The increased concentration of macrophage migration inhibitory factor in serum and cerebrospinal fluid of patients with tick-borne encephalitis

Data from animal models show that the events occurring during the peripheral virus spread and the penetration through BBB into the CNS may be crucial for the outcome of the neurotropic flavivirus infection [7, 9, 26, 33]. Especially, the BBB disruption caused by a systemic inflammatory response has been hypothesized to determine the WNV entry into the CNS [26, 33], which we have attempted to verify in human TBE in a clinical setting. Patients enrolled in our study had an increased intrathecal total protein concentration and AQ persisting long into the convalescent period, as was described in TBE previously [41]. The prominent increase of AQ points to increased BBB permeability as an important pathogenetic mechanism, but its dynamics is suggestive rather of a protracted BBB impairment during CNS inflammation and healing than to its early disruption preceding neuroinvasion. Similarly, in the mouse TBE model described by Růžek et al., the BBB dysfunction does not occur during systemic infection, but only after the establishment of encephalitis, simultaneously with the up-regulated intrathecal TNFα and IL-6 synthesis [37]. This course of events does not contradict the pathogenetic role of BBB permeability, but postpones it to the later, neurologic phase of the disease. It also implies an alternative mechanism of CNS invasion not dependent directly on BBB disruption by systemic inflammation.

To shed more light on the consequences of BBB dysfunction, we have studied if AQ was related to the clinical presentation and severity of TBE. In WNV infection, the protein concentration in CSF is higher in patients with encephalitis than in those with meningitis [42]. In our previous study, assessing retrospectively the recorded clinical and laboratory data from 687 consecutive TBE patients on admission, we have detected a modestly higher total protein concentration in patients with ME/MEM compared to uncomplicated meningitis; no CSF albumin concentration and AQ were available in most of these patients [1]. This has not been reproduced in a much smaller group included in the current study. However, the admission AQ was higher in patients with paresis than in the rest of the group, consistent with the role of the early BBB impairment in the neuropathology and hinting for a possibility of a specific pathogenesis of particular neurologic complications. We have recently confirmed this observation in another population of 250 TBE patients analyzed retrospectively, in whom CSF albumin on admission was significantly higher in MEM and in patients with spinal paresis (manuscript in preparation).

We have assessed expression of three pro-inflammatory cytokines suspected to mediate increased BBB permeability [26, 33, 35, 36], including MIF which is up-regulated by WNV but according to our knowledge has never been studied in TBE [26]. The main focus was on serum cytokine concentrations early in the course of the disease, which not only represent the early systemic response to TBEV, but might directly contribute to BBB disruption and CNS infection [26, 33]. In a clinical setting, we were not able to study cytokine expression at the most relevant time-point before or during TBEV entry into CNS, as at this stage, patients experience non-specific and usually benign symptoms of peripheral infection and rarely seek medical attention. Instead, we have studied blood samples obtained directly after the admission to hospital, during the early neurologic phase, as soon after the occurrence of neuroinvasion as practically possible. Most of the patients at that stage are febrile and present with a moderate leukocytosis and increased C-reactive protein concentrations [1], which was also true in our study group, suggesting that the inflammatory process in the periphery had not fully resolved yet. In a subgroup of patients, we have additionally studied CSF and convalescent period samples to assess the cytokine concentration gradient and dynamics. Although MIF, TNFα, and IL-1β were expected to influence the BBB function as members of a common signaling route, we found each of them following a different pattern of expression, suggesting that they are up-regulated individually and may play different pathogenetic roles. IL-1β expression was mainly intrathecal, pointing to its local synthesis within CNS, and although there was a tendency for its lower serum and higher CSF concentrations in more severely ill patients, it could not be definitively confirmed with our data. In WNV-infected mice, IL-1β is expressed independently from TNFα and MIF [26]. Interestingly, in a model studied by Palus et al., higher expression of IL-1β mRNA in the brain tissue was characteristic for the resistant mouse strain suggesting its protective role as opposed to several other pro-inflammatory cytokines, but extending this observation to a human disease requires further study [9].

The expression of TNFα in neurotropic flavivirus infections has been previously described, but its exact role, source, and dynamics is not well defined. The observation of Atrasheuskaya et al. that high serum TNFα concentrations in the first week of hospitalization correlate with a clinically severe TBE [43] and its up-regulation dependent on MIF detected by Arjona et al. in WNV-infected mice [26] prompted us to study its role alongside MIF in initiating the BBB disruption. Interestingly, TNFα may be also involved at other tissue locations and stages of the disease. In mice studied by Hayasaka et al., high TNFα expression in periphery was observed late in the course of TBE, representing the late phase of the excessive systemic inflammation, not directly linked to CNS pathology [7]. On the other hand TBEV induces TNFα synthesis in dendritic cell cultures after 24-h incubation, so its activity may begin already locally in the skin at the tick bite site with the involvement of the resident dendritic cells [44]. Finally, the role of TNFα in the intrathecal response has been suggested by its expression in brains of TBE-infected mice [9, 37]. Intrathecal TNFα expression correlated with fatal outcome in a group of 118 patients with Japanese encephalitis studied by Winter et al. as well [45]. Surprisingly in the light of the above findings, the median TNFα levels in serum and CSF were not significantly increased in our study group, unlike in the much smaller group of patients with clinically milder non-TBE meningitis, suggesting its relatively small role in TBE. This lack of an evident TNFα expression would preclude the presence of a singling pathway analogous to what was described in WNV-infected mice and make it unclear how MIF and TLR3 function could contribute to BBB disruption in the absence of TNFα [26, 33]. However, as discussed below, genotyping data show that TNFα expression in TBE varied in association with the genetic background and that it was in fact up-regulated and could be biologically active in a large subgroup of patients.

The either protective or immunopathogenic role of MIF has been described in several bacterial and viral infections. In the community-acquired pneumonia, a high-expression allele MIF-173C associates with higher IL-6 and IL-10 concentrations and a better prognosis [20]. In the CNS infections however, the trend is the opposite, as a high concentration of MIF in serum and CSF correlates with a more pronounced CNS infection and severe clinical presentation in patients with purulent meningitis [21]. The high concentration of MIF in cerebrospinal fluid (CSF) and presence of high-expression variants 7-CATT and MIF-173C correlate with an unfavorable outcome in Streptococcus pneumoniae meningitis, and MIF neutralization is beneficial in the animal model of the disease [22]. The frequency of 7-CATT is also higher in patients with pneumococcal meningitis than in the general population, suggesting that the tendency for a high MIF expression is not only unfavorable during infection, but also associated with an increased risk of its development [23, 24]. Two studies investigated but did not observe evidently increased MIF expression in patients with viral meningitis, although in the study by Østergaard et al., MIF concentration in CSF correlated with its inflammatory parameters [21, 24]. However, MIF is up-regulated in serum and correlates with a severe clinical presentation and a risk of fatal outcome in dengue, which is a systemic flavivirus infection [25], suggestive of its role in the pathogenesis of flavivirus meningitis/encephalitis as well. In fact, the existing data suggest MIF is up-regulated and plays a role in the pathogenesis of encephalitis caused by flavivirus, which may be characteristic for this particular etiology [21, 24, 26]. In Japanese encephalitis virus (JEV) infection in mice, MIF is synthesized in neurons and glial cells prior to TNFα and triggers intrathecal inflammation with resulting tissue damage [27]. In WNV encephalitis model studied by Arjona et al., MIF appears crucial for the CNS invasion, while in patients with WNV encephalitis, MIF serum concentrations were increased, the median concentration in CSF was an order of magnitude higher than in serum and a very high expression was confirmed in the brain parenchyma of two patients who died [26]. In our study, MIF was up-regulated in serum to higher concentrations than in CSF early in the neurologic phase of TBE, consistent with its implied role of a mediator of the systemic inflammation. Serum MIF concentrations tended to correlate with AQ, suggesting its link to BBB disruption. The correlation was statistically significant only in meningitis patients and after exclusion of one case, while the meningoencephalitis group was highly heterogeneous with respect to both parameters, possibly reflecting different timing of the sample collection in relation to the onset of encephalitis or the individual variability. The similarly increased serum MIF concentration in patients with non-TBE meningitis did not correlate with AQ. Although this group was relatively small, the lack of any positive trend suggests that the tendency could, if confirmed, be indeed specific for TBE, or, more generally, for neuroinfections of flavivirus etiology, and that it might reflect the mechanism of BBB impairment characteristic for them.

TLR3 is supposed to play an important protective role during CNS viral infections. It is expressed constitutively by a subpopulation of human neurons [46] and in a study by Farina et al., it was virtually the only functional TLR on human astrocytes, suggesting its importance in the intrathecal innate immunity [47]. The crucial role of TLR3 in the response to herpes simplex type 1 encephalitis has been suggested by Zhang et al. [48]. On the other hand, as TLR3 receptor seems involved in BBB disruption in the mouse model of WNV encephalitis, its effect in flavivirus neuroinfections could be paradoxically detrimental [33]. Kindberg et al. have studied two common SNPs in TLR3 gene, rs3775291 and rs5743305, in a group of 128 patients with TBE, 77 with viral meningitis/meningoencephalitis of other etiology, and 138 healthy persons from Lithuania. The homozygocity for the low-expression mutant T allele in rs3775291 was found in 19% of the controls and 21% of the patients with non-TBE neuroinfection but only in 7% of the TBE patients [39]. Similarly, Mickienė et al. detected the low-expression rs3775291 TT genotype almost twice less frequently in TBE patients than in the healthy population. However, if present in TBE, the T allele and TT genotype were associated with a more severe clinical presentation [38]. This apparent paradox could be explained by the hypothesis that TLR3 expression indeed facilitates the onset of the neurologic disease by supporting the TBEV penetration through BBB, but has a protective effect during the established CNS infection. Barkhash et al. have confirmed the protective role of the T allele in patients from Novosibirsk (Russia) [40]. The homozygocity for the wild type allele was present in 53.3% of cases of symptomatic TBEV infection and in 62.5% of cases of encephalitis or myelitis, so higher TLR3 expression was associated with a more severe manifestation [40]. This difference between the studies by Mickiene et al. and Barkhash et al. has not been explained, but it could be caused either by a difference between TBEV subtypes and strains prevalent in the locations of both studies or by a different genetic background in the studied populations.

Our study did not confirm correlation between rs3775291 genotypes and TBE frequency or severity, possibly because the low-expression allele was less frequent than previously described, on the level of 7% TT homozygocity in the healthy controls. On the other hand, the frequency of TT homozygocity in rs5743305 locus was higher in TBE group than in controls, although in a study by Kindberg et al., this particular SNP did not associate with TBE risk or severity [39]. This SNP, less studied than rs3775291, is located in the promoter region of TLR3 and supposed to affect its transcriptional activity. In a study by Dhiman et al., the AT heterozygocity in this locus correlated with a weaker serologic response to measles vaccination, confirming the association of rs5743305 with anti-viral response [49]. Interestingly, the protein concentration in the convalescent CSF was higher in the bearers of TT and TC genotypes in rs3775291 and of the TC/AT combined rs3775291/ rs5743305 genotype. The difference was present only in the convalescent CSF and not directly related to the BBB function in the early phase of the disease. The higher protein concentration was associated with allele either previously described as protective (T in rs3775291) or found to be protective in our study (A in rs5743305). As the low-expression variants protective in the context of early BBB permeability should also associate with a weaker intrathecal response during neuroinvasive disease, the observed trend may represent a less effective control of the CNS infection resulting in a delayed resolution of the inflammation.

The data on rs755622 are to our knowledge the first on MIF variability in TBE. The trend in the distribution of the high-expression C allele and CC genotype was statistically weak because of their low frequency, but consistent with our hypothesis. There were only three subjects with a high-expression CC genotype, all belonging to TBE group and presenting with uncomplicated meningitis. Such a distribution is analogous to findings of Mickiene et al. regarding TLR3 polymorphism and consistent with MIF expression not only facilitating the onset of the neurologic disease but also playing a protective role during the CNS infection. Interestingly, two meningitis patients with CC genotype had atypical composition of the CSF cellular infiltrate, with very low total and lymphocytic pleocytosis suggestive of a peculiar pathology. As the CG genotype was relatively frequent, but did not associate with TBE risk and presentation, the heterozygocity in rs755622 seems not to have a significant influence on the response to TBEV.

Finally, we have analyzed the association of the expression of the studied cytokines with SNPs in MIF and TLR3, although the strength of the analysis was limited by a relatively small size of the cytokine study group. Unfortunately, no CC homozygotes in MIF rs755622 were detected in this group so it remains to be determined to what degree this genotype actually influences MIF expression in TBE. However, there was an expected trend for lower cytokine concentrations associating with the protective TLR3 rs5743305 A allele, close to reaching the statistical significance for TNFα. Of note, TBE patients with rs5743305 TT genotype had significantly up-regulated serum TNFα levels, while in AT genotype, the increase was non-significant and in AA the serum TNFα expression was hardly detectable. As TT was identified as a high-TBE-risk genotype, this result agrees well with TNFα expression promoting the disease onset and provides a link between the TLR3 expression and the TNFα synthesis during an infection with TBEV in vivo. Even more importantly, the variability associated with rs5743305 explains the unexpectedly low median TNFα concentration in a TBE group and proves that the expression of this cytokine may play a pathogenetic role in a large subgroup of patients. The data show that the suspected pathway involving TLR3, MIF, and TNFα may be functional in human TBE, but that its expression should be highly variable individually in association with the genetic background, which in turn should contribute to a variability of clinical presentation.