Specific B- and T-cell populations are associated with cognition in patients with epilepsy and antibody positive and negative suspected limbic encephalitis

At the Department of Epileptology, University Hospital Bonn, Germany, between 2014 and 2016, 124 patients were examined for suspected autoimmune LE due to the presence of late-onset subacute epilepsies, with cognitive and/or psychiatric symptoms, and also because other reasonable alternative causes could be excluded [14].

All patients had peripheral blood (PB) and cerebrospinal fluid (CSF) extracted for autoantibodies identification and immune cells count. From the group of 124 patients, 18 patients were excluded from the analysis due to a lack of neuropsychological evaluations. In total, our study was comprised of 106 LE patients who had undergone neuropsychological evaluation at the time when CSF and PB were extracted and an MRI was performed. Demographic information (age, sex, and education), neuropsychological and clinical data (i.e., epilepsy onset, duration of epilepsy, the number of antiepileptic drugs taken, the number of psychoactive drugs, including antidepressants and neuroleptics, and structural MRI volumetric data) were collected retrospectively from medical files and records located in the databases of the Department of Epileptology, University Hospital Bonn.

This work was carried out in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki) for experiments involving humans. Written informed consent was obtained from all patients and the study was approved by our local Ethics Committee (# 222/16).

To examine the relationship between morphological structural changes in the brain and cognition and mood, brain scans were obtained utilizing a 3-T MRI scanner at the Department of Neuroradiology (Philips Medical Systems, Germany) or the Life&Brain Institute (Magnetom Trio, Siemens, Germany). The structural MRI findings were categorized as normal, swollen, hyperintense, or atrophied hippocampi and/or amygdalae. The abnormalities in the hippocampus and/or amygdala were categorized as unilateral or bilateral. Three procedures were available for categorizing the MRI data: a visual inspection by neuroradiologists; an automated volumetry based on Freesurfer as described by Wagner et al. in 2015 [49]; and MAP/07, and a morphometric analysis program developed by J. Huppertz [6].

Immune cells and their subsets were isolated and identified using the flow cytometry technique. The two main immune cells analyzed in this study were T-lymphocytes (T-cells) and B-lymphocytes (B-cells) which were categorized according to their cluster of differentiation (CD) and MHC class II, cellular surface receptors, into human leukocyte antigen DR isotype (HLA-DR+) CD4+ T-cells, HLA-DR+  CD8+ T-cells, and CD19+ B-cells from both CSF and PB. We utilized the fluorescences HLA-DR-ECD, CD4-APC, 700CD8-PacificBlue, and CD19-APC-Alexafluor700. The number of cells was counted per milliliter (cells/ml). We obtained at least 3 ml of CSF per patient. We measured the cell amounts and percentages of T-cells per total leucocytes on a BD LSR Fortessa flow cytometer (BD Bioscience, California, USA) and the data were analyzed with Kaluza software (Beckman Coulter GmbH, Life Science, Krefeld, Germany) and the following fluorochrome-conjugated antibodies (Beckman–Coulter). The gating strategy for identifying PB and CSF leukocytes subsets was performed via mainstream, cell line lineage markers. We focused on T- and B-cell populations which, according to previous investigations, pointed towards a relevant impact of these cell populations on limbic encephalitis [13]

The results of the auto-antibody testing were categorized into three subgroups: (1) antibodies against intracellular targets (IAB+); (2) antibodies against membrane surface antigens (SAB+); and (3) no detection of specific antibodies (AB−). Commercial kits were used for diagnostic procedures including: Amphiphysin, CV2, PNMA2 [Ma2/Ta; paraneoplastic antigen Ma2), Ri, Sox1, Yo, Zic4 and GAD65 (Glutamate acid decarboxylase 65] using semiquantitative immunoblots according to the manufacturer’s guidelines (EUROLINE PNS 12, Euroimmun, DL 1111-1601-7 G) coated with recombinant antigen or antigen fragments with serum diluted 1:100 and CSF 1:1. Complementary, indirect immunofluorescence was applied relying on HEK293-cells (Human Embryonic Kidney 293) with an overexpression of the individual antigens on the cell-surface (IIFT: ‘Autoimmune-Enzephalitis-Mosaik1’, Euroimmun, FA 1120-1005-1; GAD65-IIFT, Euroimmun, FA 1022-1005-50), including NMDAR-, CASPR-, LGI1- and GAD65-autoABs (dilution: serum 1:10; CSF 1:1) following the manufacturer’s protocol.

Neuropsychological memory tests were conducted on all patients to diagnose LE, determine a phenotype, and then obtain a baseline assessment for eventual follow-up examinations. While all patients underwent memory testing, 89 underwent additional assessment for attention and executive function, and 93 were assessed for mood.

To evaluate verbal memory, we the “Verbaler Lern- und Merkfähigkeitstest VLMT” [21, 32] which is a word list learning task, and is the German adaptation of the Rey Auditory Verbal Learning test (RAVLT). Our parameters of interest included learning over five learning trials, delayed free recall performance (absolute), loss of learned words over time, and recognition (minus errors).

To assess visual memory capacity, we employed the “Diagnosticum für Cerebralschädigung-Revised/DCS-R”, a design list learning task [22, 30]. We analyzed learning over five trials, performance in the last learning trial, and delayed recognition (minus errors).

Both tests, either taken alone or and in combination, are well-suited for monitoring memory in limbic encephalitis [11, 18, 34, 45, 57].

While regression analyses were performed using raw scores, individual impairments were calculated using categorized sum scores across the different visual and verbal memory subfunctions (mean value of standardized memory parameters = 100 ± 10). On a categorical level, we rated visual or verbal memory performance as impaired if the memory score was more than a one-fold standard deviation below the standardized mean value, i.e., < 90.

Executive function was assessed via the second edition of the EpiTrack [31], which merges six subtests into one total score representing attention, executive function, and working memory; its clinical utility has been proven in monitoring drug treatment [55] and in screening new-onset epilepsies [56, 58]. Here, just as with memory, patients were rated as impaired if they scored below the mean minus one standard deviation of the healthy normative sample.

Mood was assessed via the Beck Depression Inventory (BDI) I [3]. Values > 10 were rated as an indicator for depressed mood.

As already mentioned in the section describing subject selection, the timing for evaluating patients with LE is crucial because the longitudinal follow-ups show, in part, considerable fluctuations in cognition over time [15]. It was therefore essential to synchronize neuropsychology, imaging, and the flow cytometry evaluations. Furthermore, since autoimmune treatments may well have negative effects on cognition, it is important to note that neuropsychological assessments, as a rule, were never carried out if the patient was being treated for an autoimmune condition.

All statistics were performed using the IBM Statistical Package for the Social Sciences (SPSS) software, version 23. Descriptive statistics were performed to compare the three groups with regard to demographic, clinical–epileptological, neuropsychological, and flow cytometry data.

Hierarchical, multiple stepwise regression analysis was applied separately for both groups to discern prediction models for cognition. The following hierarchical levels were considered as independent predictor variables in the following order of appearance: (1) demographic data (age, gender, education); (2) epilepsy-related data (age at onset, duration, number of antiepileptic drugs, individual drugs which are known to negatively impact cognition or mood, antidepressants); (3) structural MRI pathology findings (see above); and (4) results from flow cytometry analyses. The inclusion criterion for predictor variables was p < 0.05; the exclusion criterion was p > 0.1.

Antibody screening targets were GAD65, onconeural auto-antibodies with reported positive cases in the context of LE [1, 28, 29, 33, 36‐38, 46], and neuronal surface membrane auto-antibodies. In total, 59 patients (56%) had no antibodies present in their CSF and PB, and were thus classified as antibody negative (AB−); 41 patients (38%) showed auto-antibodies against intracellular antigens (IAB+). Among the IAB+ patients, 17 (42%) were positive for GAD65, and 24 patients (58%) were positive for onconeural antibodies. It should be noted that the tumor rate is surprisingly low in onconeural antibody-positive patients, who are diagnosed in recent-onset unexplained epilepsies (3.2% in 93 patients) [18]. None of the patients included in this study had a tumor when screened with PET/CT, nor did any of the patients undergo anti-tumor treatment. The auto-antibodies with intracellular targets included those against the intracellular antigens Amphiphysin, CV2, GAD65, Ma2/Ta, Ri (ANNA2), SOX1, Yo (PCA1), and Zic4. Auto-antibodies against membrane surface antigens and synaptic proteins were identified in only six patients (6%) (SAB+). SAB+ included two patients with anti-LGI1, two patients with CASPR2, and two patients with anti-NMDAR autoantibodies.

Using the ‘Graus criteria’ for possible and definite autoimmune LE, the clinical features of the IAB+ group fulfilled the criteria for a definitive LE diagnosis. However, the cognitive impairments often extended beyond working memory, and MRI pathology was often unilateral. The AB− group remained in the possible LE category [14].

This study finally comprised 100 patients, 55% male and 45% female, with an average age of 41 (± 15) years. A majority (72%) had an education of ≥10 years. Their age of epilepsy onset was late with an average age of 34 years (± 16 years) and the average duration of the epilepsy of 7 years (± 8 years). Most patients (84%) were taking antiepileptic drugs (AEDs) at the time of evaluation; 54% of them were on polytherapy; and approximately 9% were taking antidepressants. MRI amygdala pathologies (43%) were not observed signficantly more often than hippocampal pathologies (about 37%), but 84% of the amygdala pathologies were classified as swelling, and only 16% were classified as atrophy or signal intensity. In contrast, the pathology of the hippocampus was more often an atrophy or signal intensity (59%) rather than swelling (41%). Not a single patient was seizure free. The average monthly seizure frequency was 13 with a considerable variance (SD 58). Overall, 42% of the patients displayed a depressed mood.

As Table 1 illustrates, the two study groups’ demographic and clinical characteristics were largely comparable. The IAB+ patients tended to be older and had later onset epilepsies when compared to the AB− patients.

The absolute amounts of HLADR+  CD4+ T-cells, HLADR+  CD8+ T-cells, and CD19+ B-cells in cells per milliliter (cells/ml) are reported, as well as the percentage of the circulating lymphocytes in both CSF and PB using flow cytometry (see Table 2). With the exception of the IAB+ group’s higher percentage of HLADR+ CD4 and CD8 T-cells in the PB, and a tendency of more HLADR+ CD4 in CSF, the results did not differ significantly between the patient groups. HLADR+  CD4+ T-cell, HLADR+  CD8+ T-cell, and CD19+ B-cell concentrations (cells/ml) in blood and CSF were moderately correlated with each other with correlation coefficients ranging between 0.42 (p < 0.001) and 0.58 (p < 0.001).

Analyses of the cognitive impairments in the patient subgroups revealed that between 34% (IAB+) and 40% (AB−) of the patients in the study groups displayed impaired verbal memory (total score), and between 43% (AB−) and 53% (IAB+) displayed impairments in visual memory (total score). Executive function was found to be impaired for 32% of both groups, mood was impaired for 46% of the patients in IAB+ , and for 42% in AB−. No significant group differences were observed on the level of the performance categories (% impaired). The groups did as well not differ when inference statistics were calculated (ANOVA results not specifically reported). Neither memory nor executive function or mood showed a relation to the monthly seizure frequency for the total group and individual subgroups (results not specifically reported).

When evaluating the performances as a function of the lateralization of the pathology (left/right/bilateral), the total group also failed to yield any significant difference. However, an important finding is that the verbal memory performance (more so than visual memory) differed depending on the MRI pathology. Verbal memory (but not visual memory) was significantly worse in the presence of a temporo-mesial atrophy (volume decrease) as compared to a volume increase (swelling) of temporo-mesial structures (F = 4.9 p < 0.05 verbal memory, F = 0.177 p < 0.10 for visual memory) (see Fig. 1). Unlike verbal memory, visual memory was in the impairment range (standard value < 90) with either type of pathology. As stated above, the swelling (i.e., volume increase) was diagnosed more than twice as often than atrophy in the IAB+ (45% vs. 18%) and AB− groups (37% vs. 15%). This finding is also partly reflected in the regression analyses which differentiated further as to whether or not certain pathologies were seen left or right. Accordingly, bilateral hippocampal pathology and left hippocampal atrophy had an impact on cognitive performance in verbal memory (recognition) and executive function in the IAB+ group (see Table 3).

Tables 3 and 4 summarize the hierarchical regression models to predict cognitive functions and mood based on the assumed predictors in LE patients in the two subgroups.

In the IAB+ subgroup, the percentage of circulating HLADR+  CD4+ T-cells in CSF was inversely correlated with verbal recognition (B = − 0.554, p < 0.05), verbal delayed free recall (B = − 0.419, p < 0.05), and verbal learning performance (B = − 0.404, p < 0.05), respectively. The percentage of circulating HLADR+  CD8+ T-cells in CSF negatively correlated with verbal recognition (B = − 0.424, p = 0.01) (see Fig. 2a)

The AB− patient subgroup´s regression showed that the absolute amount of CD19+ B-cells in CSF was negatively correlated with visual learning (B = − 0.303, p < 0.05) and visual learning capacity (B = − 0.315, p < 0.05). In this subgroup, the EpiTrack score representing attention and executive function correlated negatively with the percentage of circulating HLADR+  CD4+ T-cells in CSF (B = − 0.327, p < 0.05) and the absolute HLADR+  CD4+ T-cell loads in PB (B = − 0.284, p < 0.05) (see Fig. 2b) Visual memory functions did not correlate with either the absolute amount or the percentage of the circulating of HLADR+  CD4+ T-cells and HLADR+  CD8+ T-cells in either CSF or PB as well as the absolute CD19+ B-cells loads in PB (p > 0.05). Finally we noted a relation between a longer duration of epilepsy, treatment with zonisamide (ZNS) and a negative mood in this group (p < 0.01).

Regression analysis revealed no correlation between the cognitive functions and structural abnormality in the AB− patients. In those patients with auto-antibodies against intracellular antigens (IAB+) in both CSF and PB, bilateral hippocampal pathology was found to correlate negatively with verbal recognition (B = − 0.518, p < 0.001) and executive function (B = − 0.364, p < 0.001). Attention and executive functions also correlated with left hippocampus atrophy (B = −0.520, p < 0.001) and right amygdala hypertrophy (B = −0.236, p < 0.05).

As expected, the factors education (better performance with better education), gender (women displayed better verbal memory performance and worse mood), and drugs (poorer memory performance with higher drug load, poorer performance in executive function particularly with clobazam, poorer mood with zonisamide) revealed some effects on the dependent measures apart from clinical and immunological factors (see results of regression analyses in Tables 3, 4).

The following shortcomings of the present study should be mentioned. First, some findings are circular since significant predictors reflect our patient selection criteria. However, in this regard, they confirm our selection criterion. Second, our cohort is heterogeneous and the LE subgroups still comprise low patient numbers and thus limit our findings’ statistical power. Third, as this was a cross-sectional study, patients were screened at presumably very different time points during their disease course. Even when the assessments are synchronized and the patient is not under the influence of any autoimmune treatment, evaluations which are conducted only once reveal just one point in time within the suggested but unknown course of the underlying disease. The occurrence of seizures does not mark the onset of. Neuropsychiatric and neuropsychological symptoms often precede seizure onset [54, 57]. Longitudinal assessments would be ideal for discerning which disease parameters enable us to monitor this disease and assess its inflammatory acuity, and eventually, treatment success. Fourth, many other factors determine the cognitive performance and behavior in patients with epilepsy and must also be taken into consideration. It is evident (also within this evaluation) that epilepsy originating from suspected LE, differs from what has been evaluated in recent decades within the context of epilepsy surgery in chronic mesio-temporal TLE. We attempted to take this into consideration via hierarchical regression analysis which successively factored in demographic, clinical data, treatment, and finally flow cytometry results. Our results reveal several definitive signals, while also substantiating this issue’s enormous complexity. Fifth, in the present study, we focused on the subsets of T- and B-cell populations which previously appeared to have a relevant impact in limbic encephalitis [13]. In future studies, it would be worthwhile to investigate further stratified subsets of cells including “activated” B cells, in addition to all T-cell populations. Finally, it should be noted that without control data as a reference point, the results from the flow cytometry analyses are difficult to interpret with regard to the pathological status. This may in part relativize the understanding of the absolute values and the meaningfulness of some correlations as well.