The present study aimed to genetically improve growth performance under high-heat environments by specifically designing a reaction-norm animal model (RNAM) for purebred Duroc pigs in Japan. A total of 54,750 records of average daily gain (ADG) measured for pigs reared at four farms in different prefectures were analyzed. Estimated maximum daily temperatures at the respective farm locations were used to calculate the average cumulative thermal load (TL). The TL values served as an indicator of high-heat environments for pigs. The plausible cumulative period length and threshold temperature for calculating TL were determined to be 8 weeks until just before shipping and 25°C, respectively. Variance components were estimated via RNAM analysis using TL as a linear covariate. The estimated additive genetic variances under both responsive and non-responsive to TL were found to be significant. Moreover, the estimated heritability of ADG ranged from 0.38 to 0.73 for TL values of 0–8. These results suggest that the RNAM developed holds the potential for improving the genetic ability of growth under high-heat environments in pigs.

1 INTRODUCTION

Global warming has a tremendous negative impact on the overall sustainability of pig production (Mayorga et al., 2019; Ramirez et al., 2022). In Japan, Takada et al. (2008) predicted a decrease in the future growth performance of pigs in most areas because of ongoing global warming. To mitigate the impact of elevated ambient temperatures, various countermeasures have been implemented, including improving the rearing environment in piggery facilities and adjusting the nutrient composition of feeds (Liu et al., 2022; Mayorga et al., 2019). However, as global warming becomes more severe, the cost of pork production may increase owing to reduced productivity (Mayorga et al., 2019). Therefore, ensuring stable pork production under high-heat weather conditions is important.

To identify high-heat environment indicators in pigs, obtaining accurate and actual environmental information is desirable. However, comprehensive on-farm environmental data have not been continuously collected on many farms in Japan. Notably, meteorological observatory data have been accumulated at weather stations located across Japan. In other countries, various studies have been conducted on genetically improving heat tolerance using growth performance and publicly available weather records in pig populations (Fragomeni et al., 2016; Freitas et al., 2023; Usala et al., 2021; Zumbach et al., 2008a). Recently, in Japan, the availability of public weather records and their association with pig production records have been also assessed (Hara et al., 2022; Ogawa et al., 2022; Sasaki et al., 2018; Satoh et al., 2023, 2023).

Genetic improvement of heat tolerance in pigs could serve as an effective strategy for addressing the challenges posed by global warming (Mayorga et al., 2019). To achieve this, a statistical model that considers the genotype by environment interaction (G × E) is essential to evaluate the genetic transmitting ability of heat tolerance. One such model is the reaction-norm model (RNM), a type of random regression framework. The RNM can distinguish between genetic effects that respond to environmental indicators and genetic effect that do not (Cardoso & Tempelman, 2012; Mulder et al., 2006; Via et al., 1995). The Duroc is a popular terminal sire breed around the world. Previous studies have performed reaction-norm sire model analyses on crossbred pig populations with Duroc as the sire breed (Fragomeni et al., 2016; Usala et al., 2021; Zumbach et al., 2008a), as well as reaction-norm animal model (RNAM) analysis using records obtained from purebred Duroc pigs (Fragomeni et al., 2016).

In this study, we aimed to investigate the potential for genetically improving growth performance under high-heat environments using the purebred Duroc pig records in Japan. Initially, we employed regression analysis to construct a model in which average daily gain (ADG) decreased with high-heat environment indicators. Subsequently, we conducted RNAM analysis considering high-heat environment indicators to estimate genetic parameters and predict breeding values. Finally, we discussed the possibility of improving growth performance under high-heat environmental conditions.

2 MATERIALS AND METHODS

2.1 Records

Pedigree and growth performance records were collected from the four-member producer farms (Figure S1) of Global Pig Farms Incorporated (Shibukawa, Japan). The farms are located at latitudes 35.3–43.0°N and longitudes 137.4–141.6°E and are scattered from central to northern Japan. It is important to note that one of the four farms (Farm D) was not included in several results because the growth performance records were not collected after 2020. The trait studied was the average daily gain (ADG, g/day), calculated by dividing the body weight (BW, kg) by the age in days at the end of performance testing. Typically, the performance testing concludes when BW reaches 105 kg. Moreover, ADG records deviating by three standard deviations from the mean were excluded. Contemporary groups were defined by birth farm × year × bi-monthly season ( $\mathrm{FYS}$ ). Subsequently, records of contemporary groups where fewer than 10 pigs that were available per group were excluded. Finally, ADG records of 54,750 purebred Duroc pigs, born between July 2000 and July 2022 and whose BW was measured between January 2001 and December 2022, were analyzed. Pedigree records were available for 58,630 purebred Duroc pigs born between 1970 and 2022.

Weather records for the farms from 2000 to 2022 were obtained from the Agro-Meteorological Grid Square Data (https://amu.rd.naro.go.jp/), which provides estimated 1 km²-gridded meteorological data from public weather station observations with inverse distance weighting and altitude difference correction (Ohno et al., 2016; Sasaki et al., 2022). Public weather stations in Japan are located at approximately 21-km intervals (Japan Meteorological Agency, 2024).

2.2 Average cumulative thermal load

Average cumulative thermal load (TL) was defined as a high-heat environment indicator as follows:

TL=\frac{1}{Period}{\sum}_{i= Start}^{End}\mathit{\max}\left(0,{MaxT}_i- ThrT\right),

where

{MaxT}_i

is the estimated daily maximum temperature (°C) on the ith day;

ThrT

is the threshold temperature (°C), which assumes the upper limit of the thermoneutral zone (Bloemhof et al., 2008, 2013; Fragomeni et al., 2016; Zumbach et al., 2008b);

\mathit{\max}\left(0,{MaxT}_i- ThrT\right)

is the function used to calculate the daily value of

{MaxT}_i

exceeding

ThrT

;

Start

is the start date of the cumulative period;

End

is the end date of the cumulative period, that is, the day before BW measurement; and

Period

is the number of days between

Start

and

End

Various TL values were calculated using different

ThrT

and

Start

values for the following linear model (LM) analysis.

ThrT

was increased from 11 to 30°C at 1°C intervals.

Start

was increased from 7 to 112 days (from 1 to 16 weeks) at 7-day intervals before the end of performance testing. In addition, the no-threshold temperature (without the

\mathit{\max}

function) was analyzed. The TLs were fitted to LM, and the best-fitting model was determined using the lowest Akaike's information criterion (AIC) (Akaike, 1974):

\mathbf{y}=\mathbf{Xb}+\mathbf{e},\mathbf{e}\sim \mathrm{N}\left(\mathbf{0},{\sigma}_e^2\mathbf{I}\right),

where

\mathbf{y}

is a vector of the phenotypic value of the trait (ADG);

\mathbf{b}

is a vector of fixed effects;

\mathbf{e}

is a vector of the random error;

\mathbf{X}

is a design matrix relating

\mathbf{b}

\mathbf{y}

;

{\sigma}_e^2

is a random error variance; and I is an identity matrix. The fixed effects included the intercept;

\mathrm{FYS}

(4 farms × 23 years × 6 seasons per year) (419 levels); sex (male and female) (two levels); a partial regression coefficient of BW (linear covariate); and a partial regression coefficient of TL (linear covariate). The objective of performance testing is to evaluate the ADG when pigs reach a BW of 105 kg. However, actual weigh is not always measured on the day that the pigs reach 105 kg, then there is a wide range of BW. To account for this variation in BW, the linear covariate of BW was included in the model. Data were analyzed using the statsmodels 0.13.2 (Seabold & Perktold, 2010).

2.3 Expected age in days to reach shipping weight

The expected pig age in days to reach 105 kg under TL conditions, namely E(D105|TL), in 2022, was calculated as follows:

E\left(D105| TL\right)=\frac{105000}{E\left( ADG| TL\right)}=\frac{105000}{\mu +{\hat{FYS}}_j+\overline{Sex}+{\hat{b}}_0\times {\overline{BW}}_{2022}+{\hat{b}}_1\times {TL}_{\mathrm{Farm},\mathrm{Day}}},

where D105 is the age in days to reach 105 kg, calculated by dividing 105,000 g by ADG;

\mu

is the solution of the intercept;

{\hat{FYS}}_j

is the jth solution of the FYS effect;

\overline{Sex}

is the average of the sex effect solution;

{\overline{BW}}_{2022}

is the average weight of pigs in 2022;

{TL}_{\mathrm{Farm},\mathrm{Day}}

is the TL of the pigs shipped on that day and from that farm; and

{\hat{b}}_0

and

{\hat{b}}_1

are the solutions of the partial regression coefficient. All solutions used here were obtained from the LM analysis.

The maximum shipment delay in days because of TL was calculated in four steps. The average ADG was calculated backward from the mean of $E\left(D105\ | TL\right),$ and then the ADGs at TL = 0 were calculated using the product of ${\hat{b}}_1$ and the mean TL values. After that, the ADGs at the maximum TL values were calculated using the product of ${\hat{b}}_1$ and the maximum TL values, and finally, the difference in $E\left(D105| TL\right)$ between minimum and maximum TL values were calculated using the ADGs.

2.4 Variance components estimation

Variance components were estimated using a RNAM defined as follows:

\mathbf{y}=\mathbf{Xb}+{\mathbf{Z}}_0{\mathbf{a}}_{\mathbf{0}}+{\mathbf{Z}}_1{\mathbf{a}}_{\mathbf{1}}+\mathbf{e},

\left(\begin{array}{c}{\mathbf{a}}_0\\ {}{\mathbf{a}}_1\\ {}\mathbf{e}\end{array}\right)\sim \mathrm{N}\left[\left(\begin{array}{c}\mathbf{0}\\ {}\mathbf{0}\\ {}\mathbf{0}\end{array}\right),\left(\begin{array}{cc}\mathbf{G}\bigotimes \mathbf{A}& \mathbf{0}\\ {}\mathbf{0}& {\upsigma}_{\mathrm{e}}^2\mathbf{I}\end{array}\right)\right],\mathbf{G}=\left[\begin{array}{cc}{\sigma}_{a0}^2& {\sigma}_{a0a1}\\ {}{\sigma}_{a0a1}& {\sigma}_{a1}^2\end{array}\right],

where

{\mathbf{a}}_0

and

{\mathbf{a}}_1

are vectors of the breeding values that do not respond and respond to TL (intercept and slope term), respectively;

{\mathbf{Z}}_0

and

{\mathbf{Z}}_1

are design matrices relating

{\mathbf{a}}_0

and

{\mathbf{a}}_1

\mathbf{y}

, respectively;

\bigotimes

is a Kronecker product; A is an additive relationship matrix constructed using pedigree data; and

{\sigma}_{a0}^2

{\sigma}_{a1}^2

, and

{\sigma}_{a0a1}

are additive genetic (co)variances for the breeding values. The fixed effects are the same as those in LM without intercept.

Variance components were estimated via the single-trait animal model AI-REML method using BLUPF90+ software (Lourenco et al., 2022) with the variance component estimation (VCE) option method. The inbreeding coefficient was considered when calculating the inverse of the A (Aguilar et al., 2020). The standard error (SE) approximations of the variance components (Meyer & Houle, 2013) were obtained from the convergence results.

The estimated genetic variance [

\operatorname{var}\left({a}_{TL}\right)

] and heritability(

{h}_{TL}^2

) for the ADG under TL condition (at TL value) in the RNAM were calculated as follows:

\mathit{\operatorname{var}}\left({a}_{TL}\right)={\sigma}_{a0}^2+{TL}^2\times {\sigma}_{a1}^2+2\times TL\times {\sigma}_{a0a1},{h}_{TL}^2=\frac{\mathit{\operatorname{var}}\left({a}_{TL}\right)}{\mathit{\operatorname{var}}\left({a}_{TL}\right)+{\sigma}_e^2}.

The estimated genetic correlation for the ADG under different TL conditions (between

{TL}_l

and

{TL}_m

values) in the RNAM, namely

{r}_{gTLl, TLm}

, was calculated as follows:

{r}_{gTLl, TLm}=\frac{\sigma_{a0}^2+\left({TL}_l+{TL}_m\right)\times {\sigma}_{a0a1}+{TL}_l\times {TL}_m\times {\sigma}_{a1}^2}{\sqrt{\left({\sigma}_{a0}^2+{\mathrm{T}L}_l^2\times {\sigma}_{a1}^2+2\times {TL}_l\times {\sigma}_{a0a1}\right)\times \left({\sigma}_{a0}^2+{TL}_m^2\times {\sigma}_{a1}^2+2\times {TL_m}^{\ast}\times {\sigma}_{a0a1}\right)}}.

2.5 Breeding value estimation

Breeding values were estimated via the single-trait animal model BLUP method via RNAM using BLUPF90+ software.

The change in ADG deviation from the population mean (DevADG) with TL in male pigs born in 2022 was calculated to confirm the association between TL and ADG as follows:

{DevADG}_{k,l}={\hat{b}}_1\times {TL}_l+{EBV}_{k,{TL}_l},

where

{DevADG}_{k,l}

is the ADG deviation from the population mean and

{\hat{b}}_1\times {TL}_l

is a population mean at

{TL}_l

{EBV}_{k,{TL}_l}

is EBV of the kth pig at

{\mathrm{TL}}_l

in RNAM calculated as

{EBV}_{k,{TL}_l}={\hat{a}}_{0,k}+{TL}_l\times {\hat{a}}_{1,k}.

Of the 364 male pigs born in 2022, 100 pigs each of high and low ${\hat{a}}_1$ , namely high- and low- ${\hat{a}}_1$ pigs, respectively, were selected to describe changes with ${TL}_l$ in ${DevADG}_{k,l}$ . They are the most recent birth year pigs and are assumed to be the selection candidates.

Trends in EBVs by birth year were calculated for pigs with growth records.

3 RESULTS AND DISCUSSION

3.1 Basic statistics

The basic statistics of the growth performance records are presented in Table 1. The range of BW and age at the end of performance testing were 65–151 kg and 117–200 days, respectively.

TABLE 1. Basic growth performance record statistics.

Trait (unit)	Sex	N	Mean	SD	Minimum	Maximum
BW (kg)	Male	27,426	103.3	12.1	67.0	151.0
BW (kg)	Female	27,324	102.5	10.9	65.0	144.0
Age (days)	Male	27,426	147.7	10.5	117	198
Age (days)	Female	27,324	154.6	10.0	125	200
ADG (g/day)	Male	27,426	699.5	69.3	469.3	947.8
ADG (g/day)	Female	27,324	663.6	63.5	450.5	887.3

Note: BW, body weight; Age, age in days at the time of BW measurement; ADG, average daily gain.

3.2 Average cumulative TL

Changes in AIC for LM with respect to the model without TL are shown as a heatmap (Figure 1). The model incorporating the TL for the 8-week (56-day) period and a temperature threshold of 25°C had the lowest AIC; these parameters were identified as the optimal high-heat environment indicators. In this case, the mean and maximum TL values were 1.2 and 8.0, respectively. The solution for the partial regression coefficient of TL was −4.7 g/day. In the constructed model, the ADG decreased as TL increased.

Details are in the caption following the image — **FIGURE 1**
Open in figure viewer PowerPoint

Changes in Akaike's information criterion (AIC) of the linear model for the average cumulative thermal load (TL) changing the length of the cumulative period in weeks and the threshold temperature.

Note: TL, cumulative value of the maximum temperatures that exceeded the threshold temperature for the cumulative period before the end of performance testing; noThr, TL calculated without threshold temperature. The AIC of the model without TL was set to 0. The area framed in yellow is the minimum AIC.

Some information about high-heat environment indicators in purebred Duroc pigs has been previously reported. Ogawa et al. (2022) reported a daily average temperature threshold of 24.2°C and a 50-day period before reaching 105 kg. Fragomeni et al. (2016) reported a temperature threshold value of 23.7°C (the daily maximum THI was converted to the temperature at 50% relative humidity) and a 30-day period before the end of performance testing. Compared with previous studies, the difference in the threshold temperature was approximately 1°C; however, the period was considerably longer.

The distribution of TL by farm in 2022 is shown in Figure 2, and the basic statistics of TL and E(D105|TL) in 2022 are shown in Table 2. The duration and intensity of TL differed across farms, increasing around June on farm A, approximately 2 months later than that on the other farms, and decreasing around November on all farms. The intensity of TL differed considerably by farm region. Farm A is located in the northern part of Japan (Hokkaido Prefecture), and two other farms are located in the central part of Japan (Honshu Island). $E\left(D105| TL\right)$ values changed owing to ${\hat{FYS}}_i$ and TL. The TL calculated from the weather records of 2022 for each farm suggested that shipment may have been delayed by up to 1.6, 7.5, and 5.7 days from TL = 0 (corresponding to −8.0, −33.4, and −27.7 g/days in ADG) because of TL on farms A, B, and C, respectively. Notably, Ogawa et al. (2022) investigated the effect of the daily average temperature on D105 in 2020 and reported a possible shipping delay of up to 9 days on a farm south of the farms investigated in the present study.

TABLE 2. Average cumulative thermal load (TL) calculated for each farm based on estimated maximum temperatures in 2022, basic statistics for expected age in days at 105 kg [

E\left(D105| TL\right)

] and expected shipment delay in days.

Farm	A	B	C	A	B	C
	TL (°C)			$\boldsymbol{E}\left(\boldsymbol{D}\mathbf{105}\|\boldsymbol{TL}\right)$ (age in days)
Mean	0.3	1.8	1.4	145.7	151.5	144.8
SD	0.5	2.4	1.9	5.9	4.6	4.3
Minimum (min)	0.0	0.0	0.0	139.7	143.3	139.1
Maximum (max)	1.7	7.1	5.9	156.1	159.7	152.0
				Calculated average daily gain (ADG) (g/days)
TL condition [calculation formula]				A	B	C
Mean [ $105000/E\left(D105\|\ TL= Mean\right)$ ]				720.7	693.1	725.1
Min [ $ADG\left(\mathrm{TL}=\mathrm{Mean}\right)-{\hat{b}}_1\times \mathrm{TL}\left(\mathrm{Mean}\right)$ ]				722.1	701.5	731.7
Max [ $ADG\left(\mathrm{TL}=\operatorname{Min}\right)+{\hat{b}}_1\times \mathrm{TL}\left(\mathit{\operatorname{Max}}\right)$ ]				714.1	668.2	704.0
Effect of max TL [ ${\hat{b}}_1\times \mathrm{TL}\left(\mathit{\operatorname{Max}}\right)$ ]				−8.0	−33.4	−27.7
				Calculated $\boldsymbol{E}\left(\boldsymbol{D}\mathbf{105}\|\boldsymbol{TL}\right)$ (age in days)
TL condition [calculation formula]				A	B	C
Min [ $105000/ ADG\left(\mathrm{TL}=\operatorname{Min}\right)$ ]				145.4	149.7	143.5
Max [ $105000/ ADG\left(\mathrm{TL}=\operatorname{Max}\right)$				147.0	157.1	149.2
				Expected shipment delay (days)
Calculation formula				A	B	C
$E\left(D105\| TL=\mathit{\operatorname{Max}}\right)-E\left(D105\|\ TL=\mathit{\operatorname{Min}}\right)$				1.6	7.5	5.7

Note: Farm A, Kita-Hiroshima City, Hokkaido Prefecture, 43.0°N; Farm B, Nagaoka City, Niigata Prefecture, 37.4°N; and Farm C, Shibukawa City, Gumma Prefecture, 36.5°N (Only the farms with pig records are shown); TL, average cumulative value of the maximum temperatures which exceeded 25°C in the on-farm estimated weather records for the 8 weeks before the end of performance testing; ${\hat{b}}_1$ , the solution of the partial regression coefficient of TL (−4.7).

Records from 2001 to 2022 were used in the present study. A total of 22,043 records had a TL of 0, and 32,707 records had a TL > 0. The records with a TL > 0 accounted for approximately 60%, with an average TL value of 2.0. This result suggests that many pigs in Japan may have already experienced delayed shipment because of reduced ADG under high-heat environments. Therefore, improving heat tolerance in pigs is an urgent issue.

3.3 Estimates of variance components

The estimates of variance components are presented in Table 3. All estimates of ${\sigma}_{a0}^2$ , ${\sigma}_{a1}^2$ , and ${\sigma}_{a0a1}$ significantly differed from 0 (95% confidence interval). Estimated total genetic variance ( ${\sigma}_{a0}^2+2\times {TL}_l\times {\sigma}_{a0a1}+{TL}_l^2\times {\sigma}_{a1}^2$ ) ranged within 298 (TL of 2.6) to 1376 (TL of 8) (g/day)².

TABLE 3. Estimates of the variance components for average daily gain (g/day).

Parameter/item	Genetic variance			Error variance			Heritability and genetic correlation
Parameter	${\boldsymbol{\sigma}}_{\boldsymbol{a}\mathbf{0}}^{\mathbf{2}}$	${\boldsymbol{\sigma}}_{\boldsymbol{a}\mathbf{0}\boldsymbol{a}\mathbf{1}}$	${\boldsymbol{\sigma}}_{\boldsymbol{a}\mathbf{1}}^{\mathbf{2}}$	${\boldsymbol{\sigma}}_{\boldsymbol{e}}^{\mathbf{2}}$			${\boldsymbol{r}}_{\boldsymbol{a}\mathbf{0}\boldsymbol{a}\mathbf{1}}$
Estimated values	537	−93	36	505			−0.67
SE	18	5	2	8			0.02
Parameter	$\boldsymbol{\operatorname{var}}\left({\boldsymbol{a}}_{\boldsymbol{TL}}\right)$
TL	0	1	2	3	4	5	6	7	8
Estimated values	537	387	309	305	373	514	729	1016	1376
SE	18	14	13	14	17	24	36	52	73
Parameter	${\boldsymbol{h}}_{\boldsymbol{TL}}^{\mathbf{2}}$
TL	0	1	2	3	4	5	6	7	8
Estimated values	0.52	0.43	0.38	0.38	0.42	0.50	0.59	0.67	0.73
SE	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01

Note: TL, average cumulative thermal load (°C); ${\sigma}_{a0}^2$ , additive genetic variance for ${a}_0$ ; ${\sigma}_{a1}^2$ , additive genetic variance for ${a}_1$ ; ${\sigma}_{a0a1}$ , additive genetic covariance for ${a}_0$ and ${a}_1$ ; $\mathit{\operatorname{var}}\left({a}_{TL}\right)=\mathit{\operatorname{var}}\left({a}_0+ TL\times {a}_1\right)$ ; ${h}_{TL}^2$ , heritability at TL; ${\upsigma}_e^2$ , random error variance; ${r}_{a0a1}$ , genetic correlation for ${a}_0$ and ${a}_1$ .

The estimates of the heritability of the ADG in purebred Duroc pigs are moderate, ranging from 0.32 to 0.50 (Do et al., 2013; Ogawa et al., 2022, 2023; Suzuki et al., 2005). The change in heritability estimated using RNAM under varying TL conditions ranged from 0.38 (TL of 2.6) to 0.73 (TL of 8). For TL values of 0 to 2.6, estimated heritability decreased with increasing TL values. This was because the decrease caused by the negative genetic covariance of ${a}_0$ and ${a}_1$ in response to TL ( $2\times TL\times {\sigma}_{a0a1}$ ) exceeded the increase in the genetic variance of ${a}_1$ in response to TL ( ${TL}^2\times {\sigma}_{a1}^2$ ). For TL values > 2.6, heritability estimates increased with TL values. This was because the increase in genetic variance of ${a}_1$ in response to TL was greater than the decrease in additive genetic covariance in response to TL. Similar to the present study, Fragomeni et al. (2016) reported that the estimated heritability varied with high-heat environmental indicator values in a U-shape, with higher heritability at higher indicator values. They attributed this result to the fact that the genetic ability at high-heat environments has not been selected for under directional selection; thus, the additive genetic variance is retaining. As the expression of the ${\mathrm{a}}_1$ effect is dependent on TL conditions, the results suggest the feasibility of genetically improving heat tolerance using ${\mathrm{a}}_1$ in the evaluation model. However, the estimated genetic correlation between ${\mathrm{a}}_0$ and ${\mathrm{a}}_1$ was negative ( ${r}_{{\mathrm{a}}_0{\mathrm{a}}_1}=-0.67$ ). Effective and simultaneous improvement of both ${\mathrm{a}}_0$ and ${\mathrm{a}}_1$ would be challenging. Similarly, a negative genetic correlation was also estimated in a previous study involving reaction-norm sire model analysis of crossbred pig carcass weight (Zumbach et al., 2008a).

The estimated genetic correlations for ADG under different TL conditions are shown in Figure 3. Genetic correlation became lower as the difference between the two TL values increased. The minimum genetic correlation was −0.2 for the minimum (0) and maximum (8) TL values. Mulder et al. (2006) suggested that G × E should be considered in selection when the genetic correlation between different environments is less than 0.6. In the present study, the estimated genetic correlations were below this value when the differences between TLs were large. Regarding the genetic correlation between different indicators of high-heat environments, Usala et al. (2021) reported a case in which the genetic correlation was less than 0.6 for pig carcass ADG.

3.4 Estimated breeding value

The ${\hat{a}}_0$ and ${\hat{a}}_1$ of pigs born in 2022 ranged from −10.7 to 72.8 and from −13.0 to 0.22, respectively. The changes in DevADG with TL in pigs born in 2022 are shown in Figure 4. The population means of ADG (thick black line) decreased with increasing TLs. Low- ${\hat{a}}_1$ pigs (blue dotted line) tended to have higher DevADGs than those of high- ${\hat{a}}_1$ pigs (red line) at TL = 0; however, DevADGs decreased considerably with increasing TL. In contrast, high- ${\hat{a}}_1$ pigs (red line) tended to have lower DevADGs than those of low- ${\hat{a}}_1$ pigs at TL = 0; however, their declining rates with increasing TL were slower than those of low- ${\hat{a}}_1$ pigs. The change in the ranking of pigs with varying TLs suggests the need to evaluate EBV using a model that effectively considers G × E. Simms (2000) stated that the slope term quantifies tolerance in RNM. Similarly, in the present study, heat tolerance was also expressed by the slope term in the RNAM. Therefore, to genetically improve heat tolerance, it is crucial to evaluate ${\hat{a}}_1$ in RNAM and consider it in selection.

The trends in EBVs by birth year are shown in Figure 5. The values of ${\hat{a}}_0$ exhibited an increasing trend, whereas ${\hat{a}}_1$ showed a decreasing trend. [Correction added on 12 June 2024, after first online publication: Preceding sentence has been amended for correctness.] However, there is variation in the ${\hat{a}}_1$ values for pigs born in 2022, suggesting that genetically superior heat-tolerant pigs may still be present in the current Duroc pig population. The Intergovernmental Panel on Climate Change (2022) predicts that global warming will worsen. As global warming becomes more severe and economic losses increase, heat-tolerant selection in pigs is expected to become more important. To prevent economic losses and secure heat-tolerant pig genetic resources, incorporating genetic evaluation for heat tolerance into breeding programs is necessary.

3.5 Future perspective

The present study focused on purebred Duroc pigs; however, other breeds such as Landrace and Large White should also be analyzed. As modern commercial pork production is largely supported by a crossbreeding system, the utilization of crossbred pig records also remains a future challenge. Furthermore, genetic relationships between other production or reproductive traits should be clarified for comprehensive breeding (e.g., Bloemhof et al., 2013; Fragomeni et al., 2016; Freitas et al., 2021, 2023; Sevillano et al., 2016; Tiezzi et al., 2020; Usala et al., 2021; Zumbach et al., 2008a).

ACKNOWLEDGMENTS

This research forms part of the results of a project funded by the Japan Central Racing Association for the Promotion of Livestock Production, “Development of Global Warming-Adaptive Swine Breeding Methods Project.”

The authors would like to thank Mr. Kotaro Akachi, President of Akachi Genetics, Inc., Mr. Kazuyuki Tashima, and other members of Global Pig Farms, Inc., and related farms for their cooperation in collecting data for this study.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no conflict of interest.

ETHICS STATEMENT

The study did not require approval from the Animal Care and Use Committee as the data were obtained from an existing database maintained by a private breeding company.

Open Research

DATA AVAILABILITY STATEMENT

The pig records used in this study are properties of the industry partners that contributed to the study and, therefore, are not available owing to their commercial sensitivity. Kindly contact Agro-Meteorological Grid Square Data (https://amu.rd.naro.go.jp/) for information on using weather records.

Supporting Information

REFERENCES

Aguilar, I., Fernandez, E. N., Blasco, A., Ravagnolo, O., & Legarra, A. (2020). Effects of ignoring inbreeding in model-based accuracy for BLUP and SSGBLUP. Journal of Animal Breeding and Genetics, 137(4), 356–364. https://doi.org/10.1111/jbg.12470
10.1111/jbg.12470
PubMed Web of Science® Google Scholar
Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19, 716–723. https://doi.org/10.1109/TAC.1974.1100705
10.1109/TAC.1974.1100705
CAS PubMed Web of Science® Google Scholar
Bloemhof, S., Mathur, P. K., Knol, E. F., & van der Waaij, E. H. (2013). Effect of daily environmental temperature on farrowing rate and total born in dam line sows. Journal of Animal Science, 91(6), 2667–2679. https://doi.org/10.2527/jas.2012-5902
10.2527/jas.2012-5902
CAS PubMed Web of Science® Google Scholar
Bloemhof, S., van der Waaij, E. H., Merks, J. W., & Knol, E. F. (2008). Sow line differences in heat stress tolerance expressed in reproductive performance traits. Journal of Animal Science, 86(12), 3330–3337. https://doi.org/10.2527/jas.2008-0862
10.2527/jas.2008-0862
CAS PubMed Web of Science® Google Scholar
Cardoso, F. F., & Tempelman, R. J. (2012). Linear reaction norm models for genetic merit prediction of Angus cattle under genotype by environment interaction. Journal of Animal Science, 90(7), 2130–2141. https://doi.org/10.2527/jas.2011-4333
10.2527/jas.2011-4333
CAS PubMed Web of Science® Google Scholar
Do, D. N., Strathe, A. B., Jensen, J., Mark, T., & Kadarmideen, H. N. (2013). Genetic parameters for different measures of feed efficiency and related traits in boars of three pig breeds. Journal of Animal Science, 91(9), 4069–4079. https://doi.org/10.2527/jas.2012-6197
10.2527/jas.2012-6197
CAS PubMed Web of Science® Google Scholar
Fragomeni, B. O., Lourenco, D. A., Tsuruta, S., Andonov, S., Gray, K., Huang, Y., & Misztal, I. (2016). Modeling response to heat stress in pigs from nucleus and commercial farms in different locations in the United States. Journal of Animal Science, 94(11), 4789–4798. https://doi.org/10.2527/jas.2016-0536
10.2527/jas.2016-0536
CAS PubMed Web of Science® Google Scholar
Fragomeni, B. O., Lourenco, D. A., Tsuruta, S., Bradford, H. L., Gray, K. A., Huang, Y., & Misztal, I. (2016). Using single-step genomic best linear unbiased predictor to enhance the mitigation of seasonal losses due to heat stress in pigs. Journal of Animal Science, 94(12), 5004–5013. https://doi.org/10.2527/jas.2016-0820
10.2527/jas.2016-0820
CAS PubMed Web of Science® Google Scholar
Freitas, P. H. F., Johnson, J. S., Chen, S., Oliveira, H. R., Tiezzi, F., Lázaro, S. F., Huang, Y., Gu, Y., Schinckel, A. P., & Brito, L. F. (2021). Definition of environmental variables and critical periods to evaluate heat tolerance in Large White pigs based on single-step genomic reaction norms. Frontiers in Genetics, 12, 717409. https://doi.org/10.3389/fgene.2021.717409
10.3389/fgene.2021.717409
CAS PubMed Web of Science® Google Scholar
Freitas, P. H. F., Johnson, J. S., Tiezzi, F., Huang, Y., Schinckel, A. P., & Brito, L. F. (2023). Genomic predictions and GWAS for heat tolerance in pigs based on reaction norm models with performance records and data from public weather stations considering alternative temperature thresholds. Journal of Animal Breeding and Genetics, 00, 1–21. https://doi.org/10.1111/jbg.12838
10.1111/jbg.12838
Google Scholar
Hara, H., Ogawa, S., Ohnishi, C., Ishii, K., Uemoto, Y., & Satoh, M. (2022). An attempt of using public ambient temperature data in swine genetic evaluation for litter size traits at birth in Japan. Animal Production Science, 62(15), 1488–1500. https://doi.org/10.1071/AN21463
10.1071/AN21463
Web of Science® Google Scholar
Intergovernmental Panel on Climate Change. (2022). Climate change 2022: Impacts, adaptation and vulnerability working group II contribution to the sixth assessment report of the intergovernmental panel on climate change. Cambridge University Press.
Google Scholar
Japan Meteorological Agency. (2024). Automated meteorological data acquisition system (AMeDAS) [homepage on the internet]. Japan Meteorological Agency. [cited 14 February 2024]. Available from URL: https://www.jma.go.jp/jma/kishou/know/amedas/kaisetsu.html
Google Scholar
Liu, F., Zhao, W., Le, H. H., Cottrell, J. J., Green, M. P., Leury, B. J., Dunshea, F. R., & Bell, A. W. (2022). Review: What have we learned about the effects of heat stress on the pig industry? Animal, 16(Supplement 2), 100349. https://doi.org/10.1016/j.animal.2021.100349
10.1016/j.animal.2021.100349
PubMed Google Scholar
Lourenco, D., Tsuruta, S., Masuda, Y., Bermann, M., Legarra, A., & Misztal, I. (2022). Recent updates in the BLUPF90 software suite. In Proceedings of the 12th world congress on genetics applied to livestock production (pp. 1530–1533). Wageningen Academic Publishers. https://doi.org/10.3920/978-90-8686-940-4_366
10.3920/978-90-8686-940-4_366
Google Scholar
Mayorga, E. J., Renaudeau, D., Ramirez, B. C., Ross, J. W., & Baumgard, L. H. (2019). Heat stress adaptations in pigs. Animal Frontiers, 9(1), 54–61. https://doi.org/10.1093/af/vfy035
10.1093/af/vfy035
PubMed Web of Science® Google Scholar
Meyer, K., & Houle, D. (2013). Sampling based approximation of confidence intervals for functions of genetic covariance matrices. Proceedings of the Twentieth Conference of the Association for the Advancement of Animal Breeding and Genetics, NZL, 20, 523–526. Available at http://www.aaabg.org/aaabghome/AAABG20papers/meyer20523.pdf
Google Scholar
Mulder, H. A., Veerkamp, R. F., Ducro, B. J., van Arendonk, J. A., & Bijma, P. (2006). Optimization of dairy cattle breeding programs for different environments with genotype by environment interaction. Journal of Dairy Science, 89(5), 1740–1752. https://doi.org/10.3168/jds.S0022-0302(06)72242-1
10.3168/jds.S0022-0302(06)72242-1
CAS PubMed Web of Science® Google Scholar
Ogawa, S., Kimata, M., Tomiyama, M., & Satoh, M. (2022). Heritability and genetic correlation estimates of semen production traits with litter traits and pork production traits in purebred Duroc pigs. Journal of Animal Science, 100(3), skac055. https://doi.org/10.1093/jas/skac055
10.1093/jas/skac055
PubMed Web of Science® Google Scholar
Ogawa, S., Ohnishi, C., & Satoh, M. (2022). Effect of ambient temperature on average daily gain of pigs evaluated using public weather data and a plateau-linear regression model. Animal Science Journal, 93(1), e13762. https://doi.org/10.1111/asj.13762
10.1111/asj.13762
PubMed Web of Science® Google Scholar
Ogawa, S., Takahashi, H., & Satoh, M. (2023). Genetic parameter estimation for pork production and litter performance traits of Landrace, Large White, and Duroc pigs in Japan. Journal of Animal Breeding and Genetics, 140(6), 607–623. https://doi.org/10.1111/jbg.12814
10.1111/jbg.12814
PubMed Web of Science® Google Scholar
Ohno, H., Sasaki, K., Ohara, G., & Nakazono, K. (2016). Development of grid square air temperature and precipitation data compiled from observed, forecasted, and climatic normal data. Climate in Biosphere, 16, 71–79. (in Japanese with English title). https://doi.org/10.2480/cib.J-16-028
10.2480/cib.J-16-028
Google Scholar
Ramirez, B. C., Hoff, S. J., Hayes, M. D., Brown-Brandl, T., Harmon, J. D., & Rohrer, G. A. (2022). A review of swine heat production: 2003 to 2020. Frontiers in Animal Science, 3, 908434. https://doi.org/10.3389/fanim.2022.908434
10.3389/fanim.2022.908434
Google Scholar
Sasaki, Y., Fujie, M., Nakatake, S., & Kawabata, T. (2018). Quantitative assessment of the effects of outside temperature on farrowing rate in gilts and sows by using a multivariate logistic regression model. Animal Science Journal, 89(8), 1187–1193. https://doi.org/10.1111/asj.13048
10.1111/asj.13048
CAS PubMed Web of Science® Google Scholar
Sasaki, K., Nishimori, M., & Nemoto, M. (2022). User's manual for The Agro-Meteorological Grid Square Data [PDF File]. In National Agriculture and Food Research Organization. National Agriculture and Food Research Organization. (In Japanese). [cited 23 October 2023]. Available from: https://amu.rd.naro.go.jp/wiki_open/doku.php?id=docs
Google Scholar
Satoh, M., Fukuzawa, Y., Ogawa, S., & Okamura, T. (2023). Estimation of off-farm temperatures and temperature-humidity index from meteorological observation records. Animal Science Journal, 94, e13902. https://doi.org/10.1111/asj.13902
10.1111/asj.13902
PubMed Web of Science® Google Scholar
Satoh, M., Kusaka, K., Ogawa, S., & Uemoto, Y. (2023). Estimation of off-farm temperatures from meteorological observation records and their level of accuracy. Nihon Chikusan Gakkaiho, 94, 277–282. https://doi.org/10.2508/chikusan.94.277
10.2508/chikusan.94.277
Google Scholar
Seabold, S., & Perktold, J. (2010). Statsmodels: Econometric and statistical modeling with Python. In Proceedings of the 9th python in science conference (pp. 92–96). SciPy. https://doi.org/10.25080/Majora-92bf1922-011
10.25080/Majora-92bf1922-011
Google Scholar
Sevillano, C. A., Mulder, H. A., Rashidi, H., Mathur, P. K., & Knol, E. F. (2016). Genetic variation for farrowing rate in pigs in response to change in photoperiod and ambient temperature. Journal of Animal Science, 94(8), 3185–3197. https://doi.org/10.2527/jas.2015-9915
10.2527/jas.2015-9915
CAS PubMed Web of Science® Google Scholar
Simms, E. L. (2000). Defining tolerance as a norm of reaction. Evolutionary Ecology, 14, 563–570. https://doi.org/10.1023/A:1010956716539
10.1023/A:1010956716539
Web of Science® Google Scholar
Suzuki, K., Irie, M., Kadowaki, H., Shibata, T., Kumagai, M., & Nishida, A. (2005). Genetic parameter estimates of meat quality traits in Duroc pigs selected for average daily gain, longissimus muscle area, backfat thickness, and intramuscular fat content. Journal of Animal Science, 83(9), 2058–2065. https://doi.org/10.2527/2005.8392058x
10.2527/2005.8392058x
CAS PubMed Web of Science® Google Scholar
Takada, R., Yamazaki, M., Sugiura, T., Yokozawa, M., Otuka, M., & Murakami, H. (2008). Future prediction of the effect of global warming on a pig growth performance estimated from changes of the mean ambient temperature. Nihon Chikusan Gakkaiho, 79(1), 59–65. (In Japanese with English summary). https://doi.org/10.2508/chikusan.79.59
10.2508/chikusan.79.59
Google Scholar
Tiezzi, F., Brito, L. F., Howard, J., Huang, Y. J., Gray, K., Schwab, C., Fix, J., & Maltecca, C. (2020). Genomics of heat tolerance in reproductive performance investigated in four independent maternal lines of pigs. Frontiers in Genetics, 11, 629. https://doi.org/10.3389/fgene.2020.00629
10.3389/fgene.2020.00629
CAS PubMed Web of Science® Google Scholar
Usala, M., Macciotta, N. P. P., Bergamaschi, M., Maltecca, C., Fix, J., Schwab, C., Shull, C., & Tiezzi, F. (2021). Genetic parameters for tolerance to heat stress in crossbred swine carcass traits. Frontiers in Genetics, 11, 612815. https://doi.org/10.3389/fgene.2020.612815
10.3389/fgene.2020.612815
PubMed Web of Science® Google Scholar
Via, S., Gomulkiewicz, R., De Jong, G., Scheiner, S. M., Schlichting, C. D., & Van Tienderen, P. H. (1995). Adaptive phenotypic plasticity: Consensus and controversy. Trends in Ecology & Evolution, 10(5), 212–217. https://doi.org/10.1016/s0169-5347(00)89061-8
10.1016/S0169-5347(00)89061-8
CAS PubMed Web of Science® Google Scholar
Zumbach, B., Misztal, I., Tsuruta, S., Sanchez, J. P., Azain, M., Herring, W., Holl, J., Long, T., & Culbertson, M. (2008a). Genetic components of heat stress in finishing pigs: Parameter estimation. Journal of Animal Science, 86(9), 2076–2081. https://doi.org/10.2527/jas.2007-0282
10.2527/jas.2007-0282
CAS PubMed Web of Science® Google Scholar
Zumbach, B., Misztal, I., Tsuruta, S., Sanchez, J. P., Azain, M., Herring, W., Holl, J., Long, T., & Culbertson, M. (2008b). Genetic components of heat stress in finishing pigs: Development of a heat load function. Journal of Animal Science, 86(9), 2082–2088. https://doi.org/10.2527/jas.2007-0523
10.2527/jas.2007-0523
CAS PubMed Web of Science® Google Scholar

Citing Literature

Volume95, Issue1

January/December 2024

e13958

Reaction-norm animal model analysis of average daily gain heat tolerance in purebred Duroc pigs

Abstract

1 INTRODUCTION