Animals and experimental design
Initially, for this study, a total of 30 Holstein Friesian cows (ranging from three and six years-old) were analyzed. They belonged to a commercial dairy herd in which a follow-up survey on losses due to PTB was ongoing. Animals were culled in an authorized slaughterhouse following the standard methods in the current legislation. Samples from different areas of the intestine and mesenteric lymph nodes were fixed in 10 % buffered formalin of stored at -20 ºC. Infection was confirmed in 25 of them both by bacteriological culture of frozen intestinal tissues and nested-PCR for Map DNA detection [26,39]. Five samples were negative and used as uninfected controls in the study.
Tissue samples and classification of paratuberculosis lesions
Samples from ileocecal valve, ileum, jejunum (with Peyer´s patches) and associated mesenteric lymph nodes were taken from each animal. After fixation in 10 % buffered formalin, they were routinely embedded in paraffin wax, after dehydration through a graded alcohol series and xylene treatment. Tissue sections 3-μm thick were obtained from each sample and stained with Harris´s haematoxylin and eosin (H&E) and Ziehl-Neelsen method for acid-fast bacilli (AFB) detection.
No granulomatous lesion consistent with Map infection was observed in the five cattle that were negative for both bacteriological culture and nested-PCR tests, while specific lesions were detected in samples from the intestine and lymph nodes from the remaining animals. Subsequently, granulomatous lesions compatible with Map infection were classified into 3 categories -as focal, diffuse paucibacillary, and diffuse multibacillary-, according to the guidelines previously proposed [25].
Cows with focal lesions showed well defined granulomas, composed of small groups of macrophages, located exclusively in the interfollicular areas of the intestinal Peyer’s patches. They were also present in the cortical and paracortical areas of the mesenteric lymph nodes. Scant of none AFB were detected in these lesions. This type of lesion was identified in 10 infected cows. However, diffuse lesions were characterized by a widespread granulomatous infiltrate, present both in the lymphoid tissue and in areas of the lamina propria both related or not to the Peyer’s patches, and in the mesenteric lymph node. The infiltrate was formed by epithelioid macrophages, with some Langhans type giant cells. According the number of AFB present in the lesions, they were classified as diffuse multibacillary forms, detected in 9 Map-infected animals, with a predominance of epithelioid cells in the infiltrate that harbor large amounts of AFB, of diffuse paucibacillary lesions, seen in 6 infected cows, where the infiltrate was formed by large amounts of lymphocytes with scattered groups of macrophages and giant cells, with reduced numbers of AFP present in their cytoplasm. For the immunohistochemical analysis, five animals from each group, showing the most representative granulomatous changes from each type of lesion, were selected.
Immunohistochemical study
Immunohistochemical analysis was carried out in a total of 20 jejunal tissue sections (with lymphoid tissue) and other 20 mesenteric lymph node samples, one from each animal finally included in the study. As stated, these sections were representative of the lesion category assigned to each animal. Tissue sections 3-μm thick were mounted on electro charged adhesive gelatin-coated microscope slides (Thermo Scientific, Waltham, MA, USA). For the detection of Foxp3+ T lymphocytes, a primary polyclonal antibody (rabbit IgG isotype) against bovine Foxp3 (NB100-39002; Novus Biologicals®, Centennial, USA), at a 1:150 dilution, was used. The trading house had reported through a verified customer review that this Foxp3 antibody showed reactivity in sections of formalin-fixed bovine tissues (unpublished data). Heat-mediated antigen retrieval was performed by means of PT Link® system, using pH 6.0 target retrieval solution (Dako-Agilent® technologies, Santa Clara, USA) for 20 min at 95ºC. Immunohistochemical procedure was carried out as described elsewhere [22]. Appropriate species-and isotype-matched immunoglobulins were used as control. These included sections with an isotype control for the primary antibody, and the omission of the primary antibody.
Cell counting
Due to the heterogeneous distribution of positively immunolabelled Foxp3+ T cells, a differential cell count was carried out on the LP, GALT and jejunal MLN from each sample included in the study. In each slide, 30 randomly chosen fields were selected from each of the three areas analyzed and photographed at 400x (Nikon® Eclipse Ci microscope with Digital MD-E3-6-3 digital camera). The counting of positively immunolabelled cells was performed on digital images using the Cell Counting add-on of Image J program® (U.S National Institutes of Health, Bethesda, Maryland, USA). In total, 90 fields were evaluated from each animal analyzed. In each animal, the final value of cell count was obtained by calculating the average value for the 30 fields of each intestinal area analyzed (LP, GALT and MLN) together with the average value of the 90 total fields evaluated. Finally, in order to jointly assess the different types of lesion, the mean value between the five slides of each lesion type analyzed was calculated. Additionally, the distribution of Foxp3+ T lymphocytes in relation to the presence of granulomas was evaluated, when these were found in each section.
Assessment of Foxp3+ T cells immunostaining was assessed independently by two of the authors (JE and VP, an ECVP-board certified pathologist), and discordant results were reviewed in a multiheaded microscope to reach consensus.
Statistical analysis
We used linear mixed models (LMM) for longitudinal data analysis [40] to explore the effects of infection status (healthy or infected), intestinal area (LP, GALT and MLN) and type of lesion (control, focal, diffuse paucibacillary and diffuse multibacillary) on the number of positively immunolabelled Foxp3+ T cells at an individual level. In order to perform an appropriate fitting of the LMMs, we transformed the variables in the following way: the number of Foxp3+ T cells was the response variable and was log (x+1)-transformed, and infection status, intestinal area and type of lesion were the explanatory variables. In the models, the identity of each animal was included as a random factor in order to consider the inter-individual effects on the number of Foxp3+ T cells. Package “lme4” [41] was used to fit LMMs, with the “lmer” function. The dredge, get.models and model.sel functions included in the R package “MuMIn”, were employed to construct a set of candidate models with all the possible combinations of predictive variables and, according to them, we identified the best models using an automatic selection procedure based on the Akaike Information Criterion corrected for small sample sizes (AICc) provided by the dredge function [42]. The most strong and parsimonious model was selected using the combination of lower Akaike’s Information Criterion (AICc) and the Akaike weight (Wi). Subsequently, we estimated the Akaike weight (Wi) and the percentage of explained variability (R2) for each selected model [41]. The fit quality of the selected models was evaluated through the analysis of residual deviance, linearity, multicollinearity and overdispersion using diagnostic graphics (R packages “mctest” and “car”). Goodness of fit and estimation of the parameters of the model were evaluated by means of the approximation procedures of the Restricted Maximum Likelihood Estimation (REML). We also calculated between- and within-individual variability of the best selected models. Between-individual variation refers to the classic phenomenon of individual differences in a variable, and within-individual variation refers to variability from occasion to occasion within an individual. We estimated these variances using the lmer function in R package “lme4” [41]. Subsequently, to observe the differences between groups within the relevant variables included in the final fitted model, we used the Tukey’s Honestly Significant Difference adjustment for the whole pairwise comparisons using the glht.function with the “multcomp” package in R. We produced the bar-plots of the log-transformed dependent variable by groups of explanatory variable using the ggplot function in R package “ggplot2”. To observe the differences between groups within the relevant variables included in the final fitted model, we used the Tukey’s Honestly Significant Difference adjustment for the whole pairwise comparisons using the glht.function with the “multcomp” package in R. Finally, Pearson´s rank correlation test was applied to establish possible correlations between the number of Foxp3+ T cells present in each evaluated tissue area. P-values of less than 0.05 were considered statistically significant.
All statistical analyses were performed with the R software version 3.5.3 [43].