Introduction

Zebrafish (infraclass Teleostei) provide an excellent model for bone research. The elasmoid scale of teleosts is an easily accessible dermal bone plate that functions as reservoir when calcium demand is high [1, 2]. When a scale is lost or removed, the fish will quickly replace it thereby offering a unique model to study bone regeneration [3].

Scales can easily be harvested with little stress to the fish; once removed, formation of new scales is initiated immediately and, depending on ambient temperature, a new mineralized scale can be observed within a week [3, 4]. In this process of osteogenesis, i.e., osteoblast differentiation, matrix deposition, and mineralization can be studied on demand [5]. The great similarity between the original (ontogenetic) and the regenerated scale demonstrates that (ontogenetic) scale development and regeneration follows a conserved program built on similar cellular and molecular controls [6].

Mesenchymal scale-forming osteoblasts (often called scleroblasts; osteoblasts of the dermal skeleton) form two disk-shaped cell layers in between which bone matrix is secreted. Newly deposited mineral is already observed in 4-day-old regenerating scales [7, 8]. As is the case with intramembranous ossification in mammals, scales are formed without a cartilage precursor. Calcium phosphate (hydroxyapatite) is deposited in the collagenous (type Iα) matrix [8]. The acellular matrix is deposited in a highly ordered plywood-like arrangement, and contains among others osteocalcin (bone gla protein) and osteopontin (secreted phosphoprotein 1) [9–12]. The basal plate (the innermost half of the osteoid that gives flexibility to the scale) is not mineralized [8]. Scale growth involves continuous remodeling by osteoblasts and osteoclasts [3, 13]; scales keep growing peripherally by appositional bone formation as long as the fish grows [7].

Rodent models for osteoporosis studies are labor-intensive and expensive; mammalian bone formation starts in utero, which is an experimental challenge and both induction of osteoporosis and fracture repair are time-consuming studies [14]. Very few studies on mammals have addressed aspects such as gene expression and biomineralization in vivo. Clearly, the fish scale as model is exempt from these limitations; scales are easily accessible, available in large numbers (a fish typically has hundreds of scales) and osteogenesis can be induced anew by scale regeneration.

Another advantage of the use of scales over mammalian bone is their in vitro application. In a scale (ex vivo) associated osteoblasts and osteoclasts remain attached to their natural matrix and important cell–cell and cell–matrix interactions thus remain intact during culture experiments. Further, the scale matrix is transparent, which means that the bone cells are easily visualized by microscopy. In terms of complexity, scales better resemble bone compared to cultured osteoblast or osteoclast cell lines. Comparable mammalian in vitro models with “scale” features are explant cultures (e.g., calvariae), which require one animal per sample obtained, whereas a single zebrafish can provide 50 highly similar scales drastically reducing the demand on experimental animals [5, 15]. Fish scales were successfully used to assess effects of (anti-)osteogenic compounds and hormones on osteoblast and osteoclast activities in vitro [16]. Scale cells respond very similar to mammalian bone cells and this led to the (suggested) use of scales in drug discovery [5, 17–19]. The zebrafish scale, as model, can greatly benefit from an extensive genetic toolbox, e.g., transgenic reporter lines and an extensive selection of mutant lines. Yet, to the best of our knowledge, no study so far addressed whether scales could serve as disease model in for instance osteoporosis.

The glucocorticoid system is highly conserved from fish to mammals; glucocorticoids control energy flows in an organism to cope with stressful condition and are at the basis of proper acclimation to a dynamically changing environment [20]. Long-term treatment of patients with glucocorticoids (to suppress inflammation) is known to induce osteoporosis [21, 22]. Barrett and coworkers reported that glucocorticoids delay mineralization of the internal skeleton in larval zebrafish, reminiscent of an osteoporotic condition [23]. Unfortunately, osteoclasts are absent in early larval stages and the dynamics of osteoblasts–osteoclasts interactions alike [23, 24].

Glucocorticoids signal via the glucocorticoid receptor (GR) found in all tissues including bone [25, 26]. GR signaling involves genomic and nongenomic pathways and the pathophysiology of glucocorticoid-induced osteoporosis (GIOP) is therefore highly complex [27]. In general, glucocorticoids inhibit osteoblast function via decreased differentiation and increased apoptosis, and stimulate osteoclast formation and survival [21, 27]. Two phases are discriminated in human GIOP: a rapid, early phase is characterized by decreased bone mineral density (BMD) as a result of increased osteoclast activity; a slower, progressive phase is ascribed to an additional inhibition of osteoblast activity [22].

In the present study, we investigate the effects of prednisolone on scale regeneration in zebrafish. We hypothesized that prednisolone induces an osteoporosis-like state in zebrafish scales; i.e., increased scale resorption and decreased scale formation.

Materials and methods

Experimental design and animal husbandry

Wild-type zebrafish (Danio rerio) were kept in groups of eight fishes in 2-L tanks at 26 °C. Two separate recirculation systems, with a total volume of 32 L each were used to separately house control and prednisolone-treated fish. Each system contained eight tanks with predominantly males of approximately 0.4 g. Fish were fed once a day with flake food (Tetra, Melle, Germany). Scale sampling was done under mild anesthesia in 0.05 % (v/v) 2-phenoxyethanol (Sigma-Aldrich). Fish were euthanized for final sampling in 0.1 % (v/v) 2-phenoxyethanol. Experimental procedures were approved by the ethical committee of the Radboud University (RU-DEC 2008–099 and RU-DEC 2011–012).

Prednisolone exposure was started one day prior to scale removal (day −1) via the addition of 25 μM prednisolone-21-phosphate (Fagron, Capelle a/d IJssel, The Netherlands, CAS nr: 125-02-0) to the water. One third of the water volume (with or without prednisolone) was refreshed daily. On day 0, scale regeneration was induced by removal of 50 scales from the left flank of the fish. Ontogenetic scales were collected from two control and two prednisolone-exposed groups for analysis. Subsequently, fish were sacrificed after 4, 5, 6, 8, 11, 14, and 21 days of regeneration. Both ontogenetic and regenerating scales were sampled for the analysis of mineral content and gene expression. On days 8 and  21, intact ontogenetic and regenerating scales were sampled for histochemical analyses.

Our experiments showed that the mineral content of ontogenetic scales, sampled from fish with scales regenerating, did not differ from that of ontogenetic scales sampled at the start of the experiment (data not shown). This indicates that calcium balance of the fish was not at risk through removal of the 50 scales. Apparently, the branchial and dietary calcium uptake compensated calcium lost by scale removal, which was also observed in goldfish [28]. Thus, importantly, the experimental design did not influence calcium homeostasis.

Histochemistry

For whole mount studies, scales were fixed in 4 % paraformaldehyde in phosphate buffered saline for 24 h. Subsequently, samples were stained for activity of tartrate-resistant acid phosphatase (TRACP) and for presence of calcium phosphates according to Von Kossa, as described in detail by de Vrieze [13]. Morphometric analyses were done with Adobe Photoshop CS5.5 ®.

Mineral analyses

Ten scales were freeze dried, subsequently dissolved in 200 μl nitric acid (65 %); this solution was diluted 100× with MilliQ water. Total calcium, phosphorus, and magnesium contents were measured with inductively coupled plasma mass spectrometry (ICP-MS; Thermo Fisher Scientific) after which the Ca/P molar ratio was calculated.

Gene expression

The remaining scales were pooled per fish and used to determine gene expression profiles. Total RNA was isolated with Trizol (Invitrogen) according to manufacturer's instruction. RNA preparations were treated with DNase I (Invitrogen). An output of 500 ng was used for cDNA synthesis with Superscript Reverse Transcriptase II (Invitrogen) according to manufacturer's instructions. The obtained cDNA was diluted 20× in ultrapure water and used for quantitative PCR.

Quantitative PCR mix (25 μl) contained 1× Power SYBR® Green Mix (Applied Biosystems), 5 μl diluted cDNA and 5 μM of each primer, except for mmp-2 (2.5 μM forward and reverse primer). The primers used in this experiment are listed in Electronic supplementary material Table 1. Before data analysis, dissociation curves were checked for specificity and primer efficiency was confirmed to be 2 ± 0.05. The Genorm algorithm was used to select the three most stable housekeeping genes, viz. 40s, β-actin, and elf1a and calculate the mean normalized expression of target genes [29].

Statistical analysis

Normal distribution of data was assessed by the D'Agostino and Pearson omnibus normality test. Normally distributed data were tested for significant differences using the Student's t test with Welch's correction when variances were unequal. Not-normally distributed data were tested for significance of differences between means with the nonparametric Mann–Whitney test. Results from analysis of scale morphometry and mineral content were tested for interaction between prednisolone treatment and scale regeneration by means of two-way ANOVA. Columns were compared to each other using the Bonferroni post test. Statistical analysis of gene expression data was conducted on the mean normalized expression data also when set relative to control conditions.

Results

Morphometric analyses were conducted to measure size, shape, and (un)mineralized areas on Von Kossa stained scales (Fig. 1). TRACP, secreted by osteoclasts to degrade the scale matrix, was stained enzymatically in cells and matrix. Areas positive for TRACP activity were quantified and taken as a measure for osteoclast activity (Fig. 1). Regenerated scales were significantly smaller after 8 days; yet at the end of the regeneration experiment (21 days), there was no difference in size and circularity between ontogenetic and regenerated scales of the control group (Fig. 1a, b). For the control groups, the percentage of unmineralized surface of the regenerated scale was significantly larger compared to ontogenetic scales as was the percentage of TRACP activity staining (Fig. 1c, d). Regenerating scales of prednisolone-exposed fish were significantly smaller after 21 days compared to both controls and their ontogenetic counterparts. Moreover, we observed that regenerating scales of prednisolone-treated fish were more irregular in shape (reflected by a significantly lower circularity factor); either the scales were more ellipsoid or their perimeter was irregular as a result of gaps along the scale edges and of fusion processes (examples shown in Fig. 2d–g). Clear interaction effects between prednisolone treatment and scale regeneration were found for the circularity and the percentage of TRACP activity stained areas. TRACP-positive surface areas were significantly larger in scales of prednisolone-treated fish, as shown in Fig. 2.

Fig. 1
figure 1

Morphometric analyses of scales from control and prednisolone-exposed fish. Data on ontogenetic and 8- and 21-days regenerating (regen (8) and regen (21)), respectively, scales are shown for control (black dots) and prednisolone-treated (open dots) fish. a Surface area of scales in square millimeter. b Circularity factor (4πA/P2; where A is the surface of the scale and P is the perimeter of the scale). c Percentage of the scale surface that is demineralized (anticipated to be the result of osteoclast activity). d Percentage of the scale surface stained enzymatically for TRACP activity, either in cells or secreted into the scale matrix. Each replicate is displayed; the horizontal lines represent means of the replicates. Interaction effect was assessed with a two-way ANOVA. Columns were compared to each other using the Bonferroni post test. Different letters indicate statistical differences between (logical) groups

Full size image
Fig. 2
figure 2

Von Kossa (a–g) and TRACP activity (h–m) staining of zebrafish scales indicative of presence of calcium phosphates and osteoclast activity, respectively. a Ontogenetic scale with demineralized resorption pits in the focus (arrowheads). Bar 500 μm, scale circularity factor = 0.76. b Magnification of the focus of the scale depicted in a, showing demineralized areas interpreted as resulted from resorptive activity. Bar 100 μm. c Twenty-one days regenerating scales of a control fish. Resorption pits are depicted with arrowheads, scale circularity factor = 0.74. d–g Examples of misshaped regenerating scales (21 days old) resulting from prednisolone treatment. d Invagination (arrowheads) in the posterior part of the scale. Scale circularity factor = 0.48. e Extreme number and size of demineralized areas (arrowheads). Scale circularity factor = 0.75. f Fusion of two scales, boundaries are indicated with arrowheads. Scale circularity factor = 0.58. g Incompletely formed scale. Scale circularity factor = 0.57. h Twenty-one days regenerating scale from a control fish showing mild TRACP activity (arrowheads). i Eight days regenerating scale of a prednisolone-treated fish showing extreme TRACP activity. j Twenty-one days regenerating scale where TRACP activity appears slightly decreased compared to 8 days regenerating scales. k Malformation sites, such as invagination (arrowhead) coincide with excessive TRACP activity staining. l Magnification of 21-day regenerating scale showing TRACP activity following the edges of resorbed areas (arrowheads). Bar 50 μm. m Multinucleated osteoclasts (arrowheads) at the edges of resorbed areas of an 8-day regenerating scale. Bar 10 μm

Full size image

Representative scales of each group, stained for presence calcium phosphate minerals or TRACP activity, are shown in Fig. 2. Resorption pits were found on all scales, but they covered a higher percentage of the surface area in regenerated scales (Fig. 2a–g). Demineralized resorption pits (Fig. 2b) were more often found on regenerated scales and showed characteristic round boundaries as a result of osteoclast activity. Prednisolone treatment resulted in more irregularly shaped scales and even fusion of two scales was sometimes observed (Fig. 2d–g; see Fig. 2c for regularly shaped regenerated scale). TRACP activity staining revealed sporadic osteoclast activity on ontogenetic scales (not shown), but moderate TRACP activity was detected on regenerated scales (Fig. 2h). Prednisolone treatment significantly increased TRACP activity staining after 8 days of regeneration (Fig. 2i), and after 21 days TRACP activity staining was still intense (Fig. 2j). Areas of scale malformation coincide with increased TRACP activity staining (Fig. 2d, k). Resorption pits are clearly lined with TRACP activity staining, resulting from the presence of osteoclasts (Fig. 2l, m).

ICP-MS of regenerating scales revealed that calcium, phosphorus, and magnesium were incorporated in the newly formed scales and their content increased throughout the regeneration experiment (data not shown). Due to their small size, measurements required pooling of scales. Therefore, best insights in the dynamics of mineralization could be obtained by calculation of the calcium/phosphorus molar ratio (Fig. 3a, b), where each scale serves as its internal control for differences in size. In the regenerating scales (Fig. 3a), where a new mineralized matrix is formed, the Ca/P ratio increased from 1.2 at day 4, to 1.4 on day 6, and later (P = 0.0002). Prednisolone treatment resulted in significantly lower Ca/P ratios in both ontogenetic and regenerated scales throughout the experiment, with the exception of regenerating scales of 8 days old. For ontogenetic scales, the ratio was approximately 1.4 and was not influenced by new formation of scales on the opposing flank (Fig. 3b).

Fig. 3
figure 3

Mineral analysis of a regenerating scales and b ontogenetic scales. The ontogenetic scales are sampled from the same fish on the same day as regenerating scales. Scales from control fish are shown with black bars and from prednisolone fish treated are shown with white bars. All samples are expressed as means ± SD of five to eight fishes. Statistical analysis was done with two-way ANOVA. Significant effects of time and treatment were found in both cases (P < 0.001), while an interaction effect between time and treatment was found for ontogenetic scales (P < 0.05). Columns were compared to each other using the Bonferroni post-test (*P < 0.05; **P < 0.01; and ***P < 0.001)

Full size image

To assess whether the prednisolone treatment would interfere at the very start of scale regeneration, treatment was started 1 day before scale regeneration was induced. The effect of this 24-h prednisolone treatment per se on zebrafish scales was monitored through gene profiling (Fig. 4). Most genes were not influenced in their expression by the 24-h prednisolone treatment. Five bone-associated genes did respond to this short-term glucocorticoid treatment: runx2a, osteocalcin, tartrate resistant acid phosphatase (tracp; acp5 gene)m and ectonucleotide pyrophosphatase/phosphodiesterase 1 (enpp1 gene) were upregulated. On the other hand, the amount of glucocorticoid receptor-α transcripts (the glucocorticoid signaling isoform in zebrafish) [30] had decreased.

Fig. 4
figure 4

Comparison of gene expression in ontogenetic scales of control (gray bars) and 1-day prednisolone-treated fish (open bars). a Osteoblast marker genes, b osteoclast marker genes, c bone-related genes that in zebrafish scales cannot (yet) be ascribed to a specific cell population. Bars represent means ± SD of eight fishes. Significant differences are depicted with asterisks (*P < 0.05, **P < 0.01)

Full size image

We then analyzed expression of the same genes in regenerating scales. Genes known to affect bone and matrix formation are shown in Fig. 5. Osteoclast markers and genes encoding inhibitory proteins in bone formation are shown in Fig. 6. Osteoblast marker genes, involved in osteoblast differentiation (sp7 and runx2a) and production of the extracellular matrix (collagen 1a, entpd5a, and osteopontin) significantly increased during the formation of a new scale (Fig. 5a–e). Expression peaked around days 6 and 8, after which expression gradually declined, yet remained elevated compared to that in ontogenetic scales. Remarkably, no increase in gene expression was observed for the traditional osteoblast marker alkaline phosphatase (Fig. 5f). Key osteoclast marker genes cathepsin k and tartrate-resistant acid phosphatase did not increase expression during scale regeneration (Fig. 6a, b). In fact, tartrate-resistant acid phosphatase expression was suppressed. Only matrix metalloproteinases-2 and -9 (mmp2 and mmp9, respectively) significantly increased in expression (Fig. 6c, d). Some genes did not increase during early regeneration. Expression of sclerostin (sost) and enpp1 increased from day 6 (Fig. 6g, h) whereas rankl (Fig. 6e), a potent osteoclast-inducing signal and the proton channel vatp-ase (Fig. 6f), were only elevated in the last week of scale regeneration.

Fig 5
figure 5

Expression profiles of genes involved in bone formation during scale regeneration in control (gray) and prednisolone-treated fish (white). a sp7 (osterix), b runx2a (MASNS isoform), c collagen 1α, d entpd5a, e osteopontin (secreted phosphoprotein 1), f alkaline phosphatase, g osteocalcin (bgp bone gla protein) and glucocorticoid receptor α. Housekeeping gene-normalized results are expressed relative to the gene expression level found in ontogenetic scales (onto) of control fish. Data are expressed as 5–95 percentile box-plots; the horizontal line depicts the median of five to eight replicates, means are indicated by the plus sign in the box. Different letters indicate significant differences induced by scale regeneration in the untreated groups (compared to ontogenetic scales). Asterisks indicate significant differences as a result of prednisolone treatment, compared to untreated scales of the same age (*P < 0.05, **P < 0.01, ***P < 0.001)

Full size image
Fig. 6
figure 6

Expression profiles of osteoclast marker and bone-inhibitory genes during scale regeneration in control (gray) and prednisolone-treated fish (white). a ctsk (cathepsin k), b tartrate resistant acid phosphatase (TRACP), c mmp2, d mmp9, e rankl, f v-atpase, g sclerostin, and h enpp1. Housekeeping gene-normalized results are expressed relative to the gene expression level found in ontogenetic scales (onto) of control fish. Data are expressed as 5–95 percentile box plots; the horizontal line depicts the median of five to eight replicates, means are indicated by the plus sign in the box. Different letters indicate significant differences induced by scale regeneration in the untreated groups (compared to ontogenetic scales). Asterisks indicate significant differences as a result of prednisolone treatment, compared to untreated scales of the same age (*P < 0.05, **P < 0.01, ***P < 0.001)

Full size image

The effect of prednisolone treatment on gene expression during scale regeneration is also shown in Figs. 5 and 6 (as white bars). Osteoblast genes are generally activated during scale regeneration, but in some occasions further increased due to prednisolone treatment. Alkaline phosphatase, which did not increase during normal regeneration, had significantly increased as a result of the prednisolone treatment at days 6 and 11 (Fig. 5f). Only for collagen 1a, a significant increase was found for more than two time points (Fig. 5c). Osteoclast marker genes such as mmp-9 and cathepsin k had increased several times over control values, whereas vatp-ase was the only gene that got downregulated (Fig. 6a, d, f). Prednisolone also significantly increased expression of enpp1 (Fig. 6h), a gene involved in pyrophosphate generation, but which cannot yet be ascribed to a specific cell population in zebrafish scales [31].

Discussion

Three major findings from which we conclude that prednisolone induces an osteoporosis-like phenotype in regenerating zebrafish scales are provided here. (1) Histomorphometric analysis indicates that osteoclast activity and matrix resorption is higher in scales of prednisolone-treated fish. (2) Mineralization is impaired in these scales. (3) Prednisolone alters the typical phased gene expression profile of scale regeneration, especially that of the osteoclasts.

We used adult male fish for this experiment to ensure that differences in sex steroid profiles did not interfere with this study. Estrogens are known to target the RANK/RANKL/osteoprotegerin (OPG) pathway [32] and 17β-estradiol is known to stimulate recruitment of calcium from scales in rainbow trout [33, 34]. Indeed, vitellogenesis is induced in sexually mature females by elevated levels of 17β-estradiol. These females have extreme calcium demands to saturate hepatic vitellogenin (an abundant calcium- and phosphate-binding protein in the egg yolk). Sex determination upon dissection of our fish revealed that most groups contained a maximum two females (out of eight fishes). Gene expression profiles and mineral content were assessed with and without females included, but this did not lead to significant changes.

Morphometric analyses revealed that, in general, regeneration of scales under prednisolone treatment induced malformations. Regenerating scales were smaller compared to their ontogenetic counterparts, which was not the case in controls. Moreover, prednisolone significantly affected the circularity factor. Von Kossa mineral staining revealed increased osteoclast activity viz. increased demineralized scale matrix in fully regenerated scales (day 21). Prednisolone treatment increased these parameters of osteoclast activity compared to controls. This notion was further substantiated by a stronger increase in TRACP activity staining in scales from prednisolone-treated fish, which resembles the increased osteoclast surface found in femurs of prednisolone-treated mice [35]. An involvement of multinuclear osteoclasts in scale demineralization has been suggested before [3, 13], and multinucleated cells were indeed found on the scale plate. Our findings of the occurrence of round edges in resorption pits, lined with TRACP activity staining of the matrix, substantiates that the resorption pits are the obvious result of cell-mediated matrix degradation. In fish, bone resorption involves the activity of (“typical”) multinucleated osteoclasts and small mononucleated cells; both are present on zebrafish scales [24, 36]. The latter activity of mononucleated cells could be considered comparable to osteocytic bone degradation in mammals [37].

It is well established from X-ray radiography that glucocorticoids negatively influence BMD in rodents and humans alike [22, 38]. Scales are too small to rely on the commonly used techniques to assess BMD. The next best options are histomorphometry of demineralized tissue and measurement of mineral content by ICP-MS. The calculated calcium/phosphorus ratios in this study never reached 1.67 (indicative of pure hydroxyapatite), as the attached scale cells cannot be excluded from the measurement [8] and cause a shift in this ratio. The presence of hydroxyapatite as the major mineral phase in fish scales has previously been verified by means of X-ray diffraction spectroscopy [8, 39]. In the present study, we observed calcium/phosphorus ratios of approximately 1.0 in the very early regenerating scales, whereas ontogenetic and older regenerated scales had a ratio of approximately 1.4. Various calcium phosphate precursor phases (e.g., brushite; calcium phosphate ratio, 1) have been proposed as the primary calcium phosphate that transits to an apatite such as hydroxyapatite or whitlockite [40]. Even for mammals, there is little in vivo information on this, but in zebrafish fin rays (also dermal skeleton) amorphous calcium phosphate was discovered as a precursor phase [41]. In this area of research, scales may provide a valuable model as mineralization can be induced on demand (regeneration) and further benefits from easy access and temperature dependence.

Prednisolone slowed down the development of the Ca/P ratio in scales compared to controls. Moreover, from day 5 onwards, Ca/P ratios were significantly lower compared to controls for the duration of the experiment. Similar observations were made for the ontogenetic scales, sampled during scale regeneration on the opposing flank of the fish. Lowering of Ca/P ratios of biological apatites generally points towards altered crystalline phase or an impurity, both of which make them less stable and more easy to dissolve by osteoclasts [40]. We take our findings as additional evidence that (scale) mineralization is negatively affected by prednisolone.

Responses of scale cells to prednisolone treatment were observed in ontogenetic scales already after 1 day. Increased expression was found for five bone-related genes involved in osteoblast differentiation or bone resorption; clearly, the high level of conservation in glucocorticoid signaling between fish and mammals also holds for bone physiology [30]. Despite the short treatment time in terms of GIOP, this suggests that both osteoblasts and osteoclasts of zebrafish scales can elicit (quick) glucocorticoid responses.

Prednisolone effects were more pronounced in regenerating scales. Osteoclast genes were well upregulated as a result of the prednisolone treatment, which was predicted since increased osteoclast activity is a hallmark of GIOP [21, 42]. Strongest effects were found for genes encoding proteins that function in matrix and mineral degradation, such as cathepsin k, mmp-9. Highly similar upregulations of these genes were also found in bones of prednisolone-treated mice [38]. Note that in our study, mmp-9 already increased during control regeneration since osteoclasts are involved in remodeling of the scale plate [3]. Based on our combined findings, this at least appears to take place independently of enhanced cathepsin-k activity. In osteoporotic patients, increased serum markers of collagen degradation infer increased osteoclast activity [43], as this cannot so easily be assessed on a cellular level. Zebrafish are too small to collect sufficient amounts of serum, but from the observed increased osteoclastic activity it may be anticipated that collagen degradation is increased. The expression pattern of phosphate/pyrophosphate homeostasis controlling entpd5 and enpp1 genes further substantiated the notion that mineralization in prednisolone-treated fish is affected. Enpp1 expression significantly increased in regenerating scales under prednisolone tonus, while entpd5a, in fish the only known osteoblast gene not expressed in dentition [Huitema, personal communication, p. 31], was not affected. The changes in balance between these two genes suggest increased generation of calcification-inhibiting pyrophosphates, which supports the notion of a hampered mineralization. In control conditions, enpp1 expression increased from day 6 onwards, when the first unmineralized osteoid of the basal plate appears in zebrafish and other cyprinids (unpublished observation) [6].

As prednisolone attenuates/inhibits osteoblast activity in mammals [21, 27], we predicted that expression of osteoblast genes, such as collagen1a, osterix, runx2a, and osteocalcin, would be suppressed by prednisolone. An up to 20-fold strong increase in those genes was seen in scales regenerating under control conditions. Expression of osteocalcin, a marker of mature osteoblasts in mammals and zebrafish [44, 45] increased towards the end of regeneration (not statistically significant), indicating that scale osteoblasts have matured. Small effects of prednisolone were found on the expression of osteoblastic genes, especially in the first days of regeneration (when their expression is highest in the control groups). In fact, at some time points expression of collagen1a and runx2a increased rather than decreased. We take this as an indication that osteoblast activity is not directly inhibited after glucocorticoid treatment and attempts to keep up with the increased matrix resorption. Such a mechanism has been postulated before and can still result in a reduced bone mass (osteoporosis) as there is more bone degraded than formed [46]. Moreover, the 20-fold increase in osteoblast gene expression in our regeneration experiment may lower the resolution to detect inhibitory effects of prednisolone on osteoblasts. Although surprising, the negative effects of glucocorticoids on osteoblasts have mainly been described for long-term exposures. Indeed, increased osteoblast apoptosis is first measured in mammalian studies after 27 days of glucocorticoid treatment [47]. Extensive gene expression profiling of long bones of prednisolone-treated mice demonstrated that most osteoblast genes were not significantly until 28 days of treatment [38]. Expectedly, the observed decreases in osteoblast gene expression in mammals can be largely ascribed to a relative decrease in osteoblast number as a result of the increased apoptosis. Exposure in our experiment lasted 3 weeks, long enough for a scale to regenerate to its original size, but possibly too short to observe negative effects on osteoblasts. One of the few short-term effects described on osteoblasts is increased RANKL secretion, which is part of the pivotal RANKL/RANK/OPG pathway involved in osteoclast activation [48]. A different study showed that prednisolone decreases RANKL expression in cultured osteoblasts [49]. This was also observed in prednisolone-treated regenerating scales in terms of rankl gene expression, which coincides with the upregulation of other osteoclast markers. Unfortunately, no information could be obtained on the dominant negative soluble RANKL decoy receptor OPG, as genomic information on its zebrafish orthologue is inconclusive. Overall, the typical phased gene expression profile that is characteristic of scale regeneration is disturbed by prednisolone. The fine balance between bone cells has shifted towards increased and dominating osteoclast activity.

Conclusion

Our key findings of increased osteoclast activity and its reflection thereof on the mineralized layer of the scale follow the known mechanistics of (early) GIOP in mammals [21]. The regeneration paradigm (rapid new formation of a bone plate) may be compared to bone repair (new formation), an aspect of bone physiology of which very little information exists in relation to GIOP [50]. In itself, the scale regeneration process shows many similarities with bone formation/repair in higher vertebrates, both in chemical composition and in the expression of marker genes. Its response to glucocorticoid treatment is highly similar to that seen in bone of mammals, resulting in increased osteoclast activity and poor bone regeneration. In addition, we show that osteoblasts are not immediately affected by prednisolone treatment, and can still cope with the challenge of scale regeneration during early GIOP. Already, several papers addressed the conserved glucocorticoid system of zebrafish [25, 30] and now we show that glucocorticoid treatment results in a phenotype similar to (the early phase of) GIOP. This highlights zebrafish scale as a great model to study underlying mechanisms of GIOP in a bone tissue that is easily accessible. Their sensibility to glucocorticoids further confirms the potential application of scales in drug screening, for which we indeed currently have an assay in development.