Supplementary Information for "Evolutionary instability of ectomycorrhizal symbioses in basidiomycetes", by David S. Hibbett, Luz-Beatriz Gilbert, and Michael J. Donoghue.

1. Higher-level phylogenetic relationships of homobasidiomycetes inferred from analysis of 60 specieswith no missing data.

Single most parsimonious tree, 6264 steps (CI=0.344). Bootstrap frequencies over 50% are indicated along branches. Only one branch, indicated by an asterisk, isincongruent with the strict consensus of trees from the complete analysis, in which some taxa lack one or more rDNA regions(Fig. 1).

2. Support for ancestral state assignments at selected internal nodes (Fig. 1) using maximum likelihood.

node / state according to parsimony / likelihood (-logL) when node is constrained to be non-ectomycorrhizal (0) or ectomycorrhizal (1)
-logL (0) / -logL (1)
1 / 0 / 106.6921* / 113.3172
2 / 0 / 106.5871* / 112.0533
3 / 0 / 106.5712* / 112.3768
4 / 0 / 106.8415* / 111.1048
5 / 0 / 106.3764* / 116.0775
6 / 0 / 106.6197* / 115.2121
7 / 0 / 106.7855* / 119.9181
8 / 0 / 106.7224* / 115.6681
9 / 0 / 106.7797* / 113.1879
10 / 1 / 111.1033 / 106.8645*
11 / 1 / 113.5189 / 106.6552*
12 / 1 / 113.0903 / 106.6215*
13 / 1 / 116.9668 / 106.8408*
14 / 1 / 115.2030 / 106.6488*
15 / 1 / 113.0445 / 106.6582*

* optimal state;  logL>2.

3. Results of constrained phylogenetic analyses and maximum likelihood analyses of character evolution.

We used analyses under monophyly constraints to generate alternative topologies that suggest that ectomycorrhizae evolved once or more, but have never been lost. Constraint trees were constructed in MacClade1, and constrained phylogenetic analyseswere run in PAUP* 4.02 with the same settings as the unconstrained analyses. The numbers of losses and gains on the constrained trees were estimated using equally-weighted parsimony, and constrained vs. unconstrained topologies were evaluated with the Kishino-Hasegawa maximum likelihood test3. Five constraints were employed: 1) all ectomycorrhizal species forced to form a single monophyletic group (constrained trees imply one gain and no losses); 2) ectomycorrhizal species of group 1, group 5, and group 6 forced to form three separate monophyletic groups (constrained trees imply 7 gains and no losses); 3) ectomycorrhizal species of group 1 forced to form a monophyletic group (constrained trees imply 7-9 gains and 1-3 losses); 4) ectomycorrhizal species of group 5, group 6, the euagarics clade, and the bolete clade forced to form four separate monophyletic groups (constrained trees imply 8 gains and no losses); and, 5) ectomycorrhizal species of group 5 plus Sphaerobolus, group 6 plus Multiclavula, the euagarics clade, and the bolete clade forced to form four separate monophyletic groups (constrained trees imply 9-12 gains and 0-3 losses). Trees produced under monophyly constraints ranged from 45-352 steps (0.5%-3.4%) longer than the unconstrained trees, and all were rejected using the Kishino-Hasegawa test3 (p=0.0014-p<0.0001). Thus, alternative topologies that suggest that ectomycorrhizae have never been lost were rejected.

Nodes 10-13 are resolved in all equally parsimonious trees (Fig. 1), but are weakly supported by bootstrapping. To assess the sensitivity of our conclusions to the resolution of these nodes, we performed four separate analyses under inverse monophyly constraints that result in topologies that do not resolve nodes 10-13 (one constrained analysis was performed for each node, using the same settings as the unconstrained analyses). Trees produced under inverse monophyly constraints were 6-17 steps (0.05-0.2%) longer than unconstrained trees, and could not be rejected with the Kishino-Hasegawa test3 (p>0.05). Trees produced under inverse monophyly constraints suggest that there have been 7-16 gains and 1-10 losses of ectomycorrhizae (average: 12.7 gains and 4.3 losses). Therefore, the conclusion that there have been multiple gains and losses of ectomycorrhizae is not sensitive to the resolution of nodes 10-13.

Finally, we used maximum likelihood analyses in Discrete4 to evaluate alternative models of the evolution of ectomycorrhizae that either allow or do not allow losses, using one tree from the unconstrained analysis and one tree from each of the four analyses under inverse monophyly constraints (described above). Parameters of the models were either unrestricted, or the rate of loss was set to zero. A difference of two units of log likelihood was taken as a criterion of significance4. On all trees examined, models of evolution of ectomycorrhizae that allow reversals were significantly more likely than models that do not allow reversals (∆ logL>2). The view that the evolution of ectomycorrhizae is irreversible was rejected on a range of optimal and suboptimal trees.

References

1. Maddison, W. P. & Maddison, D. R. MacClade version 3 (Sinauer Associates, 1992).

2. Swofford, D. L. PAUP* 4.0b2a (Sinauer Associates, 1999).

3. Kishino, H. & Hasegawa, M. Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order in Hominoidea. J. Mol. Evol.29, 170-179 (1990).

4. Pagel, M. The maximum likelihood approach to reconstructing ancestral character states of discrete characters on phylogenies. Syst. Biol. 48, 612-622 (1999).