Stroke represents the second leading cause of mortality worldwide and a major cause of long-term disability (1). Despite the improvement of acute treatment, the overall mortality appears unchanged (2). Nowadays, revascularization is the only available approach, while specific therapies for limiting contributing mechanisms of neuronal death triggered by ischemia are lacking. Immunological factors may indeed participate in brain injury following a stroke.

Antibodies against neuronal surface antigens (NSA-Abs) might appear in circulation after a stroke as a result of neuronal damage. Brain-derived antigens have been detected in the lymphoid tissue of patients with stroke (3), and there is evidence of intrathecal antibody synthesis (4, 5). These findings suggest that NSA-Abs may appear after the acute phase, possibly modulating the long-term clinical outcome. Having been associated with cognitive impairment and neurodegeneration (6), NSA-Abs offer a tantalising additional candidate for the pathogenesis of post-stroke dementia.

Several pieces of evidence involved glutamate excitotoxicity as one of the main mechanisms of neuronal death in ischemic stroke (7–11), leading to several attempts to use N-methyl-D-aspartate receptor (NMDAR) blockers as additional therapy in stroke (12), although with discouraging results (13). Serum NMDAR autoantibodies (NMDAR-Abs), mostly against the NR2 subunit, have been implicated in stroke as a possible diagnostic biomarker as well as a marker of neurotoxicity (14–17). In 2008, NMDAR-Abs, directed against the NR1 subunit, were described in patients with an autoimmune encephalitis characterized by a specific clinical picture encompassing psychiatric disorders, cognitive and behavioral changes, movement disorders, frequently accompanied by seizures, hypoventilation and autonomic instability (18). In this clinical setting, the antibodies, belonging to the immunoglobulin (Ig) G isotype, are found in the CSF and have been proven pathogenic, inducing NMDAR internalization with a consequent decrease in NMDAR-mediated synaptic function (19). However, since this first description, NMDAR-Abs against the NR1 subunit belonging to different Ig isotypes, comprising IgA, IgM and IgG, have been found in the serum of patients with various neuropsychiatric disorders, including stroke, as well as in healthy controls (20–23). Nevertheless, their role in patients with stroke, as well as in other neurological conditions, is unclear.

Given the growing need to identify additional acute treatments for patients with stroke to reduce mortality and long-term disability and the possible role of humoral immunity both in the acute stroke phase as well in the development of vascular dementia, we conducted a systematic review and meta-analysis to investigate: 1) the frequency of NSA-Abs in patients with acute stroke (AS) compared to controls; 2) the de novo appearance of NSA-Abs after stroke; 3) the effects of NSA-Abs on the clinical outcome.

This is the first systematic review and meta-analysis investigating the possible association between NMDAR-Abs seropositivity and stroke. Despite the potential implication of autoimmune mechanisms in stroke, there is a paucity of available data and a considerable heterogeneity of study design and antibody detection methods. Albeit these limitations, we found that serum NMDAR-Abs were significantly more common in ASP than controls. This association was maintained for the ischemic stroke subgroup, while an analysis was not possible for hemorrhagic stroke. In a subanalysis, the association was confirmed only for IgA/IgM (but not IgG) serum NMDAR-Abs. Finally, no significant association was observed between serum NMDAR-Abs and discharge/day 7 NIHSS or follow-up mRS (3-12 months).

In the interpretation of these results, several methodological issues should be taken into account. Foremost, significant heterogeneity between studies and large confidence intervals for most of the effect sizes were observed. There was a substantial methodological variation between studies, including differences in patient and stroke characteristics, study design and outcome definition, antibody target (NR1 vs NR2) and antibody detection method. Therefore, to reduce these biases, we focused mainly on studies adopting CBA and investigating the presence of NR1-Abs. Second, the overall study quality was moderate and had significant bias. In particular, sex- and age-matching between patients and controls was provided only in one study (15), introducing a relevant bias since NMDAR-Abs seropositivity is known to increase with age (28, 29), and therefore possibly explaining the higher frequency of NMDAR-Abs in the stroke population. Third, only two studies (27, 28) could be included in the antibody isotype analysis, and given the low prevalence of IgG antibodies, the lack of difference could be a false negative result. NMDAR-IgGs, in serum and CSF, are associated with NMDAR encephalitis and are considered pathogenic. However, in the stroke cohorts included in the present study, antibodies were tested only in the serum. It must be noted that the significance and relevance of isolated serum IgG positivity is debated, and even less clear is the possible role of e of IgA/IgM antibodies. These non-IgG NMDAR-Abs have been associated with unclassified/atypical dementia phenotypes (20) as well as with cognitive deficits in patients with cancer (35), suggesting a possible role in cognitive impairment. However, in vitro studies showed contrasting results, with some showing no effect of these isotypes (21) and others (36) reporting a decreased NMDAR expression on the surface of cultured hippocampal neurons with a concomitant reduction of NMDAR-mediated currents similar to that obtained with NMDAR-IgG. Some Authors hypothesize that these antibodies, which have been found across multiple mammals (23), might have a role in physiological immunity (37). In the context of stroke, it has been suggested that NMDAR-Abs might act as endogenous receptor antagonists and, therefore, as a protective factor against glutamate cytotoxicity (29, 38). After a stroke, the BBB damage could allow the serum antibodies to rapidly access their target in the CNS and limit the damage, as suggested by the smaller infarct volume in NMDAR-Abs seropositive non-APOE4 carriers (29). On the other hand, a chronically altered BBB, as observed in APOE4 carriers, was associated with a worse stroke outcome (29). However, another study found a detrimental effect of NMDAR-Ab on stroke evolution (33), although the BBB integrity was not investigated. Besides the BBB status, the differential expression and role of NMDAR on extra-neuronal cells, including brain endothelial cells and platelets, might be relevant in determining the final effect of their antagonism (39).

Regarding NMDAR-Abs effects on stroke clinical outcome, our quantitative analysis was limited to the discharge/day 7 NIHSS and follow-up mRS (3-12 months). Although no difference was observed between patients with and without antibodies, these results must be interpreted cautiously due to the substantial heterogeneity. Moreover, the unavailability of individual patient data, did not allow a shift analysis; hence, only the mean value difference for mRS was calculated. In addition, given the lack of individual patient data, antibody positivity was included in the analysis as a binary variable, which could have masked the effect of antibody titers on the clinical outcome. For instance, the PROSCIS-B study (30) found that only high titers were associated with worse functional outcomes at one year and worse cognitive performance and depression at 3-year follow-up (31, 32).

In all the studies, the antibodies were assessed close to stroke onset, making a de novo formation unlikely (40) and supporting a pre-existing seropositive status instead. In this light, the observed higher prevalence of IgA/IgM NMDAR-Abs in patients compared to controls seems counterintuitive. Despite the aforementioned lack of age-matching between patients and controls might be a sufficient explanation, there is not enough evidence to draw this conclusion. Indeed, it could also be possible that, in patients with cerebrovascular disease, previous clinical (TIA) and subclinical events (i.e. asymptomatic radiological vascular lesions load) could allow the antibody formation before the major stroke event. Therefore, assessing for confounding like vascular risk factors and small vessel disease MRI markers should be performed in APS and controls in future studies. On the other hand, no sufficient data was obtained on the persistence or de novo formation of NSA-Abs after stroke to support these hypotheses, which should be the focus of future studies.

Our study has other limitations. Although we initially sought to include all NSA-Abs, the scarce number of studies investigating their presence, along with their low prevalence, forced us to limit the analysis to NMDAR-Abs. We included studies looking for NMDAR-Abs against different subunits and using different assays. Since only NR1-abs are considered pathogenic, and CBA is considered the gold standard for their detection, we focused mainly on studies using CBAs. In any case, the restricted number of studies limits the overall strength of our findings.

CBA-detected NMDAR-Abs were associated with IS; specifically, serum IgA/IgM but not IgG NMDAR-Abs were more common in patients compared to controls. Considering the significant heterogeneity, there is insufficient evidence to draw conclusions about the frequency of NMADR-Abs in HS and their influence on clinical outcomes. Additional, methodologically robust studies are needed to clarify the prevalence and significance of NMDAR-Abs in APS and, more generally, the role of the IgA/IgM antibody isotype in neurological and psychiatric disorders. Future studies should include patients with both IS and HS, as well as large age and sex-matched control groups to ascertain possible confounders.