Cognitive deficits are common in Multiple Sclerosis (MS), affecting almost half of the patients and negatively influencing social functioning and quality of life (Rao et al., 1991; Benedict et al., 2006; Chiaravalloti and DeLuca, 2008). MS is a chronic inflammatory immune-mediated disorder of the central nervous system (CNS) characterized by a variable course of clinical manifestations. Whereas the role of demyelinating white matter lesions and gray matter atrophy in motor and sensory deficits has been extensively investigated, the pathogenesis of cognitive dysfunction in MS is not completely elucidated. In recent years, neuroinflammation is emerging as a main factor possibly influencing cognitive dysfunction in MS (Gentile et al., 2015).

Different mediators released by immune cells, including inflammatory cytokines and neurotrophins, influence synaptic transmission. It has been proposed that an imbalance between proinflammatory (e.g., interleukin-1β, IL-1β, and tumor necrosis factor α, TNFα) and anti-inflammatory (e.g., IL-4 and IL-10) cytokines with a prevalence of the former, may contribute to brain damage in MS (Linker et al., 2005; Zeis et al., 2008; Ivanov and Lindén, 2009).

Different animal models of MS, including virus-induced (Theiler’s murine encephalitis virus induced demyelinating disease, TMEV-IDD) and autoimmune models (Experimental Autoimmune Encephalomyelitis, EAE) have been used to explore pathophysiological mechanisms of disease. EAE is induced by either the administration of protein or peptide in adjuvant or by the adoptive transfer of encephalitogenic T-cell blasts into naïve recipients. TMEV belongs to the cardiovirus group of the Picornaviridae and induces persistent immune demyelinating disease in mice (Miller, 1995). A common feature of both animal models is the release of proinflammatory cytokines and the recruitment of Th1 cell, monocytes and macrophages in the CNS, leading to myelin damage (Dal Canto and Lipton, 1975).

In EAE specific proinflammatory cytokines, including IL-1β and TNFα, alter both excitatory and inhibitory transmission resulting in synaptic hyperexcitability and excitotoxic neuronal damage (Centonze et al., 2009; Rossi et al., 2011; Mandolesi et al., 2013). Accordingly, the administration of either AMPA receptor inhibitors (Centonze et al., 2009) or IL-1β receptor antagonist (IL-1ra; Furlan et al., 2007; Mandolesi et al., 2013) is able to reduce both neurodegeneration and synaptic alterations in EAE mice, confirming the role of inflammation-induced excitotoxicity.

Also in MS patients, cerebrospinal fluid (CSF) levels of inflammatory cytokines are associated to analogous alterations of both inhibitory and excitatory transmission, resulting in synaptic hyperexcitability (Rossi et al., 2012a; Mori et al., 2016). Furthermore, CSF from MS patients in the active phase of disease reproduced in rodent brain slices both glutamatergic and GABAergic alterations and neuronal degeneration observed in EAE (Rossi et al., 2012a,b). Notably, these synaptic alterations did not occur when CSF from MS patients was coincubated with IL-1β inhibitors (Rossi et al., 2012a,b). In addition, different anti-inflammatory cytokines showed neuroprotective effects by normalizing glutamate (Garg et al., 2009), enhancing GABA signaling (S-Rózsa et al., 1997) and attenuating glutamate-mediated excitotoxicity (Zhou et al., 2009). Coherently, there is also evidence that anti-inflammatory cytokines may contribute to reduce synaptic hyperexcitability and neurodegeneration in MS patients (Rossi et al., 2011).

Overall, these data suggest that synaptic alterations associated to neuroinflammation may represent a critical factor inducing neuronal dysfunction in MS.

Different mechanisms have been proposed to explain cognitive impairment in MS patients. Clinical and magnetic resonance imaging (MRI) studies showed an association between white matter lesions and neuropsychological performance evaluated with different tests. In particular, lesion volume and site influence cognitive performance, highlighting the role of disconnection mechanisms (Vellinga et al., 2009; Kincses et al., 2011; Rossi F. et al., 2012). Gray matter damage is increasingly regarded as a main predictor of cognitive dysfunction in MS. Cortical lesions and in particular hippocampal CA1 region atrophy, has been associated with memory deficits in MS patients (Sicotte et al., 2008; Calabrese et al., 2009). In some cases, cognitive deficits appear already in the early phase of MS (Zivadinov et al., 2001; Olivares et al., 2005; Deloire et al., 2006) and are not associated to any substantial neuronal damage, namely isolated cognitive relapses (Coebergh et al., 2010; Pardini et al., 2014), suggesting that alternative mechanisms may be implicated. Moreover, the fact that cognitive impairment may occur at the early stage of the disease, before motor dysfunction appearance, suggests that cognitive decline associated with MS is mediated by a distinct mechanism, e.g., neuroinflammation.

Cognitive deficits have also been investigated in experimental models of MS. It has been demonstrated that spatial learning and memory deficits appear in EAE mice in different disease stages. In the late phase, for example, one study reported persisting memory acquisition and maintenance deficits after recovery of motor symptoms associated to reduced choline acetyltransferase activity in the hippocampus, cerebral cortex and basal forebrain. As memory deficits improved after anticholinesterase treatment, it was suggested that altered acetylcholine transmission could affect memory in the late phase of EAE (D’Intino et al., 2005). In addition, hippocampal degeneration and spatial learning deficits have been observed in EAE mice at a relatively late phase associated with decreased hippocampal volume and loss of GABAergic interneurons (Ziehn et al., 2010). These findings suggest that hippocampal structures seem to be particularly susceptible to inflammatory-dependent damage in EAE (Yirmiya and Goshen, 2011).

Learning and memory deficits have been also evidenced in the earlier phases of the disease, before the onset of motor and sensory symptoms, when EAE is induced in apolipoprotein E (APOE) knockout mice and human APOE ε4 (APOE4) knock-in mice (Tu et al., 2009). In particular, as early cognitive deficits and hippocampal cholinergic dysfunction were evident only in the presence of both EAE associated neuroinflammation and APOE-KO/APOE4 knock-in, a two-hit mechanism has been proposed. That is, increased susceptibility of cognitive deficit can be secondary to poor repair mechanisms associated with APOE-KO/APOE4 knock-in in the presence of a proinflammatory response (Tu et al., 2009). Interestingly, it has been recently suggested that APOE could play a critical role in neurodegenerative disorders regulating microglial function and promoting the switch to a neurodegenerative phenotype (Krasemann et al., 2017). APOE is the major apolipoprotein in the CNS, and is critically involved in neurite and synapse remodeling and synaptic plasticity (Kim et al., 2009). The APOE4 polymorphism has been associated with earlier age of onset in AD patients (Kim et al., 2009) and also associated with learning and memory deficits in MS, particularly in young patients (Shi et al., 2008). However, it should be noted that two studies found no association between APOE4 and cognitive deficits in MS (Portaccio et al., 2009; Carmona et al., 2011), therefore the role of this polymorphism in MS still requires further investigation.

Synaptic plasticity occurring within hippocampal structures is considered as one of the basic mechanisms of learning and memory processes (Stuchlik, 2014). It can be hypothesized that, even in the absence of apparent anatomical damage, functional alterations in the hippocampus could disrupt synaptic plasticity leading to cognitive deficits.

The ability of neurons to undergo functional long-term modifications at existing synapses is referred to as synaptic plasticity. Long-term potentiation (LTP), one of the most studied form of synaptic plasticity, consisting in a persistent enhancement of synaptic strength, is also characterized by structural rearrangements and neurotrophin-induced protein synthesis (Bliss and Collingridge, 1993; Cunningham et al., 1996; Murray and Lynch, 1998; Malenka, 2003). Another form of synaptic plasticity, known as long-term depression (LTD), describes long-lasting weakening of synaptic strength. Different experimental protocols have been designed to explore LTP and LTD-like plasticity in EAE animal model and in MS patients.

Preclinical studies in EAE showed that inflammatory cytokines alter synaptic plasticity. Impairment of hippocampal LTP has been reported during the initial acute phase in EAE and has been associated with a selective reduction of NMDA receptors, microglial activation and IL-1β increase (Di Filippo et al., 2013). Moreover, persistent microglial activation together with impaired hippocampal LTP was also shown during remission in EAE (Di Filippo et al., 2016). Conversely, other studies evidenced that inflammation may also subvert synaptic plasticity. A study, exploring both LTP and LTD-like hippocampal plasticity in EAE, showed that LTP induction was favored over LTD. In particular, this alteration was mediated by the proinflammatory cytokine IL-1β, interfering with GABAergic transmission (Nisticò et al., 2013). Moreover, in vivo blockade of IL-1β in EAE reduced the alterations of hippocampal synaptic plasticity (Mori et al., 2014). Altogether, these data show that inflammation alters hippocampal synaptic plasticity in vitro in EAE mouse model.

Synaptic plasticity can be also explored non-invasively in MS patients by using specific transcranial magnetic stimulation (TMS) protocols (Mariorenzi et al., 1991; Fitzgerald et al., 2004; Ziemann et al., 2008). In particular, two different theta burst stimulation (TBS) protocols have been widely used to elicit LTP-like and LTD-like effects, respectively intermittent TBS (iTBS) and continuous TBS (cTBS; Di Lazzaro et al., 2005; Huang et al., 2005).

TMS studies evidenced that in relapsing remitting (RR)-MS patients, CNS inflammation alters plasticity (Stampanoni Bassi et al., 2017) and that, during relapses, iTBS induced LTP-like plasticity is impaired (Mori et al., 2011, 2012). In remitting patients, response to the iTBS protocol was comparable to healthy controls. Conversely, in response to cTBS an abnormal LTP-like effect was observed, which showed a positive correlation with CSF IL1-β levels (Mori et al., 2014). It has been proposed that, during relapses, the lack of LTD-like effects after cTBS may rely on reduced GABAergic transmission (Caramia et al., 2004; Rossi et al., 2012b) or increased glutamatergic signaling (Rossi et al., 2012a).

These results suggest that synaptic plasticity is profoundly altered by neuroinflammation, providing a plausible substrate for cognitive deficits. In particular, some evidence suggests that acute inflammation can be associated with both cognitive impairment and altered synaptic plasticity (Mori et al., 2011, 2012). In RR-MS patients, acute inflammation, as evidenced by the presence of gadolinium enhancing (Gd+) lesions at MRI scan, was associated with both impaired LTP-like plasticity and cognitive impairment, as shown by reduced PASAT score (Mori et al., 2012). According with the inflammatory origin of these alterations, both PASAT score and synaptic plasticity improved after 6-month treatment with interferon-β (IFN)-beta 1a in Gd+ patients, whereas they did not in patients without evidence of acute inflammation at MRI (Mori et al., 2012).

LTP expression could be regulated by different dynamics acting both at receptor level and on downstream mechanisms triggered upon receptor activation (Bliss and Collingridge, 1993; Kessels and Malinow, 2009; Minichiello, 2009). Amyloid-β (Aβ) modulate synaptic functioning through different mechanisms including the modulation of other signaling systems (cytokines, neurotrasmitters/messengers), in particular the modulation of nitrergic system and involvement of IL-1 receptors could play a crucial role (Morgese et al., 2015). An emerging key factor linking together inflammation, synaptic plasticity and neurodegeneration in different neurological diseases is the alteration of Aβ metabolism. Aβ peptides derive from the proteolytic cleavage of amyloid precursor protein (APP), a transmembrane protein. In the amyloidogenic pathway, APP undergoes a first cleavage by beta-site APP-cleaving enzyme 1, followed by further cleavage by γ-secretase to release 40 or 42 amino-acid long Aβ fragments. However, APP may go through different non-amyloidogenic pathways preventing Aβ formation (Andreasson et al., 2007). The Aβ peptides are highly hydrophobic and tend to aggregate to form dimers, oligomers or amyloid fibrils and have been identified as a major insoluble component of amyloid plaques (Masters and Selkoe, 2012). The aggregation of soluble oligomers to insoluble fibrils in amyloid plaques of Alzheimer’s disease (AD) reduces the Aβ CSF concentrations (Blennow and Hampel, 2003).

Several experimental data support the view that Aβ can impair hippocampal LTP (Yamin, 2009). Injections of Aβ oligomers in rats inhibit LTP (Walsh et al., 2002) and promote LTD (Li et al., 2009) in the hippocampus. Moreover, it has been shown that Aβ_1–42 can alter both early and late LTP phases (Chen et al., 2002; Zhao et al., 2004). Altered Aβ metabolism has been consistently associated to the pathophysiology of AD and in particular with hippocampal LTP impairment (Klyubin et al., 2005; Shankar et al., 2008). Accordingly, Aβ dimers isolated from AD patients can impair hippocampal LTP and memory in mice and induce dendritic spine retraction in neurons (Shankar et al., 2008). Although through different pathophysiological mechanisms, brain inflammation is a common feature of both MS and AD (Lassmann, 2011). Under normal physiological conditions, there is a balance between Aβ production and clearance (Iwata et al., 2001; Saito et al., 2005) and inflammation can alter such equilibrium (Griffin et al., 2006; Hickman et al., 2008; Schmidt et al., 2008). Therefore, as in AD, Aβ may represent a possible player influencing both synaptic dysfunction and neurodegeneration occurring in MS as well (Gentile et al., 2015). Indeed, Aβ can be found in MS multifocal lesions (Ferguson et al., 1997; Trapp et al., 1998). Furthermore, reports of Aβ levels in CSF samples of MS patients, albeit puzzling (Hein Née Maier et al., 2008; Valis et al., 2008; Sladkova et al., 2011; Szalardy et al., 2013), mostly evidenced that a general alteration of Aβ metabolism occurs in MS (Mattsson et al., 2009; Mai et al., 2011; Mori et al., 2011; Augutis et al., 2013).

In line with the possibility that inflammation-induced alteration of Aβ homeostasis could be a key factor in cognitive dysfunction, a study explored the correlation between CSF Aβ_1–42 levels and TBS-induced plasticity in a group of cognitive impaired (CI) and cognitive preserved (CP) MS patients (Mori et al., 2011). It was found that Aβ_1–42 levels were lower in CI patients compared to both CP patients and controls. Furthermore, CSF Aβ_1–42 levels inversely correlated with the number of Gd+ lesions at MRI. Finally, altered iTBS-induced synaptic plasticity was observed in CI patients, and Aβ_1–42 CSF levels positively correlated with reduced LTP-like plasticity. Overall, these data suggest that inflammation-driven alteration of Aβ metabolism in MS could disrupt LTP and impair cognitive function. Intriguingly, in AD experimental models specific proinflammatory cytokines could alter Aβ synthesis and clearance (Wang et al., 2015). In line with this, we propose that, as in AD, a possible imbalance between proinflammatory and anti-inflammatory cytokines leading to altered Aβ homeostasis may occur also in MS and contribute to the cognitive deficit observed in this disorder.

To test the possibility that inflammation could interfere with Aβ metabolism in RR-MS patients, we investigated possible correlations between Aβ_1–42 concentrations and the levels of some proinflammatory and anti-inflammatory molecules in the CSF (IL-1β, IL1-ra, IL-8, IL-10, IL12, TNFα, IFNγ). The study, involving 103 human subjects, was approved by the Ethics Committee of the University Hospital Tor Vergata, Rome. All patients gave written informed consent to take part to the study. The diagnosis of RR-MS was established according to published criteria (Polman et al., 2011). Clinical and demographic characteristics of MS patients are shown in Table 1. No immunoactive drug was given before hospitalization and corticosteroids or immune-modulating therapies were initiated later. Lumbar puncture was performed at the time of diagnosis, during hospitalization. CSF was centrifuged and immediately stored at −80C until analyzed using a Bio-Plex multiplex cytokine assay (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer’s instructions. For the analysis of Aβ_1–42 standard procedures using commercially available sandwich enzyme-linked immunosorbent assays (Innotest β-Amyloid_1–42, Innogenetics, Ghent, Belgium) were employed (Sancesario et al., 2010). For the analysis of cytokines levels, concentrations were calculated according to a standard curve generated for each target and expressed as pg/ml. An arbitrary value of 0 pg/ml was assigned to the concentrations of the cytokines measured below the detection threshold.

Data are presented as mean (standard deviation, SD) or as median (25–75th percentiles) if not normally distributed. Kolmogorov-Smirnov test was applied to verify normality of data distribution. Non parametric Spearman’s correlational analysis was performed to evaluate the correlation between CSF levels of Aβ_1–42 and CSF levels of the main proinflammatory and anti-inflammatory cytokines. In addition, the correlation between CSF levels of Aβ_1–42 and age, Expanded Disability Status Scale (EDSS) at baseline and disease duration was evaluated. Non parametric Mann-Whitney test was applied to evaluate difference between sexes in Aβ_1–42 levels. A p value <0.05 was considered significant. Benjamini–Hochberg correction was applied to adjust the p value and control the false discovery rate in the multiple testing.

The results showed no significant correlations between CSF levels of Aβ_1–42 and age (Spearman’r = −0.05, p = 0.780), disease duration (Spearman’r = −0.17, p = 0.149) and EDSS (Spearman’r = 0.02, p = 0.955). No significant differences between genders (pMann-Whitney = 0.519) and disease activity at diagnosis (pMann-Whitney = 0.520) were found. CSF levels of Aβ_1–42 were negatively correlated with IL-8 (r = −0.417; p < 0.001) and IFNγ (r = −0.216; p = 0.034), borderline with IL1-β (r = −0.18; p = 0.073), and positively correlated with IL-10 (r = 0.381; p < 0.001) and IL-1ra (r = 0.341; p = 0.001; Table 1). Applying the Benjamini-Hochberg correction further confirmed a positive correlation between Aβ_1–42 and IL-8 (padj < 0.001), IL10 (padj < 0.001) and IL1-ra (padj = 0.002), whereas correlation with IFNγ was borderline (p = 0.077) and correlation with IL-1β was no longer significant (p = 0.132). Correlations between Aβ_1–42 levels and TNFα and IL-12 were not significant.

MS is classically considered a demyelinating disease of the CNS, primarily involving the white matter and followed by neurodegeneration in the late phases. Cognitive deficits in MS have been related to white matter damage, albeit hippocampal structures involvement is emerging as a key component of the cognitive dysfunction observed both in EAE and MS. In recent years, neuroinflammation has been identified as a factor inducing both neurodegeneration and synaptic plasticity dysfunction. How inflammation alters hippocampal plasticity in MS is still scarcely understood, although it has been proposed that altered Aβ metabolism could play a crucial role.

Our data add novelty to previous evidence showing that inflammation influences Aβ metabolism in MS. The significant negative correlation between CSF Aβ_1–42 concentrations and the proinflammatory cytokines IL-8 and IFNγ suggests that inflammatory response is associated to dysregulation of Aβ synthesis and degradation. IL-8 is a well-known biomarker of neuroinflammation (Komori et al., 2015) and its CSF levels have been found elevated in several inflammatory and non-inflammatory neurological conditions (Bielekova et al., 2012). In MS, a previous study investigated the association between IL-8 CSF levels and visual recovery 6 months after acute optic neuritis, showing that higher levels of IL-8 correlated with incomplete visual recovery (Rossi et al., 2014). In a further study, elevated IL-8 CSF levels during acute inflammation correlated with clinical progression in patients with radiologically isolated syndrome and with the risk to develop MS in those with clinically isolated syndrome (Rossi et al., 2015). IFNγ has been involved in the pathophysiology of both MS and EAE. Indeed, higher IFNγ levels within the CNS have been found during inflammation in MS (Cannella and Raine, 1995; Kahl et al., 2002) and also in EAE (Gardner et al., 2013; Hidaka et al., 2014).

Our results also show that anti-inflammatory cytokines may have a beneficial effect contributing to reduce the alterations of Aβ metabolism. In particular, IL-10 is considered one of the main anti-inflammatory cytokines involved in modulating brain inflammatory response (Kwilasz et al., 2015) and has been proposed as a useful treatment in several neurological conditions characterized by persistent neuroinflammation and neurodegeneration (Cua et al., 2001; Joniec-Maciejak et al., 2014). IL-1ra, is an anti-inflammatory endogenous molecule acting as competitive inhibitor of IL-1β (Seckinger et al., 1987). Notably, IL-1ra administration ameliorated EAE clinical manifestations (Martin and Near, 1995; Badovinac et al., 1998; Furlan et al., 2007). These results suggest that anti-inflammatory molecules may reduce the impact of neuroinflammation on Aβ_1–42 homeostasis, in line with previous results showing that altered hippocampal synaptic plasticity in EAE could be reduced by blocking IL-1β transmission (Mori et al., 2014).

Our data suggest that altered Aβ homeostasis could represent a key factor linking together inflammation, synaptic plasticity and cognitive dysfunction in MS (Figure 1). Although previous reports have shown an association between IL-1β and Aβ homeostasis in AD experimental models (Wang et al., 2015), this correlation was not found in the MS patients involved in the present study. However, on the basis of the preliminary data reported here, we may conclude that the inflammatory process underlying MS is complex and may implicate a more widespread cytokine release. Additional investigations are required to explore the relationship between CSF chemokines and Aβ metabolism in normal subjects and to further characterize the role of CSF inflammation and altered synaptic plasticity in the development of cognitive deficits in MS patients.

In conclusion, the differential modulation of Aβ metabolism by proinflammatory and anti-inflammatory cytokines could be relevant to identify novel therapeutic approaches to hinder the progression of the cognitive dysfunction in MS.

MSB: data interpretation, work conception and design, drafting the work, work revision; SG and LG: data interpretation, work revision; GAM: design of the work, work revision; IS: data analysis, data interpretation, work revision; AF and RF: data analysis, data interpretation; GMS: data acquisition and analysis, data interpretation; JDG and MS: data acquisition and analysis; FM: data acquisition and analysis, data interpretation; DC: data interpretation, work conception and design, work revision; EI: work conception and design, drafting the work, work revision.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.