Genetic and environmental e ! ects a ! ecting the variation in birth date and birth weight of reindeer calves

"e factors causing variation in birth date and birth weight were analysed from the data from an experimental reindeer herd consisting of 1136 calves with parental information. "e traits had coe+cient of variation of 37 and 14%, respectively. "e variation in both traits was a!ected by year and sex (male calves heavier) and by management factors, such as the age distribution of females and males. Early calving results from the use of older breeding males and is most apparent in prime age females. In both traits the heritability was moderate (0.23 and 0.24) with a high proportion of maternal genetic variation (0.23) in birth weight. "e North Atlantic Oscillation (NAO) indices, summarizing the major weather conditions prior to rut, explained part of the annual variation in the traits. "e amount of total genetic variation in relation to trait mean, or the evolvability, was 21% in birth date and 10% in birth weight indicating that selection could successfully be used to improve herd productivity and that the traits have substantial potential for adapting to possible changing environmental conditions. "e results on genetic correlations imply that selection on calf ’s birth weight leads on one hand to calves being born earlier and on the other hand to dams with later parturition.


Introduction
Birth date and birth weight a ect reindeer calf survival and growth and therefore productivity.Birth date (BD) is important in the subarctic region to match with the vegetation burst of short growing season (reviewed in Holand et al., 2003).Heavy birth weight (BW) is linked to a good survival and growth (Eloranta & Nieminen, 1986).
ere is much evidence from ungulates demonstrating that the variation in calf traits is inuenced by maternal e ects (Rönnegård, 2003;Bijma, 2006;Clements et al. 2011).Analysis on calf growth and survival should therefore contain direct e ect (calf ) and maternal e ect (dam) (Willham, 1972).
e maternal e ect can be separated into genetic e ect and permanent environmental e ect (Willham, 1972).As Genetic and environmental e ects a ecting the variation in birth date and birth weight of reindeer calves a practical example of the permanent non-genetic e ect, a dam could have grown large in a favourable environment and is therefore able to provide good support for the growth of calves (permanently) in di erent parities.
e estimates of heritability (h 2 ) for BD and BW are lacking in reindeer.In red deer, a species closely related to reindeer, Clements et al. (2011) have reported h 2 = 0.09 in BD and Archer et al. (2013) h 2 = 0.20 for conception date, and Kruuk & Had eld (2007) have found heritability of 0.14 for BW.
In reindeer the major determinant for BD is the conception date (Holand et al., 2003).Indeed, dam weight prior to conception a ects the calving date with the well-nourished and heavy dams calving early (Holand et al., 2003;Mysterud et al., 2009).Shortened day length and weather in autumn have been found to affect the conception time (review by Ropstad, 2000).e North Atlantic Oscillation (NAO) index (Hurrell, 1995) is commonly used in summarising the weather conditions (e.g.Weladji & Holand, 2003).
In reindeer there are still only a few studies (Varo, 1972;Rönnegård, 2003) on the genetic variation of quantitative traits.At Kutuharju experimental reindeer eld station in Finland data on individuals have been collected since the late 1960s (Eloranta & Nieminen, 1986), including the sire identi cation using DNA markers since 1997 (Røed et al., 2002).By utilising the Kutuharju reindeer data we investigated the variation in calves' birth date and birth weight in detail, including the e ect of environmental conditions, herd management factors, and direct and maternal genetic e ect, as well as the permanent environmental e ect of the dam.Further, we analysed the genetic correlations between the traits to understand how possible selection schemes would change the traits.

Data description
Study area e data were collected at Kutuharju experimental reindeer research eld station in Kaamanen,Finland (69°N,27°E).
e station, with a reindeer herd of about 100 individuals living in a fenced area of about 45 km 2 , is owned and operated by the Reindeer Herders' Association.e Reindeer Research Station of e Finnish Game and Fisheries Research Institute maintains the data.
Study animals e Kutuharju herd was founded in 1965 and the data collection started in 1969.Before 1970 there were 113 female and 51 male base animals; later 97 females and 79 males have been purchased from outside.Altogether 2980 calves have been born in 1969-2011, resulting in a total of 3320 individuals in the pedigree.During the winter the animals were supplimental fed in 1987-2011.
Available information e variables in the analyses were BD, BW, dam weight prior to conception (DW), birth year, sex and parity.e BD and BW were measured by daily observations and recordings of the herd during the calving season (Eloranta & Nieminen, 1986).BD was de ned as the number of days since 1 st of May. In years 1987, 1988, and 1990-1993 the marking harnesses were used to identify the sires of calves.From 1997 onwards the paternities have been conrmed at the Norwegian School of Veterinary Science using DNA markers giving sire information (Røed et al., 2002).e April-September NAO indices (Climate Prediction Center Internet Team, 2013) prior to conception were used to detect the e ect of weather on BD.

Exclusions
Calves without information on BD (N = 133) and paternity (N = 1700) were excluded.An additional 5 calves were removed as outliers for BD later than 1 July.Stillborn calves (N = 6) were not included in the analyses.e nal analysed data contained 1136 calves.Of the dams, 1117 had information on DW.

Statistical analyses
Analysis of xed e ects First, the signi cance of xed classi ed e ects was tested using a general linear model.Second, a regression analysis was used to nd the monthly NAO indices explaining the variation in the traits.
e xed e ects (and their levels) were birth year, parity, sire age (2,…,6 y), sex (male or female) and calf at foot during rut (present =1 or absent = 0).Among parities 10-13 were combined due to the small number of records.e sire classes for age 1 and 2 years were merged, as were those for 6 and 7 years old.
In the regression analyses using the NAO indices, for the consistency the regression variables were also calf sex, sire age, dam parity (and its squared value) and presence of calf at foot at rut. e statistical tests and preliminary analyses were performed with statistical softwares SAS EG®, version 4.2 (SAS Institute Inc., Cary, NC) and R, version 2.15.1 (R Development Core Team, 2012).
Models for variance component estimation e univariate analyses contained: 1) e (direct) additive genetic e ect of the animal (random) e ect, 2) Both direct (calf ) and maternal (dam) additive genetic e ects, and 3) Direct and maternal additive genetic e ects and maternal permanent e ect (dam), so that the maternal impact on calf is made of three di erent e ects: genetic, permanent (over parities) non-genetic and temporary non-genetic e ect.
In 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 + Zcc + e, where the vector y represented observations, b 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 the 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.
e residual e ects were assumed to be independent, random variables identically distributed around mean 0 with variance 2 e .For the random e ects V(a) = 2 a A, V(m) = 2 m A, V(c) = 2 c I, V(e) = 2 e I, where A is the addi- tive relationship matrix of individuals and I is the identity matrix.e variance of direct additive genetic e ect is 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 effect as am .
e birth year was used in the genetic analyses as it includes also management e ects varying between years, together with the climatic e ects.e random e ects were tested using likelihood ratio test (LRT) with and without the respective random e ect; calculating twice the di erence between the log-likelihood which is assumed to be distributed as ⌾ 2 with one degree of freedom, i.e. the di erence in the number of unrestricted parameters tted in the two models (Pinheiro & Bates, 2000).
Depending on the convergence, di erent types of multivariate analyses were carried out to detect the genetic (co)variation with the traits.e analyses were conducted using a Restricted Maximum Likelihood method (Patterson & ompson, 1971).e actual estimation was carried out with statistical package ASReml version 3.0 (Gilmour et al., 2009).

Data
In the Kutuharju pedigree data there were altogether 3320 animals across more than 10 generations: 101 sires (21 sires of sires and 67 dams of sires) and 566 dams (44 sires of dams and 261 dams of dams).Of the dams, 39.5% had a record on BD and BW.Percentage of dams and sires with more than one o spring was 77 and 82%, respectively (Table 1).
e BD ranged from 2 May to 16 June (mean 18 May), the BW from 1.8 to 10.4 kg (mean 6.0 kg) and DW from 54 to 107 kg (mean 82 kg).e distributions of BD and BW followed approximately the normal distribution (results not shown).e coe cients of variation (CV) for the traits were 37% for BD and 14% for BW.

Fixed e ects
e e ects of birth year and parity were substantial in both traits (Table 2).In BD and BW, the e ects of age of sire and calf sex were also signi cant.e lactation status of dam (i.e. the presence or absence of calf at autumn) seemed to have no e ect on the studied traits.
When analysing the annual variation, only the NAOs prior to oestrus in the year before calving were considered (Table 3).e weather in late spring and early autumn may a ect BD and DW (body condition), and hence the onset of oestrus cycling.e results for the other xed e ects were similar to those with classi ed factors (Table 2).

Analyses of the genetic e ects
Single-trait analyses Estimates of (co)variance components, heritabilities and correlations for the direct and maternal e ects in BD and BW together with log likelihood values (logL) are summarised in Table 4. e model (1) with no maternal effects revealed moderate to high heritabilities.Fitting maternal genetic e ects (model 2) increased logL values markedly (LRT test value was 3.02, P = 0.08 for BD and 45.80, P<0.001 for BW). e further addition of maternal permanent environmental e ect (model 3) resulted in a non-signi cant improvement (LRT test value was 0.5, P = 0.48, for BD and 0.43, P = 0.84, for BW).Consequently, the best model for both traits was model 2 that included the direct genetic e ect of animal and the maternal genetic e ect.In BD the maternal heritability was small with high standard error, suggesting that genetically dam has little in uence on the variation.In both traits the direct-maternal genetic correlations were smaller than their standard errors.
Bivariate analysis e (co)variances from the bivariate analyses are shown in Table 4 and the respective correlations in Table 5. e estimates were similar to those from the univariate analyses, with a slight increase for 2 m of BD and 2 d of BW, and considerable change of am .Among the genetic correlations only the ones related to direct e ects of BW were larger than twice their standard error.Also phenotypic correlation between BD and BW was smaller compared to its standard error.e (co)variance structure of the DW at the autumn prior to calving together with BW and BD was analysed with two separate bivariate analyses.When BD was analysed together with DW, its direct genetic variance slightly increased (from 10.22 to 11.95) and the directmaternal covariance changed from negative to positive value (from -0.38 to 0.18) compared to analysis of BD together with BW.In BW there were no changes.In DW the heritability was ( 2 a / 2 p = 28.61/ 41.44 =) 0.69 (±0.044).e correlations between DW and BD (including maternal e ects) ranged from -0.07 to 0.06 with high standard errors (from 0.04 to 0.17).
e only signi cant correlation between DW and BW was the phenotypic correlation, 0.16 (±0.046).

Discussion
e Kutuharju experimental herd provides unique data on reindeer calves' birth date and weight, life history of females and indispensable information on calves' sire (allowing the separation of maternal genetic e ect).ere is a moderate direct heritability in birth date and birth weight.e maternal heritability in birth weight is moderate and in birth date small.e evolvability (Houle, 1992), or coe cient of total additive genetic variation, including also maternal genetic variation (√( 2 a + 2 am + 2 m )) / mean) (Bijma, 2006), quanti es the potential for selection.It is 0.21 for birth date and 0.10 for birth weight, which are comparable to the expected responses in the economically relevant traits of farm animal species expressed in as a percentage of the mean (Smith, 1984). 2 a = direct additive genetic variance, 2 m = maternal additive genetic variance, am = direct-maternal additive genetic covariance, 2 c = the variance of maternal permanent e ect, 2 e = residual variance, 2 p = phenotypic variance, h 2 d = direct heritability, h 2 m = maternal heritability, c 2 = the variance due to maternal permanent environmental e ects as proportion of total variance, r am = direct-maternal genetic correlation, LogL = the log likelihood value.

Data and methods
Including all data, also the calves lacking paternal information, led to upward biased estimates (results not shown).e exclusion of the maternal genetic e ect in the analyses on the data with known sires gave direct heritability of 0.63 for birth weight, which is much higher than the estimate 0.23 from the analysis including maternal e ect. is is resonated well with what Willham (1972) predicted it to be in such a case, i.e. 0.5h 2 d + 1.5 r am + 0.5h 2 m = 0.61.Data with multiple o spring across generations and sire information allow an e cient separation of direct and maternal genetic ef- fects in the univariate analyses (e.g.Kruuk & Had eld, 2007).Elaborated models, especially across traits, resulted in poor convergence of estimated variance parameters in REML analyses.e variation in the gestation length could not be included in the analyses, because the oestrus or copulation dates were not available.Clements et al. (2011) reported in red deer a very high correlation (0.97) between the observed oestrus days and parturition date, dismissing the variation in gestation length.However, there is naturally high variation in birth date because oestrus in autumn is following a three week cycling (review by Ropstad, 2000).We did not see signs of multimodality in the birth date distribution.

Fixed e ects
Sex.Despite the reported longer gestation for the male calves (Mysterud et al., 2009), there was no evidence of calf sex e ect on birth date.
e results on birth weight agree with the common nding about male calves being heavier at birth (e.g.Eloranta & Nieminen, 1986).
Sire age is a very important trait from the management point of view with older males siring the earlier-born calves (Holand et al., 2003).
is is possibly accentuated by the prime-age females choosing to mate with heavier and older males (Holand et al., 2006).e possible di erences caused by management trials in Kutuharju (cf.Holand et al., 2003) are taken into account using sire age and birth year as xed e ects.
Annual variation in birth date and weight was explained partly by the NAO indices of months between April and September of the previous year.e dam's condition depends on weather conditions such as temperature, precipitation and insolation, as they a ect plant phenology and thereby availability of feed (review by Weladji & Holand, 2003).e NAO indices have earlier been considered in studies on reindeer herd productivity (e.g.Weladji & Holand, 2003).e season from spring to autumn prior to rut has greater e ect on the timing of parturition of caribou than conditions at late gestation (Adams & Dale, 1998).ere are comparable ndings in reindeer (Holand et al., 2004).e results suggest that there could be a way to predict the on-set of rutting from the weather statistics and consequently the optimum time for the herders to gather reindeer.
Parity a ects both birth date and weight with the prime aged (4-10 years) dams having earliest calvings and the heaviest calves in accordance with Weladji et al. (2010).
Raising a calf during the summer before the rut had no in uence on dam's weight in autumn or on calf traits in this study.is is in line with the ndings of Holand et al. (2003) and Weladji et al. (2010) from the Kutuharju herd, but in contrast to a study by Bårdsen et al. (2010) with the di erence being probably explained by the supplementary feeding at Kutuharju herd.

Genetic variation
e considerable amount of evolvability (the coe cient of additive genetic variation) suggests that the trait will respond to both articial and natural selection (cf.Smith, 1984).e genetic variation for birth date and birth weight is higher than in red deer (Kruuk & Had eld, 2007;Clements et al., 2011).Archer et al. (2013), however, found heritability 0.20 for the related conception date in the same species.e moderate h 2 m in birth weight suggests considering dam quality when designing a selection scheme for the trait.
We found a negative phenotypic correlation between birth date and birth weight, as was found earlier by Weladji et al. (2010).e results on genetic correlations imply that selection on calf 's birth weight leads on one hand to calves being born earlier and on the other hand to dams with later parturition.In traditional reindeer husbandry neither birth date nor birth weight is recorded, therefore the traits cannot directly be included in the selection criteria used by the herders.Before any recommendation are given on birth weight recording, it is important to understand how the current selection of breeding individuals based on autumn weight would be related to birth weight and date.ere are empirical views (Muuttoranta & Mäki-Tanila, 2012) and results on the advantages of early birth and medium to high birth weight (Eloranta & Nieminen, 1986;Holand et al., 2003).
With the available genetic variation, calf birth weight and timing of birth may be altered by natural selection as a response to changes in environmental conditions.Holand et al. (2004) is suggesting that there is stabilising selection in birth date, indicated by 90% of calves being born within two weeks.

Conclusions
Variation in birth date and birth weight of reindeer calves exists due to management, environment and genetic make up.e ndings could be used to improve productivity by selection based both on calf and dam traits and by herd management (e.g.optimum age distribution of males and females), including even the timing of the herding operations by the deviations from the average weather pattern.e genetic variation enables reindeer to respond to natural selection caused by possible changes in environmental conditions.

Table 1 .
e distribution of dams and sires over progeny in the Kutuharju pedigree data in the years 1987-2011.

Table 2 .
e xed e ects for birth date (BD), birth weight (BW) and dam weight (DW) in the Kutuharju reindeer herd in 1987-2011.e level values of each factor are expressed as deviations from the last level.

Table 3 .
Regression analysis on calf birth date (BD), birth weight (BW) and dam weight (DW) of the NAO indices at the Kutuharju reindeer herd in 1987-2011.

Table 4 .
Estimates of (co)variance components and genetic parameters (standard error in parentheses) for birth date and birth weight in the Kutuharju reindeer herd in 1987-2011.Models 1-3 have following random e ects: 1 = animal, 2 = animal + maternal genetic, and 3 = animal + maternal genetic + maternal permanent environmental e ect.e best model is indicated in bold and used in bivariate context giving the (co)variance components of an analysis of BD and BW together.

Table 5 .
Estimates of direct (dir) and maternal (mat) genetic correlations (standard errors in parentheses) of the bivariate analysis in birth date (BD) and birth weight (BW) of reindeer calves in the Kutuharju reindeer herd in1987-2011.