VGKC-antibodies define neurological conditions that are usually immunotherapy-responsive. As a result, these antibodies have become part of the investigation of patients with unexplained subacute onset of epilepsy, memory or cognitive problems, or peripheral nerve hyperexcitability syndromes, but it has been unclear how the antibodies could cause such a range of different clinical presentations. Here, we show that the majority of VGKC-antibodies of high titre are not directed towards the Kv1 subunits themselves but to two proteins, Lgi1 and Caspr2, that are closely associated with VGKCs in brain tissue and remain complexed with the VGKCs in 2% digitonin extracts. Lgi1 antibodies were found almost exclusively in patients with limbic encephalitis or epilepsy, all without tumours, whereas antibodies to Caspr2 were found in patients with limbic encephalitis, Morvan’s syndrome or neuromyotonia, often with thymomas. It is interesting that mutations in genes encoding both these proteins are found in hereditary epilepsy and other disorders (reviewed by Morante-Redolat et al.
; Kumar and Christian, 2009
), reflecting the fact that genetic and autoimmune conditions often target the same proteins.
An important aim of this work was to study the clinical phenotypes of the patients in parallel with their antibody specificities. We studied 96 patients with high titres of VGKC-antibodies, representative of those referred to us for testing. The clinical follow-ups confirm earlier observations in much smaller studies (Thieben et al.
; Vincent et al.
; Graus et al.
) that many VGKC-antibody positive patients (with titres >400 pM) have CNS disease, that tumours are rare, and that most respond well to immunotherapies. The majority of the patients (67%) had limbic encephalitis with typical clinical presentations; the presence of VGKC-antibodies confirmed this diagnosis in the 41% whose MRIs were normal (Bien and Elger, 2007
). Strikingly, there were no detectable active tumours in the Lgi1-antibody positive patients despite a median follow-up of over 3 years. Only five patients were identified as having Morvan’s syndrome, illustrating the rarity of this syndrome; their condition was characterized by pain, often burning in nature, as well as the typical features of neuromyotonia, autonomic disturbance and insomnia. Eleven patients were diagnosed by their neurologists as having neuromyotonia only, but four of these had pain, autonomic dysfunction or insomnia, features that could be ascribed to Morvan’s syndrome, suggesting considerable overlap between the two syndromes.
Previous immunohistological data on a small number of high titre VGKC-antibody positive limbic encephalitis and Morvan’s syndrome sera suggested co-localization with Kv1.1 or 1.2, or occasionally Kv1.6 (Buckley et al.
; Liguori et al.
; Ances et al.
; Antozzi et al.
; Kleopa et al.
). Having identified the true targets for the antibodies in these patients, it appears that these results were confounded by the very similar localization of Lgi1 and Kv1.1, particularly in the hippocampus, and of Caspr2 and Kv1.2 (Kv1.2 not shown here but can be seen in Kleopa et al.
), both in the hippocampus and cerebellum. Moreover, the precipitation of 125
I-α-DTX-VGKCs with antibodies to Lgi1 that we show here, confirms earlier experimental data that the two are closely associated (Schulte et al.
). Whereas Caspr2 is strongly associated with Kv1.1 and Kv1.2 at juxtaparanodes (Poliak et al.
), Lgi1 expression in peripheral nerves is weak but present (Ogawa et al.
, 2010; KA Kleopa, unpublished data).
It is important to recognize that the antibodies to Lgi1 and Caspr2 were first measured by binding to the surface of unpermeabilized cells, and therefore the antibodies are directed towards ‘cell-surface antigens’ on neurons, defining their pathogenic potential (Vincent et al.
; Graus et al.
), which was supported by the antibody binding to live hippocampal cultures. At present we do not have an explanation for the 19% of patients in whom we could not identify any VGKC-complex antigen; their lower VGKC-antibody titres (Supplementary Figure 3
) suggests that improvements in our antibody assays may reduce the number of sera with undefined target antigens. Conversely, it is highly likely that some patients with limbic encephalitis or Morvan’s syndrome, previously negative for VGKC antibodies, will prove to have Lgi1 or Caspr2 antibodies. Further work is clearly needed.
Lgi1 is an appropriate target for antibodies in limbic encephalitis and epilepsy, and Lgi1 antibodies have recently been confirmed by others (Lai et al.
). Mutations in Lgi1 are associated with a distinct form of temporal lobe epilepsy without peripheral nerve dysfunction (Morante-Redolat et al.
). It is notable that 91% of the Lgi1-antibody positive patients had limbic encephalitis, consistent with Lgi1’s hippocampal localization, and only one had peripheral symptoms. When we looked for Lgi1 antibodies in the sera, we were initially confused because the patients’ serum IgG bound to all of the Lgi1-transfected human embryonic kidney cells, rather than just to those that were transfected. This was explained by the fact that Lgi1 is secreted from neurons and can then bind to cell surfaces, as shown in neuroblastoma and COS7 cells (Sirerol-Piquer et al.
) and now demonstrated in HEK293 cells. Lgi1 binds ADAM22 on CNS postsynaptic membranes (Fukata et al.
; Ogawa et al.
, 2010) but we could not detect ADAM22 in our human embryonic kidney cell cultures (data not shown); further Lgi1 receptors are being identified and might be responsible for our findings (Sagane et al.
, 2008). Since, we were unable to use EGFP-tagged Lgi1 in immunoprecipitation experiments (data not shown), perhaps because the EGFP-tag interfered with its conformation, to confirm the target of the VGKC-antibodies, we expressed Lgi1 fused to the C-terminal and transmembrane domain of Caspr2; this has provided a more direct and sensitive method for detection of patients’ antibodies and will now be developed for routine use. An EGFP construct will also be made so that the serum antibodies can be quantified by immunoprecipitation.
Caspr2 is a more conventional antigenic target. It is a membrane protein with a large extracellular sequence consisting of multiple well-defined domains (Poliak et al.
), and is essential for the co-localization of Kv1.1 and 1.2 at the juxtaparanodes of the nodes of Ranvier. Caspr2-knockout mice show dispersed, non-clustered Kv1 neural expression (Poliak et al.
), similar to that seen in one study of patients with Caspr2-mutations (Strauss et al.
). It is likely that binding of Caspr2 antibodies results in down-regulation of Caspr2/Kv1.1/1.2 complexes on the peripheral nerve axon, leading to neuromyotonia, neuropathic pain and autonomic dysfunction. How these antibodies cause insomnia, one of the defining features of Morvan’s syndrome, is not clear but mutations in CNTNAP2
(which encodes Caspr2) are associated with epilepsy and cognitive decline as well as peripheral nerve involvement (Strauss et al.
), and have also been identified in various forms of schizophrenia, epilepsy and autism spectrum disorders (reviewed in Kumar and Christian, 2009
). Although in this cohort, only three of the five patients with Morvan’s syndrome had Caspr2-antibodies, in a further nine patients with Morvan’s syndrome and VGKC-antibodies, the majority had Caspr2-antibodies and six had thymomas (Vincent, 2009 and unpublished observations); it will be interesting to look for Caspr2 expression in these tumours.
Earlier studies suggested that the Kv1s were the target antigens in neuromyotonia (Arimura et al.
; Hart et al.
), but most of those sera had lower VGKC-antibody titres compared with those studied here. The targets for sera with lower VGKC antibodies should now be re-explored, but from our initial findings (unpublished observations) we anticipate that Kv1s, Caspr2 and contactin-2 will individually, or in combinations, prove to be targets in many patients with VGKC-antibody positive neuromyotonia, and also may be positive in some that are currently negative for VGKC-antibodies.
In the last few years, antibodies to ion channels and a growing number of receptors have been identified in patients with acute or subacute onset of CNS syndromes (Dalmau and Rosenfeld 2008
; Vincent et al.
; Graus et al.
). Our results help explain the diversity of clinical syndromes associated with VGKC-antibodies and suggest that proteins complexed with these receptors and ion channels may also prove to be targets for autoantibodies in patients with these and other autoimmune CNS disorders.