MBE Advance Access originally published online on October 17, 2006
Molecular Biology and Evolution 2007 24(1):26-53; doi:10.1093/molbev/msl150
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Review |
Paleontological Evidence to Date the Tree of Life
Department of Earth Sciences, Wills Memorial Building, Queen's Road, Bristol BS8 1RJ, UK
E-mail: mike.benton{at}bris.ac.uk.
 |
Abstract |
|---|
 TOP Abstract IntroductionMinimum Constraints on...The Nature of Minimum...AcknowledgementsReferences |
|---|
The role of fossils in dating the tree of life has been misunderstood. Fossils can provide good "minimum" age estimates for branches in the tree, but "maximum" constraints on those ages are poorer. Current debates about which are the "best" fossil dates for calibration move to consideration of the most appropriate constraints on the ages of tree nodes. Because fossil-based dates are constraints, and because molecular evolution is not perfectly clock-like, analysts should use more rather than fewer dates, but there has to be a balance between many genes and few dates versus many dates and few genes. We provide "hard" minimum and "soft" maximum age constraints for 30 divergences among key genome model organisms; these should contribute to better understanding of the dating of the animal tree of life.
Key Words: tree of life • paleontological dating • calibration • quality of fossil record
|
Introduction |
|---|
|
TOPAbstractIntroductionMinimum Constraints on...The Nature of Minimum...AcknowledgementsReferences |
|---|
Calibrating the tree of life has long been the preserve of paleontology but its place has recently been usurped completely by molecular clocks. Fossil data are fundamental to molecular clock methodology, providing the key means of clock calibration, but their commonplace use is far from satisfactory. We consider the utility and qualities of good calibration dates and, on that basis, we propose a number of well-supported dates, and give ages based on the best current information. In doing this, we argue that paleontological data do not provide actual age estimates for divergence events, but they can provide rather precise minimum constraints on the calibration of molecular clocks, and much looser maximum constraints. The evidence of a "hard" lower bound (minimum constraint) and a "soft" upper bound (maximum constraint) provided from paleontology can then be fed into a molecular clock analysis. It is not our aim to determine the actual timing of divergence events as we do not believe that this is possible using paleontological data alone—though paleontological data can be used to test dates estimated using molecular clock methods (e.g., Foote et al. 1999
; Tavaré et al. 2002).
Traditionally, very small numbers of calibration dates have been employed and these have been selected for utility and have rarely been defended. The most commonly used calibration node is the mammal–bird divergence, dated at 310 MYA and accepted in some 500 or more publications since 1990. This date was based on the age of the oldest members of the synapsid and diapsid clades (Benton 1990), and yet these basal fossils have been debated, as has the dating of the rocks from which they come. Recently, authors have suggested an age range from 330 to 288 MYA at most (Lee 1999; Reisz and Muller 2004; van Tuinen and Hadly 2004). So, which date is to be used, and what does that date really represent?
It is clear that the fossil record cannot be read literally (Darwin 1859). There are many gaps, and many organisms, and indeed whole groups of poorly preservable organisms that have never been preserved and are doubtless lost for ever (Raup 1972). Some have even gone so far as to suggest that the fossil record is almost entirely an artifact of the rock record, with appearances and disappearances of fossil taxa controlled by the occurrence of suitable rock units for their preservation (Peters and Foote 2001, 2002), or the matching rock and fossil records controlled by a third common cause (Peters 2005). However, the widespread congruence between the order of fossils in the rocks and the order of nodes in cladograms (Norell and Novacek 1992; Benton et al. 2000) indicates that the order of appearance of lineages within the fossil record is not a random pattern. Furthermore, a fossil of any age demonstrates the divergence of its lineage, and so provides an absolute constraint on the temporal dimension of the tree of life.
Traditionally, calibration dates have been assumed to indicate the timing of an evolutionary divergence event, as a basis for inferring rates of functionally equivalent amino acid or nucleotide substitution (in proteins or genes, respectively), from which the timing of other lineage-splitting events may be deduced (Zuckerkandl and Pauling 1965). However, paleontological data can provide good estimates only for minimum constraints on the timing of lineage divergence events (Benton and Ayala 2003; Hedges and Kumar 2004; Reisz and Muller 2004). Note that relaxed-clock methods can often require at least one point calibration or hard maximum constraint in order for the algorithm to converge on a unique solution. So, debates about the superiority of one "calibration" date or another are irrelevant in the context of a search for the most appropriate distribution of dates and minimum and maximum constraints—the only bad dates are those that predate the evolutionary event upon which they are supposed to provide a minimum constraint.
Deviations from the molecular clock may occur because of changes in selective pressures and mutation rates, and this requires that molecular clock analyses rely upon a law of large numbers in which an average rate may be derived from a data set that is sufficiently large (Rodríguez-Trelles et al. 2003). It is still debated whether an analysis based on many genes and few dates or few genes and many dates is preferable. However, multiple calibration points are particularly helpful in relaxed-clock methods where the rate is allowed to vary among branches in the tree; multiple calibrations throughout the tree act as anchor points, allowing the method to estimate the patterns and degree of rate variation more accurately. Good estimates of rate variation are required from the well-calibrated regions of the tree so that the pattern can be extrapolated to other parts of the tree that are poorly calibrated. Furthermore, molecular clock analyses are rarely, if ever, framed around the availability of reliable calibration dates. Rather, they are characterized by scientifically interesting questions and the availability of appropriate sequence data (Hedges and Kumar 2004). Together, these facts require that well-researched calibration dates are available for the majority of available sequence data and, to this end, we provide detailed assessments of the paleontological data constraining the timing of lineage splits between the main genomic models.
|
Minimum Constraints on Divergence Dates |
|---|
|
TOPAbstractIntroductionMinimum Constraints on...The Nature of Minimum...AcknowledgementsReferences |
|---|
The indicated range of minimum branching dates (table 1) reflects both uncertainty in the dates (stratigraphic error) as well as the inferred duration of the fossiliferous unit. Such a small range of dates, less than 1% in many cases, may seem startlingly low, but current geological timescales (Gradstein et al. 2004
) offer that level of precision. The quoted age range does not incorporate an estimate of uncertainty about whether the oldest fossil really belongs to the clade or about whether the clade might have originated much earlier. The date arose from a 2-step process: 1) Which is the oldest relevant fossil within the clade in question? 2) What is the best current age estimate for the geological formation that includes that fossil?
|
The first step relied on our reading of current paleontological data, and wide consultation on each date with relevant experts. We excluded all uncertain or scrappy fossils, and retained only those for which there is definitive anatomical evidence of one or more apomorphies of the clade in question. In all cases, the date is sought for branching between 2 extant species, and so we pursued each of the 2 lineages back to the point at which they shared their last common ancestor, based on current phylogenetic evidence. Having 2 lineages meant, we could select the older of the 2 oldest fossils (table 1).
The second step is to date the geological formation in which the oldest fossil, or fossils, occurs, or occur. The identity of that geological formation is clear in all cases—the earliest members of the Zebrafish (ostariophysean) and Pufferfish (euteleost) lineages, for example, both date from the lithographic limestones of the Obere Solnhofener Schichten of southern Germany. A geological formation is a well-constrained succession of rocks with a clearly marked base and top. In most cases, there is an extensive biostratigraphic literature devoted to establishing the relative age of the unit in question. For the Solnhofen lithographic limestones, ammonites and other fossils place the unit in the lower Tithonian stage (zeta 2a zone) of the Upper Jurassic. That is a relative age, refined to a zonal level that may be less than 1 Myr in duration. Absolute chronostratigraphic ages are then assigned by reference to the international standard, with precise ages established by radiometric methods. The zeta 2a zone is part of the Hyboniticeras hybonotum ammonite zone, the base of which coincides with the base of the normal-polarity Chron M22An magnetozone that is dated at 150.8 MYA ± 0.1 Myr (Ogg 2005); a minimum constraint on its age can be derived from the base of the succeeding, Semiformiceras darwini ammonite zone that coincides approximately with the M22n Chronozone, dated at 149.9 MYA ± 0.05 Myr (Ogg 2005). This is the current best estimate of the minimum date of divergence of the Zebrafish and Pufferfish genomes. Here, and in our tabulation of divergence dates, we provide minimum constraints. However, we provide the full range of error for those who wish to perpetuate the use of paleontological dates as direct substitutes for divergence times.
|
The Nature of Minimum Constraints and the Need for Maximum Constraints |
|---|
|
TOPAbstractIntroductionMinimum Constraints on...The Nature of Minimum...AcknowledgementsReferences |
|---|
Some molecular clock analyses have been calibrated using a single fossil-based date that was assumed to have no error, or with an error distributed symmetrically on either side and with uniform probability between the minimum and maximum bounds and zero probability that the date falls outside the interval (Hedges and Kumar 2004
). However, fossil calibrations are minimum dates that provide asymmetrical constraints, below which probability drops immediately to zero, but above which probability decays more gradually, and probability densities can be modeled in a variety of ways to reflect the quality of fossil calibrations (Hedges and Kumar 2004; Kumar et al. 2005; Drummond et al. 2006). Drummond et al. (2006) outline a number of parametric probability distributions for the ages of nodes, including normal, lognormal, exponential, and uniform distributions, that may be used as priors in Bayesian treatments of relaxed-clock models of sequence evolution. The shape of the probability distribution selected can then reflect current biological understanding of the shape of the base of a clade.
Probability distributions of potential ages for the origin of a clade between the maximum and minimum constraints may be modeled to reflect the postulated shape of the base of the clade in question. Paleontologists have described long, thin, spindle-shaped clades and short, fat clades (Gould et al. 1977). Empirical observations suggest that all clades, whatever their shape, expand from one species to many following a logistic curve (Gould et al. 1977; Sepkoski 1996; Tavaré et al. 2002). There may be a long or short initial phase when diversity is low, and then species are added until some kind of "equilibrium" clade species richness is achieved. The logistic model is in line with expectations from ecological models such as the Lotka-Volterra models of competition and the island biogeography model (MacArthur and Wilson 1967; Rosenzweig 1995). If the logistic model is appropriate, then the initial tail, whether long or short, would generally fall outside the maximum constraint if that were set as a 95% confidence interval.
The ends of such distributions have been termed "hard" and "soft" bounds (Hedges and Kumar 2004; Yang and Rannala 2006). A "hard bound" is absolute, and the date cannot fall beyond it, whereas a soft bound is not, and divergence dates could lie beyond it, to a degree that is dependent upon the probability density modeled. The probability density may be entirely arbitrary, or informed predictions about the shape and extent of the probability tail leading from the hard bound may be made based on the nature of the paleontological data (Hedges and Kumar 2004; Barnett et al. 2005; Yang and Rannala 2006). "Soft bounds" for maximum age constraints allow paleontologists to propose short, but realistic, time extensions below the oldest known fossil in a group; if the maximum age constraint is a hard bound, that estimate has to be very large in some cases just to allow for the faintest possibility of a very ancient fossil. Thus, soft bounds provide not only a means of reflecting the nature of the fossil record beyond providing a minimum date but they also lend themselves well to relaxed-tree algorithms in which some age constraints may be better than others, but which are good and which are bad is not known a priori. In the context of parametric probability distributions, the minimum and maximum constraints could be equated with 95% lower and upper limits, and this would allow the placement of a curve and its mean; we use the terms maximum and minimum constraint bounds for the moment because they could then be set as 99%, 95%, or 90% confidence limits for example.
A number of approaches may be taken in determining soft bounds. One approach is to consider all possible sources of error in estimating the maximum date of origin of a clade of which there are 5 broad categories of error: 1) phylogenetic topology, 2) fossil record sampling, 3) identification, 4) correlation (relative dating), and 5) exact age–date assignment (absolute dating). These errors are nonadditive but some (e.g., phylogenetic topology) may be difficult to constrain. Another approach is to model diversification pattern and preservation probability (Foote et al. 1999; Tavaré et al. 2002). Phylogenetic bracketing has also been used to provide a maximum constraint on divergence events, by bracketing the next node below and above (Reisz and Muller 2004; Müller and Reisz 2005), and even conflated with estimates of errors on each of these dates (van Tuinen and Hedges 2001; van Tuinen and Hadly 2004). However, although this method may be beguiling, all nodes used to constrain the timing of divergence are subject to the usual uncertainties of dating fossil occurrences. There is no reason why the date of the node below should be related in any way to the date of origin of the next clade above.
In seeking to determine a maximum age constraint on the origin of a clade, there is merit in modeling diversification pattern and preservation probability and in phylogenetic bracketing, but neither can ever provide a definitive answer. In practice, the degree of precision provided by some of these approaches is false and is beyond that needed to attain computational feasibility in constraining molecular clock analyses. For the moment, we prefer to use a combined, but intuitive, approach.
Our method uses aspects of phylogenetic bracketing and stratigraphic bounding, namely a consideration of the absence of fossils from underlying deposits. The line of reasoning is broadly the following: 1) the maximum age constraint for the origin of a clade will be older than the oldest definitive fossil in the clade; 2) older fossils that might belong to the clade, or to its stem lineage, can hint at (but never prove) a downward time extension; 3) older fossils in clade C, the nearest outgroup (fig. 1) could also hint at (but never prove) a downward time extension; and 4) an older fossil deposit that ought to contain fossils of the clade in question, but does not, can mark an ultimate maximum bound. We do not, here, guarantee that an older fossil will never be found, but the likelihood is low, and this will be reflected in the probability density (Yang and Rannala 2006). Probability densities have been used in deriving the confidence interval either with (Yang and Rannala 2006) or without (Kumar et al. 2005) the assumption of a molecular clock. The probability density can be modeled accurately on the basis of recovery potential functions that incorporate data on ecological distribution conflated with data on facies variation, outcrop exposure, and even taphonomic controls provided by anatomically similar organisms (Holland 1995; Marshall 1997). Alternatively, the probability density may be entirely arbitrary, for example, described by a lognormal distribution; even such simple models can be readily adapted to approximate reality by, for instance, using fossil and lithostratigraphic data to inform the position of the mean.
|
We emphasize finally that minimum and maximum constraints on calibration dates should be fully substantiated so that if any of the variables change, such as recently with the publication of the new geological timescale (Gradstein et al. 2004
), with a shift in phylogenetic hypothesis, or the discovery of an older member of the clade, the impact of the change upon the calibration date is obvious and may be refined. Thus, minimum age constraints should be justified on the basis of a phylogenetic hypothesis, with reference to the oldest integral member of the clade—on which the date is ultimately based, the justification for its membership of the clade, the means by which the correlation is achieved to a section in which a chronostratigraphic date may be obtained, and the source of the chronostratigraphic date. Maximum age constraints should likewise be justified on the basis of a phylogenetic hypothesis, with reference to fossils belonging to outgroups and to putative stem groups, and to the next oldest fossil horizon that lacks relevant fossils. Our proposed calibrations are justified below and summarized in table 1 and figure 8.
|
Dating Divergences among Mammals
Eighteen mammalian genomes have been sequenced, or are in progress (August 2006; http://www.ensembl.org/index.html), namely human (Homo sapiens), chimp (Pan troglodytes), macaque (Macaca mulatta), mouse (Mus musculus), rat (Rattus norvegicus), rabbit (Oryctolagus cuniculus), dog (Canis familiaris), cat (Felis catus), horse (Equus caballus), pig (Sus scrofa), sheep (Ovis aries), cow (Bos taurus), armadillo (Dasypus novemcinctus), tenrec (Echinops telfairi), African elephant (Loxodonta africana), kangaroo (Macropus eugenii), opossum (Monodelphis domestica), and platypus (Ornithorhynchus anatinus). These 18 consist of 1 monotreme (the platypus), 2 marsupials (the opossum and kangaroo) and 15 placental mammals, members of the clade Eutheria. According to current molecular and morphological phylogenies (Madsen et al. 2001
; Murphy et al. 2001; Huchon et al. 2002; Springer et al. 2003; Benton 2005), the Eutheria fall into 3 main clades, Afrotheria, Xenarthra, and Boreoeutheria. The 12 placental mammals include 2 afrotherians (tenrec and elephant), 1 xenarthran (armadillo), and the remaining 9 belong to Boreoeutheria that fall into 2 clades, the Laurasiatheria, containing the orders Artiodactyla (pig, cow, and sheep), Perissodactyla (horse), and Carnivora (dog and cat), and the Euarchontoglires, containing the orders Primates (macaque, chimp, and human), Lagomorpha (rabbit), and Rodentia (mouse and rat). The tree of 18 major mammalian groups (fig. 2) then contains 17 branching points: 7 within major clades (opossum–kangaroo, cow–sheep, cow–pig, cat–dog, human–chimp, human–macaque, mouse–rat) and the other 10 between orders or higher clades, namely human–mouse (i.e., Primates–Rodentia), rabbit–mouse (i.e., Glires), horse–dog (i.e., Perissodactyla–Carnivora), cow–dog (i.e., Ferungulata), human–cow (i.e., Euarchontoglires–Laurasiatheria), human–armadillo (i.e., Boreoeutheria–Xenarthra), tenrec–elephant (i.e., Afrotheria), human–tenrec (i.e., Boreoeutheria–Xenarthra–Afrotheria), human–opossum (i.e., Eutheria–Marsupialia), and human–platypus (i.e., Theria–Monotremata). Other pairings of taxa could be selected, but they are synonymous with 1 of these 10 (e.g., rat–cow is the same as human–cow; dog–opossum is the same as human–opossum).
|
These 17 branching points will be considered in order (see fig. 2).
Human–Chimp
The dating of the chimp–human split has been discussed for nearly a century. Early paleontological estimates, up to the 1970s, placed the branching point deep in the Miocene, at perhaps 20–15 MYA, but this was revised dramatically upward to about 5 MYA by early molecular studies (Sarich and Wilson 1967), and estimates as low as 2.7 MYA have been quoted (Hasegawa et al. 1985). Paleontological evidence for the branching point was distinctly one-sided until recently, since the only fossils fell on the human line, and so the question of the date of divergence of humans and chimps became synonymous, for paleontologists, with the date of the oldest certain hominin (species on the human, not chimp, line). The recent discovery of the first chimpanzee fossils (McBrearty and Jablonski 2005) does not change much, as they are dated as 545,000 years old at most.
The date of the oldest hominin has extended backward rapidly in the last 25 years. Until 1980, the oldest fossils were gracile and robust australopithecines from 3 MYA. The discovery of "Lucy", now termed Praeanthropus afarensis in Ethiopia (Johanson and Taieb 1976) extended the age back to 3.2 MYA at most. Then, 2 further hominin species pushed the age back to over 4 Myr: Ardipithecus ramidus from rocks dated as 4.4 MYA from Ethiopia (White et al. 1994) and Praeanthropus anamensis from rocks dated as 4.1–3.9 MYA from Kenya (Leakey et al. 1995). More recent finds, remarkably, have pushed the dates back to 6 Myr: A. ramidus kadabba from Ethiopia (5.8–5.2 MYA; Haile-Selassie 2001), Ororrin tugenenis from Kenya (c. 6 MYA; Senut et al. 2001), and Sahelanthropus tchadensis from Chad (6–7 MYA; Brunet et al. 2002). The last 2 taxa have proved highly controversial, with claims that one or other, or both, are not hominin, but ape like. However, the majority view is that Sahelanthropus at least is hominin (Wood 2002; Cela-Conde and Ayala 2003), and so its date becomes crucial.
Dating of the Sahelanthropus beds in Chad is not direct. Biostratigraphic evidence from mammals in particular, but with cross-checking from fish and reptile specimens, indicates that the unit is definitely late Miocene (i.e., older than 5.33 MYA), and it is older than the Lukeino Formation of Kenya, the source of Orrorin (dated at 6.56–5.73 MYA from Ar/Ar dates on volcanic layers; Deino et al. 2002), and may be equivalent to the lower fossiliferous units of the Nawata Formation at Lothagam (dated as 7.4–6.5 MYA; Vignaud et al. 2002). This might suggest a date for the sediments containing Sahelanthropus of 7.5–6.5 MYA, based on biostratigraphy and external dating. Thus, we determine a 6.5-MYA age for the minimum constraint on the human–chimp split. Kumar et al. (2005) have recently calculated a range of ages for the human–chimp divergence of 4.98–7.02 MYA; their minimum constraint (4.98 MYA) is younger than the oldest fossils (Orrorin, Sahelanthropus). However, paleoanthropogists generally accept that Sahelanthropus and Orrorin were both bipedal, upright forms, and until both are rejected by consensus view of their anatomy, we retain them as the oldest valid hominins.
A soft maximum constraint on the human–chimp divergence is hard to place because the immediate outgroups (gorilla, orang, and gibbons) lack convincing fossil records. Some late Miocene ape fossils, such as Gigantopithecus and Sivapithecus may be stem-orangs. Nonetheless, a range of such apes, Ankarapithecus from Turkey (10 MYA), Gigantopithecus from China (8–0.3 MYA), Lufengopithecus from China (10 MYA), Ouranopithecus from Greece (10–9 MYA), and Sivapithecus from Pakistan (10–7 MYA) give maximum ages of 10 MYA, early in late Miocene, and these deposits have yielded no fossils attributable to either chimps or humans. This is taken as the soft maximum constraint on the human–chimp divergence.
Human, Chimp–Macaque
The human–macaque split is equivalent to the branching of Old World monkeys (Cercopithecoidea) and apes (Hominoidea), which together form the clade Catarrhini.
The oldest cercopithecoids are Victoriapithecus macinnesi from Kenya, and 2 species of Prohylobates from Libya and Egypt. Miller (1999) surveyed all fossils of these 2 genera, and compared ages of their respective deposits. The oldest cercopithecoid fossil is a tooth identified as Victoriapithecus sp. from Napak V, Uganda (c. 19 MYA), followed by Prohylobates tandyi from Moghara, Egypt (18–17 MYA) and Prohylobates sp. from Buluk, Kenya (>17.2 MYA), P. simonsi from Gebel Zelten, Libya (c. 17–15 MYA), and V. macinnesi from Maboko, Kenya (ca. 16–14.7 MYA). MacLatchy et al. (2003) report an even older cercopithecoid, a fragment of a maxilla from the Moroto II locality in Uganda, that has been radiometrically dated to be older than 20.6 MYA ± 0.05 Myr (Gebo et al. 1997).
The oldest hominoids include Morotopithecus, also from the Moroto II locality in Uganda (Gebo et al. 1997). Young and MacLatchy (2004) determined that this taxon is a hominoid, located in the cladogram above the gibbons, and so not the most basal member of the group. Because of incompleteness of the material, Finarelli and Clyde (2004) are less certain of its phylogenetic position, but Morotopithecus is certainly a catarrhine. Even older is the first record of the long-ranging hominoid genus Proconsul from Meswa Bridge in Kenya, biostratigraphically constrained to c. 23.5 MYA (Pickford and Andrews 1981; Tassy and Pickford 1983). Even older is the purported hominoid Kamoyapithecus from the Eragaliet Beds of the Lothidok Formation of Kenya, dated at 24.3–27.5 MYA (Boschetto et al. 1992), but the material is insufficient to determine whether it is a hominoid or a catarrhine, possibly lying below the human–macaque split (Finarelli and Clyde 2004).
So, the minimum constraint on the human–macaque split is 23.5 MYA, based on the oldest record of Proconsul, or perhaps 23.5 MYA ± 0.5 Myr, based on biostratigraphy and external dating.
The soft maximum constraint is based on members of the stem of Catarrhini, namely the families Propliopithecidae (Propliopithecus, Aegyptopithecus) and Oligopithecidae (Oligopithecus, Catopithecus) that are basal to the cercopithecoid-–hominoid split (Rasmussen 2002). These are represented in particular from the rich Fayûm beds in Egypt, dated as early Oligocene (33.9–28.4 MYA ± 0.1 Myr), and so 28.3 MYA, deposits that have produced many primate, and other mammal, fossils, but no hint of a crown-group catarrhine.
Mouse–Rat
The mouse (M. musculus) and rat (R. norvegicus) are both the members of the subfamily Murinae within the family Muridae, members of the larger clade of muroid rodents. The Old World rats and mice are hugely diverse, with over 500 species, and they appear to have radiated relatively rapidly in Europe, Africa, Asia, and Australia.
The phylogeny of all genera within Murinae has not been determined, so the location of the split between Mus and Rattus is somewhat speculative at present. However, all current morphological and molecular phylogenies (Michaux et al. 2001; Jansa and Weksler 2004; Steppan et al. 2004; Chevret et al. 2005) indicate that Mus and Rattus diverged early in the evolution of Murinae, but not at the base of the divergence of that clade. A lower limit to the mouse–rat divergence is indicated by the oldest known murine fossil, Antemus chinjiensis from the middle Miocene Chinji Formation of Pakistan, dated at about 14.0–12.7 MYA on the basis of magnetostratigraphy and radiometric dating (Jacobs and Flynn 2005).
The oldest fossil example of Mus dates from 7.3 MYA, a specimen of Mus sp. from locality Y457 in the Siwaliks (Jacobs and Flynn 2005). Fossils of Rattus are not known until the latest Pliocene and the Pleistocene of Thailand (Chaimanee et al. 1996) and China (Zheng 1993), no more than 3 MYA.
The divergence of the 2 lineages leading to Mus and Rattus was stated to be 14–8 MYA by Jacobs and Pilbeam (1980), in a first review of the fossil evidence. This range was narrowed down at its older end to 12 MYA in subsequent studies (Jaeger et al. 1986; Jacobs and Downs 1994), based on the first appearance of the fossil genus Progonomys, early members of which were assumed to include the common ancestor of Mus and Rattus. The 12 MYA figure has most commonly been selected as the mouse–rat calibration point, but dates in the range from 16 to 8.8 MYA have been used in recent molecular studies.
In a thorough review of the fossil evidence, Jacobs and Flynn (2005) show that records of Progonomys in the Siwalik succession extend from 12.3 to 8.1 MYA, with the later forms (10.4–8.1 MYA) assumed to lie on the Mus lineage. The extinct genus Karnimata (11.1–6.4 MYA) is interpreted as a member of the lineage leading to Rattus. The oldest record (11.1 MYA) is uncertain, but the next (at 10.4 MYA) is unquestionable. The early species, Progonomys hussaini (11.5–11.1 MYA) is interpreted as an undifferentiated basal murine antedating the common ancestor of Mus and Rattus by Jacobs and Flynn (2005), and so they place the Progonomys–Karnimata split (equivalent to the Mus–Rattus split) at not much beyond 11 MYA, "although it may be younger." The dating is based on detailed field stratigraphic study of the long Siwaliks sedimentary sequence, with dating from magnestostratigraphy and radiometric dating (Johnson et al. 1985; Barry et al. 2002). The soft maximum constraint on this date is taken as the oldest record of Progonomys at 12.3 MYA.
Rabbit–Mouse, Rat
The rabbit–mouse basal node is synonymous with the clade Glires, comprising orders Rodentia plus Lagomorpha. The date would have been assumed traditionally to lie at 65 MYA, or younger, marking the time of purported placental mammal radiation after the extinction of the dinosaurs.
There have never been any records of Cretaceous rodent fossils, even though some molecular studies have placed the origin of the order deep within the Cretaceous. The oldest fossil rodents are known with confidence from the Thanetian (late Paleocene, 58.7–55.8 MYA), members of the family Ischyromidae from North America and Europe (Stucky and McKenna 1993), after which the clade expanded enormously to its present huge diversity. An older putative rodent might be Heomys, a eurymylid from the Danian (early Paleocene, 65.5–61.7 MYA) of China (McKenna and Bell 1997). The eurymylids may not be proper rodents, but members of a larger including clade Simplicidentata, or they may fall outside Simplicidentata, but within Glires, as outgroup to rodents and rabbits (Asher et al. 2005). Either way, the oldest members of Glires are post-Cretaceous in age (<65 MYA). Whether the Late Cretaceous zalambdalestids are related to Glires or not (see below) is irrelevant to this node.
The oldest lagomorphs are somewhat younger. Stucky and McKenna (1993) indicate several Eocene rabbits from the Lutetian: Lushilagus from China, Procaprolagus from Canada, and Mytonolagus from the United States. Meng and Wyss (2005) note an older possible lagomorph, Mimotona from the early to late Paleocene (Doumu Formation, Nonshangian, Qianshan Basin, China), the same unit that yielded the putative earliest rodent Heomys.
The minimum constraint on the age of clade Glires, and so for the rabbit–mouse split, is 61.7 MYA. The nearest outgroups of Glires (Meng and Wyss 2005) and forms such as Pseudictops, Anagale, and Hyopsodus are later Paleocene than Heomys, and so of little assistance in indicating a possible soft maximum constraint. The next outgroups, possibly the zalambdalestids, set a much older soft maximum constraint of 99.6 MYA ± 0.9 Myr to 96.2 MYA ± 0.9 Myr.
Human, Chimp, Macaque–Rabbit, Mouse, Rat
The human–mouse split is synonymous with the latest branching point between the mammalian orders Primates and Rodentia. Both orders are members of the clade Euarchontoglires. Euarchontoglires is composed of 2 clades, the Archonta and the Glires, and Primates belongs to the former, Rodentia to the latter. Thus, the human–mouse split becomes synonymous with the origin of Euarchontoglires.
Traditionally, this branching point would have been set at 65 MYA, the beginning of the Paleogene (base of Cenozoic, base of Tertiary), and corresponding to the extinction of the dinosaurs and the beginning of the radiation of placental mammal orders. This view has been challenged since 1995 as a result of 2 factors: 1) the discovery of major supraordinal clades within Eutheria, as noted earlier and 2) the repeated discovery from molecular analyses that the eutherian orders and the larger clades might have their origin at some point in the Cretaceous, whether rather early (Hedges et al. 1996; Janke et al. 1997) or rather later, and more in line with the fossils (Murphy et al. 2001; Arnason et al. 2002; Springer et al. 2003), evidence perhaps of a rapprochement between molecular and paleontological evidence (Archibald 2003; Benton and Ayala 2003).
There are no confirmed fossils of Primates or Rodentia in the Cretaceous (i.e., >65 MYA). An isolated tooth from the latest Cretaceous Hell Creek Formation (c. 67 MYA) of North America was assigned to the plesiadapiform taxon Purgatorius, and has been cited as the oldest primate (Van Valen and Sloan 1965). However, the phylogenetic position of the plesiadapiforms is debated—they were probably close relatives of primates, but not primates proper (Bloch and Boyer 2002). Further, the single tooth is arguably too little evidence for a firm record (Archibald 2003). The oldest confirmed primates are from the Paleocene–Eocene transition, some 55 MYA (Bloch and Boyer 2002), and the oldest plesiadapiform is Subengius from the late Paleocene of China (Smith et al. 2004). It comes from the Nomogen Formation, assigned to the Gashatan Land Mammal Age (latest Paleocene, 57–56 MYA).
As noted above, the oldest fossil rodents are known with confidence from the Thanetian (late Paleocene, 58.7–55.8 MYA), members of the family Ischyromidae from North America and Europe (Stucky and McKenna 1993), after which the clade expanded enormously to its present huge diversity.
Some Cretaceous fossils might be relevant to the node at the base of Euarchontoglires, however: the zalambdalestids, a group of small, long-legged jumping mammals known from excellent fossils from the Late Cretaceous of Mongolia and Central Asia (Kielan-Jaworowska et al. 2000; Archibald et al. 2001). They have been assigned numerous phylogenetic positions, but were found to be outgroup of rodents and rabbits, either members of the clade Glires or close to it (Archibald et al. 2001). Until recently, the zalambdalestids from the Bissekty Formation of Dzharakuduk, Kyzylkum Desert, Uzbekistan, were the oldest known of this clade, but they are now thought to come from the older Khodzhakul Formation at Sheikhdzheili, Kyzylkum Desert, Uzbekistan. There are 3 sets of localities in the Kyzylkum Desert that have yielded mammals. Based on biostratigraphic studies of intercalated marine units with invertebrate fossils (Averianov 2000; Archibald et al. 2001; Archibald 2003), these 3 local faunas are early Cenomanian (about 97 MYA), late Turonian (about 90 MYA), and possibly Coniacian (about 87 MYA). The age of the Khodzhakul Formation is particularly crucial: a reworked, early placenticeratid ammonite from the base of the formation suggests an early Cenomanian age, whereas an inoceramid bivalve from just above the Khodzhakul Formation suggests a late Cenomanian age (Averianov and Archibald 2005). So, the oldest zalambdalestids are from the early Cenomanian that corresponds to 99.6 MYA ± 0.9 Myr to 96.2 MYA ± 0.9 Myr.
This phylogenetic position has been challenged (Meng et al. 2003; Asher et al. 2005), and these authors place zalambdalestids outside the clade Placentalia, and certainly below Afrotheria in the cladogram of mammals. In this view, zalambdalestids would say nothing about the date of origin of either Glires or Euarchontoglires, both of which would revert to minimum origin dates of basal Paleocene (61.7 MYA). For the present, and until the contradictory views (Archibald et al. 2001; Meng et al. 2003; Asher et al. 2005) are resolved, we take a conservative view and place a minimum constraint on the human–mouse split in the early Paleocene, at 61.7 MYA. The soft maximum constraint is based on the assumption that zalambdalestids are close to Glires that corresponds to 99.6 MYA ± 0.9 Myr to 96.2 MYA ± 0.9 Myr. This soft maximum constraint is a long time before the minimum constraint.
Dog–Cat
The dog–cat split is equivalent to the branching point between the clades Caniformia (dogs, bears, raccoons, and seals) and Feliformia (cats, mongooses, and hyaenas), the major subdivisions of the Order Carnivora (Flynn and Wesley-Hunt 2005).
The oldest carnivores are members of the families "Miacidae" (paraphyletic) and Viverravidae, known from the early Paleocene onward (Stucky and McKenna 1993), but these lie outside the Caniformia–Feliformia clade (Flynn and Wesley-Hunt 2005), and so cannot provide a minimum date for the dog–cat split.
The oldest caniforms are amphicyonids such as Daphoenus and canids such as Hesperocyon, known first from the earliest Duchesnean North American Land Mammal Age (NALMA) that corresponds to magnetochron 18N, and is dated as 39.74 MYA ± 0.07 Myr, based on radiometric dating of the LaPoint Tuff (Robinson et al. 2004). Tapocyon may be an even older caniform; it comes from the Middle Eocene, Uintan, dated as 46–43 MYA (Wesley and Flynn 2003), although Flynn and Wesley-Hunt (2005) place this taxon outside the Carnivora.
The oldest feliforms may be the nimravids, also known first from the White River carnivore fauna of the Chadronian NALMA, with uncertain records extending to the base of that unit (Hunt 2004). The earliest Chadronian corresponds to the top of magnetochron 17N, and an age of 37.2–36.7 MYA (Hunt 2004; Prothero and Emry 2004).
Flynn et al. (2005) suggest a caniform–feliform split around 50 Myr, but the evidence at present suggests a minimum constraint of 43 MYA, based on magnetostratigraphy and radiometric dating of the Uintan NALMA. The soft maximum constraint is based on the occurrence of the oldest stem carnivores (miacids, viverravids) in the Torrejonian NALMA of the early Paleocene (see dog–horse below), so 63.8 MYA.
Dog, Cat–Horse
The dog–horse split is equivalent to the branching point between the orders Carnivora and Perissodactyla, that together form an unnamed clade. The minimum age will be determined from the oldest member of the carnivore and perissodactyl lineages.
Flynn et al. (2005) and others, have modified the meaning of Carnivora so that it is restricted by them to the crown clade consisting of Caniformia + Feliformia. They rename the more inclusive clade traditionally called Carnivora as Carnivoramorpha. They rename the more inclusive clade traditionally called Carnivora as Carnivoramorpha. The oldest carnivoramorphans are the viverravids. The oldest generally accepted viverravid is Protictis from the Fort Union/Polecat Bench Formation, assigned to the basal Torrejonian (To1) NALMA, and dated as 63.6–62.5 MYA (Lofgren et al. 2004). If Ravenictis from Canada is also a carnivoramorphan (Flynn 1998), and that is debated (Flynn and Wesley-Hunt 2005), it extends this date back to at least the Puercan (Pu2), 65.4–64.3 MYA ± 0.3 Myr. Most authors also agree that the extinct group Creodonta is sister group to Carnivoramorpha (Flynn and Wesley-Hunt 2005), and these date back to the Thanetian, 58.7–55.8 MYA ± 0.2 Myr, younger than the oldest carnivormorphans.
The oldest perissodactyl is represented by fragmentary teeth that resemble the brontotheriid Lambdotherium from the late Paleocene site of Bayan Ulan in China (Beard 1998), but the perissodacyl lineage may be extended further back in time. Among basal outgroups of Perissodactyla, Hooker (Hooker 2005) includes the phenacodont "condylarths" such as Ectocion, Phenacodus, and Tetraclaenodon. These all extend back into the Paleocene, and the oldest is Tetraclaenodon, known first from the basal Torrejonian (To1) of North America, the same age as the oldest creodont (above).
This places the dog–horse split minimally at the basal Torrejonian, and so 62.3 MYA. The soft maximum constraint is determined from the diverse fossiliferous units of similar facies in the North American Maastrichtian (70.6 MYA ± 0.6 Myr to 65.5 MYA ± 0.3 Myr) that have not produced remains identifiable to Carnivoramorpha or Perissodactyla, or to the stem lineages or either, providing a date of 71.2 MYA.
Cow–Sheep
The branching between the cow (Bos) and sheep (Ovis) is an intrafamilial split within the family Bovidae. Bos is a member of the Tribe Bovini and Ovis is a member of the Tribe Caprini that belong, respectively, to the subfamilies Bovinae and Antilopinae (Hassanin and Douzery 1999), although the monophyly of Antilopinae is questioned (Fernandez and Vrba 2005). These 2 subfamilies comprise the family Bovidae, so the cow–sheep split corresponds to the point of origin of the extant Bovidae.
Fernández and Vrba (2005) point to a major series of splits within Bovidae, that gave rise to the major subfamilies 25.4–22.3 MYA, and they link this to a major climatic change at the Oligocene/Miocene boundary. This date is, however, not based directly on fossil evidence, but upon a number of best-fitting dates from published morphological and molecular phylogenies.
A number of putative late Oligocene bovids (Stucky and McKenna 1993) have since been rejected. The oldest putative bovid was Palaeohypsodontus zinensis from the Oligocene of the Bugti Hills, Bolochistan, Pakistan, and the early Oligocene of Mongolia and China. This is identified as a ruminant, and was formerly at times assigned to Bovidae. However, it lacks unequivocal anatomical features of Bovidae, and is currently excluded from that family (Metais et al. 2003; Barry et al. 2005).
Fossil bovids may be identified in the fossil record by the presence of horn cores. The oldest such records, ascribed to Eotragus, come from the Early Miocene of Western Europe and Pakistan. For example, Eotragus noyi from the base of the terrestrial sequence on the Potwar Plateau is dated at approximately 18.3 MYA (Solounias et al. 1995).
Eotragus is attributed to Boselaphini, a tribe within the subfamily Bovinae consisting of the nilgai and other 4-horned antelopes. The oldest members of Antilopinae appear to come from the middle Miocene of 3 continents: Caprotragoides from Asia (India and Pakistan), Tethytragus from Europe (Spain and Turkey), and Gentrytragus from Africa (Kenya and Saudi Arabia), all dated at approximately 14 MYA (Vrba and Schaller 2000). The oldest firmly dated bovid then places the minimum constraint on the origin of the family at 18.3 MYA, and we set the soft maximum constraint as late Oligocene, the time of putative bovid fossils, so 28.5 MYA.
Cow, Sheep–Pig
The cow–pig split is equivalent to the major division in Artiodactyla between Ruminantia-Tylopoda and Suiformes. The oldest artiodactyls, such as Diacodexis from the Early Eocene of North America, fall outside this clade.
The oldest member of the Ruminantia-Tylopoda clade, the cows, deer, and camels, is the family Mixtotheriidae, represented by the single genus Mixtotherium (Theodor et al. 2005). The oldest records of Mixtotherium are from the Early Eocene (McKenna and Bell 1997), from the Cuisian mammalian fauna of France and Spain (Savage and Russell 1983). The Cuisian mammal age is the upper part of the Ypresian stage, equivalent to the Grauvian European Land Mammal Age (MP 10), dating from 51.0 to 48.5 MYA ± 0.2 Myr (Gradstein et al. 2004).
The Suiformes, or pig-like artiodactyls, include extant pigs, peccaries, and hippos, as well as the extinct raoellids and choeropotamids, of which the raoellids extend back to ca. 54 MYA (Theodor et al. 2005). The oldest raoellids include Khirtharia and Indohyus from the Early Eocene Kuldana Formation of Pakistan, dated to tethyan biozone P10, lower Lutetian, and dated as about 48 MYA (Gingerich 2003).
A confounding factor here is the suggestion that whales may be sister group to hippos (e.g., Ursing and Arnason 1998). So, if hippos are suiforms, are cetaceans also suiforms? In this case, the branching point in question would correspond to the split of whales and hippos. The alternative, and more likely, view (Theodor et al. 2005) is that whales and artiodactyls as a whole are sister groups, forming the larger clade Cetartiodactyla that split some 53.5 MYA (Gingerich 2005). This predates the cow–pig node, however.
So, based on Mixtotherium and the Indo-Pakistani raoellids, the cow–pig division is dated minimally at 48 MYA. The soft maximum constraint is selected as the putative date of splitting of Cetartiodactyla, so 53.5 MYA.
Cow, Sheep, Pig–Dog, Cat, Horse
The cow–dog split is equivalent to the branching point between the clades containing the orders Artiodactyla (even-toed ungulates) and Carnivora (flesh-eating placental mammals). This is synonymous with the point of origin of the clade Ferungulata, a clade within Laurasiatheria.
The oldest artiodactyl is Diacodexis from the Early Eocene of North America (c. 55 MYA). Artiodactyls are part of a larger clade Cetartiodactyla, with the Cetacea, whales and relatives, and these date back to the Early Eocene as well, at about 53.5 MYA (Theodor e al. 2005). The clade may also include the extinct mesonychids that are known first from the Danian/Thanetian, some 62 MYA (Stucky and McKenna 1993). The oldest carnivoramorphan is the miacoid Ravenictis from the Danian (Puercan, early Paleocene) of North America, and several carnivoran families radiated in the mid to late Paleocene of that continent (Meehan and Wilson 2002).
The clade Ferungulata includes also the orders Perissodactyla and Pholidota, but neither of these dates back before the early Eocene. The oldest fossil ferungulates by a long way may be the zhelestids from the Khodzhakul Formation of Dzharakuduk, Kyzylkum Desert, Uzbekistan. These were assigned to Laurasiatheria as basal "ungulatomorphs" (Archibald et al. 2001; Archibald 2003), that is, basal to the hoofed artiodactyls and perissodactyls. Averianov and Archibald (2005) reject Ungulatomorpha, as a polyphyletic group, and place Zhelestidae in Laurasiatheria; (J.D. Archibald, personal communication) further places Zhelestidae within Ferungulata. This then provides a minimum constraint on the human–cow split based on biostratigraphy and external dating evidence. As noted above, the Khodzhakul Formation is dated to the early Cenomanian (99.6 MYA ± 0.9 Myr to 96.2 MYA ± 0.9 Myr), hence 95.3 MYA.
The soft maximum constraint on this, and other basal dates among crown-group placentals, is the series of latest Early Cretaceous localities from North America and Mongolia, dated as Aptian and Albian. The Albian is dated as 112 MYA ± 1 Myr to 99.6 MYA ± 0.9 Myr, providing a date of 113 MYA.
Human, Chimp, Macaque, Rabbit, Mouse, Rat–Cow, Sheep, Pig, Rabbit, Dog, Cat
The human–cow divergence is synonymous with the origin of Boreoeutheria. This clade is composed of the clades Euarchontoglires (human) and Laurasiatheria (cow).
The oldest members of Euarchontoglires were noted above as the zalambdalestids from Uzbekistan, dated at 90–85 MYA (Archibald et al. 2001), although doubt has been expressed about their phylogenetic placement (Asher et al. 2005; Averianov and Archibald 2005).
A number of Late Cretaceous putative laurasiatherians have been cited. Although most orders within Laurasiatheria (Artiodactyla, Cetacea, Carnivora, Perissodactyla, Pholidota, Chiroptera) do not have fossil records older than Eocene or Paleocene, the Lipotyphla, the insectivores, may have Late Cretaceous representatives. McKenna and Bell (1997) reported the oldest lipotyphlan as Otlestes from the Cenomanian (99.6–93.5 MYA) of Uzbekistan, but Archibald (2003) regarded it as a basal eutherian, lacking apomorphies of Lipotyphla, or any other modern order. Most recently, Averianov and Archibald (2005) synonymized it with Bobolestes (from the same local fauna) and regarded it as a questionable zalambdalestoid. Next in time is Paranyctoides from the Turonian (93.4–89.3 MYA) of Asia and the Campanian (83.5–70.6 MYA) of North America, and Batodon from the Maastrichtian (70.6–65.5 MYA) of North America, both regarded as lipotyphlans by McKenna and Bell (1997). Archibald (2003) is uncertain, but retains these records pending discovery of further specimens.
More significant though are the zhelestids from the Bissekty Formation of Dzharakuduk, Kyzylkum Desert, Uzbekistan, and the even older Khodzhakul Formation at Sheikhdzhili. Zhelestids are assigned to Laurasiatheria (Archibald et al. 2001; Archibald 2003; Averianov and Archibald 2005; Wible et al. 2005), that is, basal to the hoofed artiodactyls and perissodactyls. Wherever Zhelestidae are assigned in the new conception of Laurasiatheria, they do apparently belong to that clade, and hence they provide a minimum age for the human–cow split based on biostratigraphy and external dating evidence. As noted above, the Khodzhakul Formation is dated to early Cenomanian; hence we propose a minimum constraint of 95.3.
The soft maximum constraint is, as for the cow–dog split above, 113 MYA.
Human, Chimp, Macaque, Mouse, Rat, Rabbit, Dog, Cat, Horse, Pig, Sheep, Cow–Armadillo
The human–armadillo split is equivalent to the origin of the clade comprising Boreoeutheria and Xenarthra. The oldest boreoeutherians are, as already noted, the zalambdalestids and zhelestids from the Khodzhakul Formation of Uzbekistan, dated to early Cenomanian, hence 99.6–96.2 MYA ± 0.9 Myr. The oldest reported xenarthrans are much younger, dating from the Paleocene. Riostegotherium is dated as Itaboraian (Rose et al. 2005), equivalent to the later Selandian (61.7–58.7 MYA ± 0.2 Myr).
The minimum constraint for the Boreoeutheria–Xenarthra split is then 95.3 MYA, and the soft maximum constraint is, as for the cow–dog split above, 113 MYA.
Tenrec–Elephant
The tenrec–elephant split represents a deep division within Afrotheria. According to current phylogenies, the tenrec, golden moles (Macroscelidea) and aardvark (Tubulidentata) may form one clade within Afrotheria, and the elephants, hyraxes, and sirenians form the other, termed Paenungulata. Paenungulata is widely accepted as a valid clade, having been established on morphological characters, and now confirmed by molecular analyses. Many systematists accept a grouping of tenrecs and golden moles in the clade Afrosoricida, and aardvarks may be sister to these, but that is unclear. In any case, the last common ancestor of tenrec and elephant corresponds to the base of crown-clade Afrotheria.
The oldest fossil aardvarks, tenrecs, and golden moles are all Miocene (McKenna and Bell 1997), with a possible older golden mole, Metoldobotes from the Late Eocene Jebel Qatrani Formation of Egypt. These are equaled or predated by the oldest paenungulates. The oldest hyraxes are known from the Eocene of North Africa (Gheerbrandt et al. 2005). The oldest sirenians are Prorastomus and Pezosiren from early middle Eocene of Jamaica (Gheerbrandt et al. 2005). The oldest proboscidean fossils are Phosphatherium and Daouitherium from Ypresian (lower Eocene) phosphorites of the Ouled Abdoun Basin of Morocco (Gheerbrandt et al. 2005). Extinct putative outgroups of crown-group Paenungulata such as Desmostylia and Embrithopoda (Arsinoitherium) are younger, being Oligocene in age, whereas the Anthracobunidae date back to the early Eocene.
At present, no extant clade within Afrotheria or any confirmed extinct afrothere clade predates the Ypresian (early Eocene) dated as 55.8–48.6 MYA ± 0.2 Myr, and this must be used as the basis of a minimum age constraint for the tenrec–elephant split of 48.4 MYA. Further study might reveal that certain Paleocene groups belong within one or other afrothere branch, and that could increase the minimum age constraint.
The maximum constraint is determined as equivalent to the maximum constraint on the age of Boreoeutheria and Xenarthra, because Afrotheria must be at least as old as its sister clades, although the large age extrension might in the end consist of stem-afrotherians that do not belong to either the Afrosoricida–Tubulidentata or the Paenungulata clades within Afrotheria, and hence would considerably overestimate the age of the tenrec–elephant split. The soft maximum constraint is then the soft maximum constraint on the age of Boreoeutheria, as noted above, namely 113 MYA.
Tenrec, Elephant–Human, Chimp, Macaque, Mouse, Rat, Rabbit, Dog, Cat, Horse, Pig, Sheep, Cow, Armadillo
The human–tenrec split is equivalent to the origin of the clade comprising Boreoeutheria, Xenarthra. and Afrotheria. The oldest boreoeutherians are, as already noted, the zalambdalestids and zhelestids from the Khodzhakul Formation of Uzbekistan, dated to early Cenomanian, hence 99.6–96.2 MYA ± 0.9 Myr. The oldest reported afrotherians are much younger, dating from the Eocene, as just noted. The oldest are Phosphatherium and Daouitherium from Ypresian (lower Eocene) phosphorites of the Ouled Abdoun Basin of Morocco (Gheerbrandt et al. 2005).
The minimum constraint for the Boreoeutheria/Xenarthra–Afrotheria split is then 95.3 MYA, and the maximum constraint is, as for the cow–dog split above, 113 MYA.
Opossum–Kangaroo
The opossum–kangaroo split is equivalent to the deep divergence among marsupial mammals between the clades Ameridelphia, the South American marsupials, and Australidelphia, the Australian marsupials (Amrine-Madsen et al. 2003; Nilsson et al. 2004). There are older marsupials from the Mid to Late Cretaceous, but these lie outside the split between the extant clades.
Until recently, the oldest ameridelphians came from the Tiupampa fauna from Bolivia (de Muizon and Cifelli 2000, 2001), type locality of the Tiupampan South American Land Mammal Age, and dated as 60.4–59.2 MYA ± 0.2 Myr (Gradstein et al. 2004), not 64.5–63 MYA, as is sometimes quoted (Nilsson et al. 2004). The fauna contains 11 ameridelphian marsupials, with representatives of several major lineages (didelphimorphs, sparassodonts), so the clade was already moderately diverse by this point. A new find of a possible polydolopimorphian, Cocatherium, extends the age back to Danian (Goin et al. 2006). Cocatherium is reported from the Lefipán Formation of Chubut, Argentina, a marine unit dated as basal Paleocene (basal Danian). The Late Cretaceous (possibly Campanian) La Colonia and Los Alamitos faunas do not contain marsupials, and the Laguna Umayo fauna (sometimes dated as latest Cretaceous) has been said to contain dental remains of the didelphid Peradectes in association with dinosaur eggs. However, the unit is now dated as late Paleocene to early Eocene, and it has not yielded dinosaurs. Various Cretaceous marsupials from North America have been included in Ameridelphia from time to time (Kielan-Jaworowska et al. 2005), but this is not supported by current cladistic analysis.
The oldest Australian marsupials are Thylacotinga and Djarthia from the Early Eocene Tingamarra local fauna from Murgon, Queensland, dated radiometrically at 54.6 MYA ± 0.5 Myr, indicating an Early Eocene age, and supported by biostratigraphy (Godthelp et al. 1999). The Australidephia date back further because, oddly, within this clade is the South American family Microbiotheriidae (and a putative microbiotheriid has been noted from the Murgon locality). The oldest microbiotheriid is Khasia from the Tiupampa fauna of Bolivia.
So, the oldest crown-group marsupial known to date is Cocatherium, an ameridelphian that is older than the oldest australidelphian, from the Danian (65.5 MYA ± 0.3 Myr to 61.7 MYA ± 0.2 Myr), providing the minimum constraint of 61.5 MYA on the opossum–kangaroo split.
The soft maximum constraint is determined as 71.2 MYA from the diverse fossiliferous units of similar facies in the North and South American Maastrichtian (70.6 ± 0.6 to 65.5 ± 0.3 MYA) that have not produced remains identifiable to either modern group of marsupials, or to the stem taxa, or either.
Opossum, Kangaroo–Human, Chimp, Macaque, Mouse, Rat, Rabbit, Dog, Cat, Horse, Pig, Sheep, Cow, Armadillo, Tenrec, Elephant
The human–opossum branching point is of course synonymous with the split of marsupials and placentals.
The earliest unequivocal marsupial dental fossils come from the mid Cretaceous of North America. The oldest of these is Kokopellia juddi reported (Cifelli 1993) from the Mussentuchit Member, in the upper part of the Cedar Mountain Formation, Utah, that is dated as middle to late Albian on the basis of bivalves and palynomorphs, and a radiometric date of 98.37 MYA ± 0.07 Myr was obtained from radiometric dating of zircons in a bentonitic clay layer. This suggests that the Mussentuchit Member extends to the Albian/Cenomanian boundary (99.6 MYA ± 0.06 Myr), but that the bulk of the unit is late Albian. Even older is the boreosphenidan Sinodelphys szalayi from the Yixian Formation, Liaoning Province, China, that is placed phylogenetically closer to marsupials than to placentals by Luo et al. (2003). This then has taken the root of the marsupial clade back to 125 Myr.
The oldest placentals were also, until recently, restricted to the mid and Late Cretaceous (Stucky and McKenna 1993), but subsequent finds have pushed the age back step-by-step deeper into the Early Cretaceous. First were Prokennalestes trofimovi and P. minor, reported from the Höövör beds of Mongolia (Kielan-Jaworowska and Dashzeveg 1989), and dated as either Aptian or Albian. Then came Montanalestes keeblerorum (Cifelli 1999) from the Cloverly Formation (late Aptian to early Albian, c. 100 MYA). Then, Murtoilestes abramovi was named (Averianov and Skutschas 2001) from the Murtoi Svita, Buryatia, Transbaikalia, Russia, being dated as late Barremian to middle Aptian (say, 128–120 MYA). These 3 taxa were based on isolated jaws and teeth. These were all topped by the spectacular find of Eomaia scansoria in the Yixian Formation of Liaoning Province, China (Ji et al. 2002), a complete skeleton with hair and soft parts preserved. Dating of the Jehol Group of China has been contentious, with early suggestiuons of a Late Jurassic age for some or all of the fossiliferous beds. Biostratigraphic evidence now confirms an Early Cretaceous (Barremian) age, with several radiometric dates, using different techniques, on 3 tuff layers that occur among the fossil beds of 124.6 MYA ± 0.1 Myr, 125.0 MYA ± 0.18 Myr, 125.2 MYA ± 0.9 Myr (Zhou et al. 2003). This gives an encompassing age designation of 125.0 MYA ± 0.7 Myr for the span of the 3 tuff layers, and for the fossiliferous beds of the Yixian Formation, based on direct dating. Thus, we conclude a minimum constraint of 124.3 MYA.
The soft maximum constraint is set by older fossiliferous beds with fossil mammals, but not placentals or marsupials, or members of the stem groups of either clade. For example, an older therian, neither marsupial nor placental, is Vincelestes from the La Amarga Formation of Argentina, dated as Hauterivian (136.4 MYA ± 2.0 Myr to 130.0 MYA ± 1.5 Myr. Thus, our soft maximum constraint is 138.4 MYA. Beds of similar age in North America and Europe have also produced such basal therians that are neither marsupials nor placentals according to present evidence.
Platypus–Opossum, Kangaroo, Human, Chimp, Macaque, Mouse, Rat, Rabbit, Dog, Cat, Horse, Pig, Sheep, Cow, Armadillo, Tenrec, Elephant
The base of the crown clade of modern mammals, marking the split between Monotremata, represented by the platypus, and Theria, represented by the human, might have a number of positions, depending on how many of the extinct Mesozoic mammal groups are included in the clade.
As noted above, the oldest marsupial, Sinodelphys, and the oldest placental, Eomaia, take the age of Theria back to about 125 MYA. Vincelestes from the La Amarga Formation of Argentina, as noted above, is dated as Hauterivian, and takes the age of Theria back to 136.4 MYA ± 2.0 Myr to 130.0 MYA ± 1.5 Myr.
According to a widely accepted cladogram of Mesozoic mammals (Luo et al. 2002, 2003; Kielan-Jaworowska et al. 2005), the Theria are part of a larger clade Theriimorpha that includes further extinct clades: Triconodonta, Multituberculata, Symmetrodonta, and Dryolestoidea. Most of these originated in the Late Jurassic, but triconodonts and dryolestoids began earlier, in the Middle Jurassic. Basal triconodonts include Amphilestes and Phascolotherium from the Stonesfield Slate, referred to the Procerites progracilis zone of the lower part of the middle Bathonian stage on the basis of ammonites (Boneham and Wyatt 1993), and so dated as 166.9 to 166.5 MYA ± 4.0 Myr (Gradstein et al. 2004). Tooth-based mammal taxa from the Early Jurassic of India (Kotatherium, Nakundon) and North America (Amphidon) that have been ascribed to Symmetrodonta (e.g., Asher et al. 2005) are not convincingly members of the clade (Averianov 2002), and so are ignored here. The oldest dryolestoid appears to be Amphitherium, also from the Stonesfield Slate.
The oldest monotremes are Steropodon and Kollikodon from the Griman Creek Formation, Lightning Ridge, South Australia, and dated as middle to late Albian, 109–100 MYA. Teinolophos is from the Wonthaggi Formation, Flat Rocks, Victoria, and is dated as early Aptian, 125–121 MYA.
In the new cladistic view (Luo et al. 2002, 2003; Kielan-Jaworowska et al. 2005), the Ausktribosphenida from Gondwana are immediate sister group of Monotremata, forming together the Australosphenida. Oldest are Asfaltomylos from the late Middle Jurassic (Callovian) Cañadon Asfalto Formation of Cerro Condor, Argentina (Rauhut et al. 2002), and Ambondro from the upper part of the Isalo "Group" (Middle Jurassic, Bathonian) of Madagascar (Flynn et al. 1999). The position of the Madagascar find in the Bathonian is uncertain, so the age range is 167.7 MYA ±