Box 1.

Antibodies as biomarkers: current problems and future directions

Current problems	Possible solutions
Antibodies as diagnostic biomarkers are not entirely specific
Rarely, low positive antibody titres can occur where the primary aetiology is not autoimmune. For example, GlyR antibodies were detected in Creutzfeldt-Jakob disease or genetic dystonia (Angus-Leppan et al., 2013; Carvajal-Gonzalez et al., 2014); NMDAR antibodies in serum or CSF of patients with Creutzfeld-Jakob disease (Fujita et al., 2012; Mackay et al., 2012); or MELAS syndrome (Finke et al., 2012); and GABAAR-antibodies in genetically proven Huntington’s disease (Pettingill et al., 2015). Similarly, neuronal antibodies without any clinical correlate have been found in healthy controls (Meinck et al., 2001; Dahm et al., 2014).	These findings highlight that antibody test results need to be interpreted with caution and clinical judgement.
	Methodological issues of antibody testing might be overcome by standardized tests and by international multicentre trials to establish the assays with the highest sensitivity and specificity.
	Diagnostic specificity can be increased e.g. by taking antibody titres into consideration, and by testing serum and CSF, and calculating intrathecal synthesis (particularly for GAD antibodies).
	It remains to be investigated if these antibodies could exert pathogenic effects in addition to the primary pathology.
The controversial role of IgA and IgM antibodies
Pathogenic relevance was hitherto assigned only to antibodies of IgG subclass. NMDAR-antibodies of IgA-subtype were detected in patients with slow cognitive impairment in absence of inflammatory signs in MRI or CSF. Some patients responded to immunotherapy (Pruss et al., 2012b).	There is emerging evidence of downregulation of NMDA receptors also by IgA and IgM antibodies in neuronal cell cultures (Pruss et al., 2012a, b). However, their role remains controversial (Lancaster et al., 2015) and further studies are warranted to determine if such patients should receive immunomodifying treatment.
A need for predictive biomarkers to better guide therapeutic decisions
Antibody titres and clinical course correlate only in some (typically neuronal surface) antibodies. The correlation with serum antibody titres will possibly be poorer in diseases with predominant intrathecal synthesis (e.g. NMDAR, DPPX antibodies) than in disorders where the antibody is mainly generated in the serum (e.g. LGI1 antibodies).	In NMDAR-antibody encephalitis, the B-cell-attracting chemokine CXCL13 correlated with treatment responses and relapses (Leypoldt et al., 2015).
	Another avenue to explore is FDG-PET imaging, which can show abnormalities even when the MRI is normal, and which often correspond to the clinical course (Kunze et al., 2014).
Commonly used treatment approaches are mainly empirical or based on expert opinions
There is a lack of evidence-based guidelines, mainly due to the relative rarity of these diseases.	Joint forces like international multicentre studies and registries with closely characterized patients would be desirable to investigate systematically the best treatment rationale.

MELAS = mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes.