Genetic variation in meat production related traits in reindeer ( Rangifer t . tarandus )

In reindeer husbandry, animal breeding could o%er tools for improving productivity by selection. "e traits a%ecting meat production e*ciency are primarily related to calf growth and survival, and to dam’s longevity and lifetime ability to raise heavy calves. Information on genetic (co)variation in these traits is scarce but needed in estimating the potential and e%ectiveness of selection as well as biological constraints. "e objectives of the study were to estimate these genetic parameters from the data of an experimental reindeer herd at Kutuharju (Inari, Finland). Heritabilities (h) and genetic correlations (r g ) among the average daily gain of calves (ADG), dams’ age at maturity, individual !tness (l ind ) and the cumulative sum of her calves’ weaning weight over seven years (WW7) were studied with AS-Reml software using the dataset from the experimental herd. "e pedigree included also sire information to allow the separation of the maternal e%ects. Direct and maternal heritabilities of ADG were moderate (0.24±0.09 and 0.18±0.05, respectively) and the directmaternal correlation was highly negative (-0.73±0.17). Indeed, selection on growth necessitates information on both calf and dam. "e genetic correlation of growth with birth date and birth weight could not be detected with the data. "e age at maturity and l ind had a small heritable component (0.07±0.12 and 0.10±0.06, respectively), whereas the heritability value of WW7 was 0.23±0.07. Reindeer herders’ empirical selection on calf ’s autumn weight favours fast growth (r g between growth and autumn weight = 0.35±0.24). Dam’s weight in her !rst autumn was strongly correlated with her lifetime production expressed by her individual !tness (r g = 0.71±0.23) and the cumulative sum of her calves’ weaning weight (r g = 0.63±0.12). Hence, the early information on the dam (her weight in her !rst autumn or her !rst calf ’s autumn weight) works as useful selection criteria for the traits related to lifetime production.


Introduction
In Fennoscandia, the pro tability in reindeer husbandry is based on meat production.e majority of meat is produced from calves born in spring and culled in autumn, and the num-ber and size of autumn calves determine herd productivity.From a herder's perspective the best females should mature early, have high reproductive rate and long productive life (Muuttoranta & Mäki-Tanila, 2011).In addition, they should take good care of their calves facilitating rapid growth and high survival.is will lead to the e ective use of the limited northern natural pasture areas for meat production.
Good calf growth over the summer implies heavy calf autumn weight.In practice herders use this autumn live weight as the main selection criterion (Muuttoranta & Mäki-Tanila, 2011), as fast-growing animals are thought to result in a rise in the number and quality of calves slaughtered (Muuttoranta & Mäki-Tanila, 2011).In reindeer, the heritability estimates (h 2 ) for growth are scarce: Varo (1972) reported an estimate of 0.6 based on very few animals, whereas Appel & Danell (unpubl., in Rönnegård & Danell, 2003) used a larger population and reported an estimate of 0.4.Later, Rönnegård & Danell (2003) estimated realized h 2 = 0.2 in the same population.All these estimates were based on dam-daughter pairs as none of the studies had information on sires.In reindeer, no maternal heritabilities in calf growth have been estimated and the e ect of dam's permanent environment has not been studied.
Reindeer herders consider good dams vital and dam quality is among the major selection criteria (Muuttoranta & Mäki-Tanila, 2011;2012).Quality of dam a ects productivity via her reproduction rate, maternal care and survival (Tanida et al., 1988;Brommer et al., 2002;Martinez et al., 2004a;2004b).Herders expect the dams to have a calf at foot in every autumn round-up (Muuttoranta & Mäki-Tanila, 2011).e good dams should have a long productive life which facilitates less need of recruitment animals, and hence a higher annual proportion for culling or meat production (Tanida et al., 1988;Martinez et al., 2004a).
Most of the females start calving at the age of two to four years (Rönnegård et al., 2002).Early maturation is related to a heavy body weight (see Reimers, 1983;Gaillard et al., 2000;Rönnegård et al., 2002, Weladji et al., 2008) and it is positively correlated with lifetime production (Weladji et al., 2006;2008).In addition, calving regularly and gaining experience in raising o spring improves the production performance of the reindeer female over the years (Weladji et al., 2006;2008).
e tness related traits are commonly analysed in evolutionary ecology and in some species there are heritability estimates for lifetime reproduction success (e.g.Merilä & Sheldon, 2000) and for individual tness (e.g.Blomquist, 2010).In livestock species lifetime production is gaining importance among breeders since it is related to pro tability and animal welfare (Flint & Woolliams, 2008).In beef cattle the lifetime production is measured for example by the cumulative weaning weight of calves of beef dams at the age of 7 years.Heritabilities are reported to range from 0.06 to 0.18 (Martinez et al., 2004a;2004b).Lifetime number of o spring weaned has been reported to have h 2 varying from 0.05 to 0.21 (in beef: Tanida et al., 1988;Martinez et al., 2004a;2004b;in sheep: Safari et al., 2005).Dam quality, measured as number and weight of weaned o spring, can be assessed towards the end of her productive life only.e animals for breeding purposes are mostly selected as calves, therefore early-life indicators of future production ability are needed.
Growth is a ected by genetic and environmental e ects.e genetic e ects can be separated to those due to the animal itself (direct) and those due to its dam (maternal).In addition to these genetic elements, both e ects also include an environmental element (Willham, 1963).e correlation between direct and maternal genetic e ects is often strongly negative in maternal traits, such as growth (e.g.Bijma, 2006), which, if biologically true, leads to a decline in maternal capacity if only direct growth is selected.Negative estimates have been questioned, as they may have arisen from the data structure (Meyer, 1992;Heydarpour et al., 2008), the xed e ects (Meyer, 1997;Bijma, 2006) or an ignored correlation between residual, direct and maternal e ects (Koerhuis & ompson, 1997;Bijma, 2006;Eaglen & Bijma, 2009).
For contemplating the prospects of selective breeding of meat production in reindeer, it is important to estimate the genetic variation and correlation in the relevant traits.Such analysis requires appropriate phenotypic records.Due to the semi-domesticated nature of reindeer husbandry, individual records are rare.e exceptionally detailed long term data including information on sires (Røed et al., 2002) from the Kutuharju experimental herd in northern Finland o ers a unique possibility to perform genetic research in reindeer.
e objectives of the study were to estimate the genetic variation in the traits a ecting meat production: calf growth (with related birth date, birth weight and weaning weight), females' age at maturity, and lifetime production of the dams without separating the e ects of natural selection and herders' culling; and also quantify the in uence of maternal e ects on the variation and genetic and phenotypic correlations.

Data description and subsets
e Finnish Game and Fisheries Research Institute's Reindeer Research Station maintains data from the Kutuharju experimental herd, located in Inari, Finland (69°N, 27°E), and owned and managed by the Finnish Reindeer Herders' Association.Paternities have been con rmed at the Norwegian School of Veterinary Science using DNA markers giving sire information for the calves born since 1997 (Røed et al., 2002).In addition, paternities were monitored using marking harnesses for calves born in 1987, 1988, and 1990-1993.e pedigree dataset contained 3320 animals, of which 101 were sires and 566 dams of the calves.e dataset has been described in Muuttoranta et al. (2013) Calf data e calves were weighed at birth in May (few born in June) and before selection in September.e weights are in kilograms and birth date is presented as days counting from 1 st of May.Average daily gain (ADG, in grams), was calculated by subtracting birth weight from calf autumn weight and dividing the di erence by calf age at autumn weighing.Calves lacking either birth or autumn weight record were excluded from the analyses for daily gain.e calf birth date is used to approximate the conception date (and gestation length).e other traits were calf birth date (BD), birth weight (BW) and calf autumn weight (AW1).e calf data consisted of 984 calves with known sires.
Dam data e productive life of a dam starts at maturation, which is measured in practice as the age at rst calving.Stillbirth or a calf that died during summer was counted as a calving.Females >3 y with no calving record were assumed to be late maturing and be included in the highest age class (7 years) for the age at maturation.Because data do not contain any culling reasons, there was no attempt to separate herders' voluntary culling from the involuntary mortality.A Leslie matrix (Leslie, 1945) is a square matrix of order equalling the number of age classes.e (i,j) th cell in the matrix indicates how many individuals will be in the age class i at the next time step for each individual in stage j (Leslie, 1945). is transition matrix is constructed for each individual and the individual tness (symbolised ind and used as a trait in the study) is the dominant eigenvalue of the matrix (McGraw & Caswell 1996, used in reindeer by Weladji et al. 2006).e values were estimated with statistical software R, version 3.0.1 (R Development Core Team, 2013).
e lifetime production of a dam was measured also by the cumulative sum of her calves' weaning weight over seven years (WW7) (cf.Martinez et al., 2004b).Stillborn calves and calves that died before weaning were excluded.
e trait re ects pre-weaning maternal care and is therefore in practice more important than the mere number of calves born.e weights of the calves were corrected by the birth year and calf sex.e calves with missing autumn weights were imputed by the autumn mean given the birth year and sex of the calf, instead more complicated methods (right-censored data or maximum likelihood) that could have been used.For the non-producing females the value of WW7 was 0 (similar to Kruuk et al. (2000), who estimated the number of o spring).
e dataset used for estimation of lifetime production contained 1165 females, born in 1964-2000.Only the complete age cohorts were used, as suggested by Tanida et al. (1988), hence the age cohorts with any living animals were excluded.is included the cohorts born after 2001.In the lifetime trait analysis, there were 555 female calves culled in their rst autumn and 600 females aged one year or older.Age at maturity was estimated from the dataset spanning up to the year 2007 with 1404 animals.To study the correlations among lifetime production traits and traits related to growth, we also analysed the autumn weight of the rst calf of the dam (CAW, as dam trait), dam's weight in her rst autumn (DAW) and dam's ADG when she was a calf (D_ADG).
In summary, the analysis covered the following traits: birth date, birth weight, calf 's average daily gain and calf 's autumn weight; dam's age at maturity, individual tness and the cumulative sum of dam's calves' weaning weight over seven years and the autumn weight of the rst calf of the dam.

Fixed e ects
For analysing the genetic variation in ADG we tested the model for the following xed e ects (classes in parentheses): sex of calves (male and female); year of birth (21 classes across the years 1987-2011); parity of dam (parities 1,2,…9, with the last class including also parities 10-14); sire age ( ages 1-2, 3, 4, 5, and 6-7); and calf at foot (a variable describing the reproductive status of dam, 0 or 1 for absence or presence of calf during the summer prior to calving).
For the lifetime production traits (age at maturity, ind and WW7) the xed e ect was birth year (36 year classes, across the years from 1964 to 2000).For the autumn weight of the dam's rst calf, the e ects of calf sex and birth year were tested.

Models for variance component estimation
Because of the complete pedigree information, at least for the information since 1987, the genetic variation in the population was studied analysing the di erences among individual animals using so called animal model.ree different animal models were used to estimate the genetic parameters in the ADG: 1) the only random e ect was additive (direct) genetic e ect of the animal, 2) direct (calf ) and maternal (dam) additive genetic e ects, and 3) direct and maternal additive genetic e ects and maternal permanent environmental e ect (dam).e di erent models were used to nd out the best model for random e ects.In the matrix notation the models were 1) y = Xb + Z a a +e, 2) y = Xb + Z a a + Z m m +e, and 3) y = Xb + Z a a + Z m m + Z c c + e, where y was a vector for observations; b was a vector for xed e ects; a, m and c were vectors for direct and maternal additive genetic e ects and maternal permanent environmental e ects, respectively; and e was random residual e ect.X, Z a , Z m and Z c were incidence matrices that relate the observations to b, a, m and c, respectively.
For the random e ects the assumptions were as follows: e I, where A was the additive relationship matrix between individuals and I the identity matrix.
e variance of direct additive genetic e ect was denoted as 2 a , maternal genetic variance as 2 m , maternal permanent environmental variance as 2 c , and covariance between animal's own and maternal genetic e ect as am .
In the lifetime production analyses, an animal model (for the dams) was used to estimate the variance components in the age at maturity, ind and WW7. e model and the related assumptions were similar to the model 1 for ADG.In addition to the genetic maternal effect, there may be non-genetic e ects in the dam causing advantage or disadvantage on calf 's performance.ere may be a correlation between the maternal permanent environmental e ect and the residual e ect in calf 's performance.Falconer (1965) proposed using a regression of calf 's record on dam's record in the trait for estimating such a correlation.In the multi-generation data the estimation is problematic and Koerhuis & ompson (1997) suggested including the respective covariance in the Willham's (1963) model.With a small amount of data such addition would not usually lead to satisfactory convergence.e existence of correlation between the residuals of the direct and maternal e ects could simply be detected by comparing the variance components of the animal model with direct and maternal e ects to the sire-maternal grandsire model (S-MGS), because the latter has no respective residual covariances (Eaglen & Bijma, 2009).
e sire-maternal grandsire model was y = Xb + Z s u s + Z mgs u mgs +e, where Z s and Z mgs were the incidence matrices that relate the observations to u s and u mgs (additive genetic e ects of sires and grandsires), respectively.Other terms are similar to those in the animal models.e direct and maternal variance components of the respective animal model were calculated ac-cording to Steinbock et al. (2003)  Fitting of the animal models were tested using the likelihood ratio test with a threshold value of 0.05 (Pinheiro & Bates, 2000).e bivariate analyses were conducted using the best model for ADG together with BD, BW and AW1.Age at maturity, ind and WW7 were all analysed in a bivariate context together with CAW, DAW and D_ADG.For presenting graphically the relationship between the lifetime traits and indicator traits, the predicted breeding values were computed.
e preliminary analyses were done using SAS EG®, version 4.2 (SAS Institute Inc., Cary, NC) and R, version 3.0.1 (R Development Core Team, 2013).e (co)variance components were computed by Restricted Maximum Likelihood (REML) method (Patterson & ompson, 1971).e signi cances of all genetic parameters with log likelihood tests were carried out with a statistical software package ASReml, release 3.0 (Gilmour et al., 2009).

Data
e data structure related to the calf traits has earlier been described in Muuttoranta et al. (2013).e summary statistics of the studied traits are given in Table 1.
e rst calf was lighter than those born in the later parities (CAW1<AW1).
e female calves which were selected for breeding were also lighter (DAW1<AW1) due to the sex di erence in the trait.e variation was moderate (CV about 0.15-0.25) in the growth and weight traits, higher (0.43-0.51) in birth date and age at maturation and very high (1.27-1.59) in the lifetime production traits.

Fixed e ects
e birth year was the major factor causing variation in all the calf traits.All the studied xed e ects were signi cant in the calf traits except the lactation in the previous year (Table 2).
e birth year was the only signi cant factor for the dam traits (results not shown).

Analyses of the genetic e ects
Single-trait analyses Calf traits ADG was analysed using three di erent animal models.e best t, according to the likelihood ratio test, proved to be the model including the direct and maternal genetic e ects, but the addition of the maternal permanent environmental e ect was not signi cant (P = 0.88, Table 3).Including a maternal e ect had some e ect on the variance estimates (model 1 vs. 2 and 3) but not on the direct heritability.e directmaternal genetic correlation (r am ) was strongly negative, although with a high standard error.
In addition to the animal model, a sire-ma-ternal grandsire model (S-MGS) was used to test the e ect of correlated residuals in ADG. e genetic parameters from the S-MGS analysis were very similar to the ones from the animal model analysis with slightly higher standard errors (Table 3).
Age at maturity and lifetime production e genetic parameters of the lifetime production traits are presented in Table 4.For age at maturity the heritability was smaller than its standard error.Both ind and WW7 had a signi cant heritability.However, the heritability for ind was only half of that of WW7; most likely because the latter has a scale with a much higher resolution.

Bivariate analyses
Calf traits ADG and the related calf traits were also analysed in a bivariate context using the best-tting  model with animal and maternal e ects.When analysed with BW and AW1, which are the components of ADG, the heritability estimates of ADG were similar or higher compared to the univariate analysis (Table 5).ere was a strong negative direct-maternal genetic correlation in ADG, as it was in the univariate analysis, whereas in the other traits the correlations were not signi cant.Table 6 gives the correlations among the calf traits.Birth date was the only direct e ect genetically related to growth rate.In birth weight dams with a low (or high) maternal e ect on birth weight had a high (or low) maternal effect on growth.e calves born heavy for nongenetic reasons would be expected to also grow fast.Fast growth leads to both phenotypically and genetically heavy calf autumn weight.
Age at maturity and lifetime production Age at maturity, ind and WW7 were analysed pairwise with the other traits.Gaining information from another trait resulted in an increment in the heritability of the studied trait (Table 7), which also meant obtaining signi cant estimates.However, the available amount of data relative to estimated parameters lead less often to signi cant correlations.e traits were correlated with each other and their early indicators.Genetic correlations implied that early maturation is positively correlated with high survival and reproduction ( ind ).
To demonstrate the relationship between the dams, we plotted the predicted individual genetic deviations from the population mean for dam's weight in her rst autumn (DAW1) versus the cumulative sum of her calves' weaning weight over seven years (WW7) (Figure 1).e correlation between the traits is favourable, resulting in the genetic gain in WW7 when using her autumn weight as calf as selection criterion.

Discussion
Pro table reindeer husbandry rests on good calf growth and productive dams.Genetic variation in the traits facilitates the successful selection.
e heritabilities in calf growth and dam productivity are moderate.However, the negative direct-maternal correlation in calf growth implies taking the dam properties into account in Table 3. Estimates of (co)variance components, and genetic and permanent maternal environment parameters (with their standard error in parentheses) for average daily gain (ADG) in the Kutuharju reindeer herd in 1987-2011.Models 1-3 have following random e ects: 1 = direct genetic, 2 = direct genetic + maternal genetic, and 3 = direct genetic + maternal genetic + maternal permanent environmental e ect.e best model among 1, 2 and 3 is indicated in bold.S-MGS is sire-maternal grandsire model.selection criteria.e early information on the dam (her weight in her rst autumn or her rst calf 's autumn weight) works as useful selection criteria for the traits related to lifetime production.

Data and traits
e Kutuharju reindeer data are unique, including detailed production records and pedigree, both being valuable in genetic research.
e quality and amount of data allowed uniand bivariate analyses and, in particular, the evaluation of maternal e ects.More data, like few thousand animals (cf.Kruuk et al., 2000;Martinez et al., 2004a;2004b), could have improved the power of the analyses, but using incomplete cohorts would have led to biased heritability estimates (Tanida et al., 1988;Martinez et al., 2004b).Modelling the incomplete cohorts by treating them as censored data (e.g.Ducrocq & Sölkner, 1994) or as missing data with EM-algorithm (e.g.Ducrocq & Besbes, 1993) may have in creased the power of the analysis.is was, however, not tried.e similarity in the (co)variance components estimated among other species supports the biologically sound results.e term ind was used to measure lifetime production, as it includes information on survival and reproduction and takes into account also the gaps in reproduction (Brommer et al., 2002).However, WW7 is most relevant from a meat production perspective, since it includes the calves' weights in autumn and allows  the detection of di erences among the dams.
e lifetime production traits are highly correlated (Table 7), with WW7 being more variable, probably because of more continuous nature in its variation compared to ind (cf.Brommer et al., 2002).
e time span in both of the lifetime traits was up to seven years of age.e cutting point was chosen as seven years to provide a sucient amount of data for the study and enable to detect the di erences among the prime-aged animals.
e study was dictated by production and arti cial selection purpose rather than survival and natural selection.Indeed, survival was primarily the outcome of herders' decisions with the objective of maximising meat production by selecting the female calves for reproduction.Individual tness could, in retrospect, be more appropriately called a selective advantage in production.Hence, it is not surprising that the correlations of the two lifetime production measurements with their early indicators are so high.

Fixed e ects
e annual variation in growth and reproduction (Table 2) is probably due to the weather conditions in spring and summer (e.g.review by Weladji & Holand, 2003).Also the management regime a ecting BD and BW has varied over the years.Growth of the calf depends on the dam weight in the autumn prior to calving (Lenvik et al., 2008;Weladji et al., 2003;Holand et al., 2004).Although the dam winter weight has been reported lower after rearing a calf the previous year (Rönnegård et al., 2002;Weladji et al., 2010), this had no e ect on the studied traits (Table 2). is supports the nd-Table 7. e heritabilities (the range of the estimates from the analyses with di erent traits, on the diagonal in bold), and genetic (upper diagonal) and phenotypic (lower diagonal) correlations for the age at the maturity, individual tness ( ind ), cumulative weaning weight of calves over 7 years (WW7), dam's weight in her rst autumn (DAW1), average daily gain of the dam (D_ADG) and her rst calf 's autumn weight (CAW1) in the Kutuharju reindeer herd in 1964-2000.ing of Weladji et al. (2008), how there is no evidence on reproduction costs in reindeer.
e parity e ect in the model takes into account that the very young and old females have smaller calves with lower growth rate (Lenvik et al., 1988;Rönnegård et al., 2002;Weladji et al., 2002;2003;2010).Further, the male calves grew faster than the female calves (Table 2), which is in line with the earlier studies (cf.Weladji et al., 2003), and calf sex was therefore included in the model.

Genetic variation
Growth e estimated heritability for growth is close to the realised estimates for autumn weight reported by Rönnegård & Danell (2003).e maternal heritability for growth was smaller than the direct one, a common feature for the growth traits (e.g.Heydarpour et al., 2008).Contrary to the birth weight of reindeer (Muuttoranta et al., 2013), the inclusion of the maternal e ect in the growth model did not cause reduction in the direct heritability, as also reported by Safari et al. (2005) for sheep.In reindeer this is due to the strongly negative direct-maternal genetic correlation in growth (Willham, 1972; Table 2).
Age at maturity Genetic variation of age at maturity was detected in the bivariate analysis.e low heritability value for age at maturity is common among the domestic animals (e.g.Bourdon & Brinks, 1982), and is therefore di cult to predict.In reindeer, the coarse scale makes the analysis even more challenging.In cattle, the age at rst calving is measured in days or months (e.g.Bourdon & Brinks, 1982), whereas in reindeer, due to seasonal breeding, the trait is measured in years.
Lifetime production e trait combining survival and reproduction rate, individual tness, ind , is heritable and hence selection results in genetic gain in dam quality (Table 7).We have found no previous study analysing genetic variation of ind for reindeer or other livestock species, although estimates on the other lifetime production traits exist (e.g.Kruuk et al., 2000).e found heritability for ind is smaller than that estimated in macaques by Blomquist (2010), who also noticed that the exclusion of the females, who never produced an o spring, caused lower heritabilities.is is also seen in our results: without the non-produced females ind was 0.01 (0.09) and WW7 was 0.12 (0.09).In cattle, the heritability for lifetime number of weaned calves has been estimated to be 0.16 (Tanida et al., 1988) and in red deer to 0 (Kruuk et al., 2000).Due to the coarse scale, the heritability value for ind is smaller than that of another lifetime production trait, lifetime reproductive success (Brommer et al., 2002), while WW7 is similar to it.In general, the low heritability values in the lifetime production traits, such as individual tness, could be due to a high amount of non-genetic variation that masks the additive genetic variance (Merilä & Sheldon, 2000;Brommer et al., 2002).Another explanation is that there is only little variation left in tness after selection (Falconer & Mackay, 1997).
In cattle, Martinez et al. (2004a) reported the heritability values of cumulative weaning weight for di erent dam ages (2-7) ranging from 0.06 to 0.16 (the highest values are for the longest reproductive life). is may originate from an increased opportunity to express their genetic potential.

Direct-maternal correlation
We found a strong negative direct-maternal correlation for growth of reindeer calves, commonly reported in livestock (e.g.Bijma, 2006).In contrast to what was found by Heydarpour et al. (2008), the proportion of dams with no record of the trait did not explain the high neg-ative value (results not shown).As another explanation, Bijma (2006) suggested estimating the residual covariance between the maternal and direct e ect.However, the computational challenges (Koerhuis & ompson, 1997) and the limited size of the data allows only speculation on the impact of correlated residuals on the strongly negative direct-maternal correlation.As applied by Eaglen & Bijma (2009), we compared the results from an animal model and a S-MGS model.We found similar estimates for the genetic covariance in both models and, therefore, little room for the covariance due to non-genetic e ects.
e negative direct-maternal correlation underlines the importance of including both maternal and direct e ect for calf growth in future selection criteria (Holand, 2007;Heydarpour et al., 2008).Selection on calf 's growth leads to heavy dams with possible reduced milk production and consequently poorer growth of their calves (e.g.Meyer, 1997).In a controlled situation this could be compensated with additional feeding.We suggest using large data: Among very numerous animals the ones with the least negative or even positive direct and maternal genetic e ects could be detected and selected for breeding.
Correlations between the traits When there is a favourable genetic correlation among the traits, selection on one would result in a desired gain also in the other traits.Indeed, growth has a favourable genetic correlation with survival, calf autumn weight, carcass weight, dressing percentage and early maturation (in cattle: Bourdon & Brinks, 1982;in reindeer: Petersson & Danell, 1993;in sheep: Safari et al., 2005).We found a positive genetic correlations between weight of a female in her rst autumn and both her ind and WW7, which are similar to the correlations in sheep between reproduction and growth (Safari et al., 2005).Hence, the calf autumn weight of an individual can be used as an early indicator trait for lifetime production.Apparently the herders are paying attention also to the dam information in selection.e selection intensity is very high, as only few females are needed as replacements.Knowing the negative direct-maternal genetic correlation it is retrospectively clear that the recruitment calves were from the favourable part in the distribution of selection candidate calves as the genetic correlation between the direct and maternal genetic e ect in calf autumn weight is negative.
Early maturation could serve as an another indicator of lifetime production (Table 7), as also reported by Weladji et al. (2006;2008), although some reindeer herders regard that relatively early calving, at the age of two years, leads to fewer calves in the subsequent years (Muuttoranta & Mäki-Tanila, 2012).A good environment physiologically favours the early maturation (Reimers, 2005); hence, winter feeding in Kutuharju may result in calving at too young age.However, the calves getting pregnant are very rare exceptions in Kutuharju (altogether six calves in data).Nevertheless, the di culty in nding the correlation with age at maturity and lifetime production should be noted.Brommer et al. (2002) reported about a positive e ect of age at maturity on ind , but not on lifetime reproduction success.Rönnegård et al. (2002) discussed how the late maturing females have low calving rate (despite their good condition) and relates this to reproductive inability.Indeed, a herder will cull a female with a limited productivity (Rönnegård et al., 2002;Muuttoranta & Mäki-Tanila, 2011).Martinez et al. (2004a) reported very high correlations (0.94-0.96) between the early and later production, hence the successful production in the rst parity can predict the lifetime performance.According to Weladji et al. (2008), calving at early age indicates phenotypically good lifetime production in reindeer.Our results suggest that selection of females according to their performance at the rst calving can be fruitful (Table 7).

Conclusions
Genetic variation in growth of calves and lifetime production of dams occurs.In a female reindeer the favourable correlations between calf growth, calf autumn weight, age at maturity and lifetime production make selection straight-forward, since when selecting for growth, also the other traits will be improved.Autumn weight of a female calf could therefore be used as an early indicator of her lifetime production.e production capacity of a dam can further be estimated using her rst calf 's weight.In improving growth of reindeer, the maternal e ects should be included in selection criteria due to the negative direct-maternal correlation to avoid undesirable consequences in dam quality.Intensive selection should be accompanied by parallel changes in management to compensate a possible impaired maternal care.A grant given by the Raisio Plc's Research Foundation made this study possible.Financial support was received also from Nordic Council of Ministers' Arctic Co-operation Programme and University of Helsinki.In addition, we would like to thank the Reindeer Herders' Association for maintaining the Kutuharju research station and collecting the reindeer herd data; Heikki Törmänen and Jukka Siitari at the Finnish Game and Fisheries Research Institute for checking, updating and supplying the data into our use; Sophie Eaglen and Piter Bijma on suggestions on computing the correlation between the residuals; Marja-Liisa Sevón-Aimonen for the comments on the methodology and two anonymous reviewers who gave marked input and helped to improve the manuscript substantially.
BW = birth weight, AW1 = calf autumn weight, BD = birth date, ADG = average daily gain, ind = individual tness, WW7 = cumulative weaning weight of calves over seven years, DAW = dam's weight in her rst autumn, D_ADG = average calf daily gain of a female, CAW = the autumn weight of dam's rst calf.N = number of records, SD = standard deviation, and CV = coe cient of variation.
additive genetic variance, 2 m = maternal additive genetic variance, am = direct-maternal additive genetic covariance, 2 c = the variance of maternal permanent e ect, variance due to maternal permanent environmental e ects as proportion of total variance, r am = direct-maternal genetic correlation, LogL = the log likelihood value.
Figure1.e predicted individual genetic deviations from the population mean for dam's weight in her rst autumn (DAW1) versus the cumulative sum of her calves' weaning weight over seven years (WW7).e values are standardized by mean and standard deviation.

Table 2 .
e signi cance (F test) of xed e ects for birth date (BD), birth weight (BW), calf autumn weight (AW1) and average daily gain (ADG) in the Kutuharju reindeer herd in 1987-2011.e estimated level values of xed e ects (except year) are expressed as deviations from the last level.