ABSTRACT
Wool produced (kg of fleece/shearing) by 371 sheep from the genetic groups of Romney Marsh (31) Criollo Chiapas sheep (135) and their hybrids (205) was analyzed using repeated measures models and random regression analysis. One thousand one hundred and eight production records collected over a period of 12 years, under extensive production conditions in the region of the Altos de Chiapas, Mexico, were analyzed. The genetic groups displayed wool production differences (P < 0.0001), with Romney Marsh showing the highest performance (2.21±0.94 kg of fleece/shearing), the F1 animals were intermediate (1.397±0.07) and the Criollo sheep showed the poorest performance (0.881±0.07). Heterosis estimation for fleece weight (kg) per shearing was -0.1517±0.0543, P = 0.0055. Romney Marsh animals were the most affected by the environmental effect of animal age at shearing since its negative linear slope was four to ten times steeper than the linear slopes of the F1 and the Criollo animals, respectively. The Criollo Chiapas Sheep remained the longest period of time in the flock. The environmental effects of age of animal (P < 0.0001), gender (P = 0.037), number of shearing (P < 0.001) and year (P < 0.0001) were important on wool production. Wool production for the first four years of age was similar, and it decreased from the fifth year on; males were 6.9% superior to females; the first three shearings were similar between them (P > 0.06), but lesser than the last ones (P < 0.05). Criollo sheep showed remarkable environmental adaptation; therefore, the preservation of this animal genetic resource is extremely important for the indigenous community that makes use of it.
Key words: Criollo sheep, heterosis, wool production, random regression.
The use of Criollo sheep in certain regions of Mexico is, in some cases, a viable alternative of production, due to its environmental, nutritional and management adaptation within a cultural, social and economic surrounding; as is the case of livestock in the Altos de Chiapas, Mexico. This medium-sized, double-coated sheep population is associated with the sociocultural traditions of the indigenous Totzil community, in the elaboration of handmade ceremonial wear and every day wear, as it has been described by Perezgrovas and Castro (2000).
The mating of rams and ewes of different breeds or different genetic groups has been widely used to increase reproductive and productive characteristics in the short term (Hassen et al., 2004; Malik and Singh, 2006; Mishra et al., 2007; Ghită, 2007; Kremer et al., 2010). Burfening and Carpio (1995) in Peru, observed that, because of lamb survival, Criollo sheep are well adapted to environmental conditions; however, animal growth and fleece weight can be increased by crossing this breed with specialized breeds, as long as all socioeconomic and production aspects are considered. Nawaz et al. (1992) in Pakistan, analyzed the productive response when crossing Rambouillet rams with Kaghani ewes with different genetic levels, observing a better growth and wool production response from crossbred animals than that of animals from the local breed, suggesting that the best crossbreeding strategy depends on the objective of local production.
The mating of rams and ewes of different breeds or Criollo genetic groups crossed with specialized exotic breeds, as a tool for improving production, has generated the dissolution or loss of local genetic diversity, with a loss estimated by FAO of one breed each two weeks (FAO, 2007; Köhler-Rollefston et al., 2009). Other alternative for improving production is through identification of genetically outstanding animals, used under controlled breeding programs. Castro-Gámez et al. (2008) estimated heritability value of 0.31±0.05 for fleece production in Chiapas sheep, while Gizaw et al. (2007) obtained a value of 0.393±0.016 in Menz sheep of Ethiopia for the same trait.
Due to the aforementioned, the objective of this study was to evaluate fleece production by crossing Romney Marsh rams with Chiapas Criollo Sheep ewes managed in extensive conditions in Chiapas, Mexico.
Animals
The production records of dirty fleece (kg/ewe/shearing) collected from Criollo (CR), Romney Marsh (RM) and the cross between Romney Marsh rams and Criollo ewes (F1), for the period 1983 to 1989 were used. They were obtained from the Centro de Fomento Ovino de Chiapas of the Universidad Autónoma de Chiapas, located in the municipality of Teopisca, Chiapas, at 16°32’24’’ North latitude and 92°28’19’’ West longitude and at 1, 780 ASL. (Secretaría de Gobernación del Estado de Chiapas,1988). The animals were managed extensively in tropical pastures (mainly Pennisetum clandestinum) and had free access to water. Health management of the flock consisted of deworming for the control of gastrointestinal parasites and oxytetracycline administration in case of respiratory disease. Shearing machines were used and dirty fleece weight was recorded for each animal once a year (April).
Database edition
Each animal was identified by genetic group, gender, month and year of birth, shearing number and kg of fleece collected at each shearing. The shearing number was classified from one to six or more. The unidentified records (genetic group or age) were eliminated from the analysis, giving a final number for the analysis of 1108 fleece production records (kg/ewe/shearing) of 371 sheep of the genetic groups: 135 Criollo (119 females, 16 males); 31 Romney Marsh (25 females, 6 males) and 205 from the cross between Romney Marsh males and Criollo females (128 females, 77 males).
Statistical analysis
All fleeces were weighed after each shearing of each animal within its genetic group; for which, the variation between measurements within the same individual, may present homogeneous variances or may differ throughout time and correlated between them. Measurements in the same individual throughout time, are not generally independent, for which, the adequate structure of variances and co-variances for model selection was defined by comparing the statistical data of restricted maximum likelihood: Log likelihood function [log(L)] = -2log (MLk); Akaike information criterion (AIC) AICk = -2log (MLk) + 2pk (Akaike, 1973) and the Bayesian information criterion (BIC): BICk = -2log(MLk) + pk log(n) (Littel et al., 2006).
Analysis of variance for repeated measures
Database edition
The age of the animal ranged from 1 to 12 years. Seven categories were generated for the analysis: initial, animals ≤ 2 years of age; final, ≥ 8 years of age and intermediate (3, 4, 5, 6 and 7 years of age). Fleece weight at each shearing was adjusted to a mixed-model for repeated measures using the MIXED procedure (SAS, 2011), according to the following statistical model:
Yijklmno = µ + Gi + Ej + Sk + Al + Tm + GxEij + animaln + eijklmno
Where: Yijklmno is fleece weight (kg) of the n-th animal measured at the m-th shearing; µ is the overall mean; Gi is the fixed effect of i-th genetic group (i = Criollo, F1, Romney Marsh); Ej is the fixed effect of j-th age-class (j = 2, …, 8); Sk is the fixed effect of k-th gender (k = female, male); AI is the fixed effect of I-th year in which shearing was performed (l = 1983, …, 1989); Tm is the fixed effect of m-th shearing (m = 1,…, 6); GxEij is the fixed effect of the genotype by age class interaction; animaln is the random effect of the n-th animal ~ niid(0, ); and eijklmno is the random error ~ niid(0, ).
The animal was used as subject in the REPEATED statement of the model. Following a similar procedure as suggested by Littell et al. (2000), the model was adjusted to each of the following covariance structures: compound symmetry (CS), heterogeneous compound symmetry (HCS), autoregressive (Type 1 (AR(1)), heterogeneous AR(1), Toeplitz (TOEP), heterogeneous Toeplitz (TOEPH) and unstructured (UN). After fitting the model, Akaike information criterion (AIC), likelihood function (-2 LogResL) and Bayesian information criterion (BIC) generated by PROC MIXED (SAS, 2011) were used to select the covariance structure that generated the best model. Table 1 presents each information criterion, the differences with respect to the generated by the best model, as well as to the number of parameters (that is, variances and covariances) estimated by each model.
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 1.png)
According to the Bayesian information criterion, the best adjustment was generated by including in the model a compound symmetry covariance structure (CS) or a heterogeneous compound symmetry (HCS) structure. However, the CS option was chosen because it involves the estimation of a smaller number of parameters. It must be highlighted that, either AIC as well as the -2ResLogL, showed the unstructured option, as best covariance structure, which is the one that requires the estimation of greater number of parameters. With these results, and using BIC, CS option was the most appropriate to be included in the final analysis of variance.
Once the covariance structure that generated the best adjustment was identified, the final model was adjusted and least squares means were calculated for the effects of genotype, age, gender, shearing year, shearing number and the genotype by animal’s age-class interaction. Heterosis for fleece weight at shearing was calculated using the ESTIMATE statement in PROC MIXED (SAS, 2011), where the fixed effect of the genetic group was used to generate a contrast with the coefficients -0.5, 1.0 and -0.5 for Criollo, F1 and Romney Marsh genotypes, respectively; while the difference of the F1 group with respect to the Criollo animals, was estimated using the contrast 1, -1 and 0, respectively.
Analysis of variance using random regression
The information was also analyzed adjusting Legendre orthogonal polynomials with random regression, using PROC MIXED (SAS, 2011). In this case, all animal age records were used, with a range of 1 to 12 years. The fitted random regression model is:
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 2.png)
Where: ykml is the k-th observation of the fleece weight at shearing, recorded in the m-th animal belonging to the l-th genotype; bi are coefficients from the fixed regression for age at shearing (b0 = intercept, b1 = linear effect, b2 = quadratic effect, and b3 = cubic effect); aim is the i-th random regression coefficient (a0m = intercept, a1m= linear effect, a2m= quadratic effect, and a3m = cubic effect) of the wool production curve per year of age, belonging to the m-th animal (m = 1,…, 368) of the l-th genotype (l = Criollo, F1, Romney Marsh); is the k-th observation of age, standardized, at the moment of shearing, of the m-th animal, belonging to the l-th genotype, raised to the power of 0, 1, 2, or 3; ekml is the error associated with the observation ykml. The standardized unit of time (x) was the age of the animal at the moment of shearing, with a range of -1 to +1, and it was calculated using the following expression:
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 3.png)
Where: t is the age of the animal at the moment of shearing, tmin is the youngest age at which shearing was done, and tmax is the oldest age with a record of shearing. In this study, tmin was 1 and tmax was 12 years. According to Spiegel (1971), the first three Legendre polynomials for the standardized unit of time (x) were:
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 4.png)
The adjustment of the random regression models was done following a similar procedure as suggested by Hanford (2005). The restricted maximum likelihood method was specified in the model statement. The specified covariance type for random effects was unstructured and the individual identification of the animal was used as subject in the RANDOM statement in PROC MIXED (SAS, 2011). In order to select the best-fitted model, different combinations of Legendre polynomials of degree were analyzed, either in the fixed part or in the random part of the regression model. BIC was used as comparison criterion; the best-fitted model consisted of a third degree polynomial in the fixed part of the random regression and a random intercept, as shown in the differences for BIC in Table 2.
Repeated measures
Least squares means for the main effects are shown in Table 3. There were differences between the three genetic groups (P < 0.0001), being the estimate of heterosis calculated, with its standard error, for fleece weight of -0.1517±0.05435, P = 0.0055. This result showed that the average production of wool for crossbred animals, was below the average of the specialized breed individuals, but above Criollo animals 0.5159±0.05 (P < 0.0001). An effect of age was observed, being fleece production similar to the age of four (P > 0.05), decreasing after the age of five (P < 0.05), with males being 6.9% superior in fleece production per shearing than females (P = 0.037), while shearing number (P < 0.001) showed that the first three shearings were similar between them (P > 0.06), but different to the last shearings (P < 0.05). Finally, a fluctuation between the years of study is observed (P < 0.0001), being 1983 and 1986 the more productive years, and 1989 the less productive year. The means of genetic group by age-class interaction are depicted in Figure 1. Romney Marsh (specialized genetic group) decreased (P < 0.0001) fleece production over time (-1.047 kg), while F1 and Criollo groups decreased at a slower rate (-0.2997 and -0.112 kg, respectively).
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 6.png)
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 7.png)
Random regression
The estimates of the regression coefficients for a fixed Legendre polynomial of third order per genotype is shown in Table 4. The differences between genotypes were marked, for the information of F1 genetic group, the adjustment of an intercept (p0) was enough, while for the Romney Marsh group, an intercept and a linear effect (p0 and p1) was required, and for the Criollo group, the adjustment of an intercept and linear, quadratic and cubic effects of the Legendre polynomial were needed (Table 4).
The adjusted curves per genetic group of animals are shown in Figure 2. Thicker lines are the curves per group representing the overall fixed regressions, while thinner lines represent the random regression of each individual animal in the population analyzed. The fitted overall fixed regression line for Romney Marsh is a thick continuous black line, for the F1 individuals it is a broken black line, and for the Criollo genotype it is a continuos grey line. The results showed that the Romney Marsh sheep group showed better performance for fleece production per shearing, the F1 group was intermediate and the Criollo group produced the lower amount of wool. While specialized breed animals produced greater quantity of fleece, these were more affected by the environmental effect of age, as it can be observed by the negative slopes, four times greater with respect to the crossbred group and 10 times more with respect to the Criollo animals. Conversely, there was longer stayability of the Criollo genotype in the flock, as shown by the records of shearing beyond 9 years of age, which was the level of the peak either for F1 or Romney Marsh animals (Figure 2).
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 8.png)
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 9.png)
![](/files/images/IJLP/2015/Dec/Vázquez-Peláez et al 10.png)
The decrease in fleece production per shearing in Romney Marsh, was four and 10 times the amount with respect to F1 and Criollo, respectively. This can be due to an effect of lesser adaptability of Romney Marsh breed. Likewise, the specialized breed and the crossbred group showed lesser permanence in the flock with respect to the Criollo group. There is information about poor adaptation of specialized breeds under conditions where Criollo animals have remained (Alderson et al., 1983) in Colombia and Peru (Burfening and Carpio, 1995) in). Similar to the Criollo sheep from the present study, now regarded as Chiapas Breed, that has its origins in the Spanish sheep breeds Churra, Manchega and Lacha (Mendez-Goméz et al., 2014) sheep from Chiloé´s Archipelago has its origins in the Spanish sheep breeds Churra and Castellana (De la Barra et al., 2011, 2014). Both, the Chiapas and the Chilota sheep breed are well adapted to local environmental conditions because of more than 400 years of natural selection in these respect to F1 and Criollo, respectively. This can be due to an effect of lesser adaptability of Romney Marsh breed. populations.
Martínez et al. (2012) showed that, under the agroecological conditions of the Chiloé archipelago, Chilota sheep breed evidenced a greater adaptation since, in the absence of management practices, it proved more productive and resistant than the Romney Marsh and Suffolk breeds. However, Kremer et al. (2010) in Uruguay did not find differences between breeds concerning respiratory diseases, mortality or longevity of the ewe, maybe because the environmental effect is not as severe.
The specialized Romney Marsh breed showed higher performance in fleece production than the Criollo genetic group (+1.335 kg/ewe/shearing), while F1 was intermediate showing negative value of hybrid vigour (-0.15 ± 0.05), being negative heterosis of 9%. Malik and Singh (2006) found small and insignificant heterosis values in a study using 15 genetic groups, crossing Nali ewes with Russian Merino and Corriedale. While F1 did not exceed the average of their parents, local livestock production did increase. Negative heterosis can be a reflection of a greater additive effect of this trait and of the low averages in all the groups, as product of the environmental effect.
The mixed models of repeated measures and random regression are similar; however, the use of random regression and Legendre orthogonal polynomials enables model fleece production at any time of the productive life of the individual, as well as to estimate the interaction between heterosis and environment (Su et al., 2009). The environmental effects of year, age, gender and shearing number in the animal are sources of important variation on production traits; these results coincide with the literature of Hassen et al.(2004)
The results suggest that crossing Romney Marsh with Criollo ewes under extensive systems of management at the Altos de Chiapas is not as promising as improving fleece production, since they show faster wool production decrease with age and lower stay ability in the flock when compared with F1 and Criollo animals, probably due to poor adaptation. Criollo sheep produces fleece without much variability, which makes it an important animal resource that is necessary to preserve. Random regression models may adequately model fleece production throughout the sheep’s life span.
Special thanks for financial support of projects UNAM-PAPIIT IN207707-3 and UNAM-PAPIIT IN205710-3. This research was possible thanks to the agreement between the Facultad de Medicina Veterinaria y Zootecnia of the Unversidad Nacional Autónoma de México and the Centro Universitario de Investigación y Transferencia de Tecnología of the Universidad Autónoma de Chiapas.
The authors have not declared any conflict of interests.
REFERENCES
Akaike H (1973). Information theory as an extension of the maximum likelihood 14 principle. in: Petrov BN, Csaki F (Eds.), Second international symposium on 15 information theory. Akademiai Kiado, Budapest. pp. 267-281.
|
|
Alderson A, Naranjo A, Kress DD, Burfening PJ, Blackwell RL, Bradford GE (1983). A comparison of four imported breeds of sheep and the native Criollo in Colombia. Int. Goat Sheep Res. 2:3848.
|
|
|
Burfening PJ, Carpio M (1995). Improving Criollo sheep in Peru through crossbreeding. Small Rumin. Res. 17:31-35.
Crossref
|
|
|
Castro-Gamez H, Perezgrovas R, Campos-Montes G, López-Ordaz R, Castillo- Juarez H (2008). Genetic parameters for fleece quality assesed by ancient Tzotzil indigenous evaluation system in Mexico. Small Rumin. Res. 74:107-112.
Crossref
|
|
|
De la Barra RA, Carvajal A, Uribe H, Martínez MM, Gonzalo C, Arraz J, Primitivo SF (2011). El ovino criollo y su potencial productivo, Animal Genetic Resourses FAO 48:93-99.
Crossref
|
|
|
De la Barra R, Martínez ME, Calderón C (2014). Phenotypic featrures and fleece quantitative traits in Chilota sheep breed. Livestock Sci. 5:28-34.
|
|
|
Ghită E (2007). Use of crossbreeding to produce superfine wool IV. Characterization of the hybrids obtained in the third stage of crossing. Archiva Zootechnical 10:102-110.
|
|
|
Gizaw S, Lemma S, Komen H, Van Arendok JAM (2007). Estimates of genetic parameters and genetic trends for live weight and fleece traits in Menz sheep. Small Rumin. Res. 70:145-153.
Crossref
|
|
|
Hanford K (2005). Repeated Measures Data – Random Coefficients Models. Lecture Notes –Statistics 892 – Mixed Models – Spring 2005 P 132. [Accessed: 18-february-2013] 165.
View
|
|
|
Hassen Y, Sölkner J, Fuerst-Waltl B (2004). Body weight of Awassi and indigenous Ethiopian sheep and their crosses. Small Rum. Res. 55:51-56.
Crossref
|
|
|
Köhler-Rollefson I, Rathore HS, Mathias E (2009). Local breeds, livelihoods and livestock keepers' rights in South Asia. Trop. Anim. Health Prod. 41(7):1061-1070.
Crossref
|
|
|
Kremer R, Barbato G, Rista L, Rosés L, Perdigón F (2010). Reproduction rate, Milk and wool production of corridale and East Friesian x Corridale F1 ewes grazing on natural pastures. Small Rumin. Res. 90:27-33.
Crossref
|
|
|
Littell RC, Pendergast J, Natarajan R (2000). Modelling covariance structure in the analysis of repeated measures data. Stat. Med. 19:1793-1819.
Crossref
|
|
|
Littel RC, George AM, Walter WS, Russell DW, Oliver S (2006). SAS© for mix models. Second Edition. Cary, NC: 16 SAS Institute Inc.
|
|
|
Malik BS, Singh RP (2006). Evaluation of crossbreeding effects for wool traits in sheep. Asian-Aust. J. Anim. Sci. 19(11):1536-1540.
|
|
|
Martínez ME, Calderón C, Uribe H, De la Barra R (2012). Effect of management practices in the productive performance of three sheep breeds in the Chiloé Archipielago Chile. J. Livestock Sci. 3:57-66.
|
|
|
Mendez-Gómez AC, López-Ordaz R, Peralta-Lailson M, Ulloa-Arvizu R, Pedraza-Villagómez P, Ruiz-López FJ, Berruecos-Villalobos JM, Vásquez-Peláez CG (2014). Estimación de heredabilidad de la curva de crecimiento en el Borrego de raza Chiapas en México. Animal Genetic Resources, FAO 1-7.
Crossref
|
|
|
Mishra AK, Arora AL, Kumar S, Kumar S, Singh VK (2007). Improving productivity of Malpura breed by crossbreeding with prolific Garole sheep in India. Small Rumin. Res. 70:159-164.
Crossref
|
|
|
Nawaz M, Meyer HH, Jadoon JK, Naqvi MA (1992). Results from adaptability trial of Rambouillet sheep and their crossbreeding with Kaghanis. Effects on ewe mating weight, wool production, litter size and lamb growth. AJAS 5:481-485.
Crossref
|
|
|
Perezgrovas GR, Castro GH (2000). Chiapas sheep and the traditional sheep management system of Tzotzil shepherdesses. Arch. Zootec. 49:391-403.
|
|
|
SAS Institute (2011). SAS User's Guide. Statistics, Version 9.3. SAS Inst., Inc., Cary, NC.
|
|
|
Spiegel MR (1971). Advanced Mathematics for Engineers and Scientists. McGraw-Hill, New York, NY.
|
|
|
Su G, Madsen P, Lund MS (2009). Reaction norm model with unknown environmental covariate to analyze heterosis by environment interaction. J. Dairy Sci. 92:2204-2213.
Crossref
|
|