1. Introduction
Buffalo breeding is an important economic asset in Italy, especially in the southern regions. Italy holds about 0.01% of the world buffalo population and the number of animals has increased significantly in recent years) [
1]. The reasons for the growing interest in buffalo farming are to be found in the popularity of traditional dairy products, which are obtained from the milk of these animals, and especially in the production of “Mozzarella di Bufala Campana”, a typical cheese characterized by a protected designation of origin (PDO) label [
2,
3]. Indeed, milk quality is a crucial issue for the Italian buffalo dairy industry, having a direct impact on the technological characteristics of milk itself [
4]. It is also well known that milk quality is strictly related to mammary gland health [
5,
6,
7]. Indeed, mastitis is one of the most expensive diseases in the dairy industry [
2]. Mastitic milk is characterized by a high number of somatic cells and by changes in its composition, which affects coagulation capacity with consequent production of low-quality cheeses. Diagnosis of clinical mastitis is commonly based on local and systemic reactions or on milk changes. The diagnosis of subclinical mastitis, on the contrary, is more difficult because both milk and udder do not show evidence of abnormality, although, also in this case, there is a high number of somatic cells [
8].
According to the guidelines of the International Dairy Federation, the diagnosis of mastitis is mainly based on somatic cell count (SCC) and bacteriological culture of milk, although alternative indicators from milk can be used, such as lactose [
9], differential somatic cell count [
10,
11], L-Lactate dehydrogenase [
12] and Electrical Conductivity (EC). The latter is determined by the total concentration of cations and anions in milk, whose values change during an inflammatory process of mammary gland. In fact, during mastitis the blood–milk barrier is damaged and consequently the tight junctions between secretory cells becomes leaky. This promotes the movement of extracellular fluid components, including sodium (Na) and chloride (Cl), that mix with milk and increase the Na and Cl concentrations [
13] with a concurrent decrease in milk K concentration [
14]. For this reason, EC is considered a reliable indicator of early mastitis diagnosis in dairy cattle [
12,
15]. The reliability of EC as an early indicator of mastitis has been demonstrated in several species. Milner et al. [
16] observed changes in EC after direct infusion of
Staphylococcus aureus or
Streptococcus uberis into the mammary gland in lactating Friesian-Holstein cows. Similarly, the intramammary infection of bacteria in Murciano-Granadina goats caused an increase of both SCC and EC [
17]. Recently, it has been proposed that EC alone was not sufficient to achieve the desired sensitivity and specificity targets and that improvements can be obtained by using other information (e.g., milk yield, milk flow, number of incomplete milking), that may increase accuracy of detection and ability to determine early onset of mastitis [
18]. Other authors [
19] suggested the combination of EC, milk production rate and average milk flow rate as a tool potentially useful for an early detection of mastitis.
Although buffaloes are traditionally considered less susceptible to mastitis than cattle [
2], some studies have reported high prevalence of subclinical intramammary infections [
20,
21]. Therefore, the pathology is underestimated in this species, causing both health problems to the animals and noticeable economic loss for the farmers. Because of this, the development of techniques for early detection of mastitis in the farm would allow a rapid and early management of the disease, in order to decrease the negative effect on milk quality and therefore on the livestock market economy.
Limited information is currently available on the relationships among EC, production traits and SCC in Italian Mediterranean Buffalo. In particular, a gap of knowledge is present on the value of EC as a predictive indicator of SCC increase and mastitis in this species. Hence, the aims of this study were i) to investigate the correlations among EC, SCC and production traits in buffalo species and ii) to estimate the predictive value of EC to diagnose SCC increase in buffalo species, by using data collected at a commercial Italian buffalo farm.
3. Results
3.1. Descriptive Statistics
The average observed trend for EC and SCS across stage of lactation and parity can be observed in
Figure 1.
The EC ranged from a minimum of mS/cm for primiparous within 30 DIM to a maximum of mS/cm for 3rd-parity buffalo cows at 300 dim. Overall, EC increased across lactation, especially for pluriparous buffalo cows from 90 DIM onwards. On the contrary, in primiparous buffalo cows, EC stayed stable below 8.7 across nearly all lactation approaching a nadir at the end (EC = ). The SCS increased steadily for all parities until 150 DIM, and afterwards the rate of increase changed according to parity.
The phenotypic variability of EC across class of SCS and within parity can be observed in
Figure S1. A largest variability was observed when SCS increased, especially in pluriparous buffalo cows. However, high EC values were observed even when SCS was low (<1).
3.2. Effect of Milk Conductivity on Milk Yield and Composition
Hypothesis testing for MY, FP and PP from Model 1 is summarized in
Table 2.
All effects included in model 1 were significant for all traits, with the only exception of the effect of EC nested within parity for FP. The estimated linear regression coefficients of EC nested within parity for MY and PP are in
Table 3.
EC had a significant (unfavorable) effect on MY for all pluriparous buffalo cows. Each increase in EC unit (i.e., the estimate of the linear regression coefficient) reduces milk by a minimum (absolute value) of kg in 2nd parity to a maximum of kg in 5th + parity. A similar (unfavorable) relationship was also observed between EC and PP. However, the effect of EC was only significant in 3rd and 5th + parity with a reduction in PP per EC unit increase ranging from −0.02429 to −0.02831%, respectively. The relationship between EC and FP was not significant.
3.3. Effect of Milk Conductivity on Somatic Cell Score
The average EC value at official milk recording (), EC collected 3 days before the official milk recording () as well as its mean (), standard deviation () and slope () during the 5-day interval before each test-day were 8.87, 8.85, 8.86, 0.304 and 0.013, respectively.
Hypothesis testing for all models is summarized in
Table 4.
The regression of EC on SCS at test-day using different EC parameters was always significant except when the regression parameter was the slope obtained from a linear regression of EC collected in the 5-day period. The estimates of the regression coefficients within parities are in
Table 5.
The relationship between EC and SCS was always positive regardless of the EC regression parameter used. The magnitude of the effect varied across parities ranging from a minimum of 0.245 0.074 in 3rd parity, when the regression was on EC collected 3 days before milk recording, to a maximum of 0.953 0.20 in 1st parity, when the regression was on EC standard deviation.
In order to evaluate how well the different models fit the data, three parameters were used: the Average Information Criteria (AIC), the marginal R
2, which considers only the variance of the fixed effects (i.e., without the random effects), and the conditional R
2, which takes both the fixed and random effects into account. Results, ordered by the larger conditional R
2 value (i.e., the best model), are given in
Table 6.
According to AIC and to both Marginal and Conditional R2, the best results were obtained when the regression parameter was the , followed by the model that fitted a regression on , on and on , respectively.
4. Discussion
The electrical conductivity of milk was introduced as an indicator parameter of mastitis and subsequently used in many species, attaining different aims [
29,
30,
31,
32]. Indeed, Paudyal et al. [
32] carried out a study in Holstein cows and observed that differences in EC and MY characteristic temporal patterns due to particular pathogen groups may provide indications for differentiate groups of mastitis-causing pathogens. Other authors [
29,
30,
31], instead showed that EC measurement in sheep and donkeys is a useful way to identify animals with high SSC levels and with potentially unhealthy mammary glands, thus saving time and money, reducing the cost of other analysis (e.g., bacteriological analyses). Unlike other parameters, such as bacterial culturing-based detection of pathogens, that is still considered a gold standard [
33,
34] or molecular biological tools, the EC has the advantage of being automatically measured during milking, through electrodes inserted in the milking system. The EC measurement can be converted into a computer-readable signal and therefore this method is easily applicable to on-line automatic udder health monitoring and easily installed in the milking machine [
35]. As specified above, an alteration of EC is observed when an alteration of the concentration of some ions is recorded in the milk [
13,
14]. This feature would allow the breeder to monitor in real-time the udder health of the herd and foresee the onset of problems.
Nowadays, only limited information is available on the usefulness of EC in buffalo species and no previous study investigated EC reference ranges in Mediterranean Buffaloes. The standard range suggested for EC of normal milk in dairy cattle is 4.0–5.5 mS/cm [
21]. The mean values of EC for healthy, subclinical and clinical mastitis milk suggested by Norberg et al. [
15] are 4.87 mS/cm, 5.37 mS/cm and 6.44 mS/cm, respectively. Overall, we have observed a direct relationship between EC and SCS, which agreed with previous findings in
Bos Taurus,
Bos Indicus and
Bubalus bubalis [
36,
37,
38,
39]. However, it is the relationship magnitude depended on parity, being larger for first and fourth parity cows followed by second, third and fifth parity. Observed EC values were higher than those reported by other authors in goats [
40,
41,
42], ewes [
43], dairy cattle [
38,
44,
45] or buffaloes [
39,
46] but similar to those reported in Holstein cows by Paudyal et al. [
32].
Parity and SOL were associated with different EC values, as observed in previous studies [
32,
35,
39]. Higher EC values were recorded at both the beginning and the end of lactation, contrary to what was observed in sheep, where the increase was observed only at the beginning of lactation [
43]. In particular, in this study [
43] a strong increase of EC at the beginning of lactation and only a slight, not significant increase at the end was observed. It is likely that variations in milk composition may have affected these results, since the authors recorded that the milk composition (and particularly the fat concentration) explained high variance in EC. It can be hypothesized that the differences in fat composition may explain our results. In a recent study carried out in cows [
45], an EC increase was recorded independently of parity. Interestingly, in our study EC remained almost unchanged throughout lactation only in primiparous buffaloes, whereas a rise was observed in pluriparous counterparts. In particular, for each unit increase in EC, daily milk reduces by a minimum of −0.43 ± 0.082 kg in parity two to a maximum of −0.62 ± 0.080 kg in parity 5+. This relationship between EC trend and parity has also been recorded in other species such as cows [
32], sheep [
43] and goats [
17]. Moreover, the magnitude of the differences due to parity is similar to those previously observed in other studies [
32,
41]. It cannot be ruled out that this pattern is due to the increased susceptibility of pluriparous to udder damages caused by repeated automatic milking or by some mastitis events [
47,
48,
49].
Regarding milk composition, significant changes in PP were observed in relation to EC with a similar unfavorable effect in 3rd and 5th + parity. This disagrees with previous studies carried out in cattle [
50] and sheep [
31], where EC was positively associated with milk protein content. According to them, casein would influence the milk conductivity through the insoluble salts that can be linked to the micelles in the colloidal phase. Indeed, casein shows a very low conductance compared to the milk salts. Sometimes, these micelles break down and the salts are released, causing an increase in EC. A possible explanation of the observed unfavorable relationship between EC and PP could be found in the decrease in milk protein content caused by mastitis [
39,
51]. Indeed, in such a situation a high proteolytic enzymatic activity is observed that eventually damages milk caseins in the mammary gland. It cannot be ruled out that both the higher protein content and the different casein profile recorded in buffalo milk may account for the different results. In this study, no significant relationships were observed between EC and FP, although several authors observed a negative correlation between FP and EC in dairy cows [
38], and in goats [
40] and ewes [
30,
43]. According to them, fat globules increase the real distance during ion migration and interfere with the electrodes when measurements of EC are done. Thus, EC is expected to decrease in proportion to FP. Buffalo milk shows about 8% fat content with large variability from 6% to 12% [
52] that is more than double compared to cow milk. It is likely that the high fat content and the large variability of fat percentage may have partially affected its relationship with EC. Further studies are needed to investigate this interesting aspect.
Since conductivity is a relatively simple and inexpensive inline detection technique [
53], previous researchers have studied EC for the diagnosis of mastitis. Results were controversial and in most of them it was suggested that EC alone cannot be used for this purpose, not being a useful method to determine udder health [
18,
38]. For this reason, the second aim of the present work was to evaluate the relationship between EC collected 1 (
), 3 (
) and 5 (
) days before official milk recording and SCS at official milk recording (
), to eventually develop a prediction model for the detection of the disease. Although no cross-validation models were applied, our results suggest that using data collected at different time points works better than a single data point and that their mean is the best parameter to be included in a possible prediction model. Interestingly, a parameter such as
, which is supposed to be more informative being related to a change in EC across the observed period [
54], was not significant in relation to SCS. A possible explanation of this result might be related to the reduced observed time period (i.e., 5 days before milk recording) used to estimate
. Indeed, fitting a linear regression in such a short period and with only 4 data-points could have shrunk toward zero the estimate of the regression coefficient (i.e.,
), which eventually was not significant. A similar approach, based on the assessment of electrical conductivity across several days, was also used by Kathun [
55]. In this study, a logistic mixed model was used but results showed that for the early detection of mastitis, multiple EC measurements were more informative than a single record. Moreover, Bobbo et al. [
56], in a recent study about the prediction of somatic cell count at the subsequent test-day record in the Italian Mediterranean Buffaloes, showed that EC was the 3rd most important source of variation, following SCC-traits recorded at the previous test-day and before other traits as milk production, parity and stage of lactation.