Deeper exploration of inflammatory cell populations in milk to monitor udder health in dairy cows

ABSTRACT The present study explored the predictive values of milk leukocyte differentials (MLD) as a basis for improving the diagnosis of intramammary infections (IMIs) and subclinical mastitis. Quarter milk samples were collected for bacteriological analysis, quarter somatic cell count (qSCC), and MLD. The MLD were assessed using the cytospin technique, direct microscopic smears, and flow cytometry. The predictive values of each single leukocyte population and useful potential indices that could better reflect immune complexity were also calculated. Changes in the percentage of any leukocyte alone failed to substantially improve the predictive value of qSCC in diagnosing IMIs. Although certain parameters increased the area under the receiver operating characteristic curve (ROC curve) as a result of increased specificity values, a slight reduction in sensitivity was observed. The so-called CD8 complex was a unique parameter which improved both the sensitivity (78.79 %) and the specificity (80.77 %) in IMI diagnosis, resulting in the highest area under the ROC curve (0.87). To diagnose subclinical mastitis, the percentage of macrophages and the sum of the percentage PMNLs and T CD8+ cells divided by the percentage of macrophages showed the highest predictive values (sensitivity = 79.63, specificity = 73.68, and area under the ROC curve = 0.83) in the differentiation of the inflammatory condition status of cows. In conclusion, this study provides further insights into using T CD8+ lymphocytes in diagnosing bovine IMIs, combined with PMNLs and macrophages. The antidromic trend of macrophages vs. PMNLs and T CD8+ lymphocytes due to the increasing qSCCs was crucial to differentiating quarters under both inflammatory and non-inflammatory conditions.


Introduction
The definition of bovine mastitis has not always been consistent.Markedly, the terminology intramammary infection (IMI) and subclinical mastitis are often used interchangeably.Because of this concern, IMIs require the presence of an infectious pathogen, which usually entails an increase in milk somatic cell count (SCC).Conversely, subclinical mastitis indicates an inflammatory process and does not entail an infection, although it is often caused by a bacterial infection (Andersen et al., 2010).Under this scenario, the bacteriological examination is crucial to define IMIs.At the same time, the SCC is extensively used as the gold standard for measuring inflammation and is, therefore, included as a component of the definition of mastitis (Pyörälä, 2003;Andersen et al., 2010).
Somatic cell count measures all somatic cell types in milk but it does not distinguish the different cell populations present.Thus, it has been proposed that differential cell counting could provide a more reliable udder health status of the mammary gland (Pyörälä, 2003;Koess and Hamann, 2008;Takano et al., 2018).Light microscopy and flow cytometry can obtain differential cell counts (DCCs).Microscopic DCC is a simple and cost-effective method although several researchers prefer flow cytometry analysis on account of its greater accuracy (Koess and Hamann, 2008;Pilla et al., 2013;Takano et al., 2018).
The distribution and counts of leukocytes are critical to mammary gland defenses.Therefore, as regards the complexity of the immune system and participation of all cell types in immune responses, the quantification of a single cell type may not provide the most reliable data for identifying what is different for diagnosis decisions (Leitner et al., 2015).Given this background, the present study aimed to explore the predictive values of several parameters using milk leukocyte differentials to improve the identification of IMIs and subclinical mastitis that could better stimulate immunity during infection or/and inflammation.

Materials and Methods
This study complied with the Ethical Principles in Animal Research.It was approved by the Bioethics Commission of the Faculdade de Medicina Veterinária e Zootecnica -Universidade de São Paulo (Process n. 1685(Process n. /2009)).

Animals and sampling
The present study collected 112 quarter milk samples from 28 clinically healthy Holstein dairy cows [mean daily milk yield = 24.19 ± 0.71 kg; mean parity = 2.65 ± 0.16; mean days in milk (DIM) = 196.6 ± 12.13] from a commercial dairy herd.Immediately postpartum (up to 21 DIM) animals were not used for this study.
Prior to the milk sampling, the strip cup test was performed to detect any abnormal secretions.Next, pre-dipping was carried out, and one towel was used for each teat.After discarding the first three streams the ends of the teats were scrubbed with 70 % ethanol using a piece of cotton.Quarter milk samples were aseptically collected for microbiological analysis as the NMC (1999) described.Following this, milk samples were collected by hand milking in sterile polypropylene vials (cat.n.CLS4558-300EA, Corning) for the quarter SCC (qSCC; 40 mL) and the differential cell counts by direct microscopic smears (10 mL), and in sterile polypropylene bottles (cat.n. 3120-0500, Nalgene TM PPCO Centrifuge Bottles, ThermoFisher Scientific) for the differential cell counts by cytospin and cytometric flow analyses (200 mL).Milk samples were kept at 4 °C until arrival at the laboratory.Next, milk samples were randomized and codified, and the additional milk analyses were conducted without knowing the quarter's status.

Bacteriological culture
The bacteriological culture was carried out by culturing 0.01 mL of each milk sample on 5 % sheep blood agar plates (Becton Dickinson GmbH).The plates were incubated for 24-72 h at 37 °C, followed by observation of colony morphology, Gram staining, and biochemical testing (Oliver et al., 2004).The sample was considered culture-positive when > 1 colony grew (< 100 cfu mL -1 ).

Direct microscopic smears
Milk leukocyte differentials were determined by direct microscopic smears using fresh milk (Blagitz et al., 2013;Takano et al., 2018).The milk smears in duplicate were stained with the Rosenfeld dye (Rosenfeld, 1947), a combination of May-Grunwald and Giemsa dyes, and the polymorphonuclear and mononuclear leukocytes were differentiated at a magnification of × 100.

Separation of milk cells for flow cytometry and cytospin centrifuge
For flow cytometric and cytospin analysis, 200 mL of milk from each mammary quarter was diluted with 200 mL of PBS.Milk cells were separated as previously described by Blagitz et al. (2015) and Souza et al. (2022).In brief, milk samples were centrifuged at 1,000 × g for 15 min, and the cream layer and supernatant were discarded.The cell pellet was washed once using 30 mL of PBS and centrifuged at 400 × g for 10 min.Next, the cells were resuspended in 1 mL of RPMI-1640 nutritional medium (R7638, Sigma Aldrich) supplemented with 10 % fetal bovine serum (Cultilab) and counted using a Neubauer chamber.Next, the milk cells were placed in 1 mL of RPMI-1640 nutrition medium (R7638, Sigma Aldrich) supplemented with 10 % fetal bovine serum (Cultilab).Cell viability was assessed using trypan blue exclusion and counted utilizing a hematocytometer.The milk cells were then adjusted to 2 × 10 6 viable cells mL -1 using the nutrition medium and 10 % fetal bovine serum to dilute the milk cell suspensions so as to achieve the target concentration.

Cytospin technique
The cytospin technique for differential leukocyte counts was used as previously described by Della Libera et al. (2004) and Takano et al. (2018).In short, 200 μL of milk cell suspension in triplicate was centrifuged at 28 × g for 6 min using a cytocentrifuge (Cytospin 3 SHANDON ® ).Next, the smears were stained with the Rosenfeld dye (Rosenfeld, 1947), a combination of May-Grunwald and Giemsa dyes, and 400 leukocytes per sample were differentiated into lymphocytes, macrophages and polymorphonuclear leukocytes at a magnification of × 100.

Identification of milk leukocytes by flow cytometry
Identification of milk leukocyte populations was based on their cytoplasmatic granularity and mean fluorescence intensity following 2-step fluorescent immunolabeling with primary anti-bovine monoclonal antibodies (mAbs) and the secondary antibodies (Ab) coupled to the longwavelength fluorescent probes (Table 1).In brief, 100 μL of milk cells (2 × 10 5 viable cells) were washed with PBS and incubated with the primary mAbs for 30 min on ice to detect CD21 (tube A), and combinations of CD3, CD4 and CD8 (tube B), and CH138 and CD14 (tube C) in polypropylene tubes suitable for flow cytometry as previously described (Della Libera et al., 2015;Souza et al., 2020).After washing with PBS, the cells were incubated for 30 min at room temperature with the secondary Abs.Following this, the cells were washed with PBS and analyzed by flow cytometry (FACSCalibur, BD Bioscience).Twenty thousand milk cells were analyzed in each sample, excluding most cell debris.
A gating strategy to differentiate polymorphonuclear leukocytes (PMNLs) and macrophages was used because CD14 can also be expressed to a lesser extent on bovine milk PMNLs (Souza et al., 2020).An unstained control, secondary antibody control, and single-stained milk samples were also prepared as compensation controls.FlowJo software (TreeStar Inc.) was used to examine the data.

Statistical analysis
All the statistical analyses to determine the predictive values [i.e., sensitivity, specificity, and area under the curve of the receiver operating characteristics (ROS)] of all parameters were calculated using the quarter milk microbiological culture outcomes (Dingwell et al., 2003;Ferronatto et al., 2018) as a gold standard.The predictive values of the percentage of each leukocyte population, single-cell ratios and double interactions ratio were also calculated using widely accepted milk qSCC thresholds of 100,000 cells mL -1 (Bansal et al., 2005;Pilla et al., 2013) and 200,000 cells mL -1 (Pyörälä, 2003;Schukken et al., 2003) as a gold standard.To determine the PMNLs, macrophage and lymphocyte counts, the percentage of each leukocyte population obtained by the flow cytometry method was multiplied by the qSCC.The predictive values of qSCC (total milk cells) and the milk leukocyte differentials were assessed by cytospin and direct microscopic smears.
The percentage and counts of several cell types were determined by flow cytometry, certain singlecell ratios, such as CD4 + /CD8 + T lymphocyte ratio, widely used in specific infectious diseases (Marco et al., 2018); and other single-cell ratios proposed for the diagnosis of mastitis: 1) T cells/B cells ratio (Schwarz et al., 2013); 2) PMNLs/Lymphocytes ratio (Pilla et al., 2012); 3) Phagocytes/Lymphocytes ratio (Pilla et al., 2012); and 4) PMNLs/Lymphocytes ratio (Gonçalves et al., 2017).Furthermore, potential useful indices were also calculated: double ratio interaction ; an internal complex relationship that could better reflect immunity complexity (Leitner et al., 2015), and the so-called CD* complex, a novel proposed index calculated in the present study: . This final index was positioned in accordance with the antidromic trend of macrophages vs. PMNLs and T CD8 + lymphocytes at increasing qSCCs.
The ROC area under the curve was calculated by determining the point at the minimum distance from the left-upper corner of the unit square and the point where Youden's index is at its maximum (Habibzadeh et al., 2016).In the generalized linear regression models, lactation stage and parity were considered independent while a diagnostic variable was considered dependent.The prediction accuracy of the models was evaluated using the area of curvature ROC and the model's optimal sensitivity and specificity.The "pROC" (Robin et al., 2011) and tidy (Wickham et al., 2019) packages for the R programming language were used to develop the two ROC curvature alternatives.
The correlations between qSCC and the percentage of each leukocyte population determined by flow cytometry were determined using Spearman correlation for nonparametric data.They were carried out using the GraphPad Prism 9.0 software® (GraphPad Software, Inc.).The statistical significance was set at p ≤ 0.05.
The predictive values of all investigated parameters were summarized in Figures 1 and 2 and Table 2.In the current study, the T cell/B cell ratio and T CD4 + /CD8 + ratio could not be recommended as a tool for diagnosing IMIs (Table 2, and Figures 1 and 2).It was observed that the    percentage of any leukocyte alone did not substantially improve the predictive values of the diagnosis of IMIs.In this regard, although certain parameters increased the area under the ROC curve due to an increase in the specificity values (i.e., T CD8 + counts, T cells counts, B cells counts, PMNLs counts, percentage of macrophages, and the sum of the percentage PMNLs and T CD8 + cells divided by the percentage of macrophages), they were associated with a slight reduction in sensitivity (Figures 1 and 2).In this regard, the so-called CD8 complex better reflects the immunity complexity and was the only parameter that improved both sensitivity and specificity, resulting in the highest area under the ROC curve and sensitivity values used in IMI diagnosis (Figures 1 and 2).Additionally, the percentage of macrophages and the sum of the percentage PMNLs and T CD8 + cells divided by the percentage of macrophages showed the highest predictive values in the differentiation of the inflammatory condition status of dairy cows using both somatic cell counts thresholds (100,000 and 200,000 cells mL -1 ; Figures 3, 4, 5 and 6, and Tables 3 and 4).The predictive values of all investigated parameters pertaining to the milk qSCC thresholds were summarized in Figures 3, 4, 5 and 6, and Tables 3 and 4.

Discussion
In the present study, while the predictive values of many parameters used to diagnose IMIs (pertaining to the widely used bacteriological outcomes as a gold standard) were explored, only a few parameters showed slight improvements in the predictive values when compared to the widely used qSCC.Thus, the differential somatic cell count did not robustly increase the predictive values of qSCC, which could be explained, at least in part, by the redundancy of host immune defenses (Nish and Medzhitov, 2011; Leitner et al., 2015), reinforced by the correlations between qSCC and distinct cell populations (Figure 3).In agreement with the findings of the present study, Schwarz et al. (2019), Lozada-Soto et al. (2020), and Zecconi et al. (2021) had reported quite similar ROC curve values when compared to the total leukocyte counts and differential leukocyte counts at the end of the lactation period and in fresh cows.Analogously, Schwarz et al. (2020), using many milk samples found fairly comparable predictive values of SCC alone when compared to DCC alone (using a combined proportion of PMNLs and overall lymphocytes) and in combination with SCC.Furthermore, although several studies have indicated statistical significance between diseased and healthy quarters using differential cell counts, it did not signify discrimination, as achieved significance failed to show nonoverlapping data distribution (Leitner et al., 2015).
The association between increased cell count and PMNL influx to the mammary gland during the infectious process is notorious (Paape et al., 2003;Souza et al., 2012;Gonçalves et al., 2017).Furthermore, a recently published study indicated an even more significant increase in the T lymphocyte CD8 + subpopulation (Souza et al., 2020), which could aid differentiation between healthy and diseased udder quarters.Overall, identifying T CD8 + lymphocytes in milk samples could have great implications for IMI diagnosis and prognosis (Sordillo et al., 1997;Park et al., 1993;Alnakip et al., 2014;Souza et al., 2020).While the study size was limited, these data corroborated the findings of the current study, wherein the identification of T CD8 + lymphocytes in milk samples improved the predictive values of the variables evaluated in the diagnosis of IMIs.For example, the combination of the percentage and the number of T CD8 + lymphocytes and PMNLs, as these populations increased during IMIs, divided by the percentage and counts of macrophages, which represent the major population in healthy quarters while the percentage decreases during infection, resulted in a calculated novel index, the so-called CD8 complex.Consequently, the overall strategy resulted in the highest predictive values found in the present study.However, further longitudinal studies are needed, as the immune response is not static (Leitner et al., 2015).
Conversely, the CD3/CD21 failed to distinguish culture-negatives from positives, resulting in the poorest predictive values for IMI diagnosis.In this regard, Schwarz et al. (2013), analyzing the proportions of CD2 + T and CD21 + B lymphocytes suggested the use of CD2/ CD21 index as a new marker to determine udder health, which, at least in part, was not supported by this study, using CD3 mAb instead of CD2 mAb to identify T lymphocytes.
Although there is no consensus in the literature on the predominant leukocyte population in healthy udders, as a number of studies have suggested that lymphocytes are the main leukocyte population (Dosogne et al., 2003;Schwarz et al., 2011a;Schwarz et al., 2011b;Pilla et al., 2012), the present study using a precise high-throughput flow cytometry method with specific monoclonal antibodies to differentiate leukocytes populations widely used qSCC thresholds as a gold standard).In contrast, the percentage of milk macrophages determined by cytospin had poor predictive value.In this regard, determining leukocyte populations by flow cytometry resulted in higher predictive values for the percentage of both macrophages and PMNLs applied to diagnosing IMIs and subclinical mastitis.Overall, these findings reinforce the idea of the poor repeatability of traditional methods, such as the cytospin technique and direct microscopic smears, due to the subjective evaluation of the relatively low number of milk cells (Koess and Hamman, 2008;Takano et al., 2018).Indeed, in a previous study (Takano et al., 2018), the measurement of distinct milk populations by flow cytometry resulted  ------------------------------% ------------------------------ In this regard, although many studies have investigated the fluctuations of each leukocyte percentage by microscopic differential somatic cell counts (Koess and Hamann, 2008;Schwarz et al., 2011a;Gonçalves et al., 2017) or flow cytometry (Pillai et al., 2001;Dosogne et al., 2003;Koess and Hamann, 2008;Schwarz et al., 2011b;Pilla et al., 2013;Schwarz et al., 2019), the role of the distinct T cell subpopulations (e.g., T CD4 + and T CD8 + lymphocytes) in the diagnosis of IMIs has long been neglected.In addition, even when using flow cytometry, the mAb utilized can affect the DCC outcomes (Souza et al., 2020).In this study, PMNL, monocyte/macrophage, and lymphocyte subsets were accurately identified, whereas others evaluating milk DCC by flow cytometry did not use CD14 and CH138A mAb in combination (Pillai et al., 2001;Rivas et al., 2001;Dosogne et al., 2003;Koess and Hamann, 2008;Schwarz et al., 2011a;Pilla et al., 2013).Not doing so can lead to erroneous identification of some PMNL, which can also express CD14 on their surface (Paape et al., 1996;Sládek et al., 2002;Ibeagha-Awemu et al., 2008).Even worse, a number of studies did not use a specific mAb to accurately differentiate and identify many milk cell types (Pillai et al., 2001;Dosogne et al., 2003).Other studies used CD11b mAb to identify PMNL or to differentiate PMNL (CD11b + /CD14 -) and macrophages  corroborated several studies which have reported that macrophages are the main population in milk from healthy udder quarters (Sarikaya et al., 2005;Damm et al., 2017;Gonçalves et al., 2017;Takano et al., 2018).(CD11b -/CD14 + ; Koess and Hamann, 2008;Schwarz et al., 2011a;Pilla et al., 2013) though lymphocytes and macrophages can also express CD11b (Riollet et al., 2001;Duan et al., 2016).Furthermore, comparing the DCC of milk obtained among studies can be complicated because the type of material of the sample bottle and the method of preparation could impact the leukocyte populations (Schröder and Hamann, 2005), beyond the effect of different milk fractions on milk cell populations (Sarikaya et al., 2005).
In conclusion, this study further provided the first insights into the T CD8 + lymphocytes in diagnosing bovine IMIs.Combined with PMNLs and macrophages, it improved the predictive value of differential cell counts in the diagnosis of IMIs.Furthermore, due to an increase in the percentage of PMNLs and T cells, the markedly dampened percentage of macrophages was crucial to the differentiation of udder quarters under both inflammatory and non-inflammatory conditions.

Figure 7 -
Figure 7 -Heatmap illustrating the correlations between somatic cell counts and the differential leukocytes subpopulations determined by flow cytometry in bovine milk (r value).SCC = quarter somatic cell counts; PMNLs = polymorphonuclear leukocytes.

Table 1 -
Monoclonal antibodies used for labeling and differentiation of bovine milk leukocytes in flow cytometric analysis.

Table 2 -
Predictive values and cutoff points that maximize the specificity and sensitivity of the different parameters to diagnosed intramammary infections used in the present study.

Table 4 -
Predictive values and cutoff points that maximize the specificity and sensitivity of the different parameters to diagnosed subclinical mastitis used in the present study.

Table 3 -
Predictive values and cutoff points that maximize the specificity and sensitivity of the different parameters to diagnosed subclinical mastitis used in the present study.