Susceptibility and Diversity in the Therapy-Refractory Genus Scedosporium

Abstract

Scedosporium species show decreased susceptibility to the majority of systemic antifungal drugs. Acquired resistance is likely to disseminate differentially with the mode of exchange of genetic material between lineages. Inter- and intraspecific diversities of Scedosporium species were analyzed for three partitions (rDNA internal transcribed spacer gene [ITS], partial β-tubulin gene, and amplified fragment length polymorphism profiles), with the aim to establish distribution of resistance between species, populations, and strains. Heterogeneity of and recombination between lineages were determined, and distances between clusters were calculated using a centroid approach. Clinical, geographic, and antifungal data were plotted on diversity networks. Scedosporium minutisporum, Scedosporium desertorum, and Scedosporium aurantiacum were distinguished unambiguously in all partitions and had differential antifungal susceptibility profiles (ASP). Pseudallescheria fusoidea and Pseudallescheria ellipsoidea were indistinguishable from Scedosporium boydii. Pseudallescheria angusta took an intermediate position between Scedosporium apiospermum and S. boydii. Scedosporium boydii and S. apiospermum had identical ASP. Differences in (multi)resistance were linked to individual strains. S. apiospermum and S. boydii showed limited interbreeding and were recognized as valid, sympatric species. The S. apiospermum/S. boydii group, comprising the main clinically relevant Scedosporium species, consists of separate lineages and is interpreted as a complex undergoing sympatric evolution with incomplete lineage sorting. In routine diagnostics, the lineages in S. apiospermum/S. boydii are indicated with the umbrella descriptor “S. apiospermum complex”; individual species can be identified with rDNA ITS with 96.3% confidence. Voriconazole is recommended as the first-line treatment; resistance against this compound is rare.

INTRODUCTION

Members of the fungal genus Scedosporium are increasingly recognized as opportunistic agents of disease, e.g., in transplant recipients (1). Detection has been enhanced by the introduction of semiselective isolation media (2) and molecular detection methods (3). The latter have led to the recognition of Scedosporium as the second most prevalent mold genus (after Aspergillus) colonizing the lungs of patients with cystic fibrosis (4). Scedosporium colonization is a contraindication of lung transplantation, which is the ultimate treatment option for these patients (5, 6). Scedosporium species are unique as causative agents of the near-drowning syndrome (7, 8), a cerebral infection which is life threatening due to a lack of therapeutic options. The impacts of invasive scedosporiosis on costs of patient care and mortality in Australia were found to be significant (9). Scedosporium species and relatives, particularly Lomentospora prolificans (Scedosporium prolificans), are among the fungi with the highest degrees of resistance to a wide collection of antifungal compounds (10).

During the last decade, several molecular species have been recognized in Scedosporium. Several of these have gained wide recognition, particularly Scedosporium apiospermum, S. aurantiacum, S. boydii, S. dehoogii, S. minutisporum, and L. prolificans (11–14). The most common species, Scedosporium apiospermum and Scedosporium boydii, are molecular siblings which are morphologically identical (13). Degrees of intraspecific heterogeneity are variable between species. Gilgado et al. (13) already noted that S. boydii comprises lineages within larger mutual distances than S. apiospermum. In contrast, L. prolificans, compared with the same parameters, is nearly monomorphic (15). Scedosporium boydii is believed to be homothallic (13) and produces elaborate ascigerous fruit bodies under appropriate culture conditions, classically known under the generic name Pseudallescheria. Some decades ago, with only a few isolates available for study, minute differences in ascoma morphology were judged sufficient for the description of species such as Pseudallescheria angusta, P. ellipsoidea, and P. fusoidea (16). Today, morphological differences of these deviations appear close to the range of variability of prevalent Scedosporium species.

The aim of the present study was to evaluate degrees of genetic separation of species and lineages in Scedosporium and to compare these with distribution patterns of antifungal resistance, which may be understood better in the light of sexuality and gene flow. Genetic separation is established by analysis of the number of recombination events between molecular clusters, using sequences of the rDNA internal transcribed spacer gene (ITS) and the partial β-tubulin gene (BT2) and profiles generated by amplified fragment length polymorphism (AFLP) genotyping.

MATERIALS AND METHODS

Strains.

The studied Scedosporium test set for sequence analyses comprised a total of 123 strains (see Table S1 in the supplemental material). Of this set, 72 were of clinical and 51 of environmental origin. In total, 27 countries were represented by strains (see Table S1). The following type strains were included: P. angusta (CBS 254.72), S. apiospermum (CBS 117407), Pseudallescheria boydii (CBS 101.22), Scedosporium dehoogii (CBS 117406), Pseudallescheria ellipsoidea (CBS 418.73), Pseudallescheria fusoidea (CBS 106.53), Scedosporium minutisporum (CBS 116911), and Scedosporium desertorum (CBS 489.72). The set of type strains was supplemented by several authentic strains of Scedosporium deficiens (CBS 101723), 8 strains listed by Rainer and Kaltseis (14), and a set of reference strains that were verified by sequencing the rDNA internal transcribed spacer gene region and the partial β-tubulin gene region. All strains were identified down to species level using AFLP genotyping.

In vitro susceptibility data.

In vitro susceptibility data were taken from a previous study (10). In brief, data were generated according to CLSI guidelines (17) for all echinocandins, azoles, and polyenes. For analyses in the current study, voriconazole, posaconazole, anidulafungin, and micafungin were chosen due to their promising in vitro activity against Scedosporium strains. Caspofungin was excluded as it is currently not recommended as a compound for MIC determination (18, 19).

DNA extraction.

Strains were grown on Sabouraud's agar tubes at 30°C until sporulation was visible. Conidia were harvested and mechanically lysed using a MagNA lyser instrument (Roche Diagnostics) as previously described (10). DNA extraction and purification were performed with a MagNA Pure DNA isolation kit III (used as described in the manufacturer's manual) in a MagNA Pure LC instrument. All experiments on living material were performed according to international biosafety policies (biosafety level 2).

Sequencing.

For PCR amplification and sequencing, standard procedures were used. Briefly, for ITS amplification, primers V9 (5′-TGCGTTGATTACGTCCCTGC-3′) and RLR3R (5′-GGTCCGTGTTTCAAGAC-3′) were applied, and for the partial β-tubulin gene (BT2, covering exons 3 to 69), primers Bt2a and Bt2b were used (20). As sequencing primers, ITS1 and ITS4 (21) were used for ITS, and for BT2 the same primers were used as for amplification. Sequencing was done with an ABI 3700 XL instrument in combination with a BigDye v. 3.1 Terminator sequencing kit (Applied Biosystems, Carlsbad, CA, USA) according to the manufacturer's instructions. Electropherograms were edited using SeqMan v. 8.1 (DNAStar, Madison, WI, USA). To minimize the risk of PCR amplicon cross-contamination, pre- and post-PCR procedures were performed in physically separated facilities.

DNA sequence analysis.

Sequences were exported and aligned automatically in the Muscle package (www.ebi.ac.uk/tools/msa/muscle/). Initial identification of the strains was performed by use of a BLAST search using in-house research databases (Biolomics and BioNumerics) at the CBS-KNAW Fungal Biodiversity Centre (Utrecht, The Netherlands). Sequences were deposited in NCBI GenBank (see Table S1 in the supplemental material).

Amplified fragment length polymorphism.

Genotyping by AFLP was performed as described before by Lu et al. (22). However, for the selective amplification reaction, another set of primers was used, namely, 1 μM HpyCH4IV primer with one selective residue (underlined) (5′-Flu-GTAGACTGCGTACCCGTT3′) and 1 μM MseI primer with four selective residues (underlined) (5′-GATGAGTCCTGACTAATGAA-3′). Subsequent fragment analysis was carried out as described by Lu et al. (22).

Data analysis.

AFLP data were imported in Bionumerics v. 6.6 (Applied Maths, Sint-Martens-Latem, Belgium), and DNA fragments (range, 60–300 bp) were analyzed by unweighted-pair group method using average linkages (UPGMA) clustering using the Pearson correlation coefficient. AFLP fingerprints of poor quality were excluded from analysis. Homology to type/reference strains was used for identification. AFLP groups are listed in Table S1 in the supplemental material. Genetic relatedness was visualized by constructing a minimum spanning tree in Bionumerics v. 6.6. As input data for the minimum spanning tree, the similarity matrix of the UPGMA-based tree was used and treated as categorical information. Data on antifungal resistance were taken from Lackner et al. (10).

Sequence data were analyzed in MEGA v. 5.2 (http://www.megasoftware.net) and phylogenetic trees were constructed with maximum likelihood using the Tamura-Nei model TN93 and 1,000 bootstrap replicates. The built-in model test in MEGA v. 5.2 was used with the following parameters. A neighbor-joining tree was created by the program (statistical method based on maximum likelihood), gaps/missing data were treated as complete deletions, and the branch swap filter was set to moderate. The rate variation model allowed for some sites to be evolutionarily invariable (+I) (56.85% of sites). The tree with the highest log likelihood (ln L, −868.686) is shown. Groups in Table S1 in the supplemental material were based on this tree. ITS-based subdivisions were made on the basis of the same strains clustered by Ward's averaging in Bionumerics v. 6.6. DNASP v. 5.10 (http://www.ub.edu/dnasp) was used to calculate population-related statistics. The AFLP groups were considered separate populations. The neighbor-joining network tree was created with SplitsTree v. 4.8 (http://www.splitstree.org). The pairwise homoplasy index (PHI) test was also performed with SplitsTree v. 4.8. Haplotype networks were created with Network v. 4.6.10 by using a median-joining algorithm, with all single nucleotide polymorphisms (SNPs) having the same weight (http://www.fluxus-engineering.com/sharenet.htm).

A centroid approach was applied separately to sequences of BT2 and ITS subjected to pairwise comparison, and similarity values were calculated. The method was described by T. D. Vu and V. Robert (unpublished data) and was implemented in Biolomics (23). For each of the two loci, a similarity value was predicted which optimally supported the classification of sequences based on associated provisional taxon names (i.e., AFLP groups). The centroid sequence of each group is the one that is connected with all others of the same group at the highest average similarity value. Hypothetical clustering was compared in several combinations, focusing on the complex of S. apiospermum and S. boydii (AFLP1, 3, 6, and 9). Comparisons were made by either combining AFLP3 with 1/6 (S. boydii) or with 9 (S. apiospermum) or considering AFLP3 as a separate group.

MICs of antifungals as log₂ values were analyzed according to CLSI guidelines using broth microdilution. For statistical analysis, results were categorized as low MICs (L), medium MICs (M), and high MICs (H) based on MICs published by Lackner et al. (10). As no epidemiological cutoff (ECOFF) values, breakpoints, or clinical breakpoints have been published for Scedosporium, the categories were defined as follows: voriconazole (≤2 mg/liter, L [low MICs]; 4 mg/liter, M [medium MICs]; >4 mg/liter, H [high MICs]), amphotericin B (AMB) (≤2 mg/liter, L; ≥4 mg/liter, H), posaconazole (≤2 mg/liter, L; 4 mg/liter, M; >4 mg/liter, H), anidulafungin (≤1 mg/liter, L; 2 mg/liter, M; >2 mg/liter, H), and micafungin (≤1 mg/liter, L; 2 mg/liter, M; >2 mg/liter, H). Correlation, cross-resistance, and multiresistance were plotted in minimum spanning trees.

Susceptibility data (267 records) were analyzed retrospectively. Data were summarized in cross tables, and medians were calculated for MICs of amphotericin B, voriconazole, posaconazole, anidulafungin, and micafungin. Box plots of log-transformed values were used for graphical display. Since MIC data were not normally distributed (as determined with Kolmogorov-Smirnov tests), Kruskal-Wallis H tests were applied to test for differences between AFLP clusters to test for differences in MIC values. For categorical data, chi-square tests were used. P values were calculated with a two-sided significance level of α = 0.05. Statistical analyses for susceptibility data were performed using SPSS v. 20 software (SPSS, Inc., Chicago, IL, USA). AFLP-based clusters 1, 3, 4, 5, 6a, 6b, 9a, 9b, and “no cluster” were distinguished. P values below 0.05 were regarded as statistically significant.

RESULTS

An UPGMA tree generated with AFLP electropherograms (see Fig. S1 in the supplemental material) yielded 10 main groups (AFLP1 to -10) at a cutoff at 28% similarity, three of which were subdivided into approximate subclusters (AFLP2a to -c, AFLP6a and -b, and AFLP9a and -b) (see Table S1 in the supplemental material). Lomentospora prolificans (n = 39), a phylogenetically remote species, was included for comparison. Seven of the main groups contained at least one type strain of a described species. AFLP group 1 contained the type strain of P. ellipsoidea, group 2 of S. dehoogii, group 3 of P. angusta, group 6 of S. boydii, group 7 of S. minutisporum, group 9 of S. apiospermum, and group 10 of S. aurantiacum. Groups 4 and 5 were unnamed. The AFLP groups 9a and 9b were randomly distributed in trees of BT2 and ITS within S. apiospermum, while AFLP groups 2a and 2b did not match with any BT2 group within S. dehoogii/S. deficiens (see Table S1). Some heterogeneity with two groups (AFLP5 and 6) was noted in S. boydii below the cutoff; group 5 was recognizable as an unsupported subgroup in BT2 (see Fig. S2).

The BT2 data set had a total number of sites (excluding gaps/missing data) of 232 bp and a total number of mutations of 66; 56 sites were polymorphic (7 singletons and 49 parsimony informative), and 176 sites were monomorphic. Parsimony-informative sites consisted of 41 sites with 2 variants, 7 sites with 3 variants, and 1 site with 4 variants. We also determined nucleotide diversity per site (π = 0.04008), average number of nucleotide differences (k = 9.30), number of haplotypes (h = 36), and haplotype diversity (Hd = 0.952).

Phylogenetic sorting of BT2 sequences of 121 strains using Ward's averaging demonstrated the presence of nine main clusters approximately matching with AFLP groups 1 to 9. A maximum likelihood tree of the same data (constructed with the TN93+I model and 1,000 bootstrap replications, and S. desertorum CBS 489.72 as the out-group) exhibited the same topology and clade distinction (see Table S1 in the supplemental material), but degrees of statistical support of the nodes were variable (see Fig. S2). The groups contained the type strains of all Scedosporium species and are currently referred to as S. aurantiacum (AFLP10; BT2 bootstrap support 99%), S. minutisporum (AFLP7; 100%), S. apiospermum (AFLP9; 47%), S. dehoogii including S. deficiens (AFLP2; 89%), P. angusta (AFLP3; no bootstrap support), S. boydii (AFLP5 and -6; 84%), P. ellipsoidea (AFLP1; no bootstrap support), and an unnamed group (AFLP4; no bootstrap support). In contrast, several branches were supported that did not match with known species delimitations. AFLP3 matched with an unnamed clade with 83% support, two supported clades were found paraphyletic to AFLP group 9, and AFLP group 2 showed significant diversity (see Fig. S2).

Six of the groups listed above were separated from each other in BT2 by two or more fixed mutations, whereas two (AFLP groups 1 and 3) were characterized by a combination of mutations (see Table S2 in the supplemental material). Three groups differing by limited numbers of markers contained the type strains of P. ellipsoidea (AFLP1), P. angusta (AFLP3), and S. boydii (AFLP6). The number of fixed mutations was minor relative to the number of variables, while in the unambiguous species S. minutisporum and S. aurantiacum, the number of fixed mutations was equally as large as or larger than the number of polymorphic sites (see Table S2). Recombination events were calculated in the BT2 data according to Hudson and Kaplan (39) implemented in DNASP v. 5.10. In 232 positions analyzed (excluding sites with gaps/missing data), the number of pairs of sites with four gametic types was 88, yielding a minimum number of recombination events (Rm) of 9. In contrast, the pairwise homoplasy index (PHI) test (implemented in SplitsTree v. 4.8) yielded no significant evidence of recombination (P = 0.495). The average number of pairwise differences between and within populations (Fst) was calculated considering the AFLP groups as separate populations (Table 1). Group AFLP8 was not distinct from AFLP1 and 6. At a cutoff value of 0.25, no gene flow was detected between any of the populations. In a neighbor network constructed with SplitsTree v. 4.8, AFLP groups 4 and 9 were not separated (Fig. 1). Group AFLP3 remained separate from the combined groups 1, 6, and 9. The genetic isolation of populations is confirmed by the Fst values (Table 1), which in most cases are higher than 0.25, while AFLP1, 6, and 8 did not show evidence of population isolation. The same software was used to calculate evidence of recombination events.

TABLE 1.

Fst values considering AFLP groups as separate populations^a

AFLP (no. of sequences; no of haplotypes)	Fst value for AFLP (no. of sequences; no of haplotypes):
	AFLP2 (19; 7)	AFLP1 (15; 2)	AFLP3 (15; 5)	AFLP4 (3; 2)	AFLP5 (5; 1)	AFLP6 (20; 4)	AFLP7 (4; 2)	AFLP8 (2; 2)	AFLP9 (24; 6)
AFLP1 (15; 2)	0.86529
AFLP3 (15; 5)	0.80798	0.69255
AFLP4 (3; 2)	0.85538	0.8806	0.66369
AFLP5 (5; 1)	0.89129	0.90909	0.84736	0.94737
AFLP6 (20; 4)	0.84515	0.74178	0.7555	0.86914	0.87474
AFLP7 (4; 2)	0.89874	0.96346	0.91069	0.94815	0.98246	0.94254
AFLP8 (2; 2)	0.76902	0.0	0.42052	0.63158	0.33333	0.02729	0.84211
AFLP9 (24; 6)	0.83505	0.82062	0.6045	0.67277	0.89714	0.82849	0.92038	0.60203
AFLP10 (5; 2)	0.89906	0.97573	0.94587	0.96791	0.98529	0.96007	0.97572	0.89252	0.94449

^aAverage numbers of pairwise differences between and within populations established with DNASP v. 5.10. Threshold value is 0.25.

FIG 1.

Neighbor network of Scedosporium based on BT2 data constructed with SplitsTree v. 4.8 and compared with AFLP grouping. Note that in S. apiospermum/S. boydii, two branches are present with a heterogeneous Pseudallescheria angusta group in an intermediate position.

With ITS using different algorithms, only seven (A to G, including the out-group) of the nine main groups were recognized due to lower variability of this gene (see Table S1 in the supplemental material). Groups AFLP1, -3, -4, -5, and -6 were not distinguished, and group 3 deteriorated in two unsupported branches. The species S. aurantiacum, S. desertorum, and S. minutisporum were unambiguously separated at high-bootstrap values (data not shown). Group AFLP2, containing representative strains of S. deficiens and S. dehoogii, varied 2.7% in BT2 and 0.9% in ITS. ITS differences were based almost exclusively on variations of four fixed mutations. BT2 also yielded two main subgroups, differing by five fixed mutations. Three approximate AFLP groups (2a to -c) were recognizable, with tested strains of group 2a matching with S. deficiens. However, significant conflicts between topologies of the different partitions were observed among ITS, BT2, and AFLP (see Table S1), and therefore we found the two species inseparable.

The position of group AFLP3 was verified with a centroid approach based on ITS versus BT2. For BT2, the optimal predicted similarity value was 97.6%, while for ITS this value was 99.26%. Groups were clearly separated from each other, except P. angusta (AFLP3), which was not separated from S. apiospermum or from S. boydii with BT2 or ITS data, respectively. However, when P. angusta (AFLP3) was postulated as a part of S. boydii (AFLP1/6), BT2 sequences of P. angusta merged into the group of S. apiospermum (Fig. 2) at a similarity value of 99.03%, while with ITS, P. angusta merged with S. boydii at a similarity value of 99.26%. This suggests that AFLP3 takes an intermediate position between S. boydii and S. apiospermum. In Fig. 3, sequences associated with the same taxon share the same color, i.e., have similarity values greater than or equal to the optimal predicted similarity.

FIG 2.

Distribution of BT2 sequences representing multidimensional distances from centroid. Pseudallescheria angusta takes an intermediate position between S. apiospermum and S. boydii.

FIG 3.

Antifungal susceptibilities per strain plotted on minimum spanning tree generated from AFLP data. (A) Amphotericin B (AMB). Green, low MICs (L); yellow, medium MICs (M); red, high MICs (H). (B) Cross-resistance to echinocandins anidulafungin and micafungin. Bright blue, M/H (n = 2); blue, M/L (n = 22); yellow, insufficient growth/insufficient growth (n = 10); salmon pink, insufficient growth (failed to form sufficient amount of mycelium in growth control within 72 h in three biological replicas)/L (n = 6); red, H/H (n = 39); bright pink, H/M (n = 2); bright green, L/H (n = 1); purple, H/L (n = 34); green, L/L (n = 160). (C) Cross-resistance to azoles voriconazole and posaconazole. Yellow, insufficient growth/insufficient growth (n = 10); red, H/H (n = 1); brown, H/L (n = 1); purple, M/H (n = 6); bright green, L/M (n = 3); bright blue, L/H (n = 51); dark green, L/L (n = 205).

A minimum spanning tree of AFLP data (see Fig. S3 in the in the supplemental material) limited to the S. apiospermum/S. boydii group (AFLP groups 1, 3, 4, 5, 6, and 9) showed these species as a complex of large, variable clouds, with S. apiospermum (AFLP9) in particular as a widely variable species. In the S. boydii area, some groups were individualized, including AFLP groups 5 and 6 of S. boydii, AFLP3 with the type strain of P. angusta, AFLP1 with the type strain of P. ellipsoidea, and unnamed group AFLP5. The number of AFLP clusters in relation to the total sample size was large. When plotting geographic origin on the tree, no significant structuring was observed, except for some sampling bias because of overrepresentation of European strains (data not shown). A possible relationship with strain origin (clinical versus environmental) was not possible because of a strong sampling bias, with mainly clinical isolates having been collected systematically.

Kruskal-Wallis H tests comparing antifungal susceptibilities of clinical versus environmental strains over all AFLP groups yielded no statistically significant differences in MICs for amphotericin B, anidulafungin, micafungin, posaconazole, or voriconazole. This result was supported by chi-square tests using categorization (susceptible/intermediate/resistant [S/I/R]) instead of numerical MICs (see Tables S3a to c in the supplemental material). MIC values of Lackner et al. (10) comparing AFLP groups using Kruskal-Wallis H tests were found to differ in susceptibility against amphotericin B, anidulafungin, and micafungin (see Table S3). Specifically, pairwise comparisons with correction for multiple testing (Bonferroni) revealed significant differences for anidulafungin between clusters 9a/6b and for micafungin between clusters 1/6b, 1/3, 9a/6b, and 9a/3. For amphotericin B, no pairwise comparison was significant after Bonferroni correction. No statistically significant differences were observed for amphotericin B, voriconazole, and posaconazole.

Individual strains within some of the AFLP groups differed remarkably in susceptibility to antifungal compounds, ranging from low to high MICs. Figure 4 shows median values with first- and third-quartile log₂ MICs, with outliers and extreme outliers indicated. Distributions are visualized by plotting antifungal data (S/I/R) on minimum spanning trees of AFLP groups, exemplified by amphotericin B (Fig. 3A). In amphotericin B, 111 strains were found to have low MICs and 43 to have high MICs, while only 10 had medium MICs, underlining the pronounced differences in MICs between isolates. The distribution of strains with high MICs over the S. apiospermum/S. boydii complex is more or less random (Fig. 3A), as was also found with other drugs tested (data not shown). Multiresistance was analyzed by plotting correlations for azoles (respectively for echinocandins) on the AFLP minimum spanning tree of the S. apiospermum/S. boydii complex (Fig. 3A to C). In the comparison of the azoles voriconazole/posaconazole, most strains had low MICs for both compounds (L/L; n = 205), while 51 had high MICs for posaconazole, but high MICs for voriconazole were exceptional. With echinocandins, most strains had low MICs for anidulafungin/micafungin (L/L; n = 160); of the strains with high MICs for anidulafungin, many strains had low MICs for micafungin (H/L; n = 34), and only a few isolates had high MIC values for both substances (H/H; n = 39), indicating an absence of correlation of resistance.

FIG 4.

Box plots of log₂-transformed MIC values for the AFLP-based clusters. (A) Amphotericin B (AMB), (B) anidulafungin (ANI), (C) micafungin (MICA), (D) posaconazole (POS), and (E) voriconazole (VOR). Central lines in each box denote the median, and the lower/upper rims represent first/third quartiles (Q1/Q3); the whiskers extend to 1.5 times the height of the box (or to the minimum/maximum value). Outliers are shown by circles if values are up to three times the height of the boxes or by asterisks if values are more than three times the height of the boxes (extreme outliers).

DISCUSSION

Emerging resistance against antifungal agents is likely to disseminate differentially depending on the mode of exchange of genetic material between lineages, and therefore knowledge of patterns of diversity and mating is essential for understanding the spread of resistance over populations. Gilgado et al. (24) investigated mating systems in the genus Scedosporium and suggested that there were fundamental differences between species. In particular, S. boydii and S. apiospermum were judged to be homo- and heterothallic, respectively. Borderlines between species were judged to be clearly defined in a relatively limited data set.

The present study confirms earlier investigations in that Scedosporium (including its obsolete teleomorph name Pseudallescheria [38]) comprises a number of different species, separated from each other by strictly concordant multilocus parsimony-informative sequences and AFLP data. Variability is noted in both genes studied (BT2 as well as in ITS) and is also recognized in AFLP profiles, although degrees of variability differ between groups. Unambiguous separation was achieved between S. desertorum, S. minutisporum, and S. aurantiacum, while S. prolificans is also known to constitute a separate entity, recently even at the generic level L. prolificans (11, 25). The remaining species formed two complexes, one consisting of S. deficiens/S. dehoogii and another comprising S. apiospermum, P. angusta, P. ellipsoidea, and S. boydii. ITS is known to have a lower degree of discriminatory power than BT2. Two major groups were recognized in S. dehoogii when markers were studied individually, suggesting the existence of a second species, S. deficiens (14). However, despite the significant distances in both genes, multilocus analysis (ITS, BT2, and AFLP) showed that the groups were not concordant and recombination between the groups was frequent (6 recombinants in a total set of 20 strains) (see Table S1 in the supplemental material). The latter suggests that the more recently described S. deficiens (15) belongs to the same biological species as S. dehoogii.

Differences within the complex S. apiospermum/S. boydii were even smaller. The groups were difficult to delineate with BT2 data when distance-based algorithms were used, partly due to the fact that the number of singleton mutations led to an overabundance of SNP-based haplotypes. Fixed mutations were observed only when the complex was further subdivided into four entities, provisionally indicated as S. apiospermum (AFLP9), P. angusta (AFLP3), P. ellipsoidea (AFLP1), and S. boydii (AFLP5 and -6). With BT2 data with the Tamura-Nei TN93+GI model, where G is the gamma distributed rate variation among sites and I is the proportion of invariable sites (extent of static, unchanging sites) in a dataset, S. boydii deteriorated into three groups, two of which lacked statistical support (see Fig. S2), and included P. ellipsoidea (AFLP1) and P. angusta (AFLP3). In the neighbor network in SplitsTree v. 4.8 (Fig. 1) AFLP3 was an amalgamate of deviating strains which in a centroid approach were intermediate between S. boydii and S. apiospermum. Fst values suggest complete separation of lineages (Table 1). The complex thus consists of a series of very close lineages showing limited gene flow. Although DNASP v. 5.10 showed 9 recombination events, the PHI test showed no statistically significant evidence for recombination (P = 0.495) and hence the existence of intrinsic mating barriers is hypothesized. Possibly AFLP4 and AFLP8 are recombinant, suggesting mating between S. apiospermum and S. boydii at low frequency. Gilgado et al. (24) noted that two strains of S. apiospermum crossed with both plus and minus mating types. The sexuality of Scedosporium thus remains enigmatic.

While in heterothallic fungi recombination serves to exchange traits to bring about change, in homothallic fungi sexual reproduction aims to conserve recently manifested mutations for the next generation. Our study showed that recombination is rare between closely related entities in Scedosporium. Populations seem to expand sympatrically, suggesting high degrees of homothallism. There is no match of lineages in the S. apiospermum/S. boydii group with clinical data or geography (data not shown).

Scedosporium aurantiacum is known to differ from the remaining taxa in particular for patterns of susceptibility to posaconazole (10). Lomentospora prolificans is deviant in all respects, being one of the most recalcitrant fungi known (26). Differences between susceptibility patterns of S. apiospermum and S. boydii are very small (10); only differences between AFLP subgroups were revealed, possibly biased by sampling effects and insufficient sample sizes. We analyzed the differences in more detail by plotting profiles of resistance against antifungals on minimum spanning trees generated from AFLP data, exemplified by amphotericin B (Fig. 3A). Amphotericin B has been found to have limited efficacy against activity in cerebral scedosporiosis in animal experiments (27). With most of the compounds tested, strains were classified as either susceptible or resistant, with only a few isolates in between these extremes, and distributions over AFLP groups were largely random. Figure 3B and C demonstrate that there is little correlation between susceptibility profiles (multiresistance being rare), suggesting that azole versus echinocandin resistance is governed by different mutations.

Significant clinical differences between Scedosporium species and relatives are known to exist, in that, e.g., L. prolificans is mainly involved in disseminated infections in immunocompromised hosts and is not observed in the near-drowning syndrome (8). However, differences within the S. apiospermum/S. boydii complex are small. Kaltseis et al. (28) noted that S. boydii is more frequently found in clinical specimens than S. apiospermum. This difference was not observed by Lackner et al. (29). Bernhardt et al. (30) noted that S. boydii had an incidence equal to or lower than that of S. apiospermum in pulmonary colonization of patients with cystic fibrosis. In an attempt to quantify clinical significance of species, we distinguished four main types, environmental, traumatic (sub)cutaneous, respiratory colonization and infection, and systemic and disseminated infection. No significant difference was revealed within the S. apiospermum/S. boydii complex. This result, together with limited antifungal MIC differences between the species, has led to the conclusion that as yet no medically relevant difference between these two species is apparent. In the daily hospital routine, identification with 96.3% confidence using the standard bar-coding gene ITS is sufficient for patient management. Voriconazole is recommended as the first-line treatment (32); its activity against S. apiospermum in vitro, in vivo, and in patients was demonstrated in multiple studies (31–33). Resistance against this compound was very rare. Posaconazole was found effective against S. apiospermum complex members in vitro. According to in vivo studies (murine models) and patient case reports, this drug is effective (34–37), but in vitro susceptibility testing is mandatory due to the occurrence of resistance. Anidulafungin and micafungin may represent future therapeutic options, which, however, remain to be evaluated by in vivo experiments.

The main clinically relevant species S. apiospermum/S. boydii differ in anonymous markers, but antifungal resistance has a random distribution within the species complex. Distinction of individual species is useful for epidemiological questions but does not contribute significantly to patient care. Whether or not the species should be distinguished in daily routine practice is dependent on the diagnostic question. Species distinction does not predict antifungal resistance, and therefore testing of each strain prior to therapy remains compulsory.