Ontogenetic Data Analyzed As Such in Phylogenies

(1)

HAL Id: hal-01469189

https://hal.sorbonne-universite.fr/hal-01469189

Submitted on 16 Feb 2017

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Ontogenetic Data Analyzed As Such in Phylogenies

Jérémie Bardin, Isabelle Rouget, Fabrizio Cecca

To cite this version:

Jérémie Bardin, Isabelle Rouget, Fabrizio Cecca. Ontogenetic Data Analyzed As Such in Phylogenies.

Systematic Biology, Oxford University Press (OUP), 2016, 66 (1), pp.23-37. �10.1093/sysbio/syw052�.

�hal-01469189�

(2)

ONTOGENY IN PHYLOGENIES

Ontogenetic data analyzed as such in phylogenies.

Jérémie BARDIN^1,2,3, Isabelle ROUGET^1,2,3, Fabrizio CECCA^{1,2,3, †}

1Sorbonne Universités, UPMC Univ Paris 06, UMR 7207, CR2P, F-75005, Paris, France

2Muséum National d'Histoire Naturelle (MNHN), UMR 7207, CR2P, F-75005, Paris, France

3Centre National de la Recherche Scientifique (CNRS), UMR 7207, CR2P, F-75005, Paris, France

Corresponding author: Jérémie Bardin, jeremie.bardin@upmc.fr, 01.44.27.51.77 Systematic Biology Advance Access published June 16, 2016

at UPMC on June 20, 2016http://sysbio.oxfordjournals.org/Downloaded from

(3)

ABSTRACT

Ontogeny is rarely included in phylogenetic analyses of morphological data.

When used, the ontogenetic information is reduced to one character for two or three different ontogenetic stages. Several examples show that current methods miss a major part of the information. We here propose a new method for including the ontogenetic dimension in coding schemes of phylogenetic analyses. Our goal was to maximise the phylogenetic information extracted from ontogenetic trajectories. For discrete features, we recommend including precise timings of transformation(s) from one state to another in the ontogenetic trajectories. For continuously varying features, growth laws are modelled on raw data using least-square regressions. Then,

parameters of models are included in the coding scheme as continuous characters.

This method is employed to reconstruct phylogenetic relationships using the

ammonite family Amaltheidae as a test subject. Based on the same data set, a second analysis has been performed only for characters of the adult stage. Comparisons of retention index, bootstrap support and stratigraphic congruence between the two analyses show that the inclusion of ontogeny yields better phylogenetic

reconstruction. Morphological traits in ammonites which are usually the most homoplastic show a better fit to most parsimonious trees by including the

ontogenetic dimension. In several cases, growth rates and patterns of growth have better fit to phylogeny than adult shapes, implying that paths of ontogeny can be more relevant than its products.

(4)

KEY-WORDS

phylogeny, cladistics, ontogeny, development, ammonite, Amaltheidae, Jurassic, Pliensbachian

INTRODUCTION

The interface between ontogeny and phylogeny, recently called "evo-devo", is a classical topic in evolutionary biology (e.g., Gould 1977; Alberch et al. 1979; Fink 1982; Humphries 1988; Mabee 1993; McNamara 1996; Raff 1996; Futuyma 1998; Rohlf 1998; Wiens 2005; Germain and Laurin 2009; Laurin and Germain 2011; Wolfe and Hegna 2013). One of the main cutting edges of evo-devo is that emergent properties appear from the study of both ontogeny and evolution together (Hall 2000). Thus, a major goal in this field is to represent and use ontogenetic data in phylogenies.

Several attempts have been made to include ontogenetic data in phylogenetic analyses (Mabee and Humphries 1993; Smith 1996; Velhagen 1997; Steyer 2000;

Jeffery, Bininda-Emonds, et al. 2002; Jeffery, Richardson, et al. 2002; Germain and Laurin 2009; Laurin and Germain 2011; Wolfe and Hegna 2013) and most of these studies showed that ontogenetic data contain phylogenetic signals and that their inclusion outperforms classic analyses. We briefly review current methods and some of their limitations, illustrating them with an ammonite example. These methods used ontogeny of features as discrete characters based on their shape and/or their timing or duration. However, many ontogenetic changes of morphology are

(5)

continuous or vary continuously both in shape and time. Ontogeny can be regarded as the covariation of size, shape and age; so far the best way to represent it, is the ontogenetic trajectory formalized with continuous allometric growth models (e.g.

Gould 1966; Alberch et al. 1979). We present a new approach based on these trajectories where ontogenetic data are used as such in phylogenies. This formal approach can be applied both to represent data on a given phylogeny and, in phylogenetic analyses, to increase the use of ontogenetic information. Our method avoids artificial sequencing into discrete units and codes ontogeny itself rather than its products. We assess the relevance of the method by comparing the results of two parsimony analyses: with and without including the treatment of ontogeny. The group under study is the Amaltheidae, a Pliensbachian (Early Jurassic) ammonite family which is well-known as a central model for ontogenetic studies of ammonoids (Mattéi 1985; Meister 1988). Ammonoids are a well-suited model for the study of ontogeny in relation to phylogeny because their conch records an important part of their ontogeny. Several studies emphasized how one can take advantage of the ontogeny for phylogenetic purposes (e.g. Korn 2012; Knauss and Yacobucci 2014).

Review of methods dealing with ontogenetic characters

Since Haeckel's famous adage rebuttal, several methods were proposed to include ontogenetic variation of characters in phylogenetic reconstructions. In the literature, a large terminology exists to discretize ontogeny (e.g., for butterfly: egg, caterpillar, chrysalis and adult butterfly). Most of the phylogenetic analyses using

(6)

ontogenetic information are based on these traditional ontogeny discretizations. The method proposed by Steyer (2000) separates characters for different ontogenetic stages (e.g. larval, juvenile and adult in its analysis) resulting in several matrices. An analysis is performed for each dataset. The resulting trees, called ontotrees, are compared and combined. Wolfe and Hegna (2013) proposed a somewhat different method in which each ontogenetic stage is coded as an OTU in the matrix. These methods characterize evolutionary dynamics between ontogenetic stages. However, a major limitation is that all characters have to be coded on the same divisions of ontogenetic stages. Why should we consider that all characters change at the same time during ontogeny? Some works (e.g. Minelli et al. 2006 for arthropods) have shown that conventional periodization (i.e. the categorization of ontogeny into discrete intervals) used to describe ontogeny may not illustrate accurately the evolution of the group. The transformation of different characters can occur at different times during ontogeny. This is illustrated in figure 1 where trajectories of three ammonites (amaltheid taxa) are plotted for two morphological traits: the umbilical width (Uw = U/D) and the tuberculation. Ontogenetic trajectories of Uw follow a saturating growth law (Alberch et al. 1979) according to which the shape tends to reach a constant when approaching adult size. The tuberculation develops two stages: the first one called the Perlatum stage is characterized by small tubercles on each rib (black on Fig. 1), the second one called the Gibbosum stage is

characterized by bigger tubercles which develop on every few ribs (grey on Fig. 1).

The Uw ontogenetic trajectory of A. salebrosus is more curved than the one of A.

(7)

gibbosus, this difference expresses during the whole ontogeny and is especially easy to observe at the beginning of the trajectories. The Perlatum stage is the same in both taxa whereas the Gibbosum stage has a different offset between the two species.

Comparison of these ontogenetic trajectories shows that shifts of the two characters can occur in distinct parts of the ontogeny. Moreover, this example shows that ontogenetic trends can be different for each character, as Uw is peramorphic for A.

salebrosus in comparison to A. gibbosus, whereas the offset of the Gibbosum stage is paedomorphic. The third ontogenetic trajectory presents the tuberculation for P.

bechteri (Fig. 1); it has the same offset but a different onset for the Gibbosum stage.

This example shows that timings can be independent between several events in the ontogenetic trajectory of one given morphological trait, thus constituting several characters. Such examples have been described in birds. This concept is known as

“mosaic heterochrony” or “dissociated heterochrony” (see review in Klingenberg 1998). Despite being well-known in the field of morphometrics, the possibility for the characters to evolve independently in the ontogeny has never been formalized for phylogenetic inference. Indeed, the above-mentioned methods do not precisely take into account the different timings in morphological changes.

The central question of the recognition of ontogenetic stages and their comparison between OTUs has led to the development of a suite of methods aimed at treating the order of suggested events and including it in phylogenetic analysis:

mainly "event-pairing" (Smith 1996; Velhagen 1997) and "event-pair cracking"

(Jeffery, Bininda-Emonds, et al. 2002; Jeffery, Richardson, et al. 2002; Jeffery et al.

(8)

2005). A characteristic of these methods is that they do not need a measure of age or time. This has been viewed as an advantage as the use of proxies for age has been discarded in some works (e.g. Smith 2001; Koenemann and Schram 2002). Sequence- based methods can only detect and use ontogenetic variations which make some events to change their order of appearance. This is an important intrinsic limitation.

To us, these methods should only be used when time or its proxies are too dubious to be used. Germain and Laurin (2009) proposed the continuous analysis which uses age and further showed (Laurin and Germain 2011) that the use of time rather than sequences could extract additional reliable information for phylogenetic

reconstruction.

The terms "development" and "ontogeny" are often used in very different ways. Gould (1977) recalls that ontogeny is the life history of an individual, whereas development is differentiation or any change in shape during ontogeny. Thus, development is included within the concept of ontogeny, together with maturity and growth. Most of the methods above-mentioned are presented as treating

developmental data, whereas they usually use sexual maturity or increase in size (growth). Thus, we should use the term "ontogeny" for these purposes. Developed by Alberch et al. (1979), the ontogenetic trajectory consists in representing the

expression of a given trait during the lifespan of an organism. Keeping this in mind, we realize that all the above-mentioned methods compare the results of ontogeny at different stages, rather than ontogeny itself (i.e. the trajectory). This is mainly due to discretization both in the time (e.g. hatching, sexual maturity) and/or shape

(9)

dimensions (e.g. discretization of continuous traits). When the three dimensions of ontogeny (i.e., growth, development, maturity) are available, they should all three be used and this is what we propose with our method. Referring to Gould's concepts, coding differences between modelled ontogenetic trajectories corresponds to coding dissociations between ontogenetic trajectories in these three dimensions as precisely and accurately as possible. This can be illustrated by deciphering the differences between the three ontogenetic trajectories of the character Uw provided in figure 2.

First of all, histograms of distributions show that expressed values of Uw are overlapping for the three taxa and that distributions of A. talrosei and P. solare are very similar to each other. Then, having a look at the tendencies of the character in the ontogeny, the trajectories of A. gibbosus and A. talrosei decrease while that of P.

solare increases. This would be an argument to group A. gibbosus and A. talrosei. To go further, A. gibbosus and P. solare share a curved ontogenetic trajectory while A. talrosei has a linear one. Exploiting more accurately these ontogenetic trajectories leads to an important additional source of characters. This recalls what some phylogenists did by including changes of derivatives of traits against ontogeny (e.g., Wagner 1999).

This is the information we propose to capture with the approach presented in this paper which aims at exploiting the ontogeny itself without any a priori discretization of the timing and of the ontogeny.

The Method

(10)

The objective is to code ontogenetic variation as characters and character states. We choose the theoretical framework of Alberch et al. (1979) for the

description of the ontogenetic trajectories. It consists of representing the expression of a morphological feature during the life of an individual or a population. This formalism can be applied to all different types of characters used in phylogenetic analysis. For binaries (Fig. 3a), the ontogenetic trajectory is determined by a time of onset (α) and a time of offset (β). For multistates (Fig. 3b), several onset and offset times are introduced depending on the number of states developed in the

ontogenetic trajectories. In addition to onset and offset times, growth curves of continuous characters (Fig. 3c) are determined by a growth rate (k). The latter can be involved in different growth laws (e.g., a saturating growth law in Fig. 3c). Basically, our method consists of identifying and coding homologous ontogenetic stages for each character. We discuss dependences between ontogenetic stages which can be complex and have to be integrated into the coding scheme. Steps involved in this method for coding ontogeny are summarized in figure 4.

Ontogenetic Stage Comparisons — Differences between ontogenetic trajectories can be of two types: novelties on the one hand, changes in rates and timings of parts of ontogenetic trajectories on the other, the so-called heterochronies. In organisms with complex ontogenies, the analysis of these differences requires precise

description of the ontogenetic trajectories and the distinction of ontogenetic stages.

The ontogenetic stages constitute units that we want to compare in terms of rates and

(11)

timing among taxa. The possibility of splitting and treating separately ontogenetic trajectories of different traits should be added to current phylogenetic methods. Alba (2002, p. 33-36) claims that ontogenetic trajectories have to be compared at

homologizable stages. He considers that a developmental stage is a period between two homologous shape-independent developmental events. According to him, these events should be used to standardize ontogeny instead of absolute age. However, we consider that these events can also be subject to evolutionary variation in their timing in terms of absolute age and should be treated the same way as any other characters.

These homologous events define homologous stages which can be compared between taxa in terms of shape, growth-rate and timing. Stages can be identified by remarkable events in ontogenetic trajectories (e.g. appearance of a particular feature, break in trajectory for continuous characters). Sometimes the entire ontogenetic trajectory of a given character corresponds to a stage because no specific event can be identified. When several stages can be recognized, several criteria for recognizing between-taxa homologous stages can be used. The first criterion is the order of stages, considering that they most likely appear in the same order. The second criterion is the expression of the characters. For discrete characters it means sharing the same state. For continuous characters, it corresponds to values on shape axis (Y axis) and/or values of slope. The third criterion corresponds to similar

timing/durations of some potentially identified homologous stages between ontogenetic trajectories of taxa. None of these criteria has special priority over the others.

(12)

Which Kind of Character? — The treatment of each stage depends on the type of character under study. Discrete characters provide the simplest case. During

ontogeny, an OTU can change from one state to the other once or several times. In that context, the information to code is the absence/presence of the states and the timing of this (these) transformation(s) during growth. The latter corresponds to parameters α and β of Alberch et al. (1979), respectively the onset and offset.

The case of continuous characters is more complex as the variation is

continuous in the bivariate representation of an ontogenetic trajectory (Fig. 3c). Many studies based both on empirical data (e.g. Gould 1977) and theoretical models

(Alberch et al. 1979) provide useful equations for describing ontogenetic trajectories.

In practice, ontogenetic trajectories are described with equations that fit the data. We propose to use parameters of the fitting models as continuous characters. Many methods exist to transform these raw parameters into code (e.g., gap-coding,

segment-coding, gap-weighting). In their coding scheme, to a greater or lesser extent, all available methods reduce the information expressed in the raw data (Wiens 2001;

Goloboff et al. 2006; Bardin et al. 2014). We favour methods that offer the highest resolution in continuous character treatment in order not to discard phylogenetic information, even minor, expressed from ontogenetic trajectories.

Dependences in the Coding Scheme — When entire trajectories are considered, this stage-by-stage approach leads to consider links between those stages which have

(13)

initially been considered separately. Two different types of dependences have to be taken into account. The first one concerns the coding procedure itself and the other implies biological dependences.

When several stages have been recognized and coded separately, a part of the information is redundant (Fig. 3b). Two stages that follow each other share the same timing. The end of the first stage (β1) is identical to the beginning of the second stage (α2). These two parameters have to be coded only once in the matrix. Concerning continuous characters, these redundancies exist also for the shape axis (i.e. the σ1 of a first stage is identical to the σ0 of a second stage). Again, these two parameters have to be coded only once.

In phylogenetic inference, investigations about the dependences between characters is a central topic and has received much attention (e.g. Huelsenbeck and Nielsen 1999; O'Keefe and Wagner 2001; Maglia et al. 2004). Integrating important amounts of information from ontogeny forces us to investigate dependences about timing and rates. We distinguished two cases in which the coding scheme has to be adapted because of dependences. The first one concerns transformations in different characters which occur at the same age. One should investigate whether this similar timing could be due to the same causality. If such a hypothesis is supported with evidence, this same timing between the different characters has to be coded only once in the matrix. This case corresponds to several usual developmental events such as hatching, metamorphoses, sexual maturity.

(14)

The second one involves dependences that can be integrated under the concepts of modularity, fragmentation and integration (e.g. Schlosser and Wagner 2004). Definitions of a module are really diverse (Eble 2005). Albeit in phylogenetic studies, we generally use what Eble (2005) called "Variational Morphological Modules", that is to say, modules are groups of traits co-varying and thus units we want to use as characters in phylogenetic analyses. They are of great importance in phylogenetic analysis for coding independent characters. Integration occurs when several modules start to covary, while they had an independent behaviour in earlier evolutionary history. Fragmentation is the division of a module into several

independent character sets. Integration and fragmentation can affect both shape and timing. Modules have to be recognized a priori and integrated in the coding scheme.

Extrinsic versus Intrinsic Time — By using ontogenetic trajectories, we represent shape/size (Y axis) vs age (X axis). Absolute age is called extrinsic time. In practice, it can be measured for living organisms but in wild specimens or fossils, this is rarely possible. Variables from organisms are thus selected to represent time; they are the so-called intrinsic time (e.g. Blackstone 1987; Reiss 1989; Hall and Miyake 1995;

Klingenberg 1996). Size is one of the most successful variables for representing age (Strauss 1987, 1990; McKinney and McNamara 1991: p. 41). Strauss (1987: p. 33) argued that overall body size is the best measure to standardize development of organisms. However, most of the authors who use size consider it to be an

approximation of extrinsic age (e.g. Dommergues 1988; McKinney 1988; McKinney

(15)

and McNamara 1991). McKinney (1986) introduced the term “allometric

heterochrony” for transformation deduced from shape-size representations (i.e. size as standard for age). Several examples have been published demonstrating that size was incongruent with age (Emerson 1986; Blackstone and Yund 1989; Klingenberg and Spence 1993; Leigh and Shea 1996). Determining whether size is a good proxy for age depends on the group under study and the taxonomic scale of the analysis.

For example, in ammonoids, size is generally approximated by the diameter of the shell. Ammonite shells have an accretionary growth, thus size (i.e. diameter) vs age is monotonous. As soon as their whorl expansion rate and growth rate is more or less constant, we can consider that size is a good standard for age. However, comparing phylogenetically distant ammonoid taxa with various whorl expansion rate and growth rate will make diameter unsuitable for providing a good standard for age.

The taxonomic scale of comparison is thus a major limitation of the use of size.

McKinney and McNamara’s (1991) basic assumption was that size increases identically between taxa to be compared (Klingenberg 1998). As the taxonomic breadth of the ingroup increases, size should become an increasingly poor proxy for age.

Reiss (1989) claimed that extrinsic time is not suitable for comparing developmental events because they are too dependent on several environmental parameters such as temperature as well as size itself. Thus, an intrinsic time closer to these phenomena should be selected to scale developmental events. In this paper, we are seeking to reconstruct evolutionary history through ontogeny using the

(16)

parsimony criterion. By using extrinsic time, Reiss’s criticism would correspond to detecting similarities due to environmental or size parameters. In cladistics, these correspond respectively to homoplasy and dependence. Concerning homoplasy, the power of cladistics is to find trees that minimize it. Dependence can be detected a priori or a posteriori to the analysis (O’Keefe and Wagner 2001). Thus, the main criticism against the use of extrinsic time could be avoided in the context of cladistics as well as of model-based methods. We highly recommend the use of extrinsic time.

When it is unavailable, one should select the least-biased proxy variable.

MATERIAL &METHODS

Taxa

The ingroup (Tab. 1) of our analysis is composed of all accepted species within the family Amaltheidae, following Howarth (1958) and Meister (1988). Some

additional comments on taxa are available from online appendix 1

(http://dx.doi.org/10.5061/dryad.0c1p0). The ingroup covers the entire morphological diversity of the family. Some species have been grouped into a single OTU when they were indistinguishable morphologically. The morphological analysis is based on a thorough survey of the literature (mainly Howarth 1958; Tintant et al. 1961; Repin in Efimova 1968; Mattéi 1985; Meister 1988; Morard 2004, Bardin et al. 2013) in addition to some material available in the collections of University Pierre et Marie Curie (UPMC) and of the Muséum National d’Histoire Naturelle, in Paris (MNHN) (online appendix 2: http://dx.doi.org/10.5061/dryad.0c1p0).

(17)

Outgroup

In previous evolutionary scenarios, Oistoceras figulinum (Simpson 1855) has been consensually viewed as the “ancestor” of the Amaltheidae (Howarth 1958;

Meister 1988). We choose it as the outgroup to polarize characters of the analysis.

Oistoceras figulinum is morphologically close to the Amaltheidae and expresses very few autapomorphic states in our coding scheme.

Characters

Two data matrices are compiled:

- The first one (i.e. matrix a) includes ontogenetic information (online Appendix 3: http://purl.org/phylo/treebase/phylows/study/TB2:S16472?x- access-code=2c3cb8ceba8b4a3fe755da29ff1e1713&format=html). It contains 33 characters, out of which 13 characters result from the analysis of

ontogeny (three discrete ones and ten others resulting from coding of linear and saturating growth laws). To make the reading easier, this first matrix is called matrix a, is involved in analysis a and leads to tree a.

- The second matrix (i.e. matrix b) excludes the ontogenetic information (online appendix 4:

http://purl.org/phylo/treebase/phylows/study/TB2:S16472?x-access- code=2c3cb8ceba8b4a3fe755da29ff1e1713&format=html). The characters that express ontogeny in the first matrix have been replaced by a coding of

(18)

the adult stage morphology based on the same data. The total number of characters is thus reduced to 22. To make the reading easier, this second matrix is called matrix b, is involved in analysis b and leads to tree b. Both matrices and illustrations of characters are available on MorphoBank website (O'Leary and Kaufman 2012; online appendix 5:

http://morphobank.org/permalink/?P1271).

Coding of Ontogenetic Trajectories of Continuous Characters — Ontogenetic trajectories of the continuous characters are described from measurements of specimens selected to cover the maximum range of size and ontogenetic shape variation. Regressions have been performed on continuous character data using two different growth laws: linear growth law (Eq. 1) and saturating growth law (Eq. 2).

The best fitting model is retained in further analysis. Every ontogenetic stage is treated separately.

= +0 Equation 1

Equation 1. Linear growth law. σ0: values of Y (shape) at the beginning of the stage, k: growth rate.

=1(1 − ^(∝ Equation 2

Equation 2. Saturating growth law. σ1: values of Y (shape) when a (age) tends to the infinity which corresponds to the final value of the character, α: size at the beginning

(19)

of the stage, b: equal to 1 or -1, its sign determine if the function is increasing (+) or decreasing (-), k: growth rate.

Statistically significant parameters (i.e. from regressions explaining more than 95% of the variance) have been integrated into the matrix by using gap weighting (Thiele 1993; Wiens 2001) with 1000 states (online appendices 3-4:

http://purl.org/phylo/treebase/phylows/study/TB2:S16472?x-access-

code=2c3cb8ceba8b4a3fe755da29ff1e1713&format=html); others have been coded as question marks. Ontogenetic trajectories of these characters are available from

Morphobank website (online appendix 5: http://morphobank.org/permalink/?P1271).

Shape and Ornament of Ammonite Shells

In ammonoids, characters related to shell shape and ornaments are thought to be highly homoplastic but they may still be valuable if properly defined (Yacobucci 2012). Here, we consider some of the most common parameters in ammonoid

descriptions: the umbilical width (Uw), whorl breadth against whorl height (WbWh), the ribbing frequency (Ribs) and the tuberculation. Refining their treatment in

phylogenetic analyses in order to extract a reliable signal is a major goal. Some traits appear to be correlated; table 2 indicates features that have been combined to provide independent characters. Measurements have been made twice on each specimen, one at the smallest diameter available and the other at the greatest diameter. We

measured numerous specimens for each taxon to take into account the intraspecific variation. An alternative approach is to measure few, well-preserved specimens,

(20)

which is possible in ammonoids as their conchs preserve the whole ontogeny. The advantage is that the trajectories are complete and precise but the inconvenience is that these trajectories, even if constituting good representations of specimens are poor representations of the whole species.

Five characters have been coded for Uw. Ontogenetic trajectories follow a linear model for several taxa and a saturating growth law for the others. A challenge appears because ontogenetic trajectories of Uw have different shapes. Usually, shapes have to be similar to be comparable (Klingenberg 1998). For the phylogenetic inference, we decided to code the different shape of the ontogenetic trajectories as binary character. Then, we have refined the coding scheme in including parameters of each model for the taxa involved. Thus the five characters were the type of growth, the slope, the growth rate (k) for the two types of models and the shape σ1 for SG law (Fig. 2). The coding scheme for the whorl breadth against whorl height (WbWh) is identical to that of Uw. We specify that the type of growth (i.e. linear or saturated) has been coded only once for Uw and WbWh.

For the coding of the ribbing frequency, we also used a saturating growth law in a number of ribs per half-whorl divided by the diameter against diameter. Thus, the y axis is a measure of ribbing frequency (i.e. how many ribs can be secreted on a given diameter); it constitutes a parameter of shape independent of the diameter. The selection of this model is motivated by Meister (1988) claiming that the early and middle ontogeny of Amaltheidae was very polymorphic and that a stabilization

(21)

occurred in the late ontogeny. We code the growth rate and the shape σ1 which corresponds to the ribbing frequency that specimens are tending towards.

Tubercles are part of the shell ornamentation which is well-developed and diversified within the Amaltheidae. Several characters have been coded to express this tuberculation through ontogeny (characters 17, 19 and 20). They consist of the offset diameter of the Perlatum stage, the type of complex tubercles which

characterizes the Gibbosum stage and its expression (Fig. 1). The onset and offset of this stage is subject to covariation for some species but not all. We used a step matrix to code it. The detailed procedure is presented in the online appendix 6

(http://dx.doi.org/10.5061/dryad.0c1p0). Illustrations of every character are available from online appendix 5 (http://morphobank.org/permalink/?P1271).

Analyses

We choose parsimony as an optimality criterion even if this method to treat the ontogeny can also be used with model-based methods (e.g. Bayesian inference, maximum likelihood). Most parsimonious trees were sought using branch-and- bound algorithm implemented in PAUP* version 4.0b10 (Swofford 2003).

Trees indices — The level of homoplasy can be used to estimate the quality of the homology hypotheses. To compare phylogenies resulting from both analyses, we compute Consistency Index (CI), Retention Index (RI) and Rescaled Consistency Index (RC) which are usually calculated to quantify the rate of homoplasy (Farris

(22)

1989). Node supports were evaluated by bootstrap procedure using 1000 replicates (Felsenstein 1985).

Stratigraphic congruence — Considering the high quality of the fossil record of the amaltheids (Mattéi 1985), stratigraphic congruence between phylogenetic tree and stratigraphic occurrences of the taxa is relevant to evaluate tree robustness (Rouget 2002). Consistency indices have been created to assess the congruence between the tree and the occurrences of taxa in the fossil record. None of these

indices has received unanimous support. We use the Stratigraphic Consistency Index (Huelsenbeck 1994) which has the advantage of being based on the relative position of occurrences of taxa in the fossil record rather than on absolute ages which were not available. SCI is affected by factors such as the inference accuracy, tree balance, sampling intensities and the taxonomic status of the OTUs (Sidall 1997, Wagner and Sidor 2000, Lelièvre et al. 2008). Here, we use it to compare the resulting trees of the two analyses (i.e. with and without ontogenetic information). We expect that the higher SCI indicates the best inference and thus if our new approach helps in resolving the phylogeny. To go further, we also compare two interpretations of cladograms consistent with stratigraphy by proposing ancestor-descendant relationships and counting gaps. This recalls stratocladistics (Fisher 2008) as

ancestor-descendant relationships are proposed to reduce stratigraphic debt. A major difference is that we do not use these relationships or stratigraphic debt to find most parsimonious trees but only to compare MPTs obtained with the two different analyses (i.e. with and without ontogenetic information). As it is impossible to

(23)

estimate absolute ages of appearances of taxa, the unit of time is the ammonite subzone.

Testing dependences in amaltheids coding scheme

Detecting dependences between characters is a challenging investigation. In the literature, two approaches have been tried: before the analysis (e.g. O’Keefe and Wagner 2001; Maglia et al. 2004) and after (e.g. Felsenstein 1985b; Maddison 1990;

Cheverud et al. 1995; Abouheif 1999). A covariation may be due either to common history and/or to a relation of dependence. The problem of addressing character dependences after conducting the analysis is that clades on the tree can be based on character dependences rather than on congruence of independent characters. In this case, investigating for dependence is fully biased. If dependence relationships are investigated before the analysis and no covariation is observed, no relation of dependences can be suspected between characters under consideration. If two characters are dependent, we expect them to covary highly. Thus, in every pair of characters coding ontogeny, we performed tests. For two discrete characters, we performed χ² tests; for two continuous characters, we used correlation tests and for cases of one continuous and one discrete character, we performed ANOVAs.

R^ESULTS&D^ISCUSSION

Are Ontogenetic Variations of Features Dependent or Not?

(24)

Over the 91 pairs of characters coding ontogeny, 50 tests could be performed by respecting assumptions of the tests. 17 pairs of characters over the 50 show a significant covariation (i.e. p-values < 0.05), this is more than would be expected if all characters were independent and only a few would have been correlated by chance alone. However, this result confirms that the different features mostly have

independent variations in ontogeny and, thus, have to be coded separately. As noted by Maddison and Fitzjohn (2015), investigating correlation involving discrete

characters is challenging because they usually evolved only a few times. Thus they are unsuitable for usual statistics (e.g. Pagel's test) which will conclude that a

correlation between two traits exists whereas coincidence followed by co-inheritance may easily explain the pattern. Over the 17 pairs of characters showing significant covariation, only five pairs transform enough times (more than 4) to ensure that the dependence is meaningful. These five have been kept in the analysis because no hypotheses based on shell morphogenesis or any other a priori argument could explain these relationships. We assume that their covariations are based on shared evolutionary history.

As the problem of detecting whether a covariation results from congruence or dependence is still unresolved, we prefer a priori arguments which are independent of the dataset. Here for example, we investigated the relationship between starting and ending of the Gibbosum (table 2; online appendix 6:

http://dx.doi.org/10.5061/dryad.0c1p0). The a priori study of these timings allowed us to conclude that starts and ends were dependent for several taxa but not for A.

(25)

bechteri. We built character 20 in order to express both variation between taxa and the changing relation of dependences between starting and ending of the Gibbosum stage. This coding scheme allows us to describe a fragmentation event of the module composed by these two parameters between nodes 14 and 15.

Does Including Ontogenetic Information Increase the Robustness of the Phylogeny of Amaltheidae?

The analysis with ontogenetic information (i.e. analysis a) results in three MPTs (Fig. 5a). The analysis without ontogenetic information results in one MPT (Fig. 5b). Both results are included in a biostratigraphic scheme based on the literature (Fig. 5). The two cladograms present several topological differences. The major one concerns relations between species between node 1 and 9 on both trees a and b.

Tree a has better indices (i.e. CI, RI and RC) than tree b (table 3), but the differences remain insufficient to conclude which is the more accurate of the

ontogeny-derived matrices. For both trees a and b, bootstrap supports are relatively weak. However, mean support of tree a (i.e. 66.2) is higher than for tree b (i.e. 58.8).

The stratigraphic congruence is greater using ontogenetic information (SCI = 0.77) than without (SCI = 0.61). It could indicate that tree a is better than tree b even if both values are low. However, this index is sensitive to the tree balance (Sidall 1997, Wagner and Sidor 2000, Lelièvre et al. 2008) which is different for trees a (4 balanced nodes) and b (6 balanced nodes). Thus, SCI of tree b decreases in comparison to tree

(26)

a. Finally, tree a minimizes unsampled representative with a length of 6.5 subzones compared to tree b with 16 subzones (Fig. 6, Tab. 4). Tree a is thus much more consistent with stratigraphy. To conclude, even if all indices are quite close to each other, they are all better for tree a which displays a higher robustness in comparison to tree b. Thus, for the amaltheids our method improves the quality of the

phylogenetic hypothesis.

The type of growth (i.e. saturated or linear) and the sign of the slope have a high RI suggesting that the ontogenetic pattern of variation of Uw and WbWh has a powerful phylogenetic signal (Tab. 5). However, refining the coding scheme of Uw with growth rate both for linear model and saturating growth law provided very low RIs. This could be due either to the model used which does not adequately fit the data, or to the lack of phylogenetic signal of these characters. Uw coded on the adult has a higher RI than the mean RI of all characters coding Uw with our method.

However, the character Uw σ1 which tries to code the same information as Uw adult has a greater RI than the latter. This shows that, even if only the adult stage needs to be coded, modelling ontogenetic trajectory and coding the σ1 is better than just calculating the mean of adult specimens' values.

The RI of the character WbWh in adults is very weak (Tab. 5). By including the ontogenetic dimension, RIs of the three characters (σ1, k SGL, k linear) are higher.

For this shape parameter, ontogeny has to be taken into account to extract a relevant phylogenetic signal.

(27)

Finally, ribbing frequency in adults (analysis b, Tab. 5) has a slightly higher RI than Ribs σ1 (analysis a, Tab. 5). However, the growth rate (Ribs k, analysis a, Tab.5) has a greater RI. This result suggests that concerning ribbing frequency, the

phylogenetic signal is higher in the rate of expression during ontogeny rather than in absolute value of ribbing frequency. This emphasizes that trajectory can be more relevant than its products for phylogenetic reconstruction.

Heterochrony and the Evolution of Ontogeny

The method herein proposed is not disconnected from the classic

formalisation of heterochronies (e.g. Gould 1977; Alberch et al. 1979; Alba 2002).

Alberch et al. (1979) defined the different types of heterochronies and the

relationships between these and the controlling parameter (see tab. 1 in Alberch et al.

1979). We propose here a similar table (Tab. 6) depicting how to recognize them on ontogenetic trajectories, that is to say not only the controlling parameter but also the potential inferred variation in all parameters used according to the two models in our study.

The evolution of tubercles exhibits several heterochronies (Fig. 7). Tubercles appear at node 9 which includes taxa usually included in Amaltheus margaritatus and Pleuroceras. It consists of small tubercles developed at the middle of each rib, the so- called Perlatum stage (Frentzen 1937). This stage is expressed early during ontogeny (char. 17(1)). Starting from node 10, the Perlatum stage is followed by a second tuberculated stage which consists of strong tubercles (hypermorphosis, char. 18(1))

(28)

and extends up to an average diameter of 60 mm (char. 20(0)). We called this stage the Gibbosum stage (Frentzen 1937), even if our interpretation of secondary

homology leads us to consider that the Gibbosum stage overlaps the Coronatum and Spinosum stages of Frentzen (1937). The relative strength of this stage changes through nodes 12 and 13. First, a cyclic pattern of strength appears with a period of approximately six ribs (char. 19(2)). Then, this pattern becomes stronger (node 13) and results in one strong tubercle every six ribs (char. 19(3)). At node 14, the

Gibbosum stage extends through almost the entire ontogeny up to a diameter of 80 mm (char. 20(3)). After the extension of the Gibbosum stage, it will decrease at node 38 through a post-displacement leading to the beginning of the tuberculation at a diameter of 35 mm (char. 20(4)). This post-displacement makes a clear disjunction between the stages Perlatum and Gibbosum stages. At node 16, the Gibbosum stage disappears (char. 18(0)) and the Perlatum stage extends further in the ontogeny (17(2)). Finally, the Perlatum stage extends into the entire ontogeny at node 18 (char.

16(3)).

Including Ontogeny Reveals New Clades and New Evolutionary History

Amaltheid phylogenetic relationships have been proposed in several works (e.g. Frentzen 1937; Howarth 1958; Dagis 1976; Mattéi 1985; Meister 1988). As in many ammonoid groups, relationships within the Amaltheidae have been produced using a stratophenetic approach. Only one tree formalized as a cladogram, but drawn by hand and poorly resolved, has been so far published (Hardy 2012).

(29)

Consequently, comparisons are difficult. Most of the nodes are identical. All works agree with the placement of A. bifurcus at the base of amaltheids. A. stokesi is usually invoked as the ancestor of NW European species (i.e. node 7 in tree A, Fig. 5a).

However, these previous works did not take into account Boreal species (i.e. node 3 in tree A, Fig. 5a). The only work which includes the latter (Dagis 1976) proposes a superspecies regrouping A. bifurcus and A. stokesi giving rise to a part of the Boreal species, the clade regrouping A. wertheri and representatives of the genus

Amauroceras as well as NW European amaltheids. Our tree proposes a clade (node 7) placing A. wertheri and representatives of the genus Amauroceras (node 8) with classic NW European amaltheids (node 9). A. stokesi is placed at the base of a clade (node 2) with all Boreal species (node 3). However, the weakness of support of node 2 implies that A. stokesi should be placed in a trifurcation with nodes 3 and 7. Moreover,

optimisation of characters at node 1 is very similar to A. stokesi coding. This widespread and common species may be the ancestor of both the Boreal and NW European species.

A novelty in this phylogeny is the monophyly of the Boreal clade (node 3) that refutes the hypothesis of two or three origination events from A. margaritatus

proposed by Dagis (1976). This clade results from the coding of ontogeny as it is only supported by synapomorphies from the characters herein proposed (i.e. characters 1, 3, 4, 14 and 15). The major difference between the Boreal (node 3) and NW European clades (node 7) consists in their growth patterns of UW and WbWh which are

respectively linear versus saturated. The monophyly of the Boreal clade is in

(30)

agreement with paleobiogeographic reconstruction between provinces in the Pliensbachian (Dera et al. 2011). The Boreal domain is isolated from the NW

European one during the Spinatum Zone, leading to the differentiation of the Arctic province. The monophyly of the Siberian species confirms that amaltheids were cosmopolitan during the Margaritatus Zone, and then evolved independently in the Boreal and NW European domains.

CONCLUSION

Both theoretical and empirical arguments support the idea of including ontogenetic data as accurately as possible in phylogenies and avoiding artificial discretization of ontogeny. The results showed that our method results in a tree with better retention index, bootstrap supports and fit to stratigraphy. Few pairs of

characters significantly covary showing the relevance of including ontogenetic variations in coding schemes. Classic parameters of ammonite shell shape, usually considered to be very homoplastic are less subject to homoplasy when taking ontogeny into account. The method proposed here provides tools for coding a variety of morphological traits, whether discrete or continuous and to extract a high level of information from ontogeny. This method can be used with all taxa. It will be particularly useful when ontogenetic feature variations seem determinant in a group's evolution (e.g. arthropods) and for groups in which coding ontogeny with above-mentioned methods is challenging (e.g. molluscs, brachiopods). Homology has more to do with the path followed than the destination reached.

(31)

We used classic models discussed in Alberch et al. (1979) on classic ammonite features to produce new characters. Linear ontogenetic trajectories were obviously best described by linear regressions. Saturating growth law was well-suited for several shape parameters because amaltheids show a very variable phase before reaching stable values in late ontogeny as evoked by several authors (e.g. Meister 1988). However, as noted by Klingenberg (1998), none of Alberch et al. (1979) parameters match to parameters from other classic growth functions like logistic or Gompertz functions. The selection of the growth functions is thus a key step of the method which will determine the information extracted from the raw ontogenetic data and used for the phylogenetic inference. These growth models have mainly been established because of good fit to data. Subsequently, a large literature has attempted to describe new growth models with more biological meaning. In mollusks for example, new models try to explain the mechanical bases of shell formations (e.g. Moulton et al. 2012). More generally, West et al. (2001) developed a model based on energy needs between existing and new tissues. Next steps will be to find most meaningful models to use in phylogenetic inference. It seems obvious that using models with a biological base may provide coded parameters closer to what we consider to be a good homology hypothesis. This opens very exciting insights to associate ontogeny, phylogeny and biological/physical constraints in studies of shapes.

(32)

SUPPLEMENTARY MATERIAL

Supplementary material, including data files and/or online-only appendices, can be found:

- in the Dryad data repository at http://datadryad.org, doi: 10.5061/dryad.0c1p0 for appendices 1, 2 and 6

- from the treebase website at

http://purl.org/phylo/treebase/phylows/study/TB2:S16472?x-access-

code=2c3cb8ceba8b4a3fe755da29ff1e1713&format=html for appendices 3 and 4

- from the Morphobank website, project 1271:

http://morphobank.org/permalink/?P1271 for appendix 5.

F^UNDING

This work was supported by the MNHN, the UPMC and the French ministry of research through the funding of my PhD of which this study is a part.

ACKNOWLEDGMENTS

We would like to thank Pr. Martin Pickford and Pr. David Horne who kindly reviewed and improved the English of the manuscript. We are grateful to Dieter

(33)

Korn, Peter Wagner, Norman MacLeod (AE), Frank Anderson (EIC) and Thomas Near (EIC) for the very useful comments they provided on an earlier draft of the manuscript.

REFERENCES

Abouheif E. 1999. A method for testing the assumption of phylogenetic independence in comparative data. Evol. Ecol. Res. 1: 895-909.

Alba D.M. 2002. Shape and Stage in Heterochronic Models. In: Minugh-Purvis N., McNamara J.K., editors. Human Evolution through Developmental Change.

Baltimore:The Johns Hopkins University Press.

Alberch P., Gould S.J., Oster G.F., Wake D.B. 1979. Size and shape in ontogeny and phylogeny. Paleobiology. 5: 296-317.

Bardin J., Rouget I., Cecca F. 2013. Late Pliensbachian (Early Jurassic) ammonites from Lac de Charmes (Haute-Marne, France): Systematic, biostratigraphy and palaeobiogeography. Geodiversitas. 35: 309-334.

Bardin J., Rouget I., Cecca F. Data from: Ontogenetic data analyzed as such in phylogenies. Dryad Digital Repository. http://dx.doi.org/10.5061/dryad.0c1p0.

(34)

Bardin J., Rouget I., Yacobucci M., Cecca F. 2014. Increasing the number of discrete character states for continuous characters generates well-resolved trees that do not reﬂect phylogeny. Integr. Zool. 9: 531-541.

Blackstone N.W. 1987. Size and time. Syst. Zool. 36: 211-215.

Blackstone N.W., Yund P.O. 1989. Morphological variation in a colonial marine hydroid: a comparison of sizebased and age-based heterochrony. Paleobiology. 15: 1- 10.

Bruguière J.G. 1789-1832. Encyclopédie méthodique ou par ordre des matières:

Histoire naturelle des vers, des coquilles et des mollusques. Paris.

Buckman S. 1887-1907. Monograph of the ammonites of the Inferior Oolite Series.

Paleontographical Society, London.

Buckman S.S. 1909–1930. Yorkshire Type Ammonites. William Wesley and Son, London.

Cheverud J.M., Dow M.M., Leutenegger W. 1995. The quantitative assessment of phylogenetic constraints in comparative analyses: Sexual dimorphism in body weight among primates. Evolution. 39: 1335-1351.

Dagis A.A. 1976. Late Pliensbachian ammonites (Amaltheidae) of the North Siberia.

Academy of Sciences of the U.R.S.S. Siberian branch. Publ. Ho. Nauka, Siberian Branch, Novosibirsk Acad. Sci. USSR. 309: 79 p.

(35)

Dera G., Neige P., Dommergues J. L., Brayard A. 2011. Ammonite

paleobiogeography during the Pliensbachian-Toarcian crisis (Early Jurassic) reflecting paleoclimate, eustasy, and extinctions. Glob. Planet. Change, 78, 92–105.

Dommergues J.L. 1988. Can ribs and septa provide an alternative standard for age in ammonite ontogenetic studies? Lethaia. 21: 243–256.

Eble G.J. 2005. Morphological Modularity and Macroevolution: Conceptual and Empirical Aspects. In: Callebaut W., Rasskin-Gutman D., editors. Modularity:

understanding the development and evolution of natural complex systems.

Cambridge: MIT Press.

Efimova A.F., Kinasov V.P., Paraketsov K.V., Polubotko I.V., Repin Y.S., Dagis A.S.

1968. Field atlas of the Jurassic fauna and flora of the North-East of the USSR. District Publishing House, Magadan.

Emerson S.B. 1986. Heterochrony and frogs: the relationship of a life history trait to morphological form. Am. Nat. 127: 167-183.

Farris J.S. 1989. The retention index and the rescaled consistency index. Cladistics. 5:

417-419.

Felsenstein J. 1985. Confidence Limits on Phylogenies: An Approach Using the Bootstrap. Evolution. 39: 783-791.

Felsenstein J. 1985b. Phylogenies and the comparative method. Am. Nat. 125:1-15.

(36)

Fink, W. L. 1982. The conceptual relationship between ontogeny and phylogeny.

Paleobiology. 8: 254–264.

Fisher D.C. 2008. Stratocladistics: Integrating Temporal Data and Character Data in Phylogenetic Inference. Annu. Rev. Ecol. Evol. Syst. 39: 365-385.

Frentzen K. 1937. Ontogenie, Phylogenie und Systematik der Amaltheen des Lias Delta Südwestdeutschlands. Abh. Heid. 23: 1-136.

Futuyma D.J. 1998. Evolutionary biology, 3rd ed. Sinauer Associates, Sunderland, Massachusetts.

Germain D., Laurin M. 2009. Evolution of ossification sequences in salamanders and urodele origins assessed through event-pairing and new methods. Evol. Dev. 11:

170–190.

Goloboff P.A., Mattoni C.I., Quinteros, A.S. 2006. Continuous characters analyzed as such. Cladistics. 22: 589-601.

Gould S.J. 1966. Allometry and size in ontogeny and phylogeny. Biol. Rev. 41: 587- 640.

Gould S.J. 1977. Ontogeny and Phylogeny. Harvard University Press, Cambridge.

Guex J., Koch A., O'Dogherty L., Bucher H. 2003. A morphogenetic explanation of Buckman's law of covariation. B. Soc. Geol. Fr. 174: 603-606.

(37)

Hall, B.K. 2000. Evo-devo or devo-evo — does it matter? Evol. Dev. 2: 177-178.

Hall B.K., Miyake T. 1995. How do embryos measure time? In: McNamara K.J., editor. Evolutionary Change and Heterochrony. Chichester: John Wiley & Sons.

Hammer Ø., Bucher H. 2005. Buckman's law of covariation - a case of proportionality. Lethaia. 38: 67-72.

Hardy C. 2012. Distribution phylogénétique des extinctions dans le registre fossile:

apports à la compréhension de la dynamique de la biodiversité. [Ph.D. dissertation].

Université de Bourgogne, Dijon.

Howarth M.K. 1955. Domerian of the Yorkshire coast. P. Yorks. Geol. Soc. 30: 147–

175, pls. 12, 13.

Howarth M.K. 1958. A monograph of the ammonites of the liassic family Amaltheidae in Britain. Palaeontological Society, London.

Huelsenbeck J.P. 1994. Comparing the straigraphic record to estimates phylogenetic analysis. Syst. Zool. 40: 257-270.

Huelsenbeck J.P., Nielsen R. 1999. Effects of nonindependent substitution on phylogenetic accuracy. Syst. Biol. 48:317–328.

Humphries C.J. 1988. Ontogeny and systematics. Columbia University Press, New York.

(38)

Jeffery J.E., Bininda-Emonds O.R.P., Coates M.I., Richardson M.K. 2002. Analyzing evolutionary patterns in amniote embryonic development. Evol. Dev. 4: 292–302.

Jeffery J.E., Bininda-Emonds O.R.P., Coates M.I., Richardson M.K. 2005. A new technique for identifying sequence heterochrony. Syst. Biol. 54: 230–240.

Jeffery J.E., Richardson M.K., Coates M.I., Bininda-Emonds O.R.P. 2002. Analyzing developmental sequences within a phylogenetic framework. Syst. Biol. 51: 478–491.

Klingenberg C.P. 1996. Individual variation of ontogenies: a longitudinal study of growth and timing. Evolution. 50: 2412-2428.

Klingenberg C.P. 1998. Heterochrony and allometry: the analysis of evolutionary change in ontogeny. Biol. Rev. 73: 79-123.

Klingenberg C.P., Spence J.R. 1993. Heterochrony and allometry: lessons from the water strider genus Limnoporus. Evolution. 47: 1834-1853.

Knauss M.J., Yacobucci M.M. 2014. Geographic Information Systems technology as a morphometric tool for quantifying morphological variation in an ammonoid clade.

Paleontol. Electron. 17: 20A.

Koenemann S., Schram F.R. 2002. The limitations of ontogenetic data in phylogenetic analyses. Contrib. Zool. 71:47–65.

Korn D. 2012. Quantification of ontogenetic allometry in ammonoids. Evol. Dev. 14:

501-514.

(39)

Lange W. 1932. Über ein Hammatoceras und einen Amaltheenvorläufer

(Proamaltheus wertheri Gen. Nov. Sp. Nov.) aus dem Lias Gamma+Delta von Werther in Westphalen. Z. Dtsch. Geol. Gesell. 84: 235-241.

Laurin M., Germain D. 2011. Developmental characters in phylogenetic inference and their absolute timing information. Syst. Biol. 60: 630–644.

Leigh S.R., Shea B.T. 1996. Ontogeny of body size variation in African apes. Am. J.

Phys. Anthropol. 99: 43-65.

Lelièvre H., Zaragüeta-Bagils R., Rouget I. 2008. Temporal information, fossil record and phylogeny. C. R. Palevol. 7: 2736.

Mabee P.M. 1993. Phylogenetic interpretation of ontogenetic change: Sorting out the actual and artefactual in an empirical case study of centrarchid fishes. Zool. J. Linn.

Soc. 107: 175–291.

Mabee P.M., Humphries J. 1993. Coding polymorphic data: examples from allozymes and ontogeny. Syst. Biol. 42: 166-181.

Maddison W.P. 1990. A method for testing the correlated evolution of two binary characters: Are gains or losses concentrated on certain branches of a phylogenetic tree? Evolution. 44: 539-557.

Maddison W.P., Fitzjohn R.G. 2015. The Unsolved Challenge to Phylogenetic Correlation Tests for Categorical Characters. Syst. Biol. 64: 127-136.

(40)

Maglia A.M., Ghatti V.R., Leopold J.L. 2004. Identifying Character Non-

Independence in Phylogenetic Data Using Data Mining Techniques. In: Chen Y.P.P., editor. Proceedings of the Second Asia-Pacific Bioinformatics Conference

(APBC2004), Dunedin, New Zealand: ACS.

Mattéi J. 1985. Application de méthodes d’analyse globale à l’étude des Amaltheidae du bassin sédimentaire des Causses (Carixien supérieur et Domérien). Editions du Centre National de la Recherche Scientifique, Paris, 146 p.

Maubeuge P.L. 1951. Sur quelques ammonites rares ou nouvelles du Lias moyen de Belgique. Bull. Inst. R. Sci. Nat. Belg. 27: 1-6.

McKinney M.L. 1986. Ecological causation of heterochrony: a test and implications for evolutionary theory. Paleobiology. 12: 282-289.

McKinney M.L. 1988. Classifying heterochrony: allometry, size, and time. In:

McKinney M.L., editor. Heterochrony in Evolution: A Multidisciplinary Approach.

New York: Plenum Press.

McKinney M.L., McNamara K.J. 1991. Heterochrony: the Evolution of Ontogeny.

Plenum Press, New York.

McNamara K. 1996. Evolutionary change and heterochrony. John Wiley and Sons, New York.

(41)

Meister C.1988. Ontogenèse et évolution des Amaltheidae (Ammonoidea). Eclogae Geol. Helv. 81: 763-841.

Minelli A., Brena C., Deflorian G., Maruzzo D., Fusco, G. 2006. From embryo to adult—beyond the conventional periodization of arthropod development. Dev.

Genes. Evol. 216: 373–383.

Morard A. 2004. Les événements du passage Domérien-Toarcien entre Téthys occidentale et Europe du Nord-Ouest. [Ph.D. disseration]. University of Lausanne, Lausanne.

Moulton D.E., Goriely A., Chirat R. 2012. Mechanical growth and morphogenesis of seashells. J. Theor. Biol. 311: 69–79.

O’Keefe F.R., Wagner P.J. 2001. Inferring and testing hypotheses of cladistic character dependence by using character compatability. Syst. Biol. 50: 657-675.

O'Leary, M. A., and S. G. Kaufman. 2012. MorphoBank 3.0: Web application for morphological phylogenetics and taxonomy. http://www.morphobank.org.

Phillips J. 1829. Illustrations of the geology of Yorkshire. John Murray, York.

Raff R. 1996. The shape of life. University of Chicago Press, Chicago, Illinois.

Reiss J.O. 1989. The meaning of developmental time: a metric for comparative embryology. Am. Nat. 134: 170-189.

(42)

Rohlf F.J. 1998. On applications of geometric morphometrics to studies of ontogeny and phylogeny. Syst. Biol. 47: 147-158.

Rouget I. 2002. Reconstruction phylogénétique chez les ammonites : confrontation des approches cladistique et stratigraphique. Le cas des Dayiceras (Ammonitina, Eoderoceratoidea). [Ph.D. disseration]. Université de Bourgogne, Dijon.

Schlosser G., Wagner G.P. 2004. Introduction: The modularity concept in

developmental and evolutionary biology. In: Schlosser,G., Wagner G. P., editors.

Modularity in development and evolution. Chicago: University of Chicago Press.

Schlotheim E.F.V. 1820. Die Petrefactenkunde auf ihrem jetzigen Standpunkte durch die Beschreibung seiner Sammlung versteinerter und fossiler Überreste des Thier- und Pflanzenreichs der Vorwelt. Becker, Gotha.

Sidall M.E. 1997. Stratigraphic Indices in the Balance: A Reply to Hitchin and Benton.

Syst. Biol. 46: 569-573.

Simpson M. 1855. The fossils of the Yorkshire Lias, described from nature. Whittaker, Whitby & London.

Smith K.K. 1996. Integration of craniofacial structures during development in mammals. Am. Zool. 36: 70–79.

Smith K.K. 2001. Heterochrony revisited: the evolution of developmental sequences.

Biol. J. Linn. Soc. 73: 169–186.

(43)

Sowerby J. 1812–1829. The mineral conchology of Great Britain, 1–3. Meredith, London.

Steyer J.S. 2000. Ontogeny and phylogeny in temnospondyls: a new method of analysis. Zool. J. Linn. Soc. 130: 449–467.

Strauss R.E. 1987. On allometry and relative growth in evolutionary studies. Syst.

Zool. 36: 72-75.

Strauss R.E. 1990. Heterochronic variation in the developmental timing of cranial ossification in poecilid fishes (Cyprinodontiformes). Evolution. 44: 1558-1567.

Swofford D.L. 2003. Phylogenetic Analysis Using Parsimony (*and Other Methods).

Ver. 4. Sinauer Associates.

Thiele K. 1993. The holy grail of the perfect character: The cladistic treatment of morphometric data. Cladistics. 9: 275-304.

Tintant H., Gauthier J., Lacroix, L. 1961. Les Amalthéidés de Côte-d'or et leur répartition stratigraphique. Bull. Sci. Bourg. 20: 137-161.

Tuchkov I.I. 1954. Jurassic ammonites and belemnites of northeast USSR. Materials on geology and mineral resources of North-East of USSR. 8: 98-125 [in Russian].

Velhagen W.A.J. 1997. Analyzing developmental sequences using sequences units.

Syst. Biol. 46: 204–210.

(44)

Wagner P.J. 1999. The utility of fossil data in phylogenetic analyses: a likelihood example using Ordovician-Silurian species of the Lophospiridae (Gastropoda:

Murchisoniina). Am. Malacol. Bull. 15: 1-31.

Wagner P.J., Sidor C.A. 2000. Age rank: clade rank metrics - sampling, taxonomy, and the meaning of “stratigraphic consistency”. Syst. Biol. 49: 463-479.

West G.B., Brown J.H., Enquist B.J. 2001. A general model for ontogenetic growth.

Nature. 413: 628–631.

Westermann G.E.G. 1966. Covariation and taxonomy of the Jurassic ammonite Sonninia adicra (Waagen). Neues. Jahrb. Geol. P-A. 124: 289–312.

Wiens J.J. 2001. Character analysis in morphological phylogenetics: Problems and solutions. Syst. Biol. 50: 689-699.

Wiens J.J., 2005. Can incomplete taxa rescue phylogenetic analyses from long-branch attraction? Syst. Biol. 54: 731–742.

Wolfe J.M., Hegna T.A. 2013. Testing the phylogenetic position of Cambrian pancrustacean larval fossils by coding ontogenetic stages. Cladistics. 30: 1-25.

Yacobucci M.M. 2012. Meta-analysis of character utility and phylogenetic information content in cladistic studies of ammonoids. Geobios. 45: 139-143.

(45)

Young G., Bird J. 1828. A geological survey of the Yorkshire Coast: describing the strata and fossils occurring between the Humber and the Tees, from the German Ocean to the Plain of York. 2nd edition. Kirby, Whitby.

(46)

Figures captions

Figure 1. Three ontogenetic trajectories of the umbilical width (Uw) and the tuberculation for the species: A. gibbosus, A. salebrosus and A. bechteri. U: umbilical length, D: diameter, P: Perlatum stage (black bars), C: Gibbosum stage (grey bars).

One should note that the two morphological traits show different variation through ontogeny. Even one character, here the tuberculation, can show different variation in its own trajectory. This justifies the coding of ontogenetic trajectories for each

morphological trait and that heterochronic trends can be different even in one ontogenetic trajectory.

Figure 2. Three ontogenetic trajectories and the corresponding coding scheme of the umbilical width (Uw) for the species: A. gibbosus, A. talrosei and P. solare. U: umbilical length, D: diameter, k: growth rate, σ1: final value of Uw. Growth laws are used to modelled ontogenetic trajectories; this allows the extraction of more information about the morphological traits.

Figure 3. Ontogenetic trajectories of three types of characters: a. binary, b. multistate and c. continuous. 0, 1, 2: discrete states, σ0: initial value of the trait, σ1: final value of the trait, α: age of onset, β: age of offset.

Figure 4. Logic chart of the method of coding ontogeny. Parameters from examples are those of models herein used to represent ontogenetic trajectories.