The Importance of Recent Ice Ages in Speciation: A Failed Paradigm

See allHide authors and affiliations

Science  12 Sep 1997:
Vol. 277, Issue 5332, pp. 1666-1669
DOI: 10.1126/science.277.5332.1666


Late Pleistocene glaciations have been ascribed a dominant role in sculpting present-day diversity and distributions of North American vertebrates. Molecular comparisons of recently diverged sister species now permit a test of this assertion. The Late Pleistocene Origins model predicts a mitochondrial DNA divergence value of less than 0.5 percent for avian sister species of Late Pleistocene origin. Instead, the average mitochondrial DNA sequence divergence for 35 such songbird species pairs is 5.1 percent, which exceeds the predicted value by a factor of 10. Molecular data suggest a relatively protracted history of speciation events among North American songbirds over the past 5 million years.

Evidence from molecular systematics has provided fresh insights into several long-standing controversies regarding avian evolution, including the origin of birds (1), the origin and distribution of modern-day avian orders (2), and the survival of avian lineages across the Cretaceous-Tertiary boundary (3). The timing of the origin of modern bird species remains unclear. Many authors (4,5) postulate a recent origin for North American songbird species (Order Passeriformes) and species complexes.

These origins are typically associated with Late Pleistocene glacial cycles (6-8) involving (i) fragmentation of a widespread ancestral species into refugia during periods of glacial advance and (ii) subsequent genetic divergence among small isolated populations, followed by (iii) range expansion during interglacials. Typically, one [beginning circa (ca.) 100,000 years ago] or two (ca. 250,000 years ago) such cycles are invoked. This model, here termed the Late Pleistocene Origins (LPO) model, is widely accepted today [see (9) for example].

If mitochondrial DNA (mtDNA) evolves at a clocklike (10) rate of 2% per million years (My) (11), then a plot of divergence values for sister species of Late Pleistocene origin should be strongly left-skewed and leptokurtic. Invoking either one or two glacial cycles, mtDNA sequences of species pairs should differ on average by 0.2 to 0.5%. The LPO model also predicts that phylogenetic analyses of mtDNA sequences (haplotypes) from recently separated species will result in trees that do not reflect recognized species (taxonomic) limits. That is, haplotypes in one species can be more closely related to haplotypes in the sister species than to those in their own (12). We tested the LPO paradigm directly by analyzing most of the songbird taxa used to construct it. Table1 depicts all North American songbird species (13) for which Late Pleistocene origins have been postulated (4, 5) and for which comparative mtDNA data (14) are now available (16-19). These comparisons represent the best estimates of divergence times among what are presumed to be the most recently evolved songbird species. The plot of observed divergence values (Fig.1) is neither left-skewed nor leptokurtic. The average percent divergence for the 35 taxon pairs is 5.1% (SD ± 3.0), which suggests an average Late Pliocene divergence time of 2.45 million years ago (Ma) (20). This estimated divergence time and a divergence time consistent with a Late Pleistocene origin differ by an order of magnitude. Alternatively, if the molecular clock is improperly calibrated and our average percent divergence does correctly reflect genetic change occurring since the beginning of the last glacial advance (assume 5.1% change per ca. 100,000 years), two taxa isolated for 1 My would differ by 50%. Multiple substitutions at the same nucleotide position and eventual DNA saturation make a figure this high improbable, and no such saturation effects were detected in the taxon pairs examined.

Figure 1

(left). Plot of estimated percent sequence divergence for haplotypes representing 35 pairs of North American songbird taxa (see Table 1). This distribution is not significantly non-normal (P > 0.1). Cross-hatched portions of bars represent Great Plains species pairs.

Table 1

Estimates of mtDNA percent sequence divergence and estimated ages for pairs of North American songbird species (13) that have been postulated to have differentiated during most recent Pleistocene glacial advances (4, 5). Standard errors are shown when available. ND2, NADH subunit 2; ND6, NADH subunit 6.

View this table:

An additional test of the LPO model derives from comparing the 35 “Late Pleistocene” species pairs with 13 pairs of species not specifically theorized to have evolved in the Late Pleistocene (21). The mean mtDNA pairwise distance for these presumably older songbird pairs is 5.2% (SD ± 2.3), a value not significantly different from that of the 35 “Late Pleistocene” pairs (Mann-Whitney U test, P = 0.48). Furthermore, the distributions of mtDNA distances for the two groups are not different (Kolmogorov-Smirnov two-group test; χ2 = 1.4, df = 2, P = 0.50). These results contradict the expectations of the LPO model. Overall, these data reflect a protracted history of speciation throughout the Pleistocene and Pliocene. Inspection of Fig.1, however, reveals evidence of regional and temporal pulses of diversification. For example, the North American Great Plains is a region where many morphologically and ecologically similar pairs of species that “seemingly differentiated in allopatry during the Pleistocene” (22) are in secondary contact. The mean sequence divergence among these species pairs is 5.3% (SD ± 1.8), and only one value is less than 4%. The clustering of several values (Fig. 1) near the Pliocene-Pleistocene boundary reveals a probable regional pulse of speciation, but at a time much earlier than is generally assumed.

We know of only one songbird species pair whose genetic characteristics are wholly consistent with a Late Pleistocene origin. The Timberline Sparrow [Spizella (breweri)taverneri (23)] is similar genetically [<0.1% sequence divergence (24)] to its sister taxon, the Brewer's Sparrow [Spizella (breweri)breweri]. Over a 1450–base pair (bp) span of mtDNA, 12breweri specimens differed from 7 of taverneri at a single fixed nucleotide position (24), yielding an estimated divergence time of 35,000 years ago. Extremely low divergence values, however, are not the only expected genetic signature of recent speciation. The lack of reciprocal monophyly (12) reflected in the topology of a breweri and tavernerihaplotype tree (Fig. 2) confirms recent divergence (25). That no other such topologies are known for songbird species pairs is further evidence against the LPO model.

Figure 2

(right). A neighbor-joining tree based on a Kimura two-parameter (14, 15) distance matrix for all members of the avian genusSpizella. The tree is rooted at Junco, with numbers indicating node strength over 1000 bootstrap replications. The average pairwise distance among pusilla,pallida, atrogularis, and passerinais 5.7% (SD ± 0.5). U.S. state and Canadian province abbreviations are given in parentheses.

Differing interpretations of the avian fossil record have led to controversy over the timing of the origin of songbirds. Wetmore (26) advocated Late Pliocene origins, whereas others [for example, see (7)] subsequently postulated a Late Pleistocene radiation. Recent molecular studies (16,17) of single avian genera again point to somewhat earlier origins. Our comprehensive analyses of mtDNA divergence values and haplotype trees now establish that the majority of North America's “youngest” species have Early Pleistocene or Late Pliocene origins.

The LPO model follows from the premise that glacial cycles provided conditions conducive to speciation (27). However, ice sheets grew large in the Northern Hemisphere beginning 2.4 Ma (28), and climatic oscillations sufficient to produce major changes in flora and fauna are now dated to the Tertiary (8). Thus, large-scale geographic shuffling, splitting, and bottlenecking of populations have been occurring since at least that time. The data presented here (Fig. 1) are congruent with this view in suggesting a relatively continuous history of speciation events. Periodic glacial cycles may have strongly influenced the diversification of the North American songbird fauna, but if so, these events occurred much earlier than is typically proposed (4,5). Our results show that evidence of Late Pleistocene diversification for songbirds will more likely be found among geographically segregated conspecific populations and subspecies (8) but not among traditionally recognized sister species. The most recent glaciations were not, it seems, the force driving songbird diversification so much as they functioned as an ecological obstacle course through which only some species were able to persist (29). The entrenched paradigm proclaiming that many North American songbird species originated as a consequence of these glaciations is flawed.


View Abstract

Navigate This Article