4.1 Hypogeococcus pungens species complex and host plant use
The specimens used to describe H. pungens were originally collected on an Amaranthaceae host (Granara de Willink, 1981). Further collections extended its host range to the Cactaceae, Apocynaceae and Portulacaceae families (Ben-Dov, 1994; Claps & de Haro, 2001; Zimmermann, Pérez Sandi Cuen, Mandujano, & Golubov, 2010), suggesting that H. pungens was a polyphagous species. However, our population genomics approach pointed to a clear separation between the mealybugs associated with Amaranthaceae or Cactaceae, and others that indiscriminately use Amaranthaceae and Portulacaceae species as hosts. These results confirmed that H. pungens was not a polyphagous species, but a species complex composed of cryptic species associated with different host plants as feeding resources. Population genomics, clustering and species delimitation analyses based on genome wide SNPs and the mitochondrial gene COI datasets revealed deep genetic divergence among populations formerly considered as part of H. pungens.
These results agreed with the previous studies that revealed a deep genetic divergence, asymmetrical prezygotic and postzygotic reproductive isolation between the mealybug populations associated with different hosts from Argentina, and differential preference and performance on alternative hosts (Poveda-Martínez et al., 2019; Aguirre et al., 2016). The extension of our sampling effort to a wider geographic area throughout the native and non-native ranges, along with the use of a substantially large number of nuclear markers, allowed the confirmation of the results reported for the Argentine Hypogeococcus mealybugs (Poveda-Martínez et al., 2019) and new insights of the evolutionary history of H. pungens sensu lato . In effect, our present results supported the hypothesis that genetic differentiation throughout the entire range in South America has mainly been driven by divergent selection imposed by alternate host plants rather than by isolation through distance. In fact, Mantel correlation and multiple regression analyses of genetic distance matrices on geographic and host plant distance matrices revealed that the latter was a better predictor of genetic divergence among populations (Table 2).
In this context, our results pointed to host plant use as an important driver of cryptic divergence in a priori presumed polyphagous insects. Cryptic divergence has been recurrently observed in the form of host-associated species that, very likely, diverged as the result of adaptation to alternate host plants with little morphological divergence in several insect species (Matsubayashi, Kahono, & Katakura, 2011; Bagley, Sousa, Niemiller, & Linnen, 2017; Forbes et al., 2017; Zhang, Bass, Fernández, & Sharanowski, 2018; Driscoe et al., 2019).
Genetic divergence between the cactus feeding mealybugs native to Argentina and Paraguay, and those introduced to Australia from Argentina (ArPaAu-C) and cactus feeding mealybugs from southeastern Brazil (BrPr-C) exceeded that which would be expected for conspecific populations, as indicated by our species delimitation analyses. In this context, it is worth mentioning that besides the large geographic distance separating populations of these clades and the fact that both feed on Cactaceae, these two putative species fed on cactus that belong to different genera. Mealybugs of the ArAuPa-C clade fed on species of the genera Cleistocactus and Harrisia , whereas BrPr-C thrived on species of other genera like Cereus for native Brazilian mealybugs, and various genera for non-native Puerto Rican pest mealybugs (Table S1). It may be argued that divergent selection on alternate hosts and allopatry appeared as the more likely drivers of divergence. As a matter of fact, there is evidence that divergent selection caused by alternate hosts acting on ancestral fragmented populations is sufficient to produce genetic divergence and incidental speciation (Nosil et al., 2012; Duque-Gamboa et al., 2018; Doellman et al., 2019).
Our results also elucidated a certain degree of genetic heterogeneity within the ArAuPa-C clade, which included cactus feeders sampled in Argentina, Paraguay, and those introduced in Australia for biological control of cactus weeds. Both mtDNA and nuclear SNPs revealed a certain degree of genetic structuring within this clade. Several well differentiated mtDNA haplotypes were recorded at different locations of this widespread clade, likely as the result of geographic isolation (Figure 2C). Also, the phylogenetic tree based on nuclear SNPs showed internal sub-clades concordant with the results depicted in theCOI network. An evaluation of the external morphology of specimens collected on cacti in the same localities sampled for the genomic survey in Argentina revealed subtle morphological variation (Lucía Claps, University of Tucumán, Argentina, personal communication), suggesting that genetic heterogeneity in this clade may be indicative of incipient speciation rather than within species heterogeneity. These results also helped to end the debate of the controversial origin of the mealybugs used in the biological control program in Australia (and in South Africa) (Tomley & McFadyen, 1987; Williams, 1973; Hamon, 1984). Indeed, Australian Hypogeococcus mealybugs were genetically close to the cactus feeding samples collected in northwestern Argentina and Paraguay, but not, as suggested by other authors, to Hypogeococcus festerianus Lizer & Trelles, another cactophagous species inhabiting central-western Argentina, or H. pungens (Julien & Griffiths, 1999; Zimmermann, Pérez, Cuen, Mandujano, & Golubov, 2010; McFadyen, 2012).
Overall, our data provided evidence of two factors that influenced at least 48% (Table 2) of the genetic divergence between the mealybugs considered as part of the H. pungens species complex. Host plant associations seemed to be the primary force influencing genetic divergence, followed by the limited gene flow induced by isolation by distance. Still, these results should be interpreted with caution since other factors could also affect the distribution of genetic variation that remained unexplained. Local adaptation to different environmental conditions and/or ecological interactions with natural enemies or competitors may impose varying selective pressures in different geographic locations. A recent survey of parasitoids and hyperparasitoids associated with Hypogeococcus mealybugs in South America identified Leptomastidea hypogeococci Triapitsyn (Hymenoptera: Encyrtidae) as a widespread primary parasitoid, able to attack all members of the H. pungens complex, whereasAnagyrus cachamai Triapitsyn, Logarzo & Aguirre and A. lapachosus Triapitsyn, Aguirre & Logarzo (also encyrtid primary parasitoids from Argentina and Paraguay) were only found associated toH. pungens sensu stricto and the ArAuPr-C clade (Triapitsyn et al., 2018). The influence of such differential ecological interactions with natural enemies might be affecting patterns of genetic divergence in this species complex.
Results of phylogenetic and species delimitation analyses were not entirely congruent. Analysis with mtDNA visualized four major clades whereas genome wide SNPs allowed to detect five well supported clades (Figure 3). The main difference being that Hypogeococcusmealybugs from northeastern Brazil, Puerto Rico and United States (BrPRUS-AP clade) and H. pungens sensu stricto (Ar-A) were different species according to nuclear SNPs, while these clades appeared collapsed in the same group with mtDNA data (Figure 3). Such mito-nuclear inconsistencies have often been reported in insects (Weigand et al., 2017; Hinojosa et al., 2019). Even though the mtDNA has been useful to trace the evolutionary history in many species groups (Hebert, Penton, Burns, Janzen, & Hallwach, 2004; Ball & Armstrong, 2006), incomplete lineage sorting and past hybridization events may obscure species delimitation based only on mtDNA data, particularly in recently diverged taxa (Hickerson, Meyer, & Moritz, 2006; Hinojosa et al., 2019; Després, 2019). For instance, Moreyra et al., (2019) reported a mitogenomic study in a cluster of closely related cactophagous Drosophila spp. inhabiting the southern cone of South America and found that the evolutionary history inferred by mitogenomes is not completely concordant with the phylogeny depicted by nuclear genomes (Hurtado, Almeida, Revale, & Hasson, 2019). The authors proposed that either incomplete lineage sorting (ILS) and/or introgressive hybridization could account for the pattern observed. A word of caution is needed before arriving at definitive conclusions in the present study due to limitations of the mtDNA dataset, that consisted of only a few hundred base pairs of the mtDNA gene encoding COI . In contrast, the evolutionary history depicted by the nuclear genomic dataset may be considered more reliable since it consisted of more than one thousand widely distributed SNPs.
Phylogenetic trees based on either nuclear SNPs or mtDNA indicated that feeding on Amaranthaceae was the more likely ancestral condition. However, results yielded by mtDNA and SNPs datasets were not completely congruent concerning the evolution of host plant use; both cactus feeding species formed a derived monophyletic clade in the tree based on nuclear SNPs, while cactophagy appeared to have evolved twice in the tree obtained with mtDNA data. However, to unveil the ancestral host in these clades, and to evaluate the evolution of host plant use and the biogeographic history of the genus in the continent, other well delimited cactophagous species of the genus Hypogeococcus from South America, such as Hypogeococcus spinosus Ferris and H. festerianus , should be included in an expanded analysis.