A global molecular phylogeny yields insights into the dispersal and invasion history of Junonia, a butterfly genus with remarkable dispersal abilities

The nymphalid butterfly genus Junonia has remarkable dispersal abilities. Occurring on every continent except Europe and Antarctica, Junonia are often among the only butterflies on remote oceanic islands. The biogeography of Junonia has been controversial, plagued by taxonomic disputes, small phylogenetic datasets, incomplete taxon sampling, and shared interspecific mitochondrial haplotypes. Junonia originated in Africa but its route into the New World remains unknown. Presented here is, to our knowledge, the most comprehensive Junonia phylogeny to date, using full mitogenomes and nuclear ribosomal RNA repeats from 40 of 47 described species. Junonia is monophyletic and the genus Salamis is its probable sister clade. Genetic exchange between Indo-Pacific Junonia villida and New World Junonia vestina is evident, suggesting a trans-Pacific route into the New World. However, in both phylogenies, the sister clades to most New World Junonia contain both African and Asian species. Multiple trans-Atlantic or trans-Pacificinvasions could have contributed to New World diversification. Hybridization and lateral transfer of mitogenomes, already well-documented in New World Junonia, also occurs in at least two Old World lineages (Junonia orithya/Junonia hierta and Junonia iphita/Junonia hedonia). Variation associated with reticulate evolution creates challenges for phylogenetic reconstruction, but also may have contributed to patterns of speciation and diversification in this genus.


Introduction
The balance between immigration, extinction and diversification determines species richness in geographical localities [1]. Most immigrants fail to establish populations, and the majority that colonize successfully undergo little or no diversification [2]. Only rare immigrant lineages show high diversification rates and adaptive radiation [2]. Understanding factors that make immigrant lineages successful in new habitats is important to the disciplines of biogeography, invasion biology, evolution and conservation. The butterfly genus Junonia has extraordinary abilities for dispersal and diversification. Junonia originated 15-27 million years ago (Ma) [3] and are often among the few butterfly species present on remote oceanic islands [4,5], suggesting that they can survive crossing thousands of kilometers of open water to colonize and diversify in new habitats. Some Old World Junonia (28 extant species) appear to have crossed an ocean basin 2-4 Ma to establish a radiation of 18 New World species [3]. This unparalleled ability to disperse may have fomented diversification in Junonia, but the lack of effective geographical barriers to gene flow also may have stymied previous attempts to resolve its phylogenetic and biogeographic history.
The invasion of the New World by Junonia has been the subject of decades of speculation. Based on molecular phylogenetic and biogeographic analyses, the consensus is that Junonia originated in Africa [3,19], but which lineages invaded the New World remains unanswered. Four taxa have been suggested as sister to the New World Junonia: Junonia villida (Indo-Pacific) [15], Junonia orithya madagascariensis (Africa) [12,13], Junonia orithya orithya (Asia) [11] and Junonia lemonias (Asia) [5,10]. Also unknown is whether the occupation of the New World by Junonia was a unique event [3,8,[12][13][14][15]19] or caused by multiple invasions involving one or more Old World species [10,18]. Recent full mitochondrial genome sequence data suggests a trans-Pacific route and that the same lineage (J. villida) crossed the Pacific more than once to reach the New World [5,10].
royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 289: 20212801 (a) DNA barcodes, short fragment nuclear DNA and haplotype groups Three primary mitochondrial haplotype groups are shared among New World Junonia species [3,6,9,[16][17][18][19][20]23]. This pattern was first discussed by Pfeiler et al. [19] based on the 658 base pair cytochrome oxidase I DNA barcoding fragment used to delimit many animal species [34,35]. Pfeiler et al. [19] suggested that shared mitochondrial haplotypes may be owing to the recent invasion of Junonia into the New World (approx. 2-4 Ma) [3], such that lineage sorting is incomplete and hybridization between species is ongoing. Haplotype group A is common in South American and Caribbean Junonia, while haplotype group B is common in North and Central America, and these two monophyletic groups diverged approximately 2.  [20]. Gemmell & Marcus [18] suggested that mitochondrial DNA sequences from Indo-Pacific J. villida are more similar to the New World Junonia than sequences from J. orithya, consistent with some previous hypotheses based on morphology and geographical distributions [4,14,36]. They proposed two hypotheses for the origin of haplotype B: either it evolved from haplotype group A mitochondria within the New World, or that two separate New World invasions carried haplotype groups A and B, followed by hybridization between species descended from the two invasion events [18].

(b) The use of full mitochondrial DNA genomes
Mitochondrial genome (mitogenome) sequences are a rich source of genetic variation for improved taxonomic resolution in phylogenetic studies [37][38][39][40][41]. The first two full approximately 15.2 kb Junonia mitochondrial genome sequences were reported from Old World species [40,41]. In 2016, 14 New World Junonia mitogenomes and four additional Old World mitogenomes were used to reconstruct patterns of diversification [10,42]. Subfamily Nymphalinae and genus Junonia formed monophyletic groups, but like previous findings using COI sequences, New World Junonia showed a lack of monophyly, consistent with earlier proposals of multiple Junonia invasions of the New World [10,18]. Separate invasions by different Old World species may have created each of the New World haplotype groups [5,10]. The haplotype groups A and B in Junonia were estimated to have diverged 2.31 ± 0.42 Ma, consistent with previous molecular clock estimates [3,19]. Divergence times for all other New World haplotypes were estimated: divergence of A 1 and A 2 1.52 ± 0.31 Ma and the divergence of B 1 from the rest of haplotype B 1.19 ± 0.29 Ma [10]. Peters & Marcus [5] recognized the distinctiveness of New World haplotype group C in high-elevation J. vestina, (more closely related to J. lemonias from Asia than it is to other New World haplotypes) and was estimated to have diverged 1.16 ± 0.32 Ma [5]. Consistent with previous studies, haplotypes A and B each formed monophyletic groups [10,[16][17][18][19], but the placement of haplotype group C makes the mitochondria of the New World Junonia paraphyletic [5].
Recently, Cong et al. [23] analysed next-generation sequencing libraries from 11 New World Junonia species to describe two new species (Junonia pacoma; Pacific mangrove buckeye and Junonia stemosa; South Texas dark buckeye). Fixed diagnostic characters (morphological or molecular) that allow for the consistent separation of J. stemosa from the morphologically nearly identical Junonia nigrosuffusa could not be identified, so we will treat this form as a subspecies J. nigrosuffusa stemosa nov. stat. [23]. To date, the largest mitogenome-based Junonia phylogeny consists of 28 mitogenomes (15 Junonia species (eight New World, seven Old World), and one from each of the five other Junoniini genera) [26]. This analysis concluded that Junonia was monophyletic, haplotype groups A and B form monophyletic clades, but the New World Junonia are not monophyletic. The most likely sister clade to genus Junonia (though with weak bootstrap support) contains both African Protogoniomorpha and Asian Yoma, all consistent with a recent analysis of mitochondrial DNA barcodes from Old World Junonia species [25].
Many mitogenomes have now been published for Junonia (18 of 47 species) and other genera from tribe Junoniini (one species each from the five other genera) through GenBank (Dataverse electronic supplementary material, table S1). Here, we build upon this pre-existing knowledge by assembling additional full Junonia mitochondrial genomes (22 new species: 40 of the 47 described species), tribe Junoniini (16 additional species across five genera) and additional outgroup species into a phylogenetic analysis. To complement the mitogenomes, we conducted a phylogenetic analysis based on the nuclear ribosomal RNA (rRNA) repeat sequence (an 8-10 kb sequence containing three rRNA repeat subunits (2.8S, 18S and 28S), two internal transcribed spacers (ITS1 and ITS2), and 5 0 and 3 0 non-transcribed spacers) [43] from the same group of samples. Creating phylogenies using both types of sequences will aid in reconstructing patterns of reticulate evolution and describing biogeographic distributions for a rapidly diversifying clade.

Material and methods (a) Specimen collection, preparation and sequence generation
Ninety-seven specimens were analysed, consisting of 64 Junonia (40 of 47 described species), 21 additional specimens from tribe Junoniini and 11 outgroup species within subfamily Nymphalinae (Dataverse electronic supplementary material, table S1). We generated most data ourselves, and supplemented with additional sequences from GenBank. Specimens were identified based on morphology. DNA was extracted from a single leg per specimen using a Qiagen DNEasy Blood and Tissue Kit (Qiagen, Düsseldorf, Germany) either manually as previously described [18] [26]. Sequences were assembled and annotated to previously published reference sequences with GENEIOUS 10.2.6 (Dataverse electronic supplementary material, table S1).
(c) Nuclear ribosomal RNA and internal transcribed spacer repeat phylogeny Phylogenetic reconstruction employed 90 nuclear rRNA repeat sequences from 62 Junonia specimens, 19 specimens from other genera in tribe Junoniini and nine outgroup specimens from subfamily Nymphalinae (Dataverse electronic supplementary material, table S1). Our laboratory generated all of the nuclear rRNA repeat sequences, some of which were published previously [26][27][28][50][51][52][53][54][55][56][57], but are analysed in concert here for the first time, to our knowledge. When samples in the mitogenome dataset lacked GenBank Sequence Read Archives (SRAs) containing sufficient raw sequence data to assemble a nuclear rRNA repeat, they were excluded. Sequences were aligned (Dataverse electronic supplementary material, file S2) and analysed as described above. In contrast with the New World, prior studies concluded that individual Old World Junonia species were monophyletic based on limited sampling of species and populations. The current analysis of complete mitochondrial genomes suggests a lack of monophyly in at least some Old World Junonia species (figure 2). For example, J. orithya and Junonia hierta form two separate lineages, one in Asia and one in Africa, which are more closely related to the sympatric lineage of the other species than they are to conspecific allopatric lineages. This suggests that there has either been remarkable parallel morphological and colour pattern evolution in these lineages, or that there may be lateral transfer and introgression of mitochondrial haplotypes between them. Another species pair that may be experiencing introgression and ongoing geneflow are Junonia iphita and Junonia hedonia. The J. iphita mitogenome sequence from Indonesia forms a clade with sympatric sequences from J. hedonia rather than with J. iphita sequences from elsewhere in Asia. This suggests that lateral transfer events and introgression is more widespread in Old World Junonia than previously appreciated.

(b) Junonia nuclear ribosomal RNA repeat phylogeny
The nuclear rRNA repeat phylogeny was constructed using Bayesian inference with a GTR + I + G model with a best state likelihood of −119 000 and a final average deviation of split frequencies of 0.003566 (figure 3). Tribe Junoniini and its component genera are monophyletic. Nearly all of the New World Junonia nuclear rRNA repeats form two distinct clades. One clade is restricted to North America and Bermuda. The other clade contains specimens from South, Central, and parts of North America, and the Caribbean. Some New World Junonia species include individuals assigned to both clades. As a whole, the New World Junonia are monophyletic with only a single exception: the same Australian J. villida specimen that grouped with New World mitogenomes also clusters with New World nuclear rRNA repeats. The J. vestina specimen with a distinct mitogenome (haplotype group C) that had grouped with the Old World J. villida clade in the previous analysis (figure 2) is the sister taxon to another South American J. vestina sample within the New World clade in the nuclear rRNA repeat phylogeny (figure 3).
The nuclear rRNA repeat phylogeny for Old World Junonia taxa is consistent with the mitogenome phylogeny with two major differences. The J. villida clade, for which a J. vestina sequence from the New World was a sister clade in the mitogenome tree, is more closely associated with a group of Asian Junonia species (J. lemonias, Junonia erigone and Junonia almana) in the nuclear rRNA repeat phylogeny. The second is the sister clade to this Asian clade in the mitogenome tree, is the sister clade to the New World Junonia in the rRNA repeat phylogeny ( figure 3). This clade, which includes both African and Asian lineages, contains J. orithya and J. hierta. The African lineages of J. orithya and J. hierta form a clade together and Asian lineages from these two species and are interspersed with one another, similar to the results of the mitogenome analysis. By contrast, the nuclear rRNA repeats from J. iphita form a monophyletic group and do not show the affinity with J. hedonia sequences as observed using mitogenomes ( figure 2). The relationships within tribe Junoniini based on rRNA repeats are consistent with the mitogenome analysis except for the placement of Salamis, which is placed as sister taxon to Junonia with a very high probability value (0.99). The relationships of the outgroup species are consistent with the mitogenome analysis, with the exception of tribe Kallimini, which is paraphyletic in the nuclear rRNA repeat analysis. Mallika jacksoni and Catacroptera cloanthe are assigned to tribe Kallimini as expected, but Kallima paralekta clusters with tribe Kallimoidini within subfamily Nymphalinae (figure 3). The clade defined by the oldest node within the Junoniini (figures 2 and 3) contains the monophyletic genera Precis and Hypolimnas, as reported previously [3,24,25]. The sister clade to Junonia differs between mitochondrial and nuclear analyses. The mitogenome sister clade contains the genera Protogoniomorpha, Yoma and Salamis, with Salamis as the outgroup to the other two genera, consistent with some prior phylogenetic analyses [3,[24][25][26]. Pyrcz et al. [25] found genus Salamis embedded within Junonia. Our rRNA repeat phylogeny shows Salamis as sister to genus Junonia ( posterior probability value of 0.99), with the next further outgroup containing the genera Yoma and Protogoniomorpha. Kallima (tribe Kallimini) originally served as a catch-all genus for Asian and African nymphalid butterflies masquerading as leaf mimics [58]. Based on genitalia and behavioural characteristics, it was later determined that genus Kallima is restricted to Asia, and African leaf mimics and were reassigned to three other Nymphalinae genera (Junonia (Junoniini), Mallika (Kallimini) or Kallimoides (Kallimoidini)) [29,59,60]. Tribe Kallimini comprised the genera Kallima, Mallika and Catacroptera and was viewed as the sister clade to the Junoniini, since most Kallimini and the basal character state within Junoniini is leaf mimicry [59]. Molecular phylogenetics has consistently placed tribe Melitaeini as the sister to the Junoniini, with tribe Kallimini placed as a farther outgroup [29,61]. While the mitogenome tree (figure 2) is consistent with these results, phylogenetic reconstruction using the rRNA repeat region (figure 3) shows genera Mallika and Catacroptera (both African genera) forming a monophyletic clade. Genus Kallima falls outside this grouping and as sister to Kallimoides, with a Bayesian probability value of 1 ( figure 3). These differences could be attributable to limited taxon sampling among outgroups within the rRNA repeat phylogeny ( figure 3). Both the Junoniini and Kallimini show close molecular phylogenetic associations with the Melitaeini in our mitogenome reconstruction (figure 2) and in previous studies [29,62,63], but nuclear rRNA repeat sequences, helpful for inferring higher level taxonomic relationships within the Nymphalinae, are not yet available from the Melitaeini.

(b) New World Junonia
Species-level relationships in the New World Junonia were unresolved by mitogenomes (figure 2), consistent with earlier DNA barcode and mitogenome studies [3,5,[16][17][18][19]23]. Many New World species include individuals that carry both A and B haplotype groups. Similarly, most New World Junonia species do not form monophyletic groups based on nuclear rRNA repeats. This is indicative of gene flow between New World Junonia species and is consistent with prior observations of hybridization between many of these species [20,23,30,32]. Based on the mitogenome phylogeny (figure 2), the New World Junonia is monophyletic with two exceptions. The first is a J. vestina sample that possesses haplotype group C and is most closely related to a lineage of J. villida from the Indo-Pacific. This finding differs from that of Peters & Marcus [5], which placed J. vestina as the sister taxon to J. lemonias. Peters & Marcus [5] only included single representatives of most Junonia species, including J. villida, so this discrepancy can be attributed to limited sampling. In both the current study and Peters & Marcus [5], a single J. villida sample from Australia is most closely related to the haplotype group A 2 in the New World. These findings signify that long-range dispersal and gene flow across the Pacific may be occurring and is of importance to understanding the relationships between the Old World and New World species, as well as the origins of the New World Junonia.
Complete nuclear rRNA repeats occur at high copy number in the genome and are easily recovered from the same genome skimming datasets used for assembling whole mitogenomes (figure 3; [43]). Phylogenetic analysis of Junonia nuclear rRNA repeats did not recover the same clades found through analyses of mitogenomes, nor do rRNA repeats resolve New World species-level relationships. Like the mitogenome, there is an apparent geographical signal, with most North American specimens forming one rRNA repeat lineage, while a second is made up of Central and South American, Caribbean and a few specimens from southern portions of North America. Based on nuclear rRNA repeat sequences, the New World Junonia are monophyletic, except for an Australian J. villida sample that also groups with the New World in the mitogenome phylogeny. Unlike the mitogenome phylogeny, the nuclear rRNA repeats of all J. vestina samples fall within the New World rRNA clade. This further supports the hypothesis that gene flow between New World J. vestina and the Indo-Pacific J. villida by means of long-distance dispersal across the Pacific, followed by hybridization, may be ongoing since individual specimens do not share the same level of molecular affinity with one another, with some forming clades with the other taxon.

(c) Old World Junonia
Previous phylogenetic studies of Old World Junonia could not test for monophyly or make strong statements about species-level relationships, because only a single sample of each species was used [9,[16][17][18][24][25][26]29]. Kodandaramaiah & Wahlberg [3] was exceptional in including multiple specimens from some Old World species comparable with this analysis. First, specimens from Asian and African populations of J. hierta and J. orithya form clades based on geography rather than taxonomic species assignment here (figures 2 and 3) and in Kodandaramaiah & Wahlberg [3], though they did not comment on the pattern in the text of their paper. Also consistent with Kodandaramaiah & Wahlberg [3] is the placement of this clade as sister to the New World Junonia, although our mitogenome data includes J. villida within the clade. Further, mitogenomes from J. iphita pair with either J. hedonia or Junonia atlites (figure 2). By contrast, in the analysis of rRNA nuclear repeats (figure 3), the J. iphita sequences form a monophyletic clade and none are sister to J. hedonia. A recent lateral transfer royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 289: 20212801 of the J. hedonia mitogenome to J. iphita in Indonesia seems likely as the sequences are nearly identical, but additional sampling and sequencing from these and other species in this lineage would clarify patterns of organelle capture. What seems increasingly clear is that while lateral transfer events may be most frequent in the New World Junonia, it also occurs in some Old World lineages, contributing to reticulate evolution of the genus as a whole.
Despite instances of apparent lateral transfer, the Old World Junonia form distinct molecular clades made up of species that share phenotypic features (figure 4). Although there are small differences between the two different phylogenetic analyses in respect to species placement, all Old World species can be grouped based on geography, habitat type and mimicry strategy. Some clades defined by the oldest Junonia nodes in the mitogenome phylogeny (figure 4) consist of two lineages of forest-dwelling species. The first is an Asian forest-dwelling butterfly lineage containing species (J. iphita, J. atlites, Junonia intermedia, Junonia adulatrix and J. hedonia) that at rest have closed wings that masquerade imprecisely as leaves [58]. The second lineage includes African species that are also forest-dwelling but most are considered to be very good leaf mimics [24] (figure 4). There is one exception, Junonia sophia, which is thought to be a Batesian mimic of the false diadem butterfly (Pseudacraea lucretia) [59]. Within this African Junonia lineage, there are two subclades that cluster based on similar coloration of the dorsal wing surfaces (either blue or brown).
The next large Junonia clade in the mitogenome phylogeny includes all New World Junonia species as well as some Asian (J. orithya, J. hierta, J. lemonias, J. villida, J. erigone and J. almana) and African species (J. orithya, J. hierta, Junonia oenone, J. westermanni and Junonia rhadama). These are grassland or open habitat specialists, and none are considered to be mimics (figure 4, [59,64]). Consistent with prior studies, this lineage originated in Africa, dispersed to Asia, and then returned to Africa, perhaps several times [3].
Later, this lineage established the New World Junonia radiation. The rRNA repeat phylogeny yields similar results to the mitogenome analysis, with phylogenetic trends based on habitat type, mimicry and geography, but differences do exist. The clade defined by the oldest node within Junonia recovered by this analysis supports an African origin for this genus, as all species in this clade are forest-dwelling and restricted to Africa (figure 4), reinforcing mitogenome results and an earlier three-gene analysis [3]. Like the mitogenome phylogeny, all species within this clade are mimics. The next clade consists of Asian and Indo-Pacific Junonia, with one exception, J. rhadama that occupies Madagascar and several other Indian Ocean islands. The first major split in this clade is by habitat type with forest-dwelling imperfect leaf mimics and grassland/open habitat lineages. Forestdwelling species include the J. iphita and J. hedonia complex discussed above in the mitogenome phylogeny and the same pattern is observed here. Open habitat species include J. rhadama. The remaining open habitat species are more similar morphologically to New World taxa than they are to other Old World species. A key difference in the composition of this clade exists between our phylogenetic analyses. In the mitogenome analysis, J. villida is the earliest diverging species and sister to the New World Junonia (figure 4), while in the nuclear rRNA repeat analysis, the sister clade is limited to African J. westermanni, and J. oenone and African/Asian J. orithya and J. hierta, and the remaining taxa (Asian: J. villida, J. lemonias, J. erigone and J. almana) are transferred into a separate clade ( figure 4). This difference between the mitogenomic and nuclear rRNA repeat phylogenies raises the intriguing possibility that the New World Junonia were established with contributions from two Old World Junonia lineages: one containing J. villida which crossed the Pacific Ocean and a second containing J. orithya and J. hierta, which could have crossed either the Atlantic or Pacific Ocean to reach the New World. Thus, several of the early hypotheses for the origin of the New World Junonia may royalsocietypublishing.org/journal/rspb Proc. R. Soc. B 289: 20212801 not have been mutually exclusive after all and instead may turn out to be simultaneously correct [12,13,15].

Conclusion
Compared with many other butterfly taxa, Junonia have particularly effective dispersal abilities, allowing them to colonize remote new habitats, perhaps in some cases repeatedly. This creates scenarios where resident and new immigrant Junonia are sympatric and lateral transfer is possible through hybridization and reticulate evolution. This appears to have taken place both in the Old World and New World. The large number of New World Junonia (18 species) that have evolved in a very short period of time (2)(3)(4) suggest that speciation in this group has been accelerated compared to the older lineages in the Old World (28 species) which have diverged since the origins of the genus 15-27 Ma [3]. The variety of new ecological niches (especially larval host plant associations) presented by new habitats, in combination with frequent lateral transfer of adaptive genes and traits, and reticulate evolution may have contributed to the substantially greater rate of speciation in the New World, making this system similar to the explosive species radiation events such as the Lake Victoria cichlids [65] or the Hawaiian Drosophila [66].
To address the challenges of species delimitation in Junonia in the face of reticulate evolution, one recent study successfully employed complete Z chromosome sequences [23]. Although their dataset only included specimens from some New World Junonia taxa, they were able to resolve the species into monophyletic clades. Accumulating complete Z chromosome sequences from the rest of the genus will provide an interesting comparison to results from analyses of mitochondrial genomes and nuclear rRNA repeats, and together this may permit better delimitation of the remaining species, further resolve species-level phylogenetic relationships, and should be considered for future phylogenetic studies of this genus.
Data accessibility. All sequences and SRAs have been made submitted to be publically available through GenBank and accession numbers for all data used in the analyses can be found in the Dataverse electronic supplementary material, table S1 (https://doi.org/10.34990/FK2/ O9UKCE/LQDEC8) which has been deposited through the online data repository Dataverse. Sequence alignment data is available as electronic supplementary files through the online data repository Dataverse for both the full mitochondrial genomes (https://doi. org/10.34990/FK2/O9UKCE/5S8ZJ8) and rRNA repeats (https:// doi.org/10.34990/FK2/O9UKCE/YYM2ZT).