On the regeneration of fish scales: structure and mechanical behavior

Dermal armors are a special class of structural tissues that serve a variety of functions and exhibit a wide range of mechanical properties. These tissues have been modified extensively in response to both biotic and abiotic selection pressures (Weiner and Addadi, 1997; Barthelat, 2007; Meyers et al., 2012), primarily through differences in mineral content and microstructure array (Currey, 1999; Meyers et al., 2008; Zhang et al., 2011). As a result, many dermal armors possess exceptional mechanical qualities, such as high strength and toughness to weight ratios and resistance to damage. Although armors such as highly mineralized shells primarily function to prevent or reduce predatory injury, other tissues such as fish scales, which are both tough and flexible, have also evolved to facilitate locomotion (Yang et al., 2013).

The scales of modern fish have evolved into four primary groups: cosmoid, placoid, ganoid and elasmoid (Kardong, 2006; Sire and Huysseune, 2003). Elasmoid scales have been of particular interest because they are often found on fish of high mobility requiring significant bending movement of their bodies. As such, the need for flexibility has imposed additional design requirements on scale architecture and composition (Webb, 1983; Szewciw et al., 2017). Like many biological materials, elasmoid scales possess a hierarchical microstructure that has been identified to play a key role in their mechanical properties (Khayer Dastjerdi and Bathelat, 2015; Zhu et al., 2012). Across its thickness, the elasmoid scale can be divided into two principal layers (Fig. 1) – the exterior limiting layer and the underlying elasmodine (Arola et al., 2018). The limiting layer is essentially a coating of calcium-deficient apatite reinforced with a sparse distribution of collagen fibers. The elasmodine consists of lamina (or plies) of unidirectional collagen fibrils arranged with a Bouligand helical stacking sequence (Bigi et al., 2001; Ikoma et al., 2003; Murcia et al., 2017; Torres et al., 2008; Zimmermann et al., 2013). The elasmodine can be further separated into external and the internal sub-layers differentiated by their relative mineral content. Collagen fibrils of the external elasmodine are more highly mineralized with nanocrystals of apatite; the degree of mineralization is most dense closest to the limiting layer, and then decreases to the internal elasmodine (Gil-Duran et al., 2016; Murcia et al., 2016). Correspondingly, there is a gradient in the hardness and elastic modulus through the scale thickness (Chen et al., 2012; Meyers et al., 2012; Arola et al., 2018).

Within the last decade, elasmoid fish scales have attracted substantial interest. Most efforts have evaluated the mechanical behavior of scales under uniaxial tension, transverse puncture and impact (Marino Cugno Garrano et al., 2012; Zhu et al., 2013; Yang et al., 2014; Torres et al., 2015). Recent investigations within our group have explored the effects of changes in intermolecular bonds stimulated by polar solvents on scale properties (Arola et al., 2019; Murcia et al., 2016). Collectively, these investigations are helping to develop an understanding of their unique microstructure and structure–property relationships. However, these studies have been largely comparative in nature and constrained to analysis of ontogenetic scales obtained from fish at death, with limited knowledge of the environment in which the scales developed.

When fish scales are lost or damaged, they undergo rapid regeneration through growth and mineralization to restore protective function, which places a high demand on calcium mobilization from internal and external sources (Yasuaki et al., 1989; Bereiter-Hahn and Zylberberg, 1993). The structure–property relationships of regenerated scales may be highly relevant to this process. To our knowledge, no study has been reported on the mechanics of regenerated scales, or how environmental conditions influence these relationships. Water temperature, for example, has a direct effect on both scale growth and mineral composition (Gauldie et al., 1990; Beakes et al., 2014; Thomas et al., 2019), and we hypothesized it might have a similar influence on the mechanical properties associated with these processes.

In this investigation, the microstructure and mechanical behavior of ontogenetic and regenerated scales from common carp (Cyprinus carpio Linnaeus 1758) are compared. The species has been distributed throughout the world by introduction into natural waters and from cultivation as a food fish (Kloskowski, 2011), and can tolerate temperatures over a range of approximately 2–30°C (Panek, 1987). The differences in scale properties are discussed as a function of the water temperature in which they regenerated, as well as the importance of these qualities to their performance for protective function.

Common carp (C. carpio) were captured in a flood plain channel adjacent to the Yakima River in southeastern Washington, USA, in July 2016. The fish (n=15) were caught with a beach seine net (25×3 m) and transported in an 800 liter insulated tank supplied with oxygen to the Pacific Northwest National Laboratory (PNNL), where they were transferred into a 2000 liter holding tank supplied with ambient temperature water (∼17°C at time of capture) from the Columbia River at a rate of 100 l min⁻¹. Then in March 2017, six fish were selected based on uniformity in body size and condition, placed into two additional 2000 liter tanks (n=3 fish per tank) and acclimated to the two experimental temperatures by slowly increasing the ambient water temperature from approximately 6°C to 10°C and 20°C over a period of 7 days. We used late March as the starting point for the experimental treatments because it represents the time when the ambient water temperature in the Columbia River begins to increase following the annual minimum in early to mid-February (Fig. 2A), and thus is consistent with the temperature cycle the fish would experience under natural conditions. The experimental temperatures of 10°C and 20°C were chosen because they are near the mid-point of the range of temperatures (∼2–30°C) that C. carpio occupy throughout their natural environment (Panek, 1987) and encounter seasonally in the Columbia River (Fig. 2A). Under artificial culture conditions with ad libitum diet, the temperatures of 10°C and 20°C provide maximum somatic growth rates of approximately 0.7% and 3% body mass per day, respectively (Goolish and Adelman, 1984). However, because scale growth is also highly dependent on temperature and feeding rate (Beakes et al., 2014), the fish were fed a commercial diet (Pond LE, Skretting, USA) at a rate approximating a maintenance ration (0.5% and 1.0% body mass per day at 10°C and 20°C, respectively) to minimize any confounding effect of somatic growth on scale growth. Prior to transfer, the fish were weighed, measured for length and tagged with a passive integrated transponder for individual identification. These six fish had a mean (±s.d.) length and mass of 59.7±5.2 cm and 4.2±1.1 kg, respectively.

In August and September 2017, the first set of ontogenetic scales (n=5 per fish) was extracted from three fish each held at 10°C and 20°C. Prior to extracting the scales, the fish were anesthetized by immersion in a solution of tricaine methanesulfonate (MS-222) at a concentration of 80 mg l⁻¹, after which the scales were removed with forceps. The fish were then returned to their respective experimental tanks and allowed to recover from the anesthesia. The scales were taken along two adjacent rows on the right side of each fish in the area immediately behind the gill plate and above the lateral line as shown in Fig. 2B. Each ontogenetic scale was visually examined to ensure it had not been regenerated from prior loss of the scale at that site. These could be identified upon removal by a diffuse focus and irregularly formed circuli (Bereiter-Hahn and Zylberberg, 1993). In August 2018, regenerated scales (n=5 per fish) were collected from the three carp held at 10°C and at 20°C. The regenerated scales were obtained from the same sites as the ontogenetic scales extracted in 2017. Regenerated scales were easy to discern externally because they lacked the pigmentation present in ontogenetic scales, and because each fish was photographed after scale removal to identify the specific collection sites. These can be seen as lighter colored sites in Fig. 2B. At the same time, an additional n=5 ontogenetic scales per fish were sampled from the area immediately behind the location of the regenerated scales (Fig. 2A). All of the extracted scales were placed in Hanks balanced salt solution (HBSS) and stored at 3–5°C until shipment to the University of Washington (less than 1 month). The capture, care and experimental procedures performed on the fish used in this study were performed under the PNNL Animal Care and Use Committee protocol 2017-02.

All the ontogenetic and regenerated scales possessed a diameter between 1 and 2 cm and were less than 1 mm thick. Because of their limited size, we utilized a specimen geometry that accommodated the tissue available from the scales. Conventional dog-bone shaped tensile specimens were sectioned from the scales with dedicated punch and stamping process to enable an evaluation of the tensile properties (Marino Cugno Garrano et al., 2012). A single specimen was stamped from the central area of each scale where the thickness was most uniform, with a gauge section length and width of 5.5 mm and 1.5 mm, respectively (Fig. 2B). The remaining portion of the scales were retained for further analysis. Owing to the potential for anisotropic mechanical behavior of scales from the head region (Murcia et al., 2015), all the specimens were obtained with alignment parallel to the fish length, for consistency. After sectioning, all specimens were placed in a bath of HBSS for a minimum of 24 h. Four specimens were obtained from four separate ontogenetic and regenerated scales in each fish to produce a total of 48 tensile tests overall (2 growth conditions×4 specimens per fish×3 fish per temperature×2 temperatures=48).

Tensile testing of the fish scale specimens was performed to failure at room temperature and with hydration. The evaluation was conducted using a commercial universal testing machine (Instron ElectroPuls E1000, Norwood, MA, USA) equipped with a load cell having a full-scale range of 250 N and load precision of 0.01%. The specimens were removed from the HBSS bath and mounted in the grips, and testing was conducted immediately in air. An eyedropper was used to apply HBSS over the duration of the tests to prevent dehydration. Loading was performed under displacement control at a quasi-static strain rate of 10⁻³ s⁻¹ up to failure. Though these scales are known to be highly strain rate dependent (Ghods et al., 2019), the quasi-static strain rate was chosen to be consistent with prior studies (Yang et al., 2014; Zhu et al., 2012). Failure of the specimens always occurred within the gauge section. The load and displacement data obtained from the tension tests were used with the measures of the specimen gauge section dimensions (average thickness and width measured across three points of the gauge section using a pair of digital calipers) to obtain the engineering stress–strain responses. These curves were used to estimate the elastic modulus (E), strength (S), strain at failure (e_f) and modulus of toughness (MOT). E was determined using the secant method for strains less than 1% and S was defined by the maximum stress realized by the sample up to failure. MOT was calculated by integrating the area under the stress–strain curves as a function of strain until failure.

An evaluation of the microstructure was performed on the scales of all the fish involved in this investigation to understand the differences between the ontogenetic and regenerated scales, as well as the influence of water temperature on scale growth and structure. For this purpose, the remnants of scales used for the tensile specimens were also utilized for imaging and evaluation of the chemical composition. The part of the scale most adjacent to the tensile specimen gauge section was separated from the remainder of the scale, treated using an ascending ethanol treatment and then mounted in an epoxy compound for polishing and analysis. After mounting, the samples were polished with SiC abrasive paper from mesh numbers of 800 to 4000. Final polishing was performed with a diamond liquid suspension of 1 μm, followed by a liquid suspension of 0.3 μm alumina. The polished samples were imaged using an optical microscope (Olympus BX50 Microscope, Olympus Scientific Solutions America, Waltham, MA, USA). The cross-section of the scales was evaluated in terms of the thicknesses of the principal layers, the number of plies in the elasmodine, and any general changes in the scale morphology that could be identified.

To further understand contributions from the microstructure to the mechanical behavior, an analysis of the chemical composition was performed on selected scales using Raman spectroscopy (Renishaw InVia, West Dundee, IL, USA). Specifically, scans were performed on three randomly selected ontogenetic and regenerated scales from each fish to obtain measures of the mineral to collagen ratio across the scale thickness, and as a function of distance from the scale's exterior surface. The scans were performed over the spectral range of 400 to 1800 cm⁻¹ with a Raman fluorescence microscope (50× objective; Leica, Buffalo Grove, IL, USA), resolution of 10×50 µm, laser wavelength of 785 nm and an acquisition time of 10 s. The spectrum was acquired starting from the scale exterior continuing to the interior in increments of either 6 or 12 µm for the regenerated and ontogenetic scales, respectively. The acquired Raman spectra were baseline corrected using commercial software (WiRE 3.4, Renishaw, West Dundee, IL, USA) to account for fluorescence.

The mineral to collagen ratio of regions of interest of the scales was calculated using the Raman spectra from the area ratio of the phosphate (961 cm⁻¹) and amide I (1690 cm⁻¹) peaks, following previously established methods (Arola et al., 2018). A representative Raman spectrum of a fish scale from this study with annotation of the mineral (i.e. phosphate) and collagen (i.e. amide I) peaks is shown in Fig. 2C. Measures of the thickness of the highly mineralized layer were obtained from the spatial variations in the spectra, as well as the chemical composition. Differences in the mineral to collagen ratio and composition of the regions of importance were compared as a function of distance from the limiting layer.

To test for significance, the mechanical properties, microstructural measurements and results from Raman spectroscopy were compared using a repeated-measures two-way ANOVA; P≤0.05 was used to identify significance. This method of statistical analysis accounts for the between fish variance while showing the significance in the properties of the scales between conditions. All data acquired in this investigation are available on request from the corresponding author.

Micrographs of representative ontogenetic and regenerated scales from fish held at 10°C and 20°C are shown in Fig. 1. The microstructure of the ontogenetic scales was consistent with that reported in earlier investigations of scales from common carp with well-defined limiting layer, including external and internal elasmodine regions (Murcia et al., 2018). By contrast, regenerated scales exhibited only two distinct layers, namely a relatively thick mineralized layer and a basal layer as evident in Fig. 1B,D. The mineral layer of the regenerated scales appeared as a single heavily mineralized region, with no clearly defined sub-layers or plies that would be considered the external elasmodine. The basal layer consisted of a relatively small number of plies and single distribution, with no clear transition between a low and more highly mineralized region. The mean±s.e.m. in ply thickness of the regenerated scale was 10±0.6 µm compared with values of 11±0.7 and 7±0.3 µm in the external and internal elasmodine, respectively, of the ontogenetic scales.

Measures of the scale morphology, including the layer thicknesses and the number of plies in each layer, are presented in Table 1. The regenerated scales were approximately half of the total thickness and significantly thinner than the ontogenetic scales from the same fish (P≤0.001). A comparison of the mineralized (limiting layer+external elasmodine versus mineralized layer) and non-mineralized (internal elasmodine versus basal layer) layers of the two groups showed that the mineralized layer exhibited the largest difference in the thickness between the ontogenetic and regenerated scales; the mineralized layers of the ontogenetic scales were approximately 3× thicker than in the regenerated scales (P≤0.001). By contrast, although the thickness of scales regenerated at 20°C was approximately 30% greater than those regenerated at 10°C, there was no significant difference between the two groups (P=0.052). Similarly, whereas regenerated scales had significantly fewer plies than the ontogenetic scales of the same fish, regardless of the aquatic environment temperature (P≤0.001), there was no significant difference in the number of plies between scales regenerated at 10°C and 20°C (P=0.40).

Representative stress–strain curves from the tension tests performed on scales from two selected fish are shown in Fig. 3. There was a high degree of consistency in the stress–strain responses among the multiple scales at both temperatures. The ontogenetic scales appeared to exhibit superior structural behavior with substantially higher tensile strength (S) and modulus of toughness (MOT). This quality was most evident from the scales of fish held at 10°C (Fig. 3A). By comparison, the elastic modulus (E), S and MOT of the regenerated scales were lower than those values for the ontogenetic scales, which again was most evident in the scales regenerated at 10°C (Fig. 3B).

The results for the measures of E, S, MOT and strain to failure (e_f) from the stress–strain responses for the scales of fish held at 10°C are shown in Fig. 4. The properties of the scales are presented in terms of the mean±s.d. for the three different fish that were held at this temperature (fish A, B and C). As evident from these comparisons, there was some variation in properties of the scales obtained from each fish. The coefficient of variation (CV) range for stress–strain responses of the ontogenetic and regenerated scales ranged from 0.09 to 0.21 and 0.13 to 0.32, respectively. The largest CV overall was in MOT, although this was minor when compared with the differences between properties of the ontogenetic and the regenerated scales, which were highly significant (P≤0.01), except for that in the measures of e_f being slightly less significant (P≤0.05). For all the metrics of performance, the regenerated scales exhibited inferior properties with respect to the ontogenetic scales, with the largest differences occurring in S (60%) and MOT (70%).

Comparisons of E, S, e_f and MOT of the ontogenetic and regenerated scales from the fish maintained at 20°C are shown in Fig. 5. Consistent with the data presented in Fig. 4, these properties are shown individually for scales of fish D, E, and F. The ranges in CV for E, S, e_f and MOT were 0.13–0.18 and 0.08–0.28, respectively, for the ontogenetic and regenerated scales, which were similar in variance to results for the fish held at 10°C. Although S, e_f and MOT of the regenerated scales were lower than in the ontogenetic scales, the extent of these differences were less than for scales regenerated at 10°C. Surprisingly, E of the regenerated scales for fish F was nearly 50% higher than that of the ontogenetic scales (Fig. 5A). Yet, there was no significant difference in the overall mean E of the ontogenetic and regenerated scales when all of the scales from the fish held at 20°C were included (P=0.27). However, the differences in all other mechanical properties between the ontogenetic and regenerated scales were significantly different (P≤0.05), with the largest difference represented in MOT (40%). Temperature-related effects on the mechanical properties of the regenerated scales were also present. E, S and MOT were all significantly lower in scales regenerated at 10°C than at 20°C (P<0.001), whereas e_f was not (P=0.145).

Raman spectroscopy revealed clear differences in the relative degree of mineralization among the distinct structural layers in the ontogenetic and regenerated scales. These differences can be seen from the spectra of the mineral to collagen ratio across the thickness of selected scales from fish held at 10°C (Fig. 6). For the ontogenetic scale (Fig. 6A), the mineral to collagen ratio was maximum within the limiting layer, and then underwent a marked reduction at the transition of the limiting layer/external elasmodine interface, and again at the external/internal elasmodine interface. In the regenerated scale (Fig. 6B), the highest mineral to collagen ratio was present in the outermost thick mineral layer of the scale, which was lower than the limiting layer from the ontogenetic scale but similar to that of the external elasmodine. There was a decrease in mineral to collagen ratio at the transition to the basal layer, and then a further decline through the thickness analogous to trends identified in the internal elasmodine for the ontogenetic scales.

Area ratios of the phosphate and amide I peaks used to quantify the extent of mineralization through the cross-sections of the scales as a function of normalized thickness are shown in Fig. 7. Results for representative ontogenetic and regenerated scales of fish maintained at 10°C are shown in Fig. 7A and B, respectively. Despite some differences in the absolute mineral to collagen ratios among the individual scales, all the ontogenetic scales exhibited a distinct limiting layer, external elasmodine and the internal elasmodine (Fig. 7A,C). The distributions and their transitions for ontogenetic scales of fish held at 10°C and 20°C fish were similar, with consistency in the normalized distances through the thickness of the scale where the transitions occur. By comparison, the mineral to collagen ratios of the regenerated scales (Fig. 7B,D) exhibited two characteristic regions that corresponded to the mineralized layer and basal layer. The maximum (i.e. peak) mineral content was lowest in the scales regenerated at 10°C (Fig. 7B).

The scales of bony fishes represent a specialized class of dermal armor with a hierarchical structure of collagen fibers interwoven into mineralized layers of calcium-deficient apatite (Murcia et al., 2016). This architecture and the overlapping arrangement of scales on the skin surface not only enhances physical protection, but also enhances mobility by increasing flexibility and reducing hydrodynamic drag (Yang et al., 2013). Although the composition and mechanical behavior of fish scales have been well characterized in multiple species, information about whether these properties change when scales are regenerated and the effect that environmental temperature has on this process has not been elucidated. Rapid scale regeneration is an important survival trait in fish because it restores protection against potential predatory injury and the integrity of dermis for osmoregulation (Zydlewski et al., 2010; Olsen et al., 2012). Because temperature is the major rate-controlling factor in this process, knowledge of how it affects the resulting form and function of regenerated scales can provide instructive information about the mechanisms that guide the assembly of biogenic structures and their utility for designing synthetic analogs.

In this investigation, we characterized the microstructure and mechanical behavior of ontogenetic and regenerated scales from common carp held at water temperatures of 10°C and 20°C. These temperatures approximate the mid-point and upper end of the range that carp encounter in the Columbia River basin (Fig. 2A), yet were sufficiently different to elicit changes in structure, composition and response to mechanical stress between ontogenetic and regenerated scales. Nevertheless, the mechanical properties of regenerated scales were generally inferior to those of ontogenetic scales, regardless of the rearing temperature. Strength and MOT, for example, were 60% and 70% lower, respectively, in regenerated compared with ontogenetic scales for the fish held at 10°C. For carp held at 20°C, the only property in regenerated scales that was not significantly lower than in ontogenetic scales was the elastic modulus. Moreover, regenerated scales did not have a well-defined limiting layer, nor a stratified elasmodine (consisting of external and internal elasmodine), but rather an outer mineralized layer and underlying basal layer. Although the mineral to collagen ratio was highest for the scales regenerated at 20°C, the mineralized layer occupied a larger portion of the total thickness for regenerated scales at 10°C owing to the lower overall thickness of the scales. Overall, S and MOT were linearly related to the mineralization ratio (Fig. 8).

If the primary purpose of scales is to provide protection from physical injury, microstructure development during regeneration should progress in a manner that prioritizes this function. Scale development starts in the dermis, where fibroblasts accumulate and begin to construct the mineralized layer, which then evolves into the elasmodine (Bereiter-Hahn and Zylberberg, 1993). The mineralized layer grows preferentially in diameter, eventually extending across the full surface area of the scale. In teleost fish, elasmoblasts lining the bottom of the mineralized layer begin to deposit a layer of entangled mineralized collagen (Sire and Akimenko, 2004). Then, comparatively thick unidirectional layers (or plies) of non-mineralized collagen begin to generate at the base, establishing the basal layer, and increase the scale thickness as it grows in diameter. The limiting layer develops last and involves a slow accumulation of mineral on the top of the scale throughout the life of the fish (Sire, 1986).

We suggest that when scales are lost, the highly mineralized limiting layer and external elasmodine should be preferentially synthesized from the regeneration process because they make the largest contribution to puncture resistance (Zhu et al., 2012). By contrast, although a thick collagen fiber bed is critical to toughness and tear resistance (Yang et al., 2014), we expected regeneration of this material would be prioritized less than the mineralized layers. Our results from measures of the layer topology, ply count and ply thickness showed that the regenerated scales in carp were composed of a single thick mineralized layer and plies of unidirectional collagen within the basal layer, which is consistent with previous work (Sire and Akimenko, 2004) and supports the suggestion that restoring physical protection is the proximate objective of the regeneration process. Whether carp scales organized the collagen within the mineralized layer into the serial unidirectional plies of the basal layer that become the foundation of the external elasmodine was unclear because we were not able to identify distinct collagen plies in this region by microscopic analysis. We expected to find mineralized plies of collagen in the basal layer because it is both stronger and tougher than non-mineralized collagen (Buehler, 2007) and would contribute to puncture resistance. There was also no clear demarcation between an external and internal elasmodine in the basal layer of the regenerated scales. However, based on Raman spectroscopy of the mineral to collagen ratio in the regenerated scales (Figs 6 and 7), the mineralized layer is similar in normalized thickness (approximately 0.3) and mineral to collagen ratio (approximately 2 to 6) as the external elasmodine of the ontogenetic scales. Nevertheless, the mineralized layers of the regenerated scales in our study represented a lower proportion of the total scale thickness than the mineralized zone (i.e. limiting layer and external elasmodine) of the ontogenetic scales and may reflect partial balancing for toughness and flexibility while simultaneously advancing mineralization.

Comparison of the scales that regenerated at 10°C and 20°C indicated that the water temperature had an important effect on mechanical behavior. E, S and MOT were all significantly lower in scales regenerated at 10°C than at 20°C. Although basal layer and total thickness were approximately 50% greater in the regenerated scales of carp held at 20°C, the apparent difference was significant for the basal layer thickness (P=0.047) but not total thickness (P=0.052). For the thickness of individual plies in the elasmodine layers of scales regenerated at 10°C and 20°C, the difference was not significant (P≥0.133). Hence, the basis for the differences in the mechanical properties of the regenerated scales at these temperatures appears to be unrelated to the microstructural features of the regenerated scales. However, Murcia et al. (2017) showed that S and E of the elasmodine layer increased with the ratio of external to internal elasmodine plies, which unfortunately could not be measured due to the lack of distinct external and internal elasmodine layers and absence of unidirectionally aligned plies of collagen fibrils in the regenerated scales. As an alternative, E, S, e_f and MOT can be expressed in terms of the degree of mineralization obtained by Raman spectroscopy and measured by the mineral to collagen ratio. The measures of mineral to collagen ratio in Fig. 8 show the average values across the mineralized region of the scale. Other than e_f, the average mineral to collagen ratio for each of the other mechanical properties was higher in the regenerated scales from fish held at 20°C than at 10°C, and likely contributed to their superior tensile testing performance. The E of the scales should be closely correlated with mineralization because mineralized collagen and the limiting layer are stiffer (Buehler, 2007; Arola et al., 2019). With the increased mineralization in the scales regenerated at 20°C, E nearly reached the same level as the ontogenetic scales. Also notable are the improvements with increasing mineralization for S and MOT (Fig. 8). Both of these properties exhibited an almost linear trend with the amount of mineralization, implying that mineral content has a marked role in the protection performance of the scales.

The results obtained from the compositional analysis of the scales mapped across the thickness using Raman spectroscopy further demonstrated that the maximum mineral to collagen ratio and the transition between the mineralized and non-mineralized layers differed between the regenerated scales of fish held at the two temperatures (Fig. 7). For the scales regenerated at 10°C, the peak mineral content is lower than for the scales at 20°C. Furthermore, the spatial distribution in the mineral to collagen ratio of the scales regenerated at 20°C is highly consistent with the distribution in the ontogenetic scales at 20°C. Nevertheless, although the overall thickness of this highly mineralized region and the overall thickness of the regenerated scales was lower compared with the ontogenetic scales of fish held at 20°C, it would be interesting to determine whether the regenerated scale morphology within individual fish would eventually converge to that of the ontogenetic scales, independent of temperature.

To the best of our knowledge, this is the first investigation of the structure–property relationships of both ontogenetic and regenerated fish scales obtained from the same fish. It is also the first assessment of the importance of temperature to the microstructure and mechanical properties of the regenerated scales. Despite its novelty, there are important limitations to the work that should be considered. First, the evaluation represents a snapshot of the structure and properties over the period of regeneration. According to their overall size (i.e. diameter), the regenerated scales had not reached confluence with the ontogenetic scales. Based on the lower thickness and absence of a well-defined limiting layer, the process of regeneration may not have been complete. Allowing the scales longer time to regenerate and assessing the microstructure at different stages of regeneration process could be informative. Our findings were also limited to the scales of six different fish and only three at each temperature. Larger numbers of fish held at each temperature would increase confidence in our inferences related to the variation in structure, composition and mechanical behavior. Furthermore, our study only examined scales closest to the head region. Other investigations have shown that mechanical properties of scales vary spatially over the body of fish (Marino Cugno Garrano et al., 2012; Gil-Duran et al., 2016), which could be important in comparing the process of regeneration, as would the spatial sequence of scale development over the body. For many species, including carp, scale development follows a squamation process that occurs from the posterior to the anterior (Sire and Arnulf, 1990), and it would be of interest to determine whether this process similarly affects scale regeneration.

Finally, mature carp in the Columbia River and elsewhere have relatively few natural predators. They may be highly preyed upon as juveniles by fish species, and traits that evolve in response to selection pressures early in life (e.g. high levels of predation) can be expressed throughout the life of the organism. However, their functional role may diminish or even change. Thus, although the scales in large, mature carp still serve for physical protection, this may be less critical for survival than the scales of other fish where the likelihood of predation is high over a large part of their lifespan. The regeneration timeline of scales from other fish and the development of their microstructures through the regeneration process may be quite different. Investigating scale regeneration in other species could provide key insights to the processes governing material development for physical protection. This topic deserves more consideration and will be the focus of our future studies.

An experimental investigation was performed to characterize the microstructure of the ontogenetic and regenerated scales from C. carpio, identify contributions to the mechanical properties of the scales and assess the importance of the environment in which they develop. Several fish were acquired wild from the Columbia River and then maintained live in an aquatic laboratory at either 10°C or 20°C. Ontogenetic scales were extracted from the fish for analysis, as well as the regenerated scales after a period of development and growth. Results showed that the overall mechanical behavior of the regenerated scales was inferior to that of the ontogenetic scales regardless of the rearing temperature. The largest difference between the ontogenetic and regenerated scales was for the scales that regenerated at 10°C; the properties most affected were the strength and modulus of toughness, with differences of 60% and 70%, respectively, with respect to the ontogenetic scales. For the fish held at 20°C, all properties were significantly different except for the elastic modulus. The regenerated scales did not exhibit a well-defined limiting layer, nor a stratified elasmodine (consisting of external and internal elasmodine), but rather an outer mineralized layer and underlying basal layer. Although the mineral to collagen ratio was highest for the scales that regenerated at 20°C, the mineralized layer occupied a larger portion of the total thickness for scales that regenerated at 10°C. An assessment of all scales showed that the strength and toughness of the scales was linearly related to the mineralization ratio and that it appears that the mineralized layer is developed preferentially in regenerated scales to achieve the most effective protection during the growth process.

The authors gratefully acknowledge the contributions of Dr Sandra Murcia (Intel) and Dr Arun Devaraj (PNNL) for their fruitful discussions. The authors also acknowledge the University of Washington Biostatistics Department for the confirmation on statistical methods used in this study.