Genetic correlations between growth performance and carcass traits of purebred and crossbred pigs raised in tropical and temperate climates

Abstract

In pig breeding, selection commonly takes place in purebred (PB) pigs raised mainly in temperate climates (TEMP) under optimal environmental conditions in nucleus farms. However, pork production typically makes use of crossbred (CB) animals raised in nonstandardized commercial farms, which are located not only in TEMP regions but also in tropical and subtropical regions (TROP). Besides the differences in the genetic background of PB and CB, differences in climate conditions, and differences between nucleus and commercial farms can lower the genetic correlation between the performance of PB in the TEMP (PB_TEMP) and CB in the TROP (CB_TROP). Genetic correlations (r_g) between the performance of PB and CB growing-finishing pigs in TROP and TEMP environments have not been reported yet, due to the scarcity of data in both CB and TROP. Therefore, the present study aimed 1) to verify the presence of genotype × environment interaction (G × E) and 2) to estimate the r_g for carcass and growth performance traits when PB and 3-way CB pigs are raised in 2 different climatic environments (TROP and TEMP). Phenotypic records of 217,332 PB and 195,978 CB, representing 2 climatic environments: TROP (Brazil) and TEMP (Canada, France, and the Netherlands) were available for this study. The PB population consisted of 2 sire lines, and the CB population consisted of terminal 3-way cross progeny generated by crossing sires from one of the PB sire lines with commercially available 2-way maternal sow crosses. G × E appears to be present for average daily gain, protein deposition, and muscle depth given the r_g estimates between PB in both environments (0.64 to 0.79). With the presence of G × E, phenotypes should be collected in TROP when the objective is to improve the performance of CB in the TROP. Also, based on the r_g estimates between PB_TEMP and CB_TROP (0.22 to 0.25), and on the expected responses to selection, selecting based only on the performance of PB_TEMP would give limited genetic progress in the CB_TROP. The r_g estimates between PB_TROP and CB_TROP are high (0.80 to 0.99), suggesting that combined crossbred–purebred selection schemes would probably not be necessary to increase genetic progress in CB_TROP. However, the calculated responses to selection show that when the objective is the improvement of CB_TROP, direct selection based on the performance of CB_TROP has the potential to lead to the higher genetic progress compared with indirect selection on the performance of PB_TROP.

INTRODUCTION

In pig breeding, consolidation has resulted in a reduced number of global breeding programs where selection takes place in purebred pigs (PB) raised mainly in temperate climates (TEMP) under optimal environmental conditions (Knap, 2005). However, pork production typically makes use of crossbred (CB) animals raised in nonstandardized commercial farms all over the world. As the final product of pig breeding is mostly a CB animal (Hidalgo et al., 2015), the genetic correlation between the performance of PB and CB (r_pc) is an important parameter to be considered by pig breeding companies applying crossbreeding schemes. Combined crossbred and purebred selection (CCPS) is recommended for traits presenting r_pc estimates lower than 0.8 (Wei and van der Werf, 1994), which is the case for pigs, where the average of the reported r_pc estimates is 0.63 (Wientjes and Calus, 2017).

In addition to the variable sensitivity of genotypes to changes from nucleus to commercial farming systems, differences in sensitivity to climate conditions can also lower the r_pc when PB and CB are kept in different climates. Around 50% of the commercial farms are in tropical and subtropical regions (TROP; Rosé et al., 2017). Sensitivity to ambient temperature and humidity (heat stress) has been described in growing-finishing pigs (Zumbach et al., 2008; Fragomeni et al., 2016; Rosé et al., 2017). In an international context, the sensitivity to heat stress becomes especially important because pork production is spread between TROP and TEMP. Nevertheless, to the best of our knowledge, r_g between the performance of PB in TEMP and CB growing-finishing pigs in TROP environments have not been reported yet. This is probably due to the scarcity of data in both CB and TROP. Therefore, the current study aimed 1) to verify the presence of genotype × environment interaction (G × E) and 2) to estimate the r_pc for carcass and growth performance traits when PB and 3-way CB pigs are raised in both TEMP and TROP climates.

MATERIALS AND METHODS

Ethic Statement

Data for this study were collected as part of routine data recording in a commercial breeding program. Observations from 19 farms located in 4 different countries (Brazil, Canada, France, and the Netherlands) were used in this study. All these farms are operating in line with the regulations on protection of animals of their countries.

Dataset

To verify the presence of G × E for different climates, phenotypic records were available on 217,332 PB and 195,978 CB pigs (Table 1), across 2 climatic environments, TROP (Brazil) and TEMP (Canada, France, and the Netherlands). The PB population consisted of 2 sire lines, which were located in 12 farms. The CB population consisted of terminal 3-way CB progeny generated by crossing sires from one of the PB sire lines with commercially available 2-way maternal sow crosses, which were located in 8 farms. Pedigree records were available for all animals, up to a maximum of 17 generations. A total of 535,272 pigs were included in the pedigree file with 6,229 different sires and 30,800 different dams.

Table 1.

Number of pigs with phenotypes of each line (sire or 3-way cross) by country¹

Country	Purebred farms	Sire line 1	Sire line 2	Crossbred farms	CB 1	CB 2	Total
Brazil	4	7,223	13,451	2	975	4,785	26,434
Canada	3	46,598	14,989	1	9,045	—	70,632
France	1	29,345	24,196	—	—	—	53,541
The Netherlands	4	39,134	42,396	5	80,044	101,129	262,703
Total	12	122,300	95,032	8	90,064	105,914	413,310

¹CB = three-way cross between the numbered sire line and a crossbred female (Large White × Landrace).

Traits

Each growth performance trait (ADG; lipid deposition [LD], and protein deposition [PD]) and carcass trait (back fat thickness [BF] and muscle depth [MD]) was considered as a different trait depending on the group of pigs in which it was measured (i.e., the 4 groups PB_TROP, CB_TROP, PB_TEMP, and CB_TEMP) (Table 2). All animals were weighed individually at the start of the growing-finishing period (“ontest”). All PB, and all CB in Canada, had their BW (kilogram) recorded, and BF (millimeter) and MD (millimeter) ultrasonically measured at the end of the growing-finishing period (“offtest”). In Brazil, most of CB had their BW recorded, and BF ultrasonically measured at offtest. A small number, 250, of CB pigs in one farm in Brazil, had their hot carcass weight (HCW) recorded at slaughter. All CB animals in the Netherlands had their HCW recorded along with BF and MD using the Hennessy Grading Probe (Hennessy Grading Systems, Auckland, New Zealand) or the Capteur Gras Maigre (CGM, Sydel, France) at slaughter. For the pigs with BW recorded offtest (BW_offtest), ADG_ontest (gram per day) was obtained as the difference between BW_offtest and BW_ontest, divided by the length of the test period. For the pigs with HCW recorded at slaughter, ADG was obtained as the calculated BW (CBW) minus BW_ontest, divided by the length of the growing-finishing period. The formula used to obtain the CBW based on the HCW (Handboek varkenshouderij, 2004) was the following:

Table 2.

Number of observations (No.), mean (µ), standard deviation (SD), minimum (Min), and maximum (Max) for covariates¹ and traits² used to estimate variance components and genetic correlations

	TROP						TEMP
Traits	No.	μ	SD	Min	Max	No.	μ	SD	Min	Max
Purebreds
BWbirth, g		1,658	331.5	660.0	2,670		—	—	—	—
BWontest, kg		29.6	5.6	10.0	57.0		32.1	8.4	9.0	59.0
BWofftest, kg		102.9	11.4	73.9	152.7		125.9	10.0	94.3	155.1
ADG, g/d	20,344	934.5	128.1	543.0	1,464	192,767	1,001	138.6	540.0	1,474
LD, g/d	19,987	145.9	45.3	29.1	376.6	188,250	162.9	53.8	21.3	450.1
PD, g/d	19,987	160.6	21.0	85.5	248.4	188,250	172.1	24.3	78.9	272.5
BF, mm	20,746	10.1	1.7	4.3	17.4	195,394	9.6	1.9	3.3	18.6
MD, mm	13,979	58.3	5.9	38.6	80.3	193,856	59.0	5.9	37.5	81.4
Crossbreds
BWbirth, g		1,445	325.7	450.0	2,350		1,375	312.4	440.0	2,350
BWontest, kg		24.4	4.8	10.5	40.4		25.7	4.7	10.2	40.8
BWofftest, kg		104.8	10.9	78.4	151.8		118.6	9.0	79.0	150.0
HCW, kg		93.1	7.9	68.0	113.9		92.9	6.6	72.2	114.0
CBW, kg		118.5	8.3	90.9	139.2		118.4	6.8	95.8	139.4
ADG, g/d	5,756	936.3	104.1	507.0	1,336	47,945	869.2	93.7	562.0	1,156
LD, g/d	5,227	212.3	57.6	69.3	459.2	21,205	219.8	58.8	53.7	490.0
PD, g/d	5,227	144.1	13.8	92.0	200.4	21,205	138.9	19.2	64.7	211.7
BF, mm	5,577	13.1	2.3	6.1	20.3	190,064	13.6	2.6	5.5	24.1
MD, mm	—	—	—	—	—	190,563	62.4	6.5	39.7	86.6

¹BW_birth = body weight at birth; BW_ontest = body weight ontest; BW_offtest = body weight offtest; HCW = hot carcass weight; CBW = calculated BW.

²LD = lipid deposition; PD = protein deposition; BF = back fat thickness; MD = muscle depth; TROP = tropical climate; TEMP = temperate climate.

$$\text{CBW}=1.3x\text{HCW}0.0025\text{x}{\text{HCW}}^2+0.2075x\text{HCW}.$$

Lipid deposition (gram per day) and PD (gram per day) were estimated as the increment in lipid and protein mass content during the growing-finishing period based on BW and back fat measurements (de Greef et al., 1994):

$$\%{\text{fat}}_{\text{offtest}}=\frac{\text{BF},\text{mm}-1.87}{53.3},$$

$$\%{\text{fat}}_{\text{ontest}}=\%{\text{fat}}_{\text{offtest}}\times \frac{-0.000005{({\text{BW}}_{\text{ontest}})}^2+0.0019({\text{BW}}_{\text{ontest}})+0.0665}{-0.000005{({\text{BW}}_{\text{offtest}})}^2+0.0019({\text{BW}}_{\text{offtest}})+0.0665},$$

$$\text{Protein water ratio}=5.39{(\text{BW}\times 0.14)}^{-0.145},$$

$$\text{Ash}=0.03\times \text{BW,}$$

$$\text{Lipid mass (LM)}=\%\text{ fat}\times 0.95\times \text{BW},$$

$$\text{Protein mass (PM)}=\frac{0.95\times \text{BW}-\text{LM}-\text{Ash}}{\text{Protein water ratio}+1},$$

$$\text{LD}=\frac{({\text{LM}}_{\mathrm{offtest}}-{\text{LM}}_{\mathrm{ontest}})\times \mathrm{1,000}}{\text{Test length, d}},$$

$$\text{PD}=\frac{({\text{PM}}_{\mathrm{offtest}}-{\text{PM}}_{\mathrm{ontest}})\times 1,000}{\text{Test length, d}}.$$

Genetic Parameters Estimation

Univariate analyses were performed to estimate the variance components and heritabilities for all traits. Genetic correlations were estimated using bivariate analyses. A linear mixed model implemented in ASReml (Gilmour et al., 2009) was used for the analyses as follows:

$$y=Xb+Za+Wc+Vg+Uf+e,$$ [1]

in which y is the vector of observations; X, Z, W, V, and U are known incidence matrices; b is a vector of fixed effects (Table 3); a is a vector of random additive genetic effects (breeding values), $a~N\left(0,A\otimes {\Sigma }_a\right)$; c is a vector of random nongenetic effects common to individuals born in the same litter, $c~N\left(0,I_c\otimes {\Sigma }_c\right)$; g is the vector of random pen effects (individuals grouped together in the same pen) $g~N\left(0,I_g\otimes {\Sigma }_g\right)$; f is the vector of random effects common to individuals performance tested in the same compartment of the barn within the same contemporary group, $f~N\left(0,I_f\otimes {\Sigma }_f\right)$; and e is a vector of residuals, $e~N\left(0,I_e\otimes {\Sigma }_e\right)$. A is a matrix of average additive genetic relationships among all individuals, I_c, I_g, I_f, and I_e are identity matrices of the appropriate dimensions and ${\Sigma }_a,{\Sigma }_c,{\Sigma }_g,{\Sigma }_f$, and ${\Sigma }_e$ are covariance matrices related to each effect. In the case of univariate analyses, the covariance matrix ${\Sigma }_i$ is a scalar with the variance component, ${\sigma }_i$, associated with the respective effect. In the case of bivariate analyses, the covariance matrices for PB_TROP and CB_TROP are given by:

Table 3.

Fixed effects included in the vector b of Eq. 1 for the traits¹

Model	Dependent trait(s)1
	Fixed effects2
A	ADG; LD; PD
	µ + SEXj + LINEk + HYSl + COMPm + b1 × BWbirth
B	BF and MD offtest
	µ + SEXj + LINEk + HYSl + COMPm + b1 × BWofftest
C	BF and MD at slaughter
	µ + SEXj + LINEk + HYSl + COMPm + b1 × HCW

¹LD = lipid deposition; PD = protein deposition; BF = back fat thickness; MD = muscle depth.

²SEX = the sex of the pig; LINE = the line of the pig; HYS = herd–year–season = farm × year × month of birth; COMP = compartment within barn × farm; BW_birth = body weight at birth; BW_offtest = body weight offtest; HCW = hot carcass weight (BF and MD were pre-adjusted for the covariate weight prior to the analysis).

$$\begin{array}{c}{\sum }_a=\left[\begin{array}{cc}{\sigma }_{a\_P{B}_{TROP}}^2& {\sigma }_{a\_P{B}_{TROP}C{B}_{TROP}}\\ sym& {\sigma }_{a\_C{B}_{TROP}}^2\end{array}\right];\\ {\sum }_c=\left[\begin{array}{cc}{\sigma }_{c\_P{B}_{TROP}}^2& 0\\ 0& {\sigma }_{c\_C{B}_{TROP}}^2\end{array}\right];\\ {\sum }_c=\left[\begin{array}{cc}{\sigma }_{c\_P{B}_{TROP}}^2& 0\\ 0& {\sigma }_{c\_C{B}_{TROP}}^2\end{array}\right];\\ {\sum }_g=\left[\begin{array}{cc}{\sigma }_{g\_P{B}_{TROP}}^2& 0\\ 0& {\sigma }_{g\_C{B}_{TROP}}^2\end{array}\right];\\ {\sum }_f=\left[\begin{array}{cc}{\sigma }_{f\_P{B}_{TROP}}^2& 0\\ 0& {\sigma }_{f\_C{B}_{TROP}}^2\end{array}\right];\\ {\sum }_e=\left[\begin{array}{cc}{\sigma }_{e\_P{B}_{TROP}}^2& 0\\ 0& {\sigma }_{e\_C{B}_{TROP}}^2\end{array}\right].\end{array}$$

For the bivariate analysis of the other combinations of groups PB_TROP, CB_TROP, PB_TEMP, and CB_TEMP, the covariance matrices are set up in the same manner. The r_pc estimates in the TROP $\left({\text{r}}_{{\text{pc}}_{\text{TROP}}}\right)$ and in the TEMP $\left({\text{r}}_{{\text{pc}}_{\text{TEMP}}}\right)$, the r_g between the performance of PB in both climates $\left({\text{r}}_{g_{\text{PB}}}\right)$, the genetic correlation between the performance of PB_TROP and ${\text{CB}}_{\text{TEMP}}\left({\text{r}}_{{\text{PC}}_{\text{TROP-TEMP}}}\right)$, and the genetic correlation between the performance of PB_TEMP and ${\text{CB}}_{\text{TROP}}\left({\text{r}}_{{\text{PC}}_{\text{TEMP-TROP}}}\right)$ were estimated by:

$$\begin{array}{c}{r}_{p{c}_{TROP}}=\frac{{\sigma }_{a\_P{B}_{TROP}C{B}_{TROP}}}{\sqrt{{\sigma }_{a\_P{B}_{TROP}}^2 {\sigma }_{a\_C{B}_{TROP}}^2}}; \\ r_{p{c}_{TEMP}}=\frac{{\sigma }_{a\_P{B}_{TEMP}C{B}_{TEMP}}}{\sqrt{{\sigma }_{a\_P{B}_{TEMP}}^2 {\sigma }_{a\_C{B}_{TEMP}}^2}};\\ r_{g_{PB}}=\frac{{\sigma }_{a\_P{B}_{TROP}P{B}_{TEMP}}}{\sqrt{{\sigma }_{a\_P{B}_{TROP}}^2 {\sigma }_{a\_P{B}_{TEMP}}^2}};\\ r_{p{c}_{TROP-TEMP}}=\frac{{\sigma }_{a\_C{B}_{TEMP}P{B}_{TROP}}}{\sqrt{{\sigma }_{a\_C{B}_{TEMP}}^2 {\sigma }_{a\_P{B}_{TROP}}^2}};\\ r_{p{c}_{TEMP-TROP}}=\frac{{\sigma }_{a\_C{B}_{TROP}P{B}_{TEMP}}}{\sqrt{{\sigma }_{a\_C{B}_{TROP}}^2 {\sigma }_{a\_P{B}_{TEMP}}^2}}.\end{array}$$

Responses to Selection

To assess the genetic progress a breeding program can achieve for CB_TROP performance for the traits studied here, we use the breeders’ equation to calculate the responses to selection. Phenotypes measured in PB or CB and in TROP or TEMP were considered as 4 different traits, similar to Wientjes and Calus (2017). Three different responses to selection were calculated: 1) the response for the CB_TROP trait to direct selection based on CB_TROP performance $\left({\text{R}}_{{\text{CB}}_{\text{TROP}}({\text{CB}}_{\text{TROP}})}\right)$; 2) the correlated response for the CB_TROP trait to indirect selection based on PB_TROP performance ${\text{CR}}_{{\text{CB}}_{\text{TROP}}({\text{PB}}_{\text{TROP}})}$; and 3) the correlated response for the CB_TROP trait to indirect selection based on PB_TEMP performance ${\text{CR}}_{{\text{CB}}_{\text{TROP}}({\text{PB}}_{\text{TEMP}})}$.

These 3 responses were calculated as follows (Falconer and Mackay, 1996):

$${\text{R}}_{{\text{CB}}_{\text{TROP}}({\text{CB}}_{\text{TROP}})}={\text{i}}_{{\text{CB}}_{\text{TROP}}} \text{x} {\text{h}}_{{\text{CB}}_{\text{TROP}}} \text{x }{\sigma }_a\_{\text{CB}}_{\text{TROP}}$$

where ${\text{i}}_{{\text{CB}}_{\text{TROP}}}$ is the intensity of selection on CB_TROP (assumed to be 1 in this study), ${\text{h}}_{{\text{CB}}_{\text{TROP}}}$ is the square root of the heritability of the trait CB_TROP, and ${\sigma }_a\_{\text{CB}}_{\text{TROP}}$ is the genetic standard deviation of the trait CB_TROP.

$${\text{CR}}_{{\text{CB}}_{\text{TROP}}({\text{PB}}_{\text{TROP}})}={\text{i}}_{{\text{PB}}_{\text{TROP}}} x {\text{h}}_{{\text{PB}}_{\text{TROP}}} x {\text{r}}_{{\text{pc}}_{\text{TROP}}} x {\sigma }_{\text{a}}\_{\text{CB}}_{\text{TROP}}$$

where ${\text{i}}_{{\text{PB}}_{\text{TROP}}}$ is the intensity of selection on PB_TROP (assumed to be 1 in this study), ${\text{h}}_{{\text{PB}}_{\text{TROP}}}$ is the square root of the heritability of the trait PB_TROP, ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ is the genetic correlation between the performance of PB_TROP and CB_TROP, and ${\sigma }_a\_{\text{CB}}_{\text{TROP}}$ is the genetic standard deviation of the trait CB_TROP.

$${\text{CR}}_{{\text{CB}}_{\text{TROP}}{\text{(PB}}_{\text{TEMP}}\text{)}}{\text{=i}}_{{\text{PB}}_{\text{TEMP}}}{\text{ x h}}_{{\text{PB}}_{\text{TEMP}}}{\text{ x r}}_{{\text{pc}}_{\text{TEMP-TROP}}}{\text{ x σ}}_{{\text{a\_CB}}_{\text{TROP}}}$$

where ${\text{i}}_{{\text{PB}}_{\text{TEMP}}}$ is the intensity of selection on PB_TEMP (assumed to be 1 in this study), $h_{{\mathrm{PB}}_{\mathrm{TEMP}}}$ is the square root of the heritability of the trait PB_TEMP, $P{B}_{TEMP,}{r}_{p{c}_{TEMP-TROP}}$ is the genetic correlation between the performance of PB_TEMP and CB_TROP, and ${\sigma }_a\_{CB}_{TROP}$ is the genetic standard deviation of the trait CB_TROP.

Genetic Distances

To evaluate the genetic distance between the PB and CB located at both TROP and TEMP environments, a distance plot was produced by applying principal coordinate analysis to the additive relationship matrix using the function cmdscale in R. From each combination of population and environment, 500 animals were selected at random to be included in the distance plot.

RESULTS

Variance Components

Estimates of variance components are presented in Table 4. As expected, all traits presented medium to high heritability estimates, with the larger values for CB_TROP for growth performance traits (0.38 to 0.45) and for PB_TEMP for carcass traits (0.41 to 0.46). In the TROP, all traits in CB presented larger values for heritability estimates, whereas in the TEMP, PB presented larger heritability estimates for LD, BF, and MD, and CB presented larger estimates for ADG and PD. The phenotypic variance explained by the common environment among litter mates was similar in the 4 groups, being larger for growth performance traits (3% to 7%) than for carcass traits (2% to 5%). The phenotypic variance explained by the contemporary pen effect was larger for growth performance traits (4% to 19%) than for carcass traits (3% to 13%), with the larger values (0.08 to 0.19) for PB_TROP and the lower values (0.02 to 0.09) for CB_TEMP. The phenotypic variance explained by the contemporary compartment effect was larger for growth performance traits (2% to 12%) than for carcass traits (1% to 5%), with the larger values (0.04 to 0.12) for PB_TROP, and the lower values (0.01 to 0.06) for CB in both climates.

Table 4.

Additive genetic and phenotypic variances and contribution (SE) of different random effects¹ to the estimation of the traits² expressed as percentage of the phenotypic variance

Traits	${\sigma }_A^2$	${\sigma }_P^2$	h 2	$\frac{{\sigma }_{ltr}^2}{{\sigma }_P^2}$	$\frac{{\sigma }_{pen}^2}{{\sigma }_P^2}$	$\frac{{\sigma }_{co}^2}{{\sigma }_P^2}$	$\frac{{\sigma }_{ree}^2}{{\sigma }_P^2}$
TROP
Purebreds
ADG	2,680	14,890	0.18 (0.02)	0.06 (0.01)	0.17 (0.01)	0.12 (0.02)	0.48 (0.02)
LD	492	1,822	0.27 (0.02)	0.06 (0.01)	0.12 (0.01)	0.08 (0.01)	0.48 (0.02)
PD	62.9	392.9	0.16 (0.02)	0.05 (0.01)	0.19 (0.01)	0.12 (0.02)	0.48 (0.02)
BF	0.7	2.3	0.29 (0.02)	0.05 (0.01)	0.13 (0.01)	0.05 (0.01)	0.48 (0.02)
MD	6.2	17.7	0.35 (0.03)	0.03 (0.01)	0.08 (0.01)	0.04 (0.01)	0.49 (0.02)
Crossbreds
ADG	3,788	9,238	0.41 (0.06)	0.05 (0.01)	0.06 (0.01)	0.03 (0.02)	0.45 (0.05)
LD	1,146	3,015	0.38 (0.06)	0.04 (0.01)	0.09 (0.01)	0.02 (0.02)	0.47 (0.05)
PD	89.9	199.7	0.45 (0.06)	0.03 (0.01)	0.06 (0.01)	0.05 (0.02)	0.40 (0.05)
BF	1.8	4.5	0.39 (0.06)	0.03 (0.01)	0.07 (0.01)	0.02 (0.01)	0.49 (0.05)
MD	—	—	—	—	—	—	—
TEMP
Purebreds
ADG	2,810	13,381	0.21 (0.01)	0.07 (0.00)	0.12 (0.00)	0.07 (0.00)	0.53 (0.01)
LD	862	2,462	0.35 (0.01)	0.06 (0.00)	0.09 (0.00)	0.05 (0.00)	0.45 (0.01)
PD	80.5	366.0	0.22 (0.01)	0.06 (0.00)	0.11 (0.00)	0.06 (0.00)	0.54 (0.01)
BF	1.3	2.7	0.46 (0.01)	0.04 (0.00)	0.06 (0.00)	0.03 (0.00)	0.40 (0.01)
MD	7.7	18.8	0.41 (0.01)	0.03 (0.00)	0.06 (0.00)	0.05 (0.00)	0.44 (0.01)
Crossbreds
ADG	1,739	7,563	0.23 (0.01)	0.05 (0.00)	0.07 (0.00)	0.06 (0.01)	0.58 (0.01)
LD	1,014	3,170	0.32 (0.02)	0.04 (0.01)	0.04 (0.00)	0.04 (0.01)	0.56 (0.02)
PD	59.3	219.6	0.27 (0.02)	0.06 (0.01)	0.07 (0.01)	0.05 (0.01)	0.55 (0.02)
BF	2.3	6.0	0.38 (0.01)	0.03 (0.00)	0.03 (0.00)	0.02 (0.00)	0.54 (0.01)
MD	6.1	35.8	0.17 (0.01)	0.02 (0.00)	0.02 (0.00)	0.01 (0.00)	0.77 (0.01)

¹ ${\sigma }_A^2$ = additive genetic variance; ${\sigma }_P^2$ = phenotypic variance; h² = heritability; ${\sigma }_{ltr}^2$ = variance of common litter; ${\sigma }_{pen}^2$ = variance of contemporary pen; ${\sigma }_{co}^2$ = variance of contemporary compartment; ${\sigma }_{res}^2$ = residual variance;

²LD = lipid deposition; PD = protein deposition; BF = back fat thickness; MD = muscle depth; TROP = tropical climate; TEMP = temperate climate. 0.00 = value lower than 0.005.

Genetic Correlations

Estimates of genetic correlations between climates and between PB and CB are presented in Table 5. Some estimates could only be obtained with restrained components, in all cases these estimates included data from PB_TROP, and should be treated with caution. Estimates of $r_{p{c}_{TROP}}$ (0.80 to 0.99) were higher than $r_{p{c}_{TEMP}}$ (0.71 to 0.81). G × E appears to be present for ADG, PD, and MD ($r_{g_{PB}}=0.64 \text{to}\hspace{0.17em}0.075$) and absent for BF and LD ($r_{g_{PB}}=0.97 \text{to} 0.098$). The r_pc estimates of PB_TROP and CB_TEMP were mostly moderate and r_pc estimates between PB_TEMP and CB_TROP were low (0.22 to 0.25) and had relatively high SE.

Table 5.

Genetic correlations (SE)¹

Traits	${\text{r}}_{{\text{pc}}_{\text{TROP}}}$	${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$	${\text{r}}_{{\text{g}}_{\text{PB}}}$	${\text{r}}_{{\text{pc}}_{\text{TROP}-\text{TEMP}}}$	${\text{r}}_{{\text{pc}}_{\text{TEMP}-\text{TROP}}}$
ADG	0.88 (0.14)	0.73 (0.04)	0.64* (0.25)	0.45 (0.40)	0.22 (0.58)
LD	0.97* (0.08)	0.78 (0.04)	0.97* (0.10)	0.87 (0.31)	0.24 (0.77)
PD	0.80 (0.16)	0.79 (0.05)	0.73* (0.23)	0.68 (0.37)	nc
BF	0.99* (0.10)	0.81 (0.02)	0.98* (0.08)	0.97* (0.11)	0.25 (0.52)
MD	—	0.71 (0.02)	0.75 (0.17)	0.88 (0.17)	—

¹ ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ = purebred–crossbred genetic correlation in tropical environment; ${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$ = purebred–crossbred genetic correlation in temperate environment; ${\text{r}}_{{\text{g}}_{\text{PB}}}$ = genotype by climate interaction in PB; ${\text{r}}_{{\text{pc}}_{\text{TROP}-\text{TEMP}}}$ = genetic correlation between the performance of PB in tropical climate and CB in temperate climate; ${\text{r}}_{{\text{pc}}_{\text{TEMP}-\text{TROP}}}$ = genetic correlation between the performance of PB in temperate climate and CB in tropical climate.

*Genetic variance components converged restrained (liable to change from positive definite to fixed at a boundary) and thus should be interpreted with caution; nc = not converged.

Responses to Selection

The calculated responses to selection are presented in Table 6. Direct selection for CB_TROP leads to higher responses than indirect selection on either PB_TROP or PB_TEMP. Across the different traits, the direct response ${\text{R}}_{{\text{CB}}_{\text{TROP}}({\text{CB}}_{\text{TROP}})}$ was between 1.2- and 2.2-fold higher than the correlated response ${\text{CR}}_{{\text{CB}}_{\text{TROP}}({\text{PB}}_{\text{TROP}})}$ and between 3.7- and 6.4-fold higher than ${\text{CR}}_{{\text{CB}}_{\text{TROP}}({\text{PB}}_{\text{TEMP}})}$.

Table 6.

Response to direct selection on CB_TROP and correlated response for indirect selection based on PB_TROP or PB_TEMP

Response1	ADG, g/d	LD, g/d	PD, g/d	BF, mm
${\text{R}}_{{\text{CB}}_{\text{TROP}}{(\text{CB}}_{\text{TROP}})}$	39.4	20.9	6.36	0.84
${\text{CR}}_{{\text{CB}}_{\text{TROP}}{(\text{PB}}_{\text{TROP}})}$	23	16.2	2.92	0.72
${\text{CR}}_{{\text{CB}}_{\text{TROP}}{(\text{PB}}_{\text{TEMP}})}$	6.2	4.8	1.02	0.23
${\text{R}}_{{\text{CB}}_{\text{TROP}}{\text{(CB}}_{\text{TROP}})}/{\text{CR}}_{{\text{CB}}_{\text{TROP}}{\text{(PB}}_{\text{TROP}})}$	1.7	1.3	2.2	1.2
${\text{R}}_{{\text{CB}}_{\text{TROP}}{\text{(CB}}_{\text{TROP}})}/{\text{CR}}_{{\text{CB}}_{\text{TROP}}{\text{(PB}}_{\text{TEMP}})}$	6.4	4.3	6.2	3.7

¹ ${\text{R}}_{{\text{CB}}_{\text{TROP}}{(\text{CB}}_{\text{TROP}})}$ = response to direct CB_TROP selection, ${\text{CR}}_{{\text{CB}}_{\text{TROP}}{(\text{PB}}_{\text{TROP}})}$ = correlated response for CB_TROP performance to indirect selection on PB_TROP performance, ${\text{CR}}_{{\text{CB}}_{\text{TROP}}{(\text{PB}}_{\text{TEMP}})}$ = correlated response for CB_TROP performance to indirect selection on PB_TEMP performance. LD = lipid deposition; PD = protein deposition; BF = back fat thickness.

Genetic Distances

PB_TEMP and CB_TEMP are found close together on the distance plot (Fig. 1), as expected based on the small SE that are observed for $r_{pc{}_{TEMP}}$ in Table 5. For both lines 1 and 2, the average distance between points for PB_TEMP and PB_TROP is larger than the distance between PB_TEMP and CB_TEMP. This indicates that the pedigree relationships contributing to $r_{g{}_{pb}}$ are on average smaller than the pedigree relationships for $r_{pc{}_{TEMP}}$. CB pigs in TROP show different patterns for CB1 and CB2. For sire line 1, some CB_TROP are found close to PB_TROP, some are close to PB_TEMP, whereas others are at some distance from all other groups. For sire line 2, CB_TROP are separated from the other groups, but closest to PB_TEMP.

Figure 1.

Distance plot based on the first (V1) and second (V2) principal components of the additive relationship matrix between purebred (PB) and crossbred (CB) pigs located in both tropical (TROP) and temperate (TEMP) environment for sire lines 1 and 2.

DISCUSSION

The performance of CB pigs is typically improved by applying selection in PB lines. In recent years, the use of data on CB offspring has come into play as well as the collection of performance data in the commercial environment. With this, the need for estimation of G × E and r_pc has increased. Estimates of r_pc in pigs were recently reviewed by Wientjes and Calus (2017) who reported an r_pc average of 0.63. The r_pc can, in theory, be decomposed into a G × E caused by differences between nucleus and commercial farms (Bijma and van Arendonk, 1998; Zumbach et al., 2007; Tusell et al., 2016; Wientjes and Calus, 2017; Godinho et al., 2018), and a genotype × genotype interaction (G × G) due to interaction of genes with the differences in genetic background of PB and CB. Here we aimed to separate these 2 components by estimating r_pc of the same PB and their CB in 2 environments, as well as estimating the PB genetic correlation between these 2 environments.

Data

Even though the data used herein contained records on over 400,000 pigs, it did not yield reliable estimates for some of the correlations of interest. Some estimates of ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ and ${\text{r}}_{{\text{g}}_{\text{pb}}}$ and also the r_pc estimates of PB_TROP and CB_TEMP for BF were obtained with components restrained. The data on PB_TROP are common to all these estimates. Clearly, many records alone are not sufficient for accurate estimation of genetic correlations. In addition, the records obtained in the 2 environments should be on pigs that are closely related. Typically, the relationships are closest between PB and CB in the same environment. Relationships between PB in different environments are lower, and finally, the relationship between CB in different environments is smallest because the pedigree connection would mostly be through relationships between PB in the 2 environments. A small subset of 250 CB pigs in TROP were produced as part of an experiment where semen was collected from PB sires in TEMP and exported to Brazil to produce CB in TROP. In this experiment, the relationships from PB_TEMP to CB_TROP are good but the number of sires involved was too small to estimate genetic correlations based on this experiment alone. Because importing semen is both difficult and costly, due to legislation and veterinary requirements, this is a routine practice for production of PB in TROP, but not for the production of CB in the same environment. In addition, in the last years the number of live boars imported from TEMP to TROP that are used for both PB and CB production has greatly increased. With such a close link between the TEMP and TROP populations, data that are currently been generated may contribute greatly to future better correlation estimates.

Another possibility for improving the estimation of such correlations may be the use of genomic data to more accurately measure relationships between PB_TEMP and CB_TROP. The benefits of genomic, rather than pedigree, relationships for estimation of genetic correlations are unknown. For the estimation of breeding values with genomic relationships, it is still important to have close relatives in the training dataset if a high accuracy is required (Pszczola et al., 2012). To estimate a genetic correlation with small SE may therefore still require close relatives to be present in both PB_TEMP and CB_TROP, even when using a genomic relationship matrix.

Environments

Robertson (1959) suggested that G × E is important when genetic correlations (r_g) are less than 0.8, and this suggestion is widely accepted in animal breeding. Estimates less than 0.8 for ADG, PD, and MD indicate G × E for these traits. G × E may be due to the climatic differences since sensitivity to heat stress has been described in pigs (Zumbach et al., 2008; Fragomeni et al., 2016; Rosé et al., 2017). There may also be differences in the farms for PB in TROP and TEMP that could contribute to G × E. First, PB in TROP were kept in open or semi-open barns. Second, the farm environments in different climates were also different in health status and management practices. Differences in health status of farms have been shown to cause G × E in pigs (Rashidi et al., 2014; Herrero-Medrano et al., 2015; Mathur, 2018).

The nucleus farms for the PB_TEMP probably provide the least, and CB_TROP the most challenging environment, whereas PB_TROP and CB_TEMP are in intermediate environments. The environment of PB_TROP is better controlled than commercial Brazilian farms but considerably less well controlled than the environment provided by genetic nucleus farms in TEMP. Therefore, differences between the environment of PB_TROP and CB_TEMP are probably smaller than the differences between the environment of PB_TEMP and CB_TROP, as PB_TEMP are in the better-controlled environment and CB_TROP are in the less well-controlled environment.

A more challenging environment in CB_TROP could also explain the larger differences between the heritability estimates of PB and CB in TROP in comparison with the same estimates in TEMP (Table 4). Under a less well-controlled environment in CB_TROP, animals would be challenged to express their genetic potential and consequently the genetic variance would be higher. Differences between nucleus and commercial environments are probably smaller in the TEMP what makes these estimates more similar in this environment except for the trait MD.

Estimates of r_pc Within Environment

The estimates of r_pc within environments (${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ and ${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$) were in the range of the literature (Wientjes and Calus, 2017), with higher values in TROP than in TEMP. The ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ for ADG and PD were estimated without model constraints and are in the same range as ${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$ estimates; the equivalent values for LD and BF were close to 1.0 but estimated with components restrained, so differences with the corresponding ${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$ are uncertain. Standard errors of ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ were higher, which is expected given the reduced number of records in TROP. For ADG, BF, and LD, the higher ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ could be explained by the BW being recorded offtest and the back fat measurements being done ultrasonically in both PB and CB in the TROP, whereas in the TEMP, the vast majority of CB had BW calculated based on the carcass weight and BF recorded with a probe in the carcass at slaughter.

The environment of PB in TROP is more challenging than for PB in TEMP, which could also make the difference between PB and CB environments smaller in TROP than in TEMP. This may have resulted in the higher ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ than ${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$. Moreover, selection under improved conditions has been shown to increase environmental sensitivity of genotypes (van der Waaij, 2004). Therefore, the reverse may be true in TROP whereby the challenging environment for PB may have resulted in selection for more robust PB_TROP. This increased robustness could contribute to the higher ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ when transmitted to the CB in TROP.

CCPS is recommended when the r_pc is lower than 0.8; the values of the r_pc estimates in this study are mostly around this value. With higher values of ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ compared with ${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$, we should expect less benefit of having data recorded on CB in TROP than in TEMP.

Purebred–Crossbred Correlations Between Environments

The typical situation in pig production is that genetic progress is created by selection of PB in TEMP. The performance of CB in TROP would therefore depend on the genetics of PB in TEMP. When PB and CB were in different climates, the ${\text{r}}_{{\text{pc}}_{\text{TEMP}-\text{TROP}}}$ was low (0.22 to 0.25). Besides the G × G, these estimates are also lowered by G × E given the climate differences; the differences in the health status and farm management between the TROP and the TEMP, as discussed earlier, also contribute. The r_pc between climates was higher when the PB were in the TROP with ${\text{r}}_{{\text{pc}}_{\text{TROP}-\text{TEMP}}}$ between 0.45 and 0.97. We speculate that an increased robustness of PB_TROP would also be reflected in the higher ${\text{r}}_{{\text{pc}}_{\text{TEMP}-\text{TROP}}}$ compared to the ${\text{r}}_{{\text{pc}}_{\text{TROP}-\text{TEMP}}}$. This could indicate a benefit for overall efficiency by selecting records on PB_TROP when the objective is to improve the performance of CB in both TEMP and TROP. However, most of the pigs’ nucleus farms are in the TEMP.

Genetic Distances

Figure 1 shows data on the same number of pigs for all 4 groups. However, the points for CB_TEMP are much more clustered than for the other 3 groups. The points for CB_TROP, especially in the upper panel for sire line 1, are less clustered than for the other 3 groups. The SE of ${\text{r}}_{{\text{pc}}_{\text{TEMP}}}$ are the smallest (Table 5), which is to be expected given that PB_TEMP and CB_TEMP are found closest together on the distance plots, and the largest number of records contribute to these estimates. The large pedigree distances perhaps contribute to the size of SE of ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$, although the major reason for that might be the smaller number of records in CB_TROP. Given the smaller distance of CB_TROP to PB_TEMP, a smaller SE would be expected for ${\text{r}}_{{\text{pc}}_{\text{TEMP}-\text{TROP}}}$ than for ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ but the opposite is observed. The estimated values for ${\text{r}}_{{\text{pc}}_{\text{TEMP}-\text{TROP}}}$ are however much closer to 0 than for ${\text{r}}_{{\text{pc}}_{\text{TROP}}}$ which increases the expected SE of the estimates (Bijma and Bastiaansen, 2014). Smaller, though still rather large, values are seen for the SE of ${\text{r}}_{{\text{pc}}_{\text{TROP}-\text{TEMP}}}$ compared with the SE of ${\text{r}}_{{\text{pc}}_{\text{TEMP}-\text{TROP}}}$, which is not supported by the distances in Figure 1. Especially in the lower panel for sire line 2, a large distance is observed between PB_TROP and CB_TEMP. The much larger dataset for CB_TEMP will probably have contributed to this smaller SE.

CONCLUSIONS

The current dataset collected by pig breeding programs makes it difficult to obtain accurate estimates of genetic correlations between pig performance in temperate and tropical climates. This is a fact especially when the aim is to estimate r_pc between PB in one climate and CB in the other climate because pigs in both climates are only distantly related. G × E between the temperate and tropical environment appears to be present for ADG, PD, and MD. In addition, r_pc within TEMP and within TROP are not close to 1 for most traits. Collection of phenotypes in the TROP should be included when the objective of selection is the performance of CB_TROP. On the basis of $r_{p{c}_{TEMP-TROP}}$ and the expected responses to selection, selection for CB_TROP based on the performance of PB_TEMP would compromise the genetic progress for the traits being studied. Based solely on the $r_{p{c}_{TEMP}}$ estimates, CCPS would not be necessary to increase genetic progress in CB_TROP. However, based on the calculated responses to selection, when the objective is to improve the performance of CB_TROP, direct selection based on the performance of CB_TROP has the potential to lead to the higher genetic progress for growth and carcass traits.