Modelling the faecal worm egg count curve during the periparturient period in Uruguayan Merino sheep

Disease caused by gastrointestinal nematodes (GIN) is one of the main constraints for sheep production worldwide. Although adult categories are more resistant to GIN, a temporary loss of acquired immunity is presented around parturition, which results in the so-called “periparturient rise” (PPR) in fecal excretion of GIN eggs. The objective of the present work was to study the dynamic of the PPR and to model the faecal worm egg count (FEC) curve during the periparturient period in Uruguayan Merino sheep. Faecal samples were collected from lambing ewes during 2009 and 2010 parturitions. FEC and infective larvae cultures of each sampling were recorded. A total of 2,121 records from 748 ewes, the progeny of 107 sires, were used in the statistical analysis. FEC data were normalized using two logarithmic transformations: Log FEC = Loge (FEC + 100) and Log FEC_st (Log FEC with the variance standardized within contemporary group). Three functions were compared to shape the PPR curve. Haemonchus spp. was the most prevalent parasite. Log FEC_st was the selected response variable of the model, for its better adjustment to a normal distribution and a more homogeneous residual variance. The fixed regression model with Legendre polynomials was the selected one, based on the selection model criteria (Akaike & Schwarz Bayesian Information Criteria). The highest egg output was observed between two and four weeks post-lambing. In conclusion, the PPR observed in Uruguayan Merino lambing ewes had the maximum egg output matched with the milk production peak. Additional key words: gastrointestinal nematodes; Haemonchus contortus; lambing ewes; periparturient rise.


987
The periparturient rise in Merino ewes guay, characterized by a warm and wet climate, with a mean annual temperature of 18-19°C, relative humidity of 70-72% and an average annual rainfall of 1,400-1,500 mm (Castaño et al., 2011). These conditions are favorable for the development and survival of larvae of GIN throughout the year, varying the presence and predominance of different parasite genus according to the season.
In FMN, ewes are managed in parturition groups according to the average expected day of lambing. A pre-partum strategic drenching is performed every year as a management control measure, approximately one month prior to the expected beginning of lambing. Anthelmintic used were DOVENIX ® in 2009 and TRIMIX ® in 2010. In "Talitas" and "La Gringa", ewes are managed in a single parturition group. In these stud flocks, ewes are not always dewormed before parturition because drenching is performed strategically; i.e. periodically stool samples are collected from a random sample of animals, and if FEC average is higher than 500 eggs per gram, all the parturition group is drenched immediately. For the present study, it was not necessary anthelmintic treatment in these flocks.

Ewe records
A total of 2,500 faecal samples were collected in 2009 and 2010, during lambing season. The age of ewes ranged between of 2 and 10 years-old and litter size was recorded as single or multiple (≥ two lambs). The experiment was conducted between days -50 and +68 respect to lambing (day 0). Each ewe was sampled on average three times, under natural mixed-species parasite challenge on pasture. All ewes in the same cohort were sampled on the same day. The first sample was collected in late pregnancy and the others in early and mid-lactation. The number of post-lambing measurements depended on the degree of parasitic infection; if FEC counts within a cohort reached certain levels such that animal health and welfare might be compromised, then all animals in that cohort were immediately drenched. Cohorts were defined as a parturition group in case of FMN or all the flock in case of "Talitas" and "La Gringa" studs. In FMN, ewes were sampled in the two years of the experiment (Table 1). In 2009, a total of six measurements were performed on 293 ewes divided in four parturition groups, obtaining a total of 742 records. In 2010, five measurements were performed on 345 ewes (of which 185 were also sampled

Introduction
Gastrointestinal nematodes (GIN) are one of the main constraints for sheep production in Uruguay and worldwide (Castells et al., 1995;Perry & Randolph, 1999). Although lambs are the most susceptible category to GIN infection, lambing ewes also experience a relaxation in acquired immunity around parturition through a phenomenon called periparturient rise (PPR). The PPR was firstly documented by Taylor (1935) and it can be defined as a temporary but marked increase in nematode eggs output by lambing ewes; that begins in the last weeks of gestation and reaches the maximum peak in the first weeks post-parturition. It is an important event because it represents a pasture larval contamination source for newborn lambs (Bishop & Stear, 2001;Romero & Boero, 2001). The cause has not yet been determined, but it is generally accepted that the rise occurs after an immunity depression of the host by stressful factors such as pregnancy, parturition, lactation, climate and malnutrition (Barger, 1993).
The PPR was firstly described in Uruguay by Nari et al. (1977a) in the Ideal breed, where Haemonchus spp. represented 82% of the total of the parasite genera presented. They worked with only 71 animals which were divided in three groups: lambing ewes that were dewormed before parturition, lambing ewes without a pre-partum anthelmintic treatment and non-treated barren ewes (control group). These authors observed that the maximum egg output took place between six and eight weeks post-lambing, but they did not model the shape of the PPR curve.
Because it is important to know the dynamic of the PPR for the flock management and parasite control, the aim of the present study was to model the phenotypic curve of faecal worm egg count (FEC) during the periparturient period in Uruguayan Merino sheep.

Animals and management
Animals belong to three flocks genetically connected by reference rams: the Fine Merino Nucleus (FMN) belonging to "Glencoe", a research station of the Instituto Nacional de Investigación Agropecuaria (INIA) of Uruguay, and two Merino studs: "Talitas" and "La Gringa", belonging to the same breeder. The three flocks are located in the northern part of Uru-in 2009) divided in three parturition groups, obtaining a total of 881 samples. In "La Gringa" and "Talitas" studs, only 349 and 173 females, respectively, were sampled in 2010, with two measurements performed in each flock, obtaining a total of 877 samples (596 and 281, respectively) ( Table 1).

Parasitological analysis
FEC were determined using the modified McMaster technique (Whitlock, 1948), where each egg observed represented 100 eggs per gram of faeces. In addition, in 2010, faecal cultures of infective larvae were prepared to assess the species composition of nematode infection in each flock.

Statistical analysis: Exploratory analysis of ewe FEC in the periparturient period
A total of 2121 records from 748 ewes, the progeny of 107 sires, were used in the statistical analysis. FEC data were transformed prior to analysis using the following logarithmic transformations, in order to remove skewness and to normalize data: Log FEC = Log e (FEC + 100) Log FEC_st = (Log e (FEC + 100) -m GC ) / ds GC where Log FEC_st is the natural logarithm of FEC records (Log FEC) with the variance standardized within contemporary group (CG); and m CG and ds CG are the mean and standard deviation of Log FEC of each CG, respectively.
Data were analyzed by a sire repeatability model through MIXED procedure of SAS statistical package (SAS Inst, 2004). Three functions were compared to model ewe FEC in the periparturient period ("PPR curve"): 1) fixed classes of FEC measurement period (PER); 2) days of FEC measurement respect to lambing (DML) as lineal and quadratic covariates; and 3) quadratic Legendre polynomials in function to DML. The general model common to the three functions was the following: y ijklm = CG i + LS j + EA k + f + s l + p m + e ijklm where y is FEC transformed (Log FEC or Log FEC_st). The fixed effects are: CG (defined as sampling date by flock by parturition group, with 26 levels), LS litter size (two levels: single and multiple); EA ewe age (four levels: 2, 3, 4 and ≥ 5 years). The tested function is f, that according to the model is: PER (4 levels: ≤-15, >-15, >15 and >30 days of FEC measurement respect to lambing), DML and DML 2 as covariates, or secondorder Legendre polynomials. The random effects are: s the additive genetic effect of the sire (107 levels), p the permanent environmental effect of the animal (748 levels) and e is the residual term.
The criteria used for choosing the model with the better adjustment were the Akaike Information Criteria (AIC) and the Schwarz Bayesian Information Criteria (BIC). Once the model was selected, studentized residuals of Log FEC and Log FEC_st were calculated, in order to detect possible extreme values. Records with studentized residuals ≥ to 3.0 or ≤ to -3.0 were eliminated as outliers (Ott & Longnecker, 2001).
Descriptive statistic for FEC, Log FEC, Log FEC_st and the studentized residuals was estimated, through UNIVARIATE procedure of SAS statistical package (2004). Based on residuals analysis and descriptive statistical values of Log FEC and Log FEC_st, it was selected the response variable (y) of the model.

Parasitological analysis
Results of larvae cultures are presented in Fig. 1. Haemonchus contortus was the most prevalent parasite species in all samples and flocks. It was followed in order of predominance, by Trichostrongylus columbriformis and Teladorsagia circumcinta spp. Other genera (i.e. Oesophagostomum spp., Cooperia spp.) were found in a very small proportion of the samples.

Statistical analysis: Exploratory analysis of ewe FEC in the periparturient period
The descriptive statistic of ewe FEC without and after logarithmic transformation is shown in Table 2. For FEC without transformation, the mean, median and mode values clearly differ among them, and data were strongly positively skewed. Due to the large difference among the measures of central tendency and the elevated values of the skewness and kurtosis, data were adjusted by logarithmic transformation, with and without variance standardization (Log FEC_st and Log FEC). Transformed variables presented closer values of measures of central tendency, and skewness and kurtosis values were close to zero; although Log FEC_st transformation had a better adjustment to a normal distribution than Log FEC.
The studentized residuals of Log FEC and Log FEC_st were also analyzed. The pattern of residuals of Log FEC, as a function of the predicted mean, showed heteroskedascity of the residual variance (Fig. 2a). The distribution of studentized residuals of Log FEC_st had a better adjustment to a normal distribution and they were more uniformly distributed when were plotted as a function of the  predicted mean (Fig. 2b). Thus, this transformation removed the heterogeneous residual variance pattern. Due to a better adjustment to a normal distribution and a more homogeneous distribution of the residuals, Log FEC_st was selected as the response variable of the model.

Periparturient rise
The three functions used to shape the PPR resulted in similar curves (Fig. 3); although the model with second-order Legendre polynomials was the selected one, based on the model selection criteria (lower AIC and BIC values).
The maximum nematode egg output was observed after lambing date, reaching the maximum peak in the third period of measurement (PER 3), that is, between 2 and 4 weeks post-parturition. Unfortunately, it was not possible to study the shape of PPR function beyond 68 days post-lambing, because of the high level of parasite infection in some animals began to compromise welfare and health, thus sampling end sooner than expected.

Parasitological analysis
Results of larvae cultures are consistent with Nari et al. (1977b) and Castells (2009) reports, who also found that Haemonchus spp. and Trichostrongylus spp. were the two predominant parasite genera in Uruguay, both in lambs and lambing ewes.
The cause of the increase in nematode egg output during the periparturient period has not yet been determined. The PPR may derive from an increase in adult parasite populations by an increased rate of establish-

991
The periparturient rise in Merino ewes ment of newly ingested larvae which develop to the adult stage without inhibition, as well as by the resumption of development of previously inhibited fourthstage larvae. In addition, both newly acquired and established adult female parasites may show increased fecundity (O'Sullivan & Donald, 1970;Urquhart et al., 1996). Thus, the high percentage of H. contortus could be explained by newly ingested larvae of this parasite and/or the maturation of hypobiotic larvae that arrested the development over winter.

Statistical analysis: Exploratory analysis of ewe FEC in the periparturient period
The results of the statistical analysis of ewe FEC were in agreement with those reported by several authors (e.g. Gasbarre & Miller, 2000;Stear et al., 2007), who described that the distribution of FEC is skewed and overdispersed; where most hosts have relatively low egg counts while a small proportion of hosts eliminate a large number of eggs.
Different authors found different transformations (e.g. square root, cube root, Log (FEC+25)) as the most appropriate to their data structure, being appropriate to explore and identify, which is the best transformation in each particular case (Eady, 1995;Castells, 2009). Log e (FEC+100) transformation was also used by Watson et al. (1995) and Morris et al. (1998), as well as in the sheep genetic evaluation for resistance to GIN in Uruguay (Ciappesoni et al., 2010). An alternative that has also been described is data transformation with variance standardization. Brown & Tier (2003) and Pollott & Greeff (2004) transformed FEC count through the cube root with the variance standardized within contemporary groups. Brown & Tier (2003) found a high phenotypic and genetic correlation between the variable with and without standardization of the variance (0.95 ± 0.00 and 0.95 ± 0.01, respectively).

Periparturient rise
The second reason to select the model with secondorder Legendre polynomials was that orthogonal polynomials of standardized units of time, as Legendre polynomials, have been recommended as covariates in regression models because they are easy to calculate and to use, and decrease the correlation between the estimated regression coefficients (Mrode, 2005).
The observation of the present study was consistent with the reported by Herd et al. (1983), who also found that the PPR occurs between two and four weeks after parturition, matching the milk production peak that in Merino breed occurs approximately at three weeks post-lambing (Corbett, 1968). Salisbury & Arundel (1970) observed the maximum egg output between four and six weeks post-lambing, while Crofton (1954) and Nari et al. (1977a), described this event between 6 and 8 weeks post-parturition.
In conclusion, it was observed the PPR in lambing ewes belonging to three Uruguayan Merino flocks. The maximum nematode egg output was observed between two and four weeks post-lambing, matching with the maximum milk yield peak. The pathogenicity of GIN varies with species and with the number of parasites present. Thus, it is important to take into account these results for sheep flock management around parturition, when newborn lambs will be on contaminated pastures being very susceptible to parasite infection.