Parasitoid-Host Cophylogenetic Analyses
When cospeciation between parasites and their hosts occurs, we expect
congruence between parasitoid and host phylogenies and a 1:1
relationship between the interacting species, demonstrating strong host
specialization (Balbuena, Míguez-Lozano, & Blasco-Costa, 2013;
Fahrenholz, 1913; Legendre, Desdevises, & Bazin, 2002; Page, 2003).
Given the degree of host flexibility observed within Eadyaparasitoids (Peixoto et al., 2018; Ridenbaugh et al., 2018),
cospeciation with their hosts was not expected. To test this, two
cophylogenetic methods were used, an event-based and distance-based
method. The former compares parasitoid and beetle tree topologies and
their associations to reconstruct cospeciation, duplication, host
switch, losses, and failure-to-diverge events, while the later uses
distance matrices produced from the parasitoid and beetle phylogenies
and their associations to statistically test for congruences between the
two topologies. In the event-based reconstruction cospeciation events
occur when the parasite speciates in unison with the host. Duplication
events are when speciation in the parasite occurs independently from the
host. Host switching is the result of a duplication event in which the
parasite jumps to a new host lineage. Loss and failure-to-diverge events
are very similar in that both occur when the host speciates, but the
parasite does not. However, in loss events the parasite only remains on
one of the new host lineages, and in failure-to-diverge events the
parasite remains on both host lineages. For the event-based analysis
Jane 4 (Conow, Fielder, Ovadia, & Libeskind-Hadas, 2010) was used, with
the genetic algorithm set to run for 200 generations with a population
size of 5000 using two different cost models. The first model was the
default cost model for Jane, with cospeciation set to 0, duplication and
host switching set to 2, and all other event types set to 1. An
alternative model was used with cospeciation set to -10 and all other
events set to default to test the algorithms sensitivity to the cost
settings. For this method the input was trees drawn by hand using the
tree editor within Jane 4 and with clades collapsed for each species to
control for the potential overestimation of cospeciation events due to
sampling bias (Bass, 2019). For the distance-based method, a PACo
analysis (Balbuena et al., 2013) was used as implemented in R (R Core
Team, 2016) using the R package “PACo” (Hutchinson, Cagua, Balbuena,
Stouffer, & Poisot, 2017). PACo was run for 10,000 permutations with
the r0 randomization algorithm from the R package “vegan” (Oksanen et
al., 2016).