanine at this position is conserved in S. cerevisiae, Candida albicans (Nishikawa et al., 2002), Cryptococcus neoformans GMT1 and GMT2 (Cottrell et al., 2007), and A. nidulans GmtB. Several S. cerevisiae glycosylation mutants have been identified whose mutations occur within the GALNK region, resulting in a defect in transport activity (Gao et al., 2001).

In S. cerevisiae, Candida albicans and Pichia pastoris, GMTs reside in the Golgi (Dean et al., 1997; Nishikawa et al., 2002; Losev et al., 2006; Arakawa et al., 2006). The punctate distribution pattern displayed by GmtA–GFP in our studies is consistent with the distribution of proven Golgi markers in yeasts (Dean et al., 1997; Gao & Dean, 2000; Abe et al., 2004), and of putative Golgi markers in A. nidulans (Breakspear et al., 2007; Hubbard & Kaminskyj, 2008). Of particular interest in this regard is our observation of occasional ring-shaped profiles of GmtA–GFP, which were also observed by Breakspear et al. (2007) for the A. nidulans α-COP homologue CopA, and which are consistent with the distinctive appearance of filamentous fungal Golgi-equivalent cisternae in ultrastructural studies (Howard, 1981). In Saccharomyces and other yeasts, localization of VRG4 to the Golgi depends upon an evolutionarily conserved lysine-rich C-terminal motif (Abe et al., 2004). Such a region is found in GmtA, Cryptococcus neoformans GMT1 and GMT2, and S. cerevisiae Hvg1, but not in GmtB according to the gene annotation in Version 3 of the Broad Institute A. nidulans sequenced DNA database (Supplementary Fig. S1). We have not yet been able to GFP-tag GmtB, to determine its subcellular location.

An unexpected result is the failure of GmtA to co-localize with CopA, the A. nidulans homologue of α-COP (Whittaker et al., 1999), which is a component of the seven-subunit coat protein I (COPI) coatomer complex required for retrograde transport between Golgi compartments, and between Golgi and ER in mammalian cells and yeasts (Cosson & Letourneur, 1997). Subunits of the COPI coat have been shown to localize to Golgi membranes in mammals (Griffiths et al., 1995) and Saccharomyces (Morin-Ganet et al., 2000). In S. cerevisiae, COPI is required for steady-state localization of VRG4 to the Golgi, and VRG4 has been shown to physically interact with the Ret2p component of the COPI coat (Abe et al., 2004) in a manner dependent upon the C-terminal lysine-rich targeting motif of the GMT. The unmistakable separation between the principal GmtA and CopA signals in this study does not, of course, preclude the presence of CopA in GmtA-containing compartments; it may simply be that the steady-state level of CopA in those compartments is very low compared with that in compartments nearer to the tip.

In most fungi, Golgi cisternae are widely separated, in contrast to the tightly stacked cisternae of animals, plants and the yeast P. pastoris (Shorter & Warren, 2002; Mogelsvang et al., 2003) In yeasts, evidence indicates that Golgi cisternae undergo a progressive maturation process, exhibiting different physiological properties and protein contents at different stages of development (Morin-Ganet et al., 2000; Losev et al., 2006; Matsuura-Tokita et al., 2006). On the assumption that CopA and GmtA of A. nidulans are indeed Golgi-localized, their non-overlapping distributions can be seen as providing important new evidence that the widely scattered Golgi cisternae of filamentous fungi, like the more ordered cisternae of plants and animals, represent physiologically distinct developmental stages of a single organelle. Recent evidence suggests that Golgi equivalents in A. nidulans exhibit net migration towards the growing tip during hyphal growth, at which point mature cisternae may dissipate into secretory vesicles (Hubbard & Kaminskyj, 2008). Presumably this process would be accompanied by large-scale recycling (retrograde transport) of Golgi components for reuse in newly forming cisternae further back in the cell. In keeping with this model, we propose that the GmtA-containing compartments reported in this study may represent early cis-like cisternae engaged in active biosynthesis, while the more apically located compartments exhibiting high levels of CopA may represent mature trans-like cisternae, whose high CopA content reflects retrograde transport of materials required for continued production of less mature cisternae. As further evidence for a probable trans-like nature of the CopA-containing organelles, Hubbard & Kaminskyj (2008) report colocalization of CopA with a transgenic transmembrane fragment of rat α-2,6-sialyltransferase, which in mammals localizes to trans Golgi cisternae (Roth et al., 1985).

Now that usable Golgi markers have been introduced (this study; Breakspear et al., 2007), combined with temperature-sensitive mutations in Golgi trafficking (Whittaker et al., 1999; Shi et al., 2004; Yang et al., 2008), significant strides can be made in the study of the structure and function of Golgi equivalents in A. nidulans and other filamentous fungi.

This research was supported by a Research Corporation Cottrell Science Award to L. J.-H., by NSF grant CRUI-0211600 to T. W. H. and D. M. L., by a grant to Rhodes College from the Merck Institute for Higher Education, and by funds from Rhodes College in support of undergraduate research. We thank Mary Miller for the use of her fluorescence microscope and Andrew Breakspear for helpful discussions regarding CopA.

Edited by: H. A. B. Wösten