Identification of Sindhi cows that are susceptible or resistant to Haematobia irritans

To identify susceptible and resistant Haematobia irritans cows, horn flies were counted biweekly for 3 years in a herd of 25 Sindhi cows. Repeated measures linear mixed models were created including cow as a random factor. The results were analyzed by: 1) observing horn fly counts, considering fly-susceptible cows with infestations appearing in the upper quartile more than 50% of the weeks and in the lower quartile less than 20% of the weeks, and fly-resistant cows those that the number of flies was in the lower quartile more than 50% of the weeks and in the upper quartile less than 20%; 2) by the best linear unbiased predictions (BLUPs), to evaluate the cow effect on fly counts. Fly-susceptible cows were those in which the infestation appeared in the 90th percentile of the BLUPs, whereas fly-resistant cows appeared in the 10th percentile. For the observational method the individuals identified as resistant varied between 8% and 20% and 8% to 12% were susceptible. For the BLUP method, the rates of susceptible and resistant cows were 12%. The agreement among methods suggests that susceptible cows can be identified by observations of fly counts, allowing for selective breeding, culling or treatment.


Introduction
Although Brazilian livestock production continues to increase, productivity remains low. Cattle parasitosis, including infestations by the horn fly (Haematobia irritans), is among the main factors limiting the profitability of livestock in the country (GRISI et al., 2014).
Economic losses due to horn flies have been estimated at US$ 730 million per year in the USA (FOIL & HOGSETTE, 1994) and US$ 2.56 billion per year in Brazil (GRISI et al., 2014). Additionally, control of this parasite is problematic due to the development of insecticide resistance (SPARKS et al., 1985).
Palavras-chave: Mosca dos chifres, bovinos, seleção animal, controle integrado. Horn fly resistance to pyrethroids, which are widely applied in Brazil (BARROS et al., 2013;BRITO et al., 2019), increases production costs and causes environmental contamination. Horn fly infestation of a herd is directly influenced by the proportion of susceptible animals in the herd (STEELMAN et al., 1993;JENSEN et al., 2004). Thus, the identification of individuals genetically susceptible to H. irritans would allow the selective treatment or removal of these individuals from the herd to reduce the overall level of herd infestation. The objective of this trial was to identify horn fly-susceptible and horn fly-resistant animals in a Sindhi herd by two different methods. The information acquired can be used in control programs to reduce insecticide use.

Materials and Methods
This trial was conducted for 42 months between April 15, 2011, andSeptember 20, 2014, at the Federal University of Campina Grande, Patos, Brazil. No treatment against horn flies was applied during the trial.

Data collection
The number of horn flies on 25 adult cows (aged four to 10 years) from a Sindhi herd was counted every 14 days. As it was an open herd, the trial period was divided into three stages based on cow composition, with the same cows maintained within each period The counts were performed by one trained observer between 7:00 am and 8:30 am. During the counts, the cows were contained in a squeeze chute. Total infestation was estimated by counting the flies on one side of each cow and multiplying by two according to the technique described by Morrison & Foil (1995) and Medeiros et al. (2018).

Statistical analysis
Three databases were created in Excel, and the data were reviewed for data entry errors. One database was created for each study period (2011-2012; 2012-2013; 2013-2014). Excel spreadsheets were imported into R version 3.5.2 (R CORE TEAM, 2013) for descriptive and statistical analysis, performed using the lme4 package (BATES et al., 2015).

Descriptive analysis
The variables evaluated were the number of horn flies on the cows, the sampling date and a binary variable for rainy or dry season. Descriptive statistics were calculated, including the median, the interquartile range, and the minimum and maximum number of horn flies, for each observation day. For the present analysis, fly-susceptible cows were identified as those for which the infestation of flies appeared in the upper quartile for more than 50% of the weeks and in the lower quartile for less than 20% of the weeks. In contrast, fly-resistant cows were defined as those for which the fly counts appeared in the lower quartile for more than 50% of the weeks and in the upper quartile for less than 20% of the weeks.

Inferential statistics
To identify resistant and susceptible cows for the best linear unbiased predictions (BLUPs) analysis, three repeated measures linear mixed models (one for each period) were constructed with cow as a random effect intercept. The response variable was the log ten transformed counts of horn flies per cow, and the explanatory variable were the observation date and season. As the trail took place in a semiarid region with two seasons well stablished (rainy and dry) the season was evaluated monthly as a binary outcome, considering a rainy season if it rained more or equal than 50mm or dry season if the rain was less than 50mm. From the previously modeled values, the BLUPs for the estimation of random effects were determined. The Standardized residuals and the BLUPs of the random effects were obtained and assessed for normality, heteroscedasticity and outlying observations as described by Dohoo et al. (2009). Each cow's BLUPs were plotted against the average quantile rank values that were determined as the difference between the number of weeks in the high-risk quartile group and the number of weeks in the low risk quartile group, averaged by the total number of weeks in each of the observation periods. A linear model fit for the values of BLUPS against the average rank values and the correlation between the two methods was tested using Spearman's correlation coefficient.
The animal effect values (BLUPs) were evaluated by percentiles, with 0 representing the lowest counts (or more resistant cows) and 10 representing the highest counts (or more susceptible cows). These BLUPs represented only the effect of cow and not the effect of day, season or other unmeasured counfounders.

Data availability
The data base is available at Miraballes et al. (2019).

Results
Based on the observational analysis, cows were identified as fly-susceptible or fly-resistant, and the results were compared with those of the BLUPs analysis ( Figure 1). Results for the observational and BLUPs methods are available in Figure 2 and Figure 3 respectively. The results of the two confronted methods were largely consistent, suggesting that susceptible cows can be reliably identified by observations of fly counts. In Figure 1 it can be observed the strong linear correlation between average quartile ranks, as the averaged difference between the number of biweekly counts in the high and low quartiles, and BLUPs for each cow and for each study period. The observational method results are represented in Figure 1 as point shapes, while susceptible and resistant cows identified with the BLUPs method as encircled values. For the first year period, the observational method identified cow's id number 4, 5, 18, 26 and 27 as resistant (20% of the cows) and cows 2, 9 and 20 as susceptible (12%). The BLUPs method identified the same susceptible cows but fewer resistant individuals [cows id number 4, 5 and 26 (12%)]. The agreement between methods was also partial for the period of 2012-2013: two individuals were identified as resistant in the observational method (cows id number 29 and 19) that were not identified with the BLUP method, while the BLUP method identified cow id number 30 as susceptible. For the final period of 2013-2014, observational and BLUP methods differed only in classifying cow id number 45. Overall, the rates of individuals identified varied between 8% and 20% for resistant individuals and 8% to 12%   for susceptible individuals in the observational method, while for the BLUP method, the rate of 12% was shared among years and categories (Table 1). On average, the observational method identified 16% resistant cows and 10.7% susceptible cows, while the BLUP method consistently identified 12% of the cows as susceptible and 12% as resistant. It is observed a largely consistent identification between the two methods. Furthermore, Figure 1 shows the strong linear correlation between average quartile ranks and BLUPs for each cow and for each study period. The strong correlation translates into the agreement between the two methods, with high values for the Spearman's rank correlation coefficients of 0.99 (Figure 1a), 0.97 (Figure 1b) and 0.95 (Figure 1c), and p<0.05).
In figure 4 are represented the number of horn flies for the three observed time periods, coded in accordance to the concurring results from the two confronted methods: for the first time period of 2011-2012 (Figure 4a), both methods identified 3 cows as susceptible and 3 cows as resistant; for 2012-2013 (Figure 4b), 3 cows were identified as resistant and 2 cows as susceptible; and for 2013-2014 (Figure 4c), 2 cows were identified as resistant, while 3 cows were identified as susceptible. Cow id number 4 was simultaneously identified by both methods as resistant for two time periods and as susceptible for the period of 2012-2013. Cow id number 19 was also classified among different groups resistant and susceptible for the two time periods using the observational method. The equivalent of Figure 4 with logged values is provided in Figure 5. In Figure 4, it can be verified how the number of horn flies exceeded the threshold for treatment (200 flies per cow) on numerous occasions, during peaks of infestation. After a peak, it is frequently observed that fly counts rapidly decreased to below this threshold, even though no treatment was administered.
The repeatability estimates were determined for each period, with 28.6% for 2011-2012; 21.5% for 2012-2013 and 24.7% for 2013-2014. The correlation among observations within the individual is consistent between time periods, reflecting the need for a repeated measures linear mixed model, and that the number of horn flies per cow is a trait, varying among cows.

Discussion
The mechanisms underlying individual differences in horn fly resistance or susceptibility among cows of the same breed are not well understood. Climatic and host-related factors such as breed, color, age, and behavioral defense against horn flies, among others, influence the population dynamics of horn flies (FRANKS et al., 1964;BEAN et al., 1987). In this trial, BLUPs analysis allowed the influences of these factors to be removed and the resistance or susceptibility of each animal to be evaluated, which could also be identified by observing the degree of infestation of the cows. However, one cow was considered fly resistant in two stages of the trial and fly susceptible in the third, which suggests the existence of individual variation due to unmeasured factors, such as physiological and/or immunological condition.
This methodology allows to evaluate the susceptibility of the individuals throughout the time studied. Otherwise, other reports have identified the susceptible and resistant individuals in herds at one specific day (BIANCHIN & ALVES, 1997), by counting the flies on 50% of the herd randomly selected (BARROS, 2001) or by reporting the average of the susceptible or resistant bovines in the period of study (MIRABALLES et al., 2018).
The identification of higher-susceptibility animals based on fly number allows the selective treatment of such animals, which can decrease the number of flies in the herd and the environment. Jensen et al. (2004) demonstrated that the level of infestation of a herd with H. irritans could be altered by adding or eliminating fly-resistant or fly-susceptible animals. Thus, it is suggested that selective treatment of the most infested animals may have the same effect as changing the composition of the herd. In a previous trial, the selective treatment of bulls decreased the rate  of infestation of the untreated cows during the mating period (MIRABALLES et al., 2018). Bulls have higher infestation rates than cows; therefore, selectively treating bulls in a breeding herd would be of value. However, it is important to evaluate the implications of selective treatment for the development of insecticide resistance (BARROS et al., 2013). Since information regarding the development of resistance in selective treatment programs involving long-lasting insecticides is lacking, it is important to perform bioassays before and after treatments to evaluate the susceptibility of horn fly populations to the drugs in use.
The identification of the higher-susceptibility animals may also be valuable for genetic selection. As determined by Brown et al. (1992) and Steelman et al. (1996), the heritability of fly resistance is moderate, so selection of animals based on fly resistance, among other criteria, can be expected to reduce rates of horn fly infestation over the long term. Barros (2001) identified frequencies of 13.3% and 15% of Nelore cows as susceptible, and Miraballes et al. (2018) found that 8.3% of cows were susceptible in a Braford herd. In both previous studies, the proportion of fly-resistant animals was greater than that of fly-susceptible ones. These findings align with those of this study. Nevertheless, it is recognized that the effect of variation in herd members could have contributed to the varying identification of cows as susceptible and resistant between study periods, as was the case of cow id number 4, similarly classified by both methods at one instance as susceptible and at two instances as resistant. Sindhi cows are a Bos indicus breed; thus, it is important to perform a similar study on Bos taurus breeds since horn fly infestations tend to be higher in European breeds (TUGWELL et al., 1969), and the proportion of susceptible animals may therefore differ between them.

Conclusion
Fly-susceptible and fly-resistant cows can be identified by observation of the number of flies on the animals, allowing the selective treatment or removal of the more susceptible individuals.