Conditional CD28-knockout mice with cre-expression (CD28^-/flox Cre^+/-; Cre) and without cre-expression (CD28^-/flox Cre^-/-; B6) (27) as well mice with a conventional CD28 knockout (CD28^-/-; CD28KO) (20) on a C57BL/6 background were bred at the central animal facility of the Institute for Multiple Sclerosis Research (IMSF), University of Göttingen. The study took place in the Department for Pathology of the University of Veterinary Medicine Hannover, where the mice were housed in individual ventilated cages (IVC) with ad libitum access to water and food (ssniff Spezialdiäten GmbH, DE- 59494 Soest, cat. V1534-000). All experiments were performed in accordance with German law and approved by the Lower Saxony State Office for Consumer Protection and Food Safety as the responsible authority (Niedersächsisches Landesamt für Verbraucherschutz- und Lebensmittelsicherheit (LAVES), Oldenburg, Germany; permission number: -17/2418).

The knockout of CD28 was induced in Cre-expressing FEC2/FLCK CD28KO mice (CD28^-/flox Cre^+/-) by three oral gavages of tamoxifen (3 mg tamoxifen/100µl rapeseed oil) at 5, 7 and 9 dpi, using a flexible 20 G feeding tube (Instech, cat. FTP-20-38-50) as described before (4).

All mice of the study were infected intracerebrally at the age of five weeks with 20µl of a TMEV-BeAn-1-TiHo cell culture supernatant of 2.7x10⁷ plaque-forming units (PFU)/ml (infective dose of 5.4x10⁵ PFU) under injection-anesthesia as described before (33).

Mice were visually inspected daily; clinical scoring (appearance [0-3], activity [0-3], gait [0-4]) and RotaRod^®-testing for motor skills were performed weekly as described (4). The intervals of clinical scoring were increased according to the animals’ clinical signs as described before (4).

Animals were sacrificed at 14 (n=39), 42 (n=12) or 147 (n= 11, data not shown) days post infection (dpi). Mice with CD28 knockout that reached the pre-defined termination criteria (weight loss ≥ 20%, high clinical scores, paralysis of limbs) were euthanized for humane reasons at 14 (Cre-Tam), 33 (n=1) and 35 (n=1) dpi (CD28KO-R). After euthanasia, mice were perfused with phosphate buffered saline (PBS) at a flow rate of 3.75 ml/min through the left ventricle. Brain, spinal cord, thymus, spleen, and intestine were removed and fixed in 4% formaldehyde solution followed by paraffin embedding.

For histology and immunohistochemistry, the formalin-fixed, paraffin-embedded (FFPE) tissue was cut to 2 µm thick sections and mounted on glass slides (Thermo Scientific, Superfrost^® Plus, cat. J1800AMNZ). Hematoxylin-eosin (HE) stained sagittal sections of the brain were divided into 10 regions (as seen in Supplementary Figure S1 ) and scores (0-3) for hyper-cellularity, perivascular infiltration, meningeal infiltration, vacuolization, and neuronal/axonal damage (4, 35) were added up. The scores (0-3) for hyper-cellularity in the white matter, the grey matter, perivascular infiltration, vacuolization and meningeal infiltration from three spinal cord segments (cervical, thoracic, and lumbar) each divided into 6 regions (left/right, dorsal, middle and ventral) were added up for every animal.

Immunohistochemistry was performed applying the avidin-biotin complex (ABC)-method with 3,3’-diaminobenzidine (DAB)-labelling as described before (36). The pretreatment and concentrations of primary antibodies are presented in Table 3 below:

For morphometry and figures, stained sections were scanned with extended focal image (EFI) on an Olympus Slide Scanner VS200. Tissue detection was done by thresholding (CellSens, ImageJ). Positive stained areas and cell counts were obtained according to the requirements by manual counting on glass slides (TMEV, CD45R, Caspase 3, beta amyloid precursor protein [beta-APP], S100A10) or applying a neural network (CD3 [Olympus cellSens imaging software VS200]) trained on respective tissue to detect the cell numbers per area. Brain and spinal cord sections with glial markers (ionized calcium-binding adapter molecule 1 [Iba-1] and glial fibrillary acidic protein [GFAP]) were digitalized (VS200) followed by thresholding and automatic detection of the percentage of positive areas (ImageJ, QuPath) to account for changes in cell volumes during the inflammatory response. The cell numbers and positive areas were evaluated on one sagittal brain section (including all 10 regions as defined above) or three spinal cord sections (cervical, thoracic, lumbar) combined into one evaluated area, respectively.

Photographs for the figures were obtained from scanned slides (VS200) and edited with GIMP 2.10.32 (GNU Image Manipulation Program). Graphs were created with Graphpad Prism (Graphpad Software Inc.) and show scatter plots (one dot per animal) with median (black bar) and 95% confidence intervals (black error bars). Statistical analyses were conducted using SPSS for Windows TM v. 27 (IBM^® SPSS^® Statistics, SPSS Inc., Chicago, IL, United States). Data were analyzed for normal distribution using the Shapiro–Wilk test. Significant differences between the groups were investigated via Kruskal–Wallis tests followed by Dunn–Bonferroni post hoc testing with correction of significance levels for multiple testing. Corrected p-values of ≤ 0.05 were recognized as statistically significant and indicated by an asterisk (*) within the graphs. Figures of multiple panels were generated with GIMP and Inkscape 1.2 (Inkscape Community).

Mice with tamoxifen-induced, conditional knockout of CD28 (Cre-Tam) showed a marked loss of body weight between 7 and 14 dpi ( Figure 1A ). Additionally, only Cre-Tam mice showed significantly increased clinical scores ( Figure 1B ) but without measurable effects on their RotaRod performance ( Figure 1C ) at 14 dpi.

Histopathological examination of the brain and spinal cord revealed that the sum score in HE-stained brain sections is significantly lower in Cre-Tam mice compared to the B6-Tam controls without CD28 deletion ( Figures 2A, B ). Inflammation within the spinal cord was most apparent within the lumbar region of all mouse strains ( Figure 2A ). Concerning the virus clearance, there were also differences between the groups. Whereas B6 mice without tamoxifen administration (B6-nT) completely cleared the virus by 14 dpi as expected ( Figure 2C , upper left), mice with tamoxifen application (B6-Tam) still show virus antigen within the brain at 14 dpi ( Figure 2C , upper right), as reported before (28). Virus persistence was also seen in both mouse strains in which CD28 had been deleted ( Figures 2C, D ). Remaining TMEV-antigen was detected preferably within the hippocampus of mice without CD28 knockout but tamoxifen application (B6-Tam), while in both mouse strains with CD28-knockout (CD28KO, Cre-Tam), the antigen was more likely to be spread over more brain regions, except for the olfactory bulb, forebrain, and cerebellum ( Supplementary Figure S1 ). Highest numbers of TMEV-antigen positive cells were detected within the cerebral cortex, hippocampus and thalamus of both CD28-knockout strains ( Supplementary Figure S1 ). Furthermore, only in mice with CD28-knockout (Cre-Tam, CD28KO), an unexpected virus spread into the spinal cord was seen ( Figure 2D ).

Immunohistochemical assessment of the local immune response by counting infiltrating immune cells into the CNS revealed significant differences in the T- and B cell response as well as microglial activation. It was observed that mice with conditional, tamoxifen-induced CD28-knockout mice (Cre-Tam, Figure 3A , lower right) showed increased T cell infiltration into the brain ( Figure 3B ). The T cell infiltration in B6-nT mice without tamoxifen application ( Figure 3A , upper left) and animals with conventional CD28-knockout ( Figure 3A , lower left) were at similarly low levels, despite the presence of TMEV-antigen in the CNS in the CD28KO mice. Within the spinal cord, there were no significant differences in T cell infiltration between the groups at 14 dpi ( Figure 3B ). The number of B cells in B6-nT mice without tamoxifen application ( Figure 3C , upper left) and mice with conventional CD28-knockout ( Figure 3C , lower left) were, as seen for T cell infiltration, at comparable levels ( Figure 3D ). Contrary to T cells, the B cell infiltration showed the opposite trend with non-significantly reduced B cell numbers in tamoxifen-treated B6-Tam mice ( Figure 3C , upper right) and significantly reduced B cell numbers in Cre-Tam mice with conditional CD28 knockout ( Figure 3C , lower right) in both the brain and the spinal cord ( Figure 3D ). Prominent reactive microgliosis was seen in the brain of B6-nT and B6-Tam mice without CD28-knockout ( Figure 3E , upper left and upper right). Within the brain, microglial activation was significantly reduced in Cre-Tam and CD28KO mice with knockout of CD28 ( Figures 3E, F ). Within the spinal cord, there was no significant difference between mice with or without CD28-knockout ( Figure 3F ).

The number of apoptotic cells (caspase 3 positive cells) within the brain and spinal cord was not elevated in CD28-knockout mice compared to the controls at 14 dpi ( Supplementary Figure S2 ). In addition, there was no increase in axonal damage (beta APP positive axons) within the spinal cord of CD28-knockout mice in the early phase of TMEV-infection at 14 dpi ( Supplementary Figure S2 ).

The summarized results of the clinical and histopathological findings in intracerebrally TMEV-infected mice reveal that mice lacking CD28-expression (Cre-Tam, CD28KO) show virus persistence and spreading of virus within the brain as well as into the spinal cord at 14 dpi. This is correlated with significantly less microglial activation, despite presence of virus antigen within the CNS. T- and B cell infiltration seem to be influenced by tamoxifen administration and CD28-knockout in a mutually enhancing manner. The influence of tamoxifen on the number of infiltrating lymphocytes can be seen in the direct comparison with B6-nT controls, but only reaches statistical significance in Cre-Tam mice with CD28-knockout, that show significantly increased T cell infiltration and decreased B-cell infiltration.

Because TMEV has not been cleared yet at 14 dpi, it was interesting to follow up the fate of the mice also in the chronic stage of TMEV infection. In susceptible SJL mice, the chronic phase is characterized by progressive clinical signs and loss of myelin caused by different immunopathologies of this model (1). Since the CD28KO mice were unable to clear the virus within 2 weeks after infection, a longer period was studied to see whether CD28-deficiency results in a similar fate on the otherwise resistant C57BL/6 background. Due to the detrimental effects of the tamoxifen-induced conditional CD28-knockout in the early phase of TMEV-infection (weight loss, clinical signs), Cre-Tam mice were not included in this part of the study for animal welfare reasons. Mice with a conventional CD28-knockout (CD28KO) either showed no clinical signs, similar to the control animals (B6-nT) or developed motoric deficiencies at 21 (n=1), 32 (n=1) and 33 (n=2) days post TMEV-infection. Mice that developed these clinical signs were thus evaluated separately as ‘responders’ (CD28KO-R, n= 4 out of 12). At the end of experiment, the four responders with conventional CD28-knockout had a lower body weight ( Figure 4A ) and showed increased clinical scores ( Figure 4B ). The high clinical scores in CD28KO-R mice were mainly determined by hind limb paresis (score 4) of two mice at 33 (n=1) and 35 dpi (n=1). Mice with hind limb paresis were euthanized immediately for humane reasons and thus not put on the RotaRod. Accordingly, the presented data ( Figure 4C ) show the last RotaRod performance on 28 dpi of mice prior to the hind limb paresis and 42 dpi of the mice with mild gait abnormalities. It can be seen, that anyways, the motor skills of the CD28KO-R mice are non-significantly reduced compared to the non-responder CD28KO mice ( Figure 4C ).

Histopathological examination at 42 dpi revealed that B6-nT control animals without knockout ( Figure 5A , upper left) as well as conventional CD28KO mice without gait abnormalities ( Figure 5A , upper right) showed no, or only minor inflammation characterized by single parenchymal and perivascular infiltrations of immune cells within the brain and spinal cord ( Figure 5B ). In contrast, inflammation was still visible within the spinal cord of the CD28KO-R mice and most prominent in the lumbar segment ( Figures 5A, B ). The B6-nT controls ( Figure 5C , upper left) and clinically unaffected CD28KO mice ( Figure 5C , upper right),except for one mouse with very low titer, were able to eliminate TMEV from the brain and spinal cord ( Figure 5D ) at 42 dpi. Conversely, TMEV-antigen was still detectable within the brain ( Figure 5C , lower left) of 3 and spinal cords of 2 out of 4 CD28KO-R mice, respectively ( Figure 5D ).

The low sum score of the CNS of B6-nT control animals and non-responder CD28KO mice, were affirmed by low numbers of infiltrating T cells ( Figures 6A, B ) and B cells ( Figures 6C, D ). This was in contrast to the ongoing T cell ( Figures 6A, B ) and B cell ( Figures 6C, D ) infiltration in CD28KO-R mice.

Compared to the B6-nT controls ( Figures 7A–D , upper left) the astroglial and microglial response in the brain was mildly but not significantly elevated in CD28KO mice without clinic ( Figures 7A–D , upper right). CD28KO-R mice showed significantly increased density of astrocytes (astrogliosis) within the brain and spinal cord ( Figures 7A, B ) as well as a high density of hypertrophic/hyper-ramified microglia ( Figures 7C, D ). These data indicate an ineffective and delayed immune response against TMEV-infection in clinically affected CD28KO-R mice.

Concomitant to the chronic TMEV-infection and ongoing inflammation within the CNS of CD28KO-R mice, the responders also show increased numbers of apoptotic cells ( Figures 8A, B ) and damaged axons within the spinal cord ( Figures 8C, D ). The number of neurotrophic S100A10-positive astrocytes in the spinal cord was not affected ( Figures 8E, F ). Prominent loss of myelin was not seen in luxol-fast-blue (LFB)-stained sections of the spinal cord in any of the mice (data not shown). When following the remaining B6-nT control animals (n=7) and conventional CD28KO animals (n=4) up to 147 dpi, no development of clinical signs or histopathological abnormalities were observed (data not shown).

In summary, in the chronic phase after TMEV-infection, the deletion of CD28 results in variable outcome. Control mice (as expected) and two third of conventional CD28KO mice are able to resolve the virus infection without development of clinical symptoms and histopathological abnormalities. In contrast, one third of CD28KO mice are susceptible for a chronic TMEV-infection with an inefficient inflammatory response and concomitant loss of lower motor neurons within inflamed spinal cord regions as well as increased apoptosis and axonal damage.

The infection of C57BL/6 mice with low virulent TMEV is known to cause acute and chronic seizures and is thus used as a model for human temporal lobe epilepsy (37). However, the occurrence and frequency of seizures depends on the virus strain (38). About 50%-75% of mice intracerebrally infected with the Daniel’s strain of TMEV (TMEV-DA) develop seizures, depending on the applied detection method and initial virus dose (39–41). For the TMEV-BeAn-strain the percentage differs between 0-40% (39, 41). It was seen that seizure development does not correlate with neuropathological changes (39). In the present experiment the variant BeAn-1-TiHo was used, which is not suspected to cause seizures in C57BL/6 mice (41). Concomitantly, no apparent seizures were observed after TMEV-BeAn-infection in the present study. However, it cannot be excluded that seizures might have occurred apart from the general examinations, since no specific scoring (Racine score) or continuous monitoring via video-electroencephalogram (VEEG) were applied.

One of the hallmarks after CD28 loss – both conventional and conditional - is the unexpected virus spread within the brain as well as into the spinal cord of mice with a C57BL/6 background ( Figures 2C, D ). In general, C57BL/6 mice are considered to be resistant to a chronic TMEV-infection and are able to eliminate the virus within 2 weeks after intracranial infection (9). Although it was previously already shown that the tamoxifen application is associated with delayed virus elimination (34), the tamoxifen treated mice without CD28 knockout (B6-Tam) did not show virus spread into the spinal cord, or elevated clinical scores. In both CD28-knockout strains, microglial activation and change of morphology as reaction to the acute TMEV-infection in the brain were reduced.

Brain-resident microglia and infiltrating monocytes as part of the innate immune response play a pivotal role in the acute, as well as in the chronic phase of a TMEV-infection (42). On the one hand, these cell types contribute to local damage, scarring and development of seizures in C57BL/6 mice (42). Further, microglia are susceptible for TMEV-infection and the major host cells during the chronic TMEV-BeAn infection as seen in SJL mice (43). On the other hand, a clearing of TMEV from the CNS is impossible for C57BL/6 mice after depletion of microglia (44). Similarly, the presented data indicate a link between reduced microglial activation ( Figures 3E, F ) and delayed virus elimination and virus spread within the CNS in mice lacking CD28 ( Figures 2C, D ). The molecular background of this correlation will be subject of further studies.

For infiltration of cells of the adaptive immune response (T- and B cells), the situation is different. As described in other studies on the immune system (45), conventional TMEV-infected CD28KO mice had similar levels of infiltrating T- and B cells into the brain and spinal cord as their B6-nT controls. Conversely, Cre-Tam mice with an induced knockout showed increased T-cell infiltration and almost a total lack of B cell infiltration into the TMEV-infected brain. Furthermore, it was shown that the tamoxifen application alone (B6-Tam) has a similar effect, but to a lesser extent. Thus, it seems that the tamoxifen application and the delayed anti-viral immune response is the main reason for the changed T- and B-cell infiltration, and that the actual deletion of CD28 is somehow reinforcing these effects, whereas the pure lack of CD28 from the beginning has no influence of T- and B cell infiltration. Based on these data, it also seems that T- and B cell infiltration is uncoupled from microglial activation after TMEV-infection and also from the development of neurological symptoms.

CD28 is important during the antigen presentation from myeloid cells, for instance dendritic cells (DC), but also decisive in the communication between T- and B cells (46). Antigens binding to the B cell-receptor (BCR) are processed and presented on MHC-II with concomitant expression of CD80/CD86 (B7-1/B7-2) as ligands for CD28 on the T cell (46, 47). Activated T cells secrete interleukins to further stimulate B cells, however, the interaction of CD28 with the B7 molecules itself poses a direct co-stimulatory signal for B cells (48). The presented data show, that the activation and expansion of B cells is almost completely aborted in mice with induced CD28 knockout, which is likely due to a loss of communication between the B- and T cells as well as loss of CD28/B7 mediated co-stimulation of B cells (48). The similarity in T- and B cell reaction of mice with conventional CD28 knockout in this study indicates an alternative route of activation in these mice. Although CD28 is a main molecule for the activation of naïve T cells, it is by far not the only one (49). Upregulation of other co-stimulatory molecules may have compensated the lack of CD28-signaling in the thymus of CD28KO mice with conventional CD28-knockout (45, 49, 50).

Tamoxifen as a selective estrogen receptor modulator (SERM) is known to influence the immune response under therapeutic as well as in experimental conditions (4, 34, 51–53). Although tamoxifen is designed to act on human estrogen receptors (hER) it also influences murine immune cells and has estrogen-like (agonistic) as well as antagonistic properties, depending on the expression of ER-subtypes (4, 34, 51–57). Since tamoxifen is a SERM and not a specific ER-inhibitor, and it also modulates cell responses through estrogen-independent mechanisms, reported effects on the immune response on a cellular level are divergent and highly depend on the applied method (51, 58). In the CNS, tamoxifen has mostly neuroprotective and anti-inflammatory properties, increases proliferation and differentiation of oligodendrocyte progenitor cells (OPC), decreases production of pro-inflammatory cytokines like TNF-alpha, IL-12 and IL-1 by micro- and astroglia, as well as reduces astrogliosis under non-infectious conditions (59, 60). Under infectious conditions, these effects might affect the initial immune response to be less efficient, resulting in a delayed virus elimination as seen in tamoxifen treated animals in the present study. The ongoing presence of TMEV-antigen is however suspected to stimulate the immune response in these animals, possibly covering an initially anti-inflammatory effect until 14 dpi. Further, tamoxifen is known to negatively influence the activation and maturation of B cells (54). The B cell response after a virus infection is dependent on factors of the virus itself, such as pathogenicity and cell tropism, as well as the accessibility and presentation of virus peptides and a pro- or anti-inflammatory microenvironment (61, 62). Although ER-β-mediated effects of tamoxifen are suspected to have negative effects on B cells (63), the molecular mechanism behind the reduced B cell response remains unclear for the present study and demands further investigation.

Despite differences in T- and B cell infiltration into the CNS of TMEV-infected mice with conventional and induced CD28-knockout, the virus elimination in these mice was equally impaired. The weak microglial activation in the early phase of TMEV-infection in both types of CD28-knockout mice indicates a disruption of an activation pathway, which could not be quickly overcome by other routes of co-stimulation and is somewhat independent of the B- and T cell infiltration. A comparable stimulatory function of B7-signaling on microglia (64), similar to B cells (48), might be a factor that should be addressed in future studies.

Since the cre-expression of these mice is not controlled by a cell-specific promotor, but rather expressed in all cells after tamoxifen application, a CD28 knockout in all cells is expected in the Cre-Tam mice. The variable significant difference in T- and B cell infiltrations of Cre-Tam mice compared to the TMEV-infected mice without CD28-knockout, as well as the decreased microglial reaction during the acute phase of TMEV-infection supports this assumption of a CD28-knockoutin the infiltrating immune cells.

The role of a CD28-knockout for CD4⁺ and CD8⁺ T cells separately represent an interesting topic for future studies in this model. The outcome and special contributions of MHC-I molecule H-2D^b (65, 66), the early anti-viral CD8⁺T cell response to a TMEV-infection, the reaction of CD4⁺ T helper and effector cells as well as regulatory T cells (Tregs) (67, 68) could not be differentiated in the present study, utilizing a generalized CD28-knockout. A cell specific promotor for the cre-expression might aid to shed light on the contribution of different immune cell subsets to the observed neuropathologies in the present study (69). It would further be of interest to implement this model in mice on a susceptible SJL-background to investigate the effect of a CD28-knockout on demyelination (TMEV-IDD) (1).

Due to the acute clinical and pathomorphological impairments of conditional CD28-knockout at 14 dpi, this mouse strain could not be included in long-term studies. Therefore, the chronic TMEV-infection was only investigated in mice with conventional CD28-knockout (CD28KO).

A surprising finding here was the heterogeneity in the long-term clinical outcome of the intracerebral TMEV-infection in CD28KO mice. Two thirds of CD28KO mice were, just like the B6-nT control animals, clinically unaffected until 42 dpi or even 147 dpi (data not shown). 33% of the animals showed progressive impairments of motor function (CD28KO-R, ‘responders’), starting around 21-33 dpi, leading up to hind-limb paresis in 2 of 4 affected animals.

Analysis of body weight, clinical score and RotaRod-performance test of the CD28KO-R did not reveal changes in the acute phase of TMEV-infection that would signal that these animals would develop clinical signs during the chronic phase. From 14 dpi until 42 dpi, the median RotaRod-performance of CD28KO-R mice is reduced compared to the other groups, followed by first clinical signs in some mice at 21 dpi and a lower median body weight from 28 until 42 dpi ( Supplementary Figure S3 ).

Histologically, the responder-mice (CD28KO-R) were affected by ongoing virus spread and CNS-inflammation, indicating that the virus could not be cleared due to an inefficient immune response. However, in contrast to susceptible SJL mice (4), the myelitis of CD28KO-R mice was not associated with prominent demyelination, but inflammation and possible damage of lower motor neurons in the ventral horns. Thus, autoimmunity did not develop in the absence of CD28 in C57BL/6 mice, even when the virus could not be cleared similar to SJL mice. The marked difference to the early phase for the CD28-knockout mice was the increased microglial and astroglial activation at 33-42 dpi, combined with high numbers of infiltrating lymphocytes. However, despite this inflammatory response and microglial activation, the CD28KO-R mice were not able to eliminate the virus. It would be interesting to investigate the trigger(s) that differentiate mice able versus unable to clear the virus on the long run.

CD28 signaling is not only important for T cell activation and priming but also plays an important role for effector T cells (70). In CD28^flox/flox Ox40^cre/+ mice where CD28 expression is lost after T cell priming following an influenza A infection, it was shown that T follicular helper cells (Tfh) need CD28 for survival, and T helper (Th1) require CD28-signaling for expansion after activation (70). These findings corroborate studies that showed a decreased germinal center formation and isotype switching in CD28-deficient C57BL/6 mice (20, 71). Conversely, it was shown that stimulation of the CD28-receptor on human and murine Th1 cells induces IFN-γ secretion (72). In microglia and astrocytes, IFN-γ and other Th1 effectors regulate gene expression associated with neuroinflammation (73), indicating that a loss of CD28-sinaling might have an anti-inflammatory effect. Further, the interaction of CD28 with the B7-ligands not only stimulates cytokine production in T cells but also triggers auto- and paracrine IL-6 and IFN-γ-secretion by dendritic cells (DC) (74). This bidirectional signaling of CD28 and B7 molecules has yet to be shown for microglia, as a possible factor for initially decreased microgliosis in TMEV-infected, CD28-knockout mice. However, in the present study it was seen, that a chronic TMEV-infection of CD28KO mice is associated with ongoing neuroinflammation, characterized by T- and B cell infiltration as well as astro- and microgliosis. In the early phase of TMEV-infection, the T- and B cell response of CD28KO mice does not differ significantly from the B6-nT controls. This indicates a CD28-independent activation pathway in lymphocytes of conventional CD28KO mice, and might indicate a disturbed interaction of lymphocytes and microglia in the early phase of TMEV-infection, as reported for DCs (74). The glial reaction in the late phase of TMEV-infection is likely reflective of the pro-inflammatory microenvironment, sustained by activated T and B cells in response to virus persistence.

Besides CD28, T cells own a large variety of co-stimulatory and co-inhibitory receptors whose expression depends on the cell’s environment and activation status (49). Hence, despite their narrow genetic diversity, also inbred mice own individual expression panels of TCRs and co-receptors (20). In absence of CD28, individual expression of co-receptors could explain the heterogeneity in virus persistence and clinic seen in mice with a lack of CD28 (75).