Leporidae

In our Leporidae tree (Fig 2), our small Leporidae clade (93 BP), containing the genera Nesolagus, Poelagus and Pronolagus, was found to be the sister of a larger Leporidae clade (45 species, 92 BP) that included the genera Caprolagus, Lepus, Pentalagus, Romerolagus and Sylvilagus (see also Cano-Sanchéz et al. 2022) (Fig 2). In the small clade, the monotypic Poelagus species joined (99 BP) the four-species cluster of Pronolagus (4 species 100 BP), in which P. crassicaudatus was a sister to P. randensis (94 BP) and P. rupestres clustered with P. saundersiae (100 BP). This lineage was a sister (100 BP) to the Nesolagus cluster (100 BP), which included N. netscheri plus N. timminsi.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Phylogenetic relationships and divergence times for Leporidae clade (pruned from Fig 1).

https://doi.org/10.1371/journal.pone.0307380.g002

In the large Leporidae clade (Fig 2), the monotypic genus Romerolagus (R. diazi) was a sister to a highly supported group (93 BP) with Caprolagus, Lepus, Pentalagus and Sylvilagus. In sequence, the split was between the large genus Lepus (30 species, 100 BP) and the diversity that included the Sylvilagus clade (10 species, 100 BP) plus the group (2 species, 99 BP) with the monotypic Pentalagus and Caprolagus.

According to our tree (Fig 2), the Sylvilagus cluster is monophyletic (100 BP) and divided into two lineages. The smaller lineage (89 BP) included a two-pair set of species: S. aquaticus plus S. palustris (100 BP) and S. audubonii plus S. nuttallii (100 BP). In the larger Sylvilagus lineage (100 BP), S. brasiliensis joined S. dicei (100 BP), and they clustered with another set (83 BP) of two pairs: S. floridanus plus S. robustus (100 BP) and S. obscurus and S. transitionalis (100 BP).

Our Lepus tree included 30 species (Fig 2), approximately twice as many species as the most comprehensive molecular phylogenetic work to date [23]. In this lineage, the species Lepus sinensis was a sister of the remaining Lepus diversity (29 species, 66 BP), but support was low. In the main Lepus group, the split separates the large clade (21 species, 64 BP), which includes L. arcticus, L. capensis, L. castroviejoi, L. comus, L. coreanus, L. corsicanus, L. europaeus, L. granatensis, L. habessinicus, L. hainanus, L. mandshuricus, L. nigricollis, L. oiostolus, L. othus, L. peguensis, L. sinensis, L. tibetanus, L. timidus, L. tolai, L. townsendii and L. yarkandensis from the smaller clade (nine species, 53 BP) with L. alleni, L. americanus, L. brachyurus, L. californicus, L. callotis, L. flavigularis, L. microtis, L. saxatillis and L. schlumbergeri.

In the small Lepus clade (Fig 2), the first split separated L. brachyurus from the remaining species into two lineages. In the first lineage (99 BP), L. americanus was a sister of the group (100 BP) joining two clades: one is L. californicus plus L. alleni (97 BP), and the other is L. callotis plus L. flavigularis (50 BP). In the second (93 BP), L. schlumergeri with L. microtis (100 BP) was a sister to L. saxatilis.

In the large Lepus clade, two sister lineages are shown. In the first (53 BP), L. arcticus joined (83 BP) L. castroviejoi plus L. corsicanus (100 BP). This species was a sister to L. othus (98 BP), followed by L. timidus (68 BP) and L. townsendii (97 BP). This group was a sister to L. capensis and the L. mandshuricus and L. coreanos clade (100 BP). L. tolai was a sister of this nine-species group (89 BP), and the clade (96 BP) of L. yarkandensis and L. tibetanus diverged next. This diversity was like that of a group (82 BP) containing two clusters of three species each. In the first (86 BP), L. granatensis was a sister of the L. europaeus plus L. habessinicus clade (100 BP), and in the second (100 BP) clade, L. nigricollis was a sister of the L. hainanus and L. peguensis clade (98 BP, Fig 2).

Ochotonidae

Ochotona traditionally has been divided into four subgenera (Conothoa, Lagotona, Ochotona, and Pika) that are distinguished by the type of habitat or geographical region they occupy [11, 48]. A fifth subgenus, Alienauroa, was recently proposed based on genomic data [49, 50], and in our study, Ochotona syrinx was the sole representative of this subgenus. Lagomorphs of the subgenus Pika are ochotonids with a North American distribution, and the subgenus Ochotona includes species that live in steppes and shrub environments, whereas, in the subgenus Conothoa, we find species that inhabit mountains. The subgenus Lagotona was proposed to include a single species, O. pusilla [51].

As previously proposed [11], the subgenera Conothoa (100 BP) and Pika (100 BP) are both monophyletic according to our phylogeny. In our tree, Ochotona (Alienauroa) syrinx was loosely clustered (85 BP) with the clade (87 BP) that included Ochotona (Lagotona) pusilla and the species assigned to the subgenera Pika and Ochotona. O. iliensis, O. koslowi and O. vizier were not included in previous subgenus assignments but were tightly grouped (100 BP) in the Conothoa subgenus, whereas O. cansus and O. sikimaria were clustered (100 BP) with high bootstrap support in the Ochotona subgenus (Fig 1).

The Pika subgenus (100/100) includes North American ochotonids and is arranged as follows in our tree: The first divergence observed is between O. collaris + O. princeps (100 BP), followed by the divergence of O. alpina plus the O. turuchanensis clade (100 BP), O. argentata plus O. pallasi (100 BP), which was the sister of the remaining four species (98 BP), and O. hoffmanni as a sister of the polytomy (98 BP), O. hyperborea plus O. mantchurica plus O. coreana.

The shrub and steppe species are the ochotonids of the Ochotona clade (100 BP). In the present work, O. thomasi was recovered as a sister to the remaining subgenus (100 BP), followed sequentially by O. dauurica (99 BP), O. thibetana (99 BP), O. cansus (99 BP) and O. sikimaria, which is the sister of the final O. curzoniae + O. nubrica (100 BP) clade.

The species that inhabit the mountains make up the Conothoa clade (100 BP). In this group, the first split was between the O. erythrotis and O. forresti clades (93 bp), and the remaining diversity (100 bp) of the subgenus was observed. The next split included the O. vizier and O. rufescens clades (100 BP) and the remaining Conothoa diversity (80 BP). This diversity was divided into two lineages. The first (BP 89) O. ladecensis was a sister to O. koslowi plus O. iliensis (91 BP), whereas in the second (89 BP), O. rutila was a sister to O. roylii and the O. macrotis clade (89 BP).

Lagomorpha time-resolved phylogenetic tree

The Lagomorphs first differentiated during the Palaeogene, while the Leporidae and Ochotonidae families started to differentiate in the upper Palaeogene and middle Neogene, respectively (Fig 1). The origin of Lepus dates from the Pliocene, while the age of Sylvilagus dates from the late Miocene. The internal consistency test of the calibrations detected differences between the time estimates when a specific calibration point was absent. However, in most calibration schemes, the ages were congruent. When there was a difference, usually 10% difference, the confidence intervals overlapped. Importantly, the age estimates for Lagomorpha and Leporidae remained stable when their specific calibration points were excluded. However, when the Rodentia calibration was excluded, the age of Lagomorpha recovered to the lower Eocene. Leporidae, on the other hand, originated in the upper Oligocene in every calibration scheme, while Ochotonidae was repeatedly found in the transition between the lower and middle Miocene.

The time tree with all calibration fossils indicated that the beginning of the crown group of lagomorphs was 57.2 million years ago (57.2–52.9 Ma), indicating that it appeared in the Thanetian Age, late Palaeocene, Palaeogene period (Table 2, Fig 1). The estimate by Ge et al. [23], which calibrated only the Lagomorpha node of the ingroup and with a single point in time (50.2 Ma), is younger (50.3 Ma) but consistent with our estimate. Notably, when the calibration for this node was not used in our study, the origin estimate was 62.52 Ma, increasing this difference slightly and with a wider confidence interval (74.8–52.2 Ma). The age proposed by Matthee et al. [20] of 29.0 (±3.8) Ma for Lagomorpha is considerably younger than our age, as these authors calibrated their tree with much younger ages (20–40 Ma) for the nodes. Nevertheless, their calibrations do not seem to correspond to the most recent fossil evidence, which indicates the existence of stem Leporidae and Ochotonidae since the Eocene [51, 52]. In addition, Matthee et al. [20] used a calibration for Lepus based on divergence time estimates and not on fossil occurrence. Cano-Sanchéz et al. [53] reported a Palaeocene age of ca. 60 Ma for Lagomorpha.

Among the two extant Lagomorpha clades, Leporidae was the first to diversify; it was dated to 25.4 Ma, the Chattian age, and the upper Oligocene. Ge et al. [23] found an age of approximately 18.1 Ma in the lower Miocene. Halanych and Robinson [54] used the average sequence distance between leporid genera and assumed a molecular clock to calculate the Leporidae divergence time; a slightly older age of approximately 12.2 to 16.3 Ma was found for this family. Similarly, Matthee et al. [20] reported an age of 14 (±1.5) Ma for this node. An even older age of 21.9 (± 3.8) Ma was found by Su and Nei [55] using Ig heavy chain variable region genes. The oldest estimated age for Leporidae was that proposed by Cano-Sanchéz et al. [53], with a divergence occurring ca. 47 Ma in the middle Palaeogene.

Ochotonidae, on the other hand, was recovered at 15.4 Ma in the Langhian, middle Miocene. Melo-Ferreira et al. [16] calibrated only the Lagomorpha node, testing three different ages, 31, 37 and 65 Ma, and found that the ages of Ochotonidae were 6.6, 7.8 and 13.8 Ma, respectively. Notably, no calibration points overlapped our prior distribution for this node (48–59 Ma). When Melo-Ferreira et al. [16] employed the oldest calibration (65 Ma), the age recovered was 13.8 Ma, the closest to the age estimated here. Similarly, Mohammadi et al. [54] used four calibration schemes to calibrate the Lagomorpha node, adding a 52 Ma calibration point to those previously used by Melo-Ferreira et al. [16]. The youngest age obtained was 5.7 million years ago, while the oldest was 12.5 Ma. Ge et al. [23] reported an age of 13.4 Ma for Ochotonidae, which is also the middle Miocene.

There is a clear pattern in the rich fossil record of lagomorphs; after the divergence of the Leporidae and Ochotonidae, the two families diverged quickly, producing a few stem lineages during the Eocene and the Oligocene [23, 51, 52]. The Leporidae family appeared in the fossil record at approximately 37 Ma, during the upper Eocene period, followed by Ochotonidae at approximately 33 Ma, during the Oligocene period [51]. At the Oligocene and Miocene boundaries, distinct patterns emerged in each lineage. The Ochotonidae ancestor experienced a burst of diversification at the earliest Miocene, centred in Asia, followed by range expansions to Europe, North America and Africa [51, 56]. Between the middle and late Miocene, a second wave of ochotonids that included the extant genus Ochotona replaced several primitive genera in Asia and other continents [51]. In contrast, the Leporidae diversity remained relatively low throughout the Miocene but radiated profusely by the latest Miocene [23, 51, 52]. Leporids have also expanded their ancestral range from Asia to Europe, Africa and North America since the late Miocene [23, 51]. Therefore, the rich Lagomorpha fossil record clearly indicates that Leporidae is older than Ochotonidae, but that the major diversification of Ochotonidae predated that of Leporidae.

However, such a clear-cut pattern supported by a rich fossil record on multiple continents does not fit the current hypothesis for lagomorph diversification [23]. For instance, in that study, the diversification of Leporidae predated that of the extant Ochotona radiation. Although that study dated the Ochotona clade to ~13 Ma, a fossil of an undetermined Ochotona species from the Amuwusu Micromammals Site (Miocene of China) extends the age of the genus to ~16 Ma [57]. Since fossil age uncertainty and internal calibrations were not incorporated in the Ge et al. [23] time tree, such a recent age for Ochotona might have been an artefact. In our time tree, we incorporated multiple calibration points and fossil age uncertainties as probabilistic distributions. Our age for the Ochotona ancestor was estimated to be 15.4 Ma, which better fits the minimum age for the genus in the fossil record, especially because the Ochotona fossil from the Amuwusu site was not included in our calibration set. Last, in the Ge and coworkers time tree, the diversification of crown Leporidae was dated to approximately 18 Ma. Since stem representatives of Leporidae first appeared 37 Ma, that scenario requires a 19 million-year [Myr] gap in the fossil record for stem Leporidae, which is otherwise very rich on several continents. Our age for crown Leporidae was estimated at 25.4 Ma with a small confidence interval, thereby reducing the fossil gap to only 12 Myr (Fig 2). Therefore, our time tree better fits the fossil record for both crown clades of Lagomorpha.

In the upper Neogene, adaptive radiation started in the lineage of the genus Lepus and the lineage of Sylvilagus. The large genus Lepus originated at approximately 4.5 Ma during the Zanclean stage. Matthee et al. [20] reported the origin of this genus in Zanclean, with an age of 5.2 (±0.5) Ma. Ge et al. [23], on the other hand, recovered an age of approximately 8.6 million years ago, shifting the origin of Lepus to the Miocene epoch. The most recent estimated age for Lepus was that proposed by Cano-Sanchéz et al. [53], with the most recent common ancestor dating to 2.5 Ma (95% CI: 1–4). Nesolagus diverged from Pronolagus and Poelagus by 18.4 Ma, and Pronolagus and Poelagus diverged 15 Ma in the middle Miocene. The beginning of the diversification of the clade comprising Pentalagus, Caprolagus and Sylvilagus began at approximately 14.6 Ma, with Sylvilagus separating it from the other genera at 7.8 Ma. The divergence of this clade from Sylvilagus occurred at the beginning of the Tortonian Age, with Sylvilagus appearing at approximately 6.5 Ma in the Messinian Age in the Miocene, in contrast with the age of 5.3 (±0.8) Ma reported by Matthee et al. (2004). The latter is much closer to the age recovered by Ge et al. (2013), with a Sylvilagus origin at 5.6 Ma. In addition, Cano-Sanchéz et al. [53] reported Sylvilagus as a paraphyletic taxon with respect to Brachylagus, and the age of this clade was ca. 1.5 Ma. The origin of the genus Ochotona was approximately 15.4 Ma, and its three subgenera had ages of 12.2 Ma (Conothoa), 11.8 Ma (Pika) and 6.1 Ma (Ochotona).

Among the seven time-resolved phylogenetic trees reconstructed in this study, there were some differences between the one that included all six fossils and the one in which one of the fossils was absent for the first lagomorph divergence (Table 2, Fig 1). This finding demonstrates the importance of including crucial calibration points. However, the congruence of most ages in the different calibration sets shows the robustness of the divergence time results of this part of the time tree.

The divergence time for crown Lagomorpha was consistent, approximately 58 million years ago, in almost all the sets, with only one exception. In the set without the Rodentia calibration point, there was a slight difference, with a younger age of 53 Ma. In the set without the calibration point of Lagomorpha itself, the oldest age is observed at approximately 62.5 Ma. Interestingly, the Leporidae family was consistent across all datasets, ranging from 23.7 to 27.4 Ma, and always recovered in the Chattian age. For Lagomorpha, when the Rodentia calibration was absent, the youngest age of 23.7 Ma was found. Ochotonidae showed the most stable age throughout the sets, at approximately 15.7 Ma, at the Langhian–Burdigalian boundary. The age of the Lepus genus was remarkably constant at 4.7 million years ago, always at the Zanclean age. The genus Sylvilagus, with a constant age of 6.7 Ma, differed in the set without the stem Sylvilagus calibration, with an age of 10.3 Ma.

Mammal and plant radiation

The current macroevolutionary explanation for the distinct diversification patterns in the lagomorph fossil record is also tentative. In the late Miocene, low atmospheric CO₂ concentrations, seasonal fires and aridification fuelled grassland expansion worldwide, coinciding with the extinction of most ochotonids while leporids radiated [23]. That study suggested a direct link between C4 plant diversification, which predominates in the diet of leporids, and a decrease in the diversity of C3 plants, which are preferred by ochotonids (Fig 3). However, although C4 grasses are a major component of plant biomass in grasslands, recent studies have shown that many C3 grasses (e.g., Pooideae) also exhibited great diversity in the late Miocene [58]. Furthermore, other C3 plant families, such as daisies (Asteraceae), are often as diverse as (or more diverse than) grasses in grassland biomes [59]. Analyses of diversification rates and environmental variables have demonstrated that although grasses strongly differentiated during the late Miocene (~10 Ma), the same process was mirrored by the hyperdiverse Asteraceae [59]. Most importantly, surveys of food preference revealed Asteraceae as the dominant food preference in extant pikas [23, 60]. Additionally, the great diversity of Ochotonidae in the fossil record of the early Miocene closely matches the peak in diversity of Asteraceae within the same time frame ([59], Fig 3). Therefore, food preference alone does not explain the correlation between grassland expansion and the diversification–extinction patterns observed in Lagomorpha.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Diversification scenario of Lagomorpha and grasslands during the late Cenozoic.

https://doi.org/10.1371/journal.pone.0307380.g003

As most studies in Lagomorpha have focused on comparisons between its internal lineages, the macroevolutionary relationships between lagomorphs and other small mammals remain to be explored. The species-rich rodents (order Rodentia) are the leading model for small mammals due to their sheer morphological, ecological and geographic diversity [61–63]. Most importantly, they are the sister group of Lagomorpha that forms the clade Glires, and divergence time estimates based on nuclear and mitochondrial data suggest that the burst of adaptive radiation in rodents occurred in the last 10 Ma [64]. As Rodentia and Lagomorpha share the characteristic of small body sizes and both have a pair of continually growing chisel-like incisors [61, 63], we suggest that such similarities in form and function are strong indicators of interspecific competition.

Therefore, to explain the diversity trends in Lagomorpha, we suggest a key biotic component that hinges on competition for resources with rodents. Fossils of stem Glires (e.g., Mimotonidae and Eurymylidae) indicate that small, rat-sized bodies and limb proportions are likely ancestral morphologies [61, 63, 65]. Palaeontological and molecular data agree that by the late Miocene, the Muroidea ancestor originated from the most diverse Rodentia families, Muridae and Cricetidae, which spread from Asia to other continents while consuming a variety of diets, from generalist omnivore diets (Muridae) to the strictly herbivorous diets consumed by Arvicolinae (Cricetidae) [62]. When this diversification pattern is compared with our chronological tree, we observed that the tempo and locale of the muroid radiation coincided with the drastic reduction in Ochotonidae genera in the fossil record (Fig 3).

Given that the two lineages share the ancestral Glires body plan and geographic ranges, we suggest that direct competition between members of the generalist and diverse Muroidea steered the mass extinction of the herbivorous Ochotonidae in the late Miocene. A recent study provided empirical evidence that rodent and pika populations interact enough to allow parasite spillover in North America [66]. An increase in rodent diversity has also been linked to extinction by competition with Multituberculata, a lineage closely related to the eutherian clade that disappeared when the major rodent lineages appeared in the fossil record worldwide [61, 63].

According to our hypothesis, while pikas competed for resources with muroid rodents of similar body sizes, the Leporidae ancestor would have escaped the turmoil of that competition by evolving a strikingly different body plan. Analyses of the complete fossil record of Glires in North America revealed a rise in Leporidae specialized in cursoriality (adapted to run) during the late Miocene [67]. Although Rodentia and Lagomorpha evolved burrowing habits and high crowned teeth (hypsodonty) suited to abrasive food such as that containing silica and dirt present in grasslands in the Miocene, only Leporidae showed increased cursorial specialization.

Leporidae species have overall larger body sizes than most Glires species [23, 68], and pikas and very few rodents exhibit true cursorial specializations due to their generally small size [68, 69]. Leporids, however, exhibit several unique adaptations to cursoriality, namely, a unique intracranial joint at the occipital region that alleviates spinal rebound during hopping [69] and extensive maxillary fenestration that provides mechanical resistance while reducing bone mass [70]. Although pikas exhibit some degree of maxillary fenestration, this characteristic became much more pronounced in the late Miocene Leporidae [69]. Finally, in the postcranial skeleton, the late Miocene Leporidae evolved elongated hindlimbs and metatarsals [68]. Therefore, the rapid diversification of the late Miocene Leporidae was marked by key innovations, such as powerful hindlimbs and cranium that withstand the mechanical pressures of high-speed locomotion. These are unique features of this lineage that, within our time tree, explained at least in part why leporids radiated and are more species-rich than their sister group.

The diversity trends in Lagomorpha might be related to a competition for resources with rodents. The decline in ochotonid diversity in the Late Miocene coincides with the radiation of the most diverse rodent families, Cricetidae and Muridae, which may have outcompeted the pikas particularly as the strictly herbivorous Arvocolinae murids diversified. Our findings are supported by the fossil evidence that indicate the ochotonid-rodent small body size as ancestral in Glires.

We suggest that Leporidae escaped the competition bottleneck and radiated by evolving a completely new morphospace—large bodies and an efficient cursorial locomotor system—covering wide areas for foraging and outrunning predators more easily than rodents and pikas. Our results indicate that since the Pleistocene, rabbits and hares have radiated, especially those of the genera Lepus and Sylvilagus, which tended to remain taller than most rodents and pikas. Nevertheless, successful leporid body plans have been constrained by larger body sizes, as this would move them into competitive morphospace with ungulate-type herbivores [71].