Dynamic skeletogenesis in fishes: Insight of exercise training on developmental plasticity

Authors

  • Thomas Grünbaum,

    1. Laboratoire de Biologie Évolutive, Université du Québec à Rimouski, Rimouski, Québec, Canada
    2. Département des Sciences, Université Sainte-Anne, Pointe-de-l'Église, Nouvelle-Écosse, Canada
    Search for more papers by this author
  • Richard Cloutier,

    Corresponding author
    1. Laboratoire de Biologie Évolutive, Université du Québec à Rimouski, Rimouski, Québec, Canada
    • Laboratoire de Biologie Évolutive, Université du Québec à Rimouski, 300 allée des Ursulines, Rimouski, Québec, Canada, G5L 3A1
    Search for more papers by this author
  • Bruno Vincent

    1. Département de Biologie, Chimie et Géographie, Université du Québec à Rimouski, Rimouski, Québec, Canada
    Search for more papers by this author

Abstract

Background: Through developmental and evolutionary time, organisms respond variably to their environment not only in terms of size and shape but also in terms of timing. Developmental plasticity can potentially act on various aspects of the timing of developmental events (i.e., appearance, cessation, duration, sequence). In this study, we address the developmental plasticity of median fin endoskeleton by using exercise training on newly-hatched Arctic charr (Salvelinus alpinus). Results: Developmental progress of cartilage formation (i.e., chondrification) in all fins is less influenced than ossification by an increase of water velocity. The most responsive elements, meaning those elements with greater onset plasticity owing to a water velocity increase, differ in terms of early versus late developmental events. The most responsive elements are those that chondrify and to a greater extent ossify later in the development. Conclusions: Plasticity is documented for the timing of appearance (i.e., onset) and the timing of transition from cartilage to bone (i.e., transitions of skeletal states) rather than the order of events within a sequence. Similarities of plastic response in developmental patterns could be used as a powerful criterion to strengthen the identification of phenotypic modules. Developmental Dynamics 241:1507–1524, 2012. © 2012 Wiley Periodicals, Inc.

INTRODUCTION

From a micro-evolutionary perspective, a major task of Evo-Devo is to understand how development structures, limits, and expresses the underlying genotypic architecture (and variation) to generate specific phenotypes and to produce phenotypic variation (Corley, 2002). Even if natural selection is a major driving evolutionary mechanism that acts upon phenotypes, the environment (and its changes) serves not only as a selective filter for a specific phenotype within a spectrum of trait variability (Hall, 1999; Newman and Müller, 2000; Balon, 2003). Developmental processes by means of the epigenetic control [defined as the sum of the genetic and non-genetic factors that control selectively the gene expression to produce and to organize cells that further increase phenotypic complexity during development (Hall, 1999; Newman and Müller, 2000)] transform the genetic variation and the environmental changes in phenotypic variation (Hallgrímsson et al., 2005). Thus, studies on developmental processes can address two classes of mechanisms: (1) those responsible for the production of specific/defined phenotypes (i.e., developmental stability or canalization) despite genotypic variation and/or changes in environmental conditions; and (2) those that induce alternative phenotypes (i.e., developmental plasticity, polyphenism) from a single genotype owing to changes in environmental conditions (Sholtis and Weiss, 2005). In this study, our aim was to address the latter class and specifically to provide an empirical examination of the plasticity of skeletal events through development.

Phenotypic plasticity among wild populations of numerous teleost species has been repeatedly documented, mainly in terms of adult body shape differences likely associated with trophic and/or habitat differences. Although the plasticity is unquestioned, it is frequently difficult to identify precise links between the environmental factors in a complex environment and the induced phenotypic changes. However, developmental plasticity has been tested experimentally on a variety of species, phenotypes, and environmental factors. Phenotypic plasticity has been primarily studied in teleosts [e.g., adrianichthyid (Kawajiri et al., 2011), anarhichadids (Pavlov and Moksness, 1994), cichlids (Crispo and Chapman, 2010), clupeids (Fuiman et al., 1998), cyprinids (Mabee et al., 2000), gasterosteids (Garduno-Paz et al., 2010), moronids (Georgakopoulou et al., 2007), salmonids (Pakkasmaa and Piironen, 2001; Peres-Neto and Magnan, 2004; Grünbaum et al., 2007, 2008; Fischer-Rousseau et al., 2009; Cloutier et al., 2010; Totland et al., 2011)] most likely owing to the facility of rearing experiences; however, there is no a priori reason to think that it is limited to teleosts.

Although thermally induced plasticity has been highly documented (Pavlov and Moksness, 1994, 1997; Fuiman et al., 1998; Mabee et al., 2000; Georgakopoulou et al., 2007; Schmidt and Starck, 2010; Kawajiri et al., 2011), numerous environmental/experimental factors have also been investigated such as dissolved oxygen (Schmidt and Starck, 2010), salinity (Schmidt and Starck, 2010), diet (Meyer, 1987; Huysseune et al., 1994; Muschick et al., 2011), predator odors (Frommen et al., 2011), hormonal conditions (Shkil et al., 2010), habitat heterogeneity (Garduno-Paz et al., 2010), and water velocity (Pakkasmaa and Piironen, 2001; Peres-Neto and Magnan, 2004; Grünbaum et al., 2007, 2008; Cloutier et al., 2010; Totland et al., 2011). Water velocity is of primary interest for at least two reasons. First, water velocity is a factor ubiquitous to all fish activities such as locomotion, feeding, reproduction, and predator avoidance (Azuma et al., 2002; Pakkasmaa and Piironen, 2001). Second, the main loads on tissue (muscles, bones) arise from the forces generated by the animal and the reactions of the surrounding medium (van der Meulen et al., 2006). Therefore, many species of fish control buoyancy via their swim bladder, the effects of gravitational forces are then limited compared with those of terrestrial animals.

The plastic responses themselves vary according to the factor, the range of variation of the factor, the developmental period at which the factor is applied, and the responsiveness of the species. A wide range of plastic responses has been reported in the literature. Focusing on anatomical plasticity, the following traits exemplify the types of responses: size [e.g., gills (Crispo and Chapman, 2010), brain (Crispo and Chapman, 2010), head (Meyer, 1987), body (Pakkasmaa and Piironen, 2001; Peres-Neto and Magnan, 2004; Georgakopoulou et al., 2007; Grünbaum et al., 2008; Schmidt and Starck, 2010; Frommen et al., 2011; Kawajiri et al., 2011), fins (Fischer-Rousseau et al., 2009)] and shape changes [e.g., head (Meyer, 1987), pharyngeal (Muschnick et al., 2011) and oral jaws (Meyer, 1987), body (Pakkasmaa and Piironen, 2001; Peres-Neto and Magnan, 2004; Georgakopoulou et al., 2007; Grünbaum et al., 2007; Fischer-Rousseau et al., 2009; Garduno-Paz et al., 2010; Frommen et al., 2011)], bone matrix (Totland et al., 2011), number of serially repeated elements [e.g., pharyngeal teeth, vertebrae, fin rays, spines (Arratia and Schultze, 1992; Mabee et al., 2000; Georgakopoulou et al., 2007; Shkil et al., 2010; Kawajiri et al., 2011)], and timing of ossification (Pavlov and Moksness, 1994, 1997; Fuiman et al., 1998; Mabee et al., 2000; Cloutier et al., 2010, Fiaz et al., 2012). It is also known that fin positioning is plastic with respect to changes in hydrodynamic conditions (Fischer-Rousseau et al., 2009, 2010). The functional recruitment of median fins is known to vary during fish ontogeny (Osse and van den Boogaart, 1995) as well as with water velocity (Drucker and Lauder, 2005). From a functional perspective, the dorsal and anal fins act primary as keels to stabilize the body, whereas the caudal fin participates in the generation of thrust for locomotion (Webb and Fairchild, 2001). The correlative link between plasticity in developing endoskeletal structures of median fins in relation to exercise training remains to be further addressed.

Evidently, most morphological and physiological systems are potentially responsive to experimental and environmental factors. Independently of the combination of species, phenotypes, and environmental factors, numerous studies suggest that the phenotypic response results from a modification of either the metabolism and/or developmental rates (Meyer, 1987; Fuiman et al., 1998; Mabee et al., 2000; Schmidt and Starck, 2010; Kawajiri et al., 2011). In theory, developmental plasticity could be observed on (1) the onset (i.e., timing of occurrence of a developmental event), (2) the offset (i.e., timing of cessation), and (3) the duration of an event (i.e., difference between timing of onset and offset), as well as on (4) the sequences and (5) trajectories of developmental events. In this study, our aim is to provide and to discuss an empirical examination of exercise training effect on the developing median fin endoskeleton of Salvelinus alpinus. Among salmonids, the Arctic charr (Salvelinus alpinus) is one of the most promising species for cold-water aquaculture in eastern Canada (Le François et al., 2002; Grünbaum et al., 2008). Its strong potential for aquaculture lies in high growth rate, efficiency of food conversion ratio (Le François et al., 2002) and well-known rearing practices (Johnston, 2002). We were especially interested in using endoskeletal data to define developmental events for comparative studies on the plasticity of developmental timing. Four issues have been addressed: (1) developmental progress, (2) relative developmental sequence of events, (3) transition of skeletal states (i.e., transition from cartilage to bone), and (4) developmental trajectory.

RESULTS

General Morphology of Median Fins

The dorsal fin is generally composed of 13 (12–14) elongated proximal radials (PR) and 12 (10–13) rounded distal radials (DR) that support lepidotrichia. The PR and DR are organized in a one-to-one relationship (Fig. 1A–C) except for the first proximal radial (PR1) that bears no distal radial. PR and DR are cartilaginous elements that eventually ossify perichondrally. In newly-hatched specimens (ca. 13-mm SL), no distal radial are formed and eight to nine proximal radials are already present, six being the lowest number (PR5–PR10) found in a specimen of 12.9-mm SL (Fig. 1A). Chondrification occurs initially in central proximal radials and proceeds bidirectionally in an alternate manner (Fig. 1A, B). Ossification starts at PR3–4 and proceeds posteriorly. Ossification of distal radials was not initiated even in the larger specimens (ca. 45-mm SL) (Fig. 1C).

Figure 1.

Skeletal anatomy of dorsal (A–C), anal (D–F), and caudal (G–I) fins in S. alpinus. Characteristic steps of skeletal elements formation in each fin are shown from alevin and juvenile specimens. Size and age of specimens are as follows: (A) 14.20-mm SL (0 dph), (B) 15.54-mm SL (12 dph), (C) 23.07-mm SL (64 dph), (D) 15.61-mm SL (8 dph), (E) 15.15-mm SL (20 dph), (F) 31.57-mm SL (78 dph), (G) 12.87-mm SL (0 dph), (H) 16.66-mm SL (26 dph), and (I) 31.57-mm SL (78 dph). Cartilaginous elements are stained blue (Alcian blue) and bones are stained red (Alizarin Red S) following the method described in the Experimental Procedures section. DR, distal radial; E, epural; HA, haemal arch; HS, haemal spine; H, hypural; L, lepidotrichia; NA, neural arch; NS, neural spine; PH, parahypural; PR, proximal radial; PU, preural centrum; ST, stegural; U, ural centrum; UN, uroneural. Anterior is to the left. The representation of the median fin elements schematized the skeleton without the vertebrae and their associated elements (i.e., NA, NS, HA, HS).

Morphology, relationships, direction of chondrification and ossification of PR and DR of the anal fin are highly similar to that of the dorsal fin. Generally, 12 (10–13) PR and 11 (9–12) DR are present (Fig. 1D–F). As in the dorsal fin, PR1 is never associated with a distal radial. In newly-hatched specimens, five to seven PR are already present although no radials were found in specimens under 13.5-mm SL (Fig. 1D). Central proximal radials are the first to form and chondrification proceeds bidirectionally as in the dorsal fin (Fig. 1D, E). Ossification starts at PR3–4 and proceeds from anterior to posterior; no ossified distal radials were present in the larger specimens (ca. 45-mm SL) (Fig. 1F).

The caudal fin elements analyzed belong to vertebrae known as preural (PU1–5) and ural centra (U2, U4). The anatomical distinction between these two types of centra (vertebrae) depends on their relative position in the body axis from the branching point of the caudal artery and from the parahypural; centra located anteriorly are referred to as preural, whereas centra located posteriorly are referred to ural (de Pinna, 1996). Ventrally to the notochord, the elements include haemal arches (HA) and spines (HS) of PU2–5, a parahypural (PH) (i.e., ventral component of PU1), and six hypurals (H1–6). Dorsally, two epurals (E1–2) and three uroneurals (i.e., modified neural arches) referred to as the stegural (ST) and uroneurals 2 and 3 (UN2–3) are present; the remaining dorsal components are neural arches (NA) and spines (NS) of PU2–5 (Fig. 1G–I). Generally, 28 caudal elements are present although a seventh hypural (H7) and a fourth uroneural (UN4) were found in some specimens (see Grünbaum and Cloutier, 2010). UN4 and H7 were not considered in the analyses because of their seldom occurrence. In newly-hatched specimens of 13.5-mm SL, several cartilaginous ventral elements are already present including HA of PU2–5, HS of PU2, PH, and H1–4; no dorsal elements are formed posteriorly to NA of PU3 (NA3) at this size (Fig. 1G). H5–6, E1, and ST are the next elements to form in 15-mm-SL specimens. E2, UN2, and NA of PU2-5 and NS of PU2–5 are formed in 16-mm-SL specimens; at this size, all hypurals are ossified (Fig. 1H). The last cartilaginous element to form in the caudal fin is UN3 in 17-mm-SL specimens. Most caudal elements are ossified at 20-mm SL (Fig. 1I). No clear pattern in the direction of chondrification and ossification was found in caudal elements. Even if some variations occur in the number of elements composing the three fins, especially in the caudal fin, it only reflects intraspecific variation as independent from the treatment (Grünbaum and Cloutier, 2010). The velocity treatment consisted of exposing during 100 days post-hatching, newly-hatched fish, to four different water velocities kept constant throughout the experiment (see Experimental Procedures section). The velocity treatments, given in absolute value, correspond to water velocity of 3.2 cm s-1 (A = fast treatment), 1.6 cm s-1 (B = medium treatment), 0.8 cm s-1 (C = slow treatment), and 0.4 cm s-1 (D = still treatment, normal condition/reference treatment).

Developmental Progress

Developmental progress refers to a cumulative occurrence of specific developmental events extrapolated from a series of sampled individuals. In terms of dynamic skeletogenesis, developmental progress refers to the progressive modeled onset of the cartilaginous state (chondrification) and the ossified state (ossification) for a given element through development. Modeling of onset was performed by using logistic regressions (for the description of the statistical analysis see Experimental Procedures section).

In the dorsal fin, out of 25 elements generally found (27 in maximum), the logistic model was significant for 17, 16, 18, and 16 cartilaginous elements in treatments A, B, C, and D, respectively. Elements common to all treatments (i.e., 16) are proximal radials PR1–4 and distal radials DR2–13. Independently of treatments, most elements chondrified between 13.5 and 19 mm SL except for DR2 and DR13, which chondrified between 20 and 25 mm SL. Chondrification onset is little influenced by the water velocity increase (Fig. 2A); the latest elements to form (i.e., DR2 and DR13) were slightly affected by the treatments. The logistic model was significant for 13, 12, 13, and 11 ossified elements in treatments A, B, C, and D, respectively. The elements common to all treatments (i.e., 11) are proximal radials PR1–11 (Fig. 3A). Proximal radials begin to ossify at approximately 22.5-mm SL, all of them being ossified at approximately 50-mm SL. A general trend is that ossification onset is more influenced than chondrification by the water velocity increase even if the number of PR and DR found to be significant by the logistic model differs between chondrification progress and ossification progress. Similarly to chondrification, ossification onsets for central elements are little influenced by the differential water velocity; the most responsive elements are the later ossified PR1 and PR9–11 in all treatments, PR12 in treatments A, B, and C, and PR13 in treatments A and C (Fig. 3A).

Figure 2.

Developmental progress of chondrification in S. alpinus. Modeled chondrification onsets (SL50) of elements common to all treatments are shown for dorsal (A), anal (B), and caudal (C) fins. (A) PR1–4, proximal radials 1–4, DR2-DR13, distal radials 2–13; (B) PR1–5, PR10–12, proximal radials 1–5 and 10–12, DR2–12, distal radials 2–12; (C) E1–2, epurals 1–2; HSPU3–5, haemal spine of preural centra 3–5; H5–6, hypurals 5–6; NSPU2–5, neural spine of preural centra 2–5; ST, stegural; UN2, uroneural 2. Dotted lines in A and B schematized the antero-posterior morphological organization of PR elements in dorsal and anal fins, respectively. In C, elements were organized in relation to SL50 from the slower treatment D.

Figure 3.

Developmental progress of ossification in S. alpinus. Modeled ossification onsets (SL50) of elements common to all treatments are shown for dorsal (A), anal (B), and caudal (C) fins. (A) PR1-11, proximal radials 1–11; (B) PR1–9, proximal radials 1–9; (C) E1–2, epurals 1–2; HAPU2–5, haemal arch of preural centra 2–5; HSPU2–5, haemal spine of preural centra 2–5; H1–6, hypurals 1–6; NAPU2–5, neural arch of preural centra 2–5; NSPU2–5, neural spine of preural centra 2–5; PH, parahypural; ST, stegural; UN2–3, uroneurals 2–3. Dotted lines in A and B schematized the antero-posterior morphological organization of PR elements in dorsal and anal fins, respectively. In C, elements were organized in relation to SL50 from the slower treatment D.

In the anal fin, out of 23 elements generally found (25 in maximum), the logistic model was significant for 20, 24, 23, and 23 cartilaginous elements in treatments A, B, C, and D, respectively. The elements common to all treatments (i.e., 19) are proximal radials PR1–5 and PR10–12, and distal radials DR2–12. Independently of treatments, most elements chondrified between 12.5 and 20-mm SL, except for PR13, DR2, and DR12, which chondrified between approximately 24 and 55-mm SL. Chondrification onset is little influenced by the treatments (Fig. 2B) although compared to the dorsal fin, the later chondrified elements (i.e., PR12, PR13, DR2 and DR12) are highly responsive to the water velocity increase. The logistic model was significant for 11, 11, 10, and 9 ossified elements in treatments A, B, C, and D, respectively (Fig. 3B). The elements common to all treatments (i.e., nine) are proximal radials PR1–9. Proximal radials begin to ossify at approximately 25-mm SL; all of them are ossified at approximately 45-mm SL. As in the dorsal fin, ossification onset is more influenced by an increase of water velocity than chondrification although the number of significant PR and DR differs between chondrification and ossification. Similarly to chondrification, the most responsive elements for which ossification onsets are highly influenced by the water velocity increase are the later ossified PR1, PR2, and PR8–9 in all treatments, PR10 in treatments A, B, and C, and PR11 in treatments A and B (Fig. 3B).

In the caudal fin, out of 28 elements, 13, 14, 15, and 14 cartilaginous elements were found to be significant by the logistic model in treatments A, B, C, and D, respectively. The elements common to all treatments (i.e., 13) are E1–2, H5–6, HS of PU3–5, NS of PU2–PU5, and UN2. Most elements chondrified between approximately 14 and 18-mm SL with the exception of HS of PU3 and NS of PU2, which form respectively at approximately 12.5 and 13-mm SL. As for the dorsal and anal fins, water velocity increase has little effect on the chondrification onset (Fig. 2C). However, compared to the dorsal and anal fins, most caudal elements display chondrification onset plasticity among treatments. HS of PU3 and NS of PU2 have greater onset plasticity among treatments; compared to other treatments, onsets values are lower in treatment C for HS of PU3 and in treatments B and C for NS of PU2. The logistic model was significant for all ossified elements (i.e., 28) analyzed, independently of treatments (Fig. 3C). Caudal elements begin to ossify at approximately 15.5-mm SL, all of them being ossified at approximately 21.5-mm SL. As in the dorsal and anal fins, ossification onset of caudal elements is more influenced by an increase of water velocity than chondrification. Similarly to the dorsal and anal fins, elements that ossify late in the sequence (all elements with the exception of ST, PH, H1–6, and the anteriormost uroneural UN2) are the most responsive. These elements display greater onset plasticity among treatments (Fig. 3C). Interestingly, onsets of late ossified elements show an inversion of response between treatments C and D; late ossified elements in treatment D have smaller SL50 compared to treatment C.

These results reveal four main patterns of developmental progress. First, developmental progress of chondrification is less influenced than ossification by an increase of water velocity in all fins. Second, the most responsive elements, meaning those elements with greater onset plasticity owing to a water velocity increase, are the elements that chondrify (i.e., especially peripheral proximal and distal radials in the dorsal and anal fins) and to a greater extent the elements that ossify late. Third, responsive elements show a smaller SL50 in faster velocity (treatment A) for all fins; they chondrify and to a greater extent ossify at a comparatively smaller size with a water velocity increase. These elements correspond to the hypurals in the caudal fin as well as the central proximal and distal radials in the dorsal and anal fins. Fourth, for all treatments, the onset of chondrification of the various elements show overlapping values among fins, whereas the onset of ossification has distinctly smaller values for all caudal elements than those in the dorsal and anal fins. Caudal elements are all ossified before the elements of the dorsal and anal fins initiate their ossification.

Comparison of Relative Developmental Sequences

The effect of water velocity on the changes in terms of the relative timing of endoskeletal elements of the median fins was analyzed by comparing congruence of chondrification and ossification sequences (i.e., relative order of events) by using Spearman correlation coefficients (Table 1).

Table 1. Spearman Rank Correlation Coefficients (rs) Comparisons of Developmental Sequences in S. alpinusa
FinsTreatment comparisonsChondrificationOssification
rsPrsP
  • a

    Coefficients are given for treatment comparisons of chondrification and ossifications sequences of median fin endoskeleton. Only non-redundant pairwise comparisons among treatments are given for each sequence within each fin. P is the probability associated with the respective rs. n is the number of endoskeletal elements involved within each fin for chondrification and ossification sequences.

DorsalA vs B0.929< 0.0005 (n = 16)0.954< 0.0005 (n = 11)
 A vs C0.965< 0.0005 (n = 16)0.954< 0.0005 (n = 11)
 A vs D0.931< 0.0005 (n = 16)0.925< 0.0005 (n = 11)
 B vs C0.984< 0.0005 (n = 16)1.000< 0.0005 (n = 11)
 B vs D0.988< 0.0005 (n = 16)0.984< 0.0005 (n = 11)
 C vs D0.986< 0.0005 (n = 16)0.984< 0.0005 (n = 11)
AnalA vs B0.860< 0.0005 (n = 19)0.966< 0.0005 (n = 9)
 A vs C0.859< 0.0005 (n = 19)0.974< 0.0005 (n = 9)
 A vs D0.862< 0.0005 (n = 19)0.928< 0.0010 (n = 9)
 B vs C0.993< 0.0005 (n = 19)0.974< 0.0005 (n = 9)
 B vs D0.996< 0.0005 (n = 19)0.944< 0.0005 (n = 9)
 C vs D0.996< 0.0005 (n = 19)0.970< 0.0005 (n = 9)
CaudalA vs B0.930< 0.0005 (n = 13)0.956< 0.0005 (n = 28)
 A vs C0.949< 0.0005 (n = 13)0.953< 0.0005 (n = 28)
 A vs D0.993< 0.0005 (n = 13)0.941< 0.0005 (n = 28)
 B vs C0.967< 0.0005 (n = 13)0.943< 0.0005 (n = 28)
 B vs D0.929< 0.0005 (n = 13)0.932< 0.0005 (n = 28)
 C vs D0.945< 0.0005 (n = 13)0.961< 0.0005 (n = 28)

In all fins, patterns of chondrification and ossification are strongly congruent among treatments as shown by the highly significant correlation coefficients (e.g., in all fins for chondrification and ossification P < 0.0005 for all treatment comparisons, with the exception of the anal fin ossification where P < 0.001 for comparison between treatments A versus D, Table 1). Thus, the relative timing of chondrification and ossification among endoskeletal elements within each one of the fins is fairly consistent among treatments, suggesting a high stability of median fin skeletal development (e.g., Spearman correlation coefficients are for chondrification 0.859 < rs < 0.996, and for ossification 0.928 < rs ≤ 1.000, Table 1). An increase of water velocity has induced only minor changes in the temporal order of sequences of the endoskeletal elements.

Transition of Skeletal States

In chondral bones (bony elements preformed of cartilage), the inherent switch of states from cartilage to bone provides a characteristic referred to as transition of states that can be derived from differential onsets. The transition of states is a duration calculated as the difference between the onset of ossification and the onset of chondrification for a specific developmental event. In order to evaluate changes in the transition of skeletal states with respect to differential water velocity, we used the minimum convex polygons (MCP) method to quantify the variation among the four treatments for each significant element (see details in Experimental Procedures section).

At the level of element, the MCP display plasticity in the transition of skeletal states among treatments for each median fin (Fig. 4). In relation to the area of polygons (mm2), elements were classified into six classes of variation estimated by the Sturge's rule (Scott, 2009): class 1 = [0–0.375], class 2 = [0.375–0.75], class 3 = [0.75–1.125], class 4 = [1.125–1.5], class 5 = [1.5–1.875], and class 6 = [1.875–2.25]. For purposes of synthesis, the classes were arbitrarily regrouped within three groups of variation: low (i.e., small area including classes 1 and 2), middle (i.e., intermediate area including classes 3 and 4), and high (i.e., large area including classes 5 and 6) variation (Table 2).

Figure 4.

Transition of skeletal states between chondrification and ossification onsets in S. alpinus. Treatment comparisons are reported for endoskeletal elements of dorsal (A), anal (B), and caudal (C) fins. States transitions are expressed on Y-axis by ossification minus chondrification onsets (mm) in function to chondrification onsets (mm) in X-axis. MCP's represent the amount of variation among treatments in transition of states for each element. All possible treatment comparisons are shown for those elements that are comparable between each state and common to all treatments.

Table 2. Polygon Area and Class of Variation Among Treatments for Transitions of Skeletal States in Median Finsa
FinsElementsPolygon area (mm2)Variation classbVariation groupc
  • a

    Sample size n = 22.

  • b

    The number (k) and width (h) of class of variation were defined by using Sturge's rule (Scott, 2009) (see Experimental Procedures section) as follows: k = 5.4 (6 classes) and h = 0.375 mm2, with a = 0.037 mm2 as the lowest value of the sample and b = 2.066 mm2 as the largest. Each class of variation in mm2 was denoted with numbers from 1 to 6 as follows: 1 = [0–0.375], 2 = [0.375–0.75], 3 = [0.75–1.125], 4 = [1.125–1.5], 5 = [1.5–1.875], and 6 = [1.875–2.25].

  • c

    For purpose of clarity (see Results section), the six classes of variation were regrouped arbitrarily into three groups of variation defined as follows: low = classes 1 and 2, mid = classes 3 and 4, and high = 5 and 6.

DorsalPR11.2634mid
 PR20.0491low
 PR30.0921low
 PR40.3521low
AnalPR12.0666high
 PR20.8503mid
 PR30.1221low
 PR40.0391low
 PR50.2321low
CaudalE10.1921low
 E20.2531low
 H50.0971low
 H60.1861low
 HSPU31.6675high
 HSPU40.4592low
 HSPU50.0881low
 NSPU21.3494mid
 NSPU30.2811low
 NSPU40.1071low
 NSPU50.3852low
 ST0.0371low
 UN20.0571low

In the dorsal fin, only four elements (proximal radials; PR1–4) common to all treatments allow comparisons of the transition of skeletal states (Fig. 4A). The polygonal areas of these four elements belong to two classes of variation: PR2–4 belong to class 1 (low variation) and PR1 belongs to class 4 (middle variation) (Table 2). Independently of treatments, these elements are distinguished by their size at chondrification; the elements with earlier chondrification onset display low variation, whereas elements with later chondrification onset show middle variation (Fig. 4A). Also independently of treatments, these elements collectively show a general positive linear relationship between greater size at chondrification and longer transition of states between chondrification and ossification. At the level of the element, no clear pattern appears among treatments. However, the fastest velocity shows a greater size at chondrification for three elements (i.e., PR1, PR3–4), which is associated with a shorter transition of states compared to the other treatments (Fig. 4A).

In the anal fin, five elements (PR1–5) common to all treatments allow comparisons of the transition of skeletal states (Fig. 4B). In contrast to the dorsal fin, the polygonal areas of these elements are distributed within three classes of variation: PR3–5 belong to class 1 (low variation), PR2 to class 3 (middle variation), and PR1 to class 6 (high variation) (Table 2). Similarly to the dorsal fin, the amount of variation among treatments has a positive linear relationship with respect to the size at chondrification; elements with earlier chondrification onset defined a low-mid variation group with shorter transition of states, whereas the elements with later chondrification onset display higher variation with longer transition of states between chondrification and ossification. As in the dorsal fin, no clear pattern appears at the level of each element among treatments. However, similarly to the dorsal fin, the fastest velocity shows greater size at chondrification for two elements (i.e., PR1 and PR3), which is associated with a shorter transition of states; this pattern is also observed for the medium velocity with respect to PR5 (treatment B, Fig. 4B).

In the caudal fin, 13 elements (E1–2, H5–6, HS of PU3–5, NS of PU2–5, ST, and UN2) were found to be concordant among treatments for comparisons of the transition of skeletal states (Fig. 4C). Based on the polygon areas, caudal elements belong to four classes of variation: class 1 (E1–2, H5–6, HS of PU5, NS of PU3–4, ST, and UN2), class 2 (HS of PU4, NS of PU5), class 4 (NS of PU2), and class 5 (HS of PU3) (Table 2). Elements within classes 1 and 2 display small polygonal areas that characterized the low variation group, whereas the elements of classes 4 and 5 show mostly intermediate to large polygonal areas characterizing a mid-high variation group. Contrary to the dorsal and anal fins, these groups show no linear relationship with respect to the size at chondrification but rather to the transition of states between chondrification and ossification. Elements with lower variation display shorter transition of states, whereas elements belonging to the mid-high variation groups show longer transition between states. Size at chondrification onset shows overlapping values among elements in both groups. As in the dorsal and anal fins, the low variation group displays no treatment pattern at the level of an element; elements of the mid-high variation group show a treatment response with respect to size at chondrification and transition of states. For almost all elements, a greater size at chondrification onset is associated with a shorter transition between states at least for one of the two fastest treatments (i.e., treatments A or B), whereas smaller sizes at chondrification onset are mainly associated with longer transition of states in medium velocity (Fig. 4C).

Collectively these results define two main patterns. The first pattern deals with the dorsal and anal fins that show similarities in terms of the groups of variation among treatments; these patterns differ from those of the caudal fin. Contrary to the caudal fin, the groups of variation for the dorsal and anal fins show positive linear relationship with respect to early versus late chondrification onset, with concomitant modifications of the transition of skeletal states. The second pattern applies to almost all elements, especially in later formed ones, and corresponds to a trend towards a delayed chondrification at higher water velocities (i.e., greater size at chondrification onset) but concomitantly transitions between cartilaginous and bony states are shortened. Overall, the transitions of skeletal states show plastic responses; the timing of formation (early versus late) of elements highly influences the level of plasticity among treatments.

Cartilaginous and Ossified Trajectories

A developmental trajectory deals with the relation between the onsets of a series of developmental events and the size of individuals at which these events occur. As such, a developmental trajectory is based on numerical changes marked by the incremental appearance of developmental events along a growth series (Schoch, 2004). Thus, cartilaginous trajectory (CarT) and ossified trajectory (OsiT) refer, respectively, to the cumulative number of chondrified and ossified elements observed through development.

Dorsal fin CarTs are similar among treatments and show four phases (Fig. 5A): (1) a rapid increase in the addition of cartilaginous elements from approximately 13 to 18-mm SL, (2) a short plateau extending between 18 to 22-mm SL revealing a slowdown in the addition of cartilaginous elements, (3) a renewed increase in the addition of cartilaginous elements from approximately 22 to 30-mm SL but to a lesser extent comparatively to the first phase, and (4) a plateau reached after 30-mm SL revealing completion of addition (i.e., no more elements are added). At the fourth phase, treatment A displays the highest level of completion compared to intermediate treatments B and C, whereas the slower treatment D displays the lowest level of completion (Fig. 5A). Anal fin CarTs do not differ among treatments and are highly similar to that of the dorsal fin, including phases and differential levels of completion among treatments at the fourth phase (Fig. 5B). Caudal fin CarTs are identical among treatments, but differ greatly in phases from those of the dorsal and anal fins. In all treatments, the caudal CarTs show a negative-exponential shape with three phases (Fig. 5C): (1) a rapid increase in the addition of cartilaginous elements from approximately 13 to 22-mm SL, (2) a reduction in the addition of elements from approximately 22 to 30-mm, and (3) a completion in the addition of elements after approximately 30-mm SL.

Figure 5.

Cartilaginous (CarT) and ossified (OsiT) trajectories in S. alpinus. Comparisons are reported for each water velocity treatment in dorsal (A, D), anal (B, E), and caudal (C, F) fins. Colored lines are from DWLS smoothing method for each treatment. Each symbol represents a specimen. Vertical dotted line represents phase transitions in trajectories patterns within each fin (see Results section for details).

As for the dorsal fin CarTs, dorsal OsiTs are fairly consistent among treatments but follow a sigmoid shape with three phases instead of four (Fig. 5D): (1) an absence of ossified elements from approximately 13 to 22-mm SL, (2) a rapid increase in the addition of ossified elements from 22 to approximately 30-mm SL, and (3) a plateau in the addition of elements reached at approximately 30-mm SL. Similarly to the last phase in the dorsal CarTs, the OsiT from treatment A has the highest level of completion, whereas treatments B and C have intermediate levels followed by treatment D with the lowest level of completion. As for the anal fin CarTs, the anal fin OsiTs (Fig. 5E) are highly similar among treatments as well as among those of the dorsal fin OsiTs, including the number of phases and levels in the addition of elements among treatments at the final phase (Fig. 5D). Caudal fin OsiTs (Fig. 5F) are identical among treatments and are highly similar to the caudal fin CarTs (Fig. 5C) including their negative-exponential shape and their three phases. The initiation of the caudal OsiT (Fig. 5F) seems to precede that of the CarT (Fig. 5C); however, it is only artefactual owing to the graphic representation using the smoothing method.

Four main patterns emerge from the CarTs and OsiTs. First, the chondrification and ossification trajectories are highly similar among treatments within each fin. Second, the chondrification and ossification trajectories of the dorsal and anal fins are similar but differ from those of the caudal fin. Third, the chondrification and ossification trajectories of the caudal fin are identical, whereas they differ in the dorsal and anal fins. Four, plasticity occurs among treatments during the last phase of the chondrification and ossification trajectories of the dorsal and anal fins meaning that the cumulative number of chondrified or ossified elements differs among treatments at comparatively similar size after approximately 30-mm SL.

DISCUSSION

In this study, we demonstrate plasticity of developing endoskeletal elements with respect to exercise training (i.e., differential water velocity). Developmental plasticity is here confirmed by changes in timing and developmental progress of skeletal events, whereas developmental sequences and trajectories are more conservative. In our view, these findings are of primary importance because the identification of plasticity/stability patterns of the skeletal system development shed light on uncovered patterns of phenotypic modularity.

From the results presented above, three major lines of evidences can be summarized. First, developmental progress of cartilages and bones and the transition of skeletal states are plastic in each fin; plasticity is here suggested owing to the changes in timing of several developmental events. Second, patterns of relative developmental sequences and developmental trajectories are fairly conservative among treatments in each fin; only minor changes have been experimentally (environmentally) induced in terms of the relative order of developmental events and the incremental addition of endoskeletal elements through time. Third, the anatomical and the developmental patterning of the dorsal and anal fins is highly similar; these patterns provide strong support for the hypothesis of dorso-anal fin modularity found to be a general condition within living actinopterygians (Mabee et al., 2002). In addition, anatomical and developmental patterns of the dorsal and anal fins differ greatly from those of the caudal fin. Such distinctiveness as well as the congruence of plastic responses among treatments within the caudal fin might be indicative of modularity. We will tentatively discuss aspects of skeletogenesis plasticity/stability with regards to changes of environmental conditions in the context of phenotypic modularity. However, we first address plasticity in the timing of developmental events.

The vertebrate endoskeleton is known to be plastic (Herring, 1993). Its development, growth, and maintenance are the results of dynamic interplays among genetic architecture, changes in environmental conditions, and external stresses (Herring, 1993; Mao and Nah, 2004; Lall and Lewis-McCrea, 2007; Young and Badyaev, 2007). One of the stressors acting as an epigenetic component on developing endoskeletal elements is the mechanical loading generated by the muscular contraction. These loadings are modulated by the activities necessary for functions such as mobility and locomotion (Danos and Staab, 2010, Fiaz et al., 2010). Various experimental analyses of in vivo paralysis in developing chick embryo show an increase in fusion of cervical vertebrae, smaller length of long bones (Hosseini and Hogg, 1991), and a decrease in bone surface in hind limbs (Lamb et al., 2003). In addition to these structural modifications, Hosseini and Hogg (1991) have observed a slight delay in the timing of ossification onset in several bones (i.e., thoracic and lumbar vertebrate, neural arches of cervical vertebrae, and phalanx of the third digit) of paralyzed chick embryos. This latter point is concordant with our findings on the developmental plasticity in relation to the timing of median fin endoskeletal elements. Indeed, we found out that the effect of an increase of water velocity mainly induces changes in the developmental progress in all fins; the responsive elements ossify at comparatively smaller size at faster velocity. It is suggested that rearing fish in differential water velocity may affect their locomotion and thus may change the mechanical stress exerted on developing endoskeletal elements (Fischer-Rousseau et al., 2009; Cloutier et al., 2010; Fiaz et al., 2010, 2012). In this context, several lines of evidences have shown the correlative link between environmental changes such as differential water velocity and plasticity in timing of skeletogenesis (rather than changes in the relative order of developmental events) in fishes. Zebrafish (Danio rerio) larvae subjected to swimming exercise showed an acceleration of ossification for the caudal fin elements; the ossification of hypurals and lepidotrichia occurs 15 days post-fertilization earlier in trained fish than in control fishes (Unpublished thesis). In two recent studies, it has been shown that exercise training changes the timing of formation of cartilages and bones in fishes (Cloutier et al., 2010; Fiaz et al., 2012). By using different specimens and treatments, Cloutier et al. (2010) have shown changes in the timing of ossification in specimens of S. alpinus reared at lower water velocities (i.e., slow velocity = 0.38 cm s-1 and fast velocity = 0.51 cm s-1) than tested in the present study. It was shown that endoskeletal elements of the dorsal and anal fins express a significant average acceleration of ossification of 4.8 and 5.4 days, respectively, in the fastest velocity. However, changes in timing were not accompanied by significant changes in the relative order of developmental events, which is congruent with our results. Developmental plasticity in the timing of skeletogenesis is not limited to experimentally induced changes in environmental conditions. By comparing the skeletogenetic development of yearling Brook charrs (Salvelinus fontinalis) captured in stream and lake, Fischer-Rousseau et al. (2009) have demonstrated that replace by modeled ossification events (i.e., using replace by modeled onset with logistic regressions) for the complete postcranial skeleton occur at a significantly smaller size in stream fishes; on average, the size at ossification is 0.719 mm smaller in stream fishes.

These various studies, representing a diversity of vertebrate taxa, point towards a correlative link between changes in environmental conditions, epigenetic effect of mechanical loading by modulation of locomotion exercise, and plasticity in the timing of ossification of developmental events. However, our experimental treatment influences not only the timing of ossification but also affects the timing of formation of cartilages but to a lesser extent. The developmental progress of chondrification displays differences in the timing among treatments for some elements in all median fins. These changes are indicative of plasticity. At the anatomical level, it might be hypothesized that the genetic component is more influential than the epigenetic factors for the development of cartilages, whereas even if the ossification process is genetically controlled, a greater epigenetic component is involved. The genetic-epigenetic relationship is exemplified in cranial and long chondral bones of mammals where the ossification is epigenetically dependent on the normal development of cartilages (Hallgrímsson et al., 2007) and the necessary well-coordinated molecular machinery that controls the formation of cartilage and its subsequent ossification (Nakashima and de Crombrugghe, 2003). It has been further stated, at least in guenons, that the chondrocranium is the cranial skeletal system less influenced by epigenetic factors in response to changes in environmental conditions (Cardini and Elton, 2008). However, the diversity of cartilaginous tissues in fish (Witten and Huysseune, 2007; Witten et al., 2010), the mechanisms involved in bone resorption and remodeling (Witten and Huysseune, 2009), the presence or absence of osteocytes and cell processes in basal versus derived teleosts (Parenti, 1986) that mediates mechanostimulation (Fiaz et al., 2010), as well as the lower weight-bearing role because of lower gravity pressure (Moss, 1963; Huysseune, 2000) make the fish endoskeleton highly different from that of terrestrial vertebrates. Consequently, the genetic–epigenetic relationships in developing chondral bones may differ between fish and terrestrial vertebrates.

We can hypothesize that both for chondrification and ossification, developmental plasticity may increase if induced early in development (Wund et al., 2008). The point raised here is the susceptibility of response of the developing phenotype to changes in environmental conditions. The susceptibility refers to the ability of developing systems to be receptive and to integrate environmental changes (Schlichting and Pigliucci, 1998; West-Eberhard, 2003; Young and Badyaev, 2007). In the case of developing skeletal tissues, time is necessary to cope with environmental changes. However, fish endoskeleton is remarkable in that it develops relatively late in ontogeny compared to mammals and birds; most of it occurs after hatching (Fiaz et al., 2010; Sæle and Pittman, 2010). Even post-hatching exposure to increased water velocity is sufficient to induce plastic responses in timing, primarily with respect to later formed bones and cartilages to a lesser extent. The effect of differential water velocity on the timing of developmental events increases with respect to the order of fin formation, meaning the caudal fin first, the dorsal fin second, and the anal fin third. This is consistent with our results on developmental progress of chondrification and ossification. In both cases, the most responsive elements are the latter formed peripheral elements in the dorsal and anal fins, and the neural and haemal elements of preural centra in the caudal fin. Similar lines of evidence were recently reported (Cloutier et al., 2010; Young and Badyaev, 2010; Fiaz et al., 2012) indicating that the latter formed elements in developing endoskeleton have a greater susceptibility to display plastic response with regards to changes in environmental conditions (or epigenetic factors). By analyzing the mechanical load exerted through muscle stimulation on the developing bony foraging apparatus of shrews (Sorex monticolus), Young and Badyaev (2010) found that the size and shape of the mandible is strongly influenced in late, but not in early, ossifying regions of the bone. They suggest a correlative link between differential mandible morphology of late ossifying regions and bite force functional requirements; plasticity in the timing of the ossification among regions serves as a developmental pathway for both local adaptation and diversification of mandible morphology (Young and Badyaev, 2010).

Cartilage is neither more genetically or epigenetically influenced than bone; they are both influenced differently and have different functions. It appears that the susceptibility of developing cartilages and bones to express plastic responses is a matter of “when.” For developing skeletal elements, displaying plastic responses is a matter of timing and exposure to environmental changes, and functional requirements rather than simply a question of genetic versus epigenetic predominance. Developing endoskeletal tissues are by-products of genetic–epigenetic interactions and functional requirements (Mao and Nah, 2004; Young and Badyaev, 2010).

Perhaps, the most unexpected result from our study is the plastic responses in the transition from cartilage to bone. Transitions of skeletal states are influenced by the water velocity increase. Responsive elements display a relative trend towards a delayed chondrification at higher velocity compared to the other treatments, but with concomitant shorter transitions between the cartilaginous and bony states. These observations suggest that chondrocytes of the cartilage precursor may act as mechanosensors, and thus may actively participate in the regulation of the transition of skeletal states in differential mechanical loading environments [see Fiaz et al. (2010) for a recent review on mechanosensing of vertebrate skeleton]. Various experiments on chondrocytes have demonstrated that mechanical loading promotes chondrogenesis (Takahashi et al., 1998, 2009; Wu et al., 2001; Wong et al., 2003). Mechanical loading is also required for osteogenesis (Ikegame et al., 2001; Komori, 2005; Bonewald and Johnson, 2008). While chondrocytes act in mechanosensing of vertebrate chondrogenesis, osteocytes (not osteoblasts) are recognized as the main bone mechanosensors especially in mammals (Bonewald and Johnson, 2008; Santos et al., 2009). However, in contrast to mammals, two major types of bone are encountered in fishes: acellular or anosteocytic (devoid of osteocytes) and cellular or osteocytic (with osteocytes) (Huysseune, 2000). The occurrence of acellular and cellular bones differs between young [larval teleosts often lack osteocytes (Fiaz et al., 2010)] and old fishes. Generally, salmonid endoskeleton is recognized as cellular (Witten and Huysseune, 2009). Although Arctic charr is a salmonid, we only used newly-hatched alevins up to early juveniles in our experiment suggesting that bone development has not reached its complete adult phenotype in terms of osteocyte characteristics. Thus, we hypothesize that the role of osteocytes as mechanosensors in our specimens is at best minimal. The existence of an alternate mechanosensing system, composed of osteoblasts and bone lining cells, is suggested in fishes (Witten and Huysseune, 2009; Totland et al., 2011). Totland et al. (2011) suggest that the osteoblasts on the bone surface of the vertebrae may act as a co-localized system of mechanical load detection and response. In our study, skeletal elements are preformed by cartilage contrary to vertebrae. However, we have to consider that osteoblasts at the perichondrium surface of the cartilaginous precursor may participate (may be in conjunction with chondrocytes) to add a fine-tuning mechanosensing dimension in the transition from cartilage to bone. In newly-hatched developing fish, interactivity between chondrocytes and osteoblasts may help chondral bones to sense and react to differential environmental conditions by adjusting the transition of skeletal states. However, to our knowledge no studies on chondral bones have molecularly, developmentally, and anatomically investigated these transitional aspects. A prospective regulatory system for the adaptive transition of skeletal states may involve nitric oxide (NO), an inter- and intracellular signaling molecule (Chowdhury et al., 2006; Vatsa et al., 2006, 2007). Indeed, a mechanical loading correlative link has been shown with NO down-release and chondrocytes proliferation (Chowdhury et al., 2006) and with NO production/up-release and osteoblastic/osteocytic stimulation (Vatsa et al., 2006, 2007). We suggest that in chondral bones, such a system may participate as an effective timing regulator for onset of developmental events. As such, it possibly plays a key role in the order of element formation within sequences of endoskeletal systems by adapting the development of required functional systems (such as fins) with differential environmental conditions.

We found out that chondrification and ossification sequences are highly conservative, suggesting only minor changes in the relative order of developmental events within the sequences, regardless of the treatment. This is consistent with previous findings where, even under differential environmental conditions, sequences of chondrification and ossification are fairly stable (Mabee et al., 2000; Cloutier et al., 2010); the reported variation in relative order of events was low and often characterized by shifts of one or two positions of responsive elements within the sequences (Hanken and Hall, 1984; Alberch and Blanco, 1996; Mabee et al., 2000; Cloutier et al., 2010). However in these studies, onset of events was determined by direct observation of only a few specimens instead of being modeled as such potential bias might have been introduced in the relative order of events within sequences.

In addition to the relatively limited changes in developmental sequences for each fin, developmental trajectories are also fairly conservative among treatments. However, compared to the overall stability in the caudal fin, plasticity occurs during the last phases of the CarTs and OsiTs for the dorsal and anal fins. This pattern fits well with our proposed hypothesis of increased changes in timing of latter formed elements associated with the differential order of fin formation. However, the reason is unclear for observing an overall stability in developmental trajectories and developmental sequences independently of the treatments at the level of the morphological systems (i.e., each fin), whereas timing and transition of skeletal states display plastic responses at the level of developmental events. We interpret these patterns as a result of differential scale effects on developing modular systems and the susceptibility of their components to change with respect to differential environmental conditions.

In a morphological perspective, a developmental module can be defined as part of an embryo that is quasi-autonomous with respect to pattern formation and differentiation (Wagner et al., 2007). Such embryonic (and larval) regions share more connections internally (i.e., constitutive units tightly connected to or interactive with each other) than they do externally to other regions (Hallgrímsson et al., 2009). Connection tightness or interactions among constitutive modular units result from the selective establishment of a specific developmental architecture at the interplay between integration (i.e., shared functions among units) and parcellation (i.e., absence of interactions among units) (Wagner, 1996; Hallgrímsson et al., 2009). Although modular organization is thought to reflect greater internal than external connections among constitutive units, its internal organization may be subject to a certain amount of change (Raff, 1996; Gilbert and Bolker, 2001; Callebaut, 2005). If changes in environmental conditions release cryptic genetic variation (Schlichting, 2008) or introduce newly genotypic-phenotypic variation (Hallgrímsson et al., 2009), it could selectively change the underlying developmental architectural properties of the integrated and modularized developmental systems, and thus influence their evolvability [for further discussion see Hallgrímsson et al. (2009)]. In this context, we interpret our results of plasticity in timing and transitions of skeletal states of latter formed endoskeletal elements as the possible amount of change in modular organization. As mentioned earlier, our hypothesis of developmental modular organization for the dorsal-anal fins and the caudal fin is indicated by similarity of anatomical and developmental patterns as generally found in actinopterygian fishes (Mabee et al., 2002) and recently proposed in S. alpinus (Cloutier et al., 2010). Modular organization of median fins is further supported by the overall stability of the CarTs and OsiTs and the developmental sequences. Developmental sequence stability is consistent with the expectation that sets of developmental events that constitute modules should be especially constrained in their order (Smith, 2001; Poe, 2004). We are aware that hypotheses of modularity and integration are often investigated using measures of phenotypic covariation or correlation (Hallgrímsson et al., 2009). According to Goswami (2007), the usage of developmental sequences based on ossification provides a complementary powerful proxy to these patterns of covariation/correlation of modular components for the analyses of stability/variability of phenotypic modularity and integration. However, we have demonstrated that not only are patterns of ossification informative, but patterns of chondrification as well as other properties of developmental events can be used efficiently to validate hypotheses of modularity.

Conclusions

The vertebrate endoskeleton provides a unique opportunity to study how environmental and epigenetic factors can influence the origin of novelties (Danos and Staab, 2010). Fishes offer an outstanding avenue to investigate the timing plasticity during early versus late skeleton development because (1) their cartilages and bones are highly dynamic tissues, (2) resorption and remodeling processes occur throughout their life span, and (3) they never stop growing (Witten and Huysseune, 2009). In doing so, studies on the patterns of plasticity may serve as indicators of phenotypic modularity in the endoskeleton of fishes.

Even if developmental sequences, especially those derived from chondrification and ossification of cartilaginous bones, are useful data to address variation, stability, and evolvability of development, the extent and meaning of evolutionary changes in developmental timing cannot be addressed in a replicable manner without adequate characterization of proxy for developmental time (Sæle and Pittman, 2010). Indeed, to adequately fulfill the comparative issue of development among taxa within an evolutionary perspective, objective methodologies should be used in order to disentangle the potential intraspecific variation. The use of various properties of development events (onset, offset, duration, and transition of skeletal states) in chondral bones provides a more accurate picture of the plasticity because the bony state may respond differently to environmental changes than its cartilaginous counterpart. Furthermore, in the specific case of skeletal developmental sequences, the first objective is to estimate accurately the timing of developmental events. From our perspective, this is the very first step prior further evolutionary analyses such as developmental sequence polymorphism (Colbert and Rowe, 2008), patterns of modularity (Poe, 2004; Goswami, 2007), or analyses of sequence heterochrony (Jeffery et al., 2005; Goswami, 2007; Harrison and Larsson, 2008). An environmental component to developmental studies should be integrated only after patterns of stability and variation have been assessed. Furthermore, experimental consideration of developmental plasticity would ideally be investigated with model organisms with low genetic variability. We suggest that experimental studies are crucial to investigate the extent to which stability/plasticity in timing, order, and trajectories of developmental events may change modular organization and promote morphological evolvability.

EXPERIMENTAL PROCEDURES

Specimens Examined

Approximately 13,000 eggs of the “Fraser” Arctic charr (S. alpinus) strain were obtained from four males and four females and divided by the producers (Aquarium and Marine Center of Shippagan, New Brunswick, Canada) into eight different lots of approximately 1,625 eggs each. This procedure was done to randomly distribute the potential genetic variation owing to the different parents. The low number of parents limits the potential genetic variability. After incubation, specimens were experimentally reared during 100 dph. Newly hatched fishes were submitted to four constant water velocity treatments (with laminar flow) throughout the experiment that took place between December 2004 and April 2005. Velocity treatments (two swimming canals per treatment), given in absolute values, correspond to water velocity of 3.2 cm s-1 (A = fast treatment), 1.6 cm s-1 (B = medium treatment), 0.8 cm s-1 (C = slow treatment), and 0.4 cm s-1 (D = still treatment, reference treatment). More details on the incubation conditions of fishes and the channel swimming setup have been described elsewhere (Grünbaum et al., 2007, 2008).

To address the water velocity effect on endoskeletal developmental plasticity, two specimens from each treatment were sampled every other day from day 0 to day 100 (102 specimens per treatment). Each sampled fish was then fixed in neutral buffered formalin (Presnell and Schreibman, 1997) for two days and then transferred to 70% ethanol. A total of 408 specimens were cleared and double stained with Alcian blue for cartilage and Alizarin Red S for bone following the procedure adapted from Dingerkus and Uhler (1977) with modifications from Potthoff (1984) and Mabee (1993) for larval fishes. We studied the development of 83 chondral endoskeletal elements from the dorsal, anal, and caudal fins (Fig. 2). All skeletal elements considered in this study are chondral bones, meaning that they ossify from cartilaginous precursors (Patterson, 1977; Huysseune, 2000; Bird and Mabee, 2003). The number of endoskeletal elements followed during the study was 27, 25, and 28 in the dorsal, anal, and caudal fins, respectively. These numbers reflect the maximum number of elements encountered within each fin.

Nomenclature terminology for median fins follows that of Schultze and Arratia (1989), Arratia and Schultze (1992), Arratia et al. (2001), Grünbaum and Cloutier (2010), and Cloutier et al. (2010). Figure 1 provides abbreviations used thereafter. Specimens are housed in the collection of the Laboratoire de Biologie Évolutive de l'Université du Québec à Rimouski (UQAR), Québec, Canada. All procedures were approved by the Université du Québec à Rimouski Committee on Animal Care.

Data Analyses

Cleared and stained specimens were examined with a Leica MZ16A binocular microscope. For each endoskeletal element, three developmental states were scored corresponding to: its absence (state 0), its cartilaginous state (state 1), and its ossified state (state 2). An element was coded to be cartilaginous when it took up Alcian blue and ossified when it took up Alizarine Red S; stain uptake and careful anatomical observation of each element were done to accurately code for the phase of each element. Fisher and Mabee (2004) have pointed out that the cleared-and-stained method is reliable for fine morphological description and determination of relative developmental sequences. We would like to draw attention to the fact that the Alcian blue 8GX presently sold has been changed in its chemical formula without warning, resulting in poor staining and more intensive labor. For those laboratories using cleared-and-stained methods please refer to Redfern et al. (2007) for an efficient Alcian blue dye which is the Alcian Blue-Tetrakis form.

We used standard length (SL) as a proxy for developmental time in our analyses. Age has been shown to be a metric more sensitive to environmental variation than size for a measure of developmental time (Fuiman et al., 1998) especially in skeletal ontogenesis (Bird and Mabee, 2003; Campinho et al., 2004). SL represents the distance between the tip of the snout and the posterior margin of hypurals (Mabee, 1993). The size range of the analyzed specimens (12.87 to 45.08-mm SL) encompasses transitions between late free embryo, alevin, and juvenile phases in Salvelinus (see Balon 1980, 1984) and covers most chondrification of the dorsal, anal, and caudal endoskeletal elements and their ossification, except for the distal radials of the dorsal and anal fins.

Developmental Progress

Developmental progress refers to the progressive modeled onset size of state 1 (i.e., cartilaginous state or chondrification) and 2 (ossified state or ossification) for a given element through development of our sampled individuals. Modeling of onset was performed by using logistic regressions (0.5 cut-point) with the Logit transformation (Fischer-Rousseau et al., 2009). We used the cut-point (SL50) (e.g., lethal dose in ecotoxicological studies, LD50, Mineau et al., 2008) to define, with respect to the observed onset size (SL), 50% of the individuals that show skeletal states 1 or 2 for a given element. The SL50 for an element, as defined by the logistic model, does not necessarily correspond to the SL at which the skeletal state was observed in individuals. Because skeletal states are ordinal (i.e., absent, cartilaginous, ossified), we estimated the SL50 using the proportional odds model in which the probability of an equal or smaller response was compared to the probability of a larger response (Hosmer and Lemeshow, 2000). Effect of differential water velocity on chondrification and ossification onsets was analyzed by converting treatments to dummy variables using the cell reference method (Hosmer and Lemeshow, 2000) with treatment D as the reference. Only the elements for which the logistic model was significant (P ≤ 0.05) were considered in our analyses. Significance of the logistic regression was tested using the G2 statistics, also termed the likelihood ratio χ2.

Comparison of Relative Developmental Sequences

Timing of endoskeletal elements was determined by using chondrification and ossification modeled onsets derived from the logistic model. Absolute timing of events (i.e., chondrification and ossification modeled onsets of elements) was converted into relative order by attributing rank values to each event within the endoskeletal sequences [the procedure has been described elsewhere (Grünbaum et al., 2003; Cloutier et al., 2010)]. Spearman rank correlation coefficients were then used to compare ranks congruence among treatments for chondrification and ossification sequences.

Transition of Skeletal States

The transition of skeletal states refers to the comparisons of modeled onsets among treatments. These comparisons are expressed as ossification minus chondrification modeled onsets in function to the modeled onsets of chondrification for every statistically significant element in each fin. These comparisons only use the elements common to all treatments. To evaluate changes in the transition of skeletal states, we used the MCP (also called a convex hull) method to quantify the variation among the four-treatment points. The MCP, a standard method widely used to measure species' range, estimates the smallest convex area for a set of points in which no internal angle exceeds 180° and that contains all data points (here treatments) (Burgman and Fox, 2003). The MCP areas (mm2) were calculated by using the Vegan package in R (Oksanen et al., 2011). In order to define classes of variation for the transition of skeletal states among treatments, we used the Sturge's rule to define objective classes of areas (Scott, 2009). The number of classes (k) was estimated as

equation image(1)

where n is the sample size. The width of each class was then denoted by h as follows

equation image(2)

where b and a correspond to the largest and the smallest value of the sample, respectively. The inferior limit of each class was then defined by the successive addition of the h value starting with the rounded-down value of a.

Cartilaginous and Ossified Trajectories

Cartilaginous trajectory (CarT) and ossified trajectory (OsiT) refer, respectively, to the cumulative number of chondrified and ossified elements observed through development. Each specimen displays a specific number of observed chondrified and/or ossified elements at any given length. For example, two 15-mm SL specimens of a given treatment can have five and eight elements already chondrified; note that even if two specimens have the same number of elements it does not imply that it corresponds to the same elements. For each fin per treatment, the number of skeletal elements observed for each specimen (i.e., based on 102 specimens per treatment) is reported with respect to SL. Cartilaginous and ossified trajectories were obtained within each fin for each treatment. We used the distance weighted least squares smoothing method (DWLS) (McLain, 1974) to fit a line through the data points with a local flex fixed at 0.5. Each smoothed curve represents a model trajectory of the cumulative number of developmental events with respect to size for their observed onset.

All statistical analyses were programmed and performed with SYSTAT (Version 11, SYSTAT Software Inc., Richmond, CA) with the exception of the MCP analyses performed with R (Version 2.13, R Development Core Team, Vienna, Austria).

Acknowledgements

We are deeply indebted to L. Fischer-Rousseau and A. Caron for helpful suggestions, discussions, statistical and editorial advices. Earlier versions of the manuscript have benefited from constructive comments by B.K. Hall and P.M. Mabee, as well as two anonymous reviewers.

Ancillary