Diversity of GPI-anchored fungal adhesins

PA14-type adhesins

Functionally and structurally, the best-characterized GPI-CWPs are fungal adhesins comprising an exposed PA14-type domain (PFAM entry PF07691). This group is exclusively found in ascomycetes. Functional in vivo and in vitro analysis has been performed on more than 20 different PA14-type GPI-CWPs revealing that their N-terminal A domains are lectins that confer cell-cell and cell-substrate adhesion by specific binding of diverse glycan ligands in a C-type lectin-like manner. In addition, nine different 3D structures of PA14-type domains have been reported so far from the S. cerevisiae flocculins Flo5 and Flo1 (Goossens et al. 2015; Veelders et al. 2010), the C. glabrata epithelial adhesins Epa1, Epa6 and Epa9 (Diderrich et al. 2015; Hoffmann et al. 2020; Ielasi et al. 2012; Maestre-Reyna et al. 2012), two LgFlo1 flocculins from Saccharomyces pastorianus strains (Goossens et al. 2015; Sim et al. 2013) and two Komagataella pastoris PA-14 domains, including the chitin-end binding adhesin Cea1 (Kock et al. 2018). The common features of these adhesion domains are (i) a β-sandwich core structure related to the bacterial PA14 domain from anthrax toxin (Petosa et al. 1997; Rigden et al. 2004), (ii) a calcium-dependent carbohydrate-binding pocket carrying a unique DcisD motif and two calcium binding loops (CBL1, CBL2), and (iii) the ability to recognize the non-reducing disaccharidic ends of glycans present on neighbouring cells or host surfaces.

In the current INTERPRO database (April 2020), PA14-type domains can be found in 791 fungal cell wall proteins. A previous structure-based phylogenetic analysis using a limited number of PA14-type GPI-CWP adhesins has uncovered six different clusters (Kock et al. 2018). We here provide a more comprehensive analysis of all fungal PA14-type adhesion domains by employing the Enzyme Similarity Tool (EFI-EST) (Gerlt et al. 2015), which provides a protein sequence similarity network (SSN) that visualizes broad relationships (Figure 2). The network analysis reveals that fungal PA14-type adhesion domains can be divided into two major groups, depending on the presence of two disulphide bridges, which crosslink the N- and C-terminal ends to the core domain. As can be seen in the middle panel of Figure 2B, Group I is mostly restricted to Saccharomycetes and includes the flocculin-related (Flo-type) cluster, the epithelial adhesin-related (Epa-type) cluster, a cluster comprising several Saccharomycetaceae species like Kluyveromyces marxianus or Kluyveromyces lactis (Kluy-type) and several adhesion proteins from Phaffomycetaceae (Kock et al. 2018). Group II consists of a large cluster of fungal PA14-type domains (CC-less), which lack a conserved motif of two consecutive cysteine residues present in the first group. Members of Group II are generally not located at the N-terminus like the aforementioned groups and often lack the DcisD motif required for calcium binding. The special features of these groups and clusters are discussed in the following paragraphs with a focus on the known physiological roles and structure-function relationships of well-characterized representatives.

Figure 2:

Sequence-similarity network of fungal PA14 adhesins.

(A) Fungal PA14 adhesins are part of the large PA14 domain superfamily (INTERPRO entry IPR0037524; shown as inlay, E-value cut-off 10⁻³⁰); whose 13558 members are mostly of bacterial origin and part of surface proteins or glycan-processing enzymes. 360 of the 795 fungal PA14 adhesin-like gene products (SSN shown with E-value cut-off 10⁻³⁵) belong to the fungal class of Saccharomycetes; the remainder to filamentous fungi and other, which mostly lack the DcisD motif. Currently, all known structures belong to Saccharomycetes and harbour a DcisD motif as required for putative C-lectin like functions. Abbreviations: Cg, Candida glabrata; Sc, Saccharomyces cerevisiae; Sp, Saccharomyces pastoris; Kp, Komagataella phaffii. SSN in Figures 2–4 were generated by EFI-EST using UniProt release 2019_10 (Zallot et al. 2019) and analysed by Cytoscape v3.8.0 (Shannon et al. 2003). (B) Mapping of key structural features to fungal PA14 domains like the presence of an intact DcisD motif (left), the CC motif responsible for disulphide-bridged stabilization of N- and C-termini to the core domain (middle) and N-terminal localization in cell wall protein (right).

Flo-type cluster

The Flo-type cluster currently contains 124 members from 11 species including a multitude of S. cerevisiae strains. The best characterized members of this cluster include the PA14-type flocculins Flo1, Flo5, Flo9 and Flo10 from S. cerevisiae and Lg-Flo1 from S. pastorianus, which all confer calcium-dependent aggregation of thousands of yeast cells into flocs by specific binding of mannoproteins present at the cell surface of (Brückner and Mösch 2012). Floc formation has been studied for decades and is of considerable importance in the brewing and wine-producing industry (Bauer et al. 2010; Van Mulders et al. 2010). In natural environments, the role of these multicellular aggregates is likely to confer protection against harmful agents and conditions (Smukalla et al. 2008). In addition, flocculin genes fulfil the definition of green-beard genes that confer benefits to cells or organisms that carry this allele (Dawkins 1976; Hamilton 1964). In the case of flocculins, self-nonself discrimination is ensured by the simple fact, that two-way bonding between carrier cells both presenting a FloA domain is much more efficient than one-way bonding between carrier cells and cells not expressing the adhesin (Veelders et al. 2010). This difference in binding strength causes cell sorting during flocculation and results in enrichment of non-expressers, which fail to invest into the cost of adhesin production, at the non-protective periphery of flocs and a corresponding selective disadvantage to such cheaters (Smukalla et al. 2008).

So far, crystal structures of the A domains of Flo1, Flo5 and LgFlo1 in complex with different mannosides have revealed the structural basis for ligand binding and discrimination for Flo-type adhesins (Goossens et al. 2015; Ielasi et al. 2012; Sim et al. 2013; Veelders et al. 2010). In addition to the PA14 core, key structural features include (i) a calcium-binding loop CBL1, which carries a highly conserved DcisD motif that is required for efficient binding of Ca²⁺ ions and mannose ligands, (ii) a CBL2 loop harbouring residues that are conferring ligand binding without affecting specificity, (iii) a flexible loop L3 that in certain variants can be used to enhance interactions with the mannose ligand, and (iv) a variable loop L2 that forms a Flo-specific subdomain, which is absent in other PA14-type fungal adhesin domains and confers ligand binding specificity, by e.g. discrimination against glucose and galactose ligands. Biochemical binding studies have revealed that the affinities of FloA-type A domain for mannose are in the low millimolar range, explaining why one-way interactions between carrier and cheater do not lead to efficient floc formation (Veelders et al. 2010). FloA domains are also present in high numbers at the cell surface (Bony et al. 1997), and their high degree of mannosylation, especially along the B-region, leads to efficient homophilic interaction in vitro (Goossens et al. 2015). Thus, cell-cell contacts mediated by Flo-type adhesins might be achieved by both heterophilic as well as homophilic interactions and could involve not only protein-glycan, but also glycan-glycan interactions (Goossens et al. 2015).

With respect to structural and functional diversity, it is not yet fully understood why yeast species harbour entire families of very similar Flo-type adhesins that apparently all confer the same flocculation phenotype. In S. cerevisiae, four FLO genes are able to induce floc formation to comparable degrees (Guo et al. 2000). Here, a recent study suggests that the different FLO genes govern selective association of S. cerevisiae with specific species of non-Saccharomyces yeasts, but the molecular basis for this discrimination is not known (Rossouw et al. 2015). The genome of the yeast Saccharomycopsis fermentans, an ascomycete that is distantly related to S. cerevisiae and is found in spontaneous wine fermentations, even harbours a family of 34 Flo-type adhesins with PA14-type A domains, whose functions have not been studied so far (Bernardi et al. 2018). Thus, detailed analysis of further Flo-type adhesins and A domains is required, to obtain a more complete picture of their structural and functional diversity.

Epa-type cluster

The Epa-type cluster currently contains 37 PA14-type GPI-CWPs from different Candida glabrata strains. C. glabrata is a well-known commensal of humans and some other animals, mostly thriving in mucosal tissues including the gut, but gains increasing attention as a pathogen and its surprisingly high genetic diversity (Gabaldon and Fairhead 2019). As an opportunistic human pathogen C. glabrata is now known to contain more than 80 GPI-CWP genes that are predominantly located in the subtelomeric regions (Timmermans et al. 2018; Xu et al. 2020). The founding member Epa1 was initially discovered as a fungal adhesin that is required and sufficient to confer binding to human epithelial cells (Cormack et al. 1999). Further comprehensive functional characterization on 17 different C. glabrata Epa variants included cell adhesion assays, in vitro binding studies, glycan array screening, single cell atomic force microscopy, mutational analysis and the use of animal infection models. Together, these studies have shown that Epa-type adhesins (i) confer adhesion to human epithelial and macrophage-like cells (Cormack et al. 1999; Kuhn and Vyas 2012), (ii) bind directly to diverse human proteins including mucins, fibronectin and tumour necrosis factor (Ielasi et al. 2016; Ielasi et al. 2014; Zajac et al. 2016), (iii) specifically recognize a wide variety of glycans with terminal α- or β-linked galactosides and non-galactosidic glycans present on human cells and tissues (Diderrich et al. 2015; Ielasi et al. 2016; Maestre-Reyna et al. 2012; Zupancic et al. 2008), (iv) enable efficient binding to clinically relevant plastic materials by conferring strong hydrophobic adhesion forces (El-Kirat-Chatel et al. 2015; Valotteau et al. 2019), and (v) contribute to virulence of C. glabrata in diverse murine infection models (Domergue et al. 2005; Vale-Silva et al. 2016). As such, these studies emphasize a key function of Epa-type adhesins in fungal pathogen-host interactions.

Up to date, high-resolution crystal structures of the adhesion domains of C. glabrata Epa1 (Ielasi et al. 2012; Maestre-Reyna et al. 2012), Epa6 (Diderrich et al. 2015) and Epa9 (Hoffmann et al. 2020) in complex with diverse glycan ligands have been solved. These studies show that Epa-type domains, in addition to the PA14 core, possess further peripheral structural elements for ligand binding and discrimination. These structural hot spots include two calcium binding loops, CBL1 and CBL2, which form the inner ligand binding pocket, and three further loops, L1-L3, that form the outer pocket. While the surface composition of different EpaA domains is mostly variable, a number of residues within the ligand binding pockets are highly conserved. These include a DcisD motif in CBL1 and an asparagine in CBL2, which confer coordination to a Ca²⁺ ion and together are referred to as the DD-N signature, and a tryptophan residue in loop L3, which confers high affinity ligand binding in the low micromolar range and together with a highly conserved arginine in CBL2 forms the W-R corner signature (Diderrich et al. 2015; Maestre-Reyna et al. 2012). Moreover, C. glabrata Epa-type ligand binding pockets carry several highly variable residues at three positions of CBL2 and within the loops L1 and L2. The sequence of the variable positions of CBL2 often correlates with the observed glycan ligand binding patterns of different Epa variants, suggesting that this motif confers ligand binding specificity, a conclusion that is supported by mutational analysis of the Epa1 CBL2 motif (Ielasi et al. 2014; Maestre-Reyna et al. 2012). In certain cases, however, CBL2 composition and ligand binding patterns do not correlate, indicating that other variable elements, e.g. loops L1 and L2, contribute to ligand recognition (Diderrich et al. 2015). Indeed, a comprehensive structure-based mutational analysis using six different C. glabrata Epa paralogs shows that the variable CBL2 loop is key for programming ligand binding specificity and that the variable loop L1 affects host cell binding, but not ligand binding patterns (Hoffmann et al. 2020).

The structural and functional studies on the C. glabrata Epa family indicate that variation of specific structural hot spots in the ligand binding pockets of Epa-type domains is a key driver for their diversification and evolution. This view is supported by comparative sequence analysis, which reveals that the DD-N calcium-binding signature is commonly present in 80 % of PA14 domains with conserved DcisD motif (299/376), but that the W-R signature (3%, 13/376) is mostly restricted to variants of the glabrata group of Nakaseomyces (Diderrich et al. 2015; Gabaldon et al. 2013). Moreover, nine of the functionally well-studied members of C. glabrata Epa variants carry unique sequence motifs at the positions II-IV in the variable CBL2 loop, which significantly correlate with glycan binding specificities and are generally not found in the CBL2 regions of the other known Epa-type domains that carry the DD-N and W-R signatures (Diderrich et al. 2015). This supports the view that specific adaptation of CBL2 motifs has evolved after the W-R signatures and significantly contributes to the diverse host specificities observed for different Nakaseomyces species (Bolotin-Fukuhara and Fairhead 2014).

Pwp-type cluster

The Pwp-type cluster represents another C. glabrata specific group of PA14-type GPI-CWPs. So far, functional analysis has been performed on members of the C. glabrata Pwp family consisting of seven paralogs, which in contrast to Epa-type adhesins are mostly located in centromeric regions (de Groot et al. 2008). Whereas deletion of the PWP7 gene has been reported to result in a significant reduction in adherence to human endothelial cells (Desai et al. 2011), expression of Pwp7 and most other C. glabrata Pwp paralogs in an S. cerevisiae system does not mediate adhesion to a variety of six different human endothelial or epithelial cells (Hwang-Wong 2016). In addition, a genome-wide expression study using C. glabrata axenic cultures and a number of different stress conditions shows that in comparison to housekeeping genes, the general expression levels of PWP genes are low and comparable to that of EPA genes (Linde et al. 2015). Thus, both the physiological roles of Pwp-type adhesins and the precise conditions inducing their expression, e.g. in the host, remain to be determined.

No structural or ligand binding studies are available for Pwp-type adhesins so far. Nonetheless, comparative sequence analysis indicates that Pwp-type domains might act as lectins much like other PA14-type adhesion domains, because they contain a PA14 core, CBL1 and CBL2 loops that contain the DD-N signature, and the loops L1 – L3. Like Epa-type members, they lack the Flo-type specific subdomain in loop L2, but unlike the Epa-type also the W-R signature. Another characteristic of Pwp-type domains is the lack of the CC-motif for disulphide bond formation with the N- and C-termini. Accordingly, this sets Pwp-type adhesins aside from so far studied PA14-adhesins, as they belong to Group II of PA14-dependent adhesins (Figure 2B), which otherwise are GPI-CWPs found in filamentous fungi, some in the family of Saccharomycodaceae and a group of Komagataella species. Clearly, future structural and functional studies are needed to determine the biochemical functions of Pwp-type domains such as the binding of specific glycan ligands.

Kluyveromyces cluster

As for the Pwp-cluster, no member of the Kluyveromyces cluster with its 44 PA14-type GPI-CWPs from 14 species has been structurally or biochemically characterised yet. Given their potential for biotechnology (Spohner et al. 2016) it is interesting to note that K. lactis and K. marxianus genomes encode for 6 and 8 paralogs, respectively, of PA14-dependent adhesins. All adhesins of K. marxianus have a (Y,F)-R signature in the putative carbohydrate binding site and may hence recognize similar ligands as members of the Epa-type. For the probiotic strain K. marxianus B0399, adhesion to colonic Caco-2 cells has been observed (Maccaferri et al. 2012), a characteristic that is exerted by some Epa-type adhesins like Epa1, Epa6 and Epa7, but not others (Diderrich et al. 2015; Maestre-Reyna et al. 2012).

PA14-adhesins from Phaffomycetaceae

PA14-type GPI-CWPs from Phaffomycetaceae co-cluster like those from Metschnikowiaceae with PA14-type adhesins from filamentous fungi and are hence distinct from the Saccharomycetal adhesins described so far (Figure 2A). To date, two paralogs originating from the methylotrophic yeast Komagataella pastoris, the chitin-end adhesin Cea1/KpFlo1 as the founding member and the orphan lectin KpFlo2, have been functionally and structurally characterized (Kock et al. 2018). Functional analysis employing high-throughput glycan array screening, in vitro glycan binding assays and an S. cerevisiae expression system has revealed that Cea1 specifically binds to diverse terminal β-linked N-acetylglucosamines in the low micromolar range in vitro and confers efficient binding to non-reducing ends of chitin when presented at the surface of yeast cells. For KpFlo2, no specific glycan binding was found. Thus, Cea1 represents the first fungal cell wall adhesin that strictly recognizes β-capped N-acetylglucosamine of chitinous polymers. It has therefore been proposed that Cea1 might be crucial for interaction of Komagataella species with insects (Kock et al. 2018), a hypothesis that is supported by the fact that K. pastoris was initially discovered in close association with the fruit fly Drosophila melanogaster (Shihata and Mrak 1952).

Crystal structures were solved for Cea1 in complex with N-acetylglucosamine and N,N′-diacetylchitobiose, respectively, and for KpFlo2 (Kock et al. 2018). They reveal that both domains not only consist of the β-sandwich PA14 core, but also contain the loops CBL1, CBL2, L1, L2 and L3. They also contain the DD-N signature present in Flo-type and Epa-type domains, which confers coordination to a Ca²⁺ ion and to the glycan ligands. In contrast, they lack the Flo-type specific subdomain in loop L2 and the Epa-type specific W-R signature. Two unique structural features of Cea1 are its unusually narrow binding site for terminal N-acetylglucosamine moieties and the presence of a lysine residue in loop L3 instead of the tryptophan of the W-R motif that enables a high number of ligand interactions, which explains the low micromolar binding affinity in vitro.

In K. pastoris and Komagataella phaffii, a third paralogous PA14-adhesin, KpFlo3, has been found that shares Ca²⁺-binding via the DcisD-N motif but that has not been further characterized (Kock et al. 2018). In addition to these Group I PA14-adhesins, Phaffomycetaceae comprise an additional set of Group II PA14-adhesins (Figure 2A, B), which miss not only the CC-motif, but also an intact calcium-binding site due to replacement of the second aspartate of the DcisD-motif by tyrosine. Like other Group II PA14-adhesins, they are mostly located close to the C-terminal ends of cell-wall resident adhesins, but their function is still unknown. Given the lack of typical C-type lectin signatures, the biological role of these Group II PA14-type adhesins may hence clearly differ from bona fide carbohydrate-binding adhesins.

Despite their deviations from other saccharomycetal PA14-type adhesin clusters, Cea1-like adhesins share high-affinity ligand binding and the potential for heterotypic cell-cell or cell-substrate interactions with Epa-type domains. The observation of an orphan lectin like KpFlo2 is reminiscent of many Epa-type adhesins, whose cognate (carbohydrate) ligands could not yet be unambiguously assigned (Diderrich et al. 2015; Kock et al. 2018). Clearly, detailed structure-function relationships of additional Komagataella PA14-domains must further define the general properties of this group of putative adhesins.

Further group II PA14-type adhesins

About two third of all putative PA14-type adhesins (375/564) lack the CC-motif required for disulphide-bridged stabilization of the adhesion domain and formation of a rigidified stalk region for connection to subsequent B-region repeats (Figure 2B). Despite their occurrence in many important animal and plant pathogens such as C. glabrata or diverse Fusarium or Trichoderma species, there are no functional or structural data available for this type of cell wall adhesins. Unlike found for Flo-, Epa- or Pwp-type adhesins, 71% (267/375) of these Group II PA14-type adhesion domains are not located at the N-termini of the gene products, but in contrast to the Group II members from Phaffomycetaceae possess a functional DcisD motif (Figure 2B). Accordingly, these cell wall proteins may mediate not only adhesion, but also intra-cell wall cross-linking with endogenous glycans.

Flo11-type adhesins

Fungal adhesins that carry an N-terminal Flo11-type domain (PFAM entry PF10182) are encoded by currently 251 known genes from 86 fungal species, which belong to 9 families in the ascomycetal orders of Saccharomycetales and Schizosaccharomycetales (Figure 3A). The best-characterized Flo11-type adhesin is Flo11 from S. cerevisiae (ScFlo11), which was originally discovered as a cell surface flocculin required for substrate adhesion and the formation of invasive filaments on agar surfaces (Lo and Dranginis 1996, 1998). Further genetic studies revealed that ScFlo11 confers the formation of other multicellular growth forms including structured mats on viscous surfaces (Reynolds and Fink 2001), floating flors on liquid cultures (Fidalgo et al. 2006; Ishigami et al. 2004; Zara et al. 2005), and adhesive biofilms on solid plastic surfaces (Reynolds and Fink 2001). Ultrastructural analysis employing electron microscopy suggests that Flo11 adhesins connect cells by forming an extracellular matrix that involves co-aligned Flo11 fibres (Kraushaar et al. 2015; Vachova et al. 2011). Further biochemical studies with purified ScFlo11A domains and biophysical studies employing AFM-based single cell force spectroscopy have shown that ScFlo11 mediates cell-cell adhesion by homotypic protein-protein interactions (Douglas et al. 2007; Kraushaar et al. 2015), which result in remarkably strong adhesion forces of more than 10 nN between interacting cells that must both express the adhesin (Brückner et al. 2020). These studies have demonstrated a key function of Flo11-type domains in conferring cell-cell and cell-substrate adhesion during multicellular development.

Figure 3:

The class of Flo11-like fungal adhesins.

Left, the SSN of the Flo11 domain protein family (IPR018789, E-value cut-off 10⁻¹⁵) shows that family members are restricted to both the orders of Saccharomycetales and Schizosaccharomycetales. The two structures from S. cerevisiae (red) and K. pastoris (green) show two surface-exposed aromatic bands (turquoise), which have been claimed to mediate interactions with hydrophobic surfaces as well as with other Flo11-like adhesin domains (right). Sc, Saccharomyces cerevisiae; Kp, Komagataella pastoris.

Investigation of FLO11 adhesin genes in yeasts other than S. cerevisiae has not been performed yet. However, expression of a number of Flo11A domains from diverse Saccharomycetales species in an S. cerevisiae model has shown that the capacity of these adhesins for conferring cell-cell adhesion through homotypic interactions is highly conserved (Brückner et al. 2020). These studies have further shown that Flo11-type adhesins can act as cell surface receptors, which mediate kin discrimination in mixed S. cerevisiae populations that express variants from different species or sub-species (Brückner et al. 2020; Oppler et al. 2019). In addition, Flo11 adhesins confer competitive advantages through cell differentiation in clonal biofilms (Regenberg et al. 2016). Thus, the physiological functions of Flo11-type adhesins are key not only for the formation of diverse multicellular growth forms that confer environmental protection and promote substrate colonization, but also for preventing the invasion of groups of closely related cooperating individuals by more distantly related species or sub-species.

High-resolution crystal structures of Flo11-type domains have been solved for ScFlo11A (Kraushaar et al. 2015) and for KpFlo11A (Brückner et al. 2020), a variant from the methylotrophic yeast K. pastoris of the family of Phaffomycetaceae. Although the two variants share only 32% identity, their overall structures show a high degree of similarity. They both consist of a fibronectin type III (FN3) core, a fold that forms a large family within the immunoglobulin (Ig) superfamily. In addition, both orthologs expose an unusually large number of conserved aromatic residues at the protein surface, which are arranged in two bands that have a similar but not identical configuration. Systematic mutational analysis of surface-exposed aromatic residues of ScFlo11A has shown that the aromatic bands are essential for efficient protein-protein interactions in vitro as well as for cell-cell adhesion at single cell level and in populations in vivo (Brückner et al. 2020)(Figure 3B). Comparison of homotypic and heterotypic interactions of ScFlo11 and KpFlo11 further revealed that heterotypic protein-protein binding as well as cell-cell adhesion forces are significantly reduced when compared to the respective homotypic interactions. It has therefore been proposed that discrimination between Flo11-type domains from different yeast species at the molecular level might be conferred by inefficient interaction of their differentially arranged aromatic bands.

Comparative analysis of 52 Flo11A domains from different S. cerevisiae strains has revealed the presence of a small insert of about 15 amino acids in a number of variants, that are exclusively found in the ’sake’ lineage of non-mosaic strains (Brückner et al. 2020; Liti et al. 2009). Remarkably, this insert is sufficient to confer sub-species discrimination at single cell level and in competing S. cerevisiae populations and might have originated from a single genetic event. As such, Flo11-type domains have become valuable models for investigating how the structure of microbial cell surface adhesins can evolve to permit diversification at species and sub-species levels without compromising their core functions.

Als-type adhesins

Fungal adhesins carrying an N-terminal agglutinin-like sequence (Als) region (Pfam entry PF11766) are encoded by currently 251 known genes from 83 fungal species, which are restricted to the ascomycetal classes of Saccharomycetes (74) and a few Sordariomycetes (9). This type of adhesins (Figure 4A) is a minor part of the very large adhesion domain superfamily (INTERPRO entry IPR008966) with its huge variety of bacterial and eukaryotic lectins, collagen-binding proteins and fimbrial adhesins.

Figure 4:

The class of Als-like fungal adhesins.

(A) The Als-like fungal adhesins are part of the very large adhesin domain superfamily (ADS; IPR008966, SSN shown as inlay with E-value cut-off 10⁻¹⁵), which covers all domains of life. The two Ig-domains of Als proteins from Candida albicans (Ca) bind ligands along the interface region (B) The Venn-diagram shows the co-occurrence of Als-, Flo11- and PA14-like adhesin domains in 255 fungal species as based on analysis of the corresponding INTERPRO families IPR08966, IPR018789 and IPR037524. The numbers in parentheses correspond to the number of species found for each adhesin family.

The functionally best-studied adhesins in this group are found in the most frequently isolated human fungal pathogen C. albicans, which contains the Als family encompassing eight members (Als1 to Als7, and Als9) (Hoyer 2001). The founding member ALS1 was initially discovered as a gene that is induced during hyphal development and encodes a protein with similarity to the S. cerevisiae mating alpha-agglutinin (SAG1) that belongs to the family of Als-type adhesins (SAG1 group, Figure 4A) as well and confers cell type-specific cell-cell adhesion (Hoyer et al. 1995). Since then, a plethora of functions have been described for Als family members, predominantly for Als1 and Als3, which have been reviewed in great detail (Hoyer and Cota 2016). Among others, these functions include (i) adhesion to human endothelial and epithelial cells, (ii) binding to fibronectin, fibrinogen, laminin, collagen, cadherins and further human proteins with affinities in the low micromolar to millimolar range, (iii) binding to human glycans containing fucose or N-acetyl-glucosamine with micromolar affinities, (iv) adherence to diverse abiotic plastic and glass surfaces, and (v) enhancement of C. albicans virulence in animal infection models. These functions, much like that of C. glabrata Epa adhesins, emphasize the key role of C. albicans Als proteins in host-pathogen interactions. In addition, Als1 and Als3 confer binding to bacterial pathogens such as Streptococcus gordonii and Staphylococcus aureus, indicating a role during the formation of clinically relevant microbial communities (Peters et al. 2012; Silverman et al. 2010).

High resolution structures are available for the N-terminal adhesion domains of Als9 in complex with a peptide from the C-terminal end of human fibrinogen gamma (Salgado et al. 2011) and for Als3 in complex with hepta-threonine (Lin et al. 2014). These structures provide important functional insights and show that Als domains comprise two tandem immunoglobulin G (IgG) domains (N1 and N2), which resemble MSCRAMM domains found in fibrinogen-binding adhesins from the bacterial pathogens S. aureus and SdrG from Staphylococcus epidermidis. The region between N1 and N2 forms a peptide-binding cavity (PBC), which buries up to six C-terminal residues of polypeptides and enables high-affinity ligand binding by establishing a salt-bridge between an invariant lysine at the end of the PBC and the C-terminal carboxylate of variable peptide ligands. The PBC further interacts with the backbone of peptides by forming several protein-peptide interactions through hydrogen bonding and a variable network of water molecules that contribute to the recognition of a broad range of diverse peptide sequences. The structural data available for Als-type domains so far clearly support a main function as peptide binding adhesins (Cota and Hoyer 2015). Whether Als-type domains can also act as lectins, which mediate specific binding to glycan ligands, is currently not supported by structural analysis.

Als-type fungal adhesins have also been studied with respect to conferring homotypic interactions. Purified C. albicans Als-type domains consisting of IgG domains N1 and N2 are unable to bind to each other with high affinity (Lin et al. 2014), as has been found for Flo11-type domains. In full-length Als proteins, however, Als-type domains are neighboured by amyloid forming regions (AFR), which have been demonstrated to confer homotypic interactions (Lipke et al. 2012). Moreover, AFR regions can interact with the N1 domain in a peptide ligand-dependent manner, which prevents homotypic AFR interactions. It has therefore been proposed that in the absence of suitable ligands, Als adhesins are able to confer adhesion between C. albicans cells though amyloid formation of AFRs. This mechanism could enable C. albicans to form multicellular aggregates such as protective biofilms that are less susceptible to host defence or antifungal drugs (Lin et al. 2014).

Detailed functional analysis of Als-type adhesins from species other than C. albicans has not been performed yet, but comparative genomics reveals the presence of numerous ALS genes in the Candida parapsilosis complex including the closely related C. parapsilosis, Candida orthopsilosis and Candida metapsilosis species (Lombardi et al. 2019; Oh et al. 2019). While the predicted Als proteins harbour most common features of the C. albicans Als family, including the presence of PBC and AFRs in the N-terminal regions, they also have certain unique features. Molecular modeling, for instance, reveals large differences in shape and charge of different PBCs, suggesting functional diversity with regard to the preferred peptide ligands. Clearly, future biochemical and structural analysis of a wide variety of diverse Als-type domains will contribute to a deeper understanding of their molecular and functional evolution.

Occurrences and other potential fungal adhesion domains

Of the three major GPI-adhesin types, which have been functionally and structurally characterized in detail, the PA14-type sticks out due to its wide occurrence in Ascomycetes. Unlike the other two types, i.e. Flo11- and Als-type, PA14-type adhesins occur in about three quarters of ascomycetal species without accompanying Flo11- or Als-adhesins (Figure 4B). However, this does not necessarily indicate that the largest determinant of fungal adhesion is binding to glycan structures of host surfaces. For example, Group II PA-14 adhesins lack key features required for C-type lectin function. Additionally, many Group I PA-14 adhesins with intact DcisD motif characterized so far had to be assigned as orphan lectins given the lack of detectable ligands. The delineation of the binding specificities of these adhesins may be highly relevant to understand fungal biology, e.g. in terms of host specificity. In contrast, Flo11-type adhesins as well as Als-type adhesins apparently provide additional, but not necessarily essential adherence capabilities to fungal species, because they are associated with other adhesin types in the case of 81 and 75%, respectively, of ascomycetal species (Figure 4B). Moreover, Flo11 has been claimed to act like a cell-wall tethered hydrophobin (Kraushaar et al. 2015) for mediating interactions to hydrophobic surfaces including those of other yeast cells, rather than a ligand-specific adhesin, whereas the Als-type apparently exhibits considerable specificity towards C-terminal peptide ends.

Several putative domains other than the three major GPI-adhesin types have been found to mediate adhesive function in several fungal species. De Groot et al. discovered additional adhesin-like wall proteins (Awps) in C. glabrata CBS138, besides its seven Pwp paralogs (de Groot et al. 2013; de Groot et al. 2008), which currently amount to 13 members. However, the N-terminal domains of Awp proteins themselves belong to different protein families. Awp1 and Awp3 are for instance related to fungal haze-protective-factor proteins, whereas the other Awp members belong to at least three different subfamilies (de Groot et al. 2013). Interestingly, these adhesin candidates are found among cell proteins that are up-regulated during host colonization by clinical C. glabrata strains on proteome, transcriptome and genome levels (Carrete et al. 2019; Gomez-Molero et al. 2015; Kraneveld et al. 2011). Another example is the MAD1-type of adhesins, which utilize cysteine-rich fungal extracellular module (CFEM) domains (Li et al. 2016; Wang and St Leger 2007) and apparently contribute to fungal adhesion to nematodes and insects. These and many other GPI-CWP with assigned adhesion function await further biochemical, functional and structural studies to understand their role in controlling different fungal life styles.

GPI-CWP adhesins – to be sorted or not to be sorted into the cell wall?

One unresolved issue in the context of GPI-anchored adhesins is how sorting into the cell wall occurs, as all GPI-anchored proteins initially end up at the plasma membrane (Figure 5). Some of these proteins remain there as GPI-plasma membrane proteins (GPI-PMPs), whereas others such as adhesins are transferred onto cell wall glucans as GPI-CWPs. This division is not absolutely strict, as several GPI-anchored proteins can end up in both locations (Pittet and Conzelmann 2007). In S. cerevisiae, the two paralogous transglycosidases, Dfg5 and Dcw1, which mediate membrane-to-cell wall transfer of GPI-CWPs, are themselves produced as GPI-PMPs. Here, their GPI anchorage is sufficiently stable to maintain their membrane residence for allowing Dfg5-catalyzed transfer of GPI-CWPs to cell wall glycans.

Figure 5:

Ethanolamine-phosphate modifications control GPI-based cell wall-incorporation by Dfg5-enzymes.

After GPI-precursor biosynthesis, the GPI-transamidase complex transfers target proteins onto the EtN-P of mannose-3, upon which Bst1, Per1, and Gup1 remodel the GPI-lipids. Depending on the destination, either Cdc1 removes the EtN-P from mannose-1, or Cwh43 catalyses lipid exchange from PI to IPC. Thereby, the EtN-P code for the final destination of GPI-anchored proteins, i.e. GPI-CWP vs. GPI-PMP, is generated. Subsequently, Ted1 removes EtN-P from mannose-2 to facilitate ER-exit. After further processing in the Golgi, Dfg5-transamidases decode the presence of EtN-P modifications within the GPI-core and select the correctly processed subset of GPI-anchored proteins for final maturation as GPI-CWPs.

In yeast, temperature-sensitive mutants of the cell-division cycle gene CDC1 were shown to stop growth in early G2 phase with small buds and replicated DNA, a phenomenon that was also found in mutants of DCW1 (Kitagaki et al. 2002, 2004). Cdc1 is a phosphodiesterase that removes the EtN-P added to mannose-1 of the GPI core glycan by Mcd4 during GPI-anchor biosynthesis and can be deleted in a mcd4∆ background without loss of viability (Vazquez et al. 2014). Another phosphodiesterase, Ted1, removes the EtN-P from mannose-2, and acts together with Emp24 and Erv25 at the ER exit sites (ERES) for facilitating efficient cargo exit (Haass et al. 2007). These modifications of the GPI-core glycan may disrupt binding of the GPI-anchor by Dfg5 enzymes, thus preventing GPI-CWP transfer onto cell wall glycans. This hypothesis explains the comparable phenotypes of cdc1 and dfg5 dcw1 mutants, which both exhibit growth arrest and a compromised lipid-to-wall transfer of GPI-CWPs.

In contrast to Cdc1, removal of EtN-P from mannose-2 by Ted1 is not essential, at least in S. cerevisiae, though its deletion delays ER-export of GPI-anchored proteins due to its interaction with the p24-complex during COPII trafficking (Manzano-Lopez et al. 2015). TED1 deletion mutants hence resemble deletion mutants of the p24-complex component encoding genes EMP24 and ERV25, which both show the same ER-exit defects (Belden and Barlowe 1996; Schimmoller et al. 1995). The consequences for membrane-to-cell wall transfer by Dfg5 enzymes in ted1∆ cells is hence not as fatal as that of the cdc1∆ mutants, because the 6-hydroxy group of mannose-2 in GPI-anchors may not be recognized by these enzymes. Accordingly, ted1∆ mutants exert a growth behaviour in rich medium as observed for the wild type (Yoko et al. 2018).

Cwh43 mediates a key step of GPI-lipid remodelling, namely transfer of the GPI-anchored protein precursor from phosphatidylinositol (PI) to inositol-phosphorylceramide (IPC). Both EtN-Ps of mannose-1 and -2 are required for ceramide-remodelling by Cwh43 (Benachour et al. 1999; Zhu et al. 2006). While the GPI-CWP Cwp2 possesses a PI in wild type cells, GPI-PMPs such as Gas1 harbours an IPC lipid. In cwh43∆ mutants, Gas1 is anchored instead by PI and preferably localized to the cell wall. In yeast, two models have been proposed to explain when and how Cwh43 acts on lipid remodelling. The sequential model suggests a main pathway, where Cwh43 utilizes a PI-containing diacylglycerol after Gup1. The divergent model proposes an alternative route utilizing lysoPI for ceramide remodelling and thereby separating the reactions of Gip1 and Cwh43 (Ghugtyal et al. 2007; Umemura et al. 2003). However, at least in the filamentous fungus Aspergillus fumigatus such an alternative route could be excluded (Li et al. 2018).

Apparently, the fate of GPI-CWPs and GPI-PMPs is already encoded during the maturation of the GPI-anchor in the secretory pathway, while Dfg5 enzymes decode the previous decision at the plasma membrane (Vogt et al. 2020). A sequence of events can be suggested in the ER of yeasts (and probably other fungi possessing two paralogs of the mammalian PGAP5, i.e. Cdc1 and Ted1), upon which the destination of GPI-anchored proteins is determined (Figure 5). Initially remodelling of the lipid occurs in the first three steps of the sequential pathway by Bst1, Per1, and Gup1 directly after the GPI-transamidase complex adds the target protein to the GPI-precursor (Kinoshita and Fujita 2016). GPI-lipid remodelling, but not GPI-glycan remodelling, is the crucial step for concentration and pre-sorting at ERESs (Manzano-Lopez et al. 2015). The next step determines the destination, because designated GPI-CWPs are recognized by Cdc1 (Figure 5, upper route), which not only removes the EtN-P at Man1, but also prevents ceramide transfer initiated by Cwh43 (Vazquez et al. 2014). Subsequently, Ted1 removes EtN-P from mannose-2, which results in the mature ER-form of the GPI-glycan and facilitates efficient COPII-vesicular traffic to the Golgi as initiated by the p24-complex (Haass et al. 2007; Manzano-Lopez et al. 2015). Upon export to the plasma membrane, the glycan of GPI-anchored proteins containing a PI-lipid are recognized by Dfg5-proteins and further transferred to cell wall glycans. GPI-PMPs, however, are recognized by Cwh43 (Figure 5, lower route), which remodels the lipid from PI to IPC. As the preferred substrate of Cdc1 are GPI-anchors with a diacylglycerol lipid this step is skipped and the GPI-ceramide is directly handed over to Ted1 (Vazquez et al. 2014) for removal of the EtN-P from mannose-2 and p24-mediated cargo export to the Golgi. This results in a secreted GPI-anchored protein possessing an IPC and an EtN-P at mannose-1. This modification prohibits recognition of Dfg5-enzymes and leaves GPI-PMP in the plasma membrane (Vogt et al. 2020). As animals lack GPI-CWPs, this might also explain why they possess a single PGAP5-homolog for GPI-glycan processing for ensuring efficient ER to Golgi transport by removing EtN-P from mannose-2, but leaving EtN-P at mannose-1.