Accuracy of the Aspartic Acid Racemization Technique in Age Estimation of Mammals and the Influence of Body Temperature

The aspartic acid racemization (AAR) technique has been applied for age estimation of humans and other mammals for more than four decades. In this study, eye lenses from 124 animals representing 25 mammalian species were collected and D/L ratios obtained using the AAR technique. The animals were either of known age or had the age estimated by other methods. The purpose of the study was to: a) estimate the accuracy of the AAR technique, and b) examine the effect of body temperature on racemization rates. Samples from four of the 25 species covered the range of ages that is needed to estimate speciesspecific racemization rates. The sample size from a single species of known age, the pygmy goat (Capra hircus, n = 35), was also large enough to investigate the accuracy of ages obtained using the AAR technique. The 35 goats were divided into three datasets: all goats (n = 35), goats >0.5 yrs old (n = 26) and goats >2 yrs old (n = 19). Leave-one-out analyses were performed on the three sets of data. Normalized root mean squared errors for the group of goats >0.5 yrs old were found to be the smallest. The higher variation in D/L measurements found for young goats <0.5 yrs can probably be explained by a period of continued postnatal growth of the eye lens. Normalized root mean squared errors from the leave-one-out cross-validation analyses based on goats >0.5 yrs old was for three age groups of the goats: 0.934 yrs for young goats <2 yrs (n = 16), 0.102 yrs for adult goats from 2–8 yrs (n = 15) and 0.133 yrs for old goats >8 yrs (n = 4). Thus, the age of an adult or an old animal can be predicted with approximately 10% accuracy, whereas the age of a


INTRODUCTION
The ability to accurately estimate the age of animals and humans is crucial in e.g. studies of life history parameters of wild animal populations, body condition of wildlife, or age at death in forensic science (Ohtani 1995, Ritz-Timme et al. 2000, Garde et al. 2010, Ohtani and Yamamoto 2010. Counting annual growth layers in teeth is the most widely used method of age estimation for most mammal species, however some animals lack teeth such as the baleen whales (Mysticeti) while others, like the narwhal (Monodon monoceros), has a highly specialized tooth structure. Reliable and routinely applicable age estimation methods are therefore needed for these species.
Age estimation using the racemization of aspartic acid in teeth and eye lens nuclei has been a field of research for the past 40 years Bada 1975, Nerini 1983). The aspartic acid racemization (AAR) technique has been thoroughly investigated using humans of known age (Helfman and Bada 1976, Masters et al. 1977, Ohtani et al. 1995a, Ritz-Timme et al. 2003), but only a few studies have been conducted on known-age individuals from non-human mammal species with limited sample sizes as well (Bada et al. 1980, Fujii et al. 1989, Ohtani et al. 1995b. Although it is a promising alternative to traditional age estimation methods, AAR still requires validation from known-age animals as well as a more thorough examination of the racemization rates for different species. Aspartic acid, as all amino acids except glycine, comes in two different isomeric forms, the laevo-(L) and the dextro-(D) (some amino acids can exist in four different chiral forms, e.g. isoleucine). In living proteins, only the L-form is utilized. However, in metabolically stable proteins the L-form is slowly, and at a constant rate, converted to the D-form, a process called racemization (Duin and Collins 1998). Aspartic acid racemizes the fastest of the amino acids and is as such measurable in living organisms. AAR has been detected in several types of human tissues such as teeth (Helfman and Bada 1976), eye lens nuclei (Masters et al. 1977), bone (Ohtani et al. 1998), brain (Fisher et al. 1992), and elastin (Ritz-Timme et al. 2003, Meissner andRitz-Timme 2010). Tissue from eye lens nuclei is especially suitable for AAR analysis as this is among the most stable in the human body (Masters et al. 1977). The accumulation of D-aspartic acid in the proteins of the various tissues has been found to cause significant damage and has been linked to a variety of age-related diseases such as cataracts (Masters et al. 1978), Alzheimer's (Fisher et al. 1992) and atherosclerosis (Powell et al. 1992).
The rate at which L-aspartic acid racemizes to D-aspartic acid differs among tissues depending on the physiochemical properties of the amino residues of aspartic acid and asparagine (Ritz-Timme and Collins 2002). Also, the racemization rate accelerates with increasing temperatures (Bada and Schroeder 1975). It was previously suggested that racemization rates in teeth and eye lens nuclei were similar between mammalian species (Bada et al. 1980), however rates have been found to vary in lens nuclei from species such as humans (Masters et al. 1977), cetaceans (Nerini 1983, Garde et al. 2012, Rosa et al. 2013, Nielsen et al. 2013) and seals (Garde et al. 2010), as well as in teeth from humans (Helfman and Bada 1976) and rats (Ohtani et al. 1995b). Inter-species differences in core body temperature have been proposed as the reason for these differences (Ohtani et al. 1995b, Garde et al. 2007, Rosa et al. 2013).
This study is two-fold and focuses on: a) determining the accuracy of the AAR technique in age estimation of mammals of known ages by use of the pygmy goat (Capra hircus) as a case species, and b) examining the effect of core body temperature on species-specific racemization rates using animals of known age from zoological gardens and free-ranging animals, where ages were estimated using a traditional ageing method.

Samples
Eye lenses for age estimation using the AAR technique were collected from 124 animals (25 mammal species; 7 orders) that had been held in captivity (zoos) or were free-ranging at the time of death (Appendix 1). In the period May 2007 -November 2009, eyes from 97 zoo animals (20 species, 6 orders) of approximate known age were collected. The zoo staff retrieved the eyes upon the animal's death and attained the individual ages. The remaining 27 animals were free-ranging (7 species; 3 orders) and had age estimated using other ageing techniques (Appendix 1). These included hooded seal (Cystophora cristata), harbour seal (Phoca vitulina), grey seal (Halichoerus grypus), fin whale (Balaenoptera physalus), killer whale (Orcinus orca), polar bear (Ursus maritimus) and reindeer (Rangifer tarandus).
Only a single species, the pygmy goat, constituted a sufficiently large and age-dispersed sample (n = 36) to predict the accuracy of the AAR technique for age estimation. Uncertainty regarding the age of one of the goats (ID no. Z1) led to exclusion of that particular goat in the following analyses. Of the 35 remaining goats, nine goats were <0.5 yrs old, seven were >0.5-1 yrs old, 16 were 2-8 yrs old and four goats were 9-11 yrs old.
To be able to produce species-specific racemization rates, samples covering the range of ages for each species are needed (Garde et al. 2010). Such sets of samples were available for two of the known-age zoo species, the rednecked wallaby (Macropus rufogriseus, n = 10) and the lion (Panthera leo, n = 5), and two species of the free-ranging marine mammals, the hooded seal (n = 6) and the fin whale (n = 7) (Appendix 1). Age estimates for the hooded seals were obtained by counting of GLGs in the cementum of lower canine teeth (Mansfield 1991). Results for the fin whales were previously published (Nielsen et al. 2013) and are not presented here, but the fin whale specific racemization rate established in Nielsen et al. (2013) were used in the analysis of temperature-dependent racemization rates in this study. The difficulty in obtaining large sample sizes from both captive as well as freeranging animals has resulted in small sample sizes of which the racemization rates for the red-necked wallaby, lion and hooded seal are based. These rates are therefore to be considered preliminary.

Eye lens dissection
Eyes were frozen at -20°C as soon as possible upon collection. In the laboratory, eye lenses were dissected out of the eyes. The outer lens layers were removed by rolling the lens on paper, and any remaining layers were removed under a stereoscope leaving only the lens nucleus for hydrolysis and subsequent High-Performance Liquid Chromatography (HPLC) analysis as described in detail in Garde et al. (2007;2012). Any lenses containing blood or with signs of putrefaction were discarded. The lens nuclei were parted in twoone half was used for analysis and the other half was archived at -20°C.

Hydrolysis and HPLC analysis
Procedures of Zhao and Bada (1995) and Garde et al. (2007; were followed for hydrolysis of samples and analysis by HPLC. Eye lens nuclei were hydrolysed in glass tubes containing 1 ml 6 M HCl for 6 hours at 100°C. Chromatography and data analysis was performed using an Agilent 1100 Series HPLC system (Agilent Technologies, Walbronn, Germany). Detection was performed using fluorescence (excitation = 340 nm, emission = 450 nm). The column was a Zorbax Eclipse XDB-C18, 4.6 150 mm, with particle size 3.5 µm.

 
The D/L ratios measured by HPLC were calibrated using the following D/L standards: 0.5/99.5, 1/99, 2/98, 5/95, 10/90, and 15/85, which were run at the beginning and end of each HPLC run. Measured D/L ratios from the eye lens nuclei were recalculated using calibration equations (linear regression) for the D/L standards. Two standard curves were produced for each HPLC run using D/L standards as described above. Linear regression equations for each run were calculated by regression of theoretical D/L ratios versus the measured D/L ratios from the D/L standards. The coefficient was r 2 = 0.99 for each run. Equations with a slope closest to 1 (range: 1.0 -1.3) from each run were used to recalculate D/L ratios for the samples.

D/L ratios, racemization rates (as 2k Asp values) and age estimates
Individual D/L ratios were generated for each of the 124 animals (Appendix 1). For 26 of the 124 individuals, both left and right eyes were analysed. A paired t-test showed no difference between the two eyes (t = 0.74, d.f. = 25, P = 0.46) and an average D/L ratio was used in subsequent analyses.
The pygmy goat Equations to estimate the age of a goat from an AAR measurement (D/L value) by regressing age on AAR were made following the procedures presented in Garde et al. (2012; and Ohtani and Yamamoto (2011). Three equations were made: the first equation included all goats (n = 35), the second excluded nine goats <0.5 yrs old (n = 26), and the third further excluded seven goats between >0.5-1 yrs old (n = 19). The D/L values from young individuals show higher variation than older individuals (George et al. 1999, Garde et al. 2007 and exclusion of the younger groups was done to find the best fit of the regression line. Known ages were regressed against individual ratios of (eq. (1); Garde et al. 2015). The ages from the 35 goats were not known exactly and the uncertainty (as standard deviation) differed between goats (Appendix 1) thus weighted regressions were used. Standard deviations for the known ages of each goat were assessed based on information on goat ages from Copenhagen Zoo. A leave-one-out cross-validation analysis was done for all three equations by performing 35, 26 and 19 regressions, one for each goat. Each regression was made on the remaining 34, 25 and 18 goats and the results from this regression were used to predict the age on the one remaining goat. The accuracy of an age estimate was determined by the distance between the predicted and the known age, normalized by the known age. Final prediction equations were made from 35, 26 and 19 goats. Accuracy of the ages was measured by the normalized root mean squared error from the leave-one-out cross-validation analysis, split into three age groups: young (<2 yrs, n = 16), adult (2-8 yrs, n = 15) and old (>8 yrs, n = 4).

[ ]
A goat specific racemization rate was estimated based on the group of goats with the smallest error (n = 26, >0.5 yrs old goats) and estimated following Garde et al. (2012). In this bootstrap study, the ages were drawn from uniform distributions on intervals given by the estimated age ± 2 times the standard deviation. The uniform distribution was chosen because ages were often given as rounded values, and thus, there was no information about where the true age might lie within an interval around the given age. The number of bootstrap replicates used was 10,000. The average estimate of the slope of the regression line corresponds to twice the racemization rate (2kAsp) and the intercept corresponds to twice the (D/L)0 value.

Species-specific racemization rates for the red-necked wallaby, lion and hooded seal
For the red-necked wallaby, lion and hooded seal species-specific racemization rates were made following the procedures in Garde et al. (2012;. Standard deviations for the known ages of each wallaby and lion were assessed based on information on individual ages from the zoos. For the hooded seals (n = 6) a standard deviation of one year was considered appropriate.

The effect of core body temperature on the rate of racemization
The effect of core body temperature on the rate of racemization was investigated by weighted regression of the four species-specific racemization rates (for pygmy goat, red-necked wallaby, lion and hooded seal) estimated in this study and eight published rates against the speciesspecific core body temperatures (°C). The eight published racemization rates were for humans (Masters et al. 1977 Core body temperatures instead of eyeball temperatures were chosen for the regression for three main reasons: 1) eye temperatures have been shown to be practically the same as or at least correlate with core body temperatures in several mammal species (Melero et al. 2015), 2) in situ measurements of whale eyeball temperatures are difficult if not impossible (Rosa et al. 2013),  NAMMCO Scientific Publications, Volume 10 and 3) published eyeball temperatures for all the 12 species used in this study were not available.

Accuracy of goat AAR ages and estimation of a goat specific 2k Asp and (D/L) 0 value
The slopes and intercepts of the three equations (weighted regressions, with weights proportional to the inverse of the variance) for the 35 goats, 26 goats >0.5 yrs and 19 goats >2 yrs had standard deviations of 6.9 and 0.7, 4.9 and 0.5, and 6.8 and 0.8, respectively (Table 1). Smallest standard deviations was found for the equation excluding nine goats <0.5 yrs even though this analysis builds on a smaller data set (n = 26). This equation was estimated to be: where AAR equals (Fig. 1). The three equations (Table 1) were then used to estimate how well we predict the ages by the normalized root mean squared errors in the leave-one-out cross-validation analyses. The best age predictions were obtained when excluding goats <0.5 yrs old (Eq. 1, Table 1), where errors were found to be 0.49 yrs for all 35 goats, and then 0.93 yrs for the young group (<2 yrs, n = 16), 0.10 yrs for the adult group (2-8 yrs, n = 15) and 0.13 yrs for the old group (>8 yrs, n = 4). Thus, the age of an adult or an old animal can be predicted with approximately 10% accuracy, whereas the age of a young animal is difficult to predict. Note that the adult group is expected to have a lower error since they fall in the middle of the covariate distribution.
A goat specific racemization rate and D/L0 value was estimated based on the 26 pygmy goats >0.5 yrs old. Known ages were regressed against values of and a bootstrap study was used to determine standard errors (Garde et al. 2012). The average estimate of the slope (2kAsp) was 0.0107 ± 3.8 x 10 -4 SE with an intercept of 0.0526 ± 0.0022 SE (r 2 between 0.91 and 0.96) (Fig. 2). We estimate the racemization rate to 0.00536 ± 1.9 x 10 -4 SE and the (D/L)0 to 0.0263 ± 0.0011 SE.

Estimation of species-specific 2k Asp and (D/L) 0 values for the red-necked wallaby, lion, and hooded seal
Species-specific racemization rates and D/L0 values were estimated for the red-necked wallaby, the lion and the hooded seal by regression of known or estimated ages against values of . A bootstrap study was used to determine standard errors (Garde et al. 2012).

Correlation of 2k Asp values against core body temperatures
The four species-specific 2kAsp values estimated in this study plus an additional eight published 2kAsp values were regressed against average species-specific core body temperatures (°C) (Fig. 3; Table 2). A positive relationship was found between 2kAsp values and core body temperatures (r 2 = 0.321).

Sampling
Robust estimation of species-specific racemization rates and (D/L)0 values ideally requires samples covering the range of ages including near-term fetuses or postpartum individuals. Large-scale collection of free-ranging known-age animals that grow to old ages is a challenge and inference from captive animals is the most obvious alternative. Collection of samples from captive known-age animals that grow to old ages has, however, proved difficult. Collaboration with the Copenhagen Zoo and Givskud Zoo in Denmark for a 2.5-year period resulted in samples from a range of species but only three species constituted sufficiently large sample sizes covering the range of ages to be used for estimation of species-specific racemization rates. Collection of samples from bovines from a Danish slaughterhouse was also attempted, however, most bovines are put down before the age of 8, which is not representable of longevity of about 20 years for domestic cattle (AnAge Database, September 2017). Collection of other domestic

Fig. 2.
Regression line from the bootstrap study of D/L ratios against known and estimated ages for the pygmy goat, lion, red-necked wallaby and hooded seal. The slope of the regression line corresponds to twice the racemization rate (2kAsp) and the intercept corresponds to twice the (D/L)0 value. animals e.g. horses, dogs or cats would not only require collaboration with numerous veterinarians but also assembling approvals from the owners, which was considered beyond the scope of this study.

Accuracy of the AAR technique for age estimation of known-age pygmy goats
During the sampling period only a single species, the pygmy goat, provided a large and age-dispersed sample. We found highest accuracy between predicted and known ages for the adult and old goats compared to the young. Our findings demonstrate that the AAR method is more reliable for adult and older pygmy goats, which is considered also to apply for other mammals (Garde et al. 2010;. In Garde et al. (2010), intra-specific variation in D/L ratios of 12 postpartum harp seals, used to estimate a species-specific (D/L)0 value, were also observed. Underestimation in AAR ages for young harp seals 1-7 years old was partly explained by this intraspecific variation and error in the (D/L)0 estimate. Overall, the study involving 113 harp seals found, however, good agreement between the two ageing methods used; AAR in eye lens nuclei and GLGs in teeth.
The variation found in D/L measurements for young pygmy goats as well as for young individuals of other mammalian species could be a result of eye lenses not being fully developed postpartum.

Development of the eye lens postpartum and its implication for AAR age estimation studies
The lens grows rapidly during late embryonic and early postnatal stages by cell division and differentiation (Lovicu and McAvoy, 2005). In humans, the lens grows rapidly in the embryo and during the first postnatal year (Lynnerup et al. 2008;, Bebe 2003. A recent study has shown that amino acids are incorporated into the proteins of rat eye lenses during fetal development and at pup age, revealing a postnatal growth of the rat lens. Beyond the age of 25 days synthesis of proteins in the rat eye lenses could no longer be detected and hence, no further growth of the lens was expected beyond this age. Proteins synthesized during fetal development and a few days after birth were also present later in the rat's life (>1 year) (Bechshøft et al. 2017). Fris and Midelfart (2007) also found changes in the contents of energy metabolites and amino acids in rat eye lenses postnatal. They suggested that the changes could be caused by the transformation of primary fiber cells to the fetal nucleusa transformation that occurs during the critical maturation of the rat lens from between day 12 and 16 to approximately 21 days after birth. After the age of approximately 30 days the lens will become resistant and the metabolic activity will diminish. The critical maturation period of the rat lens is related to the atrophy of a supporting vascular bed tissue (tunica vasculosa lentis) (Fris and Midelfart Garde et al. (2018) NAMMCO Scientific Publications, Volume 10 2007). By the fourth week after birth the tunica vasculosa lentis has completely regressed and delivery of several compounds to the avascular lens tissue significantly reduced (Lang 1997 : Fig. 3 for illustration of the regressing tunica vasculosa lentis). During the postnatal development of the lens, the content of specific amino acids in the lens will decline, among them aspartic acid, which is probably because aspartic acid is used in the synthesis of the lenticular proteins (Heinämäki and Lindfors 1988). The variation in D/L values observed in this study for pygmy goats <0.5 yrs old can probably be explained by a continued postnatal growth of the pygmy goat eye lens. Higher variation in D/L values of young individuals compared to adults and older individuals has also been observed in a range of marine mammals, including the bowhead whale (George et al. 1999, Rosa et al. 2013, narwhal (Garde et al. 2007;, harp seal (Garde et al. 2010), and minke whale (Nielsen et al. 2017) suggesting that eye lenses of marine mammals are also not fully developed postpartum. At which age the lens will cease growing in different species of marine mammals is, to our knowledge, unknown. The tunica vasculosa lentis, or a similar capillary structure, has been observed surrounding the eye lenses of young narwhals during dissection of the lens nucleus for AAR age estimation (pers. obs. E. Garde). Continued growth of the eye lens postnatally has significant implications for estimation of species-specific D/L0 values and the accuracy of age estimation of young marine mammals using the AAR technique. The AAR technique relies on a linear rate of racemization from birth to death of the individual, which is not the case in metabolically active tissues. Investigations in the development of the marine mammal lens as well as establishment of a model taking the early growth pattern into account will be a significant contribution to the continued use of AAR age estimation studies of young marine mammals.

Species-specific racemization rates and the effect of core body temperature
For the red-necked wallaby, lion, and hooded seal, a smaller yet representative sample size was collected. We recognize that the small sample sizes result in less precise estimation of the species-specific racemization rates and (D/L)0 values, but also emphasize the difficulty in obtaining samples from known-age captive animals of especially older age.
We found a positive relationship between the rate of racemization of aspartic acid and core body temperatures (r 2 = 0.321, Fig. 3), which is not surprising as it is well known that the racemization rate is a function of temperature (Bada and Schroeder 1975). Higher racemization rates were found for the pygmy goat and lion, which also have the highest core body temperatures (>38°C) of the species presented in Fig. 3. The harbour porpoise racemization rate was the highest of the marine mammals, which is consistent with the higher core body temperature of the harbour porpoise compared to the other marine mammals (Fig. 3). The rate of the beluga, however, seem high considering a relatively low core body temperature of 35.7°C (Melero et al. 2015). Also the narwhal, the beluga's closest relative, has a racemization rate considerably lower even though core body temperatures of the two species is almost the same (Fig. 3). The large SE of the beluga racemization rate estimate could explain the difference in rates observed for the beluga and the other marine mammals with similar low core body temperatures. The bowhead whale has the lowest racemization rate of all the marine mammals (Fig. 3) consistent with a low core body temperature. Our findings show that core body temperature indeed is a major driver of AAR, however, considering the differences in racemization rates of species with almost similar core body temperatures, e.g. the harbour porpoise (36.8°C) versus humans (36.8°C), also suggests that other factors besides temperature are involved. These other factors could include protein primary and secondary structures (Ritz-Timme et al. 2003), however, no Table 2. Core body temperature (°C), estimate of 2kAsp and (D/L)0 value and associated SE for four species in this study and published values for another eight species. Average core temperatures (°C) are from Teare (2002), except for records of fin whale (Brodie and Paasche 1985), minke whale (Folkow and Blix 1992), harbour porpoise (Desportes et al. (2003), bowhead whales (Rosa et al. 2013), narwhal (Heide-Jørgensen et al. 2014, and beluga (Melero et al. 2015). additional experiments were done in this study to further elucidate this feature, and a full discussion on this topic is beyond the scope of this paper. Rosa et al. (2013) also found a positive relationship between core body temperature and racemization rates using data from bowhead whales (Rosa et al. 2013), fin whales (Nerini 1983) and humans (Masters et al. 1977) and based on data from these three species constructed an equation for predicting 2kAsp values, when core body temperature was given. Rosa et al. (2013) suggested that age estimates can be obtained using the relationship between the rate of racemization and core body temperatures but also stated that it is better to estimate a rate for the species being studied than to use values from other species. We find that the slope of 2kAsp on core body temperatures in the range from 33.8-38.5°C was considerably greater than shown by Rosa et al. (2013) (Fig. 3).
To summarize, the accuracy of the AAR technique for age estimation of mammals was predicted using known age pygmy goats. The technique was found to be reliable for adults and older individuals, whereas it was less accurate for younger individuals, as also found for harp seals (Garde et al. 2010) and narwhals (Garde et al. 2012). This can probably be explained by a period of continued postnatal growth of the mammalian eye lens. Speciesspecific racemization rates were estimated for the pygmy goat, red-necked wallaby, lion, and hooded seal and a positive relationship was found by regression of racemization rates against core body temperatures. We found that core body temperature is a major driver of AAR, however, other factors besides temperature are involved in the racemization process in living animals. Based on our results we emphasize that non-species-specific racemization rates should be used with care in AAR age estimation studies, and that the period of postnatal growth of the eye lens be considered when estimating species-specific D/L0 values and ages of young individuals.