Porcine circovirus 2 (PCV2) is one of the most widespread pathogens affecting swine populations worldwide, contributing to a range of clinical manifestations collectively known as porcine circovirus diseases (PCVD) (1). Since its emergence in association with the disease in the late 1990s, PCV2 has become a must-control infectious agent in nearly all commercial pig-producing regions (2). However, PCV2 has been endemic in pig populations since at least the 1960s (3). In addition, wild boar populations are also endemically infected with this virus, harbouring genotypes closely related to those in domestic swine (4, 5).

Porcine circovirus 2 is a small, non-enveloped, single-stranded DNA virus belonging to the family Circoviridae (6). The virus causes systemic infections, primarily targeting lymphoid tissues, including tonsils, lymph nodes, and spleen (7, 8). Lymphoid depletion (LD) and histiocytic replacement (HR) are the primary histopathological features associated with PCV2 infection, serving as the diagnostic hallmarks of PCV2 systemic disease (PCV2-SD) (9). Diseased animals experience immunosuppression, making PCV2-SD-affected pigs more susceptible to secondary infections (7). However, PCV2 is not restricted to the immune system, as viral replication has also been observed in the lungs, liver, kidneys, heart, gastrointestinal tract, and reproductive tissues (8, 10–18). Diagnostic confirmation of PCV2-SD generally relies on a three-step approach: (1) the presence of clinical signs and a compatible herd history, (2) the identification of moderate to severe histological lesions in target organs, and (3) the detection of moderate to high amounts of PCV2 within those tissues (19–21).

Detection of PCV2 within tissues is commonly performed using immunohistochemistry (IHC) and conventional in situ hybridisation (C-ISH) to identify PCV2 antigens and genome, respectively (22, 23). More recently, RNAscope^® technology has been used for research purposes, offering higher sensitivity by detecting individual copies of single-stranded nucleic acids as distinct red dots in tissue sections (24). Additionally, digital pathology tools such as QuPath^® (25) have improved the accuracy and reproducibility of viral quantification by enabling automated image analysis of IHC-stained sections (26).

The present study aimed to evaluate a newly developed automated pixel classifier specifically designed for lymphoid tissue samples from experimentally infected animals using PCV2 ISH RNAscope^® technology, by correlating its results with visual scoring. Furthermore, the correlation between both visual and digital scores and the PCV2 load in tissue, quantified by real-time quantitative (quantitative polymerase chain reaction [qPCR]), was also analysed.

None of the animals exhibited clinical signs compatible with PCV2-SD during the study, and no significant findings were observed at necropsy. Histopathological analysis revealed that only 7.6% of animals (5/66) had LD and/or HR lesions. Two of them had an HR score of 1 in the TBLN, while another two pigs showed the same score of 1 in 3 different lymph nodes (TBLN, MSLN, and ISLN). The fifth animal had a more extensive involvement, with an HR score of 2 and an LD score of 1 in the TBLN and ISLN, as well as an HR score of 1 in the MSLN and TO.

The visual scoring system ISH identified PCV2 labelling in 34.8% (23/66) of TO, 42.4% of TBLN (28/66), 33.3% of MSLN (22/66), and 28.8% of ISLN (19/66) samples (Table 1). Based on these findings, the following thresholds for the percentage of labelled area (digitally quantified) corresponding to the visual scores were established as follows: ≤0.0025% (score 0), 0.0026–1.0% (score 1), 1.01–5.0% (score 2), and >5.0% (score 3). A representative image for each of the score values of PCV2 ISH is provided in Figure 1. The automated pixel classifier detected PCV2 in 34.8% (23/66) of TO, 42.4% of TBLN (28/66), 31.8% of MSLN (21/66), and 30.3% of ISLN (20/66) samples (Table 1). The comparison between the digital and visual PCV2 ISH scores across all tissues revealed a strong and statistically significant Spearman’s correlation (ρ = 0.96) (p < 0.05). Conversely, when both scores were correlated with LD and HR lesion scoring obtained through histopathological analysis across all tissues, lower correlations were observed for both digital and visual scores with LD (ρ = 0.16) and HR (ρ = 0.38), which were statistically significant (p < 0.05).

When digital and visual scoring were analysed separately for each tissue, the highest correlation was observed with TBLN (ρ = 0.98), while the lowest correlation was observed with TO (ρ = 0.94), and this difference was statistically significant in all cases (p < 0.05) (Figure 2). Moreover, in both scoring systems, TBLN had a significantly higher number of animals with a visual score of 2 in TO (p < 0.05), as well as in MSLN and TO for a digital score of 1.00–5.00% (p < 0.05) (Table 1). Given these results, TBLN was the tissue of choice to be processed by qPCR.

When TBLN viral load was assessed with PCV2 qPCR, 74.2% (49/66) were positive. Out of these 49 qPCR-positive samples, 28 showed positive results for both ISH and qPCR (Table 2). The remaining 21 samples that were above the LOQ were classified as score 0 by the visual scoring of ISH and showed values between 0.0 and 0.0025% by the ISH digital scoring, and these samples corresponded mostly to the lowest viral loads within the positive range (Supplementary File 1). Finally, a Spearman’s correlation analysis between PCV2 qPCR results and both ISH visual and digital scoring classification revealed a strong and statistically significant correlation among the three techniques (ρ = 0.85) (p < 0.05).

Since PCV2-SD diagnostics rely on clinical signs, histopathological lymphoid lesions, and detection of the virus in these damaged tissues, the present study developed and evaluated an automated pixel classifier for quantifying PCV2 in lymphoid tissues using ISH RNAscope^® technology. Moreover, such automated measures were correlated with the traditional visual scoring (23), LD and HR histopathology lesion scoring, and qPCR-based viral load (29) quantification (in TBLN) in PCV2 experimentally infected pigs.

The automated pixel classifier showed a very strong concordance with visual scoring across all lymphoid tissues, suggesting that digital quantification may eventually replace optical evaluation with enhanced objectivity and reproducibility. These findings align with previous research that used digital quantification of IHC results in animals with different degrees of LD and HR (26). Notably, in the present study, ISH RNAscope^® detected PCV2 loads as low as 5.3 log₁₀ DNA copies/mL of lymph node supernatant as positive (score 1), outperforming C-ISH and IHC techniques, which typically require higher viral loads (two to three more log₁₀ DNA copies) for reliable detection (29, 30). The results are consistent with a recently published study that also reported the superior sensitivity of ISH RNAscope^® to detect PCV2 when compared with C-ISH and IHC in porcine dermatitis nephropathy syndrome cases (27), a PCVD that typically displays lower viral loads compared to PCV2-SD.

Notably, the histopathological evaluation in this study confirmed that LD and HR were observed only in a subset of animals. Furthermore, no animal exhibited clinical signs compatible with PCV2-SD throughout the study, highlighting the subclinical nature of this experimental infection. These findings reinforce the relevance of using highly sensitive molecular techniques, such as ISH RNAscope^®, for detecting PCV2 in subclinical infections, where traditional histopathological approaches may be insufficient. This is demonstrated in the present study by the low Spearman’s correlation coefficients observed between histopathological LD and HR lesion scores and both digital and visual ISH scores, as well as by the lower number of animals with PCV2-compatible lesions detected through histopathology compared to those identified using ISH RNAscope^®.

A key aspect of the present study was the selection of the TBLN for viral load assessment, given its role as a primary site for PCV2 replication (20, 23). Under the present experimental conditions, TBLN consistently exhibited higher viral loads compared to other lymph nodes, probably due to the nature of the inoculation route, which was intranasal.

Although ISH RNAscope^® provides a highly sensitive method for PCV2 detection, its cost remains a limiting factor for routine diagnostic applications. However, its potential for research purposes is advisable, allowing for detailed characterisation of PCV2 distribution within lymphoid tissues. Future studies should explore the applicability of this technique across different PCVD presentations and its potential for detecting early stages of infection before significant histopathological lesions develop, as well as other subclinical presentations (i.e., PCV2-reproductive disease). Such improved sensitivity could provide a deeper understanding of the disease and its dissemination through the organism.

The automated pixel classifier demonstrated robust performance in quantifying PCV2 in lymphoid tissues, with strong correlations to both traditional visual scoring and qPCR quantification (in TBLN). By reducing operator-dependent variability and facilitating the assessment of ISH results, this approach enables a more objective and efficient evaluation of viral distribution within lymphoid tissues. These findings support the integration of digital pathology tools into PCV2 research, enhancing the accuracy and reproducibility of viral detection methodologies.