Introduction
Numerous arthropods are known to display ommatidia, which are clustered together to form compound eyes (Oakley Reference Oakley2003; Waloszek et al. Reference Waloszek, Chen, Maas and Wang2005; Harzsch and Hafner Reference Harzsch and Hafner2006; Nilsson and Kelber Reference Nilsson and Kelber2007; Aberhan et al. Reference Aberhan, Nürnberg and Kiessling2012), including the extinct trilobites (Clarkson Reference Clarkson1979; Levi-Setti et al. Reference Levi-Setti, Clarkson and Horwáth1998; Clarkson et al. Reference Clarkson, Levi-Setti and Horvath2006). Among these, phacopids are known to display a variable number of eye lenses or dorso ventral files (Thomas Reference Thomas1998; Crônier and Clarkson Reference Crônier and Clarkson2001), thereby providing variation for evolution and natural selection to act upon such a character (Darwin and Wallace Reference Darwin and Wallace1958; Hallgrímsson and Hall Reference Hallgrímsson and Hall2005; Hunt Reference Hunt2007). Separating the variation resulting from a phenotypic heredity of the variation resulting from a plastic environmental response is, however difficult, notably in extinct organisms (Urdy et al. Reference Urdy, Goudemand, Bucher and Chirat2010). Furthermore, knowledge of intraspecific variability is an important requirement for taxonomic studies as well as for studying of spatio temporal patterns of diversity through time and their relation with environmental changes (Crônier et al. Reference Crônier, Feist and Auffray2004; Brayard et al. Reference Brayard, Bucher, Escarguel, Fluteau, Bourquin and Galfetti2006; De Baets et al. Reference De Baets, Klug and Monnet2012; Månsson and Clarkson Reference Månsson and Clarkson2012).
In trilobites, only a few studies documented an intraspecific variability in the number of files or lenses (for example Novák in Perner Reference Perner1918; Clarkson Reference Clarkson1966a, Reference Clarkson1966b, Reference Clarkson1971; Clarkson and Tripp Reference Clarkson and Tripp1982; Thomas Reference Thomas1998; Budil Reference Budil1999; Crônier and Clarkson Reference Crônier and Clarkson2001). Phacopid trilobite species exhibit a relatively moderate morphological intraspecific variation in the number of dorso ventral files that can be age related (Crônier and Clarkson Reference Crônier and Clarkson2001) or not (Crônier et al. Reference Crônier, Feist and Auffray2004). In particular, some previous studies have shown that eyes in phacopine populations from deeper environments tend to be smaller and to have fewer lenses/files than those from shallower environment (Crônier et al. Reference Crônier, Feist and Auffray2004; Feist et al. Reference Feist, McNamara, Crônier and Lerosey-Aubril2009). However, the intraspecific variability in these phacopid trilobites has hardly been studied in the respect to reveal a continuous spread of morphotypes with a common intermediate shape and a bimodal or polymodal distribution.
A large sample of well preserved phacopid trilobites from the Barrandian area gives us the opportunity to study the pattern of the variation in the number of dorso ventral files of eye lenses for two Devonian species and to quantify their intraspecific variability. In order to better understand the developmental patterning, to investigate the frequency distribution change through ontogeny and to explore the evolutionary significance of intraspecific variability in the lens number, we compared more than 120 individuals in two Barrandian species.
Additionally, we investigated possible relations of phenotypic variation with environmental characteristics of the studied area in an attempt to test whether the bimodality can persist and possibly be a factor in enabling organisms to adapt to environmental changes. For example, in organisms affected by sea level fluctuations, an eye lens polymorphism matching altered environments in space and time is of selective importance. Environmental fluctuations are known to influence the evolution of natural populations and to trigger long term evolutionary diversification. According to Schluter (Reference Schluter2000), the diversification is driven by an adaptation to different ecological conditions. Thus, a divergent/disruptive selection causes adaptive divergence between conspecific groups that occupy different environments or use different resources, and suggested by a bimodal distribution of morphological traits. Such bimodal distribution can bear upon or control the origin of new species. Numerous organisms are able to maintain a high genetic variability, and a rapid alteration of phenotypic traits is necessary for their survival as a response to environmental fluctuations (Ramel Reference Ramel1998; Frankham et al. Reference Frankham, Ballou and Briscoe2002). Adaptive divergence and ecological speciation are believed to take part in the evolution of biodiversity (Skúlason and Smith Reference Skúlason and Smith1995; Schluter Reference Schluter2000; Kristjánsson et al. Reference Kristjánsson, Skúlason and Noakes2002; Rundle and Nosil Reference Rundle and Nosil2005).
Material and Methods
Data collection
For our study, we analyzed individuals of two phacopid species, Prokops prokopi (Chlupáč, Reference Chlupáč1971) and Pedinopariops insequens (Chlupáč, Reference Chlupáč1977).
Prokops prokopi is locally abundant in the current lowermost Emsian (=classical upper Pragian; Chlupáč et al. Reference Chlupáč, Havlíček, Kříž, Kukal and Štorch1998) in the near vicinity of Prague, in the Barrandian area (Figs. 1 and 2). This species is represented by two populations in our study: (1) the first population, of 22 individuals, comes from the Malá Chuchle locality (old quarry, 50°1'30.399”N, 14°22'46.758"E, in a suburb SW of Prague; Figs. 1, 2 herein; this locality has been discussed especially by Růžička Reference Růžička1940; Chlupáč Reference Chlupáč1957, Reference Chlupáč1983; Havlíček and Vaněk Reference Havlíček and Vaněk1998; Vaněk Reference Vaněk1999; Frýda et al. Reference Frýda, Ferrová and Frýdová2013), and (2) the second population originates from the Hlubočepy/St-Prokop's Quarry (called also Schwarzenberg's Quarry, 50°2'37.876"N, 14°22'34.475"E, in the Prokopské Údolí Valley, SW suburb of Prague; Figs. 1, 2 herein; this locality has been discussed especially by Woldřich Reference Woldřich1919; Růžička Reference Růžička1941; Chlupáč Reference Chlupáč1957, Reference Chlupáč1983, Reference Chlupáč1988, Reference Chlupáč1999; Kříž Reference Kříž1999; Havlíček and Vaněk Reference Havlíček and Vaněk1998; Vaněk Reference Vaněk1999; Frýda et al. Reference Frýda, Ferrová and Frýdová2013). For this study, 12 individuals were used, all coming from several paleontological collections (Růžička, Vaněk, Chlupáč amongst others) housed in the National Museum Prague and in the Czech Geological Survey, Prague. All studied material originates from “white beds,” i.e. the decalcified limestones (see Suchý Reference Suchý2002), from the Dvorce and Prokop Limestone facies, Prokop Limestone subfacies of Praha Formation (Fig. 2).
Figure 1 A sketch map of the distribution of the Devonian rocks in the Barrandian area. The position of both studied localities is indicated by stars.
Figure 2 Pragian to Eifelian stratigraphy of the Prague Basin (after Chlupáč et al. 1998, modified by Budil et al. Reference Budil, Hörbinger and Mencl2009). Conodont biozones of Pragian after Slavík (Reference Slavík2004a, Reference Slavík2004b), Slavík et al. (Reference Slavík, Valenzuaela-Ríos, Hladil and Carls2007), and Carls et al. (Reference Carls, Slavík and Valenzuela-Ríos2008). Upper Emsian to Eifelian biozones after Gradstein et al. (Reference Gradstein, Ogg, Schmitz and Ogg2012), Ferová et al. (2012), Klapper and Vodrážková (Reference Klapper and Vodrážková2013) and Vodrážková et al. (Reference Vodrážková, Frýda, Suttner, Koptíková and Tonarová2013). Key to the localities: (1) Malá Chuchle; (2) St. Prokop’s Quarry; (3) Praha-Holyně Section.
Prokops prokopi (Fig. 3A,B) is clearly distinguished from the other phacopids of the Prague Basin, including the ancestral and partially contemporaneous Prokops hoeninghausi (Barrande, Reference Barrande1846) by the characteristic configuration of the visual surface that is reduced and asymmetrically kidney shaped, with 19–34 lenses as well as by vaulted glabella and steep genae.
Figure 3 A–C, Prokops prokopi (Chlupáč, Reference Chlupáč1971) from the Malá Chuchle locality, in the vicinity of Praha: (A, B) cephalon in dorsal and lateral views respectively with 13 dorso ventral files, C, detail of an eye with 13 dorso ventral files. D–F, Prokops prokopi (Chlupáč, Reference Chlupáč1971) from the Hlubočepy/St-Prokop's Quarry, in the vicinity of Praha: D–E, cephalon in dorsal and lateral views respectively with 8/9 dorso ventral files, F, detail of an eye with nine dorso ventral files. G–K, Pedinopariops insequens (Chlupáč, Reference Chlupáč1977) from the Praha-Holyně Section, at a SW suburb of Praha: G–H, cephalon in dorsal and lateral views respectively with 18 dorso ventral files, I, detail of an eye with 18 dorso ventral files, J, cephalon in dorsal view with 12 dorso ventral files, K, detail of an eye with 12 dorso ventral files. L, Location of 21 landmarks on the cephalon. Scale: 2 mm.
Pedinopariops insequens is locally abundant in the interval of the uppermost Emsian to lowermost Eifelian in the near vicinity of Prague. This species is represented by 142 specimens in our study, including 42 individuals with well preserved eyes, coming from one population from the Praha Holyně Section in a small quarry NE of the village of Holyně and W of the Prastav quarry (50°1'57.627"N, 14°21'5.471"E) in a SW suburb of Prague (Figs. 1, 2 herein; this locality has been discussed especially by Bouček Reference Bouček1931; Růžička Reference Růžička1941; Chlupáč Reference Chlupáč1959, Reference Chlupáč1985, Reference Chlupáč1988, Reference Chlupáč1999; Kříž Reference Kříž1999; Frýda et al. Reference Frýda, Ferrová and Frýdová2013). All studied material comes from “white beds,” i.e. the decalcified limestones (see above), and was collected by J. Vaněk, partially also by F. Hanuš (both collections are housed in the Czech Geological Survey, Prague and in the National Museum, Prague). This locality yielded a rich fauna comprising about 350 invertebrate species (Bouček Reference Bouček1931; Růžička Reference Růžička1941) and comes from topmost part of the Třebotov Limestone (uppermost Emsian/lowermost Eifelian). The exact dating of these strata has been a subject of discussion (Chlupáč Reference Chlupáč1959; Chlupáč Reference Chlupáč1988, Reference Chlupáč1999; Kříž Reference Kříž1999; Frýda et al. Reference Frýda, Ferrová and Frýdová2013). The boundary between the Lower and Middle Devonian was placed by Chlupáč et al. (Reference Chlupáč, Lukeš and Zikmundová1979) and Chlupáč (Reference Chlupáč1982, Reference Chlupáč1993) within the topmost Třebotov Limestone, 2.8m below the base of the overlying Choteč Limestone, in the nearby Prastav Quarry (Chlupáč Reference Chlupáč1985). Berkyová (Reference Berkyová2009) moved this boundary here to ca. 5 m below the base of the Choteč Limestone, e.g. immediately above the supposed stratigraphical position of the “white beds” outcrop. Frýda et al. (Reference Frýda, Ferrová and Frýdová2013) suggest that the weathered “white beds” at Holyně may belong already to the Polygnathus costatus partitus conodont Biozone of early Eifelian age.
By its characteristic configuration of the dorsal exoskeleton, Pedinopariops insequens (Fig. 3C,D) is regarded as closely related to P. superstes (Barrande, Reference Barrande1852), a stratigraphically older species from the Daleje Shale and lower part of the Třebotov Limestone (upper Emsian).
Methods and Measured Parameters
The studied phacopids have been investigated using quantitative methods. Lenses and dorso ventral files for each preserved eye have been numbered using binocular microscopes NIKON MCZ 1500 and Olympus. Cephalic width and cephalic length (in mm) have been measured. Additionally, in order to describe the morphology of these organisms, a set of 21 landmarks was selected on the cephalon to summarize the general cephalic shape, along with the eye, for 33 individuals of Prokops prokopi and 88 individuals of Pedinopariops insequens. Landmarks are points among which the number and the position should be strictly identical for all individuals (Bookstein Reference Bookstein1991). The x- and y-coordinates of 21 landmarks were obtained by an optical image analyzer (TPSdig) and were automatically extracted on cephala (Fig. 3E). Only the two dimensional projections of the dorsal views have been considered here. The Barrandian trilobite material (see above) is suitable for such analyses because this material is quite abundant, preserved in three dimensions, and came from a restricted stratigraphic interval, thus minimizing time averaging.
Frequency histograms
Frequency distributions have been investigated by both ‘traditional’ and smoothed histograms (via kernel density estimation, KDE) in order to detect any multimodality in the distribution. While the ‘traditional’ parametric histograms represent categorical variables and depend on the fixed width of the bins (divided into equal subintervals in which the whole data interval is divided) and the end points of the bins, the nonparametric Kernel density estimators represent noncategorical variables, which can reveal asymmetry and mutimodality (Salgado-Ugarte et al. Reference Salgado-Ugarte, Shimizu and Taniuchi1994, Reference Salgado-Ugarte, Shimizu, Taniuchi and Matsushita2002). Plots histograms and KDE were achieved using the data analysis software PAST v2.08 (Hammer et al. Reference Hammer, Harper and Ryan2001; Hammer and Harper Reference Hammer and Harper2006). KDE uses an optimal Gaussian bin width according to the rule given by Silverman (Reference Silverman1986).
Geometric morphometry
To complete the biometrical study, a geometric morphometric method based on a set of landmarks has been used for an accurate description of specific features such as cephala. These landmarks are tracked down in a system of axes in Cartesian coordinates and have been compared by fit methods (see Adams et al. Reference Adams, Rohlf and Slice2004 for details on procedures).
Each cephalon is defined by a configuration of landmarks containing both size and shape information. Size being independently analyzed; all configurations of landmarks were scaled to unit centroid size. The centroid size is the square root of the mean squared distance between the centroid and each landmark of the configuration (Bookstein Reference Bookstein1991). The scaled configurations are then superimposed, i.e., translated and rotated through the Generalized Least Square (GLS) Procrustes procedure (Rohlf and Slice Reference Rohlf and Slice1990). The superimposed configurations constitute a data set corresponding to Procrustes residuals, i.e., x- and y-coordinates of each landmark of configurations superimposed on a reference configuration. This method requires: (1) the translation of all configurations in order to superimpose their centroid and (2) the rotation of all configurations in order to minimize the scatter of landmarks, i.e., the sum of squared distances between landmarks of each configuration to the corresponding landmarks of a consensus configuration, here the average of all configurations.
The GLS superimposition is computed by the tpsSuper software v. 1.14 (Rohlf Reference Rohlf2003a). Five replicating landmark measurements carried out on the same individual demonstrate that within individual measurement error is negligible. Univariate Analyses of Variance (ANOVAs) on each coordinate of landmarks indicate no significant morphological differentiation among replicas (P=0.05 NS).
To examine the major trend of shape variation and to estimate the intraspecific variability, the residual matrix is then subjected to a Thin Plate Spline (TPS) procedure, i.e., a Relative Warp Analysis (RWA, see Rohlf Reference Rohlf1993), which fits an interpolated function to the consensus configuration. This is a principal components analysis of the partial warp scores that describe localized departures from the average configuration (Bookstein Reference Bookstein1991; Rohlf Reference Rohlf1993). The RWA allows the emergence of principal independent axes according to a decreasing order of importance in the explanation of the shape variability. Significant RWs define a shape space into which it is possible to locate each individual according to its coordinates on these RWs. Analysis of variance is then applied on the matrix of individual scores on the significant RW in order to assess the shape difference between the populations/groups. Moreover, thin plate spline deformation grids have been plotted on each extremity of the first Significant RWs. The RWA is computed by the tpsRelw program (Rohlf Reference Rohlf2003b).
Additionally, in order to compare size and shape variations of these individuals, a morphological space defined by both the size, estimated by the centroid size of cephala, and the shape, estimated by the first axis of the relative warp, was constructed to visualize the relationships between size and shape and to determine the correlation between the scores on the first axis and centroid size. The size with the shape describes the complete morphology of the cephala.
Range of Intraspecific Variability
In order to determine trends in the amount of deviation from the mean lens number and the global shape variability, the coefficient of variation (CV) that is a measure of the range of intraspecific variability of these morphological characters, has been calculated and compared among populations by using an F-max test. For the global shape variability, the degrees of freedom have been modified to take into account the number of coordinates and the loss of four degrees of freedom due to the Procrustes Superimposition procedure (Klingenberg and McIntyre Reference Klingenberg and McIntyre1998).
Variability within and between populations of Prokops prokopi
Frequency Distributions
The frequency distribution of the number of dorso ventral files of lenses for Prokops prokopi from the Malá Chuchle locality and the Hlubočepy/St-Prokop’s quarry is presented in figure 4A. Estimations from the ‘traditional’ and the smoothed (Gaussian KDE) histograms result in a multimodal distribution; at least two modes can be distinguished (Fig. 4A). Additionally, bivariate plots of lenses versus files number, of cephalic width (tr.) versus cephalic length (sag.) and, of files versus cephalic width (tr.) are respectively presented in figure 4B, 4C, and 4D.
Figure 4 A, The frequency distribution of the file number of lenses for all individuals of Prokops prokopi (Chlupáč, Reference Chlupáč1971). The ‘traditional’ histogram and the Kernel density estimate distribution of file number using the optimal Gaussian bin width. The frequency distribution shows two modes, i.e. a bimodality in the file number of lenses; B, scatter plot of lens number versus file number for all individuals; C, scatter plot of cephalic width (tr.) in mm versus cephalic length (tr.) in mm for all individuals showing a discontinuous and a typical isometric distribution for each population; D, scatter plot of file number versus cephalic width (tr.) in mm for all individuals; E, location of 33 cranidia/cephala of Prokops prokopi (Chlupáč, Reference Chlupáč1971) in the morphospace defined according to the first two relative warps RW1 and RW2. Axes RW1 and RW2 represent 25.5% and 18.5%, respectively, of the total variance. Thin plate spline deformation grids depicting patterns of shape change between the both populations of Prokops prokopi. Double arrows in (A) and (B) underline the gap.
In comparing two independent groups, an F-test provides a measure of the probability that they have the same variance or regression model (Lomax Reference Lomax2007). However, in our groups, the cephalic length (sag.) does not meet the assumptions of normality of distribution and / or homogeneity of variances. Therefore, a nonparametric Mann-Whitney U-test was used (Mann and Whitney Reference Mann and Whitney1947), and shows that the mean of this variable does not differ between the two studied groups (Z=1.279, p=0.2NS). And, no significant differences between the ontogenetic trajectories of Prokops prokopi from the Malá Chuchle locality and the Hlubočepy/St-Prokop’s quarry, were detected by an F-test on the slopes (F=1.561, p=0.22NS).
The size range of cephala in the Malá Chuchle locality as in the Hlubočepy/St-Prokop’s quarry is almost the same from 3.5 to 17 mm for the length and from 4.5 to 22 mm for the width (Fig. 4C). The ANOVAs performed on the cephalic width and length confirms the existence of a size unity among groups previously defined according to their provenance. All tests are insignificant (Table 1A,B). Furthermore, in the Malá Chuchle locality, the lowest values of lens file number are 12, with the maximum values at 14. In the Hlubočepy/St-Prokop’s quarry, the lowest values are seven, with the maximum values at nine, with one exception at 13 (Fig. 4D). Apart from the exception, in Malá Chuchle as in Hlubočepy/St-Prokop´s quarry, the lens file number range is three. ANOVAs performed on the file and lens number confirms the existence of a differentiation among groups previously defined according to their provenance. All tests are significant (Table 1C–D). Thus, the reduction in the lens files number in the St-Prokop population is independent of body size. Additionally, the coefficient of variation in the number of eye lenses is more important in this St-Prokop population (CV=18.43%) than in the Malá Chuchle population (CV=3.50%). The comparison of the coefficient of variation by a F-max test indicates a significant heterogeneity of variance among groups (F vmax/vmin=5.27; ddl1=13; ddl2=7; p=0.018) due to the presence of one individual with 13 lenses in the St-Prokop’s quarry.
Table 1 Univariate Analyses of Variance performed (A, B) on cephalic width (in mm) and cephalic length (in mm) showing no significant morphological differentiation between the two populations of Prokops prokopi (Chlupáč, Reference Chlupáč1971); C, D, on the file and lens number showing a significant differentiation among the two populations of Prokops prokopi (Chlupáč, Reference Chlupáč1971); E–G, on the first three significant axes showing a significant morphological differentiation among the two populations of Prokops prokopi (Chlupáč, Reference Chlupáč1971) according to RW1 and RW2.
Morphological space
Morphological relationships among individuals can be displayed by their representation in the morphological space (Fig. 4E) defined by the first two principal axes, which provide a valuable approximation of the morphospace (56% of the total variance). The first relative warp explained 25.5% of the observed variation among individuals, the second 18.5%. Analyses of variance applied on the matrix of individual scores on the first three RW indicate a significant morphological differentiation among populations according to RW1 and RW2 (Table 1E–G). The set of individuals is grouped together according to their locality in the morphospace along the first two axes (Fig. 4E). The morphological interpretation of the shape changes between individuals from these two localities is indicated by the TPS warp grids showing the direction of maximum variance (Fig. 4E). The first two relative warp axes show that the main shape change in these both populations contrasts more or less developed eyes: the individuals from Malá Chuchle with more developed eyes and the individuals from Hlubočepy/St-Prokop with less developed eyes.
The morphological relationships among individuals observed in the morphological space support the same general pattern as the bivariate plot of file number versus cephalic width (tr.), towards more files in individuals from Malá Chuchle area.
Furthermore, linear regressions performed on the size (width) and shape (RW1 and RW2) suggest that there is no relationship between size and shape throughout the studied span of growth among populations previously defined according to their provenance.
Variability within population of Pedinopariops insequens
Frequency Distributions
The frequency distribution in the number of dorso ventral files of lenses in the individuals of Pedinopariops insequens is presented in figure 5A. Estimations from the ‘traditional’ and the smoothed (Gaussian KDE) histograms result in a multimodal distribution; at least two modes can be distinguished (Fig. 5A). Additionally, bivariate plots of lenses versus files number, of cephalic width (tr.) versus cephalic length (sag.) and, of files versus cephalic width (tr.) are respectively presented in figure 5B, 5C, and 5D.
Figure 5 A, The frequency distribution of the file number of lenses for all individuals of Pedinopariops insequens (Chlupáč, Reference Chlupáč1977). The ‘traditional’ histogram and the Kernel density estimate distribution of file number using the optimal Gaussian bin width. The frequency distribution shows two modes, i.e. a bimodality in the file number of lenses; B, scatter plot of lens number versus file number for all individuals; C, scatter plot of cephalic width (tr.) in mm versus cephalic length (tr.) in mm for all individuals showing a discontinuous and a typical isometric distribution for each group previously defined according to their file number (group 1 with nine to 13 lenses; group 2 with 16–20 lenses; group 3 with four lenses); D, scatter plot of file number versus cephalic width (tr.) in mm for all individuals; E, location of 88 cranidia/cephala of Pedinopariops insequens (Chlupáč, Reference Chlupáč1977) in the morphospace defined according to the first two relative warps RW1 and RW2. Axes RW1 and RW2 represent 37% and 14.5%, respectively, of the total variance. Thin plate spline deformation grids depicting patterns of shape change between the both populations of Pedinopariops insequens (Chlupáč, Reference Chlupáč1977). Double arrows in (A) and (B) underline the gap.
As detected in other phacopids (Crônier and Clarkson Reference Crônier and Clarkson2001; Crônier and Fortey Reference Crônier and Fortey2006), the scatter plot of cephalic width (tr.) versus cephalic length (sag.) shows a linear growth for all considered individuals. Within the population of Pedinopariops insequens, the relative proportions of cephala remain constant (y=ax+b, r=0.9765, p<0.001) whatever the degree of development of individuals (Fig. 5C).
However, two major groups can be distinguished according to their lens file number: a first group 1 of generally smaller individuals where the lowest values are nine, with the maximum values at 13; a second group 2 where the lowest values are 16, with the maximum values at 21. In each group, the lens file number range is from five to six (Fig. 5D). These higher values correspond well with the diagnosis and the description of adult P. insequens by Chlupáč (Reference Chlupáč1977). In addition, a third group is represented by a very small individual with only four files of lenses (Fig. 5A–D).
Within this population, a possible break in the linear dimensions between the individuals with a low lens file number versus a high lens file number seems to exist with a modification of the origin, i.e. a back and a persistence of the same slope (Fig. 5C). In order to test for structural change in these both groups, a Chow test (Chow Reference Chow1960) which is a variation of the F-test (Fisher Reference Fisher1970) and requires the sum of squared errors from three regressions, has been used to test whether a single regression line or two separate regression lines fit the data best and to detect for equality of the coefficients in these two separate linear regressions. The test indicates a structural break (F=43.41, p<0.001). Moreover, no significant difference between the two slopes while a significant difference between the constants exists (slope: F=0.2559, p=0.61NS; constant: F=49.86, p<0.0001).
Analyses of variance performed on the cephalic width and length (Table 2A,B) and on the file and lens number (Table 2C–D) confirm the existence of size heterogeneity and a differentiation among groups previously defined according to their file number (group 1 with nine to 13 lenses; group 2 with 16–20 lenses; group 3 with four lenses). All tests are significant (Table 2A–D).
Table 2 Univariate Analyses of Variance performed (A, B) on cephalic width (in mm) and cephalic length (in mm), on the file and lens number (C, D), on centroid size (E), and on the first three significant axes showing a significant morphological differentiation among groups of Pedinopariops insequens (Chlupáč, Reference Chlupáč1977) according to size, file and lens number and RW1 (F–H).
Thus, the reduction of the lens files number is dependent on body size. Additionally, the coefficient of variation in the number of eye lenses is somewhat more important in the population with 13 lenses or less (CV=8.80%) than in the population with 16 lenses or more (CV=2.44%). The comparison of the coefficient of variation by a F-max test indicates a significant heterogeneity of variance among groups (F vmax/vmin=3.62; ddl1=55; ddl2=19; p=0.016).
Morphological space
Morphological relationships among individuals can be displayed by their representation in the morphological space (Fig. 5E) defined by the first two principal axes, which provide a valuable approximation of the morphospace (61% of the total variance). The first relative warp explained 37% of the observed variation among individuals, the second 14.5%. Analyses of variance applied on the matrix of individual scores on the first three significant axes indicate a significant morphological differentiation among groups according to RW1 (Table 2F–H). The set of individuals are grouped together according to their file number in the morphospace along the first axis (Fig. 5E).
The morphological interpretation of the shape changes between individuals from the two major groups is indicated by the TPS warp grids showing the direction of maximum variance (Fig. 5E). The first relative warp axes shows that the main shape change in these both groups contrasts more or less developed eyes: the individuals from group 2 with more developed eyes and the individuals from group 1 with less developed eyes.
The morphological relationships among individuals observed in the morphological space support the same general pattern as the bivariate plot of file number versus cephalic width (tr.), towards more files in individuals from group 2.
Furthermore, a MANOVA performed on the size (CS) and the shape (RW1) suggests that there is a differentiation among groups previously defined according to their file number whereas a logarithmic regression suggests only a slight relationship (y=aln(x)+b; p<0.05) between size and shape (Fig. 6).
Figure 6 Scatter plot of size (centroid size CS) versus shape (RW1) for 88 individuals of Pedinopariops insequens (Chlupáč, Reference Chlupáč1977). A logarithmic regression suggests a slight relationship (y=aln(x)+b; p<0.05) between size and shape and shows a progressive shape change with size increase.
Discussion and conclusions
According to the number of files/lenses and the shape change, a bimodal distribution seems to differentiate individual populations of both studied species, Prokops prokopi and Pedinopariops insequens.
Changes observed in Prokops prokopi may correspond to changes in which extreme values for the number of files or lenses are favored over intermediate values. The variance of the feature increases and the population is then divided into two distinct groups. This process is known as disruptive selection, also called diversifying selection. This evolutionary process has long been viewed as important in increasing variation within natural populations (Rueffler et al. Reference Rueffler, Van Dooren, Leimar and Abrams2006) and initiating sympatric speciation (Maynard Smith Reference Maynard Smith1962; Dieckmann and Doebeli Reference Dieckmann and Doebeli1999). This disruptive selection arises from intense intraspecific competition (Martin and Pfenning Reference Martin and Pfenning2012).
Prokops prokopi is confined approximately to the upper half to the uppermost Dvorce-Prokop Limestone (see Chlupáč Reference Chlupáč1977; Vaněk Reference Vaněk1999), which is mostly represented by the subfacies of Prokop Limestone (thin bedded limestone but with supposed mud mound structures). Prokops hoeninghausi occurs especially in the lower part of the formation (Dvorce Limestone facies with thick, massive beds) but it also co occurs with P. prokopi on both studied localities (see also Vaněk Reference Vaněk1999). Both populations of Prokops prokopi come from two geographically close areas in the vicinity of Prague. There are two possible explanations for the observed bimodalities:
1. The population from Malá Chuchle is slightly older than the population from the St. Prokop’s Quarry (see localities 1 and 2 in Fig. 2). In Malá Chuchle, approximately middle levels to upper third of Praha Formation are exposed (see Chlupáč Reference Chlupáč1957), most probably still within the upper part of Nowakia acuaria Dacryoconarid Biozone. The St. Prokop’s quarry population, on the other hand, comes from the upper part of the Praha Formation (see Chlupáč Reference Chlupáč1957; Kříž Reference Kříž1999), corresponding to the Guerichina strangulata Dacryoconarid Biozone (correlation based on the occurrence of Guerichina infindibulum (Richter Reference Richter1854), P. Lukeš personal communication 2015). We thus cannot exclude the possibility of some microevolutionary shift between these two populations.
2. In the St. Prokop’s quarry, a supposed local mud mound structure characterized by a lens like body and by the thick bedded limestones with common fenestrate (Stromatactis) structure has been ascertained (see Chlupáč Reference Chlupáč1983: pp. 50; Chlupáč Reference Chlupáč1993; Galle and Parsley Reference Galle and Parsley2005); this structure was regarded as a reef by earlier authors. Disarticulated trilobite exoskeletons often form accumulations of lens like bodies with micritic matrix. It is, however, probable that similar mud mound structure existed also at Malá Chuchle (see Chlupáč Reference Chlupáč1983). The mud mound structures present a slightly different environment in comparison with the mostly muddy sea floor occurring in the NE part of the Prague Basin during middle–upper Pragian. The mud mound can offer a richer source of food and, especially caverns and cavities suitable for inhabiting. One can only speculate that some restricted populations of P. prokopi became adapted to such a protected environment as with other trilobite taxa in African Gondwana, Baltica and European Gondwana (Kaufmann Reference Kaufmann1997, Reference Kaufmann1998; Cavalazzi Reference Cavalazzi2006; Popp and Pärnaste Reference Popp and Pärnaste2011; Bignon and Cronier Reference Bignon and Crônier2014). The water depth at both localities was probably comparable (the muddy sea floor in the deeper shelf lay below the fair weather wave base, possibly also below the storm wave base but still within the photic zone, for discussion, see also Chlupáč et al. Reference Chlupáč, Havlíček, Kříž, Kukal and Štorch1998). Prokops prokopi occurs at both localities together with diverse trilobites with well developed eyes (Reedops cephalotes Hawle et Corda, Reference Hawle and Corda1847; Reedops bronni Barrande, Reference Barrande1846; Odontochile tenuigranulata Budil et al., Reference Budil, Hörbinger and Mencl2009; Zlichovaspis (Z.) rugosa Hawle et Corda, Reference Hawle and Corda1847etc.); trilobite associations established at both localities are very similar. It is interesting, that following Chlupáč (Reference Chlupáč1983), the trilobite association at both localities contain more common scutelluids than established in typical Reedops Odontochile Assemblage.
The bimodal population of Prokops prokopi thus shows several signatures of ecologically maintained adaptive divergence and reproductive isolation. The two modes differ in eye size and visual capacities, presumably owing to specialization (adaptation?) on different photic conditions.
Here we demonstrate that bimodality might also be maintained by a selective loss of trilobites with intermediate eye sizes, i.e. disruptive selection between the modes (Fig. 4). Such a disruptive selection could be expected to be important during the radiation of phacopid trilobites.
What are the possible causes for the documented disruptive selection here? One possibility is that competition for shared habitats reduces the success of individuals specializing on intermediate habitats, causing them to die or emigrate. Another possibility is that intermediate phacopids fall into a gap in the underlying habitat distribution (i.e. independent of competition). Both of these mechanisms can cause a disruptive selection in theoretical models (Schluter Reference Schluter2000; Rueffler et al. Reference Rueffler, Van Dooren, Leimar and Abrams2006; Doebeli et al. Reference Doebeli, Blok, Leimar and Dieckmann2007; Abrams et al. Reference Abrams, Rueffler and Kim2008) and could be implicated in the divergence of these phacopid trilobites. These considerations underline the importance of determining the conditions under which adaptive radiation can or cannot result by extrapolating processes observed within species.
In Pedinopariops insequens, the bimodal distribution reflects individuals from only one locality but rather two size classes. In a population, extreme values of the number of lenses and files can to be favored leading to an increase in the variance of these traits and to divide the population into two distinct groups. Could a disruptive selection cause an adaptative divergence between conspecific groups occupying different environments or using different resources?
For Pedinopariops insequens, there are two clear modes in the number of lenses and files of lenses (Figs. 5A, 7C) and there may be three modes in the cephalic width and length across the size range (Figs. 7A,B).
Figure 7 The frequency distribution of (A) the cephalic width (in mm), of (B) the cephalic length (in mm) and of (C) the number of lenses for all individuals of Pedinopariops insequens (Chlupáč, Reference Chlupáč1977). The ‘traditional’ histogram and the Kernel density estimate distribution using the optimal Gaussian bin width. The frequency distribution shows two clear major modes, i.e. a bimodality in the number of lenses and probably two modes in the cephalic length and width. Double arrows underline the gap.
The first size mode corresponds to individuals with 13 or less than 13 files of lenses. The two following modes of size correspond to individuals with 16 or more than 16 files of lenses. It seems that there is a discontinuity in the establishment of visual lenses between the first and the second mode size. We prefer to maintain the validity of this species but have to admit its notable and surprising intraspecific variability. There are several explanations of this phenomenon:
1. We can confirm the relative homogeneity of the studied material. It comes from one locality and distinct horizon at Praha Holyně (for details, see above). We thus can eliminate any collecting bias or misidentification of the samples, perhaps with the exception of hypothetical mixing of different layers within the uppermost Třebotov Limestone and the lowermost Choteč Limestone. Such unlikely mixing, in our opinion, cannot cause the observed discontinuity in establishment of the visual surface during ontogenetic development but can explain well cephalic width/length differences.
2. It is noteworthy that the population comes from the uppermost part of the Třebotov Limestone, situated very close (possibly above) to the Lower/Middle Devonian boundary. Very slightly later, the Basal Choteč Event took place, representing quite important overturn in the trilobite communities worldwide (see Chlupáč and Kukal Reference Chlupáč and Kukal1988; Chlupáč Reference Chlupáč1983, Reference Chlupáč1994; Chlupáč et al. Reference Chlupáč, Feist and Morzadec2000). The polymodality within a single population, as discussed, can be explained by a compensation of ecological stress caused by the onset of unfavorable conditions. It is noteworthy that Pedinopariops insequens is the only phacopid that crosses the Lower/Middle Devonian boundary and survived the Basal Choteč Event in the Prague Basin. It was apparently able to cope with extinctions and faunal changes accompanying this event. Perhaps, the stress conditions could cause changes of the ontogenetic trajectory and a certain type of the disruptive selection, possibly even with a short time acceleration of ontogenetic changes.
3. We have to admit the possibility that two or more species are hidden inside the traditional species of Pedinopariops insequens, especially the “large” and “small eyed” forms. One such form, however, lack adult specimens. Such an alternative explanation is, in our opinion, less probable.
Acknowledgments
The research was supported by the PHC CROISE Barrande Grant (EGIDE Project number 24465SH and MEB/021122) and the Grant Agency of the Czech Republic (Project number P210/12/2018) and by PRVOUK 44 of the Ministry of Education, Youth and Sports of the Czech Republic. This is a contribution to UMR 8198 EcoEvoPaleo du CNRS and to International Geoscience Programme (IGCP) Project 596 Climate change and biodiversity patterns in the Mid-Paleozoic. This study was financially supported also by the Ministry of Culture of the Czech Republic (DKRVO 2013/05, National Museum, 00023272).
