Homeotic and nonhomeotic patterns in the tetrapod vertebral formula

Abstract

Vertebrate development and phylogeny are intimately connected through the vertebral formula, the numerical distribution of vertebrae along the body axis into different categories such as neck and chest. A key window into this relationship is through the conserved Hox gene clusters. Hox gene expression boundaries align with vertebral boundaries, and their manipulation in model organisms often results in the transformation of one vertebral type into its neighbor, a homeotic transformation. If the variety in the vertebrate body plan is produced by homeotic shifts, then the number of adjacent vertebrae will be inversely related when making interspecies comparisons since the gain in one vertebra is due to the loss in its neighbor. To date, such a pattern across species consistent with homeotic transitions has only been found in the thoracolumbar vertebral count of mammals. To further investigate potential homeotic relationships in other vertebrate classes and along the entire body axis, we compiled a comprehensive dataset of complete tetrapod vertebral formulas and systematically searched for patterns by analyzing combinations of vertebrae. We uncovered mammalian homeotic patterns and found balances between distal vertebrae not anticipated by a Hox-vertebral homeotic relationship, including one that emerged during the progression from theropods to birds. We also identified correlations between vertebral counts and intergenic distances in the HoxB gene cluster which do not align with the common picture of a colinear relationship between Hox expression and vertebral categories. This quantitative approach revises our expectations for the diversity of a Hox-mediated vertebrate body plan.

Keywords: Hox code; evolution; vertebral column.

Fig. 1.

Vertebral formula and phylogeny. (A–C) Schematic of the approach to determine vertebral patterns. (A) Sketch of a platypus (Ornithorhynchus anatinus) skeleton reproduced from Cuvier (36). The numbers of vertebrae for each category yield the vertebral formula. (B) Schematic example of a phylogenetic tree distinguishing three different groups. Analyzing the vertebral formulas in the context of the tree yields the vertebral patterns. (C) We identify at least three types of constraints. For individual or Type-I constraints, the vertebral count in a single category is nearly constant within a specific branch of the tree. This same category can be very plastic or take on different constant values in other branches. In homeotic or Type-II constraints, adjacent vertebrae are anticorrelated and sum to a nearly constant value. In distal or Type-III constraints, nonadjacent vertebrae are positively correlated and can be numerically balanced. (D–G) Plots of the vertebral data arranged according to a phylogenetic tree (37, 38). (D) A phylogenetic tree containing all the species for which we have the full vertebral count. (We exclude some bird data to avoid unbalancing the tree, but it is used in later analysis.) The tree has been organized so that the four classes of tetrapods, Amphibia, Mammalia, Aves (birds), and Reptiles, are arranged vertically. The silhouettes to the right of the tree are a nonexhaustive representation of groups within the tetrapod classes. From top to bottom, they are (with individual examples in parentheses), within Amphibia: Urodela (salamanders), Anura (frogs), Gymnophiona (caecilians); Mammalia: Carnivora (lions), Artiodactyla (pigs), Primates (monkeys); Aves: Palaeognathae (emus), Anseriformes (ducks), Sphenisciformes (penguins), Accipitriformes (hawks); Reptilia: Crocodilia (alligators), Testudines (turtles), Iguania (lizards), Ophidia (snakes), Gekkonomorpha (geckos). (E) The individual vertebrae plotted vertically. (F) Vertical plots of two constraints and plasticities identified with our approach. $C-S$ is a global constraint found when comparing all tetrapods. $C+S$ is a plasticity found within birds. (G) Vertical plot of a constraint, $C+T-S-Ca$ found for all birds.

Fig. 2.

Representative constraints and plasticities. (A) Sketch of a prototypical Mammal redrawn from Owen (47). (B) Table with representative constraints and plasticities, including their formulas, the tetrapod class(es) within which they are found, and the constraint types (single: I, homeotic: II, distal balance: III). In order to combine all of our analyses and make the results more readable, we round the linear coefficients to the nearest integer (Materials and Methods). The coefficients are also color-coded so that red is positive and blue is negative. Several constraints (plasticities) have been identified. (C–K) Plots of representative constraints and plasticities (re)identified using our method. In each plot, the color of the data symbols indicates the tetrapod class according to the key in (C). For combination constraints (II, III) we calculate not only the Pearson correlation coefficient ($r$) between vertebral categories but also between their phylogenetic independent contrasts (PIC) (48). In (C) and (D), we plot vertebral counts which can be constraints for some classes and plasticities for others, yielding an L-shaped plot. (E–G) Mammalian homeotic patterns (II) including the previously known constrained sum $T+L$ and the constrained sums $L+S$ and $3S+Ca$ revealed by our analysis. Constraints on combined sums manifest as a negative correlation when the two categories are plotted against each other. In (F) we also plot genus-averaged data from datasets with partial vertebral formulas (13) as well as from Cetaceans (49) as open circles with higher transparency (up to $L=15$). Two additional homeotic patterns (II) are found for Testudinata (H) and Testudinata with Archosauria (birds and crocodilians) (I). In the background of (I), we additionally plot all tetrapods with a higher transparency. Excluding Amphibia and snakes yields significant correlation between $C$ and $T$ ($r\simeq -0.{84}^{\ast \ast \ast }$). (J and K) Plots of two prominent distal balance constraints (III). (J) The $C-S$ constraint spans all tetrapods, while in (K) the $C+T-S-Ca$ constraint is unique to birds.

Fig. 3.

Analysis of the bird constraint. (A) Sketch of a prototypical bird redrawn from Owen (47). (A–F) Images of the skeletons of Platalea leucorodia (Eurasian spoonbill), Upupa epops (Eurasian hoopoe), and Hypotaenidia okinawae (basionym: Rallus okinawae, Okinawa rail) from Abiko bird museum indicating $C$, $T$, $S$, and Ca. (G) Plot of the vertebral formula represented by a number of different colored circles. Here, we show birds from 21 different orders in Aves, as well as extinct Theropod species and extant species from the three other tetrapod classes. The vertical dashed line is an anterior–posterior boundary located after the thoracic ($T$). Birds show a balance between the two sides. (H) Cladogram of extinct and extant theropods extracted from a strict consensus tree (50). (I) Plot of the individual vertebral counts corresponding to the tree. (J) Plots of $C-S$ (dashed line) and $C+T-(S+Ca)$ for the species on the left. The gray band denotes three SD of $C+T-(S+Ca)$ around the average value of $\simeq$0.084 for extant birds. All taxa conform to the tetrapodian $C-S$ constraint while more phylogenetically distant theropods do not conform to the bird constraint. (K) Histograms of $(C+T)-(L+S+Ca)$ differentiated by tetrapod class to test a “modified” bird constraint ($L=0$ for Aves). The values of the modified constraint for theropods ($\square$) and pterosaurs ($▵$) are fixed at a vertical location for better visualization. All other groups of tetrapods save the putative flyers such as birds, bats (Chiroptera), and Pterosauria deviate strongly from the modified constraint.

Fig. 4.

Plots of the correspondence between the Hox genes and vertebrae. (A) Schematic inventory of the Hox genes in the four clusters present in the putative tetrapod ancestor (16). (B) Sampling of knock-out/in experiments and experimentally determined anterior expression boundaries of individual Hox genes that correlate with vertebral transitions (, , , –64). The species used in the study is indicated by the data point shape, and the specific Hox cluster is denoted by the color. Studies often focus on specific Hox regions, and so, for example, the higher density of data points near the anterior Hox genes does not represent a stronger correlation. In general, anterior Hox genes correlate with anterior vertebral boundaries, while posterior genes correlate with posterior boundaries. Notable exceptions to this are Hox9-11. (C) A plot of the Pearson correlation coefficient ($r$) between the Hox intergenic distances and the vertebral numbers for all classes. The strongest correlations we find are between the HoxB9-B13 intergenic distance and the $C$, $T$, $L$, and $S$ counts, which we indicate by arrows. (D–G) $C$, $T$, $L$, and $S$ vs. intergenic distance between HoxB9 and HoxB13. While the correlations are large and significant, they do not survive the PIC tests. Amphibians are not included in (D–G) since HoxB13 is missing in many of them (16).