Correspondence

www.thelancet.com/neurology Vol 9 April 2010 343

Complement-mediated cytotoxicity of antibodies to the GABAB receptor

We read with great interest the paper by Lancaster and colleagues,1 which describes a novel autoimmune limbic encephalitis associated with antibodies to the GABAB receptor. In all six tested patients the anti-GABAB antibodies were IgG1 and thus able to activate complement. Nevertheless, in their discussion, the authors consider the role of complement-mediated cytotoxicity in the pathogenesis of this disorder, which was sensitive to immunotherapy in most cases, as questionable. We have described a woman with recurrent acute episodes of respiratory crises, dysautonomic symptoms, and total sleep loss (agrypnia) due to an antibody directed against the GABAB receptor,2 who showed a good clinical response to immunotherapy.2,3 Similar to patients described by Lancaster and colleagues,1 our patient’s purifi ed IgG recognised a band of about 110 kDa on protein extracts of mouse cerebellum, cortex, and brainstem, and immunolabelled cultured Chinese hamster ovary (CHO) cells transfected with human GABAB1 receptors and rat GABAB2 receptors. Western blot analysis of transfected CHO homogenates showed the same band using both purifi ed IgG and anti-GABAB1 antibodies from the patient. In our case, in mice the human purifi ed IgG was able to induce a reversible neurological syndrome similar in all respects to that observed in the patient. Moreover, immunohistochemistry on brain sections of mice injected with the patient’s IgG showed the simultaneous presence of bound human IgG and C5b-9 deposits on neurons, suggesting a complement-mediated cytotoxicity. In our opinion, the clinical response to treatment does not rule out the role of antineural

complement-fi xing autoantibodies in the pathogenesis of autoimmune encephalitis.We have no confl icts of interest.

Anna Paola Batocchi, Giacomo Della Marca, Giovanni Frisulloannapaola.batocchi@rm.unicatt.it

Department of Neurosciences, Catholic University, Rome, Italy

1 Lancaster E, Lai M, Peng X, et al. Antibodies to the GABAB receptor in limbic encephalitis with seizures: case series and characterisation of the antigen. Lancet Neurol 2010; 9: 67–76.

2 Batocchi AP, Della Marca G, Mirabella M, et al. Relapsing-remitting autoimmune agrypnia. Ann Neurol 2001; 50: 668–71.

3 Frisullo G, Della Marca G, Mirabella M, et al. A human anti-neuronal autoantibody against GABA B receptor induces experimental autoimmune agrypnia. Exp Neurol 2007; 204: 808–18.

Authors’ replyWe thank Batocchi and colleagues for their comments; however, they seem to have misunderstood us:1 the antibodies of our patients did not recognise a 110 kDa band in denaturing immunoblot. Patients’ antibodies precipitated the B1/B2 subunits of the GABAB receptor, which was confi rmed with immunoblots of the immunoprecipitate by use of commercially available specifi c antibodies, indicating that the target epitope is present in its functional conformation. Additional analyses with cultured neurons showed that the epitope was on the cell surface and that patients had no other antibodies to cell-surface antigens. Our patients’ phenotype resembled that of animals with disrupted GABAB receptors. Therefore, the disorder they had was not only clinically but also immunologically diff erent from that of the patient described by Batocchi and colleagues.2,3 The authors indicate that their patient’s symptoms resulted from antibodies to an intracellular GABAB receptor epitope that competed with a commercial antibody, via complement activation. However, controls to identify the antibody specifi city and studies to rule out other antibodies were missing.

For example, reactivity of patient’s IgG with non-transfected cells was not assessed; normal IgG showed some reactivity with transfected Chinese hamster ovary cells, and competition experiments between normal IgG and the commercial antibody were not done.3 Moreover, to model the disorder the authors injected patient’s or control IgG into the cisterna magna of mice at concentrations ranging from 800 to 4000 times the normal levels of IgG in human CSF. Data from these mice showed deposits of IgG mainly inside Purkinje cells—a pattern diff erent from that of their patient’s antibody reactivity, which, according to the authors, spared Purkinje cells2,3—and deposits of complement that did not colocalise with the patient’s IgG (fi gure 5 in the paper by Frisullo and colleagues3). These fi ndings, along with the well-known property of Purkinje cells to take up proteins from CSF, raises concern about the specifi city of the IgG and complement deposits. Furthermore, the symptoms of the mice3 only partially resembled the patient’s syndrome.2 The rapid recovery of the mice is also puzzling in that Purkinje cells are unlikely to recover from supposed complement-mediated cytotoxicity and complement activation is unlikely to occur after intracellular penetration of an antibody. More consistent with the reversibility of their patient’s symptoms is that other (undefi ned) antibodies to cell-surface antigens were involved; these would not activate complement because the data of Batocchi and colleagues2 did not show cell-surface linear deposition of complement. As we suggested,1 the occurrence of antibodies that can fi x complement does not demonstrate that this mechanism is operational.JD has received royalties from a patent related to Ma2 autoantibody test and has fi led patent applications for NMDA and GABAB receptor autoantibody tests. JD has received funding from Euroimmun for projects unrelated to the study referred to in these letters. All other authors have no confl icts of interest.

Correspondence

344 www.thelancet.com/neurology Vol 9 April 2010

Medical Centre, Nijmegen Centre for Evidence Based Practice and Donders Centre for Neuroscience, Department of Rehabilitation, Nijmegen, Netherlands (ACHG)

1 Klit H, Finnerup NB, Jensen TS. Central post-stroke pain: clinical characteristics, pathophysiology, and management. Lancet Neurol 2009; 8: 857–68.

2 Lindgren I, Jönsson AC, Norrving B, Lindgren A. Shoulder pain after stroke: a prospective population-based study. Stroke 2007; 38: 343–48.

3 Roosink M, Renzenbrink GJ, Buitenweg JR, et al. Central neuropathic mechanisms in post-stroke shoulder pain. Eur J Pain 2009; 13: S44.

Eric Lancaster, Meizan Lai, Josep Dalmaujosep.dalmau@uphs.upenn.edu

Department of Neurology, University of Pennsylvania, School of Medicine, Philadelphia, PA, USA

1 Lancaster E, Lai M, Peng X, et al. Antibodies to the GABAB receptor in limbic encephalitis with seizures: case series and characterisation of the antigen. Lancet Neurol 2010; 9: 67–76.

2 Batocchi AP, Della Marca G, Mirabella M, et al. Relapsing-remitting autoimmune agrypnia. Ann Neurol 2001; 50: 668–71.

3 Frisullo G, Della Marca G, Mirabella M, et al. A human anti-neuronal autoantibody against GABA B receptor induces experimental autoimmune agrypnia. Exp Neurol 2007; 204: 808–18.

Defi ning post-stroke pain: diagnostic challenges

Recently, a new grading system for central post-stroke pain (CPSP) was proposed, which might be used to distinguish patients with stroke who have central neuropathic pain from patients who have peripheral pain.1 Accordingly, for a CPSP diagnosis, all other causes of pain have to be excluded. Although this criterion has its purpose for defi ning CPSP as a separate entity, a too rigorous distinction between central and peripheral post-stroke pain might have drawbacks as well. Most importantly, by strictly following the proposed grading system, central pain mechanisms could be missed or even disputed in patients with other types of post-stroke pain. This possibility is particularly relevant as “mixed” pain and pre-existing pain are common after stroke.1 For this reason, we would like to emphasise that peripheral nociceptive pain after stroke might coincide with symptoms characteristic of CPSP. To lend support to our concern, we present recent data on post-stroke shoulder pain (PSSP).

PSSP is commonly localised to the aff ected upper extremity and regarded as peripheral nociceptive

pain. However, unsatisfactory treat-ment and the frequent occurr ence of persistent pain2 suggest a role for other mechanisms. To try to understand the possible central mechanisms that underlie PSSP, we used some parts of the diagnostic assessment for neuropathic pain in 19 patients with chronic PSSP, none of whom could be classifi ed as having CPSP.3 Several sensory abnormalities overlapped with those observed in CPSP. Of particular interest was the high prevalence of abnormal spinothalamocortical tract function in patients with PSSP (15 of 19) compared with pain-free stroke patients (13 of 29), as abnormal function of this tract has been implicated in CPSP. Moreover, supportive criteria for a CPSP diagnosis, such as touch or cold allodynia (four of 19) and the absence of a primary relation with movement (seven of 19), were common in patients with PSSP, and PSSP was associated with abnormal sensory function in the unaff ected side.

Our data strongly suggest that central pain mechanisms have an essential role in post-stroke pain, even in patients who cannot be classifi ed as having CPSP. Therefore, central pain mechanisms should be assessed in all patients with post-stroke pain and treatments used for patients with CPSP might also be appropriate for patients with other forms of post-stroke pain. We hope that the CPSP grading system1 will not prevent clinicians and researchers in the fi elds of neurology, rehabilitation, and pain medicine from regarding and treating central pain mechanisms in patients with post-stroke pain who do not fulfi l the criteria for CPSP. We have no confl icts of interest.

Meyke Roosink, Alexander CH Geurts, Maarten J Ijzermanm.roosink@utwente.nl Biomedical Signals and Systems (MR) and Health Technology and Services Research (MJI), MIRA Institute for Biomedical Technology and Technical Medicine, University of Twente, Enschede, Netherlands; and Radboud University Nijmegen

Authors’ replyIn their thoughtful letter, Roosink and colleagues raise the point that central pain mechanisms might be overlooked in patients with stroke who have pain if our proposed defi nition for central post-stroke pain (CPSP) is used. The essential point here is how we defi ne central pain mechanisms. It is important to distinguish between central neuropathic pain and central mechanisms. When the nociceptive system is activated, physiologically short-lasting neuroplastic changes occur in the CNS. In persistent pain disorders, the molecular and cellular changes are more profound and sometimes irreversible, whether due to infl ammation or a lesion of the nervous system. In neuropathic pain, there is damage to the somatosensory systems, causing peripheral and central neuroplastic changes that can sometimes be permanent. In infl ammatory or simple nociceptive pain disorders, the somatosensory system is essentially intact, but it is in a state of heightened excitability that gradually returns to normal when the infl ammation subsides.

Specifi c sensory testing could be used to clarify whether there is a loss of sensory input to the nervous system, but such testing can be misleading. The abnormalities mentioned by Roosink and colleagues in assumed spinothalamic functions, such as temperature and pinprick response, are not necessarily an indication of central neuropathic pain. Central neuropathic pain requires a loss of