Neuronal autoantibodies associated with cognitive impairment in melanoma patients

Abstract

Background: Cancer-related cognitive impairment is an important complication in cancer patients, yet the underlying mechanisms remain unknown. Over the last decade, the field of paraneoplastic neurological syndromes has been dramatically changed by the discovery of new neuronal autoantibodies, some of them associated with cognitive impairment. We aimed to assess the prevalence of neuronal autoantibodies in melanoma patients and their association with neurological and cognitive dysfunction.

Patients and methods: A total of 157 consecutive melanoma patients with a median age of 63 years were recruited at the Department of Dermatology, Charité-Universitätsmedizin Berlin and tested for neuronal autoantibodies. A comprehensive neuropsychological assessment was carried out in a selected subgroup of 84 patients after exclusion of patients with confounding factors for a cognitive dysfunction, including brain metastases, relevant medication, and neurological disorders.

Results: Neuronal autoantibodies were found in 22.3% of melanoma patients. The most frequent antibodies were IgA/IgM anti-NMDAR antibodies. Applying the International Cognition and Cancer Task Force criteria, 36.9% had cognitive impairment, however, with a threefold higher odds in antibody-positive compared with antibody-negative patients (57.1% versus 30.2%, OR = 3.1, 95% CI: 1.1 to 8.6; P = 0.037). In patients with anti-NMDAR antibodies, this impairment increased with higher antibody titers (P = 0.007). Antibody-positive patients had a significantly impaired overall cognitive performance (z-value: -0.38 ± 0.69 versus 0.00 ± 0.56; P = 0.014) as well as significant impairments in tests of memory, attention, and executive function. In a multiple linear regression analysis, autoantibodies were an independent risk factor for cognitive impairment (B = -0.282; 95% CI: -0.492 to -0.071; P = 0.009). Autoantibody seropositivity was associated with immune checkpoint inhibitor treatment and a history of autoimmune diseases.

Conclusions: A large number of melanoma patients harbor neuronal autoantibodies that are associated with significant cognitive impairment affecting memory, attention, and executive function. Neuronal autoantibodies might represent a pathophysiological factor and possible biomarker in the development of cancer-related cognitive impairment.

Keywords: cancer-related cognitive impairment; immune checkpoint inhibitor; melanoma; neuronal autoantibodies; paraneoplastic neurological syndromes.

Figure 1.

(A) Ab+ patients showed significantly more often a cognitive impairment compared with ab− patients [57.1% (ab+) versus 30.2% (ab−), OR = 3.1 (95% CI: 1.1 to 8.6), Fisher’s exact test: P = 0.037]. Cognitive impairment was considered when a patient had ≥2 deficits whereas a deficit was a cognitive performance 1.5 SDs below that of the normative controls of the respective test systems. (B) Ab+ patients had significantly more deficits in cognitive subtests than ab− patients (2.2 versus 1.4 deficits, t-test = −2.04, P = 0.045). (C) In patients with low titers (1 : 10), the percentage of patients with cognitive impairment was comparable to that of patients without antibodies (28.6% and 30.2%), whereas the prevalence of cognitive impairment increased to 66.7% (titer 1 : 32) and 100% (titer 1 : 100) in patients with higher titers (titers ≤ 1 : 10 versus ≥1 : 32; P = 0.007) suggesting a NMDAR ab titer-dependent decline in cognitive function. Error bars: +1 SEM, *P < 0.05; ab, antibody; NMDAR, anti-NMDA receptor.

Figure 2.

(A) Compared with ab− patients, ab+ patients achieved significantly less points in the immediate recall of the ROCF (19.0 ± 5.5 versus 22.4 ± 6.1, d = 0.56, t = 2.3; P = 0.024) which indicates reduced visuospatial memory. The same applies for the subgroup of the NMDAR ab+ patients (19.0 ± 4.9 versus 22.4 ± 6.1, t = 2.1, d = 0.56, P = 0.044). (B) Decrease in working memory of ab+ patients is depicted as reduced points in the task to reversely recall digit spans compared with ab− patients (6.1 ± 1.3 versus 6.9 ± 1.9, d = 0.42, t = 2.23; P = 0.031). (C) In the phasic alertness trial, NMDAR ab+ patient’s reaction time was significantly prolonged compared with ab− patients (320.3 ms ± 93.5 versus 283.4 ms ± 47.5, d = 0.78, t = 2.2; P = 0.032). (D). Both the ab+ group and the NMDAR ab+ subgroup showed a prolonged reaction time in the visual task of the divided attention task (ab+ versus ab−: 921.5 ms ± 137.9 versus 840.6 ms ± 125.0, t = −2.5, d = 0.65; P = 0.014); for NMDAR ab+ versus ab−: 926.9 ms ± 145.1 versus 840.6 ms ± 125.0, d = 0.69, t = −2.3; P = 0.022). (E) In comparison to ab− patients NMDAR ab+ patient’s reaction time for an adequate response in the Go Nogo task was significantly prolonged which demonstrates impaired executive function (654.5 ms ± 100.3 versus 589.3 ms ± 67.1, d = 0.97, t = −2.7; P = 0.008). Error bars: +1 SEM, *P < 0.05; ab, antibody; NMDAR, anti-NMDA receptor; ROCF, Rey Osterrieth Complex Figure.

Figure 3.

(A and B) Group comparison between ab− and ab+ patient’s neuropsychological results and between the NMDAR ab+ subgroup and the ab− patients. Z-transformation of neuropsychological raw data with ab− patients as reference group: from all test results we subtracted ab− patient’s group mean and divided by ab− patient’s SD to set ab− patient’s mean test results to 0 and their SD to 1. Test scales with lower values representing a better performance were multiplied by −1 so that lower values always stand for lower performance. The composite cognitive score was calculated by averaging the obtained z-values from the 22 subtests for every patient. Significant group differences between cognitive subtests are marked with an asterisk (*P < 0.05). Ab+ patients systematically scored worse in almost all subtests and also significantly in the composite cognitive score (z-values: −0.38 ± 0.69 versus 0.00 ± 0.56; t = 2.5, d = 0.68; P = 0.014) (A). In the NMDAR ab+ subgroup, patients also received a significantly lower composite cognitive score than ab− patients (z-values: −0.40 ± 0.71 versus 0.00 ± 0.56; t = 2.3, d = 0.71; P = 0.023) (B). Error bars: 95% confidence intervals; ab, antibody; LPS, Leistungsprüfsystem; NMDAR, anti-NMDA receptor; ROCF, Rey Osterrieth Complex Figure.