Page 5 - Delicado_Machordom_2015

D. Delicado et al. Evolutionary patterns of Pseudamnicola

integrates others published for hydrobiids (Wilke 2003; Afterwards, we conducted the analysis under the models:
Falniowski et al. 2008; Hershler & Liu 2008; Wilke et al. dispersal–extinction–cladogenesis model (DEC; Ree 2005),
2009) based on geological events. The substitution rates of and dispersal–vicariance analysis (DIVA; Ronquist 1997)
16S and 28S were estimated from this COI substitution and the BayArea model (Landis et al. 2013), both in a like-
value. lihood version (referred as DIVALIKE and BayAreaLIKE;
As prior of the topology of the tree, we used birth–death Matzke 2013). These models include two free parameters
model (Gernhard 2008), suitable for species-level phyloge- (d = dispersion or range extension and e = extinction or
nies with almost complete sampling. Nucleotide substitu- range contraction); however, BioGeoBEARS also includes
tion models obtained from jModelTest were applied to the the founder event j parameter for each model, resulting in
corresponding gene partition of the data set. We ran the three additional models: DEC +J, DIVALIKE +J and Bay-
analysis with Markov Chain Monte Carlo lengths of 50 Area +J. Here, we compared the model fit of these six
million generations, sampling every 2000 generations (ini- models through AIC. The most likely ancestral range was
tial 10% discarded as burn-in). The effective sample size of estimated for each node according to the most likely model
each parameter required to reach stationarity of the poster- and subsequently plotted on the maximum clade credibility
ior distribution (above 200) was examined in Tracer v. 1.5. tree of Fig. 3.
(Rambaut & Drummond 2009). The maximum clade credi-
bility tree of all sampled trees was compiled in TreeAnno- Results
tator v. 1.7.1. The topology of this tree as well as BPPs, Phylogenetic reconstruction and species boundaries
node ages and 95% high posterior density (HPD) intervals Mitochondrial data set. After combining COI and 16S
was finally visualized in FigTree v. 1.3.1 (Rambaut 2010). fragments, a total of 1170 characters were obtained, of
As independent analyses, we performed several Mantel which 658 were for COI and 512 for 16S. The COI data
tests (Mantel 1967) to assess the possible influence of some set contained 257 variable sites, 236 of which were parsi-
environmental factors, such as water conductivity, altitude mony informative. The 16S data set contained 169 variable
and geographic distance, on the genetic divergences sites, 118 of which were parsimony informative. In all
between Pseudamnicola individuals. As this method resulted reconstructions with this data set, the two subgenera of
in no significant correlations between the environmental Pseudamnicola were revealed as two well-supported mono-
variables and genetic divergences of individuals within the phyletic groups, with the exception of the species
same subgenus (Delicado et al. 2013, 2014), here we aimed P. (P.) gasulli that constituted an independent clade
at combining the information of each variable, yielded in (Fig. 2). However, the relationships between these groups
those publications, for P. (Corrosella) and P. (Pseudamnicola) were not obvious from the study of these mitochondrial
specimens simultaneously with the same purpose. A total of genes, and this was reflected in the position of P. (P.)
52 and 88 localities of Pseudamnicola were included for con- gasulli in the different tree topologies. In the COI topol-
ductivity and geographic variables (distance and altitude), ogy, P. (P.) gasulli was clustered within the subgenus
respectively. The significance of the correlation between P. (Corrosella) with high support values in all analyses; in
environmental and genetic variables was tested based on contrast, in the 16S reconstruction, this species was more
9999 permutations using the vegan package version 2.0-4 closely related to the other P. (Pseudamnicola) species,
(Oksanen et al. 2013) for the R statistical environment ver- although this was poorly supported in the ML analysis.
sion 2.15 (R Development Core Team 2011). Overall, the relationships among Pseudamnicola species
seemed better resolved in the COI analysis (Fig. 2). The
Historical biogeography percentage of sequence divergence was higher for COI
Ancestral range estimation for Pseudamnicola species was than for 16S, with maximum genetic divergence of 16.6%
performed in the R package BioGeoBEARS version 0.2.1 for COI (between P. (P.) granjaensis and P. (C.) marisolae)
(Matzke 2013). This method infers, in a likelihood frame- and 10.6% for 16S (between P. (P.) sp5 and P. (C.) sp1).
work, the potential ancestral areas by modelling events of
range evolution along a phylogeny. As input files, we Nuclear data set. The 28S alignment consisted of 204
included the species tree obtained in the program *BEAST sequences each with 1057 characters. Of these, 815 were
and a matrix with the distribution areas of the tips (species) invariant, 54 parsimony uninformative and 188 parsimony
of this tree. To designate discrete range states for each spe- informative. Interspecific genetic variation ranged from 0%
cies, we utilized the worldwide freshwater ecoregions pre- (as between P. (P.) beckmanni and P. (P.) granjaensis)to
sented in Abell et al. (2008) and depicted in Fig. 3. We set 6.9% (between P. (P.) gasulli and P. (P.) artanensis). The
the maximum areas per lineage and node at two, because ML and BI topologies recovered Pseudamnicola as a mono-
typically Pseudamnicola species occur in just one ecoregion. phyletic group, clustering P. (P.) gasulli with the other

ª 2015 Royal Swedish Academy of Sciences, 44, 4, July 2015, pp 403–417 407