Increased Evolutionary Rate in the Z chromosome of Sympatric and Allopatric Species of Morpho Butterflies

Abstract

Divergent evolution of genomes among closely related species is shaped by both neutral processes and ecological forces, such as local adaptation and reinforcement. These factors can drive accelerated evolution of sex chromosomes relative to autosomes. Comparative genomic analyses between allopatric and sympatric species with overlapping or divergent ecological niches offer insights into reinforcement and ecological specialization on genome evolution. In the butterfly genus Morpho, several species coexist in sympatry, with specialization across forest strata and temporal niches. We analyzed the genomes of eight Morpho species, along with previously published genomes of three others, to compare chromosomal rearrangements and signs of positive selection in the Z chromosome $vs.$ autosomes. We found extensive chromosomal rearrangements in Z chromosome, particularly in sympatric species with similar ecological niches, suggesting a role for inversions in restricting gene flow at a postzygotic level. Z-linked genes also exhibited significantly higher dN/dS ratios than autosomal genes across the genus, with pronounced differences in closely related species living in sympatry. Additionally, we examined the evolution of eight circadian clock genes, detecting positive selection in Period, located on the Z chromosome. Our findings suggest that the Z chromosome evolves more rapidly than autosomes, particularly among closely related species, raising questions about its role in prezygotic and postzygotic isolation mechanisms.

Significance statement.

Using comparative genomics, we investigate the evolutionary dynamics in sex chromosomes in sympatric butterfly species of the genus Morpho, and detected increased level of divergence in both coding sequences and structural variants. We found significantly enhanced signals of positive selection in genes located on the sex chromosome Z when compared with the autosomes, in particular on the circadian gene Period, that may be involved in the adaptation to divergence diel niche in sympatric species. Our results therefore raise new questions on the role of the sex chromosome Z in ecological niche differentiation and/or reinforcement of barriers to gene flow.

Introduction

Sex chromosomes have been observed to undergo accelerated rates of sequence evolution compared to autosomes (Sackton et al. 2014), a phenomenon commonly known as fast-X or fast-Z evolution (Charlesworth et al. 1987; Vicoso and Charlesworth 2006). The faster evolution of sex chromosomes has been observed in different groups with ZW system as some birds, snakes, and the salmon louse Lepeophtheirus salmonis (Mank et al. 2009; Vicoso et al. 2013; Wright et al. 2015; Irwin 2018; Mongue and Baird 2024), although contradictory trends have also been observed (e.g. in Lepidoptera, Rousselle et al. 2016).

Within species, faster evolution of sex chromosomes can result from the effect of genetic drift limiting the purge of deleterious mutations when the effective populations size is reduced. For every mating pair, there are indeed four copies of autosomal genes and only three copies of X or Z-linked genes. Consequently, the effective population size (Ne) for X or Z sex chromosomes is reduced when compared with the autosomes ($3/4N_e$). Such reduced $N_e$ may limit the loss of slightly deleterious mutations within each species, and may also enhance the probability of fixing adaptive mutations (Vicoso and Charlesworth 2006; Johnson and Lachance 2012). Between species the fixation of divergent mutations under positive selection on the sex chromosomes X and Z may be further promoted by the hemizygosity of X and Z chromosomes in XY males and ZW females, respectively. Hemizygozysity directly exposes rare advantageous recessive alleles to selection and leads to greater rates of fixation compared to rare autosomal alleles that are sheltered in heterozygotes (Johnson and Lachance 2012). Positive selection has been suggested to promote faster-X and Z evolution in mammals, Drosophila (XX/XY sex-determination system), but also in lepidoptera (ZW/ZZ sex-determination system), like the silkmoth Bombyx huttoni, the moth Manduca sexta, and the butterfly Danaus plexippus (Meisel and Connallon 2013; Sackton et al. 2014; Mongue et al. 2022). In the Crambidae moth Ostrinia nubilalis, a presumed 4 Mb inversion in the Z chromosome appears to increase the accumulation of ecologically adaptive alleles and genetic differentiation Wadsworth et al. (2015). Inversions capturing adaptive variants located on the sex chromosomes are thus expected to disproportionately contribute to the evolution of local adaptation (Connallon et al. 2018). Different Z-located inversions may thus be observed in closely related species evolving in different ecological niches.

In sympatric or parapatric species, divergence in the sex chromosomes may also be further enhanced by their role in premating and postmating isolation (Payseur et al. 2018). Introgression in the sex chromosomes has been shown to be reduced when compared with the autosomes (Schilthuizen et al. 2011; Fraïsse and Sachdeva 2021). For instance, in sister-species of Ishnura damselflies with overlapping geographic ranges, the levels of introgression was shown reduced in the X chromosome, when compared with the autosomes (Swaegers et al. 2022). Furthermore, comparisons between species of passerine birds revealed that structural variants were more prevalent on the Z chromosomes when compared with the autosomes (Hooper and Price 2017) and were more likely to be fixed between species pairs with overlapping geographic distribution, suggesting an effect of reinforcement on the fixation of divergent structural variants. Finally, genes involved in sexual dimorphism and mate preferences are also frequently found on the sex chromosomes and may also contribute to premating isolation: for instance, in closely related species of Ficedula flycatchers, loci controlling plumage divergence and preference for plumage coloration are located on the Z chromosomes, and are likely to enhance premating barriers (Sæther et al. 2007). Sex chromosomes may thus play a prominent role in the evolution of reproductive isolation in closely related species living in sympatry. In Lepidoptera, evidence for a significant role of the Z chromosome in reinforcement has been reported. In a pair of sister-species of Colias butterflies with overlapping geographic range, increased genetic divergence in the Z chromosomes when compared with the autosomes was observed. Moreover, loci controlling for divergent coloration in males are located on the Z chromosome, suggesting a key role of the Z chromosome in the prezygotic barrier to gene flow in these species (Ficarrotta et al. 2022).

Here, we focus on the genomic divergence between species of Morpho butterflies, where multiple species currently coexist in sympatry in the Amazonian rainforest (Blandin and Purser 2013). In the genus Morpho, sympatric species show divergence in (1) flight height, with specialization in the canopy vs. understory microhabitat (Le Roy et al. 2021a) and in (2) flight time (Le Roy et al. 2021b). Males from some species only fly early in the morning, while males from other species typically patrol at noon (Gayman et al. 2016). Such specialization into different forest strata and/or temporal niches likely contributes to the coexistence of these closely related species in sympatry and may also have played a role in the speciation process itself. In the noctuid moth Spodoptera frugiperda, populations specialized on different host-plants diverge in the timing of mating, and the gene vrille, involved in the control of the circadian clock, was shown to diverge between these species (Hänniger et al. 2017). The evolution of circadian genes may thus be involved in the divergence of temporal niches in Morpho butterflies, therefore enhancing prezygotic barrier to gene flow in sympatric species. Interestingly, in Lepidoptera, some genes involved in the regulation of circadian behavior are sex-linked: for instance, the gene Period, responsible for diurnal or nocturnal activity in some butterflies and moths or the gene Clock, involved in the photoperiodic response and migration in the Monarch buterfly, are both located on the Z chromosome (Gotter et al. 1999; Iiams et al. 2019; Wang et al. 2023). We thus aimed at comparing the evolution of genomic divergence between the Z chromosomes and the autosomes in sympatric vs. allopatric species of Morpho, and specifically test for signals of positive selection.

We report the sequencing, assembly and annotation of the genomes of eight species from the genus Morpho based on long read data for individual butterflies which are near chromosome complete. We then used these new genome assemblies, as well as the previously published genomes of three other Morpho species (Bastide et al. 2023) to perform comparative genomic analyses in 11 out of the 30 species of the genus Morpho. We specifically compare the evolution of the Z chromosome vs. the autosomes, in sympatric and allopatric species, specialized in different ecological niches: we investigate inversions and ratio of synonym vs. nonsynonym variations in the region of the Z chromosome vs. the autosomes. We also investigate the location and evolution of eight genes known to regulate circadian activities in other Lepidoptera (Clock, Cycle, Period, Timeless, Cry1, Cry2, Vrille, and PDP1) (Reppert 2006; Brady et al. 2021), and specifically test for signature of positive selection on these genes, potentially involved in prezygotic isolation.

Results

 

Nuclear and Mitochondrial Genome Assembly

For the eight genomes analyzed here, GenomeScope analyses suggested very high levels of heterozygosity going from 1.01% in Morpho granadensis to 3.38% for M. menelaus (supplementary material S1, Supplementary Material online). The assemblies produced by Hifiasm showed nuclear genome sizes of the Morpho species ranging from 375 to 514 Mb (see Table 1). Consistent with other Lepidoptera and with the known genomes of other Morpho species. In all cases, BUSCO results showed a very high level of duplicated sequences, when the assembly was performed using the option $-$l0 (no purge) in Hifiasm and in most cases, we had to use the option $-$l3 (most aggressive purge). The BUSCO values then showed low levels of duplicates in the assemblies for M. telemachus (for the blue and gold morphs), M. amathonte, and M. granadensis. In contrast, M. hecuba, M. menelaus, M. eugenia, and M. marcus genome assemblies still showed high levels of duplicates despite the use of the l3 option and needed to be further purged using Purge_dups. In the case of M. rhetenor, the best assembly was obtained without purging in Hifiasm, but with the use of Purge_dups. Final BUSCO results are shown in the supplementary material figure S1a, Supplementary Material online). Mitochondrial genomes assembled using Rebaler presented similar sizes to the previously assembled genomes of other Morpho species (Bastide et al. 2023). For the nine individuals sequenced here, the Rebaler software detected mitochondrial genomes, with sizes ranging from 13,469 bp for the gray morph of $M.$  telemachus to 16,173 bp in $M.$  amathonte.

Table 1.

Heterozygosity, genome size, number of contigs, N50, and percentage of the genome covered by repetitive elements (RE) for the assemblies of eight Morpho species

	Heterozygosity (%)	Genome size	Total contigs	N50	Content of RE (%)
M. rhetenor	2.50	375.99 Mb	65	13 Mb	40.43
$M.$  hecuba	1.71	448.02 Mb	64	14 Mb	49.52
$M.$  amathonte	1.69	467.10 Mb	81	13 Mb	49.89
$M.$  menelaus	3.38	412.92 Mb	34	12 Mb	41.46
$M.$  granadensis	1.00	448.90 Mb	100	12 Mb	49.95
$M.$  eugenia	2.53	514.40 Mb	34	11 Mb	53.43
$M.$  marcus	2.36	457.33 Mb	56	11 Mb	47.92
$M.$  telemachus gold	1.19	447.12 Mb	62	14 Mb	51.14
$M.$  telemachus blue	1.11	449.72 Mb	68	13 Mb	50.09

Assemblies were performed using Hifiasm and were purged using Purge_dups when needed based on BUSCO results on the assembly. The number of total contigs corresponds to the number of contigs obtained after assembly and before any further treatment of the assemblies.

Genome Annotation

The use of RepeatMasker and RepeatModeler produced results similar to previous ones on other Morpho genomes (Bastide et al. 2023): repeated elements account for 40% to 53% of the genomes of the nine Morpho species sequenced here (Table 1). By re-annotating the three previously published Morpho genomes (M. helenor, M. achilles, and M. deidamia), we found significantly more genes predicted and better proteome BUSCO scores using BRAKER2 than with the previous methodology using Maker v2.31.10 (Bastide et al. 2023). Between 18,159 and 21,017, protein coding genes were annotated within each genome of most Morpho species studied here (see supplementary material table S3, Supplementary Material online). These values fall within the range observed in other Nymphalidae species, as Euphydryas editha (Tunstrom et al. 2022). For M. helenor, the annotation pipeline predicted 29,734 complete genes, a number significantly higher than those for other Morpho species and Nymphalidae in general. The use of ContScout revealed that the excess of annotated genes resulted from a high number of foreign sequences, principally from bacteria (see supplementary material table S3, Supplementary Material online). All these contaminations were removed from the assembly and were excluded from further analyses.

Phylogenetic Relationships between Morpho Species

We inferred the phylogenetic relationships among samples using 8,780 single-copy orthologs detected in all 12 available genomes of Morpho, using IQTree and ModelFinder. This whole-genome phylogeny is consistent with the published phylogeny of the group, inferred from mitochondrial and nuclear genes (Chazot et al. 2016, 2021). All the canopy species form a single monophyletic clade while understory species are distributed in two clades (Fig. 1). The divergence between the “blue” and the “gold” morphs within the species M. telemachus is very low ($\text{dxy}=0.02$) and the branch much shorter than those separating sister-species like $M.helenor$ and $M.achilles$, or $M.eugenia$ and $M.marcus$. This confirms that these two syntopic color morphs belong, in fact, to the same polymorphic species.

Fig. 1.

Phylogenetic relationships between the 12 Morpho samples inferred using IQ-TREE together with Q.mammal+F+I+R8 substitution model and ultrafast bootstrap with 1,000 replicates. Analyses were performed with a supermatrix from 8,780 single-copy orthologs extracted with OrthoFinder and found in all the samples. Bootstrap values are displayed next to the nodes. Additionally, the ecological niche (canopy or understory) and species distribution (SA, South America or CA, Central America) are indicated.

Synteny and Detection of Chromosomal Rearrangements between Morpho Genomes

After assembling the different contigs within each Morpho species, we identified a contig that shares synteny with the Z contig previously identified in M. helenor (Bastide et al. 2023). The size of this contig varied across species, ranging from approximately 18 to 25 Mb. Notably, it was the largest contig within each assembly, for all species except M. marcus and M. eugenia (supplementary material figure S2, Supplementary Material online). Although our assemblies did not reach chromosome-level resolution, the size of this contig was similar to the Z chromosome size observed in related species with high-quality chromosome-level assemblies, such as Maniola jurtina (17 Mb) (Lohse et al. 2021), Pararge aegeria (27 Mb https://www.darwintreeoflife.org/project-resources/), or Bicyclus anynana (24 Mb) (Saccheri 2023). This observation suggests that the identified Z contig may closely approximate the true Z chromosome size in Morpho.

D-genies whole-genome paired alignments showed strong synteny in all autosomes across the majority of closely related species (supplementary material figure S3, Supplementary Material online). On the other hand, they revealed inversions in the Z contig among pairs of closely related species (supplementary material figure S3, Supplementary Material online): three among the seven paired comparisons unveiled variable-sized inversions (Fig. 2). Further analyses using SyRI revealed, for the species living in the understory, three inversions measuring a total of more than 7 Mb positioned at the tips of the Z contig among M. granadensis and M. helenor and three inversions measuring together 2 Mb between M. helenor and M. achilles (Fig. 2a). Within the canopy clade, a 2.5 Mb inversion was identified between $M.$  telemachus and M. hecuba (Fig. 2b). While D-genies whole-genome comparisons uncovered two small inversions between M. telemachus and M. rhetenor (see supplementary material S3, Supplementary Material online), these were not observed in the paired comparisons conducted with SyRI (see Fig. 2b). SyRI plots between these two species revealed lower synteny in the Z contig compared to other species pairs. Additionally, the contig in M. rhetenor showing the highest synteny to the Z contig in M. telemachus was much smaller for M. rhetenor suggesting that the regions capturing the inversions observed in whole-genome comparisons were not included when contigs were isolated for analysis with SyRI. Comparisons also showed two pairs of closely related species without any inversions and with a very high synteny level along the entire Z contig: M. menelaus vs. M. amathonte and M. eugenia vs. M. marcus (Fig. 2c,d). Nevertheless, for these species, small inversions were found in the autosomes. The comparison between M. menelaus vs. M. amathonte revealed a 2,80,000 bp inversion in the contig ptg000019l (supplementary material figure S2a, Supplementary Material online) while for M. eugenia vs. M. marcus small inversions between 3,92,980 and 431811 bp were found when comparing the contig ptg000027l of M. eugenia vs. the equivalent contig ptg000011l of M. marcus. We also found inversions of between 7,14,789 and 8,95,465 bp when comparing contigs ptg000024l of M. eugenia and the equivalent contig ptg000013l of M. marcus (supplementary material figure S2 (b) and (c), Supplementary Material online, respectively). Overall, our results revealed a higher frequency of larger inversions on the Z contig compared to the autosomes. However, note that we restricted our inversion detection to contigs larger than 10 Mb (or 5 Mb in the case of M. hecuba, see Materials and Methods), preventing us to observe inversions occurring in smaller-sized autosomal contigs.

Fig. 2.

Paired comparisons of the region corresponding to the Z contig between closely related Morpho species. The species compared are a) understory species $M.$  granadensis and sister-species $M.$  helenor and $M.$  achilles, b) all the species belonging to the canopy clade, c) understory sister-species $M.$  amathonte and $M.$  $menelaus,$ and d) basal clade $M.$  eugenia and $M.$  marcus. Inversions are visible in dark yellow and syntenic regions in light gray

Higher dN/dS on the Z Chromosome and Signatures of Positive Selection in Circadian Genes

The dN/dS ratio (ω) of all orthologous genes located on the Z contig (148 genes) vs. the autosomal contigs (8,209 genes) was first measured throughout the 12 Morpho species. Using the gapless ortholog alignements, we found that the genes located on the Z contig have a significantly higher ω than the genes located on the autosomal contigs (Wilcoxon test: $P\text{-value}=0.0004996$, Fig. 3a). The same trend is found when calculating the ω ratio using the raw ortholog alignments, supplementary material figure S4, Supplementary Material online. Because the number of genes on the Z contig was much lower than the genes located on the autosomes, we performed permutation tests. These binomial tests showed that the P-values calculated are significantly lower than the 0.05 threshold ($P\text{-value}<2.2{\times 10}^{-16}$ for both distributions, raw et gapless orthologs), confirming that the ω ratio of the genes found on the Z contigs is indeed higher than the ω ratio of the genes found on the autosomal contigs (supplementary material figure S5, Supplementary Material online).

Fig. 3.

Comparison of the dN/dS values measured for a) the 8,357 genes located in either the autosomal contigs (in white) or the Z contig (in blue). The ω ratio values were computed using the one-ratio model implemented in PAML on gapless ortholog alignments, and are represented in logarithmic scale. b) Mean ω for autosomes and chromosome Z estimated using the YN00 program with the method of Yang and Nielsen for pairwise comparison. Different pairwise comparisons were made: between the two species from the basal clade, between species from the understory (species pairs found in sympatry (S) and allopatry (A), between species from the canopy and finally, mixed comparisons between species from different ecological niches: canopy vs. understory (DN). Asterisk represents a significant difference in the ω value between autosomes and the chromosome Z ($P<0.05$).

To test for divergent evolution in species evolving in the canopy vs. the understory, we used a clade model to compare the evolutionary rates of genes located on the Z chromosome and autosomes between species from the canopy and those from the understory. This test revealed that approximately 4.6% of the genes on the autosomes exhibited significant differences in evolutionary rates between the canopy and understory clades, while only 0.68% of the genes on the Z chromosome showed such significant differences. A Z-test for proportions showed a significant difference between the autosomal and Z chromosome groups (Z-value: 2.27, P-value: 0.02) (supplementary material figure S6, Supplementary Material online). This analysis focused on the genes shared by both clades and did not detect a prominent role of Z-linked genes in the divergent adaptation to different habitats. We then analysed of species within each clade separately (canopy or understory), allowing to study clade-specific genes as well. We estimated differences in the evolutionary rates between autosomes and the Z chromosome using the M0 model in codeml. This analysis revealed significant differences in the mean dN/dS ratio between genes located on the autosomes and those on the Z chromosome, but the direction of these differences was opposite between the two groups. For the eight understory species, the mean ω was significantly higher for genes on the Z chromosome than for those on the autosomes (Wilcoxon permutation test: $Z=-2.58$, $P\text{-value}=0.0098$, supplementary material figure S7, Supplementary Material online). In the canopy species however, the mean ω was higher for genes on the autosomes than for those on the Z chromosome (Wilcoxon permutation test: $Z=-5.14$, $P\text{-value}=1{\times 10}^{-05}$). Nevertheless, in this group, we only had three species, therefore questioning the robustness of this last result. These clade-specific analyses suggest that adaptation to the micro-habitats is probably not the main driver of the enhanced dN/dS ratio in the Z-contig when compared with the autosomes, detected at the genus scale.

We then tested whether the difference between the gene evolution in the Z-contigs and the autosomal contigs was more prevalent in closely related species living in sympatry, in line with putative effect of reinforcement of species barriers. We thus computed dN/dS ratio for different species pairs using the YN00 program. These pairwaise analyses revealed significant differences in mean ω between autosomes and chromosome Z in all comparisons involving pairs of canopy species and pairs of closely related species from the understory (i.e. pairs belonging to the granadensis achilles helenor deidamia clade), regardless of whether they were sympatric or allopatric (Fig. 3b). In contrast, no significant differences in mean ω between autosomes and chromosome Z were found for the basal clade pair (M. marcus and M. eugenia), the two closely related understory species M. amathonte and M. menelaus, or in pairs involving species from different ecological habitats (Fig. 3b). Heterospecific interactions in sympatry are thus not specifically associated with increased differences in the evolution of genes located in the Z-contigs vs. the autosomes, but this difference observed at the genus scale, is probably mostly driven by divergent evolution of the Z chromosome between closely related species.

Positive selection tests were also specifically computed on eight circadian genes. The random site models implemented in PAML applied to our 12 Morpho samples did not reveal any signal of positive selection on any of the eight tested circadian genes (supplementary material table S4, Supplementary Material online). However, using a branch site model, evidence of positive selection was found in one of these genes when comparing the canopy vs. understory species: a significant signature of positive selection in the canopy lineage was found in Period, located on the Z contig ($P\text{-value}=0.03$, see supplementary material table S5, Supplementary Material online). Such positive selection detected on sites where amino acids are different between the canopy and the understory species suggests that different selective pressures may affect the evolution of the Period gene in the two micro-habitats (Fig. 4).

Fig. 4.

Partial amino acid alignment of the gene Period in Morpho species from the understory and the canopy, showing only the amino acids with a signature of positive selection using the branch site model in the PamL analysis. The position of these amino acids along the Period protein gene are indicated above in color. The values of the posterior probability associated with each site are indicated in colors, red indicating the highest probability (95% to 100%) and green the lowest probability (50% to 65%).

Introgression among Sympatric Species

To test for gene flow among M. granadensis and M. helenor living in sympatry in Central America, we carried out ABBA-BABA tests, comparing the amount of shared genetic variation between these two species and the closely related species M. achilles and M. deidamia. All contigs show D-statistics values close to 0 ($-0.10<D<-0.03$), without any significantly departure from the mean (Z score $<3$).

Discussion

By comparing the genomes of 11 Morpho species, our study suggests an increased evolutionary rate of the Z sex contig relative to the autosomal contigs in this genus. We specifically found an elevated $dN/dS$ ratio in this contig when compared with the autosomal contigs. The prevalence of inversions was also much higher in the Z contigs, especially between closely related species sharing a similar ecological niche and similar flight hours in sympatry. We discuss below the demographic and ecological factors that may have led to this accelerated divergence of the Z chromosome when compared with the autosomes.

 

Prevalent Inversions in the Z Chromosome of Sympatric Sister-Species

In the genus Morpho, we found that big inversions of more than 1 Mb on the Z chromosome were more common among closely related species coexisting in sympatry, particularly between species with overlapping temporal niches, such as the canopy species M. hecuba and M. telemachus, or the understory sister-species M. achilles and M. helenor and M. granadensis and M. helenor (Gayman et al. 2016; Le Roy et al. 2021b). Conversely, big inversions were not observed in comparisons between closely related sympatric species with nonoverlapping temporal niches, such as the basal sister-species M. eugenia (which flies very early in the morning around 6:00 AM) and M. marcus (which flies much later around 9:00 AM) (Gayman et al. 2016), or between species found in allopatry, such as M. amathonte and M. menelaus. Chromosomal inversions have long been acknowledged as a crucial mechanism acting as barrier to gene exchange, ultimately driving speciation (Noor et al. 2001). Our results are thus consistent with a putative role of inversions located on the Z chromosome in the speciation and/or reinforcement process in the butterfly genus Morpho. Structural variations within each species now need to be investigated to test for fixation within species and better characterize the selective regime acting on the inversions within species.

The effect of the Z chromosome as a barrier to gene flow should be also reflected in the introgression rates between sympatric and closely related species, similarly to what has been found in Heliconius butterflies (Martin et al. 2013). Here, we did not detect any significant introgression between the closely related species M. helenor and M. granadensis neither in the Z-contig nor in the autosomal contigs. Given the distribution of the species studied, we could only test for introgression between M. helenor and M. granadensis, whose distribution partially overlap in Central America. But the sister-species of M. helenor, namely M. achilles, also has an overlapping geographic range with M. helenor in the amazonian basin, so that gene flow could occur both between M. helenor and M. granadensis in Central America and between M. helenor and M. achilles in South America, strongly limiting the power of the ABBA/BABA test. Investigating the intra-specific genetic variability within these three species throughout South and Central America will allow clarifying the level of introgression in the Z vs. autosomal contigs in the future.

Evolution of Circadian Genes

Since Morpho species living in sympatry display temporal segregation throughout the day (Le Roy et al. 2021b), we specifically examined signatures of selection in various circadian genes previously documented in other Lepidoptera (Merlin and Reppert 2009). We found evidence of positive selection on specific sites within the Period gene, that is located on the Z chromosome in Morpho, as well as in all Lepidoptera species studied so far (Sandrelli et al. 2008). Among circadian genes, Period has been suggested to be involved in sympatric speciation across multiple taxa: variation in the amino acid sequence of the Period protein, and variation in the regulatory regions of the Period gene have been shown to be involved in allochrony, i.e. differences in breeding time among conspecific individuals, that may facilitate speciation (Taylor and Friesen 2017). Experiments involving transgenic flies conducted by (Tauber et al. 2003) indicated that Period played a role in speciation in both D. melanogaster and D. pseudoobscura, by exerting control over the precise timing of mating behavior (Tauber et al. 2003). In the fly Bactrocera cucurbitae, Period was found to contribute to premating isolation in two lines with distinct circadian periods, one short and one long (Miyatake et al. 2002).

In Lepidoptera, differences in diel activity are assumed to contribute to prezygotic reproductive isolation, as evidenced by the temporal segregation across 400 co-occurring species of Neotropical Hesperiidae (Devries et al. 2008). Nevertheless, the role of circadian genes in regulating these differences or their putative role in speciation has been scarcely studied, except in a few moth species. In Spodoptera frugiperda, the circadian gene vrille was found responsible for allochronic differentiation in the mating time of two different strains, by acting as a premating barrier (Hänniger et al. 2017). In the moth Plutella xylostella, the CRISPR/Cas9-mediated genome editing to knockout the Period gene changed circadian rhythms patterns in pupal eclosion, mating, egg-laying and egg hatching, suggesting that Period may be implicated in the timing of mating behavior, as well as in other components of daily temporal niche in Lepidoptera (Wang et al. 2023). The detection of a signal of positive selection on the Z-located gene Period in the genus Morpho could therefore be a consequence of the specialization into different ecological niches, as a result of reinforcement. It may also suggest a role for Period in the speciation process, by contributing to a prezygotic barrier.

Divergent Evolution on the Z chromosome in Closely Related Species

The observed faster evolution of genes located in the Z chromosome throughout the Morpho genus aligns with findings in other lepidoptera species as the Carolina sphinx moth (Manduca sexta), the monarch (Danaus plexippus) or Leptidea sinapis, assessed using $dN/dS$ ratios (Mongue et al. 2022; Höök et al. 2024). However, other studies in butterflies did not detect a faster Z chromosome evolution. In the nymphalid species Maniola jurtina and Pyronia tithonus, transcriptomic data showed a slightly slower evolutionary rate on the Z chromosome compared to autosomes (Rousselle et al. 2016). In the genus Morpho, we observed a higher evolutionary rate for genes located on the Z chromosome compared to those on the autosomes. This pattern was consistent both when analyzing all species together and in most of the pairwise comparisons between closely related species. Nevertheless, no significant differences in evolutionary rates between autosomes and the Z chromosome were found in species pairs from distinct ecological niches, and the test for divergent evolutionary direction in the canopy vs. understory species did not reveal a more pronounced divergence in the Z-contigs genes as compared to the autosomal genes. In contrast, closely related species pairs, and in particular species pairs with overlapping habitats and flight times, all showed a significantly different evolution between the Z and autosomal contigs.

Nevertheless, the contribution of different selection regimes acting on the Z chromosome vs. the autosomes to the variation in $dN/dS$ observed in our study can not be fully estimated because of several limitations. The Morpho genomes studied here display a high level of heterozygosity, so that the substitutions observed between species are not necessarily fixed within species, and such intra-specific polymorphism might skew our estimation of $dN/dS$ between closely related species (Mugal et al. 2020). The difference in effective population sizes between the Z chromosome and the autosomes could thus partly explain the observed difference in $dN/dS$. Our analyses were also conducted on a subset of genes, specifically single-copy orthologs, identified through OrthoFinder and common across all genomes. This subset, representing genes found in all Lepidoptera, may exclude genes that are unique to the Morpho genus.

Interestingly, the significant differences in evolutionary rates between the Z-contig and autosomes appeared to be associated with the presence of big genomic inversions on the Z-contig. In Manduca sexta, an inversion of 1 Mb in the chromosome Z seems to evolve under positive selection, driving the inversion to near fixation. This inversion is considered adaptive as it contains a significant excess of “unbiased genes” (genes expressed in both sexes) while the rest of the Z chromosome appears to have male-biased gene expression. As unbiased genes will be in hemizygous state in females, an inversion could be fixed if recessive beneficial mutations are exposed to selection (Mongue and Kawahara 2022). The comparison of clade-specific analysis and pairwise analyses of dN/dS suggests that the acceleration divergence in Z-located genes as compared to the autosomes is particularly marked in closely related species living in sympatry and exhibiting inversions. These findings are consistent with an effect of the Z chromosome divergence, fueled by both chromosomal rearrangement and variation in protein sequences on the sex-chromosome in species divergence.

Conclusions

Altogether, our results evidence the accelerated evolution of the Z chromosome when compared with the autosomes throughout the butterfly genus Morpho, especially between closely related species living in sympatry. Whether reproductive interferences between sympatric species fuel the accelerated evolution of this sex chromosome remains an open question. The potential role of the Z chromosome in prezygotic and postzygotic barrier to gene flow needs to be investigated in further studies using sympatric and allopatric populations of different species.

Materials and Methods

The nuclear and mitochondrial assemblies of the genomes of Morpho helenor, M. achilles, and M. deidamia were previously described in Bastide et al. (2023). Here, we sequenced, assembled and annotated the whole genome of eight additional Morpho species: five understory species—M. amathonte, M. menelaus, M. granadensis, M. eugenia, and M. marcus—and three canopy species—M. rhetenor, M. hecuba, and M. telemachus. For this last species, two different morphs coexist in sympatry: males with blue/gray or golden wings, and we sequenced one individual of each morph.

 

Butterfly Sampling

The male butterflies used for sequencing in this study were collected with handnets in Panama, Veraguas district (GPS coordinates: 8.525869707300163, $-$81.13020529212375) in July 2022 for M. granadensis and M. amathonte (Export permit number: PA-01-ARB-076-2022) and in French Guiana, Roura district (GPS coordinates: 4.5917807378071975, $-$52.20357910219149) in october 2022 for M. eugenia, M. marcus, M. hecuba, M. telemachus gray and gold morphs, M. menelaus, M. rhetenor (authorization number: R03-22-10-18-0001). The butterfly samples were then SNAP-frozen and kept at $-{80}^{\circ }$C.

DNA Extractions and Genome Sequencing

For genomic analyses, the DNA extraction was carried out from the thorax muscles using the Qiagen Genomic-tip 100/G kit, following supplier instructions. Library preparation and sequencing were performed according to the manufacturer’s instructions “Procedure & Checklist—Preparing whole genome and metagenome libraries using SMRTbellⓇprep kit 3.0.” At each step, DNA was quantified using the Qubit dsDNA HS Assay Kit (Life Technologies). DNA purity was tested using the nanodrop (Thermofisher) and size distribution and degradation assessed using the Femto pulse Genomic DNA 165 kb Kit (Agilent). Purification steps were performed using AMPure PB beads (PacBio) and SMRTbell cleanup beads (PacBio). For 9 samples, 5$\mu$g of DNA was purified then repaired using “SMRTbell Damage Repair Kit SPV3” (PacBio) and sheared between 15 and 20 kb using the Megaruptor 1 system (Diagenode). For Morpho amathonte, M. granadensis and M. telemachus gold samples, a single library was prepared. For the other six samples, multiplex libraries were prepared using SMRTbell Barcoded Adaptater kit 8A for M. hecuba/M. rhetenor, M. telemachus gray/M. marcus, and M. eugenia/M. menelaus. A nuclease treatment was performed on the six libraries followed by a size selection step using a 6, 8, or 10 kb cut off on the PippinHT Size Selection system (Sage Science) with “0.75% Agarose, 6 to 10 kb High Pass, 75E protocol.” Using Binding kit 3.2 kit and sequencing kit 2.0, libraries were sequenced with the adaptive loading method onto 6 SMRTcells (+2 extra SMRTcells for M. menelaus sample) on Sequel2 instrument at 90 pM with a 2 h pre-extension and a 30 h movie.

Nuclear and Mitochondrial Genome Assembly

Heterozygosity for each PacBio dataset was estimated with Jellyfish (v.2.3.0) (Marcais and Kingsford 2011) (jellyfish count and hist options). The resulting histograms were visualized with GenomeScope online version 2.0 (Ranallo-Benavidez et al. 2020). Assemblies were performed using Hifiasm v0.19.5 (Cheng et al. 2021). Lepidoptera genomes are particularly heterozygous and high heterozygosity imposes a challenge for most genome assemblers as it may be difficult to make the distinction between different alleles at the same locus and paralogs at different loci. Raw assemblies from heterozygous genomes are thus expected to contain high levels of false duplications (Asalone et al. 2020). Hifiasm allows using different levels of purge to remove redundant contigs. We systematically produced assemblies using two different purge options: no purge and purge in the most aggressive way (options $-$l0 and $-$l3, respectively). For each assembly, we estimated basic statistics (number of contigs, size of contigs, N50) using the “stats.sh” program from the BBMap v38.93 package (Bushnell 2014). We assessed the completeness of assemblies obtained with Hifiasm using the different options with BUSCO v5.2.2 and MetaEuk for gene prediction against the $lepidoptera\mathrm{\_}odb10$ database (Manni et al. 2021). In several cases, the BUSCO results showed a high level of duplications in the final assembly, even when using the aggressive mode in hifiasm. In these cases, we used Purge_dups v1.2.5 to remove false haplotypic duplications, and then we reassessed completeness using BUSCO. For every species, we chose the final assembly among the different assemblies produced with or without the use of Purge_dups, based in the basic statistics and the BUSCO score. The mitochondrial genome for all Morpho species was assembled directly from the PacBio Hifi reads with Rebaler (https://github.com/rrwick/Rebaler) as described in Bastide et al. (2023) and using the mitochondrial genome from M. helenor as a reference.

All sequences obtained in this study have been submitted to NCBI GenBank.

Genome Annotation and Identification of Circadian Genes

For each Morpho species, we created a de novo library of repetitive elements from each assembled genome with RepeatModeler v2.0.2a (Flynn et al. 2020) with the option $-$s (slow search). This library was then used with RepeatMasker 4.1.2.p1 (Flynn et al. 2020) to produce softmasked assemblies needed for the annotation.

Structural annotation was carried out with BRAKER v2.1.6 (Hoff et al. 2016; Bruna et al. 2021), using the $arthropoda\mathrm{\_}odb10$ protein data set from OrthoDB (Kriventseva et al. 2019). BRAKER2 ab initio gene predictions were performed with AUGUSTUS v3.4.0b (Buchfink et al. 2015) and GeneMark-EP+ v4.64 (Lomsadze et al. 2005) with two rounds of training. The resulting .gff files containing predicted genes were filtered to keep only the longest isoform and were screened to remove incomplete gene coding models (i.e. genes coding with start and/or stop codon missing in its cds) with AGAT (Dainat et al. 2020). The filtered .gff files produced were used in the following steps.

The functional annotation was added to the BRAKER2 predictions in two parts, first, we ran InterProScan v5.33-72.0 (Quevillon et al. 2005) on the files produced after two rounds of BRAKER2 using the option $-$f XML. Second, the XML file produced by InterProScan and the previously filtered .gff file were used in Funannotate v1.8.13 (Palmer and Stajich 2016) with the option “annotate” to obtain a multi-fasta file of protein coding genes for each species. We followed this procedure on the nine new genomes assembled in this paper and on the three previously published genomes of M. helenor, M. achilles, and M. deidamia, to obtain comparable annotations throughout the 12 genomes. To assess the quality of the gene-set predicted for each species, we ran BUSCO v5.2.2 analyses on the protein file generated BRAKER2 with selection of the protein mode and the $lepidoptera\mathrm{\_}odb10$ database. Finally, identification and removal of foreign sequences from annotations were performed using ContScout (Bálint et al. 2024)

We then specifically studied eight different genes known to be involved in the circadian clock in butterflies: Clock, Cycle, Period, Timeless, Cry1, Cry2, Vrille, and PDP1. The position, the proteic, and nucleotidic sequences of the genes were extracted from the final gff files produced at the annotation step.

Phylogenetic Relationships between Morpho Species

To infer the phylogenetic relationships between the 12 sequenced genomes of Morpho butterflies, we followed the procedure explained in (Wright et al. 2024). Single-copy orthologs common to all 12 genomes were extracted using OrthoFinder v2.5.5 (Emms and Kelly 2019). Orthologs were aligned using Mafft with default options (Katoh et al. 2002) and trimmed using trimAl (Capella-Gutiérrez et al. 2009). Trimmed alignments were concatenated to produce a phylip supermatrix with catfasta2phyml (https://github.com/nylander/catfasta2phyml). This matrix was used in IQ-TREE (Nguyen et al. 2015) to infer the species tree under the Q.mammal+F+I+R8 substitution model and ultrafast bootstrap with 1,000 replicates (Hoang et al. 2018) based on the Bayesian information criterion (BIC) after ModelFinder analyses (Kalyaanamoorthy et al. 2017). The consensus tree was visualized using FigTree v1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/). In addition, for the two individuals of Morpho telemachus, we aligned the OPacBio HiFi reads and called SNPs as for the ABBA-BABA analysis and computed a genome-wide dxy value using the scripts available at https://github.com/simonhmartin/genomics_general.

ABBA-BABA Scans for Introgression

To test for potential introgression events between the genomes of the sympatric species in the understory clade, we applied ABBA-BABA tests (Green et al. 2010). This test relies on a four-species tree [(((P1,P2),P3),O)] to detect the site trees deviating from the expectation based on the species tree (i.e. ABBA or BABA, whereas we should expect AABB or AAAB) (Martin et al. 2015). Thus, we first selected two species groups that could be tested under this framework: Morpho achilles (P1), M. helenor (P2), M. granadensis (P3), M. deidamia (outgroup). Then, we aligned the PacBio HiFi reads to one reference genome (M. helenor) using Minimap2 v.2.8 and we called SNPs using VarScan v2.4.6. The resulting VCF file was filtered keeping only biallelic snps and removing indels. Positions associated with a read depth lower than 15 or greater than 150, with general quality lower than 30 or with missing genotype in at least one sample were converted to missing. This filtered vcf was then phased using beagle v5.4 and processed to obtain the allelic frequencies per each site using the python script freq.py available at https://github.com/simonhmartin. D statistic was then computed using the scripts for ABBA BABA https://github.com/simonhmartin and plotted for each contig of the reference genome greater than 10 Mb. This threshold was chosen to avoid the comparison between contigs that differ too much in their size (and thus in the number of sites used to obtain the chromosome-wide D mean), and facilitate the comparison with the analysis performed with syri, presented in the following paragraph.

Identification of the Z Contig, Synteny, and Detection of Chromosomal Rearrangements between Morpho Genomes

We have previously identified the contig corresponding to the Z chromosome for Morpho helenor, M. achilles, and M. deidamia, by comparing the assemblies produced by Hifiasm to the genome of the species Maniola jurtina, as this species represents the closest relative Morpho for which a high-quality chromosome-level assembly is available (Bastide et al. 2023). Here, we used D-genies to pair-map the genome of $M.$  helenor the newly assembled Morpho species genome (M. rhetenor, M. hecuba, M. telemachus, M. amathonte, M. menelaus, M. granadensis, M. eugenia, and M. marcus) treating $M.$  helenor genome as the target reference and the new Morpho species genome as the query. This alignment permitted us to visually find, for each species, the contig corresponding to the Z contig of $M.$  helenor.

To study the synteny and major rearrangements all over the genome, we aligned the genomes of different species by pair, using the online version of D-genies (Cabanettes and Klopp 2018). Given our focus on synteny among closely related species, we conducted comparisons involving all species within the “Canopy” clade. Specifically, three comparisons were performed: Morpho hecuba vs. M. telemachus blue, M. telemachus blue vs. M. telemachus gold, and M. telemachus gold vs. M. rhetenor. For the “Understory” clade, we also performed comparisons between close species: M. amathonte vs M. menelaus, M. granadensis vs M. helenor and M. helenor vs. M. achilles. Finally, we added the paired analysis of the two basal sister species: M. eugenia vs. M. marcus. Paired alignments were performed using the Minimap2 aligner. (Li 2018) in D-genies (see supplementary material figure S3, Supplementary Material online). To facilitate cross-species whole-genome comparisons, we standardized the assemblies by filtering out contigs smaller than 10 Mb, retaining approximately 90% of the total genome content for most species. However, in the case of M. hecuba, this threshold only covered 77% of the genome. Therefore, for this species, we adjusted the threshold to retain contigs larger than 5 Mb, ensuring coverage of 90% of the genome. Since we were particularly interested in genomic inversions, we visualy inspected the paired analyses produced by D-genies contig by contig. For any contig pair in which inversions were detected, we conducted additional analysis using SyRI, as this tool provides information on the size of each inversion. We were unable to conduct whole-genome comparisons with SyRI due to its requirement for the two genomes to be homologous and have matching contig names, typically necessitating chromosome-level assemblies. Since our assemblies lacked chromosome-level resolution and varied in contig number and nomenclature, we only used SyRI for contig-by-contig comparisons. Paired comparisons followed the methodology described in Bastide et al. (2023). Briefly, paired alignments of the equivalent contigs were generated using Minimap2, and the resulting SAM files were analyzed with SyRI to detect inversions (Goel et al. 2019). The genomic structures predicted by SyRI were visually represented using plotsr (Goel and Schneeberger 2022).

Detection of Positive Selection in Circadian Genes and Throughout the Whole Genome

To explore the genetic basis of the divergence in temporal niches, we searched for signature of selection in eight circadian genes (Clock, Cycle, Period, Timeless, Cry1, Cry2, Vrille, and PDP1). We first generated trees for each gene with IQ-TREE (Nguyen et al. 2015) and ModelFinder. These trees were then used in the Codeml program in the PAML package (version 4.8) (Yang 1997; Álvarez-Carretero et al. 2023) to detect signals of positive selection. PAML estimates the omega ratio (c) using maximum likelihood. The omega ratio compares nonsynonymous (dN) against synonymous (dS) substitutions per site. We used random site models to test for positive selection at the codon scale on the genes. Branch site models were also computed to estimate selective pressures at codon sites within different lineages. Three types of hypotheses were tested under the PAML branch site model: (1) we searched for differences of positive selection on the branch leading to the two most basal Morpho species: $M.marcus$ and $M.eugenia$  vs. as this clade was the earliest diverging clade in the genus around 28 MYA (Chazot et al. 2021). We labeled these two species as foreground branches and the rest of the species as background branches; (2) as the evolution of circadian genes might be linked to differences in the amount of light available, we looked for traces of positive selection on the lineages of the Morpho species living in the canopy $vs.$ the understory; and (3) because time activity is expected to be influenced by circadian genes, we looked for evidence of positive selection in the species where males display an early flight activity (patrolling from 6:00 to 10:00 AM) $vs.$ a late flight activity (patrolling after 10:00 AM). All models were always compared to a null model, assuming an omega ratio of 1 for all sites and across all lineages. To determine significance, the null model and the positive selection model were compared using a likelihood ratio test (LRT) following a Chi2 distribution, and Bayes empirical bayes (BEB) analyses were run to identify the positively selected amino acids (Yang 1997; Álvarez-Carretero et al. 2023). All the analyses were carried out using PAMLX (Xu and Yang 2013)

To identify selective regimes acting on genes located on the autosomes vs. the Z contig, we partitioned the genome of each species into two distinct segments: autosomal contigs and the Z contig. Subsequently, we applied the annotation procedure outlined in the “Genome Annotation” section to retrieve the CDS-transcript sequences predicted by Funannotate for each genomic partition. Next, we extracted single-copy orthologs shared across all 12 genomes using OrthoFinder v2.5.5.148 genes and 8,209 genes were retrieved from the Z contig and the autosomes, respectively. DNA sequences were translated into an amino acid sequences using Seqkit v2.1.0 (Shen et al. 2016) and amino acid sequences were aligned using Mafft. Finally, we used PAL2NAL for the correction of the nucleotide sequences using the corresponding aligned protein sequences (Suyama et al. 2006).

Then, we used the one-ratio model implemented in PAML to calculate the $dN/dS$ ratio (also called omega ω) on all the orthologs found on the Z contig and on the autosomal contigs. We measured the w ratio of those orthologs separately. Because large gaps in gene alignments can cause the $dN/dS$ to artificially increase when an ortholog is highly different in between species, we also measured the ω ratio of Morphos orthologs on gapless alignments. We used nonparametric Wilcoxon statistics to test whether the ω ratio measured on the orthologs found on the Z contig was higher than the ω ratio measured on the orthologs found on the autosomal contigs. To account for the unequal group sizes in the number of genes found on the autosomes and the Z contig, we performed a permutation test. We randomly drew 148 autosomal genes from the pool of 8,209 autosomal genes and compared their mean ω ratio to the mean ω ratio of the 148 Z genes. This procedure was repeated 1,000 times for both the gapless and the raw dataset.

To investigate whether the evolution of autosomes and the Z chromosome differs were linked to divergent evolution in the different ecological niches, we conducted the following analyses: first, we compared the evolutionary rates of genes located on the Z chromosome between species from the canopy and those from the understory, and we did the same for genes located on autosomes. Our goal was to identify the proportion of genes, either on autosomes or the Z chromosome, that show significant differences in evolutionary rates between the canopy and understory clades. To assess these differences, we used the Clade model in codeml ($Model=3$, $NSsites=2$), which employs the LRT to detect genes with significant rate variations between clades. We then compared whether the proportion of genes with significant differences between clades was higher for genes on the autosomes or the Z chromosome. This comparison was made using a Z-test for proportions, which accounts for the large difference in the number of genes between the two groups. Second, to investigate whether evolutionary rates differ between autosomes and the Z chromosome within each clade (canopy or understory) and to understand the potential role of these genomic regions in adaptation to the ecological niche, we analyzed species separately within each clade, allowing to include orthologous genes shared within clades but absent from the alternative. These clade-specific genes could play a significant role in adaptation to the difference niches. For each clade, we thus identified single-copy orthologous genes located on the autosomes and the Z chromosome and used the M0 model in codeml to calculate the dN/dS ratio for these gene sets. For the canopy clade, we identified 11,200 autosomal orthologs and 354 Z-linked orthologs that were common to the three species studied ($M.$  rhetenor, $M.$  hecuba, and $M.$  telemachus). For the eight species from the understory, we identified 9,299 autosomal single-copy orthologs and 189 Z-linked orthologs. With this approach, we were expecting to recover genes that may play specific roles in adaptation to particular ecological niches, which might have been overlooked in our previous analyses that included species from different niches. Finally, we used a nonparametric Wilcoxon permutation test to compare the dN/dS ratios between groups.

To test whether the differences in the $dN/dS$ ratio between autosomes and the Z chromosome stemmed mostly from divergence in specific species pairs, we estimated the substitution rates for genes in each genomic region using pairwise analyses with the YN00 program from PAML. Orthologous genes used in the genus-wide analysis were realigned by pair and used in the YN00 analysis with the method of Yang and Nielsen (2000). We conducted comparisons between pairs of closely related species within the same niche, as well as between species from distinct niches. Specifically, 11 paired comparisons were made: two pairs from the canopy, six pairs from the understory (paired close species living in sympatry or allopatry), the two species from the basal clade, and two comparisons between one canopy and one understory species. To assess differences in ω values between autosomes and the Z chromosome for each species pair, we used a nonparametric Wilcoxon test, followed by a permutation test.

Supplementary Material

Supplementary material is available at Genome Biology and Evolution online.

Funding

This work was funded by the European Union (ERC Consolidator grant OUTOFTHEBLUE, project number 101088089). Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.

Data Availability

The raw sequencing data and final genome assemblies generated in this study have been submitted to the NCBI BioProject database under accession numbers PRJNA1069011 (for raw fastq files) and PRJNA1063620 (for final assemblies of the new genomes sequenced here).