- Top of page
- Supporting Information
Patterns of vertebral variation across mammals have seldom been quantified, making it difficult to test hypotheses of covariation within the axial skeleton and mechanisms behind the high level of vertebral conservatism among mammals. We examined variation in vertebral counts within 42 species of mammals, representing monotremes, marsupials and major clades of placentals. These data show that xenarthrans and afrotherians have, on average, a high proportion of individuals with meristic deviations from species’ median series counts. Monotremes, xenarthrans, afrotherians and primates show relatively high variation in thoracolumbar vertebral count. Among the clades sampled in our dataset, rodents are the least variable, with several species not showing any deviations from median vertebral counts, or vertebral anomalies such as asymmetric ribs or transitional vertebrae. Most mammals show significant correlations between sacral position and length of the rib cage; only a few show a correlation between sacral position and number of sternebrae. The former result is consistent with the hypothesis that adult axial skeletal structures patterned by distinct mesodermal tissues are modular and covary; the latter is not. Variable levels of correlation among these structures may indicate that the boundaries of prim/abaxial mesodermal precursors of the axial skeleton are not uniform across species. We do not find evidence for a higher frequency of vertebral anomalies in our sample of embryos or neonates than in post-natal individuals of any species, contrary to the hypothesis that stabilizing selection plays a major role in vertebral patterning.
- Top of page
- Supporting Information
Awareness of mammalian vertebral diversity has a long history, dating to classic works of the 18th and 19th centuries (Buffon, 1770; Owen, 1853). A substantial body of evidence collected since then supports the interpretation that vertebral counts in mammals, plus their extinct synapsid relatives, show relatively little variation compared to other vertebrate groups (Hautier et al., 2010; Muller et al., 2010; Sánchez-Villagra, 2010). The mammalian cervical count of 7 shows a particularly high level of conservatism; to a lesser extent, so do more distal parts of the mammalian vertebral column. Based primarily on the data from Owen (1853), Narita & Kuratani (2005) observed that most mammalian species have a combined total of 19 thoracic and lumbar vertebrae. Sánchez-Villagra et al. (2007) extended this observation, but noted that a high thoracolumbar vertebral count was common among afrotherians. Most afrotherian species exhibit over 20 thoracolumbar vertebrae; this may exceed 30 in some hyraxes. Sánchez-Villagra et al. (2007) also observed frequent departures from 19 among xenarthrans, with armadillos showing as few as 14 and two-toed sloths up to 29.
The extent to which the species listed in Owen (1853) and summarized by Narita & Kuratani (2005) depart from the vertebral counts observed in one or a few individuals has rarely been tested. Assessments of intraspecific vertebral variability are limited primarily to primates (Schultz, 1930; Bornstein & Peterson, 1966; Pilbeam, 2004; Oostra et al., 2005; Galis et al., 2006) and domesticated or lab animals (Sawin, 1937; Freeman, 1939; Stecher, 1962; McPherron et al., 1999). A few exceptions have focused on intraspecific variability in afrotherians and xenarthrans (Buchholtz et al., 2007; Asher et al., 2009; Buchholtz & Stepien, 2009; Galliari et al., 2010; Hautier et al., 2010).
Vertebral variation in mammals
Following the definition of Bateson (1894: 407), changes in vertebral counts are homeotic when ‘one of the component parts of the axial skeleton assumes the morphological appearance and function of its neighbor either immediately preceding or immediately following it… in distinction from meristic variations characterized by changes in total number of component parts’. Thus, when a given individual shows variation in vertebral count, it may be because of either change of one series identity at the expense of another (homeotic) or to the addition of a segment (meristic). Asher et al. (2009) provided some documentation of the extent to which specific clades of mammals vary among these categories. They tentatively concluded that as a group, southern placental mammals (i.e. afrotherians and xenarthrans, or Atlantogenata) showed more frequent departures from median series counts, particularly meristic, compared with northern placental mammals (i.e. laurasiatheres and euarchontoglires, or Boreoeutheria). However, their samples of individuals across 20 placental species were relatively small.
Since the 19th century, biologists have noted intriguing correlations between components of the axial skeleton, such as locations of the ribcage, pelvis and sternum. For example, based on the work of Welcker (1878), Bateson (1894: 121) noted that in sloths
‘…when the sacrum is far back, the ribs also begin further back…Backward homeosis of the lumbar segments is generally, though not quite always, correlated with backward homeosis of the cervicals, and vice versa.’
‘the sacrum may begin to shift its point of attachment backward to include a part of the twenty-eighth vertebra in addition to the entire twenty-seventh. Where this tendency finds fuller expression, the extra (13th) pair of ribs is found more completely developed and the sacral shift toward the twenty-eighth vertebra is increased.’
These observations correspond to a mechanism behind variation in the sloth neck discussed by Buchholtz & Stepien (2009). They proposed that the abaxial mesoderm patterning environments of the limb girdles and part of the ribcage, distinct from the primaxial mesoderm patterning environments of vertebrae, result in covariance of the positions of the ribcage and pelvic girdle across individuals within a species. Consistent with the text quoted above, they observed that individuals of Bradypus variegatus and Choloepus hoffmanni (but not their samples of Bradypus tridactylus or Choloepus didactylus) with fewer neck vertebrae and a more proximally situated ribcage exhibited a more proximally situated sacrum, compared with conspecific individuals with more neck vertebrae and a more distally situated sacrum (Buchholtz & Stepien, 2009: Table 1). With the caveat that there is some ambiguity regarding the adult derivatives of abaxial patterning in mammals, and if the mouse is representative of other mammals (Burke & Nowicki, 2003; Buchholtz & Stepien, 2009), covariance in the positions of ribcage and sacrum along the vertebral column is consistent with the presence of abaxial and primaxial modules within the axial skeleton, as hypothesized for sloths (Buchholtz & Stepien, 2009) and across mammals generally (Burke & Nowicki, 2003; McIntyre et al., 2007; Durland et al., 2008). The existence of such modules in sloths was further supported by Hautier et al. (2010), who noted that sloths with 8–10 ribless neck vertebrae still exhibited seven cervicals using a developmental criterion (late ossification of centra) to identify vertebral homology.
Another mechanism behind variation in the mammalian axial skeleton, one that does not exclude that of primaxial–abaxial modularity, was proposed by Galis et al. (2006). Based on a sample of pre- and post-natal humans, they noted a role for stabilizing selection in the highly conserved vertebral count of the mammalian neck. Among adult humans, the proportion of adults with cervical ribs is low, around 1%. In contrast, among cases of foetal or infant mortality recorded in hospital radiographs, this frequency climbs to 30–60% and is higher in individuals with multiple pathologies (Galis et al., 2006: Fig. 3). They suggested that pleiotropic effects of Hox mutations lead to both cervical ribs and other abnormalities (e.g. cancer), resulting in a much lower proportion of individuals with cervical ribs that survive to reach sexual maturity.
Galis & Metz (2007) have further proposed that mammals deviating from typical vertebral counts may have succeeded in doing so because of a lower metabolic rate. However, they have also noted that endothermic diapsids (represented today by Aves) do not have this association between high metabolism and vertebral conservatism. In addition, synapsids acquired their high level of vertebral conservatism long before the appearance of what we now regard as a characteristic mammalian metabolic rate (Muller et al., 2010). Clearly, therefore, factors beyond metabolic rate have contributed to vertebral constraint in mammals.
Goals of present study
Here, we present and interpret data concerning intraspecific variation in vertebral counts across 42 mammalian species, focusing on placentals. Such data are necessary to critically evaluate the above issues regarding variation and constraint in the mammalian axial skeleton. In particular, we seek to address the following questions:
Do certain clades exhibit more intraspecific vertebral variation than others?
Are meristic and homeotic categories of vertebral change distributed evenly throughout mammals?
Does length of the ribcage and/or number of sternebrae correlate with position of the sacrum?
Do embryonic, foetal and infant mammals show more vertebral anomalies than post-natal specimens?
Answers to the first two questions will help address the phylogenetic hypothesis articulated by Sánchez-Villagra et al. (2007) and Asher et al. (2009) that afrotherians and xenarthrans show less axial constraint than other mammals. An answer to the third will help test the mechanism behind vertebral variation implied by Buchholtz & Stepien (2009). Although they did not directly make predictions regarding the relationship between number of sternebrae and the position of the sacrum, and while we do not deny the potential role of other factors in development influencing this relationship, such a correlation would follow from the modularity implicit in prim/abaxial patterning. If at least part of the ribcage and sternum is modular along with the pelvic girdle (abaxial), they should positionally covary relative to the vertebral column (primaxial).
An answer to the fourth question will help test the mechanism behind vertebral variation articulated by Galis et al. (2006). Implicit in their discussion is that a higher proportion of vertebral anomalies should be found among embryonic and neonatal mammals relative to post-natal mammals, because deleterious pleiotropic effects of Hox patterning resulting in vertebral anomalies do not necessarily prevent conception or birth, but show a high rate of mortality during early development. In other words, the effects of stabilizing selection should be evident in a population as individuals age.
- Top of page
- Supporting Information
We assembled counts of presacral vertebrae from museum skeletal preparations, cleared-and-stained specimens and X-ray computerized tomography (CT) scans of alcohol-preserved mammals, representing two species of monotremes, four marsupials and 24 placentals, adding substantially to the sample discussed by Asher et al. (2009). For three of these species (Amblysomus hottentotus, Eremitalpa granti, Sorex araneus), radiographs also contributed to tabulations of vertebral counts. Specimens housed at the following institutions contributed to our sample: London, UK (BMNH); King Williams Town, South Africa (KM); Madrid, Spain (MNCN); Paris, France (MNHN); Stockholm, Sweden (NRM); Oxford, UK (OUM); Zürich, Switzerland (PIMUZ); Tervuren, Belgium (RMCA), Pretoria, South Africa (TM); Cambridge, UK (UMZC); Cape Town, South Africa (ZM); and Berlin, Germany (ZMB).
The median sample size for data on each species we collected (not including literature sources), including skeletons, X-ray CT scans, cleared-and-stained specimens and radiographs, was 21 individuals. The smallest samples were Peromyscus and Phascolarctos (9) and the largest Amblysomus (121). Data from an additional 12 species were added from previous studies of vertebral diversity (Table 1): five primates and one rodent from Pilbeam (2004); two xenarthrans from Buchholtz & Stepien (2009); and domesticated horses, pigs, mice and rabbits from (respectively) Stecher (1962), Freeman (1939), McPherron et al. (1999) and Sawin (1937).
Table 1. Correlations between number of thoracolumbar vertebrae, sternebrae and rib-bearing vertebrae.
For the 30 species represented by data we collected, specimens were selected that unambiguously preserved articulations between all vertebrae. In most cases, this meant retained, natural intervertebral articulations still bound by dried connective tissue. Where breaks in the vertebral column occurred, fits between vertebrae were checked by carefully placing the adjacent elements together to ensure a natural fit. Specimens that consisted of ambiguously associated parts of the skeleton, without clear evidence of vertebrae belonging to a single individual, were excluded.
Skeletons of alcohol-preserved embryos and foetuses were analysed using high-resolution X-ray microtomography (μCT) at the engineering department of the University of Cambridge (Cambridge, UK), at the Helmholtz Zentrum (Berlin, Germany), at the Natural History Museum (London, UK) and at VISCOM SARL (Saint Ouen l’Aumône, France). 3D rendering and visualization were performed using the open source software Drishti v.1.0 (Drishti Paint and Render, Limaye, 2006). Threshold values between ossified elements and soft tissues were substantial and easily allowed osteological reconstructions. Following methods described in Wilson et al. (2010), we also examined cleared-and-stained mammals at the Paläontologisches Institut und Museum, Zürich.
Cervical vertebrae were defined as those between the skull and the first vertebra bearing large, bilateral ribs. A developmental criterion for recognizing cervical vertebrae based on the late ossification of centra has recently been discovered (Hautier et al., 2010), but is generally unavailable for most museum specimens. Among mammals, the first rib-bearing vertebra was defined as showing independent articulations via the first costal cartilage to the manubrium sterni. Hence, a neck vertebra with asymmetric, free-floating riblets, unilateral ribs or ones that join with T1 ribs to articulate with the manubrium would still be ‘cervical’.
Thoracic vertebrae were defined as those elements bearing large, bilateral ribs longer than the centrum is wide, with corresponding rib facets on or near each vertebral neural arch.
Lumbar vertebrae were defined as those elements cranial to the sacro-iliac articulation that lack conspicuous rib facets and show transverse processes. Vertebrae that show riblets similar in size to the facets that bear them, or miniscule or unilateral rib facets, were defined as ‘lumbar’. Posterior lumbar vertebrae may show some articulations with the sacrum, but not more than the first sacral vertebra.
The proximal-most sacral vertebra was defined as the first vertebral element to show complete, bilateral fusion of its transverse processes to articulate with the ilium on each side.
Sternebrae were counted as all distinct, ossified elements articulating with costal cartilages between (and including) the manubrium sterni and the xiphisternum. Elements cranial to the manubrium (e.g. the monotreme interclavicle) were excluded.
Outside of sloths, we observed no variation in cervical vertebral count. In addition, caudal vertebral counts are not reliably obtainable from museum preparations (Pilbeam, 2004). Hence, we relied primarily on thoracolumbar counts to quantify vertebral variation across mammals except for sloths. For all individuals in a given species, we took the median values for cervical, thoracic and lumbar categories. The sum of median thoracic and median lumbar values for a given species was defined as its median thoracolumbar count. This is usually, but not necessarily, equal to the median of summed individual thoracolumbar values. We recognized a ‘homeotic’ change when an individual conformed to its species’ median thoracolumbar count, but differed from its species’ lumbar and thoracic medians. We recognized a ‘meristic’ change when an individual deviated from its species’ median thoracolumbar count.
Anomalous vertebrae were those that mixed features of the above vertebral definitions, e.g. by combining small riblets and transverse processes, or that showed asymmetrical ribs, transverse or ventral processes. We tabulated occurrences of such anomalies for each species and noted if they occurred in embryonic or neonatal (before or near birth) vs. post-natal (showing ossification throughout the skeleton and at least some epiphyseal fusion) individuals. For our samples of Cryptotis and Rhabdomys with at least partial data on age, we arbitrarily recognized 10 days as the cut-off between the categories ‘neonatal’ and ‘post-natal’. Because vertebral anomalies can be difficult to detect in radiographs, we used only specimens represented by X-ray CT scans, cleared and stained preparations or macerated skeletons to calculate their frequency. We could not be certain about previous investigators’ criteria for identifying vertebral anomalies and therefore did not use our literature sources to infer their frequency. Based on these criteria, seven species (Dasypus novemcinctus, B. tridactylus, Procavia capensis, Tenrec ecaudatus, Erinaceus europaeus, Rhabdomys pumilio, Cryptotis parva) in this study had the best samples of pre/neonatal vs. post-natal individuals.
Pearson’s r was calculated to evaluate the significance of correlations between two pairs of intraspecific variables: position of the sacrum with number of rib-bearing vertebrae, and position of the sacrum with the number of sternebrae (Table 1). For all mammals but sloths, we used thoracolumbar count to infer the position of the sacrum, because their cervical count is invariant in our sample. For sloths, we used presacral count to infer position of the sacrum. Coefficients of variation (CV) represent the standard deviation of a data series expressed as a percentage of the mean and were calculated for thoracolumbar counts at intraspecific and selected suprageneric levels (Table 2), using the correction for small sample sizes (CV(1 + (1/4n))) described in Sokal & Rohlf (1995). Our original data and those derived from literature sources (Appendix S1) and summary statistics from both (Appendix S2) are available in spreadsheet format, available upon request from the corresponding author and/or journal website.
Table 2. Summary statistics for vertebral counts across clades of mammals. Nomenclature follows Asher & Helgen (2010). Data for constituent individual species are given in the Appendix S2. ‘TL CV*’ refers to the corrected thoracolumbar coefficient of variation, or coefficients of variation (CV) = stdev(100)/average, using the correction for small sample size [CV(1 + (1/4n))] as described in Sokal & Rohlf (1995).
| ||TL CV*||% homeotic||% meristic||% normal||% anomalies||n individ||n species||Notes|
- Top of page
- Supporting Information
Constraint in vertebral patterning shows marked differences among clades of mammals. In this dataset, the sampled monotremes, afrotherians, xenarthrans and primates show higher thoracolumbar CVs than other clades (Fig. 1). The rodent species in our analysis are the most conservative in the makeup of their axial skeleton, with low CVs for thoracolumbar count and a low frequency of vertebral anomalies (Fig. 1, Table 2). Afrotherians and xenarthrans depart more frequently than other clades in the proportion of individuals departing from median presacral vertebral counts, often showing more individuals with meristic variations than normal counts (Fig. 1). Both of these southern groups show fewer individuals with species-median vertebral counts than primates, even though all three groups exhibit a similarly high CV for thoracolumbar variation. Exceptions to low species-median counts include the two sampled golden moles (Amblysomus and Eremitalpa) and the sengi (Macroscelides), in which 67–79% of individuals exhibit species-median counts. Another exception may be the dasypodid Chaetophractus as discussed by Galliari et al. (2010), but not included in this study. Our sample of the closely related D. novemcinctus showed slightly over half of individuals with median vertebral counts (Fig. 1).
Interestingly, pairs of ecological analogues between afrotherians and other mammals, such as burrowing chrysochlorids (Afrotheria) vs. talpids (Lipotyphla), and insectivoran-grade Setifer (Afrotheria) vs. Erinaceus (Lipotyphla), differ substantially in level of vertebral variation. In each case, the afrotherians show fewer individuals with species-median counts, and chrysochlorids (average CV 1.98) show a much higher thoracolumbar CV than Talpa (CV 0). Although the thoracolumbar CV of Erinaceus falls within the Afrotherian range (slightly higher than that of Setifer), and although it also shows a high proportion of individuals with vertebral anomalies (43% vs. 38% in Setifer), Erinaceus shows a substantially greater proportion of individuals with a median vertebral count (69%) than Setifer (19%).
If all mammals follow the tissue-patterns exhibited in mice, in which the sternebrae, sternal ribs and limb girdles are patterned by abaxial mesoderm and vertebrae and proximal ribs by primaxial mesoderm (Burke & Nowicki, 2003), then the mechanism of prim/abaxial modularity would lead to the expectation that number of sternebrae should covary with position of the sacrum, because both are abaxially patterned. For most mammals in our sample, this is not the case. Only three (S. setosus, T. ecaudatus, D. arboreus) exhibited a significant correlation between number of sternebrae and sacral position (Fig. 2).
The more consistent (but still not universal) correlation we do recover is between number of rib-bearing vertebrae and position of the sacrum. If the model suggested by Buchholtz & Stepien (2009) for sloths is applicable across mammals, this might indicate that abaxial mesoderm-patterned structures vary across mammalian species, and that more of the ribcage in those species that exhibit this correlation is abaxially patterned, not primaxially as inferred for mice. This observation provides a framework for testing the mesodermal patterning of the axial skeleton in mammals. Species that lack a correlation between sacral position and length of ribcage, such as Procavia, may differ in the nature of the lateral somitic frontier (the structure that separates primaxial and abaxial domains following Burke & Nowicki, 2003) relative to those that exhibit this correlation, such as the shrew Cryptotis.
In our sample, vertebral anomalies among pre/neonatal individuals were less frequent than among older specimens. This contrasts with the pattern among humans discussed by Galis et al. (2006), in which the proportion of foetuses and neonates with anomalies greatly outnumbered that of adults. The foetal and neonatal humans used by Galis et al. (2006) were drawn from a hospital sample which, relative to the population at large, may overrepresent pathologies. It therefore remains to be demonstrated that the sample of Galis et al. (2006) is representative of pre/neonatal humans in general. Whatever the frequency of vertebral anomalies is among natural age-cohorts of Homo sapiens, a higher rate of such anomalies would be expected among younger individuals than older ones if the hypothesis of stabilizing selection is correct for mammals generally. This expectation is not borne out by our data (Fig. 3).
In sum, we have found that southern placental mammals share a high level of thoracolumbar vertebral variability with certain northern placental mammals, particularly primates. However, as a group, they differ from other mammals in showing frequent departures from median vertebral counts. Rodents in our sample are the most vertebrally conservative mammals quantified thus far. Covariation between the length of the ribcage and sacral position may indicate a level of modularity within prim- and abaxially patterned elements of the axial skeleton. Although the less frequent correlation between number of sternebrae and sacral position is not consistent with such modularity, it may reflect variation in the identity of prim- and abaxially patterned adult structures across mammals.
- Top of page
- Supporting Information
We are greatly indebted to Marcelo R. Sánchez-Villagra and Laura Wilson (University of Zürich) for providing helpful comments and data on cleared-and-stained specimens. We thank Johannes Müller (Museum für Naturkunde Berlin) and two anonymous reviewers for constructive critiques of the manuscript. We thank Thomas Lehmann (Senckenberg Institut Frankfurt) and Nigel Bennett (University of Pretoria) for help with data collection. Judith Chupasko (MCZ), Louise Tomsett, Paula Jenkins, Roberto Portela Miguez (BMNH), Jacques Cuisin (MNHN), Frieder Mayer, Detlef Willborn, Saskye Janke, Peter Giere (ZMB), Theresa Kearney (TM), Lucas Thibedi (KM), Malgosia Nowak-Kemp (OUM), Denise Hamerton, Margaret Avery (ZM), Emmanuel Gilissen (RMCA), Olavi Grönwall, Per Erikson and Bodil Kajrup (NRM) kindly facilitated access to museum collections. For advice and discussion, we thank Anjali Goswami and Vera Weisbecker. Alan Heaver (Cambridge) and Richie Abel (London) provided generous help and advice with X-ray CT acquisition. RJA acknowledges financial support from the Leverhulme Trust, the Royal Society, and the European Commission’s Research Infrastructure Action via the SYNTHESYS Project. KHL acknowledges a Harvard-Cambridge Summer Fellowship from the Harvard-Cambridge Scholarships Committee.