Molecular Ecology (2014) 23, 300–310
doi: 10.1111/mec.12602
Within the genome, long telomeres are more informative than short telomeres with respect to fitness components in a long-lived seabird C H R I S T I N A B A U C H , * P E T E R H . B E C K E R * and S I M O N V E R H U L S T † *Institute of Avian Research “Vogelwarte Helgoland”, An der Vogelwarte 21, 26386 Wilhelmshaven, Germany, †Behavioural Biology, Centre for Life Sciences, University of Groningen, PO Box 11103, 9700 CC Groningen, The Netherlands
Abstract Telomeres, DNA-protein structures at chromosome ends, shorten with age, and telomere length has been linked to age-related diseases and survival. In vitro studies revealed that the shortest telomeres trigger cell senescence, but whether the shortest telomeres are also the best biomarker of ageing is not known. We measured telomeres in erythrocytes of wild common terns Sterna hirundo using terminal restriction fragment analysis. This yields a distribution of telomere lengths for each sample, and we investigated how different telomere subpopulations (percentiles) varied in their relation to age and fitness proxies. Longer telomeres within a genome lost more base pairs with age and were better predictors of survival than shorter telomeres. Likewise, fitness proxies such as arrival date at the breeding grounds and reproductive success were best predicted by telomere length at the higher percentiles. Our finding that longer telomeres within a genome predict fitness components better than the shorter telomeres indicates that they are a more informative ageing biomarker. This finding contrasts with the fact that cell senescence is triggered by the shortest telomeres. We suggest that this paradox arises, because longer telomeres lose more base pairs per unit time and thus better reflect the various forms of stress that accelerate telomere shortening, and that telomeres primarily function as biomarker because their shortening reflects cumulative effects of various stressors rather than reflecting telomere-induced cell senescence. Keywords: birds, ageing, life history, lifestyle, reproduction, selective disappearance Received 5 August 2013; revision received 22 October 2013; accepted 30 October 2013
Introduction Telomeres, specialized and highly conserved DNAprotein structures, form the ends of eukaryotic chromosomes. In telomerase-negative cells, telomeres shorten gradually due to incomplete replication (Olovnikov 1996) and, enhanced by factors such as oxidative stress (von Zglinicki 2002), stochastic damage on DNA or changes in telomere-associated proteins (Karlseder et al. 2002; d’Adda di Fagagna et al. 2003). Shortened and dysfunctional telomeres lead to chromosome instability, Correspondence: Christina Bauch, Fax: +49 4421 968955; E-mail: [email protected] and Simon Verhulst, Fax: +31 50 363 5205; E-mail: [email protected]
changes in gene expression and finally cellular senescence or apoptosis (Rodier et al. 2005). Longitudinal analyses have shown that telomere length decreases with age in mammals (Br€ ummendorf et al. 2002; Aviv et al. 2009) and birds (Hall et al. 2004; Salomons et al. 2009; Barrett et al. 2013; Bauch et al. 2013), but there is a large variation between individuals of the same age (Benetos et al. 2013). In addition to a strong genetic basis (Broer et al. 2013), this variation has been linked to differences in lifestyle (Monaghan & Haussmann 2006; Bakaysa et al. 2007). Short telomeres have been associated with premature ageing syndromes (Chang et al. 2004; Hofer et al. 2005) and reduced survival in humans (Boonekamp et al. 2013), nematodes (Joeng et al. 2004) and birds (Haussmann et al. 2005; Bize et al. 2009; © 2013 John Wiley & Sons Ltd
L O N G T E L O M E R E S B E S T P R E D I C T F I T N E S S P R O X I E S 301 Salomons et al. 2009; Angelier et al. 2013; Barrett et al. 2013). Thus, evidence is accumulating that telomere length is a biomarker of health and remaining lifespan. Telomere length varies between chromosomes and cells within samples (Lansdorp et al. 1996), and hence, each sample is characterized by a telomere distribution rather than a single estimate. The commonly used measures are the average of the distribution and the proportion of short telomeres (e.g. Canela et al. 2007), but other characterizations are also possible and it is not well known how the performance of different measures compares as biomarker of health and remaining lifespan. In vitro studies show that replicative senescence and apoptosis are triggered by the short telomeres in the cell (Martens et al. 2000; Hemann et al. 2001; Zou et al. 2004; Capper et al. 2007; Bendix et al. 2010). One can therefore hypothesize that short telomeres may be the better predictors of survival, health status or fitness components than the average telomere length. There are, however, few tests of this hypothesis, and results are inconsistent. Kimura et al. (2008) evaluated, additionally to average telomere length, the average of the shorter 50% and 25% of the telomere distribution and found an indication that the shorter telomeres better predict mortality in humans, but there were no significant differences and telomeres longer than the median were not considered. Vera et al. (2012) investigated the relationship between longevity and telomere length in mice, comparing average telomere length and the proportion of ‘short’ telomeres. Neither measure was a significant predictor of lifespan, while telomere dynamics in form of shortening rate and subsequent increase in proportion of short telomeres did predict remaining lifespan, and the two measures of telomere shortening predicted lifespan approximately equally well. In contrast, Salomons et al. (2009) investigated the entire telomere distribution subdivided into percentiles and found that the longest telomeres best predicted survival in free-living jackdaws Corvus monedula. Thus, despite the evidence from in vitro studies that the shortest telomeres in a cell trigger cell senescence, little is still known about the relative predictive value of long versus short telomeres in a genome. Resolving whether the long or short (or average) telomere length is the best predictor of health and lifespan is of interest, because the result can be used to improve the value of telomere length as biomarker in epidemiological studies. Furthermore, it may shed light on the mechanistic connection between telomere dynamics and health and lifespan. We therefore investigated the predictive value of different characterizations of the telomere length distribution measuring telomeres with terminal (telomere) restriction fragment analysis in erythrocytes of common terns Sterna hirundo, a long-lived, © 2013 John Wiley & Sons Ltd
migratory and philopatric seabird species, in a colony continuously studied since 1984. Birds are transpondermarked, which allows recording of annual survival, arrival at the breeding site (a parameter of phenotypic quality in birds) and individual identification on the nest (Becker et al. 2001, 2008). We previously showed that in this population, individuals that performed best with respect to multiple fitness-related traits had shorter average telomere length and this effect was more pronounced in males (Bauch et al. 2013). Here, we extend these analyses, first, by examining the same relationships for different characterizations (percentiles) of the telomere distribution and, second, by investigating the association between telomere length and adult survival. We show that the longer telomeres in a genome were a better predictor of all available fitness proxies.
Materials and methods Study species and data collection We studied common terns breeding in Wilhelmshaven, Germany (53°30′40′’N, 08°06′19′’E). The common tern is a long-lived (annual adult survival 90%; Szostek & Becker 2012), migratory seabird species. Fledglings in this colony have been ringed since 1984 and additionally fitted with subcutaneous, passive transponders since 1992 (TROVAN ID-100; Becker et al. 2001). Thus, by means of an antennae system on resting places and temporarily all nests at the colony site, attendance and hence survival of these highly philopatric birds is remotely and reliably recorded (Szostek & Becker 2012). Arrival date was defined as the first day of the year an individual was recorded at the colony site. Data on reproductive parameters (e.g. laying date, brood size) were collected during colony visits at intervals of 2–3 days during the entire breeding seasons. Birds were initially sexed by observation of copulations and since 1998 using molecular methods.
Blood sampling and telomere length analysis Birds were bled in 2007 and 2008 during incubation using haematophagous bugs (Dipetalogaster maxima, Triatominae, 3rd larvae stage; Becker et al. 2006; Arnold et al. 2008). Blood samples were stored in 2% EDTA buffer at 3–7 °C for up to 3 weeks before transferred into a 40% glycerol buffer and snap-frozen for permanent storage at 80 °C. We collected blood samples from 184 common terns (88 males and 96 females) between 3 and 22 years of age (mean SEM: 8.92 0.25 (years)). 55 of these birds were sampled twice, in 2007 and 2008 (including one male with a minimum age of 17 and 18 years); 76 were breeding partners (38 pairs).
302 C . B A U C H E T A L . We measured class II telomeres (Delany et al. 2000) by terminal (telomere) restriction fragment analysis, without denaturing of DNA double strands, as previously described (Salomons et al. 2009; Bauch et al. 2013), which yields the genome-wide telomere distribution of telomere length of individuals. In contrast to telomere analysis using qPCR, our method requires larger DNA quantities and more time, but excludes interstitial telomeric repeats (Nakagawa et al. 2004), which occur in most bird species (Delany et al. 2000), including the common tern (Bauch, Becker & Wink, unpubl. data). Interstitial telomeric sequences, which do not shorten with age, can vary strongly between individuals of the same species and can make up a significant part of all telomeric repeats in a genome (Foote et al. 2013). For telomere analysis, we removed the glycerol buffer and washed the blood cells with 2% EDTA. DNA was isolated from erythrocytes (7 lL cells) using the CHEF Genomic DNA Plug kit (Bio-Rad, Hercules/CA, USA) and subsequently digested simultaneously with Hind III (60 U), Hinf I (30 U) and Msp I (60 U) in NEB2 buffer (New England Biolabs, Inc., Beverly/MA, USA) for 18 h at 37 °C. Two-thirds of the DNA from each sample was separated by pulsed field electrophoresis on a 0.8% nondenaturing agarose gel (Pulsed Field Certified Agarose, Bio-Rad, Hercules/CA, USA) for 22 h at 14 °C (3 V/cm, initial switch time 0.5 s, final switch time 7.0 s). For size calibration, we added 32P-labelled size ladders (1 kb DNA ladder, New England Biolabs, Inc., Ipswich/MA, USA; DNA Molecular Weight Marker XV, Roche Diagnostics, Basel, Switzerland). Gels were dried with a gel dryer (model 538, Bio-Rad, Hercules/CA, USA) and hybridized overnight at 37 °C with a 32 P-labelled oligonucleotide (5′-C3TA2-3′)4 that binds to the single-strand overhang of telomeres. Subsequent washing of the gel with 0.259 SSC buffer for 30 min at 37 °C removed unbound oligonucleotides. Finally, the gel was exposed to a phosphor screen (MS, PerkinElmer, Inc., Waltham/MA, USA) for 4 h to detect the radioactive signal, which was then visualized by a phosphor imager (CycloneTM Storage Phosphor System, PerkinElmer, Inc., Waltham/MA, USA). Telomere lengths distributions were quantified using Image J (version 1.38x, open source) as previously described (Bauch et al. 2013). Besides the measurement of average telomere length, we classified the telomere distribution of every sample into every 10th percentile from 10th to 90th. The lower limit of the measurement was lane-specifically set at the point with the lowest signal, with the shortest telomeres being located between 1500 and 2875 bp (mean SEM: 1800 20 (bp), n = 239). At the upper end, we chose to set a fixed limit at 30 kb as the background noise in the region of the longest telomeres was more variable. The selected
range represents the average telomere measurement, which yielded the highest intraindividual repeatability (Bauch et al. 2013). The background value was lanespecifically calculated as a mean outside the telomere region between 40 – 50 kb and subtracted from the optical density measurements. Between-gel coefficient of variation of a standard sample run on all gels was 2.0% (Bauch et al. 2013).
Statistics Data analysis was performed in JMP (version 7.0.1, SAS Institute Inc.). Unless stated otherwise, we applied linear mixed-effects models and tested specific predictions. To account for assay effects, gel identity was included as random effect. As telomere length shortens with age, as it has been shown previously for common terns (Haussmann et al. 2003; Bauch et al. 2013), age was always included as covariate. Further variables included when testing specific hypotheses were as follows: for the covariate ‘survival’, the number of years that birds survived after sampling in 2007 or 2008 was tracked until 2011 and 2012, respectively (up to 4 years for both cohorts to keep the analyses comparable). As in Bauch et al. (2013), the individual mean value from the sampling years 2007 and 2008 was included for the covariates ‘arrival date’ or ‘laying date’, and brood size at different brood ages from hatching to fledging (0, 10, 18, 26 days) as measures of reproductive success. Additionally, we calculated mean values of brood size at brood age day 10 for the individuals’ entire lives until blood sampling. At brood age day 10, the impact of environmental effects or year effects is comparatively low, whereas parental quality differences are already detectable (Bollinger 1994).
Results Telomere lengths and age in different parts of the telomere distribution We previously showed that, as in many other species, common terns’ average telomere length shortens with age (Bauch et al. 2013). Here, we compare telomere shortening rate between different telomere subsets (percentiles). To this end, we used the mean-centred telomere lengths (residuals) at the percentiles (increments of 10) as dependent variable and included bird ID as random effect (for individuals sampled more than once, we used only the first sample). We first tested the hypothesis that telomere shortening differed between percentiles (covariate), which was confirmed by a significant interaction percentile * age, revealing a higher telomere shortening in the range of longer telomeres © 2013 John Wiley & Sons Ltd
L O N G T E L O M E R E S B E S T P R E D I C T F I T N E S S P R O X I E S 303 (a) 0 Slope correlation telomere length [bp]/age [years]
(Table 1a). Next we tested whether this pattern differed between the sexes, by adding sex and its interaction with age and percentile to the model in Table 1a. This interaction was significant (Table 1b), indicating that the pattern of percentile-dependent telomere attrition differed between the sexes. To investigate the implications of this three-way interaction, we analysed the sexes separately (Table 1c and 1d), which revealed that in males the longer telomeres (higher percentiles) showed higher rates of telomere shortening, whereas in females telomere shortening rate was independent of percentile (Fig. 1a).
–20 –40 –60 –80 –100 –120
Males Females
–140 10
20
30
40
50
60
70
80
90
Average
70
80
90
Average
Percentiles
The rate of telomere shortening within individuals may differ from cross-sectional shortening rates if there is selective disappearance of individuals with long or short telomeres (e.g. Salomons et al. 2009). To compare between- and within-individual telomere shortening, we analysed the data for every 10th percentile separately, now including the second samples of birds that Table 1 Telomere length at the 10th, 20th, etc., to 90th percentile in relation to (a) age, (b) age and sex (coded as: 1 = male, 2 = female), and analysed separately for (c) males and (d) females. Bird ID and gel were included as random effects. n = 184 individuals (88 males and 96 females)
(a) Intercept Percentile Age Percentile * (b) Intercept Percentile Age Sex Age * sex Percentile * Percentile * Percentile * (c) Intercept Percentile Age Percentile * (d) Intercept Percentile Age Percentile *
t-ratio
Estimate
SEM
prob>|t|
age
347.98
doi: 10.1111/mec.12602
Within the genome, long telomeres are more informative than short telomeres with respect to fitness components in a long-lived seabird C H R I S T I N A B A U C H , * P E T E R H . B E C K E R * and S I M O N V E R H U L S T † *Institute of Avian Research “Vogelwarte Helgoland”, An der Vogelwarte 21, 26386 Wilhelmshaven, Germany, †Behavioural Biology, Centre for Life Sciences, University of Groningen, PO Box 11103, 9700 CC Groningen, The Netherlands
Abstract Telomeres, DNA-protein structures at chromosome ends, shorten with age, and telomere length has been linked to age-related diseases and survival. In vitro studies revealed that the shortest telomeres trigger cell senescence, but whether the shortest telomeres are also the best biomarker of ageing is not known. We measured telomeres in erythrocytes of wild common terns Sterna hirundo using terminal restriction fragment analysis. This yields a distribution of telomere lengths for each sample, and we investigated how different telomere subpopulations (percentiles) varied in their relation to age and fitness proxies. Longer telomeres within a genome lost more base pairs with age and were better predictors of survival than shorter telomeres. Likewise, fitness proxies such as arrival date at the breeding grounds and reproductive success were best predicted by telomere length at the higher percentiles. Our finding that longer telomeres within a genome predict fitness components better than the shorter telomeres indicates that they are a more informative ageing biomarker. This finding contrasts with the fact that cell senescence is triggered by the shortest telomeres. We suggest that this paradox arises, because longer telomeres lose more base pairs per unit time and thus better reflect the various forms of stress that accelerate telomere shortening, and that telomeres primarily function as biomarker because their shortening reflects cumulative effects of various stressors rather than reflecting telomere-induced cell senescence. Keywords: birds, ageing, life history, lifestyle, reproduction, selective disappearance Received 5 August 2013; revision received 22 October 2013; accepted 30 October 2013
Introduction Telomeres, specialized and highly conserved DNAprotein structures, form the ends of eukaryotic chromosomes. In telomerase-negative cells, telomeres shorten gradually due to incomplete replication (Olovnikov 1996) and, enhanced by factors such as oxidative stress (von Zglinicki 2002), stochastic damage on DNA or changes in telomere-associated proteins (Karlseder et al. 2002; d’Adda di Fagagna et al. 2003). Shortened and dysfunctional telomeres lead to chromosome instability, Correspondence: Christina Bauch, Fax: +49 4421 968955; E-mail: [email protected] and Simon Verhulst, Fax: +31 50 363 5205; E-mail: [email protected]
changes in gene expression and finally cellular senescence or apoptosis (Rodier et al. 2005). Longitudinal analyses have shown that telomere length decreases with age in mammals (Br€ ummendorf et al. 2002; Aviv et al. 2009) and birds (Hall et al. 2004; Salomons et al. 2009; Barrett et al. 2013; Bauch et al. 2013), but there is a large variation between individuals of the same age (Benetos et al. 2013). In addition to a strong genetic basis (Broer et al. 2013), this variation has been linked to differences in lifestyle (Monaghan & Haussmann 2006; Bakaysa et al. 2007). Short telomeres have been associated with premature ageing syndromes (Chang et al. 2004; Hofer et al. 2005) and reduced survival in humans (Boonekamp et al. 2013), nematodes (Joeng et al. 2004) and birds (Haussmann et al. 2005; Bize et al. 2009; © 2013 John Wiley & Sons Ltd
L O N G T E L O M E R E S B E S T P R E D I C T F I T N E S S P R O X I E S 301 Salomons et al. 2009; Angelier et al. 2013; Barrett et al. 2013). Thus, evidence is accumulating that telomere length is a biomarker of health and remaining lifespan. Telomere length varies between chromosomes and cells within samples (Lansdorp et al. 1996), and hence, each sample is characterized by a telomere distribution rather than a single estimate. The commonly used measures are the average of the distribution and the proportion of short telomeres (e.g. Canela et al. 2007), but other characterizations are also possible and it is not well known how the performance of different measures compares as biomarker of health and remaining lifespan. In vitro studies show that replicative senescence and apoptosis are triggered by the short telomeres in the cell (Martens et al. 2000; Hemann et al. 2001; Zou et al. 2004; Capper et al. 2007; Bendix et al. 2010). One can therefore hypothesize that short telomeres may be the better predictors of survival, health status or fitness components than the average telomere length. There are, however, few tests of this hypothesis, and results are inconsistent. Kimura et al. (2008) evaluated, additionally to average telomere length, the average of the shorter 50% and 25% of the telomere distribution and found an indication that the shorter telomeres better predict mortality in humans, but there were no significant differences and telomeres longer than the median were not considered. Vera et al. (2012) investigated the relationship between longevity and telomere length in mice, comparing average telomere length and the proportion of ‘short’ telomeres. Neither measure was a significant predictor of lifespan, while telomere dynamics in form of shortening rate and subsequent increase in proportion of short telomeres did predict remaining lifespan, and the two measures of telomere shortening predicted lifespan approximately equally well. In contrast, Salomons et al. (2009) investigated the entire telomere distribution subdivided into percentiles and found that the longest telomeres best predicted survival in free-living jackdaws Corvus monedula. Thus, despite the evidence from in vitro studies that the shortest telomeres in a cell trigger cell senescence, little is still known about the relative predictive value of long versus short telomeres in a genome. Resolving whether the long or short (or average) telomere length is the best predictor of health and lifespan is of interest, because the result can be used to improve the value of telomere length as biomarker in epidemiological studies. Furthermore, it may shed light on the mechanistic connection between telomere dynamics and health and lifespan. We therefore investigated the predictive value of different characterizations of the telomere length distribution measuring telomeres with terminal (telomere) restriction fragment analysis in erythrocytes of common terns Sterna hirundo, a long-lived, © 2013 John Wiley & Sons Ltd
migratory and philopatric seabird species, in a colony continuously studied since 1984. Birds are transpondermarked, which allows recording of annual survival, arrival at the breeding site (a parameter of phenotypic quality in birds) and individual identification on the nest (Becker et al. 2001, 2008). We previously showed that in this population, individuals that performed best with respect to multiple fitness-related traits had shorter average telomere length and this effect was more pronounced in males (Bauch et al. 2013). Here, we extend these analyses, first, by examining the same relationships for different characterizations (percentiles) of the telomere distribution and, second, by investigating the association between telomere length and adult survival. We show that the longer telomeres in a genome were a better predictor of all available fitness proxies.
Materials and methods Study species and data collection We studied common terns breeding in Wilhelmshaven, Germany (53°30′40′’N, 08°06′19′’E). The common tern is a long-lived (annual adult survival 90%; Szostek & Becker 2012), migratory seabird species. Fledglings in this colony have been ringed since 1984 and additionally fitted with subcutaneous, passive transponders since 1992 (TROVAN ID-100; Becker et al. 2001). Thus, by means of an antennae system on resting places and temporarily all nests at the colony site, attendance and hence survival of these highly philopatric birds is remotely and reliably recorded (Szostek & Becker 2012). Arrival date was defined as the first day of the year an individual was recorded at the colony site. Data on reproductive parameters (e.g. laying date, brood size) were collected during colony visits at intervals of 2–3 days during the entire breeding seasons. Birds were initially sexed by observation of copulations and since 1998 using molecular methods.
Blood sampling and telomere length analysis Birds were bled in 2007 and 2008 during incubation using haematophagous bugs (Dipetalogaster maxima, Triatominae, 3rd larvae stage; Becker et al. 2006; Arnold et al. 2008). Blood samples were stored in 2% EDTA buffer at 3–7 °C for up to 3 weeks before transferred into a 40% glycerol buffer and snap-frozen for permanent storage at 80 °C. We collected blood samples from 184 common terns (88 males and 96 females) between 3 and 22 years of age (mean SEM: 8.92 0.25 (years)). 55 of these birds were sampled twice, in 2007 and 2008 (including one male with a minimum age of 17 and 18 years); 76 were breeding partners (38 pairs).
302 C . B A U C H E T A L . We measured class II telomeres (Delany et al. 2000) by terminal (telomere) restriction fragment analysis, without denaturing of DNA double strands, as previously described (Salomons et al. 2009; Bauch et al. 2013), which yields the genome-wide telomere distribution of telomere length of individuals. In contrast to telomere analysis using qPCR, our method requires larger DNA quantities and more time, but excludes interstitial telomeric repeats (Nakagawa et al. 2004), which occur in most bird species (Delany et al. 2000), including the common tern (Bauch, Becker & Wink, unpubl. data). Interstitial telomeric sequences, which do not shorten with age, can vary strongly between individuals of the same species and can make up a significant part of all telomeric repeats in a genome (Foote et al. 2013). For telomere analysis, we removed the glycerol buffer and washed the blood cells with 2% EDTA. DNA was isolated from erythrocytes (7 lL cells) using the CHEF Genomic DNA Plug kit (Bio-Rad, Hercules/CA, USA) and subsequently digested simultaneously with Hind III (60 U), Hinf I (30 U) and Msp I (60 U) in NEB2 buffer (New England Biolabs, Inc., Beverly/MA, USA) for 18 h at 37 °C. Two-thirds of the DNA from each sample was separated by pulsed field electrophoresis on a 0.8% nondenaturing agarose gel (Pulsed Field Certified Agarose, Bio-Rad, Hercules/CA, USA) for 22 h at 14 °C (3 V/cm, initial switch time 0.5 s, final switch time 7.0 s). For size calibration, we added 32P-labelled size ladders (1 kb DNA ladder, New England Biolabs, Inc., Ipswich/MA, USA; DNA Molecular Weight Marker XV, Roche Diagnostics, Basel, Switzerland). Gels were dried with a gel dryer (model 538, Bio-Rad, Hercules/CA, USA) and hybridized overnight at 37 °C with a 32 P-labelled oligonucleotide (5′-C3TA2-3′)4 that binds to the single-strand overhang of telomeres. Subsequent washing of the gel with 0.259 SSC buffer for 30 min at 37 °C removed unbound oligonucleotides. Finally, the gel was exposed to a phosphor screen (MS, PerkinElmer, Inc., Waltham/MA, USA) for 4 h to detect the radioactive signal, which was then visualized by a phosphor imager (CycloneTM Storage Phosphor System, PerkinElmer, Inc., Waltham/MA, USA). Telomere lengths distributions were quantified using Image J (version 1.38x, open source) as previously described (Bauch et al. 2013). Besides the measurement of average telomere length, we classified the telomere distribution of every sample into every 10th percentile from 10th to 90th. The lower limit of the measurement was lane-specifically set at the point with the lowest signal, with the shortest telomeres being located between 1500 and 2875 bp (mean SEM: 1800 20 (bp), n = 239). At the upper end, we chose to set a fixed limit at 30 kb as the background noise in the region of the longest telomeres was more variable. The selected
range represents the average telomere measurement, which yielded the highest intraindividual repeatability (Bauch et al. 2013). The background value was lanespecifically calculated as a mean outside the telomere region between 40 – 50 kb and subtracted from the optical density measurements. Between-gel coefficient of variation of a standard sample run on all gels was 2.0% (Bauch et al. 2013).
Statistics Data analysis was performed in JMP (version 7.0.1, SAS Institute Inc.). Unless stated otherwise, we applied linear mixed-effects models and tested specific predictions. To account for assay effects, gel identity was included as random effect. As telomere length shortens with age, as it has been shown previously for common terns (Haussmann et al. 2003; Bauch et al. 2013), age was always included as covariate. Further variables included when testing specific hypotheses were as follows: for the covariate ‘survival’, the number of years that birds survived after sampling in 2007 or 2008 was tracked until 2011 and 2012, respectively (up to 4 years for both cohorts to keep the analyses comparable). As in Bauch et al. (2013), the individual mean value from the sampling years 2007 and 2008 was included for the covariates ‘arrival date’ or ‘laying date’, and brood size at different brood ages from hatching to fledging (0, 10, 18, 26 days) as measures of reproductive success. Additionally, we calculated mean values of brood size at brood age day 10 for the individuals’ entire lives until blood sampling. At brood age day 10, the impact of environmental effects or year effects is comparatively low, whereas parental quality differences are already detectable (Bollinger 1994).
Results Telomere lengths and age in different parts of the telomere distribution We previously showed that, as in many other species, common terns’ average telomere length shortens with age (Bauch et al. 2013). Here, we compare telomere shortening rate between different telomere subsets (percentiles). To this end, we used the mean-centred telomere lengths (residuals) at the percentiles (increments of 10) as dependent variable and included bird ID as random effect (for individuals sampled more than once, we used only the first sample). We first tested the hypothesis that telomere shortening differed between percentiles (covariate), which was confirmed by a significant interaction percentile * age, revealing a higher telomere shortening in the range of longer telomeres © 2013 John Wiley & Sons Ltd
L O N G T E L O M E R E S B E S T P R E D I C T F I T N E S S P R O X I E S 303 (a) 0 Slope correlation telomere length [bp]/age [years]
(Table 1a). Next we tested whether this pattern differed between the sexes, by adding sex and its interaction with age and percentile to the model in Table 1a. This interaction was significant (Table 1b), indicating that the pattern of percentile-dependent telomere attrition differed between the sexes. To investigate the implications of this three-way interaction, we analysed the sexes separately (Table 1c and 1d), which revealed that in males the longer telomeres (higher percentiles) showed higher rates of telomere shortening, whereas in females telomere shortening rate was independent of percentile (Fig. 1a).
–20 –40 –60 –80 –100 –120
Males Females
–140 10
20
30
40
50
60
70
80
90
Average
70
80
90
Average
Percentiles
The rate of telomere shortening within individuals may differ from cross-sectional shortening rates if there is selective disappearance of individuals with long or short telomeres (e.g. Salomons et al. 2009). To compare between- and within-individual telomere shortening, we analysed the data for every 10th percentile separately, now including the second samples of birds that Table 1 Telomere length at the 10th, 20th, etc., to 90th percentile in relation to (a) age, (b) age and sex (coded as: 1 = male, 2 = female), and analysed separately for (c) males and (d) females. Bird ID and gel were included as random effects. n = 184 individuals (88 males and 96 females)
(a) Intercept Percentile Age Percentile * (b) Intercept Percentile Age Sex Age * sex Percentile * Percentile * Percentile * (c) Intercept Percentile Age Percentile * (d) Intercept Percentile Age Percentile *
t-ratio
Estimate
SEM
prob>|t|
age
347.98
