Patterns of host–parasite associations in tropical lice and their passerine hosts in Cameroon

Abstract Coevolutionary processes that drive the patterns of host–parasite associations can be deduced through congruence analysis of their phylogenies. Feather lice and their avian hosts have previously been used as typical model systems for congruence analysis; however, such analyses are strongly biased toward nonpasserine hosts in the temperate zone. Further, in the Afrotropical region especially, cospeciation studies of lice and birds are entirely missing. This work supplements knowledge of host–parasite associations in lice using cospeciation analysis of feather lice (genus Myrsidea and the Brueelia complex) and their avian hosts in the tropical rainforests of Cameroon. Our analysis revealed a limited number of cospeciation events in both parasite groups. The parasite–host associations in both louse groups were predominantly shaped by host switching. Despite a general dissimilarity in phylogeny for the parasites and hosts, we found significant congruence in host–parasite distance matrices, mainly driven by associations between Brueelia lice and passerine species of the Waxbill (Estrildidae) family, and Myrsidea lice and their Bulbul (Pycnonotidae) host species. As such, our study supports the importance of complex biotic interactions in tropical environments.


TA B L E 1 (Continued)
(intrahost speciation), and cohesion (failure to speciate), comparisons of host and parasite phylogenies can be used as a cue for revealing the role of cospeciation and host switching in a given host-parasite system (Page, 2003).
Feather lice represent a convenient, repeatedly used model for cospeciation studies as they are regularly host specific, their entire life cycle takes place on the body of a single host, their survival outside the host is limited, and they are predominantly transmitted vertically between parents and offspring (Price, Hellenthal, Palma, Johnson, & Clayton, 2003). Cospeciation analysis has frequently been applied to feather lice and their avian hosts (de Vienne et al., 2013; Table 1), the results indicating a wide spectrum of potential processes that drive the patterns of host-parasite associations. While incongruences between phylogenies of some feather lice and their hosts suggest that host-parasite associations were mainly established through host switching (e.g., Banks, Palma, & Paterson, 2006;Johnson, Adams, & Clayton, 2002;Weckstein, 2004), phylogenies of other louse groups strongly mirror the phylogenies of their hosts and hence advocate a predominant role for cospeciation (e.g., Page et al., 2004;Paterson, Wallis, Wallis, & Gray, 2000). In addition to differences in the methodological approaches used in cospeciation studies, various parasite species' life-history traits may affect the ratio between cospeciation and host switching during the formation of host-parasite associations (Clayton, Bush, & Johnson, 2004). For example, while parasite physiological adaptations to the host apparently support cospeciation (Clayton, Bush, Goates, & Johnson, 2003), phoresis (mechanical transport by louse flies) favors host switching (e.g., Harbison & Clayton, 2011;Johnson et al., 2002). On the other hand, host life-history traits may affect the frequency and pattern of host switching. According to the "resource tracking hypothesis," a parasite should switch to a new host on which it can continue to exploit the same resources (Timm, 1983). Exploitation of the new host may be thwarted, however, by a difference between the former and new host that increases with their phylogenetic distance (Engelstädter & Hurst, 2006). The importance of host relatedness has been demonstrated by "natural" experiments, in which lice fail to establish on brood parasites (e.g., cuckoos and indigobirds) despite close contact between the young brood parasites and foster parents in the nest (Balakrishnan & Sorenson, 2007;Brooke & Nakamura, 1998 (Clayton, Bush, et al., 2003;Tompkins & Clayton, 1999). On the other hand, as lice are parasites with limited dispersal ability, patterns of host shifting will be greatly affected simply by the probability of encountering new hosts (Clayton et al., 2004 (Sweet, Chesser, & Johnson, 2017).
Tropical host populations are also typically less dense and abundant than temperate zone ones (e.g., Brown, 2014) and may not represent a reliable or abundant resource. This may favor generalist parasites in the tropics which makes cospeciation less likely (Combes, 2001;Vázquez, Poulin, Krasnov, & Shenbrot, 2005).
Lice may also be significantly limited by abiotic factors (Malenke, Newbold, & Clayton, 2011;Moyer, Drown, & Clayton, 2002;Rai & Lakshminarayana, 1980); hence, the high humidity and temperatures of the tropics may increase louse survival off the host, thereby facilitating host switching. Conversely, the stable conditions prevalent in the tropics (i.e., less pronounced seasonality and glacial periods), along with the higher longevity of tropical birds (Snow & Lill, 1974;Wiersma, Muñoz-Garcia, Walker, & Williams, 2007), could result in tighter parasite-host specialization, which would decrease the success of new host colonization.
The prevailing role of host switching in the tropics for forming feather lice and bird associations is supported by the study In this study, we analyze the coevolutionary processes that drive the patterns of host-parasite associations in two feather louse groups and their hosts in tropical lowland and montane forests in Cameroon (West-Central Africa). We assess the congruence of parasite and host phylogenies and attempt to find associations that contribute to the cophylogenetic structure.

| Sample collection
Birds were mist-netted and blood-sampled at two locations in the Lice were collected from the hosts using the "fumigation chamber method" (Clayton & Drown, 2001), followed by manual inspection of the host's head plumage. Lice were stored in ethanol and subsequently classified into genera using morphological criteria (Price et al., 2003).  (2017)), each representing one of the two feather lice suborders, that is, Amblycera and Ischnocera, respectively.
Myrsidea lice are host-specific parasites found predominantly on tropical passerine species (Figure 1), though they were found also on toucans and hummingbirds (Price et al., 2003). Including more than 380 mostly neotropical described species, Myrsidea is one of the most specious phthirapteran genera (Kolencik et al., 2018). They seem to be intolerant to low humidity (Bush et al., 2009), feed on host feathers, and partially utilize host body fluids, including blood (Marshall, 1981).
On the contrary, lice of the Brueelia complex are common in both the tropics and temperate zones, and they are less host-specific and, in addition to passerines, parasitize other bird groups, including Coraciiformes, Trogoniformes, and Piciformes (Gustafsson & Bush, 2017;Price et al., 2003). So far, over 426 species of this complex have been described (Gustafsson & Bush, 2017). Some Brueelia complex species are also capable of phoresis (horizontal transfer by hitchhiking) on louse flies (Hippoboscidae), which may eventually result in transport between different avian species due to the low specificity of louse flies (Keirans, 1975).

| Molecular methods and species delimitation
Louse DNA was extracted using the Qiagen DNeasy Blood and Tissue Kit (Qiagen), following the manufacturer's protocol. To increase the DNA yield and preserve the parasite's morphological features, each louse was pierced with an entomological pin prior to incubation in proteinase K solution at 56°C for 36 hr. The exoskeleton was then removed and kept as a voucher specimen.
For species delimitation, we used partial sequences of cyto- Distance matrices, histograms of pairwise nucleotide distances, and COI trees are provided in File S1-S6. According to ABGD results, we classified lice into groups characterized by intragroup COI sequence distances up to 3%. The groups were considered as unique evolutionary units and are hereafter referred to as species. A single individual of each species was used for subsequent cophylogenetic analyses. A description of new species will be given elsewhere (Sychra O., Gajdosova M., Andresova P., Albrecht T. & Munclinger P.,unpublished data).
Partial sequences of COI, wingless (wg), and 18S rDNA were sequenced in lice of both groups. In addition, partial sequences of the elongation factor 1 alpha (EF1α) and hypothetical protein EOG9X3HC5 (hyp) were obtained from Myrsidea and the Brueelia complex, respectively (see Table 2 for primer details). PCR conditions were identical for all loci. Amplification began with 1 min of denaturation at 94°C, followed by 35 cycles of 30 s of denaturation at 92°C, 40 s of annealing at 54°C, and 90 s of elongation at 65°C, the final step comprising 10 min of final extension at 72°C.
Owing to amplification problems, we used both original and redesigned forward primers for amplification of 18S rDNA and wingless (

| Cospeciation analysis
Cophylogenetic history was reconstructed in Jane 4 (Conow, Fielder, Ovadia, & Libeskind-Hadas, 2010), which accepts multihost parasitism. Jane implements a reconciliation algorithm to find the most optimal scenario of cophylogenetic past. By assigning costs to events which could possibly happen during the host-parasite cophylogenetic history (e. g., cospeciation, sorting events, lineage duplication, host switching, parasite's failure to diverge), Jane finds the least costly scenario that explains the observed situation.
Event costs were left as default, that is, cospeciation 0, duplication 1, duplication with host switching 2, loss 1, and failure to diverge 1. The analyses were run for 30 generations with a population size of 1,300. To test whether the reconstructed solution was better than scenarios expected by chance, we compared the cost of the reconstructed scenario with costs of 999 pseudorandom replicates generated using the "random tip mappings" approach. Tanglegrams visualizing host-parasite associations and phylogenies were created in TreeMap3 (Charleston & Robertson, 2002). Codivergence between both groups was further tested using the PACo script (Balbuena, Míguez-Lozano, & Blasco-Costa, 2013), using the APE (Paradis, Claude, & Strimmer, 2004) and VEGAN (Dixon, 2003) packages in R version 3.5.1 (R core Team, 2017). PACo is a specific case of Procrustean analysis, which generally assesses the level of congruence between two (or more) ordinations of multivariate data sets. More specifically, PACo is designed to test for congruence between genetic divergence of hosts and parasites. First, we calculated cophenetic distances separately for hosts and parasites based on branch lengths in corresponding phylogenetic trees.
Subsequently, principal coordinate analysis (PCoA) with Cailliez correction for negative eigenvalues was applied to extract orthogonal gradients (i.e., PCoA axes) from the two distance matrices. Scores for PCoA axes were used as an input for Procrustean superimposition assessing phylogenetic codivergence between hosts and parasites. Significance of the codivergence was tested by permutations of PCoA-scaled distances (100,000 random rearrangements with significance level being set a priori as 0.05) as described in Balbuena et al. (2013). We also extracted squared residuals from the PACo fit to assess contributions of individual host-parasite links to the final Procrustean superimposition.
As cophenetic distances were not available for K. poliothorax host species due to correction of its position in the tree, we omitted this species and its parasites from the PACo analysis.

| RE SULTS
In total, 626 birds of 78 passerine species were examined for lice.
Thirty-nine birds were parasitized by Myrsidea lice (prevalence 6.2%) and 52 by lice of the Brueelia complex (prevalence 9.9%; File S12). Parasite loads were relatively low and varied between 1-38 for the Brueelia complex and 1-10 for Myrsidea. The majority of parasite species were found on a single host species; however, 1 of 14 Myrsidea species was found on two bird species, which involved hosts belonging to the same family ( Figure 3). More cases of multihost parasites (4 of 15) were found within the Brueelia complex and involved associations with hosts from different families in two cases ( Figure 4). One species from the Brueelia complex was even found on hosts of different orders, that is, the Bangwa Warbler

F I G U R E 1 Cryptospiza reichenovii and its Myrsidea parasite
The most parsimonious scenario found for the Brueelia complex and its hosts comprised 5 cospeciation events, 0 duplications, 9 host switches, 4 sorting events, and 4 failures to speciate ( Figure 4). Hence, the frequency of parasite cospeciation (29%) appears to be slightly lower than in Myrsidea, though the overall cost of the scenario was significantly lower than expected by chance (i.e., Jane did not find the same or lower cost in any of 999 randomly permuted samples). There was a significant congruence between host and parasite distance matrices (the goodness-of-fit value was 34,205.59 with p < .001 based on 100,000 permutations; Figure 5), with the association between Waxbills (Estrildidae) and their parasites contributing most strongly to the overall congruence pattern (File S9).

| D ISCUSS I ON
Here, we analyze for the first time the host-parasite associations between lice and their avian hosts in the Afrotropical region.
Several species of lice were detected on more than one host species; moreover, it should be noted that our sample was geographically restricted, and hence, the actual number of parasite multihost interactions may have been underestimated. The lower specificity of Brueelia complex lice, which were even found on phylogenetically distant hosts, can be at least partly ascribed to their ability to transfer horizontally between hosts (Keirans, 1975). We also found one Myrsidea species (7%) on two host species. Our study was lim-  Schemske, Mittelbach, Cornell, Sobel, & Roy, 2009). Under strict cospeciation scenarios, one would expect unique (one-to-one) parasite-host associations (Lyal, 1986).
However, the number of host switches found in this study was higher than the number of cospeciation events, even though the event costs were set higher for host switching than cospeciation. Thus, our results are in agreement with previous evidence TA B L E 2 Primers used for obtaining partial sequences of the elongation factor 1 alpha (EF1α) and hypothetical protein EOG9X3HC5 (hyp)  Cho10 ACRGCVACKGTYTGHCKCATGTC Danforth and Ji (1998)  hole nesters (Timm, 1983;Weckstein, 2004 (Dumbacher, 1999). Furthermore, the survival of lice during such horizontal transfers may be higher in the tropics due to increased temperature and humidity. Finally, lice may also be transmitted through direct contact between hosts in mixed-species feeding flocks or at watering places.

TA B L E 3 Models used for alignment subsets
The apparent incongruence between parasite and host phylogeny in Myrsidea lice and their hosts appears rather surprising.
Myrsidea lice feed partially on blood (Marshall, 1981) and thus come into direct contact with the host's immune system. This may reinforce parasite coadaptation to a particular host and, as a result, lower the possibility of new host colonization. On the other hand, Clayton, Bush, and Johnson (2016)  and their louse parasites has rarely been studied, the few analyses undertaken thus far mostly show substantial incongruence between their phylogenies (Bueter et al., 2009;Johnson et al., 2002;Štefka et al., 2011;but see Sweet et al., 2018), in accord with our own results. Further, the concept of risk of extinction on small-bodied hosts fits well with our own findings, which suggest sorting as the prevailing event in the most parsimonious scenarios related to Myrsidea lice.
Despite the general incongruence between parasite and host phylogenies, PACo analysis showed a significant correlation between host and parasite phylogenetic distances, which may be at least partly interpreted through the prevalence of host switching to closely related hosts. The existence of such clade-limited colonization has already been suggested, for example, in brood parasites of genus Vidua and their passerine hosts (Sorenson, Balakrishnan, & Payne, 2004) or in Monogenoidea (Platyhelminthes) and their Neotropical fish hosts (Braga, Razzolini, & Boeger, 2015). Presumably, limited phylogenetic distances between hosts also reflect sharing of host traits, which allows the parasite to utilize the same resources on a new host.
As such, our results appear to be in accord with the "resource tracking hypothesis" (Timm, 1983). Nevertheless, the exact traits that facilitate host shifts remain unknown as related species tend to be similar in morphological, physiological, and behavioral features. On the other hand, congruence appeared to be higher in some host-parasite clades. Similar variation in host-parasite phylogenetic congruence has previously been recorded in Brueelia by . In our case, the congruence mainly concerned hosts, and Brueelia complex lice and Waxbills (Estrildidae).
Species within both these avian families are of similar size and body shape and have similar biology. They are also known to form flocks and sometimes even mixed-species flocks. While our analysis suggested only one cospeciation event in the Bulbul clade with Myrsidea lice, the majority of host speciations were accompanied by parasite cospeciation in lice from the Brueelia complex and Waxbills. Hence, it would appear that congruence was established through different evolutionary processes in these two parasite-host association groups.

ACK N OWLED G M ENTS
We thank the Congo Basin Institute/UCLA for facilitating the research in Cameroon. We also greatly acknowledge the help of our colleagues and local guides in the field and the assistance of Zdeňka Csibreiová in the laboratory. This study was supported by the Czech Science Foundation (GA ČR), project no. 17-24782S.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest. Tomáš Albrecht https://orcid.org/0000-0002-9213-0034