|Home | About | Journals | Submit | Contact Us | Français|
We report four young women who developed acute psychiatric symptoms, seizures, memory deficits, decreased level of consciousness, and central hypoventilation associated with ovarian teratoma (OT) and cerebrospinal fluid (CSF) inflammatory abnormalities. Three patients recovered with treatment of the tumor or immunosuppression and one died of the disorder. Five other OT patients with a similar syndrome and response to treatment have been reported. Our patients' serum or CSF showed immunolabeling of antigens that were expressed at the cytoplasmic membrane of hippocampal neurons and processes and readily accessed by antibodies in live neurons. Immunoprobing of a hippocampal-expression library resulted in the isolation of EFA6A, a protein that interacts with a member of the two-pore-domain potassium channel family and is involved in the regulation of the dendritic development of hippocampal neurons. EFA6A-purified antibodies reproduced the hippocampal immunolabeling of all patients' antibodies and colocalized with them at the plasma membrane. These findings indicate that in a young woman with acute psychiatric symptoms, seizures, and central hypoventilation, a paraneoplastic immune-mediated syndrome should be considered. Recognition of this disorder is important because despite the severity of the symptoms, patients usually recover. The location and function of the isolated antigen suggest that the disorder is directly mediated by antibodies.
Paraneoplastic limbic encephalitis (LE) often associates with brainstem dysfunction and predominantly affects older individuals with lung cancer.1 A review of 137 patients with LE showed that young individuals with germ-cell tumors of the testis or ovarian teratoma (OT) were more frequently affected than patients of any age with more prevalent tumors such as cancer of the breast, prostate, or colon.2 Subsequent studies defined the LE associated with germ-cell tumors as a syndrome with dominant limbic, diencephalic and upper brainstem dysfunction, and the Ma proteins as the main autoantigens.3 Because germ-cell tumors often contain teratoma elements, we reasoned that a similar disorder may occur in women with OT. This led us to investigate the neurological and immunological features of four women with OT and encephalitis examined by us, and to review the clinical features of similar cases in the literature.4-12 None of our four patients had antibodies to Ma proteins, but we were impressed by the similarity and severity of the neurological symptoms, which often resembled an acute psychotic episode, malingering, or drug abuse. These patients often had cerebrospinal fluid (CSF) inflammatory abnormalities and the neurological syndrome improved after tumor resection, immunotherapy, or both. On the basis of these observations, we postulated that teratoma-associated encephalitis is an immune-mediated disorder and that if antibodies are involved they are not detected by conventional testing. We report the clinical features of this disorder along with the associated antibodies and preliminary characterization of one of the antigens.
Four patients were examined by the authors, and sera or CSF was obtained at symptom presentation (three cases) or recurrence (one case) and kept frozen at −80°C until use. A brief description of Patient 1 has been reported previously (Case 4 in Ances and colleagues12); this patient and Patient 2 are fully reported here (see online supplementary information). The clinical features of Patients 3 and 4 have been previously reported.9-11
Rats were anesthetized and euthanized by decapitation and the brain removed and processed as reported.12 Frozen 7μm-thick sections were directly mounted on slides and the patients' sera (diluted 1:250) or CSF (1:10) were tested using the avidin-biotin-peroxidase technique.12 To determine whether patients' antibodies targeted the same epitopes, we used immunocompetition assays between IgG biotinylated from one patient's serum and whole serum of other patients.13
Rat hippocampal neuronal cultures were prepared as reported.14 Neurons were grown on coverslides, fixed with paraformaldehyde (PFA), and serially incubated with patients' sera (1:250) for 1 hour and fluorescein-labeled goat anti–human IgG for 30 minutes. After washing, slides were incubated with biotinylated IgG from control patients with voltage-gated potassium channels (VGKCs) or normal individuals or mouse antibodies to the VGKC Kv1.2 (1:50; Upstate Biotechnology, Lake Placid, NY) or ARF6 (1:25; Chemicon International, Temecula, CA). The reactivity of biotinylated human IgG was developed with avidin-rhodamine (1:2000; Vector, Burlingame, CA) and the reactivity of mouse antibodies was developed with Alexa Fluor rhodamine-labeled goat anti–mouse IgG (1:2000; Molecular Probes, Eugene, OR).
To determine whether the target antigens were accessible in live neurons, we added patients' antibodies to the neuronal cell cultures for 30 minutes. Afterward, the media was removed and the neurons were gently washed in phosphate-buffered saline and fixed in PFA, and the bound antibodies were revealed with Alexa Fluor fluorescein-labeled anti–human IgG (1:2,000). Live neurons similarly incubated with normal human IgG and anti–Hu IgG served as controls. All human IgGs were used at concentration 0.37μg/ml obtained from patients' CSF.
A uni-ZAP-XR rat hippocampal library (Stratagene, La Jolla, CA) was probed with sera from Patients 1, 2, and 3, as reported.15 Positive clones were purified with several rounds of antibody screening and subcloned into pBluescript using the in vivo phage rescue protocol (Stratagene) and sequenced. Filters with purified relevant or irrelevant phage plaques were incubated with patients' or control sera for 12 hours at 4°C. After acid elution and neutralization, antibodies were concentrated with a column of protein A–Sepharose and used in immunohistochemical studies as indicated above.
The four patients had in common the development of acute behavioral changes and prominent psychiatric symptoms that were followed by rapid neurological deterioration, seizures (in three cases), decrease of level of consciousness or hypersomnia, and central hypoventilation, requiring admission to an intensive care unit and extended ventilatory support. This neurological syndrome is similar in many respects to that of five previously reported patients with OT (Table 1). When considered together, the clinical picture of these nine patients was dominated by severe behavioral and personality changes (eight patients), short-term memory deficits (six at presentation, and two noted after recovering from seizures or decreased level of consciousness), central hypoventilation (five), seizures (five), and hallucinations (three). The neurological disorder preceded the tumor diagnosis in seven patients (median, 1 month; range, 1–18 months) and developed shortly after the tumor diagnosis in two patients. The associated OT was benign or mature in four patients (one diagnosed radiologically as a dermoid cyst), immature in four, and mixed in one.
In six of nine patients, brain magnetic resonance imaging (MRI) was abnormal and in three normal (Fig 1, Table 2). Abnormalities usually were seen with fluid-attenuated inversion recovery (FLAIR) sequences and involved several brain regions (three patients), brainstem (two), cerebellum (one), pituitary gland and infundibulum (one), and spinal cord (one). Brain 18Ffluorodeoxyglucose (FDG) positron emission tomography (PET) or single-photon emission CT (SPECT) obtained in four patients demonstrated abnormalities not seen with MRI in temporal lobes (three patients), basal ganglia (one patient), and brainstem (one patient; see Fig 1). The CSF was abnormal in seven patients: four with pleocytosis and increased protein concentration, two with pleocytosis, one with increased protein concentration. Electroencephalogram was abnormal in all eight patients examined: seven had diffuse or focal slow wave activity, and one had bilateral temporalparietal epileptic activity (Patient 3).
All patients except Patient 1 had surgery and tumor resection; four patients received corticosteroids, two received intravenous immunoglobulins (IVIg), one received plasma exchange, and one received chemotherapy (see Table 2). Eight patients had neurological improvement (four full recovery) and one died as a result of the neurological disorder. In six patients with available information, the median time from development of the main neurological syndrome until initial evidence of improvement was 12 weeks (range, 7–16 weeks), and for those requiring ventilatory support the median time on a ventilator was 49 days (range, 30–90 days).
The sera of the four patients showed identical reactivity with the neuropil of hippocampus (Fig 2). The CSF available from two patients showed more intense and hippocampal-specific reactivity than the sera; minimal or barely visible reactivity was seen in other areas of the brain or cerebellum. In the hippocampus, the most intense immunolabeling localized in the inner aspect of the molecular layer adjacent to the granular neurons of the dentate gyrus, sparing the cell bodies (Figs (Figs22 and and3A).3A). The stratum lacunosum moleculare of the cornu ammonis 1 region of the hippocampus adjacent to the hippocampal fissure was spared by the antibody reactivity of the four patients.
Because some patients with LE have antibodies to VGKCs,16 we compared the reactivity of our patients' antibodies with that from patients with VGKC antibodies. This demonstrated clear differences; sera of patients with VGKC antibodies predominantly react with the outer aspect of the molecular layer of the hippocampus without sparing the stratum lacunosum moleculare, and also react with the molecular layer of the cerebellum (see Figs Figs22 and 3A-D). Double immunolabeling with human VGKC antibodies or Kv1.2 antibodies in neuronal cultures did not show colocalization with the teratoma-associated encephalitis antibodies (see Fig 3E).
The appearance of cell surface immunolabeling lead us to investigate whether the antigens of teratoma-associated encephalitis were readily accessed by antibodies in live neurons. For these studies, live neuronal cultures were exposed for 30 minutes to antibodies added to the media followed by extensive washing and then fixation in PFA. These studies showed intense immunolabeling of the neuronal cell membrane by patients' antibodies, but not by normal IgG or antibodies to intracellular antigens (Hu), strongly suggesting that the autoantigens are exposed on the cell surface (Fig 4).
Immunocompetition assays between patients' antibodies showed that they partially blocked the hippocampal reactivity of each other, indicating that some, but not all epitopes are common targets (data not shown).
Initial serum antibody titers were 1 to 2,000 for Patient 3, 1 to 1,000 for Patients 1 and 2, and 1 to 100 for Patient 4. Follow-up titers, after neurological improvement, were obtained in Patients 1, 2, and 4; no antibody reactivity was detected in any of the patients.
To further determine the target antigens of teratoma-associated encephalitis antibodies, we probed a rat hippocampal expression library combining sera of three patients (Patients 1–3). This resulted in the isolation of two overlapping clones. A homology search of Gen-Bank databases demonstrated that the consensus sequence formed by the two clones corresponded to a 330–amino acid portion of EFA6A (exchange factor for ARF6) and included part of the Sec7 domain, the entirety of the PH (pleckstrin homology) domain, and the coiled-coil motifs.17,18
When tested individually against EFA6A-expressing phage plagues, one of the four patient's sera reacted with EFA6A (Fig 5A). We then used these phage plaques to purify EFA6A-specific antibodies from the patient's serum, and the reactivity of these antibodies was compared with that of the other three patients' sera or CSF. These studies demonstrated that affinity-purified EFA6A antibodies had an identical pattern of hippocampal reactivity to that of the other three patients' antibodies (see Fig 5B). Furthermore, in a double immunolabeling assay with hippocampal neurons, the reactivity of all patients' antibodies colocalized with EFA6A (see Fig 5C), indicating that the patients' antibodies targeted several epitopes contained in EFA6A or proteins interacting with EFA6A at the plasma membrane. Among these interacting proteins, we explored whether ARF6 (ADP-ribosylation factor 6) was a target antigen; no significant colocalization of reactivities was identified between patients' antibodies and ARF6 antibodies (the latter mainly identifying intracytoplasmic ARF6; data not shown).
In addition to the hippocampal immunolabeling, highly concentrated EFA6A affinity-purified antibodies also reacted with the neuropil and cytoplasm of neurons of the cerebral cortex, brainstem, and cerebellum, indicating that the expression of EFA6A predominates but is not restricted to the hippocampus (data not shown).
We report the clinical and neuroimaging features of a disorder that occurs in women with OT and associates with subacute psychiatric symptoms, seizures, decrease level of consciousness, and frequent central hypoventilation. CSF findings suggest an encephalitic process that in prior accounts was suspected to be immune mediated. This hypothesis is now strongly supported by the detection in patients' sera and CSF of an immune response to antigens that colocalize with EFA6A, a brain-specific protein involved in the regulation of dendritic development of hippocampal neurons.19
Clinical recognition of this syndrome is important for two reasons: (1) the subacute presentation of the indicated symptoms in a young woman often leads to an initial diagnosis of acute psychosis, malingering, or drug abuse, and (2) despite the severity of symptoms patients usually recover. Because the symptom presentation is usually acute with prominent confusion, delusional thinking, agitation, and low level of consciousness, the recognition of the short-term memory deficits characteristic of limbic dysfunction can be difficult. Furthermore, the neuroimaging findings may differ from classic LE: combining MRI, FDG-PET, and SPECT only one third of the patients had involvement of the limbic system (cingulum, orbitofrontal region, temporal lobes), and none had the characteristic FLAIR or T2 medial temporal abnormalities seen in LE; the other two thirds of the patients had abnormalities in other areas of the central nervous system (CNS) or normal MRI.
Central hypoventilation has been reported in a few patients with paraneoplastic disorders involving brainstem, hypothalamus, or both.20-22 The high frequency of this complication in our group of patients (young women without a history of respiratory problems) is remarkable and suggests a pathogenic mechanism in common with the associated encephalitis. The development of hypoventilation was unrelated to seizures or antiepileptic medication; all patients required prolonged ventilatory support (median, 49 days) until recovery or in one case until death (Patient 3). The autopsy of this patient showed nonspecific microglial activation throughout gray and white matter of the brain, brainstem, and cerebellum, with scattered monocytes in perivascular spaces.11 Involvement of the brainstem was also shown by FDG-PET or MRI in three other patients (Patients 1, 4, and 5). Overall, these findings and other transient symptoms in some patients (ie, aphasia, diplopia, myelopathy) suggest that teratoma-associated encephalitis is an extensive or multifocal encephalitis, in which brainstem involvement is frequent and may be the cause of hypoventilation.
Despite the severity of the clinical features, three of our four patients and five patients reported by others survived, four with total recovery and four with substantial improvement. All but one patient had tumor resection, and six received immunosuppressants or chemotherapy. Because this is a small series, the contribution of each therapy to the neurological improvement is unclear. The close temporal association between presentation of teratoma-associated encephalitis and tumor diagnosis (median, 3 months) and the striking correlation in one case (Patient 4) between three episodes of neurological relapse and tumor recurrence suggest that the tumor plays a role in triggering the immune response. Frozen tumor tissue (not available from our patients) is needed to determine whether immunological activation depends on the tumor expression of neuronal proteins.
What is the cause of teratoma-associated encephalitis? The identification in four of four patients of antibodies with a unique pattern of reactivity with the CNS provides strong evidence that the disorder is immune mediated. Several features put these antibodies in a recently reported category12 that is different from the group of classic paraneoplastic antibodies to intracellular neuronal antigens (ie, HuD, CRMP5, or Ma2): (1) the immunohistochemical reactivity is highly restricted to the neuropil of rat hippocampus, sparing neuronal cell bodies; (2) the reactivity is missed by immunoblot analysis (data not shown) or if rat brain tissue is fixed with acetone or methanol-acetone; (3) studies with hippocampal neurons demonstrate that the immunolabeling occurs at the plasma membrane and dendritic processes suggesting that the antigens are on the cell surface; and (4) serum antibody titers become undetectable after neurological improvement.
To identify the autoantigens of the disorder, we probed a cDNA library of rat hippocampus with patients' sera, resulting in the isolation of EFA6A. When each individual serum was reacted with EFA6A expressed in bacteriophage plaques, only one recognized the protein, but affinity-purified EFA6A antibodies reproduced the same unique pattern of hippocampal reactivity as that of all patients' antibodies, with striking colocalization of neuronal immunolabeling. This finding indicates that the disorder's autoantigen is EFA6A itself but some epitopes are conformational or absent in the isolated partial sequence or that there is an additional coantigen that forms part of the complex of proteins that bind EFA6A. When EFA6A interacts with ARF6 at the plasma membrane,23 it binds to TWIK-1,24 a member of a new family of K(+) channels that is composed at least of 14 related subunits, each characterized by two-pore domains.25,26 In preliminary studies, we have found that ARF6 is not a coautoantigen, but the immunolabeling pattern of patients' antibodies resembles that of commercially available antibodies to some members of the two-pore-domain K(+) channels (ie, TASK-3; data not shown). These channels (also known as “leak K+ channels”) differ in structure and function from the shaker channels recently reported as autoantigens of nonparaneo-plastic LE.16,27 A general property of the two-pore K+ channels is that they show little time and voltage dependence and play critical roles in controlling the resting membrane potential.
The ARF6/EFA6A/TWIK-1 complex is important for endocytosis, channel internalization, and recycling.17,24 Because the reactivity of patients' antibodies colocalizes with EFA6A, it is reasonable to speculate that an immune-mediated disruption of the ARF6/EFA6A/TWIK-1 complex could result in severe deficits, similar to those of our patients. These would have broad effects on the limbic system where EFA6A is enriched and other areas of the CNS where at least three EFA6 isoforms (B, C, D) are also expressed.18,24,28 Pertinent to the frequent hypoventilation and low level of consciousness of our patients, the function of at least two-pore channels (TASK-1 and TASK-3) is important for control of breathing and arousal.29,30 A next step to determine other autoantigens is to precipitate EFA6A-interacting proteins using patients' antibodies.
On the clinical side, this study has the following implications: (1) when confronted with a young woman with a new-onset psychiatric syndrome associated with seizures, memory loss, decreased level of consciousness or central hypoventilation physicians should suspect a paraneoplastic disorder; (2) the tumor search should focus in the ovaries, and if a tumor is detected (even with a benign appearance) removal is recommended; (3) clinical improvement is slow; the development of central hypoventilation does not necessarily imply an ominous prognosis, but patients usually require extended ventilatory support; (4) based in a recently reported animal model demonstrating CNS dysfunction caused by antibodies to membrane-associated antigens,31 the finding that the antigens of teratoma-associated encephalitis are readily accessed by antibodies in live neurons suggests that early use of corticosteroids and IgG-depleting strategies (IVIg or plasma exchange) may improve outcome. This could not be determined in our patients because immunotherapies were used empirically when the patients were critically ill. The hippocampal antibodies were retrospectively identified in archived sera and CSF examined after recovery in three patients, and after death in one case.
Analysis for EFA6A-specific antibodies by immuno-precipitation or affinity purification from patient's serum is complex, and we consider the use of these methods as a diagnostic test premature. However, immunohistochemical analysis for the antibodies reported here (anti-EFA6A or anti-EFA6A–interacting proteins) is simple, and the pattern of reactivity characteristic enough to allow its use to identify the disorder. Analysis of the functional effects of the antibodies in live neurons is in progress.
This work was supported by the NIH (National Cancer Institute, RO1CA89054 and RO1CA107192, J.D.).
We thank A. Zupa-Fernandez and M. Maronski for their excellent technical assistance. The authors are indebted to Drs R. H. Muni and D. S. Liebeskind for providing information and serum, and Drs F. Graus, L. Bataller, and M. R. Rosenfeld for critical review of the manuscript.