*2.2. Immunohistochemistry*

Serial para ffin-embedded sections (3 μm) were prepared from the CNS samples listed in Section 2.1. and stained with primary antibodies against the following antigens: ionized calcium-binding adaptor molecule 1 (Iba1) to detect microglia cells [9], CD3 to detect T lymphocytes, CD20 to detect B lymphocytes, and glial fibrillary acidic protein (GFAP) to detect astrocytes (Table 1). The sections were de-para ffinized, and endogenous peroxidase activity was blocked by incubating them with 0.5% H2O2 in distilled water for 30 min. Following antigen retrieval (Table 1), the samples were incubated for 20 min in a humidified chamber with 5% normal serum and 0.1% bovine serum albumin in Tris-bu ffered saline to prevent non-specific binding. The remaining steps in the protocol and the reagents used throughout immunohistochemistry staining were identical to those described by Martínez et al. (2020) [6]. As a negative control, the slides were subjected to the standard procedure in the absence of the primary antibodies. As a positive control, lamb lymph node tissue was stained against Iba1, CD3, and CD20. As an overall negative control, samples of CNS from healthy lambs were stained with anti-GFAP antibody, and, as positive control, samples from a sheep with CNS degenerative chronic disease were used.

**Table 1.** Immunohistochemical protocols used for cellular type characterization.


GFAP, glial fibrillary acidic protein; Iba1, ionized calcium binding adaptor molecule 1.

#### *2.3. Evaluation and Quantitation*

The stained slides were analyzed under a light microscope (Eclipse E600, Nikon, Japan) equipped with a digital camera (DS-Fi1, Nikon). Cells were counted using NIS-Elements BR imaging software (Nikon). Sections were examined in five randomly selected parenchymal fields at a magnification of 200× [9]. The abundance of microglia, T lymphocytes, and B lymphocytes was expressed as a proportion (%) of all immunolabeled cells within the perivascular cu ffing and glial foci. Astrocytes were evaluated using the following semi-quantitative scoring system adapted from a previous study on reactive astrogliosis in human disorders [13]: 1, astrocytes present in healthy CNS tissue, some astrocytes do not express detectable levels of GFAP; 2, mild to moderate reactive astrogliosis; 3, severe di ffuse reactive astrogliosis; and 4, severe reactive astrogliosis with compact glial scar formation.

In order to estimate the total area occupied by glial foci, the areas with glial foci were summed and divided by the total area of the CNS section [6] using Image J software (https://imagej.nih.gov/ij/ download.html). To evaluate the spatial distribution of glial foci, we chose medulla oblongata sections because they were consistently a ffected and had a similar size in all lambs examined [4] and we studied them using light microscopy (BX51, Olympus, Japan) and the imaging software OlyVIA 2.91 (Olympus, Japan). The images were analyzed using VS-ASW 2.8 software (Olympus, Japan).

#### *2.4. Statistical Analysis*

The proportions of immunolabeled cells that were microglia or T or B lymphocytes were submitted to an arcsin square-root transformation in order to meet the normality criterion for statistical analyses. Then, the results were subjected to cluster analysis by the centroid method in the SAS CLUSTER

procedure (SAS, Cary, NC, USA), which generated two clusters: cluster A comprised lambs 19, 21, 25, and 29, which showed a relatively loose relationship; and cluster B comprised lambs 14, 18, 26, and 28, which showed a relatively tight relationship. Microscopic lesions in each cluster did not fit nearly within the four-grade scoring system: cluster A lesions were mild to moderate (grades I and III), while cluster B lesions were moderate to severe (Figure 1) (grades III and IV) [4]. These clusters were considered to better represent the severity of lesions in the study sample of lambs. As a consequence, "cluster" was added as a variable in order to capture population variability potentially more accurately than only with individual values.

**Figure 1.** Cluster analysis of lambs experimentally infected with Spanish goa<sup>t</sup> encephalitis virus (SGEV) in terms of lesion severity using the centroid method in the SAS CLUSTER procedure. Lambs 19, 21, 25, and 29 were assigned to cluster A (mild to moderate lesions), while lambs 14, 18, 26, and 28 were assigned to cluster B (moderate to severe lesions).

A general linear model was constructed using the GLM procedure in SAS with the following dependent variables: arcsin-transformed proportions of microglia, T lymphocytes, and B lymphocytes; log-transformed semi-quantitative scores for astrocytes; and proportion of total stained area that was positive for Iba1. The independent variables were: CNS region, type of lesion (perivascular cuffing or glial foci), cluster, and their first-order interactions.

Pairwise least-square means comparisons were carried out using Student's t test as implemented in the LSMEANS statement in SAS. Statistical significance was accepted at *p* < 0.05, while *p* = 0.05–0.1 was taken to indicate a tendency [6].

Potential differences in the number of lesions (glial foci) between the infected side of the animal and the opposite side (laterality effects) were examined using the chi-squared test in the FREQ procedure in SAS and confirmed using the log-transformed foci counts submitted to analysis of variance of the generalized linear model. Laterality effects were also confirmed using a least-squares mean Student's t test with the LSMEANS statement in a model with three independent variables: cluster, side, and their interaction [6].

## *2.5. Ethics Approval*

Sampling procedures and SGEV challenge in the previous study by Salinas et al. (2017) [4] were approved by the Animal Research Ethics Committee of the Community of Junta de Castilla y León, Spain (reference number ULE\_010\_2015). Experiments were conducted in accordance with Spanish and European legal requirements and guidelines on animal experimentation and welfare.
