Despite progress in diagnostic techniques and laboratory methods, a considerable number of encephalitides with unknown cause, sharing similar histopathological pattern, remains unidentified (1–3). Viruses are particularly known to induce non-suppurative inflammatory processes in different organ systems, including the central nervous system (CNS) (4). In clinic, the diagnostic process for encephalitis begins with a general and neurological examination based on the patient’s medical history, followed by an analysis of the cerebrospinal fluid (CSF) (5). The CSF is tested for glucose, protein and cell content, and various polymerase chain reaction (PCR) methods are used to detect potential pathogens (5). Since pathogen detection is highly dependent on the timing of CSF collection, it is also recommended to assess antibodies against common encephalitis pathogens in both CSF and serum using enzyme immunoassays (5). The next step in the diagnostic process is neuroimaging, preferably with magnetic resonance imaging (5). In approximately 30–60% of the cases, the etiology remains undetermined, and brain biopsies or post mortem tissue samples are further analyzed in pathology (6). The first step involves a histopathological evaluation of the tissue. The distribution and extent of inflammation, the type of inflammatory cells and specific cytological features such as inclusion bodies provide crucial diagnostic insights (5). Particularly, non-suppurative encephalitis often suggests a viral cause (7). In cases, suspicious of having a viral etiology, tissues will then be tested for known viruses using conventional methods including immunohistochemistry (IHC), various PCR techniques or in situ hybridization (ISH) (3, 8). Instead of singleplex PCR tests, multiplex PCR and microarray are utilized, as they enable the detection of different viruses in a single run (9). Microarrays also provide the advantage of analyzing the pathogen’s genotype (5). While multiplex PCR and microarrays are often considered less sensitive than real-time PCR, all three methods exhibit comparably high specificity (10, 11). Even after further examination and exclusion of all common well-known viruses, a large proportion of cases remains undetermined (12–14). In a previous study investigating non-suppurative meningoencephalitides in dogs and cats, a viral cause could not be demonstrated in the majority of cases (4). This might be related to the fact that most conventional methods are virus-specific and may not detect viral variants or newly emerging viruses. Next Generation Sequencing (NGS) provides a sequence-independent approach that offers the possibility to identify so far overlooked viral variants or newly emerging viruses. Nevertheless, the generated data must possess a certain degree of similarity to already known viruses in order to achieve a match by comparing the obtained results with data available (8, 15). Additionally, this method is very cost-intensive and time-consuming (16–18). The success in discovering new viruses relies not only on the methods employed but also on the material utilized for the investigations. Formalin-fixed, paraffin-embedded (FFPE) specimens are the most commonly used material in pathology and offer several advantages in the field of virus discovery (19, 20). To access viral antigens and nucleic acids in FFPE specimens for investigating potential viral causes, various pre-treatment techniques can be employed to retrieve viral structures. Furthermore, FFPE material can be stored at room temperature (RT) for extended periods, allowing retrospective analyses of past disease outbreaks even several years after sampling (19, 21, 22). In this context, a previous study from 2023 identified the newly discovered rustrela virus (RusV) as the cause of encephalitis in captive lions from the 1980s (23). The presence of RusV RNA and antigen in the FFPE tissues of these lions was detected using the panRusV-2 a broadly reactive quantitative reverse transcription PCR (qRT-PCR) assay, ISH and IHC (23). In addition, a part of the RusV genome was determined by Sanger sequencing (23). Moreover, the maintained tissue morphology of FFPE samples facilitates the examination of viral structures in association with histopathological lesions and enables the investigation of the cell tropism (24–27). Although the process of formalin fixation and embedding in paraffin can have a negative impact on the integrity of these structures, e.g., the degradation of the nucleic acids, the development of new screening methods that are easily applicable to stored FFPE material may help to overcome the frequent unavailability of native or properly frozen organ samples (19, 20, 28–32). Immunohistochemical detection of double-stranded ribonucleic acids (dsRNA) in cases with suspected viral etiology of unknown origin represents a promising approach to enhance the identification of previously unknown viruses, due to its broad applicability to commonly used FFPE tissue samples. Antibodies against dsRNA are considered to represent a pan-viral detection method, as their use allows virus infections to be identified independently of the genomic constitution of the causative virus (33, 34). Because dsRNA molecules are synthesized independently of the virus genome, the detection of dsRNA is considered to be virus-independent and not limited to already known viral proteins or sequences, as in the case of antibodies, probes or primers. The generation of viral dsRNA is determined by the viral genome and distinct replication strategies employed by various virus groups. Single-stranded RNA (ss RNA) viruses use their genome as a template to synthesize complementary negative-sense and positive-sense RNA strands, respectively. During the replication of positive-sense single-stranded RNA viruses (+ss RNA viruses), the complementary negative-sense RNA strand functions as a template for generating new positive-sense RNA genomes. In contrast, negative-sense single-stranded RNA viruses (-ss RNA viruses) must initially convert their genome into complementary positive-strand RNA before protein synthesis can take place. In both cases, the reverse-oriented copies occur simultaneously with the genomic RNA, and accidental hybridization of both strands can result in the formation of dsRNA intermediates (15, 34–39). However, previous studies indicated that significantly smaller amounts of dsRNA intermediates are generated during -ss RNA virus infections compared to infections caused by +ss RNA viruses (34, 37). The genome of dsRNA viruses is the primary source of dsRNA in infections caused by these viruses. Although this group of viruses specifically attempts to conceal dsRNA from the immune system detection by replicating within viroplasms, these dsRNA molecules can still be found in the cytoplasm of the infected cells. DNA viruses generate dsRNA by overlapping transcription (15, 34–39). The present study investigated the applicability of antibodies directed against dsRNA as a sensing tool for virus infections in encephalitides of suspected viral origin, regardless of the specific virus or host species. Therefore, this study examined cases of encephalitis with known causes including Borna disease virus 1 (BoDV-1), canine distemper virus (CDV) and Rift Valley fever virus (RVFV) as representatives of -ss RNA viruses as well as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), tick-borne encephalitis virus (TBEV) and Theiler’s murine encephalomyelitis virus (TMEV) as examples of +ss RNA viruses.

Immunohistochemical investigation of brain tissue from animals displaying virus-induced encephalitis (n = 23) and non-infected control animals (n = 22) revealed immunopositive reactions for dsRNA in 43 of 45 animals investigated, including dsRNA detection in 21 virus-infected animals and all 22 non-infected animals. In addition, the expression of dsRNA was also detected in DH82 cells, independently of whether the cells were CDV-infected or non-infected. The immunohistochemical analysis of dsRNA yielded negative results in two CDV-infected dogs by the application of all three anti-dsRNA antibodies. This lack of detectable dsRNA is likely due to several factors. Variations in viral load and the timing of sampling may contribute to the negative results. Histopathological findings, such as demyelination and lymphocytic perivascular inflammation, suggest a subacute to chronic inflammatory response (51). As the disease progresses, the detectable antigen decreases, which could explain the absence of dsRNA signals (52). Second, CDV, like other -ss RNA viruses, produces only small amounts of dsRNA and the detection of dsRNA relies on ongoing viral replication (34). If the viral load is low or if replication-defective viruses are present, this could explain the negative results observed. Additionally, the post-mortem processing of tissue samples collected over the past 30 years may also influence the outcome. Some samples may have undergone prolonged fixation in improperly concentrated or unbuffered formalin, which is known to fragment nucleic acids and mask epitopes (20). Furthermore, RNA degradation could have occurred during the post-mortem interval before fixation and extended storage of FFPE tissues may have led to further degradation, all of which could hinder the detection of dsRNA (53, 54). For measles virus, another member of the Paramyxoviridae family, dsRNA formation has already been detected in experimentally infected cells (34). Interestingly, the intensity of the immunopositive signal was considerably weaker than that of +ss RNA viruses (34). It has been proposed that the C protein, which modulates innate immune responses, may contribute to reduced dsRNA production during measles virus infection (34). This was further supported by demonstrating a considerable decrease in the amount of dsRNA in cells infected with a wild type measles virus compared to cells infected with a mutant measles virus with an induced knockout of the C protein (55). Similarly, for CDV, the expression of the C protein is known to result in significantly fewer dsRNA-positive cells, which helps the virus evading the host’s innate immune system (56). Attenuated variants of CDV, often with a defective C protein, produce excessive amounts of defective RNA, which aggregates into dsRNA (56). Given this, a stronger dsRNA signal would have been expected in the CDV Ond infected DH82 cells. The morphological appearance of the immunopositive anti-dsRNA signal differed considerably, being present either as diffuse homogeneous cytoplasmic, granular cytoplasmic, intranuclear, or as an extracellular signal. In three CDV-infected dogs as well as in non-infected controls, cerebral cortical neurons and Purkinje cells showed an almost diffuse immunopositive signal for dsRNA. Surprisingly, the dsRNA signal in CDV-infected dogs and CDV-infected DH82 cells was homogeneously distributed within the cytoplasm, whereas in non-infected dogs and DH82 cells, the immunopositive signal appeared predominantly granular. The homogeneous cytoplasmic signal observed in CDV-infected animals is likely nonspecific. CDV replicates in the cytoplasm through RNA-dependent RNA polymerase (57). This process takes place in membrane-less organelles, enriched with high levels of N- and P-proteins (58). Given this mode of replication, a granular dsRNA signal would be anticipated, reflecting the formation of large cytoplasmic aggregates, rather than the homogeneous signal that was observed. The use of J2 and K1 in BoDV-1-infected horses resulted in a convincing granular reaction co-localized to viral antigen. A similar signal was detected in non-infected horses using K1, occurring in areas where virus antigen expression would typically be expected. Furthermore, the immunopositive dsRNA reaction was observed in both the cytoplasm and the nucleus, although BoDV-1 replicates within the nucleus (59). This finding is somewhat surprising, as –ssRNA viruses like BoDV-1 are known to produce relatively low amounts of dsRNA during viral replication (34, 37). Nevertheless, the usefulness of dsRNA antibodies in BoDV-1 appears to be limited, as similar positive signals were also observed in non-infected horses. In contrast, it was shown that dsRNA was successfully detected in cells infected with Nyamanini virus, a member of the order Mononegavirales and therefore related to BoDV-1 (34, 60, 61). The study by Son et al. (34) also used a panel of three anti-dsRNA antibodies, including J2 and 9D5, while Thomsen et al. (62) applied J2 and 9D5, and Richardson et al. (8) used J2 and K1. These studies differed in terms of protocols and, in some cases, the antibody manufacturers. One notable difference is that Son et al. (34) and Thomsen et al. (62) detected a strong 9D5 signal, unlike the present study. Additionally, the present study and the investigation by Thomsen et al. (62) detected dsRNA signals in control animals, suggesting possible non-specific binding. While the current study examined both +ss RNA and –ss RNA viruses, only Son et al. (34) investigated both types of viruses. Similar to our findings a weaker dsRNA signal was detected for –ss RNA viruses compared to +ss RNA viruses. The present study and the investigations by Thomsen et al. (62) and Richardson et al. (8) included tissues from naturally infected animals and cell lines from various animal species. In contrast, Son et al. (34) used cell lines, fixed with 10% formalin or 3.7% paraformaldehyde, and viruses that likely underwent multiple in vitro passages. It is important to note that viral attenuation in cell culture may influence the expression of virulence factors and dsRNA production, as demonstrated in studies of measles and CDV (55, 56). In all RVFV-infected mice as well as in all non-infected control mice, a predominantly homogeneous signal, almost diffusely distributed throughout the brain, was detected. It remains questionable whether the weak signal represents dsRNA or a non-specific background staining. Nonetheless, the staining pattern does not allow any differentiation between RVFV-infected and non-infected mice. As already described for RVFV, the application of J2 and K1 resulted in cytoplasmic homogeneous signals in the brain of SARS-CoV-2-infected K18 mice. However, in contrast to RVFV, the application of K1 resulted in a more intense, inconsistently granular signal in non-infected control K18 mice. On the other hand, no difference in J2 signaling was detected in infected and non-infected K18 mice. Coronaviruses have developed strategies to circumvent the innate immune system, which are partially based on the fact that dsRNA recognition by host factors initiates an antiviral response (63–67). Nevertheless, detectable amounts of dsRNA were demonstrated in experimentally SARS-CoV-2-infected cells (67). TBEV and TMEV represent the only two viruses with a granular signal co-localized to viral antigen in all infected animals using all three anti-dsRNA antibodies. Based on the morphology of the signal and the positivity of all three antibodies, it was assumed that this represents a true positive signal for the expression of dsRNA. However, even in these cases it must be noted that in non-infected control mice, staining with J2 resulted in a homogeneous cytoplasmic signal, while K1 displayed a partially granular reaction. The few to no signals observed with the 9D5 antibody, compared to the positive immunoreactivity seen with J2 and K1 in non-infected control mice could be explained by the fact that the low number of positive cells observed by J2 and K1 antibodies may reflect an unspecific signal. In contrast, the absence of a similar number of 9D5-positive cells could indicate the higher specificity of this antibody. Granular immunopositive anti-dsRNA reactions were exclusively observed in BoDV-1-infected horses and TBEV- and TMEV-infected mice, suggesting that these antibodies might indeed bind to dsRNA in the context of viral infections. However, the presence of similar signals in non-infected control animals raises questions about the specificity of these antibodies in the broader context of viral encephalitides. This finding is consistent with previous studies that have highlighted the limitations of dsRNA antibodies in distinguishing between infected and non-infected states, particularly in diseases caused by -ss RNA viruses (34, 37). In these previous studies, dsRNA was demonstrated mainly in mammalian FFPE cell pellets, e.g., infected with Coxsackie virus (+ss RNA virus), measles virus and influenza A virus (−ss RNA viruses), and minute virus of mice, lymphocytic choriomeningitis virus, vaccinia virus and adenovirus (DNA viruses) (8, 34, 37). Similarly, dsRNA could be successfully visualized in neonatal mouse FFPE organ samples and human neonatal FFPE heart samples infected with Coxsackie virus (8). It therefore remains undetermined whether species-specific differences may also represent factors influencing dsRNA expression. Similarly, in own studies analyzing Usutu virus-infected birds and rustrela virus-infected lions, immunopositive signals were observed in different organs of infected animals and virus-negative control animals (22, 42).The hypothesis that the virus and animal species have a potential impact on dsRNA immunoreactivity is supported by the considerable differences observed in the results from BoDV-1-infected horses and CDV-infected dogs, despite both viruses belonging to the order Mononegavirales (61, 68). Similar observations have been made in SARS-CoV-2 infections: While previous investigations on the demonstration of dsRNA in experimentally infected cells were successful, the present results indicate that dsRNA expression was lacking in SARS-CoV-2 infected brains of K18 mice (63). However, it is important to consider whether the species-specific differences might also be partly due to the use of different detection kits, including ABC kit and Mouse-on-Mouse Polymer IHC kit for immunohistochemical staining of dsRNA in the different species. Furthermore, it remains unclear whether the anti-dsRNA antibodies sensed exclusively viral dsRNA intermediates. It is assumed that the anti-dsRNA antibodies recognize viral dsRNA intermediates regardless of the nucleotide sequence, but with a nucleotide length of at least 50 base pairs (bp) (8). This would also reduce the possibility of the detection of endogenous dsRNA, as these molecules are usually shorter than 50 bp (8, 69). In order to be used as a reliable screening tool for potential virus infections, these antibodies should exhibit a higher specificity for viral dsRNA. For example, virus-specific epitopes like the 5′-triphosphate end could be incorporated into the recognition process of the antibody (70). Additionally, the sensitivity of the antibody should be increased so that even small amounts of dsRNA, such as those produced during the replication of –ssRNA viruses, can be reliably detected in the future.

The investigation of the expression of dsRNA in animals and cells infected with various –ss RNA viruses and +ss RNA viruses has produced some notable findings. Specifically, the study observed the co-localization of anti-dsRNA signals with viral antigens in BoDV-1-infected horses. The latter is a -ss RNA virus with only low levels of dsRNA expression during viral replication. Additionally, granular immunopositive signals in TBEV- and TMEV-infected animals, co-localizing with viral antigens using all three anti-dsRNA antibodies, suggest the potential of these antibodies as markers for viral infection. However, the usability of anti-dsRNA antibodies as an early detection marker of viral infections was considered very limited due to immunopositive signals in non-infected control animals and cells. While the reactions in controls are disturbing, they also present a point for discussion and further refinement. These findings emphasize the need to enhance the specificity of anti-dsRNA antibodies, which could help to distinguish better infected from non-infected tissues in future studies.