Several components of the HI reaction encoded by the het genes, the HI suppressor genes, and the HI-induced genes have been characterized at the molecular level in P. anserina and N. crassa. In natural populations of P. anserina, N. crassa and Cryphonectria parasitica, the functional het genes exist as two or more polymorphic alleles conferring alternative specificities 19202122. Different loci may be responsible for triggering the HI reaction in these two and other fungal species. Heterocomplex formation between alternative het gene products is thought be a common theme in nonself recognition during allelic and non-allelic HI 2324. One of the genetically (and potentially physically) interacting partners is often a HET domain protein 15 that may interact with well-conserved proteins playing important roles in development and differentiation (Dementhon and L. Glass, unpublished).

Although at least eight loci have been implicated in HI in A. nidulans 25, none of them has been characterized at the molecular level. To identify putative HI-associated proteins in the Aspergilli, we first searched completely sequenced fungal genomes using known inducers and mediators of heterokaryon incompatibility as BLASTp queries. We examined the domain composition and phyletic distribution of the BLASTp hits and built phylogenetic trees for several protein families. We also applied domain fusion analysis to several so called Rosetta Stone proteins with unusual domain composition to infer protein domain interactions and functional linkage between putative HI-associated proteins identified in the Aspergilli proteomes.

Our database BLASTp searches identified orthologs of P. anserina HET-C2 26, in all filamentous fungal genomes sequenced to date (Table 1). The family also has a high level of sequence conservation and wide phyletic distribution in other taxa. HET-C2 orthologs are found in several Saccharomycetes, including Debaryomyces hansenii and Kluyveromyces lactis, (Fig. 1). Yet no homologs are detected in S. cerevisiae and S. pombe, suggesting a gene loss in some yeast lineages. HET-C2 homologs are also present in most animals and plants.

The high level of conservation among the HET-C2 family members is consistent with the important role these proteins may play in the glycosphingolipid and sphingosine metabolism and possibly in regulation of cellular stress responses. HET-C2 shows significant similarity to human GLTP 27 and Arabidopsis thaliana ACD11 28 proteins, which catalyze the intermembrane transfer of glycosphingolipids and sphingosines, respectively. ACD11 has also been shown to function in PCD and pathogen defense in plants. In Aspergilli, sphingosines have been also shown to induce an apoptosis-like PCD 4 and to affect cell cycle progression 29. P. anserina HET-C2 was proposed to act as a glycolipid metabolite sensor in addition to its role in glycolipid transfer, regulation of ascospore maturation, and triggering HI 2630. The high level of sequence conservation in this family, suggests that the role of HET-C2 orthologs in the Aspergilli PCD pathway is likely to be similar.

Further analysis showed that all Aspergillus species have direct orthologs of N. crassa HET-C 31 and its close homolog of unknown function (Table 1). The phylogenetic tree has a bipartite division resulting from an early gene duplication event predating the separation of Eurotiomycetes and Sordariomycetes (Fig. 2). HET-C is orthologous to A. nidulans HetC and homologous to TinC 32. The HET-C domain is found in many ascomycetes and basidiomycete species, but surprisingly in only one yeast species, Yarrowia lipolytica, which clusters together with HET-C homologs from basidiomycete species. Unexpectedly, a partial HET-C domain is also present in several epiphytic and symbiotic bacteria including two gamma-proteobacteria, Pseudomonas syringae [GenBank:

AAY39263

AAO58004

ZP_00106220

ZP_00567177

Based on the current tree topology, the origin of the bacterial homologs is not clear. It can be attributed to vertical inheritance from the last common ancestor between bacteria and fungi followed by massive gene loss in most bacterial and yeast lineages. Alternatively, it can be explained by horizontal transfer of the ancestral het-C gene from epiphytic fungi followed by rapid divergence in bacteria. In the latter case, the gene must have persisted in bacterial populations by conferring a selective advantage to the recipients. Since heterologous expression of an N. crassa het-C allele was also shown to trigger an HI-like growth defect in P. anserina 21, the het-C homologs in P. syringae or related species may induce growth inhibition in epiphytic filamentous fungi and thus facilitate substrate defense.

N. crassa polypeptides encoded by the het-C alleles of alternative specificity were shown to form a heterocomplex localized to the plasma membrane during the HI reaction 23. It has a putative signal peptide, a conserved HET-C domain and a divergent C-terminal glycine-rich region, often found in extracellular glycoproteins. The biological role of the HetC proteins in the Aspergilli is unknown. A. nidulans TinC has been shown to stabilize the NimA mitotic kinase required for mitotic entry 32. A. nidulans strains lacking tinC displayed cold and osmotic sensitivity and overexpression of its truncated form produced growth inhibition, defects in nuclear envelope fission and cell cycle 32. It is unlikely that either protein triggers the HI reaction in the A. nidulans. Moreover, het-C may not act as a bona fide het gene in other fungal species, since no het-C polymorphism was observed in A. flavus (K. Ehrlich and P. Cotty, unpublished), A. nidulans 32, and P. anserina 21 isolates. Nonetheless, expression of the N. crassa het-c(PA) allele triggers a growth inhibition similar to the HI reaction in P. anserina. If N. crassa HET-C has a similar role to TinC, it may explain the growth inhibition effects caused by expression of N. crassa (and possibly bacterial) het-c in P. anserina 21.

HET domain 1533 proteins were identified using the HMMer package as described in Methods. Unlike the ubiquitous HET-C2 family, the HET domain appears to be limited to filamentous ascomycetes and is not detected in yeasts or basidiomycete species (Table 1). In the Aspergilli, the number of HET domain proteins varies from seven in N. fischeri to 38 in A. oryzae. The tree topology delineates multiple duplication events in filamentous ascomycetes species followed by rapid diversification and gene loss in several Aspergillus spp. (data not shown). Orthologous relationships within this Aspergillus family are difficult to establish, except for a subfamily of HET and Ankyrin domain proteins, which appear to be related by direct vertical descent (data not shown).

The HET domain expansion in filamentous ascomycetes may represent a niche adaptation strategy to process a large number of similar stimuli associated with defense against pathogens, self/nonself recognition, differentiation, or analogous roles. It is found in N. crassa HET-6 and TOL and in P. anserina HET-D and HET-E, and so appears to be critical to the HI reaction in both species (for review see 15). In P. anserina HET-D and HET-E, HET domains are followed by a NACHT domain and multiple WD repeats, while N. crassa proteins contain a coiled-coil domain and LRR repeats, instead (see Figure 3). In addition to HET-6 and TOL, N. crassa has about 50 other HET domain proteins, whose role in the HI reaction if any is as yet unknown.

Initial BLASTp searches using the P. anserina HET-s sequence 34 as a query revealed homologs in A. nidulans, P. chrysogenum, M. grisea, N. crassa and G. zeae (Table 1). Iterative PSI-BLAST searches identified a new domain that includes more proteins from the same species plus a pathogenicity protein, LopB, from the Dothideomycete fungus Leptosphaeria maculans 35. For LopB and most other members of this family, sequence similarity is limited to the N-terminal globular domain of HET-s (Fig. 4) 36. Two members from A. nidulans and N. crassa have an adjacent NACHT domain (described below) at the N- and C-terminus, respectively.

As mentioned earlier, HI was proposed to act as a self/nonself recognition system responsible for limiting the spread of numerous infectious elements in natural populations 151617. Coincidentally, HET-s prion behaves as a non-conventional infectious element capable of propagation during anastomosis and sexual reproduction in P. anserina 37. HET-s can exist in two forms: as a normal protein 18 and as an infectious prion 38, capable of propagating as a self-perpetuating amyloid aggregate 1836. Its rather unexpected similarity to LopB implies that members of the family may have another function unrelated to HI. Although its specific role in L. maculans is unknown, LopB^- mutants showed impaired ability to form lesions on oilseed rape 35. LopB contains a predicted signal peptide suggesting that it is secreted and might contribute to the L. maculans pathogenicity by compromising host membranes. The fusions between HET-s/LopB and NACHT domain in N. crassa and A. nidulans suggests that, in other species, proteins containing one of these two domains may physically interact.

Using P. anserina HET-E as a BLASTp query, we identified several proteins containing NACHT domain in the Aspergilli and other filamentous ascomycetes. Further HMMer searches detected two Aspergillus-specific expansions of the STAND domain 39: NACHT NTPases and NB-ARC ATPases (Table 1). These NTP-binding proteins are often linked to various protein-binding modules such as WD40, Ankyrin or TPR at the C-terminus and a highly divergent nucleoside phosphorylase (Pfs) domain at their N-terminus (Fig. 3). A different type of domain composition is found in several other STAND NTPases. NB-ARC can fused to a LipA domain found in putative serine esterases and in the SesB protein from Nectria haematococca 40. Some NACHT NTPases are linked to the HET-s/LopB domain described above. The orthologous relationships within the two STAND domain expansions are difficult to establish since the proteins are highly divergent and exhibit uneven phyletic distribution. They also appear to have undergone multiple domain shuffling events as well as lineage-specific gene loss and expansions during the evolution of the Aspergilli as well as other filamentous fungi.

As mentioned earlier, one of N. crassa NACHT domain protein is linked to the HET-s/LopB domain. P. anserina HET-D and HET-E are fused to the HET domain and 11 WD40 repeats, which determine their allelic specificity 41. HET-E has been shown to genetically (and potentially physically) interact with HET-C2 to trigger incompatibility in P. anserina suggesting that the interaction may activate the ceramide stress response pathway 24. Similar to the HET domain expansion, the two STAND domain expansions may represent a niche adaptation strategy in filamentous ascomycetes. The multiple fusions involving STAND domains in filamentous fungi may be responsible for the enhancement of their repertoire of signal-transducing interactions, linking preexisting signaling pathways, or integrating multiple signals.

Although their specific biological role is unknown, several functional inferences can be made regarding the role of the STAND domain proteins in the Aspergilli. Despite the variability of their domain architecture, the regulatory/signaling function of STAND NTPases seems to be conserved from fungi to man to possibly bacteria. The domain has been implicated in hetero-oligomerization and signal transduction during apoptosis, inflammatory and pathogen responses in animals and plants, and in transcriptional regulation of secondary metabolism in bacteria 424344. Similar to their fungal counterparts, animal and plant members of the superfamily tend to be fused to death effector domains at the N-terminus and to repetitive protein binding/regulatory modules at the C-terminus 39.

Other observations suggest that Pfs and STAND domain fusion proteins in the Aspergilli may play a regulatory or signaling role. In plants, the Pfs domain was found in several stress-inducible enzymes 42. The Pfs domain is also found in bacterial methylthioadenosine/S-adenosylhomocysteine (MTA/SAH) nucleosidases and phosphoribosyltransferases 45. The bacterial nucleosidases function in methionine salvage pathway and control intracellular levels of MTA/SAH and, in several cases, production of a quorum-sensing signaling molecule 46. In animals, MTA has been shown to affect many critical responses including regulation of gene expression, proliferation, differentiation and apoptosis 47.

The NB-ARC and LipA as well as NACHT and HET-s/LopB domain fusion proteins may function as bistable switches in signaling cascades. A putative serine esterase domain, LipA, is found in the SesB protein implicated in bistability of developmental signaling cascades in Nectria haematococca 40. Incidentally, sesB is adjacent to a gene, which encodes a NACHT and Ankyrin domain protein. HET-s is also associated with bistability of the HI reaction resulting from the spread of the infectious prion in P. anserina 38.

Another line of evidence implicating this family in integration of developmental, stress, and nutrient availability signals, comes from expression data. At least five NACHT domain proteins appear to be regulated by LaeA in A. fumigatus (N. Keller, S. Kim and W. Nierman, unpublished), a putative chromatin-dependent regulator of secondary metabolism, virulence and conidiation 4849. A few other putative PCD-associated proteins seem to be affected by LaeA in A. fumigatus, including both metacaspases, bZIP transcription factor JlbA, BAX Inhibitor family protein, Cu²⁺/Zn²⁺ superoxide dismutase SOD1, histone chaperone ASF1, and AMID-like mitochondrial oxidoreductase, some of which are described below.

In addition to the putative HI inducers, the Aspergilli possess homologs of HI suppressors from P. anserina and N. crassa (Table 2) 15. Thus, sequence similarity searches detected orthologs of N. crassa VIB-1 50 and P. anserina MOD-A, MOD-D and MOD-E 515253. VIB-1, a putative regulator of conidiation and HI in N. crassa, is orthologous to Penicillium chrysogenum PhoG and A. nidulans PacG, which were annotated as putative non-repressible acid phosphatases via transformation experiments 505455. There is no apparent orthologs in yeasts, although, VIB-1 is a distant homolog of the Ndt80p transcription factor from S. cerevisiae 56. The distribution of MOD-A orthologs also appears to be limited to filamentous ascomycetes. Only A. oryzae, but not other Aspergilli, contains an ortholog of MOD-A, implicated in ascomycete-specific functions such as regulation of growth arrest during HI and female organ formation in P. anserina 51. On the contrary, MOD-D and MOD-E display a very high level of sequence conservation and have a much wider phyletic distribution with homologs in all fungi and higher eukaryotes. MOD-D, a G protein alpha subunit, is orthologous to GpaB and GanB from A. fumigatus and A. nidulans, respectively 5758.

The degree of sequence conservation seems to be linked to the relative importance of the biological function and more conserved proteins appear to be functional orthologs. The similarity between VIB-1 and Ndt80p, a transcription factor involved in regulation of sporulation and meiosis in S. cerevisiae 56, suggests that phoG and pacG may encode a transcriptional regulators of the acid phosphatase, rather than the enzyme itself 59. MOD-D and MOD-E have been shown to function as an alpha subunit of heterotrimeric G protein, and an HSP90 family molecular chaperone, respectively 5253. P. anserina MOD-E, in addition to suppressing HI, is involved in regulation of development and the sexual cycle 5360. The MOD-E/HSP90 function during the sexual cycle appears to be conserved from fungi to mammals 61. Mammalian HSP90 family chaperones also mediate the unfolded protein response to endoplasmic reticulum stress through regulation of the secretory pathway, cell cycle and programmed cell death 6162.

Besides HI suppressors, Aspergilli also possess orthologs of P. anserina idi-2, idi-4, idi-6 and idi-7 genes (Table 2) induced by heterokaryon incompatibility and implicated in autophagy in response to starvation and sporulation and in regulation of sexual differentiation 8636465. IDI-6 and IDI-7 proteins seem to be highly conserved across fungal species; while IDI-2 and IDI-4 are poorly conserved and their distribution is limited to filamentous fungi. No detectable homologs of IDI-1 and IDI-3 are found in Aspergilli and an IDI-2 ortholog is only present in A. oryzae. Autophagic serine protease IDI-6 is orthologous to A. fumigatus Alp2 that has been shown to function in regulation of sporulation as well as pathogenesis in A. fumigatus 66. Likewise, IDI-4 an ortholog of the A. fumigatus JlbA, a putative bZIP transcription factor induced by amino acid starvation 67.

Many HI-associated genes have wide phyletic distribution and are well-conserved across filamentous fungi. Some are involved in autophagy, suggesting that the incompatibility function might have evolved by recruiting components of the cellular system controlling adaptation to starvation 11. In addition, cytological alterations during HI in P. anserina are similar to those observed during starvation and treatment with rapamycin, an inhibitor of the TOR (

t

o

r

In addition to HI, filamentous fungi appear to possess a wide range of PCD reactions triggered by various death stimuli. In Aspergilli, the apoptotic-like phenotypes and are observed during entry into stationary phase and sporulation and on exposure to certain antifungal agents, peptides, and sphingosines 4676970. Similarly, in N. crassa, morphological changes during the HI reaction, starvation, and DNA damage response resemble apoptosis 91071. The apoptotic machinery of filamentous fungi may share some key components with yeast and mammalian systems. First, we looked for homologs of apoptotic proteins in S. cerevisiae, which can undergo PCD in response to nutritional and oxidative stresses, plant antifungal peptides, hydrogen peroxide, or during aging and mating 72737475.

To identify candidate apoptosis-associated proteins in the Aspergilli, BLASTp similarity searches were performed with yeast apoptotic proteins as queries. Homologs of more than 30 yeast proteins were detected including metacaspases and caspase-regulating serine protease HtrA2 (Table 3). Phylogenetic analysis shows that many Aspergillus proteins are in one-on-one orthologous relationships with S. cerevisiae, S. pombe, and basidiomycete proteins (data not shown).

BLASTp and HMM searches identified homologs of ~50 human and mouse apoptotic proteins in the Aspergilli [see Additional file 1 ]. Similar to yeasts, filamentous ascomycetes lack the upstream metazoan apoptotic regulators including members of the bax/bcl-2 family and p53, while downstream components of the apoptotic machinery appear to be shared. Interestingly, many Aspergillus proteins are more similar to their human counterparts such as AMID, BIR1, HtrA, and CulA, than to yeast homologs. Thus, the tree topology of the AIF family confirmed that Aspergilli proteins are more closely related to human AMID than to S. cerevisiae Aif1p, which clustered together with plant homologs [see Additional file 2 ]. Moreover, homologs of several key components of the mammalian apoptotic machinery, including AmsH and Poly(ADP-ribose) polymerase (PARP), are not detected in S. cerevisiae (Table 4).

At least two Aspergilli proteins appear to be functional homologs to their mammalian counterparts. The enhanced PARP and caspase-like activity reported in A. nidulans during sporulation-induced PCD is consistent with the presence of both metacaspase-dependent and -independent apoptotic pathways 7076. In addition, the metacaspase-independent apoptosis pathway was shown to operate in A. fumigatus during stationary phase and treatment with fungicidal sphingoid bases and antifungal agents 469. For the rest of the fungal proteins, further experimental characterization is required before any conclusions can be drawn regarding their involvement in PCD. Many yeast and mammalian apoptotic proteins appear to be involved in regulation of cell programs monitoring the cell status such as maintenance of genome integrity, cell cycle control, glycolipid metabolism, and ubiquitin-dependent proteolysis [see Additional file 2] 77. It is likely that at least non-apoptotic function is conserved in both fungi and metazoa. The results also support the idea that complex development and differentiation in filamentous fungi may require additional PCD pathways or their components not found in unicellular yeasts 78.

To identify human and mouse proteins implicated in PCD, we searched the Gene Ontology (GO) database 81 and the Apoptosis database 82. Then, sequence similarity searches were performed using PSI-BLAST and Gapped BLAST against selected fungal genomes downloaded from GenBank. The searches were also performed against an in-house database composed of whole-genome sequences of several fungal species from finished and ongoing sequencing projects. The N. fischeri genome sequence has been generated in the course of the genome sequencing projects at TIGR, The Institute for Genomic Research (Rockville, MD). Conserved protein domains were identified using the HMMer package 83.

Protein sequences were re-aligned using MUSCLE 84 and columns of low conservation removed manually. The alignments were then used to infer bootstrapped neighbor-joining and maximum-likelihood trees. The neighbor-joining trees were constructed using QuickTree 85 and the maximum-likelihood trees were constructed using the PHYLIP package 86, applying the JTT substitution model with a gamma distribution (alpha = 0.5) of rates over four categories of variable sites. In general, the maximum-likelihood and neighbor-joining trees were congruent.