Assessing a relationship between bone microstructure and growth rate: a fluorescent labelling study in the king penguin chick (Aptenodytes patagonicus)

The relationships between structure and function in compact bone tissues of vertebrates have been more thoroughly studied in secondary remodelled bone(e.g. human Haversian bone) than in primary bone (see review in Currey, 2002). A first and essential contribution came from Amprino(1947), who predicted that the structure of primary bone tissue can be understood as the expression of growth rates. Experimental fluorescent labelling has documented this functional relationship in some species (e.g. de Buffrénil and Pascal, 1984; Castanet et al., 1996, 2000). Accordingly, the current structural classification of primary bone tissues(de Ricqlès, 1975; de Ricqlès et al.,1991) relies on distinctions of tissue types mainly underlain by differing growth rates (e.g. slow-growing `lamellar-zonal' tissues vsfast-growing `fibrolamellar' tissues). However, some recent results challenge Amprino's theory; de Margerie et al.(2002) found that different types of fibro-lamellar bone tissue can grow at similar rates during the skeletal ontogeny of the mallard (Anas platyrhynchos), for instance. Conversely, Starck and Chinsamy(2002) reported a wide range of growth rate in the Japanese quail (Coturnix japonica), without any correlative shift of bone tissue type. These findings raise doubts about the validity of the relationship between growth rate and bone tissue type during skeletal ontogeny, especially within fibro-lamellar (i.e. fast-growing) bone tissues. Believing that some clarification of this controversy could arise from the study of a model species with extreme adaptations, we have investigated the periosteal (i.e. diametric) long bone growth of the king penguin (Aptenodytes patagonicus Miller).

King penguins are flightless seabirds with a unique growth strategy(Stonehouse, 1960; Barrat, 1976): between hatching and fledging (i.e. departure to sea), chicks spend almost one year on land in the breeding colony. They experience a first phase of rapid growth(approximately 3 months between February and early May). During this phase,chicks are intensively fed by their parents; they achieve a 40–60-fold increase in body mass (from 0.2 kg to 9–12 kg) and almost reach the adult size, which is among the largest in extant neognathe birds (1 m tall,∼12 kg; Prévost and Mougin,1970; Barrat,1976). A period of very low feeding rate follows, during the austral winter (4.5 months, May to mid-September). Chicks lose half of their body mass, and only chicks that have large initial body energy reserves have a high survival rate (Barrat,1976; Cherel et al.,1987). The last 3.5 months in the colony (mid-September to late December) start with the recovery of an intensive parental feeding rate, see ponderal recovery of chicks and end with chick moulting, immediately followed by their departure to open sea.

Our study focused on the first – very active – growth phase. Chicks were studied between their third and fifth week, i.e. when the growth rate of their flipper and leg bones was the highest(Stonehouse, 1960). Bone tissue growth rates were measured in the appendicular skeleton, coupled with the investigation of associated bone tissue types, which yielded appropriate data to test experimentally Amprino's theory(Amprino, 1947).

The field study was performed in the colony of the Baie du Marin,Possession Island, Crozet Archipelago (46°26′ S, 51°52′ E;Indian Ocean), where ∼40 000 pairs of king penguins breed(Weimerskirch et al., 1992). Four wild chicks, hatched in the first half of February, were triple-labelled with vital fluorescent dyes at the ages of 2, 3 and 4 weeks to measure bone growth rate. Fluorescein and alizarin solutions (1 g per 100 ml of sterile saline) were injected intraperitoneally with sterile single-use material at a dose of 30 mg kg^–1 body mass and 80 mg kg^–1body mass, respectively. Prior to the injections, solutions were sterilised using 0.2 μm Minisart high-flow filters (Sartorius, Goettingen, Germany). Fluorescein was injected at the end of weeks 2 and 4 and alizarin was injected at the end of week 3.

During these first weeks of life, chicks were either always with a parent or, during the time (∼5–10 min) a chick was manipulated in a nearby shelter for injections and body measurements, were exchanged with a dummy egg,a procedure well accepted by the parents. Until the chick was returned, the`incubating' parent was continuously observed. Each time they were manipulated, birds were weighed (±10 g) and their bill, flipper and foot lengths measured (±0.5 mm). Between each weekly manipulation,chicks were left undisturbed in the breeding colony with their parents and were observed daily. They were identified with coloured fish-tags implanted dorsally in the skin under local anaesthesia (1 ml xylocaine 2%) at the first manipulation. Each chick's parents were identified through dye marks on the chest. The growth curve (body mass and body measurements) of the four labelled chicks was compared with that of 30 unlabelled chicks born in the same area of the colony at similar dates (identified in the same manner as labelled chicks).

The labelled chicks were sacrificed by cervical vertebrae dislocation after being emancipated (i.e. left alone in the colony by their parents between two feedings) when they were 5 weeks old (except for one chick, who died accidentally during a storm 4 days after the third labelling and suffered partly from predation). They were kept frozen (–20°C) until further analysis.

The authorisation to perform these experiments in the breeding colony and to sacrifice chicks was obtained from the Ethics Committee of the Institut Français pour la Recherche et la Technologie Polaire (present name Institut Polaire Paul-Emile Victor). The study followed the `Agreed Measures for the Conservation of Antarctic and Subantarctic Fauna'.

Four long bones (humerus, radius, femur and tibiotarsus; Fig. 1) were removed per individual and embedded in polyester resin, after dehydration in graded ethanol and defatting in acetone and trichloroethylene. A 500 μm-thick mid-diaphyseal cross section was sawed out of each bone using a Presi P-100 diamond saw (Grenoble, France). Sections were glued to a glass object-holder with epoxy glue (Devcon, Riviera Beach, FL, USA) and ground to 80 μm using graded abrasive material.

Periosteal growth rates were measured optically under ultraviolet light on a fluorescence microscope (Zeiss Axiovert 35; Jena, Germany; magnification 50×), using the following method. Each section was observed in four orthogonal anatomical directions (e.g. anterior, posterior, dorsal and ventral). In each field, three successive fluorescent labels and the periosteum delimited three bone tissue `patches' (examples in Fig. 2), corresponding to week 3, week 4 or week 5 in the chick's growth. Each patch was characterised by a bone tissue type, identified under natural light, following de Ricqlès'classification of bone tissues (de Ricqlès, 1975; de Ricqlès et al., 1991). Using an optical micrometer, the thickness of each patch was measured once (±10 μm) as the distance between the outer borders of labels (Fig. 2) and then divided by the time elapsed between two labels to yield a local value of growth rate. Out of a theoretical sum of 192 (three patches per direction, four directions per section, four sections per chick,four chicks), we were unable to analyse 52 patches because, occasionally, the first (and even second) label was locally destroyed by perimedullar resorption. Moreover, the periosteum was sometimes difficult to locate precisely, invalidating some measures of week 5. The data from the 140 analysed patches are presented in Table 1.

We analysed our data using the following analysis of variance (ANOVA)design: growth rate (the quantitative variable) was considered as the dependent variable. Growth rates were log-transformed to satisfy the`homogeneity of variance' assumption of parametric ANOVA. This transformation also conferred a normal distribution to the 140 growth rate values. In agreement with the histological level of integration of our study, bone tissue patches were the statistical elements of the design (and not the four chicks,which constitute a grouping variable here). Four fixed factors (categorical grouping variables) were assumed to have a possible effect on growth rate:

1. chick (four levels: chick 1, chick 2, chick 3, chick 4);

2. week of growth (three levels: week 3, week 4, week 5);

3. long bone (four levels: humerus, radius, femur, tibiotarsus);

4. bone tissue type (four levels: longitudinal, reticular, radial, laminar;see below).

Unfortunately, as a consequence of working on limited and precious material, our data were insufficient to test the effects of all factors and interactions simultaneously in a complete controlled four-way ANOVA design. Indeed, our data set contains too few replications of measurements per cell in the table and a substantial amount of missing data. To circumvent this restriction, we first tested the difference between chicks and between weeks of growth using one-way ANOVAs. Since these tests returned no significant effects of the two factors (see Table 2), we pooled growth rate data across chicks and weeks of growth. This reduced the number of cells in the data set and therefore increased replication of measurements within each cell. This enabled us to test the effects of the two remaining factors (long bone and bone tissue type, which are of main interest here) in a complete controlled two-way ANOVA design. Post-hoc multiple comparisons were carried out using HSD Tukey test for unequal replication (5% experiment-wise error rate). They allowed groupings of non-significantly different levels within each factor.

Body mass and measurements at the beginning and end of the experiment were compared using paired t-tests. Ages of emancipation between labelled and non-labelled chicks were compared using t-test. Growth curves of labelled and non-labelled chicks (Fig. 3) were compared using the 95% confidence intervals of their linear slopes.

STATISTICA and STATVIEW software were used to perform the analysis.

In the labelled chicks, body mass increased from 680±160 g (mean± s.d.) at the first injection to 2370±165 g at sacrifice, a 3.6±0.6-fold increase (paired t-test, d.f.=2, t=–29.721, P=0.001, final body mass missing for the predated chick). Bill, flipper and foot lengths increased from 39±2 mm,86±4 mm and 75±6 mm, respectively, to 52±3 mm (d.f.=3, t=–8.385, P=0.004, no missing data), 151±20 mm(d.f.=3, t=–6.901, P=0.006) and 123±10 mm(d.f.=3, t=–12.325, P=0.001). Ponderal and statural growth curves of the four labelled chicks were very close to those of 30 other chicks that were not labelled (Fig. 3; only bill length daily increase differed between labelled and non-labelled chicks). Moreover, labelled chicks were emancipated at the same age (32.3±2.1 days, N=3; emancipation date missing for the predated chick) as non-labelled chicks (32.2±3.2 days, N=30)(d.f.=31, t=0.089, P=0.930). These results indicated that the labelling procedure we used had no strong adverse effect on the chicks'development.

Each bone section comprised two concentric regions: a peripheral crown, or`cortex', encircling a central area of lesser compactness, the `medulla'(Fig. 1).

Five-week-old chicks had bone cortices made of fine cancellous bone(Fig. 1). This structure results from periosteal growth of `fibro-lamellar bone', the rapidly growing type of bone tissue (see de Ricqlès et al., 1991). Fluorescent labels(Fig. 2) demonstrated rapid periosteal bone growth, i.e. apposition of a highly lacunar woven network at the bone periphery, completed by a progressive filling by primary osteons in the depth of the cortex. Such fine porosity of the whole cortex is due to fast centrifugal progression of the periosteum, exceeding the primary osteon filling process, which thus remains temporarily incomplete. Primary osteon orientation was quite variable, even within a single cross section. Accordingly, four fibro-lamellar bone tissue types could be distinguished:

1. `longitudinal bone' (Fig. 2A): longitudinally oriented primary osteons (i.e.“ tissu fibro-lamellaire à ostéones primaires longitudinaux dispersés”, according to de Ricqlès, 1975; p. 89);

2. `radial bone' (Fig. 2B):radially and longitudinally oriented primary osteons (i.e. “tissu rayonnant”, according to de Ricqlès, 1975; p. 91);

3. `reticular bone' (Fig. 2C):obliquely and irregularly oriented primary osteons (i.e. “tissu réticulaire”, according to de Ricqlès, 1975; p. 92);

4. `laminar bone' (Fig. 2D):circularly and longitudinally oriented primary osteons (i.e. “tissu laminaire sensu stricto”, according to de Ricqlès, 1975; p. 90).

All four tissue types were common, exhibited by every bone of our study,with various frequencies (Fig. 4).

Finally, the chicks' cortices were thick, occupying the outer third of the shaft's radius, on average. This high cortical thickness is the result of limited perimedullar resorption (osteoclasts progressing centrifugally at the inner surface of the cortex), which proceeds slowly and/or starts late compared with periosteal apposition.

A rather loose medullary spongiosa occupied the central area of the sections (Fig. 1). It was made of remnants of cortical primary bone, spared by perimedullar resorption. These trabeculae had been remodelled and consolidated by some endosteal bone apposition, as shown by fluorescent labels(Fig. 5).

The 140 growth rate values ranged from 7 μm day^–1 to 171 μm day^–1 (mean 55 μm day^–1). There was no significant heterogeneity of growth rate among chicks or weeks of growth (P=0.475 and 0.093, respectively; Table 2). However, the `long bone' factor had a highly significant effect on growth rate(P<10^–5; Table 2), mainly because of growth rates 38% lower in the radius than in other bones (P<0.002; multiple comparisons; Table 2; Fig. 4).

The `bone tissue type' factor had the strongest effect on growth rate(P<10<sup>–6</sup>; Table 2). Despite the four types having overlapping ranges(Fig. 4), laminar bone had significantly lower growth rates (P<10<sup>–3</sup>;multiple comparisons; Table 2),while radial bone had significantly higher growth rates (P<0.002;multiple comparisons; Table 2). Radial bone was the only tissue apposited at a rate above 109 μm day<sup>–1</sup>. Longitudinal and reticular bone had similar intermediate growth rates. It is noteworthy that the effect of `bone tissue type' was measured while controlling for the `long bone' effect (and vice versa; principle of a two-way ANOVA). Moreover, a significant interaction between `long bone' and `bone tissue type' was detected(P<10^–3; Table 2), which means that the two effects were not simply linearly additive. Graphical representation (Fig. 4) illustrates the results of the ANOVA; while the radius indeed has lower growth rates than other long bones, a shared pattern of increasing growth rate from laminar to radial bone is perceptible in all four long bones.

Fibro-lamellar bone is typical of early growth in endothermic vertebrates(review in de Ricqlès et al.,1991). Previous studies using fluorescent labelling have measured its growth rate (e.g. de Buffrénil and Pascal, 1984; Castanet et al., 1996, 2000; de Margerie et al., 2002; Starck and Chinsamy, 2002) and all found high rates of periosteal bone growth (>5–10 μm day^–1) compared with other bone tissues (i.e. lamellar-zonal tissues).

The range we report in king penguin chicks (7–171 μm day^–1) is congruent with those previous findings. Moreover,it extends the known range of periosteal growth rates up to previously unreported values; the highest rate was previously 112 μm day^–1, observed locally in the Mallard chick(de Margerie et al., 2002). High-speed osteogenesis in early growth of the king penguin might be required by peculiar life history traits in this species (i.e. large body size and only 3 months available to reach that size before winter starvation). Even the Emperor Penguin (Aptenodytes forsteri), which is larger and hatches further south on sea ice during austral winter, has a more evenly spread growth, and thus may not grow as fast as the king penguin(Stonehouse, 1960; p. 60).

According to our results, a relationship between growth rate and bone tissue type does exist in king penguin chicks, even after controlling for other factors (e.g. long bone); a special type of fibro-lamellar bone tissue(i.e. radial bone) grows faster, while another type (i.e. laminar bone) has comparatively moderate rates of growth. Nevertheless, aside from this central result supporting Amprino's prediction, some other findings indicate that Amprino's theory must be applied carefully.

First, we found a significant effect of `long bone' on growth rate, even after controlling for the effect of `bone tissue type'. This means that a given bone tissue type can have different absolute growth rates in different parts of the skeleton. Starck and Chinsamy(2002) also reported a `long bone' effect on growth rate of longitudinal bone tissue in the Japanese quail. Moreover, the significant interaction we observed between factors attests once more that the relationship between bone tissue type and growth rate can vary somewhat across parts of the skeleton.

Second, growth rate within each tissue type is highly variable, and ranges extend widely. This point had been raised in other species(Castanet et al., 2000; de Margerie et al., 2002; Starck and Chinsamy, 2002). For instance, Starck and Chinsamy(2002) observed a range of 10–50 μm day^–1 for longitudinal bone in the Japanese quail. Consequently, when several tissue types are observed (as in the present study), there is a consequential overlapping of growth rate ranges between tissue types (Fig. 4), and some differences fall below the significance threshold (e.g. comparison between longitudinal and reticular bone; Table 2). Variations in porosity might explain such variation of growth rates within tissue types (de Margerie et al., 2002; Starck and Chinsamy, 2002).

Far from nullifying Amprino's theory, which becomes more widely documented after the present work, the preceding points still have restrictive outcomes. A single and precise growth rate value can hardly be extracted from bone type. Moreover, extrapolation of extant growth rates to fossils on the basis of a shared bone tissue type (e.g. Curry,1999; Horner et al.,2000; Padian et al.,2001) should be conducted very carefully, as already emphasised(de Margerie et al., 2002; Starck and Chinsamy,2002).

Radial bone has the fastest growth and is the only tissue type found at rates of 109–171 μm day^–1(Fig. 4). Radial bone has already been observed in the bones of some wild or domesticated tetrapods(Table 3), especially in immature specimens, but its growth rate had never been measured. Nevertheless,in the light of comparative paleohistological observations, de Ricqlès(1977; p. 139) already hypothesised that radial primary osteonal orientation could be the signature of very rapid bone growth. Moreover, in aviculture, radial bone is observed in the limb bones of birds artificially selected for rapid growth(Dämmrich and Rodenhoff,1970; Itakura and Yamagiwa,1970; Leblanc et al.,1986; Leterrier and Nys,1992; Wyers et al.,1993). Finally, radial bone has not been observed in growth series of mallard (Castanet et al.,1996; de Margerie et al.,2002), ostrich (Struthio camelus) or emu (Dromaius novaehollandiae; Castanet et al.,2000), where growth rates did not attain the highest values measured here in the king penguin. These results suggest that radial bone could be a microstructural adaptation permitting higher rates of diametric bone growth, within an already fast-growing tissue type (i.e. fibro-lamellar bone). Functionally, the way through which the radial microarchitecture yields faster growth remains unclear. This would require fine comparative ultrastructural analysis of bone production by the periosteum, which was not conducted here. Nevertheless, we do notice that radial struts of woven bone can be produced continuously by the periosteum, contrary to a discontinuous`saltatory' pattern in other bone tissue types (as illustrated in Fig. 6).

It is well-known that “mature function and growth are antagonistic... features”(Ricklefs et al., 1994). A“ trade-off between growth rate and functional maturity”is a general constraint in various tissues of vertebrate organisms (review in Ricklefs et al., 1998). For example, Carrier and Leon(1990) addressed a“ conflict between development and skeletal function”. Therefore, the question of the biomechanical outcome of very rapid bone growth in the king penguin chick arises. Although this point would deserve experimental investigations (i.e. mechanical testing), two facts already support the idea that king penguin chick bone tissue has indeed poor mechanical attributes: (1) high cortical thickness in chicks, as observed in our material, is known to partly compensate for low resistance of the bone tissue, as Carrier and Leon(1990) pointed out in the California gull; (2) during their first month (i.e. before emancipation), king penguin chicks have a parsimonious locomotor behaviour: a chick always remains in the immediate vicinity of its parents (curled up in the brood patch or just standing close to it). The needs for active locomotion are very low, as they do not have to escape from predators, for example. As in other altricial birds, growth itself, rather than body support and transmission of forces,becomes the primary function of the bird's skeleton(Starck, 1998). Moreover,research in aviculture has shown that birds artificially selected for high growth rates tend to have weaker long bones (e.g. Leterrier and Nys, 1992),sometimes resulting in pathological conditions(Dämmrich and Rodenhoff,1970; Itakura and Yamagiwa,1970).

It is likely that radial bone has the most detrimental effect on mechanical resistance, because radial cavities between bone struts interrupt the shear flow around bone (Fig. 7) and concentrate stresses at their corners. These `open section effects' are known to dramatically reduce stiffness and strength, as has been modelled for torsional loads (Elias et al.,2000). Conversely, it has been proposed that laminar bone, which incidentally grows more slowly, would have better mechanical properties(de Margerie, 2002).

Adult king penguins exhibit strong(Currey, 2002; p. 130), matured bone (completed primary osteons, peripheral layer of lamellar-zonal bone,intracortical Haversian bone, perimedullar layer of endosteal bone; Meister, 1962). The timing of the bone maturation process in juveniles with regard to winter fast and to the onset of full locomotor activity (i.e. departure to sea after the first year on land) remains to be studied.

The authors would like to thank N. Lambert and J.-C. Fournier for their expert advice in the tagging and manipulating of young chicks in the breeding colony and M.-M. Loth for technical help during the histological preparations. We are grateful to V. de Buffrénil, M. Laurin and A. de Ricqlès for comments on earlier versions of the manuscript and to A. Chagnon and A. Camproux for advice on statistical methods. Financial support was provided by the Institut Français pour la Recherche et la Technologie Polaires(present name Institut Polaire Paul-Emile Victor; Programme 119) and logistical support by the Terres Australes et Antarctiques Françaises.