Information

Does the metabolic rate determine how fast the telomeres shorten?

Does the metabolic rate determine how fast the telomeres shorten?


We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

In many papers one can read that telomeres may play an important role in longevity. According to Calado et al.1 the telomeres of mice are much longer than the telomeres of humans. However, mice have an average life expectancy of about 2 years and humans about 80 years.

Does the metabolic rate therefore determine how fast the telomeres shorten? Can one say that the higher the metabolic rate, the faster the telomere shorten?

  1. Telomere dynamics in mice and humans

Telomere shortening and survival in free-living corvids

Evidence accumulates that telomere shortening reflects lifestyle and predicts remaining lifespan, but little is known of telomere dynamics and their relation to survival under natural conditions. We present longitudinal telomere data in free-living jackdaws (Corvus monedula) and test hypotheses on telomere shortening and survival. Telomeres in erythrocytes were measured using pulsed-field gel electrophoresis. Telomere shortening rates within individuals were twice as high as the population level slope, demonstrating that individuals with short telomeres are less likely to survive. Further analysis showed that shortening rate in particular predicted survival, because telomere shortening was much accelerated during a bird's last year in the colony. Telomere shortening was also faster early in life, even after growth was completed. It was previously shown that the lengths of the shortest telomeres best predict cellular senescence, suggesting that shorter telomeres should be better protected. We test the latter hypothesis and show that, within individuals, long telomeres shorten faster than short telomeres in adults and nestlings, a result not previously shown in vivo. Moreover, survival selection in adults was most conspicuous on relatively long telomeres. In conclusion, our longitudinal data indicate that the shortening rate of long telomeres may be a measure of ‘life stress’ and hence holds promise as a biomarker of remaining lifespan.


Abstract

Many organisms are capable of growing faster than they do. Restrained growth rate has functionally been explained by negative effects on lifespan of accelerated growth. However, the underlying mechanisms remain elusive. Telomere attrition has been proposed as a causal agent and has been mostly studied in endothermic vertebrates. We established that telomeres exist as chromosomal-ends in a model insect, the field cricket Gryllus campestris, using terminal restriction fragment and Bal 31 methods. Telomeres comprised TTAGGn repeats of 38 kb on average, more than four times longer than the telomeres of human infants. Bal 31 assays confirmed that telomeric repeats were located at the chromosome-ends. We tested whether rapid growth between day 1, day 65, day 85, and day 125 is achieved at the expense of telomere length by comparing nymphs reared at 23°C with their siblings reared at 28°C, which grew three times faster in the initial 65 days. Surprisingly, neither temperature treatment nor age affected average telomere length. Concomitantly, the broad sense heritability of telomere length was remarkably high at

100%. Despite high heritability, the evolvability (a mean-standardized measure of genetic variance) was low relative to that of body mass. We discuss our findings in the context of telomere evolution. Some important features of vertebrate telomere biology are evident in an insect species dating back to the Triassic. The apparent lack of an effect of growth rate on telomere length is puzzling, suggesting strong telomere length maintenance during the growth phase. Whether such maintenance of telomere length is adaptive remains elusive and requires further study investigating the links with fitness in the wild.


1 INTRODUCTION

Variation in life histories is thought to result from differential allocation of limited resources to competing life history traits. Such trade-offs and the resulting optimal resource allocation may vary with environmental conditions (Stearns, 1992 ). For example, tropical environments have favored a slow pace of life, that is, reduced fecundity but increased life span, in many vertebrates (Ricklefs & Wikelski, 2002 ). This is especially well studied in birds where tropical species produce fewer, but higher quality offspring (Jetz, Sekercioglu, & Böhning-Gaese, 2008 Martin, 2015 ), have lower basal metabolic rates (Tieleman et al., 2009 Wiersma, Muñoz-Garcia, Walker, & Williams, 2007 ) and live longer (Møller, 2007 Peach, Hanmer, & Oatley, 2001 ) than temperate species. Therefore, a comparison between tropical and temperate species may reveal physiological constraints that may limit the evolution of alternative combinations of life history traits (Ricklefs & Wikelski, 2002 ).

An important candidate mechanism with respect to physiological constraints of growth, reproduction and survival are telomeres (Haussmann & Marchetto, 2010 ). Telomeres are noncoding DNA—protein caps at the end of eukaryotic chromosomes that protect genomic integrity, but shorten during cell division and potentially when exposed to oxidative stress (Boonekamp, Bauch, Mulder, & Verhulst, 2017 Reichert & Stier, 2017 Zglinicki, 2002 ). Critically, short telomeres eventually lead to cell senescence or death (Blackburn, 2000, 2005 ), and the accumulation of cells with short telomeres may be one of the factors that causes aging and senescence in vertebrates (López-Otín, Blasco, Partridge, Serrano, & Kroemer, 2013 ).

Both longitudinal and cross-sectional studies in birds show that, in general, older individuals have shorter telomeres than younger ones with the greatest loss in telomeres occurring early in life (Heidinger et al., 2012 Pauliny, Larsson, & Blomqvist, 2012 Salomons et al., 2009 Spurgin et al., 2018 Tricola et al., 2018 ). Furthermore, an increasing number of studies in birds show that individuals with longer telomeres or little telomere attrition have better survival prospects than individuals with short telomeres or high levels of telomere attrition (reviewed in Wilbourn et al., 2018 ). This has been especially well studied in zebra finches (Taeniopygia guttata), for which it has been shown that long telomeres in early life are associated with increased survival and a long life span (Heidinger et al., 2012 ). In addition, studies in a variety of species show that telomere dynamics are sensitive to environmental influences such as variations in food availability (Spurgin et al., 2018 ), parasitic diseases (Asghar et al., 2015 ), and exposure to stress (Hau et al., 2015 ). In particular, conditions experienced during development can influence telomere dynamics. For example, exposure to poor or stressful environments can lead to accelerated telomere loss in young birds (Costanzo et al., 2017 Haussmann, Longenecker, Marchetto, Juliano, & Bowden, 2012 Herborn et al., 2014 Nettle et al., 2017 Salmon, Nilsson, Nord, Bensch, & Isaksson, 2016 Soler et al., 2017 Young et al., 2017 ), which can be predictive of decreased survival as nestlings or fledglings (Boonekamp, Mulder, Salomons, Dijkstra, & Verhulst, 2014 Salmon, Nilsson, Watson, Bensch, & Isaksson, 2017 Watson, Bolton, & Monaghan, 2015 ). Thus, telomere length and the rate of telomere loss are considered biomarkers of individual health and quality (Young, 2018 ).

Fewer studies have compared telomere length between taxa that vary in their life histories and life span. In mammals, a comparative study found that short-lived, small species have longer telomeres and higher telomerase expression than long-lived, large species (Gomes et al., 2011 ). In a study on rodents, no relationship between maximum lifespan and telomere length was detected (Seluanov et al., 2007 ). In birds, absolute telomere length does not seem to relate to variation in lifespan between species however, longer-lived avian species seem to have lower rates of telomere shortening than shorter-lived species (Dantzer & Fletcher, 2015 Haussmann et al., 2003 Sudyka, Arct, Drobniak, Gustafsson, & Cichoan, 2016 Tricola et al., 2018 ). This relationship between rate of telomere loss and maximum lifespan in birds may be caused by variation between species in how well telomeres are maintained throughout their lifespan. In addition, it may reflect selective disappearance of low-quality individuals with short telomeres. In longer-lived species, that experience lower levels of extrinsic mortality, individual condition, and thus telomere dynamics, may play a greater role as determinants of mortality (Kirkwood & Austad, 2000 ). Therefore, selective disappearance of individuals with short telomeres may be more apparent in long-lived species (Tricola et al., 2018 ).

Tropical species live in less seasonal environments with lower levels of adult extrinsic mortality than temperate ones (Brown, 2014 ). Consequently, tropical songbirds have higher survival probabilities than temperate birds (Martin et al., 2017 Muñoz, Kéry, Martins, & Ferraz, 2018 ). Therefore, stronger selective disappearance of individuals with short telomeres is expected in tropical compared to temperate birds. However, mortality rates are age-specific, and therefore, the strength of selective disappearance may vary with age. In birds, mortality is usually highest during the first year of life, especially directly after fledging (Cox, Thompson, Cox, & Faaborg, 2014 Naef-Daenzer & Grüebler, 2016 ). As predicted by life history theory (McNamara, Barta, Wikelski, & Houston, 2008 ), juvenile survival is in general higher in tropical compared to temperate birds (Lloyd, Martin, & Roskaft, 2016 Remes & Matysiokova, 2016 ). Tropical parents take care of their fewer fledglings for considerably longer than temperate birds and may thereby be able to lower extrinsic mortality in juveniles (Styrsky, Brawn, & Robinson, 2005 ). We, therefore, hypothesize that differential survival of high-quality fledglings should be more apparent in tropical compared to temperate birds. Assuming that telomeres are bioindicators of somatic state and individual quality we expect that in tropical birds, individuals with short telomeres disappear faster from a population than in temperate birds both during the critical first year of life and later as adults.

In addition, there is good evidence that tropical species invest more into self-maintenance, but are less fecund than temperate species. For example, tropical species exhibit stronger sickness behavior after infection during the breeding season than temperate species (Owen-Ashley, Hasselquist, Raberg, & Wingfield, 2008 ). Furthermore, in tropical species, reproductive workload is reduced as they lay smaller clutches, thereby caring for fewer young and expending less energy than temperate birds (Nilsson, 2002 Tieleman et al., 2006 ). Thus, they may reduce high levels of oxidative stress (Noguera, 2017 ) and potential telomere loss associated with breeding (Reichert et al., 2014 ). In addition, tropical songbirds seem to have lower post-natal metabolic rates and slower, more sustained growth despite similar nestling times than temperate birds (Martin, 2015 Ton & Martin, 2016 ). These slower growth trajectories in combination with increased parental care per offspring may favor lower levels of telomere attrition during early life in tropical birds, which in turn may be important determinants of their longer life spans (Monaghan & Ozanne, 2018 ). Thus, longer-lived tropical species, which invest more in growth and self-maintenance than in fecundity, are expected to show longer telomeres as nestlings or a slower rate of telomere loss than short-lived temperate species. To determine how life history variation has shaped variation in telomeres, comparisons between the same or closely related species in different environments are necessary.

We collected samples from Afrotropical (referred to as tropical) and temperate European (referred to as temperate) male stonechats from different individuals at four age classes: as nestlings, as pre-breeding first-year birds (within their first 6 months of life), during their first breeding season (

1-year-old), and during their further adult life (≥2 years old). Because of their slow pace of life, tropical stonechats are expected to prioritize somatic maintenance over fecundity. Thus, we expected longer telomeres in tropical compared to temperate stonechats. Further, because of their higher juvenile survival and extended parental care we expected a shallower decrease in telomere length during the first year of life in tropical compared to temperate stonechats.


2 MATERIALS AND METHODS

2.1 Study organism

9 months old) female dragons (Ctenophorus pictus, W. Peters, 1866) were caught by noose or hand at Yathong Nature Reserve, NSW, Australia (145°35′E 32°35′S) and taken to holding facilities at the University of Sydney in October 2015 where they were housed for the duration of the experiments. Animals were collected under a permit issued by NSW National Parks and Wildlife Service (SL100352), and experiments were conducted in accordance with University of Sydney ethics approval (AEC-2013/6050). Females were housed in pairs in opaque plastic tubs (330 × 520×360 mm) with sandy substrate and exposed to a 12 hr light: 12 hr dark cycle. The lizards were fed mealworms and crickets, dusted with calcium and multivitamins, to satiation every day, and the cages were misted with water once a day. Heat lamps and ceramic hides were provided to allow the lizards to thermoregulate to their preferred body temperature (36°C M.O., unpublished data obtained from cloacal temperature readings in the wild). A mound of moist sand was available in each tub to allow females to burrow and lay eggs. Females were checked for oviposition each day, as apparent from skin flaps around the abdomen. Overall, 18 females produced 22 clutches and 89 eggs in total between 4 November 2015 and 6 January 2016. The eggs were removed and weighed to the nearest 0.1 g. As part of an investigation on the effects of temperature on TL, the individual eggs from each clutch were then placed sequentially in one of four incubators (27, 30, 32, 36°C, ±0.5°C) to separate clutch and temperature effects. Eggs were half-buried in a 1:7 mix of water to vermiculite in sealed, transparent, containers in each incubator. Containers were sealed to reduce evaporation of water but aerated weekly. Eggs were checked daily for nonviable eggs, which were removed. Seventy-three dragons hatched and 16 nonviable eggs were removed (most from the 36°C incubator, which is warmer than estimated incubation temperature in the wild: approximately 30°C, M.O., unpublished data). Hatchlings were sexed (determined by the presence or absence of hemipenes), weighed to the nearest 0.1 g, and snout–vent length (SVL) and total length were measured to the nearest mm. Each individual was marked with a unique toe-clip pattern for identification. Individuals were then placed in groups in tubs with the same conditions as the adult females, with the exception of the sand mound. Hatchling dragons were fed pinhead crickets, dusted in calcium and multivitamins, to satiation every day. Mortality is high in juvenile painted dragons but tubs were checked multiple times a day and any dead individuals were promptly removed.

2.2 Sample collection

In March 2016, the 24 remaining juveniles (14 female, 10 male, ranging in age from 97 to 140 days, 1 female and 3 males incubated at 27°C, 6 females and 5 males incubated at 30°C, 5 females and 2 males incubated at 32°C, 2 females incubated at 36°C) were euthanized by decapitation. Prior to decapitation individuals were first sedated by cooling in a refrigerator for 10 min and in a freezer for 2 min. They were sexed again to confirm prior assignations, weighed to the nearest 0.1 g, and SVL and total length measured to the nearest mm. Whole blood, whole brain, heart, liver, and spleen were then collected. Each tissue type was placed in 300 µl of RNAlater (Sigma-Aldrich, Castle Hill, NSW, Australia) and stored at −80°C until DNA extraction.

2.3 Quantifying relative telomere length

To analyse relative telomere length (RTL relative to the 18S gene), we first purified DNA from the collected tissues. The brain, heart, and liver were sliced into small pieces, and 50 µl of the diluted blood was used in the extraction. A DNeasy Blood and Tissue Kit (Qiagen, Chadstone, VIC, Australia) was used for extractions, according to the manufacturer's instructions. The protein kinase digestion step was run for 10 min for the blood, but the brain, heart, liver, and spleen required overnight incubation for complete digestion. RNase A (Qiagen, Chadstone, VIC, Australia) was added at the recommended concentration. The DNA concentration (ng/µl) and A260:A280 ratio of each sample were measured in duplicate using a Nanodrop (Thermo Fisher Scientific) and aliquots diluted to 10 ng/µl using the AE buffer provided in the DNA extraction kit. Only samples with a A260:A280 ratio between 1.7 and 1.9 and a concentration above 10 ng/µl were considered high enough quality to be used in analyses. Ultimately, 22 blood (12 female, 10 male), 18 liver (9 female, 9 male), 17 heart (8 female, 9 male), 16 brain (11 female, 5 male), and 10 spleen (6 female, 4 male) samples were of sufficient quality for use in the study. DNA was stored at −30°C.

Telomere length was measured using real-time quantitative PCR (qPCR) using SensiMix SYBR No-ROX Kit (Bioline, Sydney, NSW, Australia) and a Rotor-gene 6000 thermocycler (Qiagen, Chadstone, VIC, Australia) according to published protocols (Rollings, Friesen, et al., 2017 Rollings, Uhrig, et al., 2017 ) using techniques developed by Criscuolo et al. ( 2009 ) and Plot, Criscuolo, Zahn, and Georges ( 2012 ) with the 18S ribosomal RNA (18S) gene as the nonvariable copy number reference gene. The telomere primers used were Telb1 (5′-CGGTTTGTTTGGGTTTGGGTTTGGGTTTGGGTTTGGGTT-3′) and Telb2 (5′-GGCTTGCCTTACCCTTACCCTTACCCTTACCCTTACCCT-3′, Cawthon, 2002 ).

The 18S gene (92 bp amplicon in Anolis) was selected as the reference gene as it had previously been validated in reptiles (Plot et al., 2012 Rollings, Uhrig, et al., 2017 ). The primer sequences used were 18S-F (5′-GAGGTGAAATTCTTGGACCGG-3′) and 18S-R (5′-CGAACCTCCGACTTTCGTTCT-3′). Reactions were run in triplicate for each sample, with each run containing either Telb or 18S primers. Amplifications were carried out in a Rotor-Gene 6000 thermocycler (Qiagen, Australia) using an initial Taq activation step at 95°C for 10 min and a total of 40 cycles of 95°C for 15 s, 60°C for 15 s, and 72°C for 15 s. Each reaction had a final volume of 20 µl with 10 ng of DNA, forward and reverse primers used at a concentration of 250 nM, and MgCl2 added for a concentration of 1.7 mM. 11.25 µl of the SensiMix SYBR No-ROX Master Mix was added per reaction. A melt curve was generated after each run over the temperature range of 60 to 95°C to ensure that there was no nonspecific product amplification (see appendix Figure S1 and Figure S2 for examples). All of the DNA samples for a given individual were included in the same run. No-template control reactions were run in triplicate for each primer set during every qPCR run to ensure that there was no contamination. Standard curves were produced, using the pooled DNA from three randomly selected lizards, for both telomeres and 18S using fourfold serial dilutions to ensure consistent rates of amplification over a wide range of concentrations (60 ng/µl down to 0.05859 ng/µl with 6 different concentrations in total: Appendix Figure S3 and Figure S4) giving a linear dynamic range of 0.05859 to 60 ng/µl. The reaction was considered consistent when the linear correlation coefficient exceeded 0.985. The efficiency of the telomere amplification was 1.17 and the efficiency of the 18S amplification was 0.97, and samples all fell within the same concentration range as our standards. All runs included the same “golden standard” and also a no-template control to detect contamination. LinRegPCR 2016.0 (Heart Failure Research Centre, Ruijter et al., 2009 , Tuomi, Voorbraak, Jones, & Ruijter, 2010 ) was used to analyse the qPCR data. The starting concentrations of telomere (T) and control gene (S 18S) as determined with LinRegPCR were used to determine the relative telomere length with the calculation T/S. Telomeres and the control gene were assessed in separate runs. The mean interassay coefficient of variation for qPCR runs for telomere (n = 4) and 18S (n = 4) amplification were 2.98% and 0.59%, respectively, calculated using the golden standard. The intra-assay coefficient of variation for telomere and 18S runs were 1.09% and 0.90%, respectively. A general linear model found no significant difference in the distribution of the sexes among the plates (F1,23 = 1.211, p = 0.283). To investigate possible run effects, we took the mean value for each triplicate as produced by LinRegPCR for the blood data to simplify the analysis. An ANOVA of the ln-transformed (for normality) 18S data showed no significant difference among the runs (F3,18 = 0.795, p = 0.513) and neither did the runs containing telb (F3,18 = 0.888, p < 0.466).

2.4 Statistical analyses

Analyses were conducted with SAS 9.4 (SAS Institute, Cary) and SPSS 25.0 (IBM, Armonk). RTL was ln-transformed in order to conform to normality, as verified with a Shapiro–Wilk test. Potential effects of body condition (BCI) were assessed by generating the residuals of a regression analysis of mass versus SVL at death. First, to test whether incubation at different temperatures had affected the results, a mixed model analysis of the relationship between RTL and incubation temperature, with individual ID and maternal ID as random factors, was conducted and found not significant (F1,15.1 = 0.17, p = 0.6818, n = 78). To further investigate potential effects of incubation temperature, Pearson's correlation coefficients of temperature, incubation time, and hatching mass were calculated. Pearson's correlations were conducted between RTL and SVL at hatching and death, mass at hatching and death, residual BCI at death, age at death, and growth rate (calculated as the difference between hatching and death SVL, divided by age). Pearson's correlation coefficients were also calculated for the telomeres of all combinations of the tissue types to test for similarity in telomere dynamics across the tissues. We chose not to apply a correction (e.g., Bonferroni) to this analysis despite the number of correlations tested for as the low sample sizes available limit our statistical power and increase the probability of type II errors. Application of a correction would only further increase the chance of type II errors and reduce our probability of detecting real effects (Nakagawa, 2004 ). To test for sex- and organ-specific telomere effects a mixed model analysis of the relationship between RTL and sex, organ type and a sex×organ type interaction was tested with ID included to control for multiple measures from the same individual. The sex×organ type interaction was not significant (F4,73 = 1.206, p = 0.316) but was retained in the final model as it resulted in a smaller −2 Res log likelihood. As organ type effects were detected, pairwise contrasts between each organ type were conducted. Sequential Bonferroni adjustments were performed for all pairwise contrasts. As sex-based differences in RTL were found, GLMs testing sex-based differences in SVL at hatching and death, mass at hatching and death, residual BCI at death, age, and growth rate were performed to determine whether size differences might account for the variation in RTL.


The rate of telomere loss is related to maximum lifespan in birds

Telomeres are highly conserved regions of DNA that protect the ends of linear chromosomes. The loss of telomeres can signal an irreversible change to a cell's state, including cellular senescence. Senescent cells no longer divide and can damage nearby healthy cells, thus potentially placing them at the crossroads of cancer and ageing. While the epidemiology, cellular and molecular biology of telomeres are well studied, a newer field exploring telomere biology in the context of ecology and evolution is just emerging. With work to date focusing on how telomere shortening relates to individual mortality, less is known about how telomeres relate to ageing rates across species. Here, we investigated telomere length in cross-sectional samples from 19 bird species to determine how rates of telomere loss relate to interspecific variation in maximum lifespan. We found that bird species with longer lifespans lose fewer telomeric repeats each year compared with species with shorter lifespans. In addition, phylogenetic analysis revealed that the rate of telomere loss is evolutionarily conserved within bird families. This suggests that the physiological causes of telomere shortening, or the ability to maintain telomeres, are features that may be responsible for, or co-evolved with, different lifespans observed across species.

This article is part of the theme issue ‘Understanding diversity in telomere dynamics'.

1. Introduction

With advancing age, organisms experience gradual, functional deterioration that leads to diminished performance and a rising risk of mortality. While ageing, or senescence, is a common occurrence across taxonomic groups, the pattern and the pace of the ageing process are variable both within and among species [1]. Understanding the processes that underlie this variation remains a central question across diverse fields of biology. Within evolutionary biology, a better understanding of the underpinnings of the ageing process can ultimately give insight into life-history trade-offs and the evolution of lifespans [2].

One factor that contributes to the ageing phenotype is cellular senescence [3,4]. This process causes an irreversible change to a cell's state, in which the cell ceases to divide and undergoes distinctive phenotypic alterations, including an altered secretory profile that can damage nearby healthy cells [3,5]. Because the number of senescent cells rises during the ageing process, there is a growing loss of the regenerative capacity of cells with age [6]. Cellular senescence occurs as a complex response to excessive extracellular or intracellular stresses, including, but not limited to, severe DNA damage, mitochondrial deterioration, oxidative stress and telomere dysfunction [5]. The progressive erosion of telomeres, the noncoding DNA sequences at the end of linear eukaryotic chromosomes, ultimately triggers a permanent DNA damage response that causes cells to enter senescence [7]. While telomere shortening is only one contributing factor to cellular senescence, it has become a biomarker for senescent cells [8] and the physiological state of an organism [9,10].

The structure of telomeres is generally consistent across eukaryotes, suggesting that telomeres are an ancient and effective guardian of the genome [11]. Consistent with this, telomeres play a broad role in the maintenance of chromosomal genomic stability. Normally, the very end of the telomere folds back on itself to form a structure referred to as the ‘T-loop’, and along with associated proteins, effectively caps the ends of chromosomes. However, as cells replicate, their telomeres shorten due to replication restrictions of DNA polymerase at the ends of chromosomes [12]. Telomere shortening during DNA replication may also be propagated by single-stranded breaks in telomeric regions due to oxidative stress [13]. This progressive telomere shortening will ultimately expose an uncapped free chromosome end that leads to permanent cell cycle arrest [14]. Telomere dysfunction, then, augments the ever-growing pool of senescent cells and could thereby contribute to the decline in tissue function and integrity that is a hallmark of ageing [15].

The ties between telomere loss and cellular senescence suggest an important role of telomere shortening in age-related declines of physiological function. In support of this, a number of human studies have found that individuals with shorter telomeres also have reduced life expectancy [16–19], though other studies do not report this relationship [18,20,21], and see [22] for a discussion of the causal role of telomere shortening in ageing. One particularly interesting within-pair analysis of Swedish twins reported that the twin with the shorter telomere length also had a mortality rate that was three times higher than their co-twin [19]. Consonantly, a growing number of nonhuman studies in natural populations, particularly in wild birds, also show that survival prospects are related to telomere length [23–30]. And a notable report on zebra finches suggests that early life telomere length (at 25 days of age) is a very strong predictor of longevity [30].

While the number of studies exploring links between telomere length and survival within species continues to grow, there have been relatively few comparative studies exploring how telomere biology is associated with ageing rates or lifespans among species. Comparative studies provide a powerful tool to explore vertebrate evolution of ageing in the wild as these studies take advantage of the wealth of variation in lifespans among species generated by long-term selection across different environments. In particular, birds are an especially attractive model for comparative studies of ageing as they are conveniently monitored using ringing and exhibit wide variation of life-history traits [31].

Comparative work has revealed that while absolute telomere length does not seem to relate to longevity among species ([32–35], but see [11]), the rate at which telomeres shorten may offer better insight into the evolution of the lifespan of a species [35]. The first study, done almost 15 years ago, found that species-specific rates of telomere degradation are predictive of maximum lifespan among five avian species and eight mammalian species [35]. To our knowledge, since that time only two additional studies have explored this question, and both supported the original finding, suggesting that variation in telomere degradation rates among species is indicative of distinct levels of telomere maintenance [36,37]. These studies made some advances in experimental design and analysis as Dantzer and Fletcher [36] increased the number of species studied while controlling for phylogeny, and Sudyka et al. [37] focused on longitudinal datasets. However, both studies also relied on pre-existing data that were generated by a variety of techniques to measure telomere length and even within a method, different laboratories can produce widely different results [38,39]. In addition, many of these studies apply the relative qPCR method of telomere measurement. While this technique offers quick and comparable telomere measurements within a study, it is not appropriate for quantitative telomere length comparisons among laboratories or species [39]. Alternatively, the telomere restriction fragment (TRF) assay is more time-consuming but provides an absolute telomere length based on a physical molecular marker that produces more commensurable results across species [39]. Studies based on data that were all generated using the same telomere measurement methodology and analysis while controlling for phylogeny are still sorely needed [9,40].

(i) Our main hypothesis was that rates of telomere loss are associated with maximum lifespan of species, so that those species with slower rates of telomere erosion also have longer maximum lifespans [35]. Because of the comparative nature of this study, patterns of telomere length and age in a population or species may also be caused by selective disappearance of individuals with short telomeres [41] and we discuss this as well.

(ii) A recent comparative study using phylogenetic analyses of over 60 mammalian species reported that mean telomere length of a species is inversely correlated with lifespan [11]. To our knowledge, this relationship has yet to be tested in a phylogenetically controlled study of birds.

(iii) The telomeric brink hypothesis [42] postulates a causal role for telomere shortening in shaping longevity. If critically short telomeres increase the risk of mortality, then a corollary to this hypothesis is that species with shorter mean telomere lengths and faster telomere loss rates should also have shorter lifespans, which we test here.

(iv) Another recent hypothesis in telomere biology is that long telomeres shorten more quickly than short telomeres, possibly because longer telomeres are more sensitive to telomere-damage events [26]. To our knowledge, this hypothesis has only been evaluated within species, and we tested whether there is any support across species of birds.

2. Methods

(a) Species

We explored telomere shortening in cross-sectional blood samples from 19 avian species representing 5 orders (table 1 and figure 1). We chose species in which long-term study populations were available allowing us to sample individuals over a wide range of their predicted maximum lifespan (table 1). Maximum lifespan estimates for these species range from 7 to 50 years, and were based on natural, long-term study populations. Sex was unknown for a substantial number of individuals in many of the species, and thus, sex was not included in the analysis. We acknowledge that a potential bias in our results may arise from sex differences in mortality as males and females often differ in mortality and lifespan [48] and telomere attrition rates can also differ by sex [49]. Because sex was unknown, the average maximum lifespan between males and females was used in our analyses. While there may be some error in the maximum lifespan estimates, we do not think it would be enough to change our conclusions. Although some studies have suggested that median or mean lifespan may reflect differences in the ageing process more accurately, these values were not available for all species. In addition, maximum lifespan estimates allowed us to be consistent with previous studies [35–37].

Figure 1. Telomere length (from whole blood measured by TRF analysis) as a function of age in 19 bird species included in the comparative analysis. The lines are linear regressions, and the slope of the regression line for telomere length versus age was used as the telomere rate of change (TROC). The slope of the regression, its standard error and the r 2 are printed within the panel of each species.

Table 1. Cross-sectional estimates of blood cell telomere rate of change (TROC) from 19 avian species studied (including order and family), observed maximum lifespan (with literature references) and body mass. Sample size (number of individuals) is included for each species along with the range of ages sampled. *Personal communication from D.W.W., 2011.

(b) Laboratory methods

All samples analysed in this study were measured using the TRF assay following previously established methodology [26,50,51]. The TRF assay was developed over 25 years ago and is still widely used to validate other telomere measurement techniques [39]. The assay uses restriction enzyme digestion of genomic DNA followed by Southern hybridization to a radioactive probe containing a terminal repeat to measure mean telomere length from a distribution of TRFs. The traditional TRF assay does have some disadvantages, among them that along with the terminal telomeric repeats, it also measures interstitial telomeres and is biased against the detection of short telomeres. A variant of this, the in-gel TRF assay that we used here, resolves both of these issues by only probing the short G-strand overhang [39,52]. It is important to note that telomere measurements for 17 of the 19 species were measured in a single laboratory (M.F.H.), while the remaining two species (Haematopus ostralegus and Parus major) were measured with consistent methodology adopted from the aforementioned laboratory. Removing these two species from the analysis does not change any of the following results. In addition to consistency within the TRF technique, blood sampling methods and storage, and DNA extraction are known to influence telomere length estimates [39]. In the light of this we used consistent storage and extraction protocols, and if small changes were made, we were able to validate that, within this sample set, it did not influence telomere measurement. For specific methodology on the in-gel TRF assay used in this study see Haussmann et al. [51].

(c) Telomere length analysis

Gels from all 17 species were imaged on a phosphor screen with a Typhoon Variable Mode Imager (Amersham Biosciences, Buckinghamshire, UK) to visualize telomeres. The amount of radioactive signal (optical density, OD) in each lane corresponds with the amount of telomere at that position on the gel (i), and was quantified by densitometry in ImageJ (v. 1.51). Background signal from nonspecific binding of the radioactive probe was subtracted from all OD measures. The specific molecular markers on each gel differed because gel conditions were optimized based upon a species' particular telomere distribution (1 kb DNA ladder (1–10 kb) 1 kb plus DNA ladder (1–12 kb) Lambda DNA/EcoR1 +HindIII (1–21 kb) 1 kb DNA extension ladder (1–4 0 kb), Invitrogen λ DNA Monocut (2–49 kb) New England Biolabs DNA Marker XV (2–49 kb), Roche PFG Marker 1 (15–200 kb), New England Biolabs). However, regardless of the molecular marker used, the distance each band of the molecular marker migrated (i) was plotted against the molecular weight in kilobases and converted into molecular weights (L) using a three-parameter log-linear function. The mean TRF length (called mean telomere length hereafter for simplicity) for each individual was calculated using: mean TRF = ∑(ODi * LI)/∑(ODi), where ODi is the densitometry output at position i, and Li is the length of the DNA (kB) at position I [52].

(d) Statistical analysis

For each species, we estimated the species-specific telomere rate of change (TROC, bp yr −1 ) as the slope of the linear regression line for telomere length versus age [35]. We did this because while we tried to sample as widely across the age range for each species as possible (table 1), we were limited to what was available. Thus, we made a heuristic assumption of a linear rate of telomere loss across all ages as it allowed for unbiased TROC estimates due to differences in the range of the species lifespan sampled. While longitudinal studies report that telomere loss rate is faster early in life [40], interestingly, we did not observe clear deviations from this assumption (figure 1) but as more telomere comparative data become available for telomere length (TL), nonlinear relationships should be explored. Differences among species' mean telomere length and TROC were assessed in a linear model. The relationships between telomere length or TROC with maximum lifespan among species were assessed using comparative analysis.

In comparative analysis, shared ancestry violates the assumption of independence among data points, and including phylogenetic information statistically accounts for such dependence. To this end, phylogenetically corrected regressions were analysed using generalized least squares assuming a Brownian correlation structure in package ape in R [53]. The phylogeny we used was extracted from a bird supertree [54]. The most parsimonious tree was selected from a Bayesian distribution of a 1000 trees using BEAST [55]. Our models therefore do not incorporate phylogenetic uncertainty [56]. The phylogenetic signal—the amount of variation among species explained by shared ancestry [57]—in both TROC and mean TL was analysed in phytools [58]. Both Pagel's λ and Blomberg's K were tested (with only K presented, as both yielded similar results [59]). Trait evolution was plotted on the phylogeny in phytools [58] in R.

Our main hypothesis centred on if and how TROC is associated with maximum lifespan across species (hypothesis i). However, we also explored whether mean telomere length is associated with maximum lifespan across species, as is the case in mammals (hypothesis ii [11]). To this end, we fitted several models around these predictors, and also included body mass as a covariate [60]. Body mass (log10-transformed) was included because of the clear associations between body mass and longevity [61]. Mean telomere length and TROC were log10-transformed prior to analysis. Telomere length increased with age in two species (H. ostralegus and Oceanodroma leucorhoa), resulting in a negative TROC, and hence TROC values were log10 (x + 100)-transformed. Including or removing these covariates allowed us to investigate any sensitivity of our results to mean telomere length and body mass, but the results were similar in all models. We prefer the presentation of full models rather than performing model selection, since full models show the full range of the predictors investigated (both significant and nonsignificant) and are not as likely to inflate type I error [62].

In addition, our comparative dataset allowed us test two other specific hypothesis presented in the literature. First, we considered the telomeric brink hypothesis (hypothesis iii [42]), which suggests that telomere shortening is causal in the ageing process, and when telomeres become too short they cause death. Here, the prediction is that species with both short average telomere length and faster telomere shortening rates would also have shorter maximum lifespans. If such a relationship is present in the data this should result in an interaction between mean TL and TROC against maximum lifespan of a species. Second, we tested the prediction that species with longer telomeres may exhibit faster telomere shortening, which was suggested previously within-species (hypothesis iv [63]). We performed a phylogenetic regression of mean TL against TROC, where a positive relationship would suggest that those species with longer telomeres also show more rapid telomere loss.

3. Results

Mean telomere length and TROC differed substantially among species (figure 1, linear model: both p < 0.0001). Some species show very sharp declines in telomere length with age, while others even increase with age. Interestingly, TROC was strongly associated with maximum lifespan across species, with species of shorter maximum lifespan having greater TROC (figure 2a, hypothesis i). This relationship is robust to the inclusion of the different covariates tested (table 2). There was no relationship between mean telomere length of a species and maximum lifespan (table 2 and figure 2b, hypothesis ii). Additionally, the interaction between mean telomere length and TROC against species-specific maximum lifespan was not significant (hypothesis iii, interaction: −0.60 ± 0.55 p = 0.29, also without the inclusion of body mass, p = 0.21), suggesting that TROC is not more determinative of species' maximum lifespan in species with short absolute telomere lengths. A species' TROC was also not associated with the species’ mean telomere length (table 2 and figure 2c, hypothesis iv).

Figure 2. Maximum observed lifespan as a function of (a) telomere rate of change (TROC) and (b) mean telomere length in 19 bird species. (c) Mean telomere length plotted against TROC in 19 bird species. The dashed lines represent the regressions from the phylogenetic regressions without any other covariates included.

Table 2. Tests for the first two hypotheses (see §§1,2) in a phylogenetically corrected regression (* indicates p < 0.05, ±denotes s.e.). Models were tested with and without body mass (log10-transformed) as a covariate, and with and without mean telomere length as a covariate. Telomere rate of change (TROC) is the only significant and strong predictor of maximum lifespan variation among species, with greater telomere loss rates associating with shorter maximum lifespan (hypothesis i). TROC was not related to mean telomere length of a species (hypothesis ii).

A phylogenetic signal was detected for TROC (K = 1.21, p < 0.001 figure 3), but not for absolute telomere length (K = 0.35, p = 0.60). Note that when K is larger than 1 it indicates that phylogenetically related species are more similar than expected under Brownian motion [59].

Figure 3. Trait evolution of telomere rate of change (TROC) mapped to the phylogeny in 19 bird species. Colours indicate different levels of the trait value (transformed values were used for mapping, but linear values are depicted for illustrative purposes in the legend). TROC shows a strong phylogenetic signal and the major families or clades of species which were included in this analysis show similar rates of telomere loss with age.

4. Discussion

Our study confirms previous reports that species with greater TROC have shorter maximum lifespans (hypothesis i [35–37]). By contrast, among the species sampled here, mean telomere length was not associated with longevity (hypothesis ii). Because TROC so accurately explains maximum lifespan (table 2, figure 2), the physiological causes of telomere shortening, or the ability to maintain telomeres, could be partially responsible for the different lifespans observed across species. In support of this suggestion, TROC shows a strong phylogenetic signal, whereas absolute telomere length does not, although it does vary widely among species. TROC, in contrast to absolute telomere length, therefore, appears evolutionarily conserved and selected within bird families. We acknowledge that this inference is less firm when phylogenies are small, but the difference in phylogenetic signal is striking when considering the potential biological significance of telomere length loss compared to absolute telomere length.

The pattern reported here between TROC and lifespan may be caused in part by selective disappearance, in which certain phenotypes are preferentially removed from a population [41]. Since our study is cross-sectional in nature due to the lack of long-term study populations and the long lifespans of some species, the relationship between TROC and lifespan could be a result of selective loss of short-lived individuals from the environment. This selective disappearance of particular individuals [41,64]—those with short telomeres—can cloud the relationship between telomere loss and age in a cross-sectional context [65]. For example, the positive relationship between telomere length and age seen in Leach's storm petrels (O. leucorhoa) is most likely due to the longest-lived individuals starting with the longest telomeres and variation in telomere length decreasing with age owing to the selective disappearance of individuals with short telomeres [65]. It is also possible that selective disappearance is responsible for the positive relationship seen in the Eurasian oystercatcher (H. ostralegus). It is difficult to translate a cross-sectional pattern within species to within-individual processes for that species. Future longitudinal studies will allow us to distinguish between population differences that result from the removal of certain phenotypes earlier in life than others. At the moment, longitudinal studies of ageing in free-living populations are rare, but are needed because of their greater power to identify age-related changes compared to cross-sectional studies [31,66,67]. However, in a comparative context, cross-sectional studies can still broadly inform us about the biology of ageing and lifespan, though we lose the ability to firmly conclude that these patterns are resulting from processes within individuals.

The degree to which selective disappearance differs among species can be caused by differences in extrinsic mortality rates and differences in how age-related mortality is influenced by telomere biology. Classical evolutionary theories of ageing predict that extrinsic mortality levels should be inversely correlated with evolved lifespan [68]. In other words, short-lived species generally face higher risk of death due to predation, starvation or accident. Given this, one interpretation of our results is that if there is selective disappearance of individuals with short telomeres, this pattern may be partially concealed in populations of species where individuals are removed from the population due to random processes regardless of their condition. This is not to say that selective disappearance based on telomere length is not occurring in short-lived species, only that it may be more readily obscured. Conversely, if telomere erosion does in fact increase mortality risks, then one might expect that this would be more evident in long-lived species with lower rates of extrinsic mortality where functional senescence is more easily observed. If extrinsic mortality differences are a major driver of selective disappearance based on telomere length this may explain why the species displaying patterns that most closely resemble selective disappearance are two of the longest-lived species we studied, the storm petrel and oystercatcher.

However, regardless of species differences in extrinsic mortality, the degree to which telomere biology affects intrinsic ageing processes may also differ across species, thereby affecting selective disappearance. In other words, some species might have a stronger relationship between telomere biology and survival prospects compared to others. Hence, selective disappearance could be stronger in species that are long-lived and for whom telomere biology is more important. Therefore, selective disappearance may be due to both differential association of intrinsic mortality with telomere length and differences in extrinsic mortality that reduce the importance of telomere length in determining mortality at the population level. Regardless of either of these causes, the comparative pattern we find suggests that as the longevity of a species increases, telomere biology becomes increasingly important. Moving forward, more longitudinal data are sorely needed to disentangle the possible scenarios outlined above that can result in the cross-sectional relationship we report here. Such efforts will allow us to understand more details of the deteriorative process of senescence in general [67], and how selective disappearance occurs in species of differing lifespan in particular.

While selective disappearance may be partially responsible for the pattern we observe between TROC and lifespan, another possibility is that telomere erosion is a potential mechanism underlying the evolution of lifespan in birds, with short-lived birds losing more telomeres each year compared to long-lived birds. A recent meta-analysis of 14 avian species reported that the rate of telomere loss is correlated with maximum lifespan estimated from a composite measure of life-history traits [36]. Another recent study, using existing data from longitudinal studies in bird species, confirmed a negative relationship between the rate of telomere shortening and maximum longevity [37]. Both of these recent studies provide additional support for the hypothesis that telomere attrition is correlated with interspecific rates of ageing. The underlying mechanisms responsible for this relationship are still unknown, and selective disappearance, physiological mechanisms that ameliorate telomere loss, or some combination of the two may be at play.

In search for physiological mechanisms that underlie the evolution of lifespan, comparative analyses have revealed that across avian and mammalian species, those species with longer lifespans also have cells that are both more resistant to external stressors [69] and have lower rates of mitochondrial free radical generation [70]. One possible physiological explanation for the different rates of telomere loss in the avian species in our study is that they also had different levels of oxidative stress. Oxidative stress can increase single-stranded breaks in telomeric regions of DNA that can cause telomere shortening during DNA replication due to a proposed pausing of the replication fork [13], though this work is mainly based on in vitro evidence, and whether it holds in vivo has recently been questioned [71]. Nevertheless, species with higher levels of oxidative stress may experience more rapid telomere shortening. Interestingly, this may be due in part to free radicals' preferential damage to the guanine-rich telomeric sequence in comparison to other regions of DNA [72]. This may allow telomeres to act as a free radical magnet, soaking up damaging free radicals while protecting coding regions of the chromosome. Accordingly, if long-lived species experience lower levels of oxidative stress, then their telomeres may have less exposure to the damaging attack of free radicals.

Another mechanism that could underlie the relationship between the rate of telomere loss and maximum lifespan of species is differential levels of telomerase expression. In mammals, a phylogenetically controlled comparison in rodents found that telomerase activity relates to body mass [33]. Another comparative mammalian study confirmed the relationship between telomerase activity and body mass and also found an inverse relationship between mean telomere length and lifespan [11]. The study authors suggest that short telomeres along with telomerase suppression are necessary for the evolution of large body size and longevity in mammals, apparently as a cancer suppression mechanism. Interestingly, while mean telomere length varies by an order of magnitude in the species we studied (5–50 kb), we do not find a relationship between mean telomere length and lifespan (hypothesis ii). The lack of a phylogenetic signal may suggest that telomere length evolves rapidly within a lineage, though this scenario is unlikely as the analysis includes some closely related species (Aphelocoma and Tachycineta species pairs, for example). Another possibility is that telomere shortening is more important on specific chromosomes, but that selection is neutral for other chromosomes. Every chromosome end contributes to the mean telomere length of a species with the in-gel TRF assay, and methods such as the single-telomere length analysis could better determine the role that telomere length on single chromosome ends might play in lifespan.

Previous studies in birds actually showed higher levels of telomerase expression in cells of species with longer maximum lifespans [73]. This suggests that longer-lived avian species may have evolved mechanisms that promote telomere maintenance through telomerase expression. Compared to mammals as a whole, birds have reduced body mass and relatively long lives. And the idea was proposed that a smaller body size and fewer cells may allow birds to have higher telomerase activity and longer telomeres without the associated high cancer risk [40]. While this idea is intriguing, more phylogenetically controlled work on interspecific variation in telomerase expression as well as comparative experimental work on telomerase activity in tumour cells is critically needed [74,75].

The direct role of telomeres in organismal ageing is still unclear [22]. While telomere shortening may directly contribute to senescence [76], there also may be no causal relationship between telomere biology and ageing [22,77]. Recently, the telomeric brink hypothesis (hypothesis iii [42]) postulated a causal role for telomere shortening in shaping longevity. Specifically, the authors note that individuals who are born with short telomeres also have shorter telomeres in adulthood, which results in a greater degree of cellular senescence and mortality, potentially through atherosclerosis. We tested this hypothesis in our comparative study of birds, but did not find a stronger relationship between TROC and maximum lifespan for species with short average telomere lengths. Given that atherosclerosis is relatively common in avian species [78], this raises the possibility that telomeres may not play a causal role in ageing, but rather serve as a biomarker of other ageing-related physiology [18,69]. However, data exploring telomere loss over an individual's lifespan within a species are necessary to better evaluate this hypothesis. Another recent hypothesis in telomere biology is that long telomeres shorten more quickly than short telomeres, possibly because longer telomeres are more sensitive to telomere-damage events or even random DNA damage (hypothesis iv [26,79]). To our knowledge this has only been evaluated within species, but across species we did not find support for this hypothesis. But, while we might expect species with longer average telomere length to lose telomeres more rapidly we cannot account for differences in telomere maintenance mechanisms among species, which could cloud these relationships.

5. Conclusion

Identifying the evolutionary importance of how different physiological mechanisms cause the ageing process can be aided by comparative studies. The results of this study clearly show that rates of telomere loss are strongly associated with species longevity in birds. In addition, this relationship is evolutionarily conserved and selected within bird families. Avian species that are better able to maintain their telomeres or conversely for which telomeres are more strongly associated with survival, causing selective disappearance, may experience lower rates of cellular and organismal ageing. While this study highlights the connection between telomere biology and the pace of life among species, we need to continue to uncover how within-individual processes that affect ageing in an ever-changing environmental backdrop relate to telomere loss.

Ethics

All procedures were conducted with approval of the appropriate local Animal Ethics Committee.


2 Methods

We quantified the telomere lengths of mature male painted dragons (Ctenophorus pictus) of four different head color morphs (Figure 1). The dragons are ideal models for this research as they are short-lived and age rapidly (surviving

1 year in the wild, Olsson, Healey, Wapstra et al., 2007 ), making it more likely that attempts to manipulate telomeres will be detected (Olsson, Tobler, Healey, Perrin, & Wilson, 2012 ).

This work was performed under the Animal Ethics permit 2013/6050 at the University of Sydney. Mature (

9 months old) male lizards were caught by noose or hand at Yathong Nature Reserve, NSW, Australia (145°35′E 32°35′S) and taken to holding facilities at the University of Sydney in October 2014 where they were housed for the duration of the experiments. Adult males were housed individually in opaque plastic tubs (330 × 520 × 360 mm) with sandy substrate and exposed to a 12-h light: 12-h dark light cycle. The males were housed in three different rooms for logistical reasons, but morphs were randomly distributed among rooms. The lizards were fed mealworms and crickets, dusted with calcium and multivitamins, to satiation every day and the cages were misted with water once a day. Heat lamps and ceramic hides were provided to allow the lizards to thermoregulate to their preferred body temperature (36°C M.O., unpublished data obtained from cloacal temperature readings in the wild).

2.1 Blood sampling

150 μl) was sampled using a capillary tube prior to and after the completion of the treatment by rupturing the vena angularis (in the corner of the mouth) with the tip of a syringe. Collected blood was mixed with heparin and centrifuged, the plasma removed and the remaining cells resuspended with 200 μl of PBS. One milliliter of RNAlater (Sigma Aldrich, Australia) was added and the diluted blood cell solution for qPCR stored immediately at −80°C.

2.2 Quantifying telomere length: qPCR

To analyze telomere length, we first purified DNA from 50 μl of the using a DNeasy Blood and Tissue Kit (Qiagen, Australia), according to the manufacturer's instructions. RNase A (Qiagen, Australia) was added at the recommended concentration. The DNA concentration (ng/μl) of each sample was measured in duplicate using a Pherastar FS (BMG, Labtech, Germany) and aliquots diluted to 10 ng/μl using the AE buffer provided in the DNA extraction kit. Samples were then stored at −30°C. Telomere length was measured using real-time quantitative PCR (qPCR) using SensiMix SYBR No-ROX Kit (Bioline, Sydney, Australia). The telomere primers used were Telb1 (5′-CGGTTTGTTTGGGTTTGGGTTTGGGTTTGGGTTTGGGTT-3′) and Telb2 (5′-GGCTTGCCTTACCCTTACCCTTACCCTTACCCTTACCCT-3′) (Criscuolo et al., 2009 ). The 18S ribosomal RNA (18S) gene (92 bp amplicon in Anolis) was selected as the reference gene as it had previously been validated in a reptile (Plot, Criscuolo, Zahn, & Georges, 2012 ). The primer sequences used were 18S-F (5′-GAGGTGAAATTCTTGGACCGG-3′) and 18S-R (5′-CGAACCTCCGACTTTCGTTCT-3′). The melt curves produced for both telomere and 18S after amplification by qPCR displayed a single peak, indicating specific amplification of the DNA sequence. The qPCR was performed in a final volume of 20 μl for both telomeres and 18S. DNA of 10 ng was used per reaction, and the primers were used at a concentration of 250 nM. SensiMix SYBR No-ROX Master Mix (Bioline, Australia) of 11.25 μl was added per reaction and MgCl2 was added for a reaction concentration of 1.7 mM. Reactions were run in triplicate for each sample. Amplifications were carried out in a Rotor-Gene 6000 thermocycler (Qiagen, Australia) using an initial Taq activation step at 95°C for 10 min, and a total of 40 cycles of 95°C for 15 s, 60°C for 15 s and 72°C for 15 s. A melt curve was created after each run over the temperature range of 60–95°C to ensure no non-specific product amplification. No-template control reactions were run in triplicate for each primer set during every qPCR run to ensure no contamination. Standard curves were created, using the blood of a randomly selected lizard, for both telomeres and 18S to ensure consistent rates of amplification over a wide range of DNA concentrations. The reaction was considered consistent when a straight line with an R 2 exceeding .985 could be fitted to the values obtained. Threefold serial dilutions were created, starting at a concentration of 26.67 ng/μl down to 0.037 ng/μl, with seven different concentrations in total (Supplemental Figure S1) giving a linear dynamic range of 0.037–26.67 ng/μl. The efficiency of the telomere amplification was 1.05 and the efficiency of the 18S amplification was 0.96. All samples fell within the concentration range generated by the standard curve. In all runs, the no-template controls had a Cq value at least 10 times higher than the lowest sample measured, indicating that contaminant DNA made up a maximum of approximately 0.001% of the original DNA concentration. LinRegPCR 2015.2 (Ruijter et al., 2009 Tuomi, Voorbraak, Jones, & Ruijter, 2010 ) was used to analyze the qPCR data. LinRegPCR calculates individual starting concentrations based on the average efficiency of an amplicon, the baseline fluorescence, and the threshold cycle values. The starting concentrations of telomere (T) and control gene (S 18S) were used to determine the relative telomere length with the calculation T/S. Telomeres and the control gene were assessed in separate runs. The mean inter-assay coefficients of variation for qPCR runs for telomere (n = 13) and 18S (n = 13) amplification were 16.73% and 38.21%, respectively, calculated using a reference sample at 10 ng/μl that was included in all runs. Due to the high inter-assay coefficient for the 18S runs, we checked individual values and found an outlier which inflated the inter-assay coefficient. When this outlier was removed, the mean inter-assay coefficient was 23.52%. To test whether this value was more reliable, we calculated a new mean inter-assay coefficient based on a standard from the standard curve that was included in all runs at a concentration of 8.89 ng/μl. Under these conditions, the mean inter-assay coefficient was 22.30%. This suggests that the actual mean inter-assay coefficient for 18S was approximately 23%. The mean (±SD) intra-assay coefficients of variation for telomere and 18S runs were 14.63 ± 0.1% and 12.52 ± 0.09%, respectively. Mean (±SD) amplification efficiencies generated by LinRegPCR across telomere and 18S qPCR runs were 1.89 ± 0.04 and 1.98 ± 0.01, respectively. LinRegPCR efficiency values can be compared with the efficiency obtained by a standard curve by subtracting 1. The difference in efficiency between the standard curve and LinRegPCR method is likely due to the manual setting of the Cq in the standard curve method.

2.3 Statistical analysis

For analysis of morph-specific effects on telomere length, the qPCR data were first entered into a mixed model analysis in Proc MIXED SAS 9.4 (SAS Institute) using telomere length in February (end date of experiment) as the response variable, the predictor fixed factors morph (yellow, orange, red, and blue), and bib (bibbed and not bibbed), and with the room in which a lizard was held (numbered 1–3) as a random factor. When “room” was not significant in a likelihood ratio test (χ 2 < 1, p > .9), we performed the corresponding analysis in Proc GLM (in order to obtain R 2 values for our analyses). Telomere length in December was used as a covariate, which controlled telomere length at the onset of the experimental period. This also constrains the analysis to non-interstitial telomeres, since it effectively measures the change in telomere length (under the assumption that interstitial telomeres do not change during the experimental period). That said, we do not suggest that interstitial telomeres are irrelevant to viability and other fitness effects under selection in association with telomere dynamics. The data obtained from this experiment will be archived in Dryad.


The possible role of telomeres in the short life span of the bay scallop, Argopecten irradians irradians (Lamarck 1819).

Several theories of aging have been developed over the years. Pearl (1928 cited in Carlson & Riley 1998) saw an inverse relationship between longevity and metabolism, with the harmful results of an excessive metabolic rate leading to an earlier demise. The free radical theory, put forth by Harman (1956), states that highly reactive products of metabolism such as reactive oxygen species (ROS) can be damaging to tissues. Another theory assumes the constant onslaught of environmental insults (e.g., solar radiation, toxic effects of ingested materials, and the like) cause DNA damage that outstrips the body's ability to repair it (Carlson & Riley 1998). A fourth suggests that genes producing favorable characteristics in early life may become harmful after reproduction (Rose 1991). Recent progress in the field of molecular biology has led to new methods of studying aging in eukaryotes. Telomeres, nucleotide repeats found at the end of chromosomes, have been shown to act as a "mitotic" clock that may define the life span of a species (Blackburn 1991, Harley 1991, Wright & Shay 2005). With each round of cell replication, a number of telomeres are lost (Blackburn 1991, Levy et al. 1992).

Senescence may occur when telomere length reaches a critical length, inducing changes that resemble DNA breaks and subsequent checkpoint arrest (Zou et al. 2002). Although the exact connection between telomere loss and cell senescence has not been elucidated, it is known that the cell cycle control protein, p53, is located near telomeres (Levine et al. 1993), and other studies have shown a relationship between telomere loss and aging (Kulju & Lehman 1995, Vaziri & Benchimol 1996, Whikehart et al. 2000). There have been some studies that correlate longer telomeres with longer survival. Experiments have shown that reconstituting active telomerase in cells yields elongated telomeres and an extended life span in human tissue culture (Bodnar et al. 1998, Vaziri & Benchimol 1998). Joeng et al. (2004), overexpressed the telomere-binding protein, HRP-1 in the worm, Caenorhabditis elegans, resulting in longer telomeres and subsequent longer life. Tree swallows (Tachycineta bicolor) with longer telomeres have been shown to live longer than those with shorter telomeres (Haussmann et al. 2005). Conversely, experiments that prevented telomerase from adding more telomeres to the ends of chromosomes resulted in telomere shortening and cell death (Herbert et al. 1999).

Aging in many invertebrates and some vertebrates does not exist (i.e., these animals simply continue to grow and reproduce until environmental conditions, disease or predation finally end their lives). Klapper et al. (1998a) discovered the enzyme telomerase in all tissues of the lobster, Homarus americanus. Similarly, Koziol et al. (1998) found high levels of telomerase in the tissues of sponges studied. The rainbow trout, Oncorhynchus mykiss is considered "immortal" because of its similar telomerase distribution (Klapper et al. 1998b). Normally, species with a defined life span do not exhibit telomerase activity in their somatic cells, and the subsequent telomere loss determines the length of their lives (Harley et al. 1992, Shay & Wright 2000). The ages of survival of some commercial bivalve molluscs range from less than two years in the genus Argopecten, (Belding 1910) to the deep-ocean clam, Artica islandica, living specimens of which have been aged at more than 150 y (Thompson et al. 1980, Ropes 1985). Another long-lived deep-sea bivalve, Tindaria callistifomis, attains an average length of 8.4 mm after 100 y and is sexually mature after 50 60 y (Turekian et al. 1975). It is currently unknown whether these molluscs have telomerase in all of their tissues as is seen in the lobster, or have a much lower cell turnover because of their deep-sea environment coupled with a sufficient complement of telomeres. Research has shown that representative molluscs demonstrate the same telomeric repeat, [(TTAGGG).sub.n] as is found in all vertebrates studied to date (Meyne et al. 1989). For example, the oyster. Crassostrea gigas, (Guo & Allen 1997), the bay scallop, Argopecten irradians irradians (Estabrooks 1999), the wedgeshell clam, Donax trunculus, (Plohl et al. 2002), and the land snails, Cantareus aspersus and C. mazzullii, (Vitturi et al. 2005), all share this same sequence. However, there have been no studies to date, that demonstrate a correlation of age and senescence with telomere length in the Mollusca. Epp et al. (1988) looked for possible causes of senescence in the bay scallop, A. irradians and postulated that there might be a connection between protein metabolism and neurosecretory cycles resulting in a possible neuroendocrine disturbance. Barber and Blake (1986), speculated that the high cost of reproduction may accelerate the process of senescence in the southern subspecies, A. irradians concentrieus. Bricelj et al. (1987b) concluded that senescence in the second year bay scallop was not linked to the metabolic costs of an upcoming second reproductive effort.

The current study compared the telomeres of two closely related species of Argopecten. namely A. purpuratus, capable of surviving 7 10 y or more (DiSalvo et al. 1984, Alarcon & Wolff 1991), and A. irradians irradians, with a life span less than 2 y. The genus, Argopecten, split off around 19 million years ago during the middle Miocene (Waller 1969), whereas Argopecten purpuratus, found off the coasts of Peru and Chile, arose approximately 6 million years ago and became separated from the Atlantic stock with the closing of the Atlantic-Pacific connection at the end of the Miocene, whereas Argopecten irradians split off around 1.5 million years ago during the Pleistocene when rising water levels produced the bays and sounds of today (Waller 1969). After first confirming that the telomeric sequence of A. purpuratus was the same as that for A. irradians [(TTAGGG).sub.n] their telomere lengths were compared as a possible explanation for the difference in their respective life spans. In addition, the subtelomeric sequences of A. irradians were compared with those of Placopecten magellanicus, a species with the modal number of n = 19, that arose some 65 million years ago. It has been shown that the more recent the species, the more numerous the subtelomeric patches that are sites of high rates of interchromosomal recombination and with both positive, and negative outcomes (Rocco et al. 2001, Linardopoulou et al. 2005).

An argument is presented for the possible benefits of extending the life span of Argopecten irradians through telomere elongation as an additional approach to population recovery and maintenance of adequate stocking densities in this species.

Three to four bay scallops belonging to each cohort of A. irradians were collected at approximately six-month intervals from the same locale in Nantucket Harbor, MA. This ensured that all comparisons of telomere lengths were of scallops approximately one year apart in age. Specimens of Argopecten purpuratus were obtained through the generosity of Karin Lohrmann Sheffield from the Universidad Catolica del Norte in Chile and Louis DiSalvo from Chile. DNA was extracted from the digestive gland, kidney, adductor muscle, heart and gill tissues either by the method using DNAzol (Molecular Research, Inc., Cincinnati, OH), see Estabrooks (1999). or a slight modification of the method of Sokolov (2000) that yields large amounts of DNA free of mucopolysaccharides often found in molluscan tissue. In the latter method, small amounts of tissue samples, 50-70 [micro]g in 1.5-mL tubes, were homogenized briefly in 1.0 mL of lysing reagent (50 mM Tris-HCl, 100 mM NaCl, 10 mM EDTA, 1% sodium dodecyl sulphate (SDS), 0.2-0.4 mg/mL proteinase K) and incubated for 2 h at 55[degrees]C. 100 [micro]l of saturated KCl were added, mixed and the tubes placed on ice for 5 min. After spinning at x 14,000g, the supernate was transferred to new tubes and extracted twice with an equal volume of phenol:chloroform:isoamyl alcohol (25:24:1) using the Eppendorpf Phase-Lok system (Westbury, NY). The supernate was then transferred to a new tube and the DNA precipitated with an equal amount of 100% ethanol, washed twice in 70% ethanol, drained and reconstituted with TE buffer, pH 8.0. RNase A was added at 0.2 mg/ml and the DNA was then quantified at 260/280 nm with a Beckman DU7 spectrophotometer. Known amounts of DNA were cut with the restriction enzymes RsaI (Sigma-Aldrich, St. Louis, MO) and HinfI (Amersham Biosciences, Piscataway, NJ) according to manufacturers recommendations and electrophoresed on 0.8% agarose gels for approximately one hour at 105V. Each electrophoretic run consisted of DNA from year 1 and year 2 bay scallops taken at different times from Nantucket Harbor. Scallops in each age group were from the same area of the Harbor and hence from the same cohort, helping to ensure a 12 mo difference in age at each sampling time. The DNA was then transferred to nylon membranes by Southern blotting in 0.4M NaOH, rinsed briefly in 2x SSC and air-dried between sheets of 3MM filter paper. The membrane was then baked at 80[degrees]C for 30 min to fix the DNA to the membrane. Telomeres were detected using the chemiluminescent procedure described in Estabrooks (1999). Average telomere lengths were estimated using molecular weight markers (kb) included in each run.

Telomeres are expressed as telomere restriction fragments (TRF) because they are cut from the ends of the chromosome by restriction enzymes at a point below the innermost telomeric (TTAGGG) segment and may contain a small amount of nontelomeric DNA.

Figure 1 follows the decline in the telomere lengths of the digestive gland of Argopecten irradians as compared that of A. purpuratus. A year 1 A. irradians is from 0-12 mo at which point it is generally sexually mature and ready to spawn. Semiquantatative measurements determined that the average telomere lengths for A. purpuratus (2-y-old) to be approximately 9.4 kb. A representative telomere run for A. irradians at approximately 6 mo old yielded a length of 2.3 kb, and 0.75 kb for A. irradians at 18 mo. Samples of year 1 and year 2 cohorts of A. irradians were collected together from Nantucket harbor. This ensured that all comparisons of telomere lengths were of scallops approximately one year apart in age and all demonstrated a large drop in telomere lengths between the two cohorts. At no time did a year 2 scallop have telomeres longer than those of any year 1, and only the digestive gland tissue demonstrated the dramatic differences in length, most likely because of a much higher rate of cellular turnover.

A year 2 scallop is from 13 24 mo, but will generally not survive to complete a second year. The TRFs of a bay scallop entering its 3rd year (24 mo plus) are shown in Figure 2, comparing those of its cardiac tissue with those of the digestive gland. The heart muscle does not divide, and so demonstrates a uniform group of long (high MW) telomeres that do not shorten with age, whereas the digestive gland has very few telomeres of any length. The level of telomeres in the adductor muscles of two cohorts of A. irradians also remains the same as would be expected, as muscle tissue generally does not replicate (Fig. 3).

Figure 4 compares the digestive gland telomeres of the long-lived deep-sea scallop, Placopecten magellanicus (4-y-old) with those of a year 1 bay scallop. Note the scarcity of subtelomeric sequences in P. magellanicus, giving support to the theory that the more recent the species, the more subtelomeric sequences may be found (Meyne et al. 1990, Rocco et al. 2001).

The chromosome number of the extant species of scallops has the modal haploid number, n = 19, whereas the bay scallops, A. irradians and A. purpuratus have n = 16 (Wada 1978, Gajardo et al. 2002). This reduction in chromosome number may be because of Robertsonian fusion, wherein telocentric chromosomes fused, reducing the total chromosome number (Wang & Guo 2004). They also noted that A. irradians incurred chromosome arm losses of nearly 50% as compared with the modal number 2 n = 38, and suggested that an ancestral bivalve may have become tetraploid at some point in time as an explanation of how Argopecten was able to sustain such extensive chromosome losses and still survive.

Most of the scallop species studied to date have the modal chromosome number n = 19, and live longer than the bay scallop (Beaumont & Zouros 1991). It is suggested that key DNA material involved in longevity may have been lost in some scallop species along the way, leaving A. irradians with just enough to survive a single reproductive effort.

As further evidence, Rocco et al. (2001) demonstrated an increasing number of subtelomeric sequences of TTAGGG in several Chondrichthian species at different stages of evolution that resulted in chromosome reduction. This may be the case with the bay scallop. The more recent the species, the more internal telomeric sequences are seen (Meyne et al. 1990). Interestingly, no interstitial TTAGGG sequences were found in the oyster, Crassostrea gigas (n = 10), a more primitive bivalve (Guo and Allen 1997), or in C. angulata (Cross et al. 2005), whereas the current study (Fig. 4) was able to demonstrate their presence in the more recent A. irradians (1.5 million years old) and only slightly in the deep-sea scallop, P. magellanicus (65 million years old). Zou et al. (2002) found in a species of deer, internal telomere sequences (subtelomeric) are also sites of fragility that may be leftover remnants of Robertsonian fusion that may actually contribute to further chromosome instability. In humans, approximately 50% of all subtelomeric sequences were generated after the chimpanzee/human divergence (Linardopoulou et al. 2005).

A. irradians is often labeled semelparous (Barber & Blake 1986, Bricelj et al. 1987a, Estabrooks 1999, Tettelbach et al. 1999 among others), but there appears to be no selective advantage to the bay scallop having such a short life span. A semelparous existence is usually found when there is a need to put all of one's energy into a single reproductive effort with death resulting shortly thereafter. This is seen, for example, in the Pacific salmon, Oncorhynchus spp., in which all of the energy needed for the reproductive process is brought in from the ocean in the form of stored nutrients to be used upstream (Morbey et al. 2005). To the contrary, the bay scallop enters into its second year post spawning by doubling in weight from September to the end of November (Bricelj et al. 1987b, Tettelbach et al. 2002). Lipids, carbohydrates, and protein are stored for the winter, and many may survive to at least initiate gametogenesis in the early spring, though most rarely survive to complete a second spawning (Belding 1910). Bricelj & Krause (1992) found a high rate of survival to a second reproductive season without completion (90% surviving until March). A. irradians appears to be caught by surprise, so to speak, let down by telomeric deletions, perhaps better labeled as a case of "iteroparous interruptus."

Darwinian theory would infer that this single life history trait (reproduction) has been optimized over time, but a more realistic view may suggest that the physiological constraints of living under species-specific opportunities might yield a strategy that, whereas not optimal, is still "good enough" (Darlington 1977, Tuomi et al. 1983). It is postulated that at least part of the key chromosomal material lost over time were telomeres. The energy-producing digestive gland with its rapidly dividing cells appears to be the key organ to lose sufficient telomeres to initiate senescence and the subsequent death of the bay scallop. In comparing the telomeres of the digestive gland of a two-year-old Argopecten purpuratus with those of the first and second year Argopecten irradians, it can be seen that A. purpuratus has longer telomeres (Fig. 1). Semiquantitative estimates show that a two-year-old A. purpuratus contains telomere restriction fragment (TRF) lengths that are much longer than those of a first year A. irradians. Figure 3 shows even shorter telomeres in the digestive gland of A. irradians entering its third year. The telomeres of the heart and adductor muscle do not shorten as these tissues do not generally undergo cell division (Fig. 3,4).

The gill and kidney also demonstrates telomere losses as would be expected, but not to the extent seen in the digestive gland (data not shown). This might be explained in the kidney, as this organ stores numerous excretion granules that are laid down layer by layer over an extended period to be finally expelled in the urinary tract (George et al. 1980, Morse 1987). This process would result in a slower cellular turnover and fewer lost telomeres.

Whereas the final time of death may be caused by variations in the environment coupled with the condition of individual scallops as suggested by Bricelj et al. (1987b), it is most likely that telomere loss is the coarse adjustment of the mitotic clock that determines the average life span of this species. The fine tuning may be modulated by several factors such as the initial number of telomeres, and the individual rate of cell turnover. Scallops in deeper and colder water may have a slower rate of metabolism, whereas those spawned later in the season will face different conditions at different points in their lives. Those that store fewer nutrients for the second upcoming winter may be in a weaker condition, as would be the case of those infected with parasites.

This could also account for the observations that senescence in scallops may be size related as well as age related (Gutsell 1930, Orensantz 1986, Bricelj et al. 1987b) with smaller scallops surviving longer for having had fewer cellular replications and thus fewer telomeres lost.

Bay scallop mortality can take place over a period of up to several months, usually beginning in early winter and running through to spring in the Northeast (Belding 1910, Bricelj et al. 1987b). It is believed that other theories of aging, including the accumulation of genetic errors over time, the damage caused by free radicals or an excessive metabolic rate may have their effects on defining the date of death of individual scallops, but not on the overall life span of the species.

It is proposed that A. irradians has the potential of a longer life, and only the unfortunate loss of key chromosomal material from an original tetraploid relict prevents it from doing so. A consequence of evolving into a species that has adapted to shallow bays may have been the loss of key telomeric sequences. It seems to make all of the preparations necessary for survival through the upcoming winter and beyond, but rarely lives to complete a second reproductive season. Genetic manipulation to increase the number of telomeres in A. irradians could possibly grant the bay scallop several additional reproductive opportunities that could help ameliorate the impact of some larval catastrophes in which there were no concomitant catastrophic losses of the spawner population.

This leads to the question that if an extended life span could be the answer to some of the problems facing A. irradians, then how successful is A. purpuratus in Chile? The industry there faces a different set of problems maintaining a successful threshold level of scallops in the wild, mainly that of illegal extraction. All year cohorts are considered seed and are taken indiscriminately by divers (Stotz & Gonzalez 1997), leaving aquaculture to account for the majority of scallop production today in Chile (von Brand et al. 2006). Genetic crosses of A. irradians irradians and A. purpuratus may have the possibility of extending the life span of the former. Both are cold water species with n = 16 and identical karyotypes, 5st (subtelocentric) and 11t (telocentric) (Gajardo et al. 2002, Wang & Guo 2004). Waller (1969), states that A. purpuratus, which is 4.5-5 million years older than A. irradians, has undergone little change over time. In addition, Chen et al. (1991) found 90% of crosses between Chlamys farreri (n = 19) and A. irradians (n = 16), survived up to 12 days. It seems reasonable to expect that a cross between A. irradians and A. purpuratus would fare better. The El Nino event of 1983, when ambient water temperatures dramatically increased, resulted in a sixty-fold increase in the normal scallop population. Wolff (1987), has suggested that A. purpuratus, a normally cold-water species, may have retained many of its warm water characteristics from the tropical/subtropical Miocene. If this true, a cross between A. purpuratus and A. irradians concentricus (Say) or A. irradians amplicostatus (Dahl), may yield similar results. Currently, research is underway to determine the viability of crosses between A. irradians irradians and A. purpuratus and to measure the resulting telomeres, if such crosses are successful.

The author thanks Dr. Sarah Oktay of the University of Massachusetts Field Station and Valerie Hall for their critical input. This research was supported by grants from the Nantucket Marine and Coastal Resources Department and the PADI Foundation.

Alarcon, E. & M. Wolff. 1991. Estudio biologico pesquero sobre el recurso de ostiones (Argopecten purpuratus) de bahia Tongoy durante el fenomeno El Nino 1982-83. Invest. Pesq. 32:167-173.

Barber, B. J. & N. J. Blake. 1986. Reproductive effort and cost in the bay scallop, Argopecten irradians concentricus. Int'l. J. of Invert. Reprod. and Develop. 10:51-57.

Beaumont, A. R. & E. Zouros. 1991. Genetics of scallops. In: S.E. Shumway, editor. Scallops: biology, ecology and aquaculture. Developments in aquaculture and fisheries science. Vol. 21. Amsterdam: Elsevier. pp. 585-623.

Belding, D. L. 1910. A report upon the scallop fishery of Massachusetts. The Commonwealth of Massachusetts. Boston. 150 pp.

Blackburn, E. H. 1991. Structure and function of telomeres. Nature 350:569-573.

Bodnar, A. G., M. Ouellette, M. Frolkis, S. E. Holt, C.-P. Chiu, G. B. Morin, C. B. Harley, J, W. Shay, S. Lichtsteiner & W. E. Wright. 1998. Extension of life-span by introduction of telomerase into normal human cells. Science 279:349-352.

Bricelj, V. M., J. Epp & R. E. Malouf. 1987a. Comparative physiology of young and old cohorts of bay scallop Argopecten irradians irradians (Lamarck): mortality, growth and oxygen consumption. J. Exp. Mar. Biol. Ecol. 112:73-91.

Bricelj, V. M., J. Epp & R. E. Malouf. 1987b. Intraspecific variation in reproductive and somatic growth cycles of bay scallops. Argopecten irradians. Mar. Ecol. Prog. Ser. 36:123-137.

Bricelj, V. M. & M. K. Krause. 1992. Resource allocation and population genetics of the bay scallop, Argopecten irradians irradians: effects of age and allozyme heterozygosity on reproductive output. Mar. Biol. 113:253-261.

Carlson, J. C. & J. C. M. Riley. 1998. A consideration of some notable aging theories. Exp. Gerontol. (1/2):127-134.

Chen, Q., J. Xiang, Y. Qin, B. Kou & H. Wang. 1991. Ref. in: Wang, Y. & X. Guo. 2004. Chromosomal rearrangement in Pectinidae revealed by rRNA loci and implications for bivalve evolution. Biol. Bull. 207:247-256.

Cross, I., E. Diaz. I. Sanchez & L. Rebordinos. 2005. Molecular and cytogenetic characterization of Crassostrea angulata chromosomes. Aquaculture 247:135-144.

Darlington, P. J., Jr. 1977. The cost of evolution and the imprecision of adaptation. Proc. Natl. Acad. Sci. USA 74:1647-1651.

DiSalvo, L. H., E. Alarcon, E. Martinez & E. Uribe. 1984. Progress in mass culture of Chlamys (Argopecten) purpuratus Lamarck (1819) with notes on its natural history. Rev. Chil. Hist. Nat. 57:34-45.

Epp, J., V. M. Bricelj & R. E. Malouf. 1988. Seasonal partitioning and utilization of energy reserves in two age classes of the bay scallop Argopecten irradians irradians (Lamarck). J. Exp. Mar. Biol. Ecol. 121:113-136.

Estabrooks, S. L. 1999. The telomeres of the bay scallop, Argopecten irradians irradians (Lamarck). J. Shellfish Res. 18:401-404.

Gajardo, G., M. Parragez & N. Colihueque. 2002. Karyotype analysis and chromosome banding of the Chilean-Peruvian scallop, Argopecten purpuratus (Lamarck, 1819). J. Shellfish Res. 21:585-590.

George, S. G., B. J. S. Pirie & T. L. Coombs. 1980. Isolation and elemental analysis of metal-rich granules from the kidney of the scallop, Pecten maximus (L.). J. Exp. Mar. Biol. Ecol. 42:143-156.

Guo, X. & S. K. Allen, Jr. 1997. Fluorescence in situ hybridization of vertebrate telomere sequence to chromosome ends of the Pacific oyster, Crassostrea gigas Thunberg. J. Shellfish Res. 16: 87-89.

Gutsell, J. S. 1930. Natural history of the bay scallop. Bull. U.S. Bur. Fish. 45:569-632. Harley, C.B. 1991. Telomere loss: mitotic clock or genetic time bomb? Murat. Res. 256:271-282.

Harley, C. B. 1991. Telomere loss: mitotic clock or genetic time bomb? Mutat. Res. 256:271-282.

Harley, C. B., H. Vaziri, C. M. Counter & R. C. Allsopp. 1992. The telomere hypothesis of cellular ageing. Exp. Gerontol. 27:375-382.

Harman, D. 1956. Aging-A theory based on free radical and radiation chemistry. J. Gerontol. 12:298-300.

Haussmann, M. F., D. W. Winkler & C. M. Vieck. 2005. Longer telomeres associated with higher survival in birds. Biol. Lett. 1:212-214.

Herbert, B. S., A. E. Pitts, S. I. Baker, S. E. Hamilton, W. E. Wright, J. W. Shay & D. R. Corey. 1999. Inhibition of human telomerase in immortal human ceils leads to progressive telomere shortening and cell death. Proc. Natl. Acad. Sci. USA 96:14276-14281.

Joeng, K. S., E. J. Song, K.-J. Lee & J. Lee. 2004. Long lifespan in worms with long telomeric DNA. Nat. Genet. 36:607-611.

Klapper, W., K. Heidorn, K. Kuhne, R. Parwaresch & G. Krupp. 1998a. Telomerase activity in 'immortal' fish. FEBS Lett. 434: 409-412.

Klapper, W., K. Kuhne, K. K. Singh, K. Heidorn, R. Parwaresch & G. Krupp. 1998b. Longevity of lobsters is linked to ubiquitous telomerase expression. FEBS Lett. 439:143-146.

Koziol, C., R. Borojevic, R. Steffen & W. E. G. Muller. 1998. Sponges (Porifera) model systems to study the shift from immortal to senescent somatic cells: the telomerase activity in somatic cells. Mech. Age. Dev. 100:107-120.

Kulju. K. S. & J. M. Lehman. 1995. Increased p53 protein associated with aging in human diploid fibroblasts. Exp. Cell Res. 217:336-345.

Levine, A. J., A. Chang & D. Dittmer. 1993. The p53 tumor suppressor gene. J. Lab. Clin. Med. 123:817-823.

Levy, M. Z., R. C. Allsopp, A. B. Futcher. C. W. Greider & C. B. Harley. 1992. Telomere end-replication problem and cell aging. J. Mol. Biol. 225:951-960.

Linardopoulou, E. V., E. M. Williams, Y. Fan, C. Friedman, J. M. Young & B. J. Trask. 2005. Human subtelomeres are hot spots of interchromosomal recombination and segmental duplication. Nature 437:94-100.

Meyne, J., R. J. Baker, H. H. Hobart, T. C. Hsu, O. A. Ryder, O. G. Ward, J. E. Wiley, D. H. Wurster-Hill T. L. Yates & R. K. Moyzis. 1990. Distribution of non-telomeric sites of the [(TTAGGG).sub.n] telomeric sequence in vertebrate chromosomes. Chromosoma 99: 3-10.

Meyne, R. K., R. L. Ratliffe & R. K. Moyzis. 1989. Conservation of the human telomere sequence (TTAGGG)n among vertebrates. Proc. Natl. Acad. Sci. USA 86:7049-7053.

Morbey, Y. E., C. E. Brassil & A. P. Hendry. 2005. Rapid senescence in Pacific salmon. The Amer. Naturalist 166:556-568.

Morse, M. P. 1987. Comparative functional morphology of the bivalve excretory system. Am. Zool. 27:737-746.

Orensantz, J. M. 1986. Size, environment, and density: the regulation of a scallop stock and its management implications. Can. Spec. Publ. fish. Aquat. Sci. 92:195-227.

Plohl, M., E. Prats, A. Martinez-Lage, A. Gonzalez-Tizon, J. Mendez & L. Cornudella. 2002. Telomeric localization of the vertebrate-type hexamer repeat, [(TTAGGG).sub.n], in the wedgeshell clam Donax trunculus and other marine invertebrate genomes. J. Biol. Chem. 277:19839-19846.

Rocco, L., D. Costagliola & V. Stingo. 2001. [(TTAGGG).sub.n] telomeric sequence in selachian chromosomes. Heredity 87:583-588.

Ropes, J. W. 1985. Modern methods used to age oceanic bivalves. Nautilus 99:53-57.

Rose, M. R. 1991. Evolutionary biology of aging. New York: Oxford University Press.

Shay, J. W. & W. E. Wright. 2000. Hayflick. His limit, and cellular ageing. Nat. Rev. Mol. Cell Biol. 1:72-76.

Sokolov, E. P. 2000. An improved method for DNA isolation from mucopolysaccharide-rich mollluscan tissues. J. Moll. Stud. 66:573-575.

Stotz, W. B. & S. A. Gonzalez. 1997. Abundance, growth and production of the sea scallop Argopecten purpuratus (Lamarck 1819): bases for sustainable exploitation of natural scallop beds in north-central Chile. Fish. Res. 32:173-183.

Tettelbach, S. T., C. F. Smith, R. Smolowitz, K. Tetrault & S. Dumais. 1999. Evidence for fall spawning of northern bay scallops Argopecten irradians irradians (Lamarck 1819) in New York. J. Shellfish Res. 18:47-58.

Tettelbach, S. K., C. F. Smith, P. Wenczel & E. Decort. 2002. Reproduction of hatchery- reared and transplanted wild bay scallops, Argopecten irradians irradians, relative to natural populations. Aquacult. Internat. 10:279-296.

Thompson, I., D. S. Jones & D. Dreibelbis. 1980. Annual internal growth banding and life history of the ocean quahog Artica islandica (Mollusca:Bivalvia). Mar. Biol. 57:25-34.

Tuomi, J., T. Hakala & E. Haukioja. 1983. Alternative concepts of reproductive effort, costs of reproduction, and selection in life-history evolution. Am. Zool. 23:25-34.

Turekian, K. K., J. K. Cochran, D. P. Kharkar, R. M. Cerrato, J. R. Vaisnys, H. L. Sanders, J. F. Grassle & J. A. Allen. 1975. Slow growth rate of a deep-sea clam determined by [sup.228]Ra chronology. Proc. Natl. Acad. Sci. USA 72:2829-2832.

Vaziri, H. & S. Benchimol. 1996. From telomere loss to p53 induction and activation of a DNA-damage pathway at senescence: the telomere loss/DNA damage model of cell aging. Exp. Gerontol. 31:295-301.

Vaziri, H. & S. Benchimol. 1998. Reconstitution of telomerase activity in normal human cells leads to elongation of telomeres and extended replicative life span. Curr. Biol. 8:279-282.

Vitturi, R., A. Libertini, L. Sineo, I. Sparacio, A. Lannino, A. Gregorini & M. Colomba. 2005. Cytogenetics of the land snails Cantareus aspersus and C. mazzullii (Mollusca:Gastropoda:Pulmonata). Mieron 36:351-357.

von Brand, E., G. E. Merino, A. Abarca & W. Stolz. 2006. Scallops: biology, ecology and aquaculture. In: S.E. Shumway & G.J. Parsons, editors. Amsterdam: Elsevier B.V 1293-1314.

Wada, K. 1978. Chromosome karyotypes of three bivalves: the oysters, Isognomon alatus and Pinctada imbricata, and the bay scallop, Argopecten irradians irradians. Biol. Bull. 155:235-245.

Waller, T. R. 1969. The evolution of the Argopecten gibbus stock (Mollusca:Bivalvia), with emphasis on the Tertiary and Quaternary species of eastern North America. J. Paleon. 4, part II of II, supplement No. 5.

Wang, Y. & X. Guo. 2004. Chromosomal rearrangement in Pectinidae revealed by rRNA loci and implications for bivalve evolution. Biol. Bull. 207:247-256.

Whikehart, D. R., S. J. Register, Q. Chang & B. Montgomery. 2000. Relationship of telomeres and p53 in aging bovine corneal endothelial cell cultures. Invest. Ophthal. Visual Sci. 41:1070-1075.

Wolff, M. 1987. Population dynamics of the Peruvian scallop Argopecten purpuratus during the El Nino phenomenon of 1983. Can. J. Fish. Aquat. Sci. 44:1684-1691.

Wright, W. E. & J. W. Shay. 2005. Telomere biology in aging and cancer. J. Am. Ger. Soc. 53:S292-S294.

Zou, Y., X. Yi, W. E. Wright & J. W. Shay. 2002. Human telomerase can immortalize Indian Muntjac cells. Exp. Cell Res. 281: 63-76.

STEPHEN L. ESTABROOKS * Nantucket Marine Laboratory, 0 Easton Street, Nantucket, Massachusetts 02554


Acknowledgements

We thank Johanna Borlid for assistance with molecular work, Benjamin Goh, Krista Woodward, and Geordia Rigter for keeping fish stocks and sampling, Britt Wassmur for sharing beta-actin primer sequences, and the Editor and two anonymous reviewers for constructive comments. The study was supported by Formas (21.5/2002-1037 and 215-2009-463, DB 2008-383, JIJ), funding from Oscar och Lili Lamms Minne (FO2009-0007 and FO2012-0039, AP), Carl Tryggers Stiftelse (CTS 09:294, AP), as well as by the Canadian Regulatory System for Biotechnology (RHD).


Results

Age determination by GLG counts

Chronological ages estimated for 24 tusks by counts of GLGs ranged from 5 to 69 years (Table 1) and were used to determine correlations between age, rTL and body morphometrics.

ID Stranding information Body information Age* (year)
Date (M/D/Y) Location Sex Length (cm) Body (kg)
307 24 January 2010 Libong Island, Trang Female 277 358 54
300 5 October 2009 Samran Beach, Trang Female 224 235 11
299 28 September 2009 Libong Island, Trang Female 281 335 63
292 2/2009 Ao Nang Beach, Krabi Female 282 305 41
291 1 February 2009 Ko Sarai Island, Satun Female 263 317 67
285 26 July 2008 Map Ta Phut, Rayong Female 200 24
283 9 April 2005 Ko Lanta Yai Island, Krabi Female 300 38
272 30 August 2007 Sattahip, Chonburi Male 245 13
258 12 August 2006 Bang Sare, Chonburi Female 231 200 12
243 26 December 2004 Ko Pa Yung Island, Phangnga Male 275 310 23
236 1 June 2004 Ko Klang Island, Phangnga Female 250 69
234 22 April 2004 Ko Yao Island, Phangnga Female 256.5 297 27
143 28 November 2001 Libong Island, Trang Female 199.5 142 5
129 6 December 2000 Khao Mai Kaeo, Trang Female 225 258 43
098 14 October 1998 Kram, Rayong Male 214 180 16
084 3 August 1998 Wichit, Phuket Male 219 184 8
078 19 May 1998 Lamae, Chumphon Female 231 151 34
058 6 January 1997 Libong Island, Trang Male 250 245 15
057 2 January 1997 Libong Island, Trang Female 258 281 14
048 3 March 1996 Suk Samran, Ranong Female 271 293 43
047 25 January 1996 Ko Yao Island, Phangnga Male 221 143 6
038 21 June 1995 Libong Island, Trang Male 257 250 14
036 31 March 1995 Bo Hin, Trang Female 273 272 34
016 11 May 1993 Muang Krabi, Krabi Male 254 16

Determination of variability of rTL measurement

In the present study, we tested the variability of rTL measurement using real-time PCR and if it was caused by differences in experimental design or biological factors as shown in Table 2. Our results revealed lower CVs obtained from intra-individual samples across both age groups compared to inter-individual values. These results suggest that although the experiment error had a relatively large variability, the biological factor (ages) appeared to be the main factor accounting for the variability in rTL.

Stage Intra-individual (%) Inter-individual (%)
Premature (<20 years) 38.97 47.02
Mature (≥20 years) 37.44 91.65
Overall mean 38.20 69.34

Correlation between age and GLG counts, rTL and body morphometrics

DNA was isolated from 24 dugong skin samples to measure telomere length by real-time PCR however, there was positive amplification for only 14 specimens (56%). But all 24 dugong tusks were aged based on counts of GLGs. Using logistic regression, positive correlations were found between age estimated by GLG counts and both body weight and length, with limiting values of approximately 309 kg and 266 cm, respectively (Figs. 3A and 3B Table 3). By contrast, the relationship between age by GLG counts and rTL fit a negative logistic curve, so increasing age was associated with decreasing telomere lengths (Fig. 3C Table 3). Negative correlations between rTL and body weight and length were observed, with an adjusted R 2 of 0.43 (p < 0.0001) and 0.51 (p < 0.0001), respectively, for the model logarithm function (Figs. 3D and 3E Table 3). Body weight and length were positively correlated with an adjusted R 2 of 0.78 (Fig. 3F Table 3). In addition, we evaluated the performance of the model obtained from age determination by rTL using a linear regression to test the correlation between the predicted age from rTL and age from GLG counts, resulting in an R 2 of 0.86 and residual standard error of 6.33 (Fig. S1).



Comments:

  1. Arat

    Bravo, this is just a great thought.

  2. Callum

    Bravo, what suitable words ..., the magnificent thought

  3. Abdul-Bari

    In my opinion it is obvious. I recommend to look for the answer to your question in google.com



Write a message