1. Introduction
The current most simple, practical and sustainable method to monitor udder health in dairy herds is represented by individual somatic cell count (SCC), even if it does not have the same accuracy as microbiological analysis [
1,
2]; SCC can only suggest the presence of inflammation, but not the presence of a pathogen. Usually, under field conditions, a level of 200,000 cells/mL is considered the threshold to identify subclinical mastitis [
3,
4,
5,
6]. However, the progressive decrease in mean SCC in dairy herds worldwide has resulted in a reduction in the use of SCC to identify diseased cows. There is a consensus on the association between an increase in SCC in milk with a change in the proportion of inflammatory cells in the cell population, and it has been suggested that the amount of polymorphonuclear neutrophils (PMN) may be a more useful indicator in the evaluation of udder health than SCC [
7,
8,
9,
10,
11]. Indeed, milk composition changes were observed even below 100,000 cells/mL [
9,
12,
13,
14].
This evidence supported the research aiming to apply differential somatic cell count (DSCC) as a tool to identify mastitis in combination with SCC or alone. However, one of the major obstacles to the application of DSCC in practice was the unavailability of high-throughput milk analyzers with an acceptable cost for the analysis. Microscopy is the conventional method to perform DSCC, but it is time consuming and it has poor repeatability, while flow cytometry is more efficient, but the cost of the analysis and its accuracy are critical points that prevent its application outside the research field [
15].
The recent availability of a high-throughput milk analyzer (Fossomatic™ 7DC, Foss A/S, Hillerød, Denmark), able to perform a partial DSCC [
16] and fully integrated with the instruments currently used to analyze individual milk samples, opened new opportunities in bovine mammary gland investigations. This analyzer is now used in several countries, but it has a potential limitation related to the accuracy of measurements when the SCC is below 50,000 cells/mL [
16]. Indeed, DSCC values for samples below 50,000 cells/mL are not usually reported by the instrument, even if they are recorded.
Our recent investigation in the Lombardy region (Italy) [
17] on nearly 46,000 milk test records showed that around 25% of primiparous cows and 19% of all cows has a lactational SCC average below 50,000 cells/mL. Therefore, if DSCC data are not available for the farmer, a significant proportion of information on herd health status is missing, thus impairing the application of routine tests of this technique and the potential improvement provided by the new technology.
Thanks to the cooperation with the producer of the instrument (Foss A/S, Hillerød, Denmark), we were able to retrieve data from the analyzer for cows with an SCC ≤ 50,000 cells/mL and to develop a study under field conditions with the aim to assess the relationship between DSCC and udder health status, milk composition and milk yield in these cows. We hypothesized the that the use of DSCC is a marker for early detection of intramammary infections (IMI) and for changes in milk secretion. To support this aim, we also explored the repeatability of a DSCC in samples with a SCC ≤ 50,000 cells/mL, as well as the characteristics of the sampled cows classified according to the SCC threshold of 50,000 cells/mL.
2. Materials and Methods
2.1. Herd and Cow Selection
This study considered 3022 cows from 24 randomly selected dairy herds in the Lombardy region (Italy) among the 150 herds that at the time of the study already applied for DSCC testing. The sample size was estimated based on the proportion of cows with a SCC ≤ 50,000 cells/mL (20%) in Lombardy as reported in a recent study [
17]. More than 95% of the cows included in the study were Italian Holstein Frisian; most of the remaining ones were Italian Brown Swiss while other breeds or cross breeds represented less than 1% of the sample. Milk samples to determine their composition and the presence of bacteria by real-time PCR (qPCR) assay were collected from September to December 2018. All lactating cows in each herd were sampled only once for this purpose.
2.2. Sample Collection
To reduce the risk of contamination of the milk samples in individual cow testing, teats were carefully cleaned as required for milking before applying clusters and milk meters [
18]. Individual cow sampling was performed by certified methods currently applied by the Italian Breeders Association at the laboratories of the Regional Breeders Association of Lombardy (ARAL, Italy) by means of a Lactocorder™ (WMB AG, Balgach, Switzerland). Samples were delivered refrigerated to ARAL labs the same day and divided into two aliquots: one for the milk composition analysis and the other for the qPCR assay, kept refrigerated until the analyses were performed (within 30 h from the sampling).
2.3. Milk Composition Analysis
Milk analyses included protein, fat, lactose and casein content, as well as the SCC and DSCC, and were carried out on a Fossomatic™ 7DC (Foss A/S, Hillerød, Denmark). The analysis of DSCC is based on the Foss DSCC Method Cell Staining (international patent PCT/EP2010/065615-Holm, 2012) as described by Damm et al. [
16]. The method allows to identify within a milk sample the macrophages (MAC) and the combination of polymorphonuclear neutrophils (PMN) and lymphocytes (LYM). Diagnostic characteristics and performances were described by Damm et al. [
16]. DSCC is expressed as the combined proportion (%) of PMN and lymphocytes on the overall count of milk cells.
2.4. Repeatibility of the SCC and DSCC Measures
To assess the repeatability of the SCC and DSCC measures in milk test samples of ≤ 50,000 cells/mL, the coefficient of variation (CV%) was estimated by testing 5 different samples having SCC ≤ 50,000 cells/mL. The 5 samples were analyzed 20 times each on two identical Fossomatic™ 7DC instruments (Foss A/S, Hillerød, Denmark) for a total of 10 repeated measures.
2.5. Molecular Analysis
All samples with an SCC ≤ 50,000 cells/mL were considered for qPCR assay. Individual milk samples were analyzed using Mastit M4BDF kit (DNA Diagnostic A/S, Risskov, Denmark), following the producer’s instructions (
https://dna-diagnostic.com/files/Downloads/Mastit4/Instruction_protocol_M4BDF_2017.11..01.pdf). This kit allows bacterial DNA extraction, identification and quantification of
Staphylococcus aureus,
Streptococcus agalactiae,
Streptococcus dysgalactiae,
Streptococcus uberis,
Mycoplasma bovis,
Mycoplasma spp., coagulase-negative staphylococci,
Prototheca spp.,
Escherichia coli,
Klebsiella species,
Enterococcus and
Lactococcus lactis using qPCR. The qPCR reactions were performed on a Stratagene Mx3005P (Agilent Technologies Inc., Santa Clara, CA, USA). To avoid the potential overestimation of positive samples due to carry-over effects during sampling, when the same bacteria species was detected in two consecutive milk samples taken from the same milking slot, the second one was considered as uninfected as suggested by Mahmmod et al. [
18].
2.6. Cow and Milk Test Record Data
Milk and cow data were recorded by ARAL and included herdID, cowID, number of lactations (n), days in milk (d), milk yield (kg/d), SCC (cells/mL), DSCC (%) and grams per 100 g of milk (%) of protein, fat, lactose and casein. The SCC was log10-transformed to a somatic cell score (SCS, units). Cow, milk and molecular analysis data were combined in a database and statistical analyses were performed.
2.7. Statistical Analysis
Pearson’s correlation tests, χ2 tests and Armitage–Cochrane trend tests were applied to compare the distributions of SCC and DSCC across the whole study population by means of XLSTAT 19.4.1 software (Addinsoft, New York, NY, USA). Agglomerative hierarchical clustering, applying Euclidean distances and the unweighted pair-group average as the agglomeration method, was used to classify samples with ≤50,000 cells/mL based on their composition, and an ANOVA LSD test was applied for comparison of the means among clusters, both performed in XLSTAT 19.4.1 software (Addinsoft, New York, NY, USA). Data from cows with ≤50,000 cells/mL were also analyzed by generalized linear models applying the GLM procedure in SAS 9.4 (SAS Institute Inc., Cary, NC, USA), to identify the factors associated with the different milk traits considered.
The models were:
(a) DSCC as a marker of IMI:
where Y = DSCC; µ = general mean; Hi = effect of IMI (i = negative, major pathogens, coagulase negative Staphylococci—CNS); Dj = effect of days in milk (DIM) (j = 5–60; 61–120; 121–180; 181–240; >240 d); Pk = effect of parity (k= 1; 2; 3; 4; >4); and eijk = residual error.
(b) DSCC, milk composition and yield:
where Y = dependent variables (SCS, milk yield, fat, protein, casein, lactose); µ = general mean; H
m is the random effect of the mth herd (m = 1 to 24); Fi = effect of DSCC (i = 12.2–41.2; 41.3–51.7; 51.8–60.6; 60.7–91.6%); Dj = effect of DIM (j = 5–60; 61–120; 121–180; 181–240; >240 DIM), Pk = effect of parity (k= 1; 2; 3; 4; >4), Iw = effect of IMI (i = negative, major pathogens, CNS); and eijkw = residual error.
3. Results
3.1. Repeatability
The first step of our study aimed to estimate the CV% of the SCC and DSCC in samples with ≤50,000 cells/mL, analyzing 5 different samples below that threshold on two identical instruments (Fossomatic™ 7DC, Foss A/S, Hillerød, Denmark). The results are reported in
Table 1. The data shows that the CV% ranged between 3% and 7% for DSCC, values that are close to the ones calculated for SCC in the same samples. Both the DSCC and SCC CV% observed were very close to the reference values reported by the producer (<6% for samples in the range 100,000–299,000 cells/mL).
3.2. Field Trial—Data Description
Overall, 3022 milk samples were collected from 24 dairy herds.
Table 2,
Table 3 and
Table 4 summarize the main characteristics of the samples. The results showed that about one third of the samples (901; 29.8%) had an SCC ≤ 50,000 cells/mL. The mean DSCC was higher in samples with an SCC > 50,000 cells/mL when compared to samples with ≤50,000 cells/mL, being respectively 68% and 51%. The frequency of cows with ≤50,000 cells/mL (
Table 3) showed a statistically negative trend as the number of lactations increases (Cochran–Armitage trend test;
p < 0.0001). Indeed, nearly 40% of the cows in first lactation have ≤50,000 cells/mL, whereas only 17% of the cows with >4 parturitions were below this threshold. Analogously, when data were analyzed by DIM, we observed a significant decrease of cows with ≤50,000 cells/mL when DIM are >60 d (
p < 0.0001, as per the Cochran–Armitage trend test).
The relationship between the SCS and DSCC in the two subsamples (≤50,000 cells/mL and >50,000 cells/mL) was also assessed by Pearson’s correlation analysis. The results showed that the correlation was 0.16 (p < 0.05) when the ≤50,000 cells/mL data were considered. When samples with >50,000 cells/mL were considered, the correlation coefficient was 0.52 (p < 0.05).
3.3. Intramammary Infections and DSCC
The samples with ≤50,000 cells/mL were analyzed by qPCR to identify the presence of bacteria. Overall, 20.75% of the samples (187) were positive and the distribution of the results is reported in
Figure 1. Data show that one third of the positive samples harbored contagious pathogens, while about half of them were positive for coagulase-negative staphylococci (CNS).
The relationship between IMI and DSCC in samples ≤50,000 cells/mL were analyzed by a simple GLM model, including health status (negative, CNS and major pathogens), parturition and DIM. The model has a low R2 (0.05) even if significant (p < 0.0001), and the health status did not show any significant effects on DSCC variability. Indeed, mean values were the following: healthy cows 49.61% ± 0.80%, CNS 48.95% ± 1.46% and major pathogens 48.50% ± 1.47%.
These results were also confirmed by the analysis of the contingency table (
Table 5) comparing health status and DSCC classified by quartiles. The distribution of the samples was nearly uniform (statistically not significant following a χ2 test, α = 0.05) among the four quartiles, respectively, for the negative, major pathogen and CNS-positive samples.
3.4. DSCC, Milk Composition and Yield
The relationships between DSCC and milk composition, SCS and yield were analyzed by a general linear model (GLM), also including herd as random factor, while parity, days in milk and IMI were fixed factors (
Table 6). The model was statistically significant for all the traits considered, with an R
2 in the range between 55% (milk yield) and 15.4% (SCS). Herd factor was confirmed to have a significant effect on the variance of these traits and its contribution to the R
2 was in the range 5.3%–22.9%. IMI did not show any statistical influence on the variance of the traits considered. Protein, casein and SCS were the traits where herd had the lowest influence on their variance (≤6%), while for milk yield it has the strongest effect (22.9%). DSCC values were classified in quartiles, based on the observed distribution of the values. The influence of DSCC on the variation in milk components was always highly significant (
p < 0.0001), whereas it was not significant for milk yield and lactose. Despite our dataset included only milk samples ≤50,000 cells/mL, the statistical analysis confirmed a significant positive association between DSCC and SCS. Indeed, samples in the lowest DSCC classes also had the lowest SCS mean, whereas the highest DSCC classes showed the highest SCS mean.
The analysis of the influence of DSCC on milk fat, protein and casein showed the same pattern with a significant decrease in the proportion of the three components as DSCC increased. The decrease in the fat proportion was the most evident with an absolute decrease of 0.29% when the samples with a DSCC in the range of 12.2%–41.2% were compared to samples with DSCC in the range of 60.7%–91.6%. Protein and casein percentages showed a significant decrease when the DSCC was >51.8%, with an absolute difference between the lowest and highest DSCC class of 0.12% and 0.10%, respectively.
To confirm these relationships, data were analyzed by agglomerative hierarchical clustering, resulting in the definition of three characterized clusters, as reported in
Table 7. The statistical analysis showed that Cluster A is characterized by statistically significant, lower means for DSCC, SCS and lactose, and higher means for fat, protein and casein compared with clusters B and C. Cluster C, on the other end, is characterized by statistically significant higher means for DSCC, but the opposite for fat, protein and casein when compared with the other clusters.
5. Conclusions
This study confirms that cows with a very low SCC (≤50,000 cells/mL) may harbor IMI; however, the DSCC did not increase our capability to identify them. Despite the low level of the SCC, milk fat, protein and casein significantly declined as the DSCC was raised. Therefore, a DSCC in low SCC cows may be suggested as a marker to identify early changes in milk composition, as a result of an alteration in the milk secretion mechanism.