β-Tubulin gene tree of Penicillium subgenus Penicillium , sect. Viridicata, ser. Corymbifera, Camemberti, and Solita , including all sequenced strains. One of 5000 equally most parsimonious trees of 95 steps based on a heuristic search with P. gladioli as outgroup. The branches in bold occur in 100% of the equally most parsimonious trees. The numbers represent bootstrap percentages > 50 %. (CI= 0.811 RI= 0.940 RC= 0.762, HI= 0.189). Ex-type cultures are indicated with T. 

Contexts in source publication

Context 1

... 98%, Fig. 1, see also Fig. 5). The β-tubulin analysis was consistent with the ITS analysis of the same group of species by Skouboe et al. (1996) and Boysen et al . (1996). According to Fig. 1, ser. Roqueforti is a sister group to Eupenicillium osmophilum , with which it shares few phenotypic characters except the ability, widespread in the subgenus, to produce roquefortine C. Given the phenotypic similarities to species in ser. Urticicolae , Expansa and Claviformia , ser. Roqueforti may eventually be shown to be more closely related to these series. Most of these species are able to grow on creatine as the sole nitrogen source, and produce patulin and roquefortine C. However, the coprophilous species in ser. Claviformia compete well in this alkaline habitat and grow along with the associated alkali-tolerant bacteria. In contrast, species in ser. Roqueforti appear to have co-evolved with lactic acid bacteria and tolerate lactic acid bacteria (and yeast) products such as lactic acid, carbon dioxide and ethanol. The phylogeny suggested for sect. Penicillium provided good support for the species recognized by Frisvad & Samson (2004), but variable support for the proposed infra-section classification (Figs. 6, 7). One problem was the position of P. sclerotigenum , which was classified in subgenus Furcatum by Pitt (1979). This species has some affinities with ser. Expansa by its high growth rate, plant rot, production of roquefortine C and patulin, and was therefore placed by Frisvad & Samson (2004) in series Expansa. According to the β-tubulin phylogeny, however, it may be more appropriately classified with the citrus- loving Penicillia P. italicum and P. ulaiensis . Despite this, P. sclerotigenum shares very few phenotypic features with the species included in ser. Italica . Series Claviformia comprises mostly synnematous species, many of which occur on dung. In Fig. 7, ser. Claviformia appeared to be paraphyletic with ser. Urticicolae derived within it , although this was strongly supported by strict consensus but not bootstrap. However, the phylogeny supports some of the ecological hypotheses of Frisvad (1998). Perhaps the primarily soil-borne species P. griseofulvum was derived from coprophilous fungi associated with rodent dung. Its closest relative, P. dipodomyicola , is found primarily in seed caches of desert rats. Similarly, perhaps P. glandicola , generally associated with Quercus (especially acorns), was derived from a fungus that originally lived on the pellets of a rodent collecting these seeds. The phylogenetic structure of sect. Chrysogena was disrupted by ser. Mononematosa and Persicina , which rendered ser. Chrysogena polyphyletic (Fig. 8). The backbone of this phylogram, which represents an ecologically and phenotypically distinctive group of species, was not well-supported by bootstrap analysis although strict consensus support was relatively strong. Isozyme data indicated that the four species in ser. Chrysogena are closely related (Banke et al ., 1997), with P. confertum a possible outgroup. Isozyme analyses suggested close relationships between P. chrysogenum and P. flavigenum, and P. dipodomyis and P. nalgiovense , which were confirmed by the β-tubulin data presented here and extrolite profiles. While P. confertum shares the production of roquefortine C, meleagrin and secalonic acid with P. chrysogenum and P. flavigenum , P. mononematosum has no extrolites in common with any species of ser. Chrysogena (Frisvad and Samson, 2004). Penicillium confertum and P. mononematosum should be examined to determine whether they produce penicillin, which is common to all species in ser. Chrysogena . Section Coronata was the phylogenetically most distant section included in subgenus Penicillium , and includes species that are themselves also phylogenetically rather distant from one another (Fig. 9). The position of this group of species at the base of subgenus Penicillium was clear in the ITS-D1/D2 analysis by Peterson (2000). However, the structure of this clade suggested either that several species remain undiscovered, or the possibility that it might be pulled into subgenus Penicillium by long branch attraction. Perhaps the discovery and phylogenetic analysis of additional taxa will demonstrate that sect. Coronata should be classified with another subgenus, or be regarded as a subgenus on its own. The β-tubulin sequence analysis provided excellent support for the species concepts adopted by Frisvad & Samson (2004), with a few exceptions. The species that were problematic were: a) P. freii and P. neoechinulatum , which had identical sequences (Fig. 2); b) P. viridicatum , which despite some sequence differences among its strains, was not phylogenetically distinct from P. freii/P. neoechinulatum (Fig. 2); c) P. camemberti and P. caseifulvum , which had identical sequences (Fig. 3); d) P. commune , which had strains that were either paraphyletic with or identical to P. camemberti and P. caseifulvum (Fig. 3); e) P. nordicum , which was paraphyletic because P. verrucosum was derived within it, and f) P. confertum, represented by one strain with an identical sequence to the two strains of P. mononematosum (Fig. 8). The three species P. freii , P. neoechinulatum and P. viridicatum are easily separated by conidium ornamentation, conidium colour and extrolite profiles (Frisvad & Samson 2004). Additional similarities are the production of the closely related diketopiperazines extrolites aurantiamine and viridamine. A synonymy of these three species cannot be considered based on the available data and it is clear that β-tubulin lacks sufficient resolution to support the phenotype-based classification. More genes should be sequenced in order to determine whether a DNA-based phylogenetic species concept can be applied to these species. The lack of consistent sequence differences between P. camemberti and P. caseifulvum could be a reflection of the hypothesis that they are two domesticated derivatives of P. commune (Polonelli et al. 1987). However, they differ in macroscopic morphology and extrolite profiles and can be regarded as distinct species. Perhaps P. caseifulvum has a silent gene cluster for cyclopiazonic acid, a characteristic metabolite of P. camemberti, but this extrolite has never been detected in that species. The few sequence differences between P. verrucosum and P. nordicum were in agreement with the results of Castella et al . (2002), who also found few sequence differences based on ITS1-5.8S-ITS2. However, they found a clear distinction between P. verrucosum and P. nordicum based on RAPD and AFLP patterns. Penicillium nordicum itself presents an interesting situation for additional work. Originally, Frisvad & Samson (2004) had considered the possibility that the two variants evident in Fig. 4 were different species. One variant of P. nordicum was mostly found on salted lumpfish roe and produced the extrolite lumpidin (Larsen et al ., 2001a, b). Multigene analysis will be necessary to determine whether this possibility warrants further consideration. Only one strain of P. confertum is available, but it differs fom the related P. mononematosum by its less complicated penicilli and thin often sinuous conidophore stipes. Furthermore it produces asteltoxin and meleagrin which are absent in P. Apart from these examples, all other species were monophyletic in our phylogenetic analyses. Branch lengths leading into the species were variable in length, from 1-40 bp substitutions or indels, being particularly short in some taxa (eg. sect. Viridicata , Fig. 2-4), where the longest branches were about 5 bp leading into P. venetum and P. allii ), to very long in sect. Coronata (Fig. 8). In general, the longer supporting branches also had highest bootstrap support. With sampling usually limited to three or four strains per species, only preliminary observations can be made on infraspecific variation in the β-tubulin allele with a species. Many species were invariant, including P. aethiopicum , P. allii, P. aurantiogriseum, and Eupenicillium osmophilum . In many species, only one bp substitution was noted among the strains, including P. albocoremium, P. cavernicola, ...

Context 2

... relationships and providing molecular support for phenotypically based species concepts. In general, the overall relationships suggested by the ITS were similar to those suggested by β-tubulin, but the β- tubulin analyses had more resolution in the terminal branches. Even though there was little bootstrap support for most species in the ITS study by Skouboe et al. (1996, 2000) and the rDNA study by Peterson (2000), the small sequence differences found were in agreement with our results in most cases. Since its initial use in phylogenetic studies of Epichloe (Schardl et al., 1997), β-tubulin sequences have been used as putative species markers or for phylogenetic analysis in a variety of ascomycete and hyphomycete genera, including Neocosmospora and Gibberella (anamorphs Fusarium, O’Donnell et al . 1998, 2004), Calonectria (anamorphs Cylindrocladium , Schoch et al ., 2001), Neonectria (anamorphs Cylindrocarpon , Seifert et al . 2003), Bionectria (anamorphs Clonostachys , Schroers, 2001), Phaeoacremonium (Dupont et al ., 2002), Botryosphaeria (Slippers et al ., 2004), Ophiostoma (various anamorphs, Jacobs and Kirisits, 2003), Stachybotrys (Andersen et al ., 2003), Aspergillus (Geiser et al ., 1998; Peterson 2001), Ascochyta (Fatehi et al ., 2003), Sphaerophorus (Hognabba and Wedin, 2003), Pseudocercospora (Beilharz and Cunnington, 2003), and Parmelia (Molina et al ., 2004). Despite its demonstrated utility in these genera, problems with multiple gene copies and failure of primers in some groups restrict its applicability. Furthermore, the sometimes highly variable introns at the 5’ end of the most frequently sequenced part of the β-tubulin gene can make alignment even across a genus a serious challenge for a computer, resulting in an alignment almost impossible to evaluate using the human eye. Fortunately, our data set for subgenus Penicillium is free of paralogs, and the taxon is sufficiently homogenous that reasonable alignments could be assembled with confidence. The β -tubulin sequences generated in this study were intended to serve two functions. First, we hoped to find phylogenetic support for the phenotypically based classification of sections and series proposed by Frisvad & Samson (2004). Second, we wanted to derive DNA sequence data that would complement the species concepts suggested by phenotypic data, and facilitate species discovery and identification, and eventually the development of molecular diagnostics. The cladification presented in Figs. 1-9 is generally complimentary to the classification of species into sections and series in subgenus Penicillium proposed by Frisvad & Samson (2004) (Table 2) and is discussed in detail below. The limitations of a single gene phylogeny for producing conclusive evidence for a well-resolved and supported phylogeny are clear in our analysis. Our decision to emphasize collecting sequence data from a large set of strains (180) required that we focus on a single gene. Our phylograms sometimes lack satisfactory bootstrap support for clades representing sections or series in the Frisvad & Samson (2004) classification. Questions of monophyly of these higher taxa cannot be conclusively settled because of these limitations. The constraint analyses we ran to test the monophyly of groups that were in conflict between the classification and the cladification did not reject these alternative hypotheses. Sequencing and analysis of additional genes, such as calmodulin (Peterson 2001, 2004), elongation factor 1-alpha (Peterson 2004), and ribosomal polymerase B2 (Seifert et al ., unpublished), should be explored. The recent assertion by Rokas et al . (2003) that analysis of as many as thirty genes may be necessary to obtain a fully resolved, completely supported phylogeny of a group of eight yeast species is certainly sobering in the context of a speciose taxon like subgenus Penicillium . The taxonomic classification of the largest section of subgenus Penicillium , sect. Viridicata was generally supported. Three series were monophyletic, and the monophyly of ser. Camemberti, Solita and Corymbifera were not evident in the most parsimonious trees, but not rejected by constraint analyses (Figs. 1-4). Even for those clades lacking strong bootstrap support, we identified several phenotypic characters that are consistent with the cladification. The phylogram for the mostly grain- or seed-borne species in ser. Viridicata was similar to that for many of the same species presented by Seifert & Louis-Seize (2000). The addition of P. tricolor to the data set confirms its phylogenetic position in this series. Seifert & Louis-Seize (2000) did not include this species in their analysis because only unalignable β-tubulin paralogs were amplified and sequenced at that time. The lipid- and protein-loving species in ser. Camemberti and Solita have several phenotypic characters (e.g. strong growth on CREA, large conidia) that support the branch that distinguishes them from the other series (Fig. 3). However, both series as defined by Frisvad & Samson (2004) include species whose classification could not be confirmed or rejected by the β-tubulin gene tree. Penicillium cavernicola , included by Frisvad & Samson (2004) in ser. Solita, appeared more closely related to ser. Camemberti in Fig. 3, but there was no bootstrap or strict consensus support for this alternative classification. Based on strict consensus but not bootstrap support, P. crustosum was closer to ser. Solita than its proposed classification in ser. Camemberti , but constraint analyses suggested that either relationship was possible. Earlier studies using rDNA sequences indicated that P. crustosum was closely related to P. camemberti and P. commune (Skouboe et al ., 1996, 2000; Peterson, 2000). The production of viridicatins and ability to rot apples is shared by P. solitum and P. crustosum , but P. crustosum also has phenotypic similarities to P. expansum (growth rate, apple rot). As delimited by Frisvad & Samson (2004), ser. Corymbifera appeared phylogenetically dispersed in the β-tubulin analysis. This series, united by the tendency of the species to attack plant bulbs and formation of feathery synnemata, was also differentiated from the other series by its smaller conidia and weak growth on CREA. However, the results suggesting polyphyly for this series were not particularly strong. A more focussed multigene phylogeny may be necessary to conclusively prove or reject the monophyly of this series. Section Roqueforti ser. Roqueforti was defined by a long list of phenotypic characters and had a completely resolved, well-supported species phylogeny (bootstrap support, 98%, Fig. 1, see also Fig. 5). The β-tubulin analysis was consistent with the ITS analysis of the same group of species by Skouboe et al. (1996) and Boysen et al . (1996). According to Fig. 1, ser. Roqueforti is a sister group to Eupenicillium osmophilum , with which it shares few phenotypic characters except the ability, widespread in the subgenus, to produce roquefortine C. Given the phenotypic similarities to species in ser. Urticicolae , Expansa and Claviformia , ser. Roqueforti may eventually be shown to be more closely related to these series. Most of these species are able to grow on creatine as the sole nitrogen source, and produce patulin and roquefortine C. However, the coprophilous species in ser. Claviformia compete well in this alkaline habitat and grow along with the associated alkali-tolerant bacteria. In contrast, species in ser. Roqueforti appear to have co-evolved with lactic acid bacteria and tolerate lactic acid bacteria (and yeast) products such as lactic acid, carbon dioxide and ethanol. The phylogeny suggested for sect. Penicillium provided good support for the species recognized by Frisvad & Samson (2004), but variable support for the proposed infra-section classification (Figs. 6, 7). One problem was the position of P. sclerotigenum , which was classified in subgenus Furcatum by Pitt (1979). This species has some affinities with ser. Expansa by its high growth rate, plant rot, production of roquefortine C and patulin, and was therefore placed by Frisvad & Samson (2004) in series Expansa. According to the β-tubulin phylogeny, however, it may be more appropriately classified with the citrus- loving Penicillia P. italicum and P. ulaiensis . Despite this, P. sclerotigenum shares very few phenotypic features with the species included in ser. Italica . Series Claviformia comprises mostly synnematous species, many of which occur on dung. In Fig. 7, ser. Claviformia appeared to be paraphyletic with ser. Urticicolae derived within it , although this was strongly supported by strict consensus but not bootstrap. However, the phylogeny supports some of the ecological hypotheses of Frisvad (1998). Perhaps the primarily soil-borne species P. griseofulvum was derived from coprophilous fungi associated with rodent dung. Its closest relative, P. dipodomyicola , is found primarily in seed caches of desert rats. Similarly, perhaps P. glandicola , generally associated with Quercus (especially acorns), was derived from a fungus that originally lived on the pellets of a rodent collecting these seeds. The phylogenetic structure of sect. Chrysogena was disrupted by ser. Mononematosa and Persicina , which rendered ser. Chrysogena polyphyletic (Fig. 8). The backbone of this phylogram, which represents an ecologically and phenotypically distinctive group of species, was not well-supported by bootstrap analysis although strict consensus support was relatively strong. Isozyme data indicated that the four species in ser. Chrysogena are closely related (Banke et al ., 1997), with P. confertum a possible outgroup. Isozyme analyses suggested close relationships between P. chrysogenum and P. flavigenum, and P. dipodomyis and P. nalgiovense , which were confirmed by the β-tubulin data presented here and extrolite profiles. While P. confertum shares the production of ...

Context 3

... by a long list of phenotypic characters and had a completely resolved, well-supported species phylogeny (bootstrap support, 98%, Fig. 1, see also Fig. 5). The β-tubulin analysis was consistent with the ITS analysis of the same group of species by Skouboe et al. (1996) and Boysen et al . (1996). According to Fig. 1, ser. Roqueforti is a sister group to Eupenicillium osmophilum , with which it shares few phenotypic characters except the ability, widespread in the subgenus, to produce roquefortine C. Given the phenotypic similarities to species in ser. Urticicolae , Expansa and Claviformia , ser. Roqueforti may eventually be shown to be more closely related to these series. Most of these species are able to grow on creatine as the sole nitrogen source, and produce patulin and roquefortine C. However, the coprophilous species in ser. Claviformia compete well in this alkaline habitat and grow along with the associated alkali-tolerant bacteria. In contrast, species in ser. Roqueforti appear to have co-evolved with lactic acid bacteria and tolerate lactic acid bacteria (and yeast) products such as lactic acid, carbon dioxide and ethanol. The phylogeny suggested for sect. Penicillium provided good support for the species recognized by Frisvad & Samson (2004), but variable support for the proposed infra-section classification (Figs. 6, 7). One problem was the position of P. sclerotigenum , which was classified in subgenus Furcatum by Pitt (1979). This species has some affinities with ser. Expansa by its high growth rate, plant rot, production of roquefortine C and patulin, and was therefore placed by Frisvad & Samson (2004) in series Expansa. According to the β-tubulin phylogeny, however, it may be more appropriately classified with the citrus- loving Penicillia P. italicum and P. ulaiensis . Despite this, P. sclerotigenum shares very few phenotypic features with the species included in ser. Italica . Series Claviformia comprises mostly synnematous species, many of which occur on dung. In Fig. 7, ser. Claviformia appeared to be paraphyletic with ser. Urticicolae derived within it , although this was strongly supported by strict consensus but not bootstrap. However, the phylogeny supports some of the ecological hypotheses of Frisvad (1998). Perhaps the primarily soil-borne species P. griseofulvum was derived from coprophilous fungi associated with rodent dung. Its closest relative, P. dipodomyicola , is found primarily in seed caches of desert rats. Similarly, perhaps P. glandicola , generally associated with Quercus (especially acorns), was derived from a fungus that originally lived on the pellets of a rodent collecting these seeds. The phylogenetic structure of sect. Chrysogena was disrupted by ser. Mononematosa and Persicina , which rendered ser. Chrysogena polyphyletic (Fig. 8). The backbone of this phylogram, which represents an ecologically and phenotypically distinctive group of species, was not well-supported by bootstrap analysis although strict consensus support was relatively strong. Isozyme data indicated that the four species in ser. Chrysogena are closely related (Banke et al ., 1997), with P. confertum a possible outgroup. Isozyme analyses suggested close relationships between P. chrysogenum and P. flavigenum, and P. dipodomyis and P. nalgiovense , which were confirmed by the β-tubulin data presented here and extrolite profiles. While P. confertum shares the production of roquefortine C, meleagrin and secalonic acid with P. chrysogenum and P. flavigenum , P. mononematosum has no extrolites in common with any species of ser. Chrysogena (Frisvad and Samson, 2004). Penicillium confertum and P. mononematosum should be examined to determine whether they produce penicillin, which is common to all species in ser. Chrysogena . Section Coronata was the phylogenetically most distant section included in subgenus Penicillium , and includes species that are themselves also phylogenetically rather distant from one another (Fig. 9). The position of this group of species at the base of subgenus Penicillium was clear in the ITS-D1/D2 analysis by Peterson (2000). However, the structure of this clade suggested either that several species remain undiscovered, or the possibility that it might be pulled into subgenus Penicillium by long branch attraction. Perhaps the discovery and phylogenetic analysis of additional taxa will demonstrate that sect. Coronata should be classified with another subgenus, or be regarded as a subgenus on its own. The β-tubulin sequence analysis provided excellent support for the species concepts adopted by Frisvad & Samson (2004), with a few exceptions. The species that were problematic were: a) P. freii and P. neoechinulatum , which had identical sequences (Fig. 2); b) P. viridicatum , which despite some sequence differences among its strains, was not phylogenetically distinct from P. freii/P. neoechinulatum (Fig. 2); c) P. camemberti and P. caseifulvum , which had identical sequences (Fig. 3); d) P. commune , which had strains that were either paraphyletic with or identical to P. camemberti and P. caseifulvum (Fig. 3); e) P. nordicum , which was paraphyletic because P. verrucosum was derived within it, and f) P. confertum, represented by one strain with an identical sequence to the two strains of P. mononematosum (Fig. 8). The three species P. freii , P. neoechinulatum and P. viridicatum are easily separated by conidium ornamentation, conidium colour and extrolite profiles (Frisvad & Samson 2004). Additional similarities are the production of the closely related diketopiperazines extrolites aurantiamine and viridamine. A synonymy of these three species cannot be considered based on the available data and it is clear that β-tubulin lacks sufficient resolution to support the phenotype-based classification. More genes should be sequenced in order to determine whether a DNA-based phylogenetic species concept can be applied to these species. The lack of consistent sequence differences between P. camemberti and P. caseifulvum could be a reflection of the hypothesis that they are two domesticated derivatives of P. commune (Polonelli et al. 1987). However, they differ in macroscopic morphology and extrolite profiles and can be regarded as distinct species. Perhaps P. caseifulvum has a silent gene cluster for cyclopiazonic acid, a characteristic metabolite of P. camemberti, but this extrolite has never been detected in that species. The few sequence differences between P. verrucosum and P. nordicum were in agreement with the results of Castella et al . (2002), who also found few sequence differences based on ITS1-5.8S-ITS2. However, they found a clear distinction between P. verrucosum and P. nordicum based on RAPD and AFLP patterns. Penicillium nordicum itself presents an interesting situation for additional work. Originally, Frisvad & Samson (2004) had considered the possibility that the two variants evident in Fig. 4 were different species. One variant of P. nordicum was mostly found on salted lumpfish roe and produced the extrolite lumpidin (Larsen et al ., 2001a, b). Multigene analysis will be necessary to determine whether this possibility warrants further consideration. Only one strain of P. confertum is available, but it differs fom the related P. mononematosum by its less complicated penicilli and thin often sinuous conidophore stipes. Furthermore it produces asteltoxin and meleagrin which are absent in P. Apart from these examples, all other species were monophyletic in our phylogenetic analyses. Branch lengths leading into the species were variable in length, from 1-40 bp substitutions or indels, being particularly short in some taxa (eg. sect. Viridicata , Fig. 2-4), where the longest branches were about 5 bp leading into P. venetum and P. allii ), to very long in sect. Coronata (Fig. 8). In general, the longer supporting branches also had highest bootstrap support. With sampling usually limited to three or four strains per species, only preliminary observations can be made on infraspecific variation in the β-tubulin allele with a species. Many species were invariant, including P. aethiopicum , P. allii, P. aurantiogriseum, and Eupenicillium osmophilum . In many species, only one bp substitution was noted among the strains, including P. albocoremium, P. cavernicola, ...

Context 4

... bootstrap support for most species in the ITS study by Skouboe et al. (1996, 2000) and the rDNA study by Peterson (2000), the small sequence differences found were in agreement with our results in most cases. Since its initial use in phylogenetic studies of Epichloe (Schardl et al., 1997), β-tubulin sequences have been used as putative species markers or for phylogenetic analysis in a variety of ascomycete and hyphomycete genera, including Neocosmospora and Gibberella (anamorphs Fusarium, O’Donnell et al . 1998, 2004), Calonectria (anamorphs Cylindrocladium , Schoch et al ., 2001), Neonectria (anamorphs Cylindrocarpon , Seifert et al . 2003), Bionectria (anamorphs Clonostachys , Schroers, 2001), Phaeoacremonium (Dupont et al ., 2002), Botryosphaeria (Slippers et al ., 2004), Ophiostoma (various anamorphs, Jacobs and Kirisits, 2003), Stachybotrys (Andersen et al ., 2003), Aspergillus (Geiser et al ., 1998; Peterson 2001), Ascochyta (Fatehi et al ., 2003), Sphaerophorus (Hognabba and Wedin, 2003), Pseudocercospora (Beilharz and Cunnington, 2003), and Parmelia (Molina et al ., 2004). Despite its demonstrated utility in these genera, problems with multiple gene copies and failure of primers in some groups restrict its applicability. Furthermore, the sometimes highly variable introns at the 5’ end of the most frequently sequenced part of the β-tubulin gene can make alignment even across a genus a serious challenge for a computer, resulting in an alignment almost impossible to evaluate using the human eye. Fortunately, our data set for subgenus Penicillium is free of paralogs, and the taxon is sufficiently homogenous that reasonable alignments could be assembled with confidence. The β -tubulin sequences generated in this study were intended to serve two functions. First, we hoped to find phylogenetic support for the phenotypically based classification of sections and series proposed by Frisvad & Samson (2004). Second, we wanted to derive DNA sequence data that would complement the species concepts suggested by phenotypic data, and facilitate species discovery and identification, and eventually the development of molecular diagnostics. The cladification presented in Figs. 1-9 is generally complimentary to the classification of species into sections and series in subgenus Penicillium proposed by Frisvad & Samson (2004) (Table 2) and is discussed in detail below. The limitations of a single gene phylogeny for producing conclusive evidence for a well-resolved and supported phylogeny are clear in our analysis. Our decision to emphasize collecting sequence data from a large set of strains (180) required that we focus on a single gene. Our phylograms sometimes lack satisfactory bootstrap support for clades representing sections or series in the Frisvad & Samson (2004) classification. Questions of monophyly of these higher taxa cannot be conclusively settled because of these limitations. The constraint analyses we ran to test the monophyly of groups that were in conflict between the classification and the cladification did not reject these alternative hypotheses. Sequencing and analysis of additional genes, such as calmodulin (Peterson 2001, 2004), elongation factor 1-alpha (Peterson 2004), and ribosomal polymerase B2 (Seifert et al ., unpublished), should be explored. The recent assertion by Rokas et al . (2003) that analysis of as many as thirty genes may be necessary to obtain a fully resolved, completely supported phylogeny of a group of eight yeast species is certainly sobering in the context of a speciose taxon like subgenus Penicillium . The taxonomic classification of the largest section of subgenus Penicillium , sect. Viridicata was generally supported. Three series were monophyletic, and the monophyly of ser. Camemberti, Solita and Corymbifera were not evident in the most parsimonious trees, but not rejected by constraint analyses (Figs. 1-4). Even for those clades lacking strong bootstrap support, we identified several phenotypic characters that are consistent with the cladification. The phylogram for the mostly grain- or seed-borne species in ser. Viridicata was similar to that for many of the same species presented by Seifert & Louis-Seize (2000). The addition of P. tricolor to the data set confirms its phylogenetic position in this series. Seifert & Louis-Seize (2000) did not include this species in their analysis because only unalignable β-tubulin paralogs were amplified and sequenced at that time. The lipid- and protein-loving species in ser. Camemberti and Solita have several phenotypic characters (e.g. strong growth on CREA, large conidia) that support the branch that distinguishes them from the other series (Fig. 3). However, both series as defined by Frisvad & Samson (2004) include species whose classification could not be confirmed or rejected by the β-tubulin gene tree. Penicillium cavernicola , included by Frisvad & Samson (2004) in ser. Solita, appeared more closely related to ser. Camemberti in Fig. 3, but there was no bootstrap or strict consensus support for this alternative classification. Based on strict consensus but not bootstrap support, P. crustosum was closer to ser. Solita than its proposed classification in ser. Camemberti , but constraint analyses suggested that either relationship was possible. Earlier studies using rDNA sequences indicated that P. crustosum was closely related to P. camemberti and P. commune (Skouboe et al ., 1996, 2000; Peterson, 2000). The production of viridicatins and ability to rot apples is shared by P. solitum and P. crustosum , but P. crustosum also has phenotypic similarities to P. expansum (growth rate, apple rot). As delimited by Frisvad & Samson (2004), ser. Corymbifera appeared phylogenetically dispersed in the β-tubulin analysis. This series, united by the tendency of the species to attack plant bulbs and formation of feathery synnemata, was also differentiated from the other series by its smaller conidia and weak growth on CREA. However, the results suggesting polyphyly for this series were not particularly strong. A more focussed multigene phylogeny may be necessary to conclusively prove or reject the monophyly of this series. Section Roqueforti ser. Roqueforti was defined by a long list of phenotypic characters and had a completely resolved, well-supported species phylogeny (bootstrap support, 98%, Fig. 1, see also Fig. 5). The β-tubulin analysis was consistent with the ITS analysis of the same group of species by Skouboe et al. (1996) and Boysen et al . (1996). According to Fig. 1, ser. Roqueforti is a sister group to Eupenicillium osmophilum , with which it shares few phenotypic characters except the ability, widespread in the subgenus, to produce roquefortine C. Given the phenotypic similarities to species in ser. Urticicolae , Expansa and Claviformia , ser. Roqueforti may eventually be shown to be more closely related to these series. Most of these species are able to grow on creatine as the sole nitrogen source, and produce patulin and roquefortine C. However, the coprophilous species in ser. Claviformia compete well in this alkaline habitat and grow along with the associated alkali-tolerant bacteria. In contrast, species in ser. Roqueforti appear to have co-evolved with lactic acid bacteria and tolerate lactic acid bacteria (and yeast) products such as lactic acid, carbon dioxide and ethanol. The phylogeny suggested for sect. Penicillium provided good support for the species recognized by Frisvad & Samson (2004), but variable support for the proposed infra-section classification (Figs. 6, 7). One problem was the position of P. sclerotigenum , which was classified in subgenus Furcatum by Pitt (1979). This species has some affinities with ser. Expansa by its high growth rate, plant rot, production of roquefortine C and patulin, and was therefore placed by Frisvad & Samson (2004) in series Expansa. According to the β-tubulin phylogeny, however, it may be more appropriately classified with the citrus- loving Penicillia P. italicum and P. ulaiensis . Despite this, P. sclerotigenum shares very few phenotypic features with the species included in ser. Italica . Series Claviformia comprises mostly synnematous species, many of which occur on dung. In Fig. 7, ser. Claviformia appeared to be paraphyletic with ser. Urticicolae derived within it , although this was strongly supported by strict consensus but not bootstrap. However, the phylogeny supports some of the ecological hypotheses of Frisvad (1998). Perhaps the primarily soil-borne species P. griseofulvum was derived from coprophilous fungi associated with rodent dung. Its closest relative, P. dipodomyicola , is found primarily in seed caches of desert rats. Similarly, perhaps P. glandicola , generally associated with Quercus (especially acorns), was derived from a fungus that originally lived on the pellets of a rodent collecting these seeds. The phylogenetic structure of sect. Chrysogena was disrupted by ser. Mononematosa and Persicina , which rendered ser. Chrysogena polyphyletic (Fig. 8). The backbone of this phylogram, which represents an ecologically and phenotypically distinctive group of species, was not well-supported by bootstrap analysis although strict consensus support was relatively strong. Isozyme data indicated that the four species in ser. Chrysogena are closely related (Banke et al ., 1997), with P. confertum a possible outgroup. Isozyme analyses suggested close relationships between P. chrysogenum and P. flavigenum, and P. dipodomyis and P. nalgiovense , which were confirmed by the β-tubulin data presented here and extrolite profiles. While P. confertum shares the production of roquefortine C, meleagrin and secalonic acid with P. chrysogenum and P. flavigenum , P. mononematosum has no extrolites in common with any species of ser. Chrysogena (Frisvad and Samson, 2004). Penicillium confertum and P. mononematosum should be examined to determine whether they produce penicillin, ...

Context 5

... sequences are available in a searchable format as part of an identification application at the CBS website, at the following URL . knaw.nl/penicillium.htm Nine separate parsimony analyses were made. The first included the type strains of all accepted species, and gave an overall impression of the phylogenetic structure of the β-tubulin gene tree. Subsequent analyses included all strains of particular sections or series of subgenus Penicillium , in groupings derived from the first analysis. Constraint analyses were run to test conflicts between the proposed classification of Frisvad & Samson (2004) and the cladification suggested by the β-tubulin gene trees, using the Kishino-Hasegawa test in PAUP 4.0b10. The constraints implemented are discussed in the Results section below. The results of the maximum parsimony analyses are presented as phylograms in Figs. 1-9. Each figure represents one of the most parsimonious trees (MPTs) from each of the nine analyses, with lines in bold designating branches present in the strict consensus tree (i.e . 100%) of the MPTs. Because of the large numbers of identical or very similar sequences in some data sets, there were often many topologically equivalent trees among the MPTs, differing in the arrangement of terminal or zero-length branches. Therefore, we generally restricted heuristic searches to 5000 MPTs, to avoid saturating the computer’s memory with redundant trees. Prior to proceeding with the analyses presented here, a preliminary analysis was undertaken using an alignment including all the 180 sequences present in the complete data set. This analysis gave a similar overall species topology to Fig. 1 (results not shown). All species remained monophyletic (except as noted below for other analyses) and the relationships among them were consistent with those shown in Fig. 1. The difficulty of effectively presenting such an analysis in print form, combined with problems inherent in assessing the reliability of a computerized alignment of this size, led us to the series of analyses presented in this paper. Several preliminary parsimony analyses were run on subsets of the data to assess the reliability of alignments. Separate analyses were run for the exons (which clearly had no alignment ambiguities) and for the introns (which tended to include areas that were more difficult to align satisfactorily). The results of these analyses (not shown here) suggested that in the data set covering the whole subgenus, many of the characters in the introns were responsible for the fine structure of the phylogram, in particular the resolution in the terminal branches. For all analyses reported here, both exons and introns were included. A first analysis was made for the entire subgenus Penicillium ; with each species represented by its type (or other reliably identified) strain (Fig. 1). Eight separate analyses were subsequently undertaken to examine the relationships of three or four strains per species for clades identified from the larger data set. These clades did not necessarily receive strong bootstrap support in Fig. 1, but all received 100% consensus support. For each of these analyses, the alignments were individually optimized manually to maximize homology. In most cases, the alignments were unambiguous in the smaller data sets, and separate analyses of exons and introns were unnecessary. The phylogenetic analyses for sect. Viridicata , the most speciose section in the subgenus, were divided into three sets of species, and the results are shown in Figs. 2-4. Penicillium gladioli (sect. Penicillium, ser. Gladioli ) was used as outgroup for these analyses based on its position in Fig. 1. The species formed a monophyletic group, and was apparently related to sect. Viridicata based on consensus but not bootstrap support. Section Viridicata ser. Viridicata formed a monophyletic group in Fig.1 (86% bootstrap support), and was divided into three subclades supported by strict consensus in Fig. 2, which shows one of eight MPTs. The largest clade included P. tricolor, P. freii, P. neoechinulatum, P. viridicatum and P. aurantiogriseum. Penicillium melanoconidium formed its own clade, and the sister group relationship with the P. polonicum/P. cyclopium clade suggested in Fig. 2 has no bootstrap or consensus support. All but two species in the series had 100% consensus support and there was strong bootstrap support for P. tricolor (100%), P. melanoconidium (99%), P. polonicum (87%) and P. cyclopium (99%). The sequences of the three strains of P. aurantiogriseum (79%) were identical. This clade was basal to a poorly supported clade (65 %) including strains of three species, P. freii , P. neoechinulatum and P. viridicatum . All strains of P. freii and P. neoechinulatum had identical sequences. The three strains of P. viridicatum each had unique sequences, with the two extra strains differing from the type in two and three positions. These three strains did not form a monophyletic clade. Fig. 3 represents one of 5000 saved MPTs. The topology of the presented phylogram suggests that ser. Camemberti, Solita and Corymbifera may be polyphyletic as currently delimited, but the back bone of the tree had weak bootstrap and consensus support. All species had 100% consensus support, except as noted below. Within ser. Camemberti , P. palitans (83%) was reasonably well-supported by bootstrap. All strains of Penicillium camemberti and P. caseifulvum had identical sequences to each other, and to the type strain of P. commune . Two strains of P. commune differed from the type by one bp substitution and thus formed a clade separate from the type. A constraint analysis enforcing the monophyly of strains identified as P. commune was accepted (3 steps longer than the MPTs, P=0.0832). Penicillium crustosum , a strongly supported species (99%) was considered a member of ser. Camemberti by Frisvad & Samson (2004), but based on consensus support was a part of ser. Solita in this analysis. The strongly supported species P. cavernicola (95%), considered part of sect. Solita by Frisvad & Samson (2004), appeared phylogenetically related to ser. Camemberti here, but without bootstrap or consensus support. Within ser. Solita , P. echinulatum (98%) was also a strongly supported, monophyletic species. Penicillium discolor and P. solitum both formed monophyletic groups in all MPTs, but received weak bootstrap support. These three species formed a distinct clade in the strict consensus analysis, which was weakly supported by bootstrap. Their separation of this clade from ser. Camemberti was apparent with consensus support. Species assigned to ser. Corymbifera occurred in four different monophyletic clades, all emerging from the backbone of the strict consensus tree. All the species of ser. Corymbifera in this phylogram were monophyletic based on strong bootstrap and strict consensus support . Penicillium albocoremium and the well-supported P. allii (99%) together comprised a fairly well-supported monophyletic pair of sister species, on a relatively long branch. Similarly, P. tulipae and P. radicicola were a well-supported clade (93%). Penicillium radicicola CBS 112425 occurred basal to the main P. radicicola clade that included the ex-type, differing by 3 bp. The β-tubulin sequence of this strain was more similar to P. tulipae than to the ex-type of P. radicicola. Penicillium hordei forms a well-supported monopheletic clade and is the only species included in ser. Corymbifera by Frisvad & Samson (2004) that does not attack plant bulbs. Constraint analyses were run to enforce the monophyly of ser. Camemberti, Corymbifera and Solita as defined by Frisvad & Samson (2004). The constraint for ser. Camemberti resulted in trees 3 steps longer than the unconstrained MPTs, and was accepted (P=0.0832). The Corymbifera constraint was identical to the consensus of the MPTs. A constraint enforcing the monophyly of ser. Solita was also accepted (trees 2 steps longer than the MPTs, P=0.1587). The phylogenetic analysis of sect. Viridicata ser. Verrucosa is presented in Fig. 4, which is one of 6 MPTs. This series was monophyletic in Fig. 1 (96%) and comprised two well-supported subclades in Fig. 4. One of these was P. thymicola (87%). The other consisted of P. nordicum , paraphyletic because of a well-supported, nested monophyletic clade consisting of two strains of P. verrucosum (87%). Some of the strains of P. nordicum (CBS 109541, 109538) differed from the type of P. nordicum (CBS 112573) by one bp, and had identical sequences to another strain isolated from jam (CBS 109538). Section Roqueforti was supported with 100% bootstrap support in Figs. 1 and 5. Each of the three species in Fig. 5, the single MPT from this analysis, was well-supported by bootstrap, as was the structure of the species phylogeny. The three strains of P. roqueforti isolated from blue cheese (CBS 135.67, 234.38, 221.30) were clearly conspecific with a strain from mouldy bakers yeast (CBS 479.84). The relationship between sect. Roqueforti and Eupenicillium osmophilum was supported by 72% bootstrap in Fig. 1. Eupenicillium egypticum was included as an outgroup in Fig. 5 based on its position in Fig. 1, but the species is not particularly close phylogenetically to sect. Roqueforti . The phylogenetic relationships among the five sections of sect. Penicillium are shown in Figs. 6 and 7. Fig. 6 represents the single MPT from the analysis. Ser. Expansa was paraphyletic as circumscribed by Frisvad & Samson (2004), with the monophyletic ser. Italica (89%) derived from within it. A constraint analysis enforcing the monophyly of ser. Expansa resulted in an acceptable tree 3 steps longer (P=0.2630). The species of ser. Expansa were generally strongly ( P. marinum , P. sclerotigenum , both 100%) or moderately well-supported ( P. expansum 70%) by bootstrap. Of the two species in series Italica, P. ulaiense (98%) received strong bootstrap support. The phylogeny of ser. ...