An examination of species limits in the Aulacorhynchus “prasinus” toucanet complex (Aves: Ramphastidae)

Taxonomic history

In the second edition of his monograph on the Ramphastidae, Gould (1854) recognized 10 species in the genus Aulacorhamphus (now Aulacorhynchus by priority), but several taxa remained undescribed at that time. Between them, Salvin & Godman (1896) and Brabourne & Chubb (1912) recognized 15 species in the genus. Ridgway (1914) followed this treatment. Cory (1919) considered one of these species (A. erythrognathus) to be only a subspecies, and he treated the genus as having 14 species. Although a new species was described in 1933 (A. huallagae, Carriker Jr, 1933), Peters (1948) reduced many of the formerly recognized species to subspecific status and presented the species-level diversity of the genus as just seven taxa. Haffer (1974) followed Peters (1948), except that he reduced one species (A. calorhynchus) to subspecies status (after Schwartz, 1972), leaving just six. Sibley & Monroe Jr (1990) followed this treatment, but they presented a further reduction of the apparent diversity by mentioning only two subspecific groups below the species level. Although more comprehensive subspecific inclusion and discussion were given by Short & Horne (2001) and Short & Horne (2002), six full species were retained. The treatment of this genus since 1891 is summarized in Table 1.

The massive lumping of Peters (1948) proceeded with neither the presentation of data nor with discussion. Careful study of one taxon, A. calorhynchus, by Schwartz (1972) supported the single species-level change that was made between Peters (1948) and Haffer (1974), and Haffer’s (1974) important work has been cited to support maintaining a broad A. prasinus (sensu lato; AOU, 1998). Haffer (1974), Short & Horne (2001) and Short & Horne (2002) used evidence of hybridization and intergradation to support their conclusions that the whole A. “prasinus” complex represented one biological species with many subspecies (14 and 13 subspecies, respectively); at the species level across the genus Short & Horne (2001), Short & Horne (2002) treatment reflected that of Dickinson & Remsen Jr (2013) in Table 1 except that they considered A. whitelianus a subspecies of A. derbianus, as Peters (1948) did. Presumably similar reasoning was behind Peters (1948). But subspecies, even those distinctive enough to have been considered full species for a century, can get lost in the shuffle. For example, oversimplification of subspecific variation led Sibley & Monroe Jr (1990) and the American Ornithologists’ Union (AOU, 1983; AOU, 1998) to completely omit mention of the very distinct form A. “prasinus” wagleri from southwestern Mexico. The AOU (1983; 1998) considered Middle American diversity in the genus as being just two subspecific groups of a single species, A. prasinus.

Renewed interest in this complex (Navarro et al., 2001; Puebla-Olivares et al., 2008; Bonaccorso et al., 2011; Del Hoyo & Collar, 2014) is beginning to rectify the absence of data, but the ensuing taxonomic changes recommended have either been based on a different species concept (Bonaccorso et al., 2011) or have inadequately considered the hybridization and intergradation (e.g., Navarro et al., 2001; Puebla-Olivares et al., 2008; Del Hoyo & Collar, 2014) that have been integral to supporting the “post-Peters” taxonomy. These latter works have recommended elevation of numerous A. “prasinus” taxa to species status (Table 1), but they did not address the reasons for lumping in the first place: evidence of hybridization. There has also been heavy reliance on a single molecular marker (mtDNA) for species delimitation in A. “prasinus” (Puebla-Olivares et al., 2008; Bonaccorso et al., 2011). This is problematic because mtDNA can be misleading about species limits and relationships between populations due to gene-tree/species-tree mismatches and because genetic distance is not a reliable indicator of species limits (Avise & Wollenberg, 1997; Irwin, 2002; Funk & Omland, 2003; Degnan & Rosenberg, 2006; Cheviron & Brumfield, 2009; Galtier et al., 2009; Ribeiro, Lloyd & Bowie, 2011; Toews & Brelsford, 2012; Pavlova et al., 2013; Peters et al., 2014; Dolman & Joseph, 2015; Morales et al., 2015). Thus, species limits in the group remain uncertain (Table 1). Most disagreement has been in the A. “prasinus” complex, and it is on this group that I focus.

The A. “prasinus”complex.—As currently treated (e.g., Table 1, Dickinson & Remsen Jr, 2013; Del Hoyo & Collar, 2014), A. “prasinus” either is a widely distributed and highly variable biological species or it comprises multiple species (Fig. 1). So far as is known, Aulacorhynchus toucanets are nonmigratory (AOU, 1998), but as O’Neill & Gardner (1974) and Navarro et al. (2001) noted, members of the genus can wander widely during the nonbreeding season. Short & Horne (2001) considered Central American forms from Mexico to Panama to be partially migratory, with downslope movements to lowlands also occurring during the nonbreeding season (but less commonly) throughout the A. “prasinus” range. The sexes are alike by plumage (sexually monochromatic), but sexual size dimorphism is apparent in all taxa examined.

In past work there has been too little discussion of the fact that different levels of differentiation occur among the subspecies of A. prasinus, sensu lato. All of the named forms do not represent equally differentiated populations; there are major subspecific groups of one or more described subspecies. Objectively determining what these groups are can be done by following taxonomic history, coupled as it is with a color-based clustering. Cory (1919) recognized eight species that were later lumped by Peters (1948) into A. prasinus (Table 1). It seems that four of these (A. wagleri, lautus, cyanolaemus, and dimidiatus) were not available for Cory (1919) to examine, however (he noted which taxa were in the Field Museum of Natural History at that time). Two of these taxa, A. lautus and A. dimidiatus, may have been included as full species through inertia. This was not uncommon: Ridgway (1914) considered that the genus had 15 species, but he was uncertain because he had only been able to examine seven of them. The subsequent rediscovery and examination of A. dimidiatus (O’Neill & Gardner, 1974) showed intergradation with A. “prasinus” atrogularis, and the Santa Marta isolate A. “prasinus” lautus is, by plumage, clearly a relatively minor derivative of the A. “p.” albivitta group. These considerations reduce the number of major, color-based subspecific groups in the A. “prasinus” complex to six (A. “p.” wagleri, prasinus, caeruleogularis, albivitta, cyanolaemus, and atrogularis).

The rather pronounced differences among these major groups are illustrated in Fig. 1, together with examples of some of the less-pronounced variation (though still between named subspecies) occurring within two of these groups. Del Hoyo & Collar (2014) presented an analysis of colors that reaffirms this approach, coming back to the same six color-based subspecific groups and treating them as full biological species. Diagnostic characters of these six groups are given in Table 2. It is these six groups that form the basis for my comparisons. They are the color-based groupings that have been recognized by students of the birds themselves.

The troubled taxonomic history of Aulacorhynchus (especially A. prasinus, sensu lato) reflects disagreement among revisors over the ability of plumage and bill colors and patterns to represent species limits. Members of the A. “prasinus” group in particular are effectively ecologically similar geographic replacements (Mallet, 2007), and where named taxa have been found to come together hybridization has been described (e.g., Haffer, 1974; Short & Horne, 2001). Given evidence of hybridization, it is unclear how the birds themselves perceive these differences. Is this group (A. “prasinus”sensu lato) really one in which head and bill colors and patterns are very plastic, resulting in species that include a high degree of color variation? Or are full species being overlooked? If bill and head color characteristics are plastic and not indicative of species limits, as many treatments since 1948 suggest, then strikingly color-based taxa, most of Cory’s (1919) species and the six of Del Hoyo & Collar (2014), would not likely show a great deal of morphometric distinctiveness. If, on the other hand, substantial, concordant morphological changes are also occurring, then perhaps the concept of ecologically similar geographic replacements, which seems broadly applicable to A. prasinus, sensu lato, is masking important group-specific evolutionary adaptations that go beyond the existing color changes, the latter of which likely reflect social selection. Such morphological differences might suggest adaptive changes among groups that would make immigrants and hybrids less fit (Price, 2008). My question then is simple: Are these color-based subspecific groups just “cookie-cutter” (i.e., morphologically nearly identical) toucanets bearing different throat and bill colorations, or are there also significant morphological changes occurring among them? If substantial morphometric changes occur concurrently with dramatic color changes, then species limits should probably be reconsidered, as several have suggested (Navarro et al., 2001; Puebla-Olivares et al., 2008; Bonaccorso et al., 2011; Del Hoyo & Collar, 2014).

Morphometrics alone are unlikely to be important components of species limits in forms like these where colors are obviously important, and this is not meant as a study of how morphology varies within the group independently of color-based clustering. Navarro et al. (2001) did a morphological analysis based on 17 allopatric groupings, and this is not meant to repeat those analyses. Geography alone can affect morphology (e.g., Bergmann’s rule; Mayr, 1963), and my analyses account for this. In this study I will (1) test for univariate differences between pairwise groups that are geographically closest to each other; (2) compare these groups in multivariate, principal component space (because univariate measures can be correlated with each other); and (3) visually examine specimens for evidence of hybridization because such evidence has been historically important in the taxonomy of the group.

Hybridization

Because members of the genus are known to wander rather widely during the nonbreeding season, the opportunity for gene flow does exist between these largely allopatrically breeding groups. Among the six major subgroups I examined there are theoretically five pairwise instances of possible gene flow between any two of the groups, particularly across some of the narrower zones of separation, A. “prasinus” prasinus-wagleri (Oaxaca, Mexico), A. “p.” caeruleogularis-albivitta (in W Colombia), A. “p.” albivitta-cyanolaemus (Ecuador), and A. “p.” cyanolaemus-atrogularis (Peru); the fifth, prasinus-caeruleogularis (in Nicaragua), is a larger distance, on the order of about 240 km. Note that closest approach distances are not accurate in Fig. 2, which is based on the specimens I examined; ranges given by other sources (e.g., Haffer, 1974; Hilty & Brown, 1986; Binford, 1989; Howell & Webb, 1995; Short & Horne, 2001; Ridgely & Greenfield, 2001; Restall, Rodner & Lentino, 2006; Schulenberg et al., 2007) include more records (including sight records) and are more accurate.

Specimens were examined carefully for phenotypic evidence (i.e., intermediate phenotypes in plumage and bill coloration) of hybridization between these major groups, but it was found to occur in just one of these pairwise comparisons: between A. “prasinus” cyanolaemus and A. “p.” atrogularis in Peru. Four specimens representing possible F₁ hybrids (due to intermediacy of characters) were found; one from La Libertad, Utcubamba (25 October 1979, D. Wiedenfeld, LSUMNS 92029; Fig. 3), and three from La Lejía in NE Peru (11 & 19 March, 16 April 1925, H. Watkins, AMNH 234533, 234532, and 234535). All four specimens show obvious intergradation between these two taxa, particularly in bill coloration (see Fig. 3 and Haffer, 1974, Fig. 16.8), and all four are males (these individuals were not included in the morphometric analyses). In addition, there are another five specimens that seem to show evidence of intergradation to a lesser degree, two females that are phenotypically mostly A. “p.” cyanolaemus (both H. Watkins: La Lejía, 19 March 1925, AMNH 234534; and Uscho, Dept. Amazonas, 3 October 1925, AMNH 234531), and three males that are phenotypically A. “p.” atrogularis but seem to have some A. “p.” cyanolaemus influence (e.g., primaries edged in russet). These latter three birds are from three localities: Divisoria, Cordillera Azul, Dept. Huanuco (17 August 1967, J. P. O’Neill, LSUMNS 62227), Huanhuachayo, Dept. Ayacucho (6 May 1971, J. P. O’Neill, LSUMNS 69410), and Abra Divisiona, Dept. Loreto (14 Aug 1977, J. W. Eley, LSUMNS 84550). It is of interest that this evidence of hybridization occurs between the subspecific pair with the fewest morphometric differences (Table 4) and close genetic affinity (Puebla-Olivares et al., 2008).

Morphometrics and hybridization in a genetic context

While divergent selection should produce phenotypic differences between species, this observation makes predictions neither in direction nor degree as far as morphometrics are concerned (i.e., it only predicts accumulating differences). Nevertheless, where differences occur between groups on a phylogeny, and the genetic distances involved, might be of further aid in inferring species limits—qualified, of course, by the many known ways in which mtDNA can be misleading about species limits (Avise & Wollenberg, 1997; Irwin, 2002; Funk & Omland, 2003; Degnan & Rosenberg, 2006; Cheviron & Brumfield, 2009; Galtier et al., 2009; Ribeiro, Lloyd & Bowie, 2011; Toews & Brelsford, 2012; Pavlova et al., 2013; Peters et al., 2014; Dolman & Joseph, 2015; Morales et al., 2015). The mtDNA topology of the phylogenetic relationships among the six major subspecific groups is given in Fig. 4 (after Puebla-Olivares et al., 2008). To examine the results of my study in relation to what is presently known about relationships among and genetic distances between the groups examined, I downloaded the mtDNA data of Puebla-Olivares et al. (2008) from GenBank, concatenated and aligned the sequences using Geneious (ver. 7.1; Kearse et al., 2012), and calculated genetic distances between the groups for which I made pairwise comparisons using MEGA (ver. 6; Tamura, Nei & Kumar, 2004; Tamura et al., 2013). Three of these groups are not monophyletic in their mtDNA (not uncommon; see Funk & Omland, 2003), but I treated the haplotypes of A. “p.” atrogularis that have introgressed into A. “p.” albivitta (see discussion below) as A. “p.” atrogularis for the calculation of genetic distances.

Phenotypic evidence of hybridization occurs only between the most closely related pair on this tree, but the presence of A. “p.” atrogularis mtDNA in birds that are phenotypically A. “p.” albivitta with no outward evidence of hybrid characteristics indicate that gene flow can occur between groups that are on average 4.2% divergent (Fig. 4, bottom clade). Contrasting the number of morphometric differences that have accumulated between the pairwise comparisons of major groups that might hybridize due to proximity (Table 4, excluding albivitta-atrogularis) with genetic distance reveals a positive correlation (linear regression, F = 6.07, P = 0.04, R² = 0.67; Fig. 5).

Hybridization

Fortunately, with respect to gene flow we do have larger sample sizes if we use diagnostic morphological attributes as a surrogate (i.e., the plumage and bill characteristics upon which the named forms have been based). Despite what is likely to be ample opportunity for gene flow through dispersal across the five zones of contact or “nearest approach”, evidence of hybridization among the major groups in the A. “prasinus” complex presently exists in just two cases: between A. “prasinus” atrogularis and cyanolaemus (phenotypic only, as detailed above), and in the genetic results of Puebla-Olivares et al. (2008), which showed two A. “p.” albivitta specimens from NE Ecuador with mtDNA more closely related to A. “p.” cyanolaemus/atrogularis. This latter case suggests that historical crossing may have occurred across this zone, although morphological evidence of this is not yet evident. Aulacorhynchus “prasinus” is uncommon in this region, and the ranges of A. “p.” albivitta and cyanolaemus are not known to come into contact (Ridgely & Greenfield, 2001; Restall, Rodner & Lentino, 2006). These two A. “p.” albivitta specimens also demonstrate that phenotypic evidence of hybridization, which does occur and has been useful in past evaluations in A. “prasinus”, can be absent despite gene flow (but even nuclear genomic evidence of hybridization can disappear over a few generations of backcrossing; Lavretsky et al., 2016).

From this dataset, therefore, we know that hybridization in toucanets can be visible and invisible, the latter probably after repeated backcrossings to one parent taxon. We might, however, consider the visible hybrids to be roughly indicative of a hybridization rate. Current evidence thus suggests that hybridization between A. “p.” albivitta and cyanolaemus is rare. In the case of A. “p.” cyanolaemus and atrogularis, however, my results expand the scope of hybridization recognized, both in number of possible F₁ specimens (those most intermediate in characters) and in the broader distribution of specimens likely exhibiting intergradation (and note that it is bi-directional). The fact that all four putative F₁ specimens are males suggests the possibility that the two forms are sufficiently divergent that genetic incompatibilities are preventing viability of the heterogametic sex, which is the female in birds (Haldane’s rule; Price, 2008). However, this is not a significant departure from the sex ratio of the rest of the A. “prasinus” sample I examined (P = 0.13, Fisher’s exact test), and the two forms are not very divergent genetically in mtDNA (<1%; Fig. 4 and Puebla-Olivares et al., 2008). The matrilineal passage of A. “p.” cyanolaemus/atrogularis mtDNA into A. “p.” albivitta in NE Ecuador (Puebla-Olivares et al., 2008) also suggests that Haldane’s rule is not operating among major subspecific groups in South America. In the case of A. “p.” cyanolaemus and A. “p.” atrogularis, hybrids and intergrades represented a substantial percentage of the specimens I was able to examine of these taxa (as many as 10 specimens, with remaining sample sizes of 19 each of cyanolaemus and atrogularis). Given the small sample, it remains unclear whether this reflects the true incidence of hybrids between these taxa, but given present evidence it is substantial.

Species limits

Despite considerable combined evidence from coloration, morphometrics, and mtDNA data, comprehensive and accurate species limits for this group remain elusive, no matter what species concept one chooses to use. Using the biological species concept, I suggest that consideration of all of the available evidence indicates that we should recognize five species in the A. “prasinus” complex (A. wagleri, prasinus, caeruleogularis, albivitta, and atrogularis), each with any associated named subspecies (Appendix). Further study could raise this number (e.g., by splitting A. atrogularis cyanolaemus from atrogularis again). Under a phylogenetic species concept (PSC), one could probably raise every allopatric population in A. “prasinus” to the species level, resulting in at least 12 taxa using morphology alone. So how many species of toucanets are there in the A. “prasinus” complex?

Historically, evidence of hybridization has often driven taxonomic decisions under the biological species concept (BSC), as has apparently occurred in this case. My interpretation of the taxonomic history of Aulacorhynchus “prasinus” is that evidence of hybridization and intergradation between named forms (among forms of the A. “p.” albivitta group and between A. “p.” albivitta and cyanolaemus) caused all named forms to be lumped together as one species (Peters, 1948). But hybridization is not uncommon between full species (Grant & Grant, 1992), and avian taxonomists have used a working definition of the BSC that recognizes this (e.g., Short, 1969; Johnson, Remsen Jr & Cicero, 1999; Winker et al., 2007). Gene flow, reproductive isolating mechanisms, and lineage reticulation remain fundamentally important evolutionary phenomena affecting species diversity and the process of evolutionary divergence, and thus they require consideration. Effective lineage reticulation requires that hybrid offspring have equal or greater fitness than offspring of pure parental forms. Also, gene flow must occur frequently enough to overcome the differentiating selective factors likely to be operating on largely allopatric populations (and this relationship is nonlinear; see Winker, 2010 for discussion). The more differences there are between populations in morphology, the more differences there are likely to be in selective factors operating on these populations and the more difficult effective gene flow is likely to be between populations; at larger scales this results in the general correlation between morphological difference and reproductive isolation (Mayr, 1963; Price, 2008).

Classic systematics and taxonomy (Mayr, 1969; Mayr & Ashlock, 1991) uses a comparative approach to determine species limits among allopatric taxa, examining what occurs at contact zones (if available) and/or what occurs in similar cases in closely related taxa. In previous work on A. “prasinus” taxonomy I do not think enough credit has been given to the dispersal abilities of these birds. And yet despite that ability there is a lack of evidence for gene flow (using phenotype as an indicator) between five of the major subspecific groups (A. prasinus-wagleri, prasinus-caeruleogularis, caeruleogularis-albivitta, albivitta-cyanolaemus). For example, in south-central Mexico (Oaxaca), A. prasinus and A. wagleri breed within about 100 km of each other, a distance that A. prasinus individuals appear to move routinely away from their breeding areas, e.g., at the base of the Yucatan Peninsula (e.g., Land, 1970; Jones, 2003), which does not seem unusual for an arboreal frugivore (see also discussions in O’Neill & Gardner, 1974, and Navarro et al., 2001). Hybridization per se is not sufficient evidence for conspecificity, and in this group I find the lack of hybrids at most zones of potential crossing of major subspecific groups to be more compelling in the determination of species limits than its clear and seemingly routine presence at one—particularly in light of the repeated evidence of varying suites of morphological characters changing abruptly across these zones. However, I do consider that the apparent frequency of hybridization between A. atrogularis cyanolaemus and A.a. atrogularis warrants a conservative approach to their separation at the species level, and thus I do not recommend doing so without more evidence. In short, morphologically there is no evidence for hybridization between five of the major subspecific groups, despite likely opportunity, especially in northern Middle America. This is coupled with pronounced morphometric differences between these groups, suggesting group-specific ecological adaptation in addition to whatever social selection factors have likely caused the rather dramatic head and bill color differences (Fig. 1 and Table 2).

The populational processes of lineage divergence and the hierarchical nature of differentiation that accrues as gene flow decreases and divergent selection produces increasingly different phenotypes (anagenesis) have produced gradations of differentiation in the genus Aulacorhynchus. This is seen from population-level differences of little significance (e.g., among some questionably recognizable subspecies; see Appendix), to diagnosable isolated populations within biological species, to full biological species, to a recognizable subgeneric group (members of A. “prasinus” sensu lato). Genetic data also support this subgeneric group (Puebla-Olivares et al., 2008), and the name Ramphoxanthus, which is particularly fitting (i.e., yellow-bills), is available for it (Bonaparte, 1854). This group diverged from other members of the genus during the Miocene (approximately 6–9 million years ago; (Bonaccorso & Guayasamin, 2013). Within the five groups comprising A. “prasinus” that I consider full biological species (A. wagleri, prasinus, caeruleogularis, albivitta, and atrogularis) there are a number of diagnosable subspecific taxa that are clearly evolutionarily significant units, and some, being 100% diagnosable, could be called phylogenetic species (e.g., A. p. warneri, A. c. cognatus, A. albivitta lautus). These, however, do not represent major phenotypic differences, and I consider this continued lumping to be warranted given present evidence, which includes hybridization and intergradation between other subspecies with similar degrees of differentiation within the major subspecific A. “prasinus” groups (e.g., A. p. prasinus-virescens, A. albivitta phaeolaemus-griseigularis; Short & Horne, 2001 and pers. obs.).

Voice has not diverged among these five groups as much as it has in other species in the genus (Schwartz, 1972; Short & Horne, 2001). Indeed, Short & Horne (2001:327) related that “Calls of forms in Peru, Venezuela, Panama, Costa Rica and Mexico are much alike...” However, more work is warranted in this area. For example, A. wagleri has a slower pace to its vocalizations than A. prasinus (from www.xeno-canto.org, 4 A. wagleri average 1.85 calls/sec while 6 A. prasinus average 2.13 calls/sec; XC 274798, 219401-2, and 177515 vs. XC 96724, 256673, 256311, 233097, and 138132-3). Other differences may be apparent with increased sample sizes.

Considering my suggested taxonomy (Appendix) in relation to the mtDNA tree of Puebla-Olivares et al. (2008), there are two paraphyletic species (Fig. 4). First, A. albivitta is paraphyletic with respect to A. atrogularis-cyanolaemus. Second, my treatment of A. caeruleogularis is paraphyletic with respect to A. prasinus and A. wagleri because it includes A. “p.” cognatus; that has been the norm since its description, only Navarro et al. (2001) and Puebla-Olivares et al. (2008) have treated it as a full species thus far. Paraphyletic species are not uncommon (Funk & Omland, 2003; Oyler-McCance, St. John & Quinn, 2010), but there is clearly work remaining to be done on species limits in this complex, especially in South America. For example, the distributions of A. atrogularis cyanolaemus, A. a. atrogularis, and the hybrid zone between them warrant further study, as does the apparently rare instance of crossing between A. albivitta and A. atrogularis cyanolaemus in Ecuador (mtDNA evidence of (Puebla-Olivares et al., 2008). Also, the relationship between A. c. caeruleogularis and A. c. cognatus in Panama bears further investigation; they are phenotypically relatively similar (Short & Horne, 2001:325 also noted their close resemblance) in contrast to, e.g., A.a. atrogularis and cyanolaemus. Larger sample sizes, more loci, coverage of hybrid zones, and continued recognition that there are relatively major and minor phenotypic variants among these named taxa will be needed to finally and fully resolve species limits in this group.