Synchronization in allometric and morphological changes during metamorphosis : comparison among four sparid species

Understanding critical ontogenetic transitions in the life of fish, such as metamorphosis, is of prime importance for determining the factors that affect survival probabilities and control recruitment levels. Metamorphosis involves a plethora of changing characters and its definition requires a multi-parametric approach. In the present study, mean lengths (Lm) at allometric and morphological change during the transition from larval to juvenile phenotype were estimated with a recently introduced multi-character protocol and compared between Diplodus sargus (settling in spring), Oblada melanura (summer), D. puntazzo (autumn) and D. vulgaris (winter). Lm from allometry closely matched Lm from morphology in D. sargus and D. puntazzo but not in the winterand summer-settling species, in which morphological transition occurred later or earlier than allometric change, respectively. Lm from allometry was very similar in the 3 Diplodus species but larger in O. melanura. The coefficients of variation of Lm from morphometric and morphological characters were negatively correlated. Morphological transitions were less synchronized but changes in allometric growth were more synchronized in D. vulgaris (winter) while the opposite was true for O. melanura (summer). This study highlights for the first time the relative importance of taxonomic relatedness and season (temperature) in determining the size at metamorphosis and the degree of synchronization (abruptness of change) both within and between mensural and morphological characters.


INTRODUCTION
Metamorphosis, i.e. the transition from larval to juvenile phenotype is an important event in the life cycle of fishes (Thorisson 1994, Copp & Ková 1996, McCormick et al. 2002, Nikolioudakis et al. 2010).There is a lack of a generally agreed definition of the end of larval (onset of juvenile) period because larval development is very varied and the plethora of changing characters during metamorphosis cannot synchronize perfectly.Definitions based on single mor phological criteria (e.g.squamation) are ultimately arbitrary (Urho 2002, Ditty et al. 2003) and may impede interspecific comparisons of morphological, sensory and behavioral development (e.g.Fuiman 1994, Fuiman et al. 1998, Higgs & Fuiman 1998, Koumoundouros et al. 2009).
In a recent publication (Nikolioudakis et al. 2010), we emphasized the importance of multi-character approaches for defining transitions, such as metamorphosis, in fish ontogeny.We used techniques to calculate length-at-metamorphosis that were based on scoring a suite of morphological characters ('multivariate morphometry') and on principal component analysis of morphometric traits ('multivariate allometry').These techniques were applied to newly settled white seabreams Diplodus sargus caught in spring and a high agreement was demonstrated between size-at-change in multivariate allometric growth and mean size-at-morphological transformation associated with metamorphosis (Nikolioudakis et al. 2010).
In the present paper, we extend this analysis to 3 more species of the family Sparidae (Oblada melanura, D. puntazzo and D. vulgaris) sampled from the same location in different seasons (summer, autumn and winter respectively).We compare patterns of morphological and allometric change among species (including D. sargus) as well as variability in the relative timing of change of the different characters in an effort to identify seasonal (i.e.temperature) effects and differences between the 2 genera.
Temperature affects the timing and rates of ontogenetic processes, with decreasing temperature slowing down the developmental rate and causing an increase in body size at specific developmental events (e.g.Fuiman et al. 1998, Koumoundouros et al. 2001).It has been hypothesized that when different morphological characters require dissimilar ontogenetic times to complete transformation, synchronization in their change will decrease with decreasing temperature (Somarakis & Nikolioudakis 2010).In that sense, thresholds that 'saltatorists' recognize in fish ontogeny will be more abrupt (easy to define) at higher temperatures but more 'gradual' (difficult to recognize) at lower temperatures (for the 'saltatorist'-'gradualist' debate, see Ková & Copp 1999).
Under the general framework of multi-character approaches to study metamorphosis (Ditty et al. 2003, Nikolioudakis et al. 2010, Ben Khemis et al. 2013), and using a comparative approach, we aimed at addressing 2 main questions expected to improve our understanding of ontogenetic transitions: (1) Is the close coupling of morphological transformation and allometric growth change, demonstrated in Diplodus sargus (Nikolioudakis et al. 2010), consistent across species?(2) How high is the degree of synchronization of the different morphological and morphometric traits during metamorphotic change in species from different seasons/temperature regimes?

Sampling
The 4 sparid species, saddled seabream Oblada mela nura (Linnaeus, 1758), sharpsnout seabream Dip lodus puntazzo (Walbaum, 1792), common two-banded seabream D. vulgaris (Geoffroy Saint-Hilaire, 1817) and white seabream D. sargus (Linnaeus, 1758) were sampled in July 2005, November 2005, February 2006 and May 2006, respectively, during the period of their settlement (García-Rubies & Macpherson 1995, Vigliola et al. 1998, Kiparissis et al. 2008) (Fig. 1).Sampling was carried out at a coastal site of the Ionian Sea (western Greece), east of the estuary of Acheloos River (38°20' N, 21°07' E, Nikolioudakis et al. 2010).On each sampling occasion, the site was visited for 5 to 7 d during which a 1 to 2 h dive was performed daily and fish were caught using hand nets.More details are provided in Nikolioudakis et al. (2010).Fish were transferred live to the Laboratory of Zoology in the University of Patras, Greece and each individual was anaesthetized, photographed and finally fixed and stored in 5% borax-buffered formalin.
Patterns of multivariate allometry for each species were studied through Principal Component Analysis (PCA) of the covariance matrix of the log-transformed morphometric measurements (Jolicoeur 1963a,b, Shea 1985, Klingenberg & Froese 1991).The point of change in oblique orientation of the second principal component (PC2) scores in relation to SL was estimated using a piece-wise linear regression (Fig. 2) and considered to define mean length-at-metamorphosis, L m (morphometric characters).This methodology was developed and applied first to D. sargus (Nikolioudakis et al. 2010) based on the work of Shea (1985).According to the latter, in an ontogenetic series in which shifts in allometric growth patterns between sequential stages (different growth stanzas) exist, these shifts would be reflected in the multistage PCA as divergent PC2 trajectories (changes in oblique orientation) in relation to the PC1 (or SL).
Comparisons of multivariate allometric growth patterns of individual morphometric characters were carried out by plotting bootstrapped confidence intervals of allometric (PC1) coefficients and evaluating their departure from the theoretical value for multivariate isometry 1/√p, where p denotes the number of measured characters (Klingenberg & Froese 1991).Multivariate allometry was compared for fish <L m (Group 1) and fish ≥L m (Group 2), both intra-and inter-specifically.

Analysis of morphological characters
The same set of characters examined in D. sargus (Nikolioudakis et al. 2010) were also studied in O. melanura, D. puntazzo and D. vulgaris (Table 1).This included 6 osteological characters, relating to fin calcification (proximal radials of each the dorsal, anal and pectoral fin [Prx], distal radials of the dorsal fin [Rd], epurals of the caudal fin [Ep], and basiptery-Fig.2. Schematic of the steps adopted to define mean lengthat-metamorphosis (L m ) based on morphometry (Nikolioudakis et al. 2010).A Principal Component Analysis is performed on the covariance matrix of log-transformed morphometric measurements and a piece-wise regression is fitted on the PC2 versus SL relationship.L m and its 95% confidence interval is estimated as the breakpoint of the piecewise regression gium of the pelvic fin [Bp]) and 4 characters related with external morphology (body coloration, lateral line canal, coverage of ventral body region with scales, coverage of fish body with scales).In each species, an ontogenetic series consisting of one randomly selected fish per ~0.5 mm SL, covering the entire size range of sampled fish was examined (O.melanura: n = 41, D. puntazzo: n = 33, D. vulgaris: n = 25, D. sargus: n = 38).The calcification state of osteological characters was assessed after staining the fish with Alizarin Red and Alcian Blue solutions (Park & Kim 1984).Every studied morphological character was assigned equal weight, i.e. it had 3 possible states (0: larval, 1: intermediate, 2: juvenile) (Table 1).
For body pigmentation pattern all specimens used in the morphometric analysis were scored.More details are provided in Nikolioudakis et al. (2010).
The mean length at state 1 (intermediate state) of each morphological character was calculated.Mean lengths were subsequently averaged and the resulting value and coefficient of variation (CV) were used to index 'mean length at morphological change', L m (morphological characters) and its variability, respectively.
CV of L m can be considered as a measure of synchronization of changing characters during the transition from larval to juvenile sate, i.e. the higher the CV the lower the synchronization of examined characters.
For each morphological character, the length at which state 1 starts, ends as well as the respective range (end minus start) were determined.Furthermore, these lengths were expressed in terms of the ontogenetic index of Fuiman (1994): where O L is the ontogenetic index, a numerical representation of developmental state of a fish with length L and L m is the length at metamorphosis as defined from the analysis of multivariate allometry L m (morphometric characters).

RESULTS
In total, 674 fish were caught during the settlement period of each species (Fig. 1).Each species settled during a different period of the year: Diplodus vulgaris in winter (low temperatures), Oblada melanura in summer (high temperatures), and D. sargus (Nikolioudakis et al. 2010) and D. puntazzo in spring and autumn, respectively (intermediate temperatures) (Fig. 1).

Multivariate allometry
The PCA of morphometric characters and the plots of PC2 scores against SL (Table 2, Fig. 3) revealed the significant (p < 0.01) change in allometric growth (change in oblique orientation of PC2 scores) of D. vulgaris, O. melanura and D. puntazzo as has already been described for the sample of D. sargus (Nikolioudakis et al. 2010).The fit of piece-wise re gressions estimated L m at 17.97 mm for O. melanura, 14.46 mm for D. puntazzo and 14.82 mm for D. vulgaris (Table 2, Fig. 3).The change in slope (b 2 ) was higher in D. vulgaris and estimated L m was more precise (Table 2).The respective estimate for D. sargus was 14.46 mm with CV = 0.023 (Nikolioudakis et al. 2010).1. Morphological characters (external and osteological) and their states.S: Hard Spines, R: Lepidotrichia, SCR: Dermatotrichia, Ep: Epurals, Prx: Proximal radials, Rd: Distal radials, Bp: Basypterygium (see also Fig. 2 in Nikolioudakis et al. 2010).Characters marked with an asterisk were also examined but finally excluded from subsequent analyses as they had attained their definite number and were fully calcified in all specimens.Dermatotrichia of the caudal fin presented high variability in number to be useful in this study and were also excluded from analyses.
The comparison of allometric vectors (PC1 coefficients) between groups (Group 1: fish < L m and Group 2: fish ≥ L m ) (Fig. 4) revealed that, in general, D. puntazzo and D. vulgaris exhibited patterns similar to that described for D. sargus (Nikolioudakis et al. 2010).Inter-specific differences for fish < L m and fish ≥ L m , including D. sargus analyzed in Nikolioudakis et al. (2010), are illustrated in Figs. 5 & 6.Characters related to the horizontal axis of the body, and, more specifically, to general somatic length and tail length (SL, FL, PostDFL, PostAFL, PreAFL and PostAL), exhibited negative allometry that was more pronounced in Group 1 (Figs. 4 & 5).In Group 2, the respective coefficients were, generally, closer to isometry (Figs. 4 & 6).Measurements of the middle part of the body (HD2, BD, PreAL and MidBL) exhibited positive allometry in both groups (though slightly lower in Group 2).Characters of the cephalic region (HL, EyeD, SNL, HD1, PreDFL and PrePlvFL) were more or less isometric in both groups.A contrasting difference was observed in snout length (SNL) of D. puntazzo that turned from positive (Group 1) to negative allometry (Group 2).Finally, the depth at the caudal peduncle (CpD) grew isometrically in both groups.Regarding O. melanura, individual morphometric characters had similar patterns to those of Diplodus but with allo metric coefficients closer to isometry.The 2 genera differed mainly with respect to characters of the cephalic region that were negatively allometric in O. melanura Group 2.
The acquisition of adult coloration ('juvenile pattern') in O. melanura was attained at approximately L m (morphometric characters), whereas D. puntazzo and D. vulgaris exhibited intermediate pigmentation patterns, i.e. that between transparency of body flanks ('larval pattern') and adult pigmentation ('juvenile pattern'), at estimated L m (Fig. 3).
The completion of formation of the lateral line canal and coverage of ventral region with scales took place over very short length intervals whereas complete squamation was more prolonged in terms of size range (Table 3, see also Nikolioudakis et al. 2010 for D. sargus).Prolonged completion was also noticed for the osteological characters examined (Table 3).The overall mean length at character state 1 for the 10 characters considered was 15.75 mm for O. melanura, 14. 16.17 mm for D. vulgaris (Table 3).The respective length for Diplodus sargus was 14.42 mm with CV = 0.092 (Nikolioudakis et al. 2010).

Matching of changes during metamorphosis
A comparison of L m (morphometric characters), with the overall mean length at morphological change, i.e. mean length at character state 1 L m (mor-phological characters), demonstrates a close match between the 2 L m s for D. puntazzo and D. sargus (Fig. 7A).L m (morphological characters) was lower for O. melanura (summer) and larger in D. vulgaris (winter) (Fig. 7A).The CV of L m (morphological characters) was negatively related to the CV from morphometric characters (Fig. 7B).
Considering L m (morphometric characters) as being length-at-metamorphosis, no relationship was found between ontogenetic start (O L start) and ontogenetic end (O L end) (Table 4).However, a significant negative relationship existed between O L start and 'ontogenetic range' (i.e.O L end minus O L start) in each species separately or after pooling all data (Table 5).These relationships indicate that the closer a morphological character starts to change in relation to the allometric change the shorter the developmental time needed to complete transformation (acquisition of 'juvenile' state).O. melanura displayed the strongest correlation, D. vulgaris the weakest, and the remainding species intermediate correlation between O L start and ontogenetic range (Table 5).ANCOVA models showed that the slopes of the relationships were homogeneous (F = 2.01, p = 0.133) but intercepts were statistically different between species (F = 10.36,p < 0.001), indicating that the overall morphological change in D. vulgaris (winter) was late and in O. melanura early with regard to allometric change (Fig. 8).

DISCUSSION
Understanding ontogenetic transitions, especially those that are critical for the life of fish, is of prime importance in our attempt to determine factors that affect survival probabilities and control recruitment levels.the sense that they can effectively capture 'synchronization', a condition necessary for defining thresholds (Ková et al. 1999).In the present study, we provide evidence that synchronization of changes in the same subsets of morphological and morphometric characters vary largely between the 4, closelyrelated sparid species, both among the same set of characters (as indexed by the CV of L m ) and between the morphological and morphometric subsets (differences in mean L m ).L m , as defined by morphometric and morphological characters, was similar for the species developing at intermediate temperatures in spring and autumn, namely Diplodus sargus and D. puntazzo.In contrast, L m (morphological characters) was smaller than L m (morphometric characters) for Oblada melanura, developing in warm waters during summer, and higher for D. vulgaris, developing in cold waters during winter.L m (morphometric characters), however, was almost identical for the 3 species of Diplodus, indicating that size of metamorphotic changes in allometry is less dependent on temperature than morphological change, mostly reflecting evolutionary relatedness (i.e.congeneric species had similar L m s).In that sense, L m (morphological characters) seems highly dependent on temperature and can be  Gadus morhua [Jordaan et al. 2006]).
Considering that L m (morphometric characters) better expresses the size at metamorphosis, and comparing ontogenetic 'timing' (O L start) (sensu Fuiman, 1994) versus 'duration' of transition (O L end − O L start) for the different morphological characters examined here, a negative linear relationship emerged, implying that the later a morphological change occurs with regard to timing of allometric change, the shorter is its ontogenetic range (i.e. it is more abrupt).The slope of the relationship was not significantly different between the 4 species but the intercept was higher for D. vulgaris and lower for O. melanura, indicating that, besides the aforementioned differences in overall timing of morphological transition, ontogenetic range (abruptness of change) of individual morphological characters was generally independent of species (temperature).However, the strength of the relationship (r 2 in Table 5) was higher for O. melanura, intermediate for D. puntazzo and D. sargus, and lower for D. vulgaris, implying that decreasing temperature induces increasing variability in morphological change, i.e. the change is less synchronized between characters.The negative relationship (Fig. 7B) between variability in allometric change and morphological transformation highlights a yet unknown aspect of ontogenetic thresholds.Synchronization of allometric change of mensural characters might be higher at low temperatures but that of morphological characters higher at high temperatures.
Relative allometric growth patterns of most morphometric characters examined here were generally to isometric growth (juvenile).Occasionally, there were certain species-specific patterns such as the strong contrast of snout (SNL) allometry in D. puntazzo, which possesses a pointed head in its metamorphosed stage (Loy et al. 2001).The positive and negative allometric patterns of the middle and posterior part of the body, respectively, are associated with changes in body form related to requirements of the benthic life.The profiles of the middle part of the body were always found to deepen, something that is necessary for maneuvering (quick starts and rapid turns) (Webb 1984).Furthermore, pronounced growth in the abdominal region (MidBL) implies enhanced development of the intestine (Elbal et al. 2004), a finding similar to other studies that have shown that the middle part of the body increases later throughout ontogeny (after head and tail) in bilateral species (Osse et al. 1997, van Snik et al. 1997, Gozlan et al. 1999), as opposed to asymmetrical species (Paralichthys californicus, Gisbert et al. 2002).In summary, the use of standard multivariate protocols to compare allometric and morphological changes associated with metamorphosis highlighted the relative importance of taxonomic relatedness and season (temperature) in determining the size at metamorphosis and the degree of synchronization (abruptness of change) both within and between mensural and morphological characters.The results obtained have implications for current views of ontogenetic processes and life history theory (Copp et al. 1999)

Fig. 1 .
Fig. 1.Mean daily water temperature at the study area.Sampling dates (arrows) and the mean water temperature at the week of capture for each species is given (parentheses).The number of individuals caught as well as their standard length range are also given in parentheses Fig. 3. Estimation of mean length-at-metamorphosis using morphometric characters.Fitted piece-wise regressions of PC2 scoreson-standard length (SL).Pigmentation state of individuals is also indicated.Δ: larval, d: intermediate, and h: juvenile Fig. 4. Intraspecific comparisons.Allometric coefficients for morphometric characters for Group 1 (s) and Group 2 (d).Dashed lines indicate multivariate isometry (0.242).Error bars: 95% bootstrapped confidence intervals.SL: Standard Length, FL: Fork Length, SNL: Pre-Orbital Length, EyeD: Eye Diameter, HL: Head Length, PreDFL: Pre-Dorsal Fin Length, PostDFL: Post-Dorsal Fin Length, PreAFL: Pre-Anal Fin Length, PostAFL: Post-Anal Fin Length, PrePlvFL: Pre-Pelvic Fin length, PreAL: Pre-Anal Length, MidBL: Mid-Body Length, PostAL: Post-Anal Length, HD1: Head Depth 1, HD2: Head Depth 2, BD: Body Depth, CpD: Caudal peduncle Depth

Fig. 7 .
Fig. 7. Comparison between mean morphological and mean morphometric change in the 4 sparid species (see Fig. 6).(A) Plot of the mean length-at-metamorphosis (L m ) for morphological vs. morphometric characters.Error bars: 95% confidence intervals.Dashed line is y = x.(B) Relationship between coefficients of variation (CVs) for morphological vs. morphometric estimates of L m .D. sargus data are from Nikolioudakis et al. (2010) and warrant further investigations under controlled laboratory conditions.

Table 3 .
Morphological characters.Start: minimum length for state 1.End: minimum length for state 2. Mean: mean length of fish in state 1.Range: End−Start.n:number of specimens observed in state 1 for the respective character.Abbreviations of the osteological characters listed are given in Table1.For the overall mean of each species CVs are also given in parentheses

Table 4 .
Nikolioudakis et al. (2010) start O L end O L start O L end O L start O L end Ontogenetic start (O L start) and ontogenetic end (O L end) for character state 1, based on the ontogenetic index of Fuiman (Fuiman 1994) with lengths-at-metamorphosis as defined by the analysis of morphometric characters.See Table1for abbreviations.D. sargus data are fromNikolioudakis et al. (2010)

Table 5 .
Results of the linear regressions of (O L start)-on-[(O L end) -(O L start)] and (O L end)-on-[(O L end) -(O L start)].for the 3 species of Diplodus (D. puntazzo and D. vulgaris: this study, D. sargus: Nikolioudakis et al. 2010, see also Vigliola & Harmelin-Vivien 2001 and Loy et al. 2001, as well as for O. melanura).The PC1 coefficients of most characters of fish with SL ≥ L m (Group 2) always exhibited values closer to isometry, indicating a transition from intense allometric (larval) ns: non-significant.D. sargus data are from Nikolioudakis et al. (2010) Fig. 8. Plot of the fitted ANCOVA model for the morpho-logical characters of the 4 species.D. sargus data are from Nikolioudakis et al. (2010) similar