Diversification and secondary contact in the magpie-jays (Calocitta) throughout the pacific lowlands of Mesoamerica

2.1 Sampling and laboratory procedures

To assess patterns of genetic variation, we obtained samples of 54 individuals from 25 localities of C. colliei and all three currently recognized subspecies of C. formosa (Table 1, Figure 1). We used taxa from some of the major New World corvid clades: Aphelocoma coerulescens, Cyanocorax yncas, Cyanocitta cristata, we also included Psilorhinus morio, the putative sister taxon to Calocitta (Bonaccorso & Peterson, 2007; Bonaccorso et al., 2010; Fernando et al., 2017).

DNA sequences were obtained from samples of muscle, heart, or liver tissue stored in ornithological collections (Table 1; see Acknowledgements). DNA extraction was performed using a phenol–chloroform method (da Silva et al., 2012) or using the Qiagen DNeasyTM Blood and Tissue Kit (Qiagen Inc.) following manufacturer's protocol. We amplified two mitochondrial DNA loci (mtDNA): the mitochondrial encoded NADH dehydrogenase 2 (ND2), using the primers H6313 and L5219 or L5216 (Sorenson et al., 2004) and mitochondrially encoded cytochrome c oxidase I (COI), using primers COIBirdF1-COIBirdR2 (Hebert et al., 2004). We also amplified two nuclear DNA (nuDNA) loci: transforming growth factor beta 2 intron 5 (TGFB2), using the primers TGFB2.5F and TGFB2.6R (Sorenson et al., 2004), and glyceraldehyde-3-phosphate dehydrogenase intron 11 (GAPDH), using the primers G3P13b and G3P14b (Fjeldså et al., 2003). Primer sequences are included in Table S1.

The mtDNA and nuDNA loci were amplified by PCRs containing PCR buffer 1X, 1.5 mM MgCl₂, 0.2 mM of dNTP, 0.4 to 0.5 µM of ePach primer, 1–2 µl of DNA polymerase, and 2 µl of the DNA extract. All PCR procedures included negative controls to rule out contamination, and products were visualized on agarose gels stained with ethidium bromide. Sequencing was performed in two different laboratories, depending on the original of the tissue samples. PCR products of tissues housed at the MZFC were sequenced at the High Throughput Genomics Center of the University of Washington (USA). PCR products from tissues from other collections were prepared for sequencing using BigDye^TM, v3.1 terminator chemistry (Qiagen Inc.), using the same primers as in the PCR, purified with ethanol precipitation, and sequenced using an ABI 3730XL automated sequencer at the Biodiversity Institute of the University of Kansas (USA). The sequences were edited and aligned manually using Sequencher 4.8 (GeneCodes, An Arbor, MI). The nuDNA chromatograms were inspected for double peaks, which were coded using standard IUPAC codes for ambiguities. The allelic phase of the nuclear loci was determined using PHASE (Stephens & Donnelly, 2003; Stephens et al., 2001), a Bayesian method based on coalescence with default parameters as implemented in DnaSP v5 (Librado & Rozas, 2009). For all subsequent analyses, we used only individuals for which all four study loci could be successfully sequenced.

3.1 Phylogenetic analyses

We obtained mtDNA sequences for 38 samples of C. formosa and 14 of C. colliei. These sequences varied in length for the ND2 gene; 22 sequences had 646 base pairs (bp) and 30 sequences had 1041 bp. The COI gene sequences had 494 bp. We concatenated the mtDNA genes into a single alignment of 1535 bp analyses. For the GAPDH gene, we obtained 38 sequences of C. formosa and 14 of C. colliei, with a length of 305 bp; for TGFB2, we obtained 35 sequences for C. formosa and eight sequences for C. colliei, with a length of 432 bp. All sequences are deposited in GenBank (Table 1).

The Bayesian phylogenetic tree showed two well-supported major clades (Figure 2). One clade included all the C. formosa individuals, and the other included all C. colliei individuals. The C. formosa clade was further divided into three well-supported lineages, indicating clear phylogeographic structure (Figure 2), although the relationships among them were unclear due to low support values. The first sub-clade grouped samples from the southern Mexican Pacific coast including Oaxaca and the Isthmus of Tehuantepec regions (OAX). A second sub-clade included samples from the Mexican Central Pacific Coast (MCP), mainly from the state of Guerrero, but also including samples from Jalisco, Michoacán, and Oaxaca. The third sub-clade grouped the samples from southeastern Mexico (Chiapas) and the Pacific coast of Central America (Nicaragua and Salvador; MCA) (Figure 2). Interestingly, although we recovered three highly supported lineages, the distribution of the samples in these lineages did not correspond to the delimitation of the three currently recognized subspecies.

The mtDNA haplotype network showed 45 mutations between C. colliei and the OAX population of C. formosa. There were another 24 mutations between the haplogroups conformed by the OAX and MCA populations of C. formosa, and 29 mutations between the MCA and MCP populations (Figure 3d), indicating a marked divergence among the lineages within C. formosa. Using the nuDNA markers, GAPDH showed seven alleles, with a one private allele for C. colliei, three private alleles in the OAX and one in the MCA populations of C. formosa. The MCP population of C. formosa was monomorphic for the most common allele at this locus (Figure 3e).

3.2 Genetic diversity and structure

Patterns of genetic diversity for mtDNA and nuDNA in each of the main lineages are summarized in Table 2. For both datasets, MCA presented the highest number of haplotypes and alleles, but this population did not exhibit high levels of nucleotide diversity, which varied according to the locus analyzed. For the mtDNA, OAX had the highest nucleotide diversity (π = 0.0017) and MCP had the lowest values (π = 0.0004) and low haplotype diversity. In the case of nuDNA, C. colliei also had high values of nucleotide diversity (π = 0.001 and π = 0.0017 for GADPH and TGFB2, respectively.

The AMOVA revealed that 39.2% of the variation was in the two major groups defined for this analysis (C. formosa and C. colliei) and the three subgroups corresponding to the clades within C. formosa that were used in the analysis (Table 3). Most of the variance (56.45%) occurred within the three subclades of C. formosa. The F_ST values for the mtDNA dataset were higher than 0.91, indicating remarkably strong genetic structure among all of the populations of Mesoamerican Calocitta (Table 4). Genetic differentiation distances (D_xy) among C. colliei and the lineages within C. formosa were about 4.3% (range: 4.1– 4.7%), while among the subclades of C. formosa the average genetic differentiation was about 2.9% (range 2.5–3.5%; Table 4).

Based on the species tree (Figure S1), gene flow estimates using the IM model showed a dynamic demographic history within Calocitta (Figures 4 and 5). Based on the peak HPD values of N_e, the initial divergence in Calocitta occurred during the Middle Miocene (HPD 8.09 mya) leading to a population increase in NWM, while the rest of the lineages, currently grouped within C. formosa, decreased in population size. These three lineages then increased in N_e toward the Early Pliocene (HPD 4.3 mya), during which MCP diverged from OAX and MCA. The most recent divergence, the split between OAX and MCA, occurred during the Late Pliocene (HPD 2.79 mya). In general, our estimated migration rates showed non-significant values slightly above zero (0.03–0.08, i.e., less than one migrant every fourth generation) for geographically adjacent groups (Figure 5). Gene flow was highly limited (i.e., less than one migrant every fourth generation) in a north to south direction (from NWM to MCP to OAX to MCA), but not in the opposite direction (Figure 5). In addition, we detected limited ancestral gene flow from NWM to the ancestor of all of the C. formosa lineages (1.34; at least one migrant every generation), as well as from MCP to the ancestor of OAX and MCA (0.34), but no gene flow in the opposite direction (Figure 5).

3.3 Population clustering

Population clustering analyses from the mtDNA and nuDNA suggested an optimal value of K = 4. These four genetic clusters are geographically well-delimited (Figure 6) and correspond to the same clades recovered by the phylogenetic analyses.

4.1 Divergence in Calocitta

Recent studies on different vertebrate groups in the dry forests of western Mexico (Arbeláez-Cortés et al., 2014; Arbeláez-Cortés & Navarro-Sigüenza, 2013; Arcangeli et al., 2018; Montaño-Rendón et al., 2015; Reyes-Velasco et al., 2013; Zarza et al., 2008) have revealed unexpected geographic structure in spite of current habitat continuity and the apparent absence of contemporary geographic or ecological barriers in Mesoamerica since at least the Holocene (Castillo-Chora et al., 2021). This pattern suggests that past climate and geological changes have deeply impacted the biotic history of this tropical lowland region. It has been proposed that these changes were most favorable during the Middle Miocene, which allowed for the development, coalescence, and radiation of seasonally drier communities and lineages (Graham, 2011).

This dynamic history is also reflected in the evolutionary history of Calocitta Magpie-jays. The clear phylogeographic structure in our analyses showed that the two currently recognized species (C. colliei and C. formosa) grouped into well-supported and reciprocally monophyletic clades. According to the estimated divergence dates, this initial split may have occurred during the Middle Miocene, coinciding with the genesis of the Trans-Mexican Volcanic Belt and a period of high vulcanism (from 5 to 3 mya) near the border between Jalisco and Nayarit (Frey et al., 2007). Furthermore, strong phylogeographic structure was recovered within C. formosa, in which three well-supported and reciprocally monophyletic groups are also apparent (Figure 2). The estimated divergence dates for these splits (Figure 5) are consistent with the transition between the Late Pliocene and the onset of the Pleistocene, suggesting a significant role of climatic change. Divergence in these three clades may have been maintained through periods in which favorable environmental conditions were highly localized in isolated climatic stable areas (Castillo-Chora et al., 2021).

In terms of demographic history, after the first split, the population in NWM experienced an increase in the effective population size, while the ancestor of the three groups within C. formosa apparently decreased in effective population size, probably due to a reduction in the dry forest area throughout western Mesoamerica during moister periods. Later climatic changes promoted the expansion and reconnection of previously isolated forest patches, allowing populations to established contact zones and gene flow, at least between adjacent populations (Castillo-Chora et al., 2021). However, several of our results support the idea of genetic divergence. Our results using the IM model suggest that gene flow is highly limited and not significant (Figure 4), with less than one individual every fourth generation in modern lineages. Furthermore, the large F_ST and D_xy values for all Calocitta lineages and the star-like haplotype network (Figure 3) suggest population expansion. Further studies incorporating data from more nuclear loci would provide additional support for our conclusions.

Our analyses suggest that the current distribution of the four lineages detected in the Calocitta magpie-jays has resulted from secondary contact, allowing for recent but highly limited and non-significant gene flow among formerly separated populations. Both intra- and interspecific gene flow and hybridization (sensu Moore, 1995; Phillips, 1986) have been reported as locally common in Calocitta (dos Anjos et al., 2009; Pitelka et al., 1956). However, observations by Phillips (1986) of phenotypic differences in individuals of flocks including both C. colliei and C. formosa in western Jalisco led the author to suggest that hybridization may not be as common as previously thought, and that phenotypically intermediate individuals may rather represent individual variation or past interbreeding, and that the reported variation in plumage patterns may be restricted to young individuals that do not yet express their definitive adult plumage. Our results also seem to support this, since gene flow is highly limited in recent times but was higher at the time of the initial split. Therefore, areas of present sympatry among the different lineages seem likely to be the product of secondary contact. Further studies using novel sequencing and analysis techniques are ongoing and will be published elsewhere (Castillo-Chora et al., unpublished data).

The strong genetic differentiation suggesting population isolation was less evident from the analysis of the nuclear dataset, which was expected given the different coalescent times for nuDNA versus mtDNA because of their population sizes (Hung et al., 2016; Zink & Barrowclough, 2008), Nonetheless, the geographic distribution of the nuclear alleles showed several similarities to the distributions of the mtDNA clades. When the nuDNA is grouped according to the previously identified mtDNA clades, the fact that the distribution of only a few alleles in these nuDNA groups shows geographically sorted private haplotypes suggests gene flow (Figure 2). The nuDNA allele frequencies among populations suggest incomplete lineage sorting, probably due to the slow mutational rates, but probably also to processes such as migration and introgression between the two taxa. However, IM coalescent analyses suggest that gene flow is highly limited in modern lineages but was somewhat higher in past lineages (NWM to OAX-MCA-MCP; Figure 5), suggesting a lag on the path to reciprocal monophyly (Omland et al., 2006). In addition, we detected differences in the divergence dates in this study compared with those estimated by Castillo-Chora et al. (2021), which may be due to the ability of IMa2 (Hey, 2010) to accommodate gene flow into a single coalescent analysis.

Factors may account for maintenance of the observed genetic divergence in Calocitta: (1) ecological barriers and (2) behavioral traits and natural history. No clear modern geographic barriers are associated with the distribution of the lineages in Calocitta. Therefore, ecological barriers may have acted first as promoters of diversification, and later contributed to maintaining the differentiation among adjacent lineages in more recent times (Castillo-Chora et al., 2021). Thus, after habitat modification due to orogenic and volcanic activity, diversification between colliei and formosa may have evolved due to habitat differences between them. These two species can form mixed flocks in western Jalisco, where oak forests mix with tropical dry forest vegetation (Phillips, 1986), suggesting secondary contact. Populations of OAX would have been prevented from expanding toward the range of the MCA by the humid lowland rainforests of southern Chiapas (the Soconusco region), as well as by a low-altitude mountain range near Rizo de Oro, which may have limited dispersal to the Central Depression in Chiapas, as suggested by Campbell (1999). In the case of MCP and OAX, no contemporary geographic barrier is apparent, placing a stronger emphasis on the significance of ecological barriers for genetic differentiation in these populations.

Sampling individuals from presumed suture zones would be highly informative, as they could reveal novel population dynamics (e.g., Harrison & Larson, 2016) between the different genetic groups in our study group. In addition, sampling is also needed to cover geographic gaps within some of the recovered clades, such as in Nayarit (for C. colliei) and from southern Chiapas to western El Salvador (for C. [f.] pompata). More complete sampling in these areas may reveal further details on the demographic history, as well as on the significance of past interbreeding that resulted in phenotypic intermediates or hybridization. On the other hand, behavioral differences and natural history may have also played a role in differentiation (Barber & Klicka, 2010); however, studies on these aspects throughout the distribution of Calocitta are virtually completely lacking and are presently limited to populations that are presumably outside of the secondary contact zones (e.g., Innes & Johnston, 1996; Langen, 1996b).

4.2 Taxonomic implications

Our research revealed unexpected levels of phylogenetic structure, historical demography, and genetic distinctiveness among the lineages found within Calocitta which may have taxonomic implications. High levels of genetic differentiation and the strongly reduced gene flow detected among the three C. formosa lineages are comparable to the levels of genetic divergence between the two currently recognized species in Calocitta (Table 4) and are also consistent with deep divergence times. The unmatched geographic distribution of the recognized subspecies in C. formosa as compared to the genetic lineages found in our study is especially complex in coastal Oaxaca and the Isthmus of Tehuantepec, where up to three recognized subspecies (plus one unrecognized taxon) meet, and for which our analyses seem to provide a clearer picture. Following our results, the MCP lineage may be assigned to C. [f.] formosa, which ranges from Jalisco to Guerrero and the Balsas Basin in southern Mexico; the MCA lineage includes two currently accepted subspecies: C. [f.] pompata, distributed from southern and central (the Central Depression) Chiapas in Mexico south to northwestern Costa Rica, and C. [f.] azurea from the SE Oaxaca to western Guatemala. This lineage should be recognized as azurea, which has priority over pompata since it was assigned to birds described from SE Chiapas in 1897, while the second was assigned to birds described from Costa Rica in 1914. This situation leaves the OAX lineage without a name, because although the unrecognized name impudens may be available, the type locality for birds under this name is Conchagua, El Salvador, which is within the range of azurea, therefore invalidating the use of impudens. The formal description of this putative new taxon from Oaxaca will be published elsewhere (Sánchez-González et al., unpublished data).

4.3 Evolutionary differentiation in the dry forests of Mesoamerica

Our results add to a growing body of evidence suggesting that the dry forests of the Mexican Pacific slope have had a highly dynamic history (Graham, 2011; Prieto-Torres et al., 2018, 2019) that has influenced the diversification and differentiation of a large number of plant taxa (e.g., De-Nova et al., 2012), and animal taxa, including frogs (Zaldı́var-Riverón et al., 2004), reptiles (Bryson et al., 2011; Suárez-Atilano et al., 2014; Zarza et al., 2008), birds (Arbeláez Cortés et al., 2014; Vázquez-Miranda et al., 2009), and mammals (Arcangeli et al., 2018), all of which exhibit clearly structured monophyletic lineages. These patterns highlight a very dynamic region, which was first isolated by tectonic processes, such as the rise of the Sierra Madre Occidental and the Trans-Mexican Volcanic Belt (Becerra, 2005; Sosa et al., 2018), allowing for the diversification of several plant and animal taxa. These patterns, however, differ widely in the age of diversification for different groups, which suggest that taxa in the region have also experienced further diversification events due to climatic changes since at least the Miocene (De-Nova et al., 2012; Navarro-Sigüenza et al., 2017). These climatic shifts seem to be the most likely cause of divergence and differentiation in other dry forest birds in western Mexico (see Arbeláez-Cortés et al., 2014; Castillo-Chora et al., 2021; Montaño-Rendón et al., 2015; Navarro-Sigüenza et al., 2017) as well as in the Calocitta magpie-jays explored here.