Host Plant Variation and Lack of Genetic Differentiation in Populations of Dione (Agraulis) dodona Lamas & Farfán (Lepidoptera: Nymphalidae)

Simple Summary In this study, we analyzed the population genetics of Dione (A.) dodona Lamas and Farfán (Nymphalidae), a recently described butterfly from xerophytic environments of the western slopes of the Central Andes, associated with Malesherbia tenuifolia (Passifloraceae). Searches for additional host-plant species were carried out throughout its known distribution. In total, seven species of Malesherbia have been reported as host plants for D. (A.) dodona, extending its distribution as far north as 50 km. We obtained samples of the last larval instar in six localities on different host-plant species that have disjointed distributions. We inferred the genetic diversity and structure between populations by using microsatellites and mitochondrial COI markers. The results showed a weak genetic differentiation between the evaluated populations, suggesting a continuum of gene flow through high dispersal, along with the disjoint occurrence of the host plants. Abstract Dione (Agraulis) dodona (Nymphalidae: Heliconiinae) is a butterfly restricted to the western slopes of the Andes of Peru and Chile and is associated with Malesherbia tenuifolia in xerophytic environments. In this study, we found six additional species of host plants for D. (A.) dodona belonging to the genus Malesherbia (Passifloraceae). We used mitochondrial DNA sequences (COI) and microsatellites to screen genetic variation and investigate population structure in six geographic disjointed populations of D. (A.) dodona associated with distinct Malesherbia species. The PCoA analysis based on allele frequencies evidenced a lack of differentiation among populations and a low FST. The Bayesian cluster analyses revealed the existence of three genetically distinct groups, but almost all individuals present an admixture ancestry. An absence of isolation by distance pattern was observed. Possible scenarios are discussed: a bottleneck or recent colonization from interconnected populations from the south, and ongoing gene flow among local populations by high dispersal through a landscape formed of isolated populations of Malesherbia.


Introduction
The Neotropical butterfly genus Dione Hübner (Nymphalidae: Heliconiinae) includes a newly described species, Dione (Agraulis) dodona Lamas & Farfán, restricted to xeric slopes of the Andes in Peru and Northern Chile, associated with the host plant Malesherbia tenuifolia D. Don (Passifloraceae) [1]. The females of D. (A.) dodona oviposit on flowers or leaves near them, a behavior that is unique among Heliconiines, because the first instar feed on the internal structures of the flowers [1]. Malesherbia R. and P. is a genus distributed in the Central and Southern Andes in Peru, Chile, and Argentina, mostly in the lower parts of the western slopes on rocky and sandy soils, in arid and semi-arid habitats [2][3][4][5]. The subgeneric classification of Malesherbia comprises five sections, of which only the Malesherbia section is found in Peru and Northern Chile, with M. tenuifolia being a member, along with thirteen other species [6], mostly known from single localities and with discontinuous or isolated distributions [4][5][6]. The geographic range of D. (A.) dodona appears to be narrowly based on the description records besides the type locality; this butterfly has been considered rare [1,7], with adults collected only in two Regions in Northern Chile and four Departments in Peru, with about 1400 km between the most distant records [1].
As the northernmost occurrence of D. (A.) dodona in the Peru-Chile desert [1] exceeds the northern limit of its only documented host plant, M. tenuifolia [4], and the presence of additional populations of several species of Malesherbia between the Pacific Coast and the Andes of South-Central Peru and Northern Chile provides a potential for (i) wider geographic and potential additional host-plant ranges for this butterfly and (ii) the occurrence of isolated populations due to a discontinuous distribution of host plants. Accordingly, to verify the entire occurrence of D. (A.) dodona (considered a rare species) and to assess its host-plant range, it should be screened in habitats where Malesherbia species occur. Considering the patchy distribution of Malesherbia throughout extensive arid areas in the desert landscape of Western South-Central Peru and Northern Chile, together with the highly specialized egg-laying-site selection of the females, some degree of genetic differentiation might be expected for putative populations of D. (A.) dodona.The scale at which intraspecific gene flow occurs is an important property of a species, with implications for conservation. However, the spatial scale of genetic connectivity is often unknown since its assessment either requires large-scale mark-recapture studies or genetic data [8]. A simple marker used to infer genetic variation is the mitochondrial DNA (mtDNA) due to its suitable rate of mutation. However, the nature of inheritance (strictly maternal) and absence of recombination limit its use for gene-flow estimates. A more appropriate marker for population genetics and conservation studies is the microsatellites, given their high variability, biparental inheritance, and co-dominance [9,10]. However, microsatellites have been not widely used in studies of lepidopterans because of a paucity of these markers in their genomes [11]; that is, they are difficult and costly to develop. In Heliconiini, microsatellites have been developed for three species, two belonging to the genus Heliconius [12,13] and one to Dione [14]. Since cross-amplification was successful for Dione (Agraulis) vanillae [14], we consider it practicable to amplify polymorphic loci in populations of D. (A.) dodona.
In phytophagous insects, the distribution of intraspecific genetic diversity can be determined by the intrinsic and extrinsic factors of the organism. Life history can be of significant importance [15], but environmental factors mostly promote the relative isolation of populations, such as geographic distances, physical barriers to gene flow, habitat suitability and/or fragmentation, and the use of host plants [16]. Particularly, the host-plant fidelity has been demonstrated to play a major role in genetic isolation between populations [17].
Considering that the distribution of D. (A.) dodona [1] is much wider (ca. 1200 km) than that of M. tenuifolia [3,5], one would expect this butterfly is not monophagous but feeds on additional host plants. Several Malesherbia species were recorded by Beltrán et al. [4] for the area, but such plants have not been searched regarding usage by this butterfly. Thus, in this work, we studied the geographic and host-plant ranges of D. (A.) dodona and evaluated the   Table S1 for details of host plants' localities). Flowers of host plants with fifth instar larvae (blue close arrow) are shown on both sides of the map.

Sample Collection and Host-Plant Records
Fieldwork was performed from November 2020 to May 2022, screening for Malesherbia plants along the roadways, covering a latitudinal gradient of seven degrees (∼1200 km), from Central Peru to Northern Chile, on western slopes of the Andes, between 1000 and 3000 m elevation ( Figure 1). An intensive search for immature stages of D. (A.) dodona on flowers of Malesherbia spp. was carried out in each locality where plants were found. Larvae of the fifth instar of D. (A.) dodona were collected and preserved in 70% ethanol. Samples of Malesherbia from each locality were also collected and identified at the species level, based on the literature [3][4][5].  Table S1 for details of host plants' localities). Flowers of host plants with fifth instar larvae (blue close arrow) are shown on both sides of the map.

Sample Collection and Host-Plant Records
Fieldwork was performed from November 2020 to May 2022, screening for Malesherbia plants along the roadways, covering a latitudinal gradient of seven degrees (∼1200 km), from Central Peru to Northern Chile, on western slopes of the Andes, between 1000 and 3000 m elevation ( Figure 1). An intensive search for immature stages of D.   Table S1).  Table S1).

DNA Extraction, Polymerase Chain Reaction Amplification, and Sequencing
Genomic DNA was isolated from each specimen, using larvae tissue stored in ethanol 70%, according to Huanca-Mamani et al. [18]. Genomic DNA was sent to Macrogen Inc. (Seoul, South Korea) for PCR amplification, purification, and sequencing of the mitochondrial gene Cytochrome oxidase subunit 1 (COI), using primers HCO2198 and LCO1490 with conditions described by Folmer et al. [19]. Sequence alignments were performed by using CodonCode Aligner (CodonCode Corp., Centerville, OH, USA) and inspected manually.  Genomic DNA was isolated from each specimen, using larvae tissue stored in ethanol 70%, according to Huanca-Mamani et al. [18]. Genomic DNA was sent to Macrogen Inc. (Seoul, South Korea) for PCR amplification, purification, and sequencing of the mitochondrial gene Cytochrome oxidase subunit 1 (COI), using primers HCO2198 and LCO1490 with conditions described by Folmer et al. [19]. Sequence alignments were performed by using CodonCode Aligner (CodonCode Corp., Centerville, OH, USA) and inspected manually.
We surveyed 14 microsatellites developed in the genus Dione, which also proved to be polymorphic in the subgenus Agraulis [20]. An M13 tail (5 -TGTAAAACGACGGCCAGT-3 ) was linked at the 5 of the forward primer of each polymorphic loci. A primer mix consisting of M13 labeled with a fluorescent dye (6-FAM, HEX, or TAMRA) was used for PCRs were performed by using the following conditions: 95 • C for 2 min, followed by 35 cycles of 30 s of 94 • C, specific annealing temperature for 30 s, 1 min of 72 • C, and a final extension of 72 • C for 4 min [14]. All individuals were genotyped (multiplexed by using different fluorescence and size ranges), using a Liz GeneScan 600 (Life Technologies, Carlsbad, CA, USA) standard marker of molecular weight in an ABI Genome Analyzer by the commercial facility of Macrogen.

Data Analysis
Genetic diversity within each population was estimated by using the mean number of alleles per locus (allelic richness (A)), the percentage of polymorphic loci (%P), observed heterozygosity (H O ), and the heterozygosity expected from Hardy-Weinberg equilibrium (H E ) [21], using the software GenAlEx v 6.5 [22]. Genotypes were checked for the presence of null alleles, using the software Micro-Checker 2.2 [23]. Linkage disequilibrium and deviations from the Hardy-Weinberg equilibrium [24] were tested by using ARLEQUIN 3.5 [25]. Sequential Bonferroni corrections were applied to correct for multiple comparisons [26], with p = 0.05 to adjust the statistical significance levels. Standard genetic diversity indices, such as the number of COI haplotypes and haplotype diversity per population, were estimated for mtDNA sequences, using ARLEQUIN 3.5. The topological relationship between haplotypes was estimated by using the median-joining approach [27] implemented in NETWORK 4.1.
Genetic differentiation between the six populations was characterized by exact tests [28] and by estimating pairwise F ST [29], within ARLEQUIN 3.5. The levels of significance for multiple tests were adjusted by the sequential Bonferroni method [26]. We investigated whether inter-population genetic distances (based on microsatellite loci) increased linearly with geographic distance. Thus, genetic (F ST ) and geographic (km, linear) distances between populations were calculated in GenAlEx 6.5 and Google Earth 7.3, respectively. Pairwise genetic and geographic distance matrices were correlated by using a Mantel test, with a test for a significant relationship by 9999 random permutations that was also implemented in GenAlEx 6.5. To investigate the existence of hierarchical levels of population structure, an analysis of molecular variance (AMOVA) was performed by using allele frequencies with ARLEQUIN 3.5, following a geographic pattern of north (LIM, ARE, MOQ, and TAC) to south (CH1 and CH2) structure.
As an alternative approach to represent the genetic relationship among the six populations of D. (A.) dodona, a principal component analysis (PCoA) was applied by using gene frequencies of all six variable microsatellite loci. The frequencies of all the alleles at a single locus were considered independent variables. Accordingly, correlation matrices were computed from the gene frequencies of all the loci, as well as the eigenvalues of all the principal components, the proportions of individual eigenvalues to the total variance (contribution rates of components), and the raw scores of every individual for each of the principal components. PCoA was performed by using the program PCAGEN 1.2.1.
A Bayesian clustering procedure implemented in STRUCTURE v 2.3.3 [30] was used to infer the number of distinct genetic populations represented in our samples. The number of K clusters was set from one to six (i.e., the total number of sampled populations). MCMC simulations were performed by using the admixture model and correlated genetic frequencies among individuals. Simulations ran for 1,000,000 iterations, using 100,000 iterations as a burn-in period. We evaluated the length of burn-in, as well as the stabilization of the parameters of the analysis, using the plot function present in STRUCTURE 2.3.4 software. Then we ran results through STRUCTURE HARVESTER (https://taylor0.biology.ucla.edu/structureHarvester/, accessed on 1 May 2022) [31], using the maximum delta K as the selected value of K (Evanno method) [32]. Finally, we looked for a genetic signature of past demographic events (e.g., population expansion or decline) in populations of D. (A.) dodona by using COI sequences. We performed Fu's Fs test by comparing Fs statistics against a distribution generated from 1000 random samples, using Arlequin v.3.5. The statistic Fs tend to be negative under an excess of recent mutations, and a significant negative value was then taken as evidence of population growth. We also estimated Tajima's D statistics, and the significance of the D statistic was tested by generating random samples under the hypothesis of selective neutrality and population equilibrium, using a coalescent algorithm. A mismatch distribution analysis was also performed for the species, using the population growth-decline model in the software DNAsp 6.12.03 [33].

Genetic Differentiation
Eight of the fourteen microsatellite loci were correctly amplified for all 52 individuals; six of them were polymorphic (Supplementary Materials Table S2). Exact tests of genotypic linkage disequilibrium (either global or for populations) yielded no significant values (p = 0.05), thus suggesting that loci are independent. Significant positive departures from the Hardy-Weinberg equilibrium were found in two loci (Dione9 and Dione27) for two populations (LIM and CH2) (Supplementary Materials Table S3). Using Micro-Checker 2.2, we detected a signal for null alleles at these two loci, but there was no indication of errors caused by stutter bands or large allele dropout. Values for allelic richness, polymorphic loci (%), and observed and expected heterozygosities for each population are given in Supplementary Materials Table S4.
As two of our six loci were possibly affected by null alleles, we performed all statistics twice: once for all loci, and a second time for only the four loci without null allele signals. In the following analysis, we only show the values based on all six loci (the results based on four loci are given in Supplementary Materials Table S5). Because all analyses based on the four loci without null allele signals (i.e., genetic diversities and structures) showed similar results to those based on all loci, we conclude that the result is reliable population-genetic and phylogeographic information that has not been blurred by null alleles.
In the alignment of 654 bp of COI gene, we found nine variable sites (Supplementary  Materials Table S6). All the observed polymorphisms were single base-pair mutations; five were transitions and four were transversions. The haplotype diversity values range from 0.22 to 1, with an average value of 0.28, and nucleotide diversity ranged from 0 to 0.005, with a global value of 0.007 (Supplementary Materials Table S4). In total, we identified five haplotypes (Figure 3 and Supplementary Materials Table S4). The median-joining network was star-shaped, in that some less-frequent haplotypes (H1, H4, and H5) were closely related to a single common haplotype (H3). The haplotype H3 was shared between all populations, except for CH2. LIM presented three haplotypes (H2, H3, and H4), and CH2 showed two haplotypes (H1 and H5).
In the alignment of 654 bp of COI gene, we found nine variable sites (Supplementary  Materials Table S6). All the observed polymorphisms were single base-pair mutations; five were transitions and four were transversions. The haplotype diversity values range from 0.22 to 1, with an average value of 0.28, and nucleotide diversity ranged from 0 to 0.005, with a global value of 0.007 (Supplementary Materials Table S4). In total, we identified five haplotypes (Figure 3 and Supplementary Materials Table S4). The median-joining network was star-shaped, in that some less-frequent haplotypes (H1, H4, and H5) were closely related to a single common haplotype (H3). The haplotype H3 was shared between all populations, except for CH2. LIM presented three haplotypes (H2, H3, and H4), and CH2 showed two haplotypes (H1 and H5). Patterns of genetic differentiation were represented by a multivariate PCoA plot (Figure 4). The PCoA individual scores plotted into the first two principal component axes, Patterns of genetic differentiation were represented by a multivariate PCoA plot ( Figure 4). The PCoA individual scores plotted into the first two principal component axes, coordinate axis 1 and coordinate axis 2, explained 19.64% and 27.63%, respectively, of the total genetic diversity.  Samples presented a lot of mixture in the PCoA plot, without a clear segregation pattern for any population. These trends were confirmed by the AMOVA results. When looking at a possible structure by geographic distance (north vs. south), the differences among groups were not significant. Accordingly, 13.5% (p = 0.1241) of the total variance component was explained by the north (LIM + ARE + MOQ + TAC) and south (CH1 + CH2) groups of populations (see Table 1). The greater proportion of variation (74.7%, p = 0.000) was identified within populations. The FST based on allele frequencies between pairs of populations ranged from 0.10 to 0.30 (Figure 5a), with a mean FST = 0.19. Increased differentiation between north (LIM, ARE, MOQ, and TAC) and south (CH1 and CH2) groups were observed. However, Mantel tests did not show a significant correlation between geographic distance and estimates of gene flow, thus failing to support a pattern of isolation by distance (r 2 = 0.01, p > 0.05) (Figure 5b).  Samples presented a lot of mixture in the PCoA plot, without a clear segregation pattern for any population. These trends were confirmed by the AMOVA results. When looking at a possible structure by geographic distance (north vs. south), the differences among groups were not significant. Accordingly, 13.5% (p = 0.1241) of the total variance component was explained by the north (LIM + ARE + MOQ + TAC) and south (CH1 + CH2) groups of populations (see Table 1). The greater proportion of variation (74.7%, p = 0.000) was identified within populations. The F ST based on allele frequencies between pairs of populations ranged from 0.10 to 0.30 (Figure 5a), with a mean F ST = 0.19. Increased differentiation between north (LIM, ARE, MOQ, and TAC) and south (CH1 and CH2) groups were observed. However, Mantel tests did not show a significant correlation between geographic distance and estimates of gene flow, thus failing to support a pattern of isolation by distance (r 2 = 0.01, p > 0.05) (Figure 5b). groups of populations (see Table 1). The greater proportion of variation (74.7%, p = 0.000) was identified within populations. The FST based on allele frequencies between pairs of populations ranged from 0.10 to 0.30 (Figure 5a), with a mean FST = 0.19. Increased differentiation between north (LIM, ARE, MOQ, and TAC) and south (CH1 and CH2) groups were observed. However, Mantel tests did not show a significant correlation between geographic distance and estimates of gene flow, thus failing to support a pattern of isolation by distance (r 2 = 0.01, p > 0.05) (Figure 5b).  A population structure analysis that was conducted by using the software Structure identified three Bayesian groups (K = 3) ( Figure 6 and Supplementary Materials Figure S1). All three groups and all populations were found to share alleles, with slight evidence of two cluster formations, mainly composed of (i) LIM + ARE + MOQ + TAC and (ii) CH1 + CH2.  A population structure analysis that was conducted by using the software Structure identified three Bayesian groups (K = 3) ( Figure 6 and Supplementary Materials Figure  S1). All three groups and all populations were found to share alleles, with slight evidence of two cluster formations, mainly composed of (i) LIM + ARE + MOQ + TAC and (ii) CH1 + CH2.  Table S4. The tests of neutrality were not statistically significant for any populations. Similarly, the mismatch distribution suggests that these populations are in demographic equilibrium or decline (Supplementary Materials Figure S2).

Evolution of Heliconiinae-Passifloraceae in the Andes
The evolutionary history of Malesherbia and D. (A.) dodona is intimate, both in space and time, considering the overlap in their distribution and origin of the clades (around 5 Mya) [7,34]. The increasing aridity of the Atacama Desert, associated with the Andes uplift, seems to have had an important effect on the flora [35][36][37]. Thus, the gradual isolation of species due to climatic fluctuations, hampered by geomorphologic events, may explain the geographic pattern found in Malesherbia [2]. From a geographic perspective, diversification of Heliconiinae is broadly tied to the area of occurrence of D. (A.) dodona. In fact, radiation of heliconiine butterflies occurred predominantly on the eastern slopes of the Andes in Colombia, Ecuador, and Peru, as well as in the upper/middle Amazon basin [38]. Thus, the Andes uplift may have had consequences on such a process, with the diversity of heliconiine butterflies resulting in being greater on the east side. The evolution of Hel-  Table S4. The tests of neutrality were not statistically significant for any populations. Similarly, the mismatch distribution suggests that these populations are in demographic equilibrium or decline (Supplementary Materials Figure S2).

Evolution of Heliconiinae-Passifloraceae in the Andes
The evolutionary history of Malesherbia and D. (A.) dodona is intimate, both in space and time, considering the overlap in their distribution and origin of the clades (around 5 Mya) [7,34]. The increasing aridity of the Atacama Desert, associated with the Andes uplift, seems to have had an important effect on the flora [35][36][37]. Thus, the gradual isolation of species due to climatic fluctuations, hampered by geomorphologic events, may explain the geographic pattern found in Malesherbia [2]. From a geographic perspective, diversification of Heliconiinae is broadly tied to the area of occurrence of D. (A.) dodona. In fact, radiation of heliconiine butterflies occurred predominantly on the eastern slopes of the Andes in Colombia, Ecuador, and Peru, as well as in the upper/middle Amazon basin [38]. Thus, the Andes uplift may have had consequences on such a process, with the diversity of heliconiine butterflies resulting in being greater on the east side. The evolution of Heliconiinae is closely tied to Passifloraceae, the only plant family they use as a host. Benson et al. [39] were the first to point out that older lineages of Heliconiinae tend also to use older Passifloraceae as host plants, as in the case D. (A.) dodona larvae-Malesherbia plants. Such a co-evolutionary scenario predicts a progression from heliconiine species that feed on woody, long-lived, and presumably ancient food plants to those that specialize in herbaceous recent passion vines [39]. However, Malesherbia is a fragile herbaceous plant, not showing a scandent growing habitat. Furthermore, ecological opportunism and host and habitat shifts also seem to have played a major role in Heliconiinae speciation (for a review, see also References [40,41]). There is no record of additional herbivores associated with such plants in the area. The system provided by D. (A.) dodona and associated Malesherbia plants-which, in most cases, are geographically isolated-where populations of both seem to be very localized, restricted to single valleys on the western slopes of the Andes, offers a unique opportunity for future studies on heliconiine evolution of host-plant usage.
We found that Dione (A.) dodona uses seven distinct Malesherbia species as host plants, revealing it as an oligophagous species. This oligophagy was evidenced through the field collection, with each population presenting larvae that use distinct plants. Some of these individuals were brought to the laboratory and reared, exchanging the host plants. Flowers of M. ardens from Torata locality (MOQ) were offered to the first instar collected on M. tenuifolia from the type locality (ARE); the larvae accepted the flowers and fed during the following days without any negative behavior (unpublished data). All Malesherbia species reported here belong to the Malesherbia section, a lineage distributed from Northern Chile to Central Peru [6]; more than half of all the species in this section recorded for the Andean western slopes were used as host plants by D. (A.) dodona (7 vs. 13 spp.). Although the search for Malesherbia species during this study was restricted to the altitude range from 1000 to 3000 m elevation (due to the fact that all of these species in the study area are distributed between these altitudes), it is not ruled out that D. (A.) dodona can use as host plant Malesherbia section species that are found at lower elevations, as is the case of M. tocopillana, which is only known from its locality type (Tocopilla) in Chile, at an altitude range of 150 to 400 m elevation [5]; or also those species that could not be found during the intensive searches distributed on western slopes of the Andes, such as populations of M. haemantha, M. scarlatiflora, M. splendens, and M. turbinea.
The occurrence of D. (A.) dodona is strongly associated with the presence of Malesherbia, since all known records of adults match the distribution of this plant genus on the western slopes of the Central Andes [1]. Interestingly, the occurrence of D. (A.) dodona in a given area seems to relate to the availability of plant population belonging to the section Malesherbia, which is mostly allopatric. Changes in host plants among localities allow the occurrence of this butterfly in a vast area in the xeric western slopes of the Andes. However, the genus Malesherbia has a much wider distribution. Thus, in a broader sense, D. (A.) dodona may be limited in expansion due to the absence of Malesherbia species accepted as hosts, a subject which should be further explored. Heliconiinae, in general, is selective regarding intraand inter-specific attributes of plants used as hosts, particularly regarding oviposition behavior and larval feeding, which may lead to local specializations [42][43][44][45]. Although most species are oligophagous and, thus, have relatively broad distributions since they alter hosts among localities, there is variation in this trait. Some have a narrower diet breath, using a few passion vines as hosts throughout their distribution range. For example, Dione (Dione) moneta and Heliconius nattereri are monophagous and, thus, have narrow geographic distributions in a given area of occurrence, which are limited by that of their host plants [39,42,46].

Genetic Connectivity through Suitable Areas
Our study shows that populations of D. (A.) dodona contain relatively little genetic variation and are not genetically subdivided. The overall haplotype and nucleotide diversity in the sister species Dione (A.) vanillae was 0.68 and 0.0143, respectively [47], which is quite higher than the 0.28 and 0.0009 found in D. (A.) dodona. Considering that the D. (A.) vanillae study included subspecies (further considered species) in addition to the population level, and that variation in sampling size and length of sequences likely influence such differences between studies, we looked to the microsatellites within the genus. Similarly, the other species of Dione showed little variation (13 alleles in 19 loci) [14].
It remains unknown whether D. (A.) dodona migrates or stays in a kind of dormancy when plants are not in active growth. Given that one congeneric species, D. (D.) moneta, is putatively migratory [48], we might consider that D. (A.) dodona could have a migratory potential. This aspect should be explored in conjunction with variation in phenology of Malesherbia species used as hosts, since females are highly selective regarding oviposition, laying eggs on flowers [1]. Thus, the little genetic variation of microsatellites observed can result from repeated localized extinctions and recolonizations, typical of a metapopulation structure [49]. Accordingly, patterns of variation found in D. (A.) dodona could result from lower historical population sizes and perhaps more recent colonization and/or bottleneck, as is supported by neutrality tests and mismatch distribution.
Dione (Agraulis) dodona is the oldest lineage among the extant Heliconiinae, ca. 8.59-2.8 My according to estimates of Nuñez et al. [7]; and several Malesherbia section species (e.g., M. tenuifolia and M. arequipensis) are relatively young (~5 My), with the genus Malesherbia having only recently colonized hyperarid habitats in deserts (~10 My) [34], thus possibly indicating that the population of D. (A.) dodona originated in the north of Chile and spread to Central Peru, together with the species of the Malesherbia section, being relatively recent colonization that does not allow us to detect genetic divergence between the populations. There exists a slight difference in the genetic structure about the southernmost population evaluated of D. (A.) dodona in the Arica region (CH2); however, this could be studied in greater detail in future studies including more southern populations.
In the second scenario, probably connectivity between the different populations of D. (A.) dodona (attested by low F ST and admixture ancestry of individuals) is promoted by the distribution of the Malesherbia species. Historically, the species of Malesherbia recorded in Peru were considered isolated populations in their majority, due to the few records [4]. However, this may be largely due to a constraint in evaluating areas of difficult access on the western slopes, which resulted in gaps of information along with large areas, leading to the assumption that Malesherbia species were absent. An example is the Department of Ica, which connects ARE and LIM populations, where no Malesherbia species were recorded. However, probably due to the lack of access to areas where this plant genus grows, we assume that if a species of Malesherbia is recorded in the future in this Department, it could be the link between the populations of D. (A.) dodona from Lima and Arequipa that are currently separated by approximately 600 km, without a record of species.
The evidence of connectivity between populations using different Malesherbia species, based on the individual admixture of alleles and shared haplotypes of D. (A.) dodona, allows us to infer a high dispersal capability in this species. Likely using a stepwise pattern, individuals keep an ongoing gene flow between populations along 1200 km.

Conservation of Suitable Areas
Desert areas separating the currently documented D. (A.) dodona populations are almost completely devoid of vegetation, representing potential barriers to the dispersal of flying insects, as recently reported for a leaf miner micromoth whose geographic range partially overlaps that of this butterfly [50]. High connectivity between populations despite potential barriers suggests that a given species could be less affected by anthropogenic modifications [51]. However, under the accelerated rate of habitat modification and destruction throughout the distribution range of D. (A.) dodona, its dispersal ability may not be enough to maintain population connectivity. Accordingly, habitats in which this butterfly and its host plants occur must be protected as a first step toward contributing to the conservation of the still insufficiently known biodiversity of the arid environments of the western slopes of the Andes.

Conclusions
Dione (Agraulis) dodona is an oligophagous species which uses at least seven different species of Malesherbia (Passifloraceae) as host plants within its distribution range on the western slopes of the Andes of Peru and Northern Chile. Such a distribution coincides with that of species located within the section Malesherbia, recorded as the only host plant group, suggesting that these are selected for oviposition, thus conditioning the geographic distribution of this butterfly. We did not find significant genetic differentiation between population samples from different host-plant species. Our results indicate high genetic exchange among populations between the different localities and host plants. It remains unknown how such a gene exchange is made among such geographically isolated populations located in the valleys of the Andes.
Supplementary Materials: The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/insects13090819/s1. Table S1: Collection data for larvae of D. (A.) dodona analyzed in this study. Table S2: Summary of basic information of the eight amplified microsatellites, with repeat motifs; annealing temperatures (Ta); primer sequences; size range for PCR products; number of alleles per loci; and estimates for F ST , F IS , and expected (H E ) and observed (H O ) heterozygosity.    Table S6: Polymorphic sites (except ambiguous ones) at the COI gene for populations of Dione (Agraulis) dodona (n = 34). Nucleotides are numbered from 1 to 654. Each haplotype distribution is represented per locality and by the total number of individuals per haplotype. Abbreviations for populations are given in Supplementary Materials Table S2; periods represent an absence of the correspondent haplotype. Figure S1: Bayesian analysis of individuals of Dione (Agraulis) dodona based on microsatellite data. Likelihood values for each number of groups tested (i.e., the K value) with Structure 2.3. The empty circles represent the exact likelihood values for each run, with K ranging from 2 to 6. Filled circles and complete lines represent the absolute value of the approximate second derivative of the mean likelihood for each tested K value according to Evanno et al. [32]. Figure