Open Access

Research Article

• Download Article XML
• Download Article PDF
• Download Citation
• E-mail this Article
• Order Reprints
• Print this Article

Introduction

At the dawn of modern genetics, William Bateson's [1] vision of the new field he had named led him to follow his exploration of Mendel with an exploration of traits underlying serially repeating elements in biology. For ninety years however, his definition of “homeotic” variation along the body axis led to little or no academic interest while the broader field he coined as “genetics” grew to dominate biology.

Among the questions that Bateson sought to address by studying homeotics was the way in which genetic change could lead to the emergence of new body plans. Neither classical morphology nor standard Darwinian analysis has provided truly satisfying explanations of such major body plan innovations as the origin of the Bilaterians by symmetric right/left replication of the organism or the origin of the vertebrates by body axis inversion of the original Bilaterian design [2]. These appear to be abrupt massively pleiotropic [3], [4] consequences of single or small number gene changes that have little to do with gradual shifts in population gene frequencies under drive from natural selection.

The discovery of the homeobox in the 1970s [5], [6], [7] and the subsequent growth of interest in developmental genetics [8], [9], [10], [11], [12], [13], [14], [15], [16] has led to a revolution in evolutionary biology. There is a new understanding of terminal addition and the emergence of a wide variety of genetic mechanisms of segmentation in the Bilateria [17], [18], [19], [20]. The recent identification of extensive similarities in the genes mediating the mechanisms of segment formation in the embryos of spiders and vertebrates [21] has further revealed the ancient nature of body axis segmental morphogenesis.

It is now reasonable to return to Bateson's project. Evolutionary change in the system of homeotic genes seems to be involved in body plan transformation. Modularity theory [22], [23] and a reexamination of mutationism in the light of modern morphogenetics [24], have opened the door to a major revision of evolutionary theory to accommodate this new understanding of body plan innovation.

Can the study of homeotic change help show how morphogenetic evolution relates to the emergence of new body plans [25], [26], [27], [28]? Do similar considerations apply to the more modest alterations in “body configuration” as it may apply to changes at the level of infraclass, order and family within the Mammalia? The advance of comparative genomics has accelerated our understanding of the way in which duplications of genes play a critical role in evolution [29], [30]. When a gene is present in a second copy, evolutionary constraints are relaxed–one copy may be altered without depriving the organism of the existing effects of the original gene. It has not been clear whether morphologies display similar patterns of change. If morphologies do evolve in this fashion, are the effects of these changes of minor or major theoretical, systematic and biological importance?

This report examines the question of whether duplications and homeotic changes have played a role in new body configuration change in three events of special biological interest-the emergence of mammals among the synapsid amniotes, the diversification of mammal groups in the Late Cretaceous, and the emergence of “hominiforms” among the catarrhine primates in the Early Miocene.

The study of axially arrayed serial homeotic characters in a group such as the mammals necessitates the study of vertebrae. This is a topic that has been relegated to limited sub-specialist and medical interest for more than 150 years. However, before Darwin, many of the major attempts to assemble a biological explanation for similarity among animals involved vertebrae explicitly. Most prominently, the widely attended zoological works of Goethe [31], [32], Geoffroy [33], [34], [35], and Owen [36] represented spinal repetition series as central to understanding biology. Recently, our new understanding of morphogenetics has triggered a new interest in this complex anatomical arena [37], [38], [39], [40], [41]. Still, the published literature on the evolutionary biology of mammalian axial structures is remarkably sparse.

In addition to the progress of axial skeletal fossil discoveries, the remarkable advances in our understanding of the embryologic development of axial structures and their relationships to Hox, Pax and other Bilaterian homeotic and morphogenetic gene families have further increased the relevance of attention to evolution of axial structures [39], [40], [42]. As we explore the hominoid genome [43], [44], we need careful analysis on where to look among the thousands of genetic differences among these species [30] to best identify critical events in the genetic genesis of human form. There is tantalizing evidence that the major changes that distinguish human vertebrae from those of Old World monkeys follow a pattern that may leave a distinct and identifiable trace in the genome.

The hominiform example is particularly compelling. Proconsulid hominoids differed from old world monkeys in having a Y-5 pattern of molar cusps but were otherwise similar to them in body form and ecological niche–most appear to have been generalized quadrupeds [45], [46], [47], [48]. A significant subsequent homeotic transformation is correlated with the emergence of novel upright (orthograde) locomotor patterns in a new hominiform clade. That makes this clade particularly interesting as a biological transformation [37], [38], [39] in addition to its importance in understanding the relationship of homeotic change to human origins.

For most of the past two hundred years, models of the origin of human upright posture and bipedalism have been based primarily on evidence from the appendicular and cranial skeleton, but evidence from the spine has played little or no role in our understanding. A series of discoveries of axial skeletal fossils from species including Morotopithecus bishopi [47], [49], Proconsul nyanzae [45], Oreopithecus bambolii [50], [51] and Pierolapithecus catalaunicus [52] have now provided evidence that is remarkably inconsistent with models that have not considered axial structures in understanding posture.

Given the many unique aspects of load bearing and movement requirements, it is not at all surprising that the lumbar vertebrae of modern humans are strikingly different in structure and function from typical mammalian vertebrae. However, the appearance of most of the unique features of the Homo sapiens lumbar vertebra in UMP 67-28, a hominoid fossil from 21.6 million years ago [37], [47], [49] is very surprising. This is particularly true since the apes of the Early and Middle Miocene have been generally considered to have few or none of the modifications of body plan that characterize modern apes and humans.

For a variety of reasons, the term “human” has been applied to a clade of hominoids commencing at the split from the chimpanzee lineage about six million years ago [53]. The basis for this distinction has been upright bipedalism exclusively in the human lineage. However, when the evidence from serial axial structures and homeotic events are considered, the anatomical basis for upright posture and bipedalism appears to have arisen far earlier–it is the axial anatomy first seen in Morotopithecus. Upright bipedalism plays a significant role in all the species of a clade that share the morphogenetic transformation with Morotopithecus.

The significance of the anatomical adaptations to upright posture and varying degrees of bipedalism seem among the hominoids has been a matter of ongoing interest [54], [55] [56]. Nonetheless, it has been widely accepted that specialization for full time primary bipedal locomotion did not occur in the direct human lineage until the split from chimpanzees had taken place about six million years ago.

However, when the various components of axial anatomical specialization in hominoids are fully identified, and their context in the broader setting of mammalian homeotic evolution is made clear, an alternate sequence of events becomes increasingly compelling. This is the possibility that a distinct and ancient clade within the hominoids can be identified that share a major modification of axial architecture that underlies the upright posture and primary bipedalism of modern humans. This morph appears to persist across the succeeding 21 million yeas to be preserved in primitive form in modern humans. The various other types of specialized locomotion seen among existing hominoids are made possible by comparatively minor secondary and tertiary modifications of the original primitive upright, bipedal architecture. This is the basis for asserting a homeotic transformation is the basis of the origin of humanity.

Results and Discussion

General Patterns of Homeotic Change in the Mammalia

This study revealed that body configuration modification in the Mammalia often involves emergence and change of homeotic gradients. In a number of instances clusters of multiple different homeotic gradient changes occurred at the stem of a major systematic radiation (Figure 1).

Figure 1. Systematic and temporal distribution of homeotic character transitions in Mammalian groups.

Divergence data after Springer et al[58], Flynn et al[90], Kielan-Jaworowska et al[76]. K-T-Cretaceous-Tertiary boundary.

doi:10.1371/journal.pone.0001019.g001

These clusters of homeotic change generally qualify as body plan changes and often relate to significant alterations in the adaptive zone of the descendant groups. These clusters of changes are often preserved as a fixed homeotic set in the descendant group across tens of millions or hundreds of millions of years.

Within individual lineages many of the gradients demonstrate alterations on a sporadic basis (at the level of species or higher level clades). Some lineages (e.g. hominiform hominoids, pilosan xenarthrans) show a very high frequency of homeotic change for some gradients. Other lineages show little or no homeotic change over hundreds of millions of years (Monotremata).

Some homeotic alterations appear to be relatively highly conserved–they fluctuate in their expression among more ancient lineages but eventually become fixed (e.g. lumbar rib suppression). A few homeotic features never change after their initial appearance (e.g. emergence of the laminapophyis, septo-neural approximation).

At a finer level, some gradients clearly are subject to independent alteration in rate and tempo of expression along the body axis–some progress incrementally along the segmental series, some commence abruptly and then progress slowly and these properties vary across taxa. The gradients may respect medio-lateral and dorso-ventral positional relationships relative to each other or they may cross as they progress down the body axis. The segmental locations of onset of gradient change do not follow rigidly fixed sequences relative to each other.

Once established, the expression pattern of these gradients and of the morphological substrates upon which the gradients act then diversify (Figure 1). Some appear to have major functional impacts on the organism, others may have become fixed (uniformly present in descendant lineages) solely due to morphogenetic constraints.

One remarkable aspect is the mirroring and duplication of homeotic gradients. A gradient series usually seen with a given polarity and location recurs with opposite polarity at a different location. New gradients may act along the entire body axis or in replicated form within each segment. The emergence of new types of structures by duplication with subsequent diversification of the new version mimics the pattern of change often seen with gene duplication at the level of the genome.

Segment Identity–the Primary Gradient

The basic homeotic distinction of five major spinal regions (Table 1) is apparent in the earliest land vertebrates [57] and can be assessed by boundary transitions. Seven cervical segments are standard and readily identifiable in mammals and seven to nine in most amniotes with the prominent exception of the extensive duplication and alteration of the cervico-thoracic region at the emergence of the avian winged archosaurs (birds). A very small number of mammalian species have alteration in cervical vertebral numbers on a sporadic basis.

Table 1. Primary Gradient–Segmental Identity and Boundaries

doi:10.1371/journal.pone.0001019.t001

The thoraco-lumbar transition within the vertebral series of mammals, however, depends on a variety of gradients that defy simple counting and categorization (Table 1)–this issue is explored in detail below. The components of this transition are stably arrayed in some higher taxa but subject to frequent generation of new versions in others (Figure 2).

Figure 2. Thoracic and lumbar segmental homeotic trait patterns in mammalian species.

doi:10.1371/journal.pone.0001019.g002

The lumbo-sacral boundary collectively affects multiple gradients in concert and is therefore a discreet phenomenon like the cervico-thoracic boundary. The recent advent of a molecular resolution to the deep relationship of mammalian groups [58], [59] provides an opportunity for observing phylogenetic patterns in the segmental position of the lumbar/sacral boundary. Some groups are very stable for this boundary position, some demonstrate occasional small shifts, others are quite unstable with either significant increases or decreases in number of segments (Figure 2). There are a few species with highly unusual thoraco-lumbar or lumbo-sacral boundary effects.

Scutisorex provides the most dramatic example of morphogenetic disruption of the homeotic system among the mammals [37] having scores or hundreds of facet pairs and a seeming duplication of the entire lumbar region. Although most mammals–including the numerous other species of the Soricidae-have six or seven lumbar vertebrae, Scutisorex has twelve lumbar vertebrae.

Another informative homeotic character state is the replication of the “diaphragmatic” thoraco-lumbar transition vertebra in a specimen of the macroscelid Petrodromus tetradactylus (USNM 241593)–a species with a remarkably accelerated rate of morphological evolution [60]. There is an elongated lamina with a double neural spine. The more posterior “third” half of the lamina replicates the anatomy of the last pre-diaphragmatic vertebra. This represents discontinuous homeotic change and shows that the joint surface reorientation seen in the diaphragmatic vertebra is indeed a homeotically determined aspect of serial morphology.

Reduction in the number of dorsal (thoracic+lumbar) segments is relatively uncommon. It is typical of the Order Chiroptera and the Order Cingulata. Among hominoids this occurs in all of the species of the hominiform clade (Figure 3, 4) but not among the proconsulid hominoids. Some proconsulids may have tail loss without reduction of dorsal segment numbers [61], [62], [63] but full details of the sequence of these events remains unclear.

Figure 3. Homeotic shifts in the catarrhines.

The data show the average segmental midpoint of nerve and plexus origins relative to vertebral segment regionalization (after Filler 1993 [38], some data from Keith 1902 [91]).

doi:10.1371/journal.pone.0001019.g003

Figure 4. Secondary sacral boundary shifts within the hominiform clade.

(A) Humans appear to retain the original hominiform longer flexible lumbar region. (B) Anatomical reconfiguration results in effective elimination of the lumbar region in Gorilla. (C) Despite a modal number of 5 lumbars, humans may have 4 lumbar but maintain a long flexible lumbar region. (D) Molecular phylogeny suggests that lumbar region shortening in Pan occurred independently and convergently (X-ray in D after Filler 1979 [92]).

doi:10.1371/journal.pone.0001019.g004

The initial reduction in number of lumbar vertebrae in the hominiforms appears to be a shift from the catarrhine modal number of seven down to a modal number of five or six (Figure 3). Modern humans typically have five lumbar vertebrae, the only known complete australopithecine lumbar spine has six [38], [64], [65].

Reduction to a modal number of four lumbar segments may have occurred separately in Pongo, then Gorilla, and then Pan, with the longer more flexible lumbar spine retained in primitive form in hominines such as Australopithecus and Homo (Figure 1,4). Alternately, the entire “great hominiform” group shared a single common secondary event causing reduction to four lumbars, but hominines subsequently reversed the trend to regain the modal fifth lumbar segment [39]. This may be consistent with the presence of upright bipedalism in the stem hominiforms, that is transformed to diagonograde postures in the common ancestor of great apes and humans, followed by rapid re-establishment of bipedalism early in the course of an independent hominine lineage.

However, as explored below, the secondary reductions of the lumbar region may be independent, parallel convergent adaptations to the various non-upright, “diagonograde” postures employed by the large apes. This interpretation, requiring an independent lumbar shortening in Pan after divergence from the hominines six million years ago gains some support from recent fossil evidence. Sahelanthropus-a candidate pre-split common ancestor of chimps and humans dated to seven million years ago-was very likely an upright biped [66], [67], [68], [69] There is also evidence for bipedalism in Orrorin [70] [71], [72], another hominoid dated to a period quite close to the chimp-human split. This model suggests that the upright bipedal body plan of the hominiforms arose in the Early Miocene and that since that time, there has been a continuous lineage including upright bipeds of which Homo sapiens is only the most recent species to demonstrate this primitive hominiform body plan.

Frame Shifting and Rib Suppression in the Second Gradient

The two major types of segmentally repeating structures in tetrapods are ribs and vertebrae. Among mammals however, this study showed that these represent two separately determined segmental systems that may be frame shifted relative to each other. The pattern of frame shifts strongly suggests that a separate gradient for the segmental identity of ribs had emerged before the emergence of the therian group 150 million years ago.

As in most tetrapods, the contact point of the rib with the vertebra has been duplicated in the dorso-ventral plane (Table 2). The more dorsally placed rib head and articulation seems to have its segmental identity determined by the original primary segmental gradient since it never demonstrates frame shifting (Figure 5).

Figure 5. Frame shifting between rib and vertebral segments.

Evidence for independent formation of a parallel segmental identity gradient for ribs that may differ from the vertebral gradient is demonstrated by frame shifting. The synapsid (sy) primitive condition has a principal (capitular) rib head articulating on a pararthrum on the intercentrum (ic) which seems to serve as a morphogenetic “target”. In basal therians (th), there is no intercentrum, but the rib head still articulates between the two centra (pleurocentra) as if the lost intercentral morphogenetic target were still present. The articulation is divided into a pre-pararthrum (red) on the anterior end of the following vertebra (iso-segmental) and a post-pararthrum (orange) on the posterior end of leading vertebra. In the posterior thorax of many eutherians (e.g. Euarchontoglires, the Xenarthran Order Pilosa) and some metatherians, the post-pararthral articulation is lost (post1)-“pre-pararthral dominance”-and the diarthral (blue) articulation is also suppressed in many groups (post2). However in metatherians, the Xenarthran Order Cingulata, Hippopotamidae and Cetacea, it is the pre-pararthrum that is lost-“post-pararthral dominance”-in the posterior thorax so that the capitulum articulates only with the post-pararthrum (ant1). The post-pararthrum may move away from the intervertebral space (ant2). In some groups, the diarthrum is also lost so that the rib (e.g. r9) articulates only with the leading vertebra (T8)-this is seen sporadically in the posterior thorax in myomorph, hystricomorph and anomaluromorph rodents and perissodactyls.

doi:10.1371/journal.pone.0001019.g005

Table 2. Second Gradient–Rib Head Suppression and Frame Shifting

doi:10.1371/journal.pone.0001019.t002

Monotremes (e.g. Ornithorhynchus) can have mobile ribs on all of their lumbar vertebrae. In fact, many groups of Mesozoic mammals also have mobile uniarticulate ribs on their lumbar vertebrae. It is only among the therian mammals that lumbar ribs are lost definitively [73]. Since some therian mammals demonstrate suppression of the dorsal rib head (Table 2) and others demonstrate suppression of the ventral rib head (Table 2) it is not clear whether complete suppression of the lumbar ribs in therians is due to both of these traits becoming fixed in a common ancestor or due to a separate set of changes.

The more ventrally placed rib head is the principal in-line end point of the rib and is generally considered to reflect the original primitive vertebrate rib head. This rib head, and therefore the mammalian rib itself, appear to be controlled by the new, independent secondary segmental identity gradient that may be frame shifted anterior or posterior to the primary gradient (Figure 5, Table 2).

This common class of segmental ambiguities shows that numerical segment correlation between costal and vertebral elements is not a fundamental morphogenetic principal in mammals. These represent two separate seriation systems that may shift relative to each other by as much as a full segment.

Duplications & Mirroring in the Tertiary Gradient Set

Much of the homeotic plasticity among mammals involves a number of gradients acting on a “neomorphic” or newly established structure on the dorsal (laminar) part of the vertebra that was revealed and characterized by this study. Some of these gradients appear to have profound functional significance, others seem to be best valued as windows into the morphogenetic mechanisms in play in mammalian evolution. The neomorphic element can be termed the laminapophysis. The evidence for either functional or morphogenetic significance comes from the widespread fixation of the character. It has been universally present in all mammalian species since it first appeared 220 million years ago.

The neomorph appears to arise by a medio-lateral duplication on the lamina of the vertebra. A single primitive extension or process seen in most tetrapods (the diapophysis) becomes two side by side extensions (Table 3, Figure 1, 6, 7, 8, 9, 10, 11). This effects a fundamental body configuration change in the mammalian clade.

Figure 6. Antecedents and relations of the neomorphic laminapophysis in mammals.

(A)-Configuration of diarthrum, pararthrum and intercentrum in synapsids (Ophiacodon) with the entire pararthrum (orange+red)) on the intercentrum (after Williston[80]) (A1), and diapsids[82] (Crocodylus) upper (A2) and (Alligator) lower (A3) thoracic. (B)-Muscle attachments of the laminapophysis. (C) New nomenclature of vertebral articular surfaces and processes in mammals. Blue-diarthrum, red-pre-pararthrum, orange-post-pararthrum, green-NLM.

doi:10.1371/journal.pone.0001019.g006

Figure 7. Diversity of lumbar transverse processes (LTP) serial homology and NLM morphology in therians.

(A)-There is an independent laminapophysis (NLM) in Erinaceus (Eulipotyphla) that does not split at the thoraco-lumbar transition and is unrelated to the LTP. Erinaceomorphs have no pre-pararthrum on the last ribbed vertebra (post-pararthral dominance) and have a diapophysial LTP. (B)-Typical transition from tri-articulate rib to uni-articulate rib to LTP in Superorder Euarchontoglires. Note splitting of laminapophysis (NLM) (green), loss of the diarthrum (blue), and suppression of the post-pararthrum (orange) to yield a pre-pararthral base for parapophysial LTP (red)–drawing of Macaca (Primates). (C)-Post-pararthral dominance with anterior segmental frame shift in metatherians. (C1)-Diapophysial LTP with absence of prepararthrum and no participation of the post-pararthrum (orange). The last rib articulates only on the vertebra of the preceding segment. Note that the diarthrum transposes from dorsal to the neuraxis to ventral (diarthro-neural transposition). Drawing of Thylacinus. (C2)-Thoraco-lumbar transition in Thylacinus cynocephalus (Metatheria) MCZ 36797 (photo of specimen drawn in C1).

doi:10.1371/journal.pone.0001019.g007

Figure 8. Segmental frame shifting.

(A)-Anterior shift at thoraco-lumbar transition: pararthrum entirely on preceding segment with diarthrum on iso-segment. First lumbar transverse process (LTP) (on L1) is bi-segmental (T13+L1). Transitional vertebra (T13) has no capitular rib articulation and no LTP. Macropus rufus (Metatheria) MCZ 6930. (B)-Anterior shift at cervico-thoracic transition: pararthrum entirely on preceding segment (C7) in Sotalia fluviatilis (Cetacea) FMNH 99612. (C)-Posterior shift in the thoracic region: pararthrum entirely on iso-segment and migrated dorsal to the border between the neural arch and the centrum (neuro-central suture)-these two features together are analogous to the condition in archosaurian reptiles. Myrmecophaga tridactyla (Pilosa) FMNH 49342. Oc–occipital, C-cervical, T-thoracic, L–lumbar, di–diarthrum, pa–pararthrum, nc–neuro-central suture, LTP–lumbar transverse process.

doi:10.1371/journal.pone.0001019.g008

Figure 9. Distinction of laminapophysis from diapophysis (A)–Relation of diarthrum to laminapophysis in Zaglossus (Monotremata) and Erinaceus (Eulipotyphla).

(B)–Relation of diapophysis to laminapophysis in Potamogale (Afrosoricida). (C)–Distinct diapophysis and laminapophysis in Rhizomys sumatrensis (FMNH 98534) (Rodentia). T-thoracic, L–lumbar, di–diarthrum, la–laminapophysis. Blue–diarthrum, red–pre-pararthrum, orange–post-pararthrum, green–NLM.

doi:10.1371/journal.pone.0001019.g009

Figure 10. Body configuration change in mammalian axial anatomy.

(A)-Monotonous laminapophysis in Monotremata (Tachyglossus aculeata) with no lumbar transverse process. (B)-Laminapophysis split into anteriorly directed metapophysis that slowly drifts medially to engage in sagittalization of the L4/S1 facet and posteriorly directed anapophysis. Large orthapophysial lumbar transverse processes from “third tubercle” of laminapophysial condyle on the arch (Tapirus bairdii, Perissodactyla). m–metapophysis, a–anapophysis, s–sagittalization.

doi:10.1371/journal.pone.0001019.g010

Table 3. Third Gradient–Duplications and Mirroring

doi:10.1371/journal.pone.0001019.t003

Once established it actually becomes more constant than the primitive extension that it replicates. In the posterior thoracic region of many mammals, the diapophysis is suppressed along with the dorsal rib head, but the laminapophysis still appears. It is therefore clear that its morphology is determined by a new homeotic gradient that is not necessarily subject to events that alter the old homeotic gradient responsible for the diapophysis.

The laminapophysis disassociates most of the trunk musculature from the ribs, thus significantly disengaging the rib cage from the locomotor musculature of the body (Table 3, Figure 6B). This is a critical major body configuration transformation that allows mammals to progressively increase ventilation as they run at faster speeds. It establishes a “mammaliform” clade and is in many ways a defining event in mammalian origins.

At its earliest appearance there are no additional homeotic gradients affecting it. In monotremes it proceeds with monotonous uniformity of shape through all dorsal vertebrae (Figure 11A). Independently in many metatherian and eutherian mammals, however, a homeotic gradient appears to have emerged that splits it into two progressively separated halves. The extent of the split increases in more posterior segments (Table 3, Figure 7B, 12). The split also reveals an intra-segmental antero-posterior gradient replicated in each segment.

Figure 11. Laminapophysis and lumbar transverse processes emergence in mammals.

(A)-Emergence of laminapophysis at T3 in Monotremata (Tachyglossus aculeatus) with no lumbar transverse processes (MCZ 25438). (B)-Emergence of orthapophysial lumbar transverse process (arrow) on vertebra also bearing a rib in small ferungulate (typical adult weight 1.5 kg) Tragulus javanicus subrufus (Artiodactyla) FMNH 62824. T-thoracic, L-lumbar.

doi:10.1371/journal.pone.0001019.g011

Figure 12. Laminapophysial splitting sequence in non-hominiform and hominiform catarrhines.

(A & B)-The laminapophysis splits into anterior metapophysis (**) and posterior anapophysis (*). The anapophysis forms a posteriorly directed styloid process on the arch and does not participate in the emergence of the pre-pararthral positioned parapophysial LTP. Typical euarchontogliran style anatomy in Macaca (Primates) Harvard Peabody N/3587. (C)-The anapophysis (*) forms the lumbar transverse process rather than a styloid process in hominiforms (e.g. non-proconsulid apes and humans)-juvenile Pan troglodytes. NLM-neomorphic laminapophysis, LTP-lumbar transverse process.

doi:10.1371/journal.pone.0001019.g012

The two halves are typically also subjected to opposite medio-lateral position effects (Figure 7, 10). In eutherians the anterior half shifts medially and forces the facet to rotate 90 degrees onto its medial surface to form the “diaphragmatic” joint (Figure 12A). The posterior portion may shift laterally. The few therian species that do not display splitting of the laminapophysis (Figure 7A) have most likely lost it secondarily.

The most striking and widespread mirroring phenomenon among eutherians produces a separate series of “splitting of the laminapophysis” proceeding anteriorly (thoraco-cervical direction) along the spine (reverse polarity) (Figure 13A) in addition to the standard posterior progression (thoraco-lumbar direction) (Figure 7B). Most interestingly, this is associated with a mirror of the diaphragmatic joint as well. The normal one appears as part of the thoraco-lumbar transition and the mirrored one occurs at the thoraco-cervical transition.

Figure 13. Homeotic mirroring of axial character elements.

(A)–Mirrored repetition of splitting of laminapophysis into anterior metapophysis and posterior anapophysis with associated sagittalization of facet in thoraco-cervical direction in addition to the usual eutherian thoraco-lumbar gradient polarity for this sequence-Myrmecophaga tridactyla (Pilosa) FMNH 49338. (B)–Medio-lateral mirroring of recurved lumbar facet joints–Dasypus novemcinctus (Cingulata) FMNH 60493.

doi:10.1371/journal.pone.0001019.g013

Anterior mirroring also occurs in most carnivores, all pholidotans (pangolins), many artiodactyls and some perissodactyls suggesting that this is an echo of a single homeotic gene-based replication event in an ancie