International Journal of
Livestock Production

  • Abbreviation: Int. J. Livest. Prod.
  • Language: English
  • ISSN: 2141-2448
  • DOI: 10.5897/IJLP
  • Start Year: 2009
  • Published Articles: 264

Full Length Research Paper

Milk composition of Chiapas sheep breed under grazing conditions

Karin Nicole Carrillo-Pineda
  • Karin Nicole Carrillo-Pineda
  • Departamento de Genetica y Bioestadistica, Facultad de Medicina Veterinaria y Zootecnia, Universidad Nacional Autónoma de Mexico. Av Universidad 3000, Ciudad Universitaria Mexico 01710 D.F.
  • Google Scholar
Reyes Lopez-Ordaz
  • Reyes Lopez-Ordaz
  • Departamento de Producción Agrícola y Animal, Unidad Xochimilco, Universidad Autónoma Metropolitana, Calz del Hueso 1100, Coyoacán, Villa Quietud, 04960 Ciudad de México, D.F.
  • Google Scholar
Adimelda del Carmen Mendez-Gomez
  • Adimelda del Carmen Mendez-Gomez
  • Facultad de Medicina Veterinaria y Zootecnia, Universidad Autonoma de Chiapas. KM 8, carretera Teran, Ejido Emiliano Zapata, Tuxtla Gutierrez Chiapas, Mexico.
  • Google Scholar
Raúl Ulloa-Arvizu
  • Raúl Ulloa-Arvizu
  • Departamento de Genetica y Bioestadistica, Facultad de Medicina Veterinaria y Zootecnia, Universidad Nacional Autónoma de México. Av Universidad 3000, Ciudad Universitaria Mexico 01710 D.F.
  • Google Scholar
Carlos Gustavo Vázquez-Peláez
  • Carlos Gustavo Vázquez-Peláez
  • Departamento de Genetica y Bioestadistica, Facultad de Medicina Veterinaria y Zootecnia, Universidad Nacional Autónoma de México. Av Universidad 3000, Ciudad Universitaria Mexico 01710 D.F.
  • Google Scholar

  •  Received: 04 November 2014
  •  Accepted: 12 January 2015
  •  Published: 20 February 2015


The objective of this study was to estimate and model the percentage of protein, fat, lactose, non-fatty solids and protein-fat relationship variation in a 90-day milking period, from 136 test-day milk yield records of 46 Chiapas ewes. Least square means were estimated using a mixed model with repeated measures considering year (2006 to 2007), lactation (2, 3, 4, 5), variety (white, brown and black) and the interaction between lactation and variety. The relationship between days of lactation and daily milk yield (ml) and composition was modeled using random regression techniques. Least square means were 14.2 ± 0.36 kg for milk yield per lactation, 169.12 ± 4.97 ml/ewe/day, 5.49 ± 0.04% for protein, 4.37 ± 0.17% for fat, 4.53 ± 0.03% for lactose, 11.08 ± 0.04% for non-fat solids and 1.56 ± 0.07 for protein-fat relationship. Daily milk yield showed constant decreasing, while milk components presented quadratic trend during milking period. The component percentages of protein, fat, lactose, non-fatty solids and protein: fat relationship remained constant during the first five lactations and varieties showed similarity between milk composition studied traits, except in fat content, where the white variety had the highest proportion and the black variety the lowest, with a difference of 30%, whereas the brown variety was intermediate between these two. The results of the present study show the feasibility of selecting the Chiapas sheep breed for milk production and for a dual-purpose animal (wool-milk) under grazing conditions in the Altos de Chiapas, Mexico.


Key words: Protein in sheep milk, fat in sheep milk, lactose in sheep milk, Composition of sheep milk, Chiapas sheep breed, modeling milk composition, random regression.


Sheep milk production worldwide has been used for consumption, cheese and yoghurt. Elaboration through local breeds with particular production standards has a direct relationship between milk composition and final products, such is the case of the Sarda breed for production of ricotta cheese; Lacaune, for Roquefort cheese; Manchega, for manchego cheese; Latxa, for Idiazabal cheese; Saloia, for Nisa cheese, etc., (Scintu and Piredda, 2007). Composition in sheep milk has been widely studied (Table 1) and variation has been attributed to different effects such as breed, lactation, management, among others for protein (5.13 to 6.6%), fat (4.68 to 11.8%), lactose (3.7 to 5.3%) and non-fatty solids (10.33 to 21.24%). Protein percentage is two times higher than goat or cow (Jandal, 1996) and fat, protein, ashes, non-fatty solids and total solids are superior than goat, cow and human milk; lactose content however is superior to goat milk, but inferior to cow and human milk (Pandy and Ghodke, 2007). In the High land of Chiapas, Mexico, a production system prevails based on a local breed, defined as Chiapas sheep breed (Pedraza et al., 1992; FAO, 2009), with three colour phenotypes (white, brown and black), which have been maintained as a separate closed flock. Traditional management is done by the indigene of the region, mainly for wool production (Perezgrovas and Castro, 2000). Studies on milk yield show that the Chiapas sheep breed can be considered for milk and cheese production. Peralta et al. (2005), using nonlinear models and Vázquez et al. (2014), using different random regression models, characterized the lactation curve of this breed. Perezgrovas and Castro (2000), presented from a random sample of ewes of this breed has a range of 5.5 to 5.9% for protein, 5.8 to 5.9% for fat, 4.3 to 4.6% for lactose and 16.7 to 17% for non-fatty solids. Pedraza and Peralta (2003) mentioned that there is a relationship of 4:1 l/kg of produced cheese. Raynal-Ljutovac et al. (2008) mentioned that the effects of lactation status, season of the year, breed, genotype and nutrition are important factors to be considered in sheep milk production. Therefore, the objective of the present study was to model the variation in protein, fat, lactose, non-fatty solids and protein: fat relationship during lactation and other environmental effects in Chiapas sheep breed, using random regression techniques.


A total of 136 test-day milk yield records of 46 ewes [white (13), brown (10) and black (23)] of Chiapas sheep breed were measured in two consecutive years (2006 and 2007). Because of the number of observations, parities 2, 3, 4 and 5 were considered where records of first and second parity animals were grouped as a class. Records were obtained in random days between ewes covering the total milking period, hand milking was performed in rustic facilities once a day and at every milking, milk yield was recorded.
Two samples of approximately 125 ml per ewe in each sampling were homogenized by agitation and fat, protein, lactose and non-fatty solid values were recorded, using specialized equipment. Management and feed was described previously (López-Ordaz et al., 2012); briefly the study was conducted at the University Centre for Technology Transfer - Autonomous University of Chiapas, located in the Highlands (1780 masl) of Chiapas, Mexico. Ewes were free range grazing during the day on native grasses (Pennisetum clandestinum) and grain supplemented during the afternoon with free access to water. A mixed model with repeated measures (SAS, 2011) was used to estimate environmental effects on daily milk yield (ml), protein (%), fat (%), lactose (%), non-fatty solids (%) and protein: fat relationship, according to the following statistical model:
Yijkno = µ + Ai + Pj + Bk + PxBjk + animaln(jk) + eijkno
Where: Yijkno was the value of each analyzed variable; µ was the general mean; Ai was the fixed effect of i-ith year of study (i = 2006, 2007); Pj was the fixed effect of j-ith class of parity number (j = 2, …, 5); Bk was the fixed effect of k-ith animal colour (k = white, brown, black); PxBjk was the fixed effect of the interaction of parity number and variety; animaln(jk) was the random effect of n-ith animal nested in j-ith parity number and k-ith variety of the ewe ~ niid (0, ); and eijkno was the random error ~ niid (0, ). The animals were used as subject in the repeated-measures command of the model. Following a similar procedure as suggested by Littell et al. (2000), the statistical model was fitted specifying each of the following structures of covariance: compound symmetry (CS), compound symmetry with heterogeneous variances (CSH), autoregressive type 1 [AR(1)], autoregressive type 1 with heterogeneous variances [ARH(1)], Toeplitz (TOEP), Toeplitz with heterogeneous variances (TOEPH) and unstructured (UN).
The goodness-of-fit of the models with different structures of specific covariance was compared using the Akaike information criterion (AIC), minus two times the restricted log likelihood function (-2 LogResL), and the Bayesian information criterion (BIC) using the MIXED procedure (SAS, 2011). The covariance structure that yielded the lowest value was considered as the one generating best fit in analyzed data.
A univariate model was used for analyzing the characteristics of: milk yield, protein, fat, lactose and non-fatty solids, using a model with both fixed effects and linear regression random effects for analyzing the test-day production records in Chiapas sheep breed. The relationship between DIM and DMY (ml), P (%), F (%), L (%), NFS (%) and P:F relationship were determined using n-th degree Legendre polynomial of the best model found with random regression (SAS, 2011).
The animal within the variety was considered as experimental unit in the random command of the analysis of variance and the first three order Legendre polynomials were fitted for each analyzed characteristic. Modeling was suspended when the parameter of the new term in the random regression model did not show statistical significance (P > 0.05). For each variable, the best model was selected comparing the restricted maximum likelihood (Mc Ardle, 2012). -Log likelihood function [log(L)] = 2log(MLk); Akaike information criterion (AIC) AICk = -2log(MLk) + 2pk (Akaike, 1973) and Bayesian information criterion (BIC): BICk = -2log(MLk) + pk log(n) (Littell et al., 2006). The random regression model was represented as:
Where:  was the k-th observation of the variable studied at lactation day when the measurement of the m-th animal was made; were fixed regression coefficients for day in milk function (b0 = intercept, b1 = linear effect, b2 = squared effect and b3 = cubic effect); was the i-th random regression coefficient; (a0m = intercept, a1m= linear effect, a2m= squared effect, y a3m = cubic effect) of the milk production curve of the variable studied per day of lactation belonging to m-th animal (m = 1,…,54); is the k-th observation of the standardized lactation, at the moment of sampling m-th animal, raised to the power 0, 1, 2 and 3; ekm was the error associated with observation ykm. The standardized unit of time (x) was lactation day ranging from -1 to +1 and was calculated using the expression:
Where: x represnt the standarized unit of time from -1 to 1; t was day in milk at the moment of sampling;  was the earliest day in milk (9 in this study) and the latest day of recorded sample (83 in this study). According to Kirkpatrick et al. (1990), the first three Legendre polynomials for the standardized time (x) are: 
Where: x represnt the standarized unit of time from -1 to 1; t was day in milk at the moment of sampling;  was the earliest day in milk (9 in this study) and the latest day of recorded sample (83 in this study). According to Kirkpatrick et al. (1990), the first three Legendre polynomials for the standardized time (x) are:
The fit of the random regression models was carried out following a procedure similar to that suggested by Burnham and Anderson (2004). The restricted maximum likelihood method was specified in the command of the MIXED procedure model. Different order combinations of Legendre polynomials were analyzed to fit the best model and -2 Res Log Likelihood was used as comparison criterion.



Table 2 shows the likelihood criteria for the comparison of the models, where the best fit consisted of a second order polynomial in the fixed part of random regression and a random intercept for protein (%), fat (%), lactose (%), non-fatty solids (%) and protein: fat relationship and linear for daily milk yield (ml).
Least square means for the environmental effects over the milking period are presented in Table 3. These values are within the ranges presented by Ochoa-Cordero et al. (2002), who in their review presented the average of 20 breeds between 1979 and 2002, with milk composition ranging from 3.4 to 6.5 for protein (%), 5.1 to 12.6 for fat (%), 4.4 to 5.5 for lactose (%) and 14.5 a 23.4 % for non-fatty solids. Parity effect showed lower milk production on average per day (29%) on ewes from second parity with regard to older ones (P ≤ 0.09), there was no significant effect (P ≥ 0.22) of parity number on milk composition traits.
The breed variety effect showed significant effect (P ≤ 0.02) on fat (%), where the white ewes was 30% more than the black variety; while the brown variety showed to be intermediate, although the difference with the white variety was smaller than with the black variety. Protein was not statistical different (P ≥ 0.65) among the three varieties, protein: fat relationship was greater in the black variety with regard to the other two (P ≤ 0.10). The interaction of parity number between varieties for lactose percentage (P ≤ 0.01) was found, where the brown variety showed 3% more in the second lambing with respect to the other two varieties in the same lambing, being similar in subsequent parities.
Previous studies on this breed showed that the brown variety has greater milk yield per lactation/day (Perezgrovas and Castro, 2000; Peralta et al., 2005). In the present study, there were no statistical differences in yield between varierities, although the same behavior and lack of significance can be attributed to the reduced number of observations.
Traditionally, the production of hard or semi-hard cheese is based on fat, protein and lactose content. The values found in Chiapas breed show that protein, fat and lactose are within the range presented by Raynal-Ljutovac et al. (2008),  while the non-fat solid content is 3% below the inferior limit. Milk protein and lactose content for breed varieties are similar to the previous report (Perezgrovas and Castro, 2000), while fat and non-fat solid content were lower in all varieties; these differences can be because of the results presented by Perezgrovas and Castro (2000), the evaluation was carried out in only one sample, while in this study, the average was obtained across different phases of lactation.
Table 4 shows the phenotypic correlations between traits. In general, the direction of the correlations, although with different magnitude in Chiapas breed was similar to those shown in literature (Simos et al., 1996, in Mountain Epirus ewes; Sanna et al., 1997, in Sarda ewes; Ochoa-Cordero et al., 2002, in Rambouillet ewes). The difference in magnitude can be explained by differences between breeds and environmental factors, such as production system and nutrition.  The estimates of fixed and random regression for daily milk yield, protein (%), fat (%), lactose (%), non-fat solids (%) and protein: fat relationship are shown in Table 5. The scatter plot and the fixed model are shown in Figure 1 and random regression of variety is shown in Figure 2. The best model for daily milk yield was of first order Legendre polynomials, while for protein (%), fat (%), lactose (%), non-fat solids (%) and protein: fat relationship, the best fit model consisted in a second order polynomial.
In a previous study, Vázquez et al. (2014) observed third order Legendre polynomials for lactation curve with daily milk measurements in 120 days period of time in this same breed, with values of: 115.67 (2.46), -49.34 (1.58), 4.61 (1.6) and -6.57 (1.43) for estimates of α0, α1, α2 y α3, respectively, for which the difference can be explained due to the sampling number between both studies, being the estimates of α1 with the same trend in both studies.
The results of the present study, show that Chiapas sheep breed presents higher percentage of fat and protein than those reported by Jandal (1996) in goat and cow milk, but lower than in sheep milk, while lactose percentage being higher in goat and sheep milk and lower in cow milk. Finally, non-fat solids percentage is higher in goat, sheep and cow milk than in Chiapas breed ewes for the current work.
By their origin, ewe varieties (white, brown, black) from the Chiapas sheep breed are descendants of the Spanish breeds Churra, Lacha and Manchega, respectively; however, it has been considered many years ago that Spanish breeds were selected for milk production. Churra breed is superior to the white variety of Chiapas breed in fat %, lactose (%) and non-fatty solids (%) and lower in protein (%) (Ochoa-Cordero et al., 2002). Lacha breed is 1.6 percentage points higher in fat (%) and similar in protein (%), with respect to the brown variety of Chiapas breed. Whereas, Manchega breed is 3.3 and 6.6% points higher than the black variety in fat (%) and non-fat solids (%) and similar in protein (%) and latose (%), in accordance with Ochoa-Cordero et al. (2002).
The quadratic behavior trend for protein, lactose and non-fat solids in the Chiapas sheep breed are similar to those presented by Simos et al. (1996), in Mountain Epirus ewes and Ochoa-Cordero et al. (2002), in Rambouillet ewes. On the other hand, fat (%) showed similar trend to Rambouillet ewes, but different to Epirus ewes, because this breed decreased to the 4th month and increased in the 5th and 6th month and decreased again in the 7th month.


Component percentages of protein, fat, lactose, non-fat solids and protein: fat relationship remained constant during the first five lactations. Chiapas sheep breed varieties showed similarity between milk composition characteristics, except in fat content (%), where the white variety had the highest proportion and the black variety the lowest, with a difference of 30%, whereas the brown variety was intermediate between these two. The results of the present study show the feasibility of selecting the Chiapas sheep breed for milk production and for dual-purpose animal (wool-milk) under natural conditions in the Altos de Chiapas, Mexico.


The authors have not declared any conflict of interest.


Partial information of this study was used as a dissertation project by the first author for her Veterinary Medicine and Husbandry licensing process in the FMVZ-UNAM. Special thanks for financial support of projects UNAM-PAPIIT IN207707-3, UNAM-PAPIIT IN205710-3 and SAGARPA-CONACYT 2004-CO1-111/A1 and to the collaboration between the Facultad de Medicina Veterinaria y Zootecnia of the Universidad Nacional Autónoma de México and the Centro Universitario de Investigación y Transferencia de Tecnología of the Universidad Autónoma de Chiapas.


Akaike H (1973). Information theory as an extensión of the máximum likelihood principle. In Petrov BN, Caski F (Eds.) Second international symposium on information theory. Akademiai Kiado, Budapest, pp 267-281.
Burnham KP, Anderson DR (2004). Multimodel inference understanding AIC and BIC in model selection. Soc Methods Res. 33: 261-304.
Dario C, Carnicella D, Dario M, Bufano G (2008). Genetic polymorphism of β-lactoglobulin gene and effect on milk composition in Leccese sheep. Small Rumin. Res. 74: 270-273.
FAO (2009). 
Jandal JM (1996). Comparative aspects of goat and sheep milk. Small Rumin. Res. 22:177-185.
Jaramillo DP, Zamora A, Guamis B, Rodríguez M, Trujillo AJ (2008). Cheesemaking aptitude of two Spanish dairy ewe breed: Changes during lactation and relationship between physicochemical and technological properties. Small Rumin. Res. 78: 48-55.
Kirkpatrick M, Lofsvold D, Bulmer M (1990). Analysis of the Inheritance, Selection and Evolution for Growth Trajectories. Genetics 124: 979-993.
Leitner G, Chaffer M, Caraso Y, Ezra E, Kababea D, Winkler M, Glickman A, Saran A (2003). Udder infection and milk somatic cell count, NAGase activity and milk composition – fat, protein and lactose – in israeli-Assaf an Awassi sheep. Small Rumin. Res. 49: 157-164.
Littell RC, Milliken GA, Stroup WW, Wolfinger RD, Schabenberber O (2006). SAS© for mix models. Second Edition. Cary, NC: SAS Institute Inc, pp. 159-202
Littell RC, Pendergast J, Natarajan R (2000). Modeling covariance structure in the analisys of repeated measures data. Stat. Med. 19(13): 1793-1819.
López-Ordaz R, Olivera-Vega I, Berruecos-Villalobos JM, Peralta-Lyson M, Ulloa-Arvizu R, Vásquez-Peláez C (2012). Genetic parameters for birth and weaning weights in the local Chiapas sheep breed from Mexico. Rev. Mex. Cienc. Pecu. 3: 113-123.
Marrtini M, Scolozzi C, Cecchi F, Mele M, Salari F (2008). Relationship between morphometric characteristics of milk fat globules and the cheese making aptitude of sheep's milk. Small Rumin. Res. 74. 194-201.
Mc Ardle JJ (2012). Latent curve modeling of longitudinal growth data. in: Hoyle, R.H. (Ed.), Handbook of structural equation modeling. The Guilford Press, New York, pp. 547-571.
Nudda R, Bencici R, Mijatovic S, Pulina G (2002). The yield and composition of milk in Sarda, Awassi and Merino sheep milked unilaterally at different frequencies. J. Dairy Sci. 85: 2879-2884.
Ochoa-Cordero MA, Torres-Hernández AE, Ochoa-Alfaro AE, Vega-Roque L, Mandeville PB (2002). Milk yield and composition of Rambouillet ewes under intensive management. Small Rumin. Res. 43. 269-274.
Pedraza P, Peralta M, Perezgrovas-Garza R (1992). El borrego Chipas: una raza local mexicana de origen espa-ol. Arch. Zootec. 41: 355-362.
Pedraza VP, Peralta LM (2003). Elaboración de queso de oveja bajo un sistema artisanal. In: Memorias XII Congreso Nacional de Producción Ovina. Asociación Mexicana de Técnicos Especialistas de Ovinocultura. Tulancingo Hidalgo, México, pp 115-120.
Peralta-Lailson M, Trejo-González AA, Pedraza-Villagómez P, Berruecos-Villalobos J M, Vásquez CG (2005). Factors affecting milk yield and lactation curve fitting in the creole-sheep of Chiapas-México. Small Rumin. Res. 58: 265-273.
Perezgrovas G, Castro GH (2000). El borrego Chiapas y el sistema tradicional de manejo de ovinos entre las pastoras tzotziles. Arch. de Zootec. México. 49: 391-403.
Ploumi K, Belibasaki S, Triantaphyllidis G (1998). Some factors affecting daily milk yield and composition in a flock of Chios ewes. Small Rumin. Res. 28: 89-92.
Pugliese C, Acciaioli A, Rapaccini S, Parisi G, Franci O (2000). Evolution of chemical composition, somatic cell count and renneting properties of the milk of Massese ewes. Small Rumin. Res. 35. 71-80.
Raynal-Ljutovac K., Lagriffoul G, Paccard P, Guillet I, Chillard Y (2008). Composition of goat and sheep milk products: An update. Small Rumin. Res. 79 57-72.
Sanna SR, Carta A, Casu S (1997). (Co) variance component estimates for milk composition traits in sarda dairy sheep using a bivariate animal model. Small Rumin. Res. 25. 77-82. d
SAS (2011). Institute Inc., Cary, NC, USA
Scintu MF, Piredda G (2007). Typicity and biodiversity of goat and sheep milk products. Small Rumin. Res. 68 221-231.
Simos EN, Nikolau EM, Zoiopoulos PE (1996). Yield, composition and certain physicochemical characteristiscs of milk of the Epirus Mountain sheep breed. Small Rumin. Res. 20: 67-74.
Vázquez-Peláez CG, García-Mu-iz JG, Lopez- Villalobos N (2014). Empirical models used for lactation curve analysis in the Chiapas sheep breed using randon regresión models. Int. J. Livest. Prod. 5(3): 55-64.