Thymoma and Autoimmune Encephalitis: Clinical Manifestations and Antibodies.

OBJECTIVE: To report the clinical, neuroimaging, and antibody associations in patients with autoimmune encephalitis (AE) and thymoma.
METHODS: A retrospective cohort study of 43 patients was conducted. Antibody determination and immunoprecipitation to characterize novel antigens were performed using reported techniques.
RESULTS: Patients' median age was 52 years (range: 23-88 years). Forty (93%) had neuronal surface antibodies: gamma-aminobutyric acid receptor A (GABAAR) (15), amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) (13), contactin-associated protein-like 2 (CASPR2) (4), leucine-rich, glioma inactivated 1 (LGI1) (3), glycine receptor (GlyR) (3), and unknown antigens (2). Concurrent antibodies against intracellular antigens occurred in 13 (30%; 9 anti-collapsin response mediator protein 5 [CRMP5]) and were more frequent in anti-AMPAR encephalitis (54% vs 20%; p = 0.037). The most common clinical presentation was encephalitis with multiple T2/fluid-attenuated inversion recovery hyperintense lesions in 23 (53%) patients (15 GABAAR, 5 AMPAR, and 1 unknown neuropil antibody), followed by encephalitis with peripheral nerve hyperexcitability in 7 (16%; 4 CASPR2, 2 LGI1, and 1 unknown antibody), limbic encephalitis in 6 (14%; 4 AMPAR, 1 LGI1, and 1 antibody negative), progressive encephalomyelitis with rigidity and myoclonus in 4 (9%; 3 GlyR and 1 AMPAR antibodies), and encephalitis with normal MRI in 3 (7%; AMPAR antibodies). Anti-GABAAR encephalitis was more prevalent in Japanese patients compared with Caucasians and other ethnicities (61% vs 16%; p = 0.003). In anti-AMPAR encephalitis, 3/4 patients with poor and 0/6 with good outcome had concurrent CRMP5 antibodies (p = 0.033). Immunoprecipitation studies identified metabotropic glutamate receptor 3 antibodies that were additionally found in 5 patients (3 with and 2 without encephalitis).
CONCLUSIONS: AE in patients with thymoma include several clinical-radiologic syndromes that vary according to the associated antibodies. Anti-GABAAR encephalitis was the most frequent AE and occurred more frequently in Japanese patients.

Thymomas frequently associate with systemic and neurologic autoimmune disorders, mainly myasthenia gravis (MG).[1,2] Although these disorders can be considered paraneoplastic, there are some features that differentiate them from classical paraneoplastic neurologic syndromes. First, the autoimmune neurologic syndromes associated with thymoma occur more frequently as non-paraneoplastic syndromes. For example, in MG, only 15%–20% of patients have an underlying thymoma.[3] Second, with the exception of MG, up to 40% of the remaining autoimmune neurologic syndromes develop months to years after the diagnosis and treatment of the thymoma.[4] Third, compared with other tumors, the higher frequency of autoimmune disorders in patients with thymoma suggests a loss of self-tolerance that normally occurs in the thymus, resulting in autoreactive T cells that also affect B-cell function.[3,5] This mechanism is different from that proposed for classical paraneoplastic neurologic syndromes in which the immune response is likely triggered by tumor expression of the antigen.[6]

The occurrence of autoimmune encephalitis (AE) in patients with thymoma was first recognized in 1988,[7] and since then, around 40 cases have been reported, showing a great heterogeneity in clinical and brain MRI features, as well as in the antibody repertoire.[8] Previous series on thymomas and autoimmune neurologic disorders focused on the analysis of neuronal antibody profiles in patients with multiple neurologic manifestations.[9,10] Moreover, the frequency and distribution of the most recently characterized neuronal surface antibodies have not been previously described in a well clinically defined series of patients with AE with thymomas.[11]

Here, we describe the spectrum of clinical and neuroimaging features in patients with AE with thymoma and the associated neuronal antibodies.

Methods

Patients, Controls, and Clinical Definitions

We retrospectively assessed the clinical information of patients with suspected AE and histologic diagnosis of thymoma whose serum or CSF samples were sent to the Neuroimmunology Laboratory of the Institute of Biomedical Research August Pi i Sunyer, Hospital Clínic, for determination of neuronal antibodies, between January 2000 and October 2020. Demographic and clinical data were obtained by review of clinical records, and a structured written questionnaire was sent to the referring physicians. Patients were included in the study if they had clinical manifestations compatible with AE[12] and serum or CSF was available for immunologic studies.

The encephalitis was classified into 5 groups: (1) limbic encephalitis with MRI T2-weighted/fluid-attenuated inversion recovery (FLAIR) hyperintense lesions highly restricted to medial temporal lobes[12]; (2) encephalitis with concomitant peripheral nerve hyperexcitability (PNH)[13]; (3) encephalitis and multiple T2/FLAIR hyperintense lesions in the cerebral hemispheres or basal ganglia; (4) encephalitis with normal brain MRI; and (5) encephalitis with prominent brainstem symptoms and rigidity compatible with the diagnosis of progressive encephalomyelitis with rigidity and myoclonus (PERM).[14]

In patients with multiple neuronal surface antibodies, a predominant antibody was chosen on the basis of its presence in the CSF, titers, and type of AE. Controls for antibody studies included paired serum/CSF (23) or serum (16) samples of 39 patients with thymoma and isolated neuromuscular disorders (23 with MG, 1 of them with concomitant optic neuritis, and 16 with PNH).

Antibody Studies

Samples were investigated for cell surface, synaptic, and onconeural antibodies by immunohistochemistry and in-house cell-based assays (CBAs). Rat brain immunostaining was performed with patients' serum and CSF using a standard avidin-biotin immunoperoxidase technique adapted to cell surface and intracellular antibodies, as reported.[15,16] The presence of antibodies against amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR), aquaporin 4, contactin-associated protein-like 2 (CASPR2), dipeptidyl-peptidase-like protein 6, gamma-aminobutyric acid receptors A (GABAAR) and B (GABABR), glycine receptor (GlyR), immunoglobulin-like cell adhesion molecule 5 (IgLON5), leucine-rich, glioma inactivated 1 (LGI1), metabotropic glutamate receptor 1 (mGluR1), 2 (mGluR2), 3 (mGluR3), and 5 (mGluR5), neurexin 3α, and NMDA receptor was determined with in-house CBAs with live (serum 1:50 and CSF 1:5) or fixed (serum 1:40 and CSF 1:2) human embryonic kidney (HEK)-293 cells expressing the corresponding aforementioned antigens, as reported.[17] The commercial immunoblot kit EUROLINE Paraneoplastic Neurological Syndromes 12 Ag (DL 1111-1601-7 G; Euroimmun, Lübeck, Germany) was used following the manufacturers' instructions to confirm glutamic acid decarboxylase (GAD) and onconeural reactivities seen in the immunohistochemistry studies. Immunoprecipitation using cultures of rat cerebellar granule neurons and patient's serum was used to characterize the target antigen in a patient who had antibodies against an unknown cell surface protein, as described.[17] The serum of this patient was preabsorbed with HEK293-mGluR3 cells and tested on rat brain sections to determine whether the preabsorption prevented brain immunoreactivity. The immunoabsorption studies were conducted as previously reported.[18]

Statistical Analysis

Variables comparing patients with thymoma-associated AE and isolated neuromuscular disorders were analyzed with the Fisher exact test. All tests were 2 tailed, and statistical significance was set at <0.05. All analyses were performed with SPSS Statistics 26.0 (IBM, New York City, NY).

Standard Protocol Approvals, Registrations, and Patient Consents

The Ethic Committee of the Hospital Clinic approved the study. All patients or proxies gave written informed consent for the storage and use of serum, CSF, and clinical information for research purposes.

Data Availability

All data are reported within the article and available anonymized by request from qualified investigators.

Results

We identified 61 patients with thymoma and suspected AE, 18 of them were excluded due to inadequate clinical information or because an alternative diagnosis was established; 43 were lastly diagnosed with AE and included in the study. Eleven of them had been previously reported (including patients 5 and 6 of reference 15, patients 13 and 15 of reference 19, and patients 1 and 5 of reference 20).[21-25] Median age of the 43 patients at disease onset was 52 years (range: 23–88 years), and 26 (60%) were female. Twenty (46%) patients were Caucasian, 18 (42%) Japanese, and 5 (12%) had other ethnicities. Thirteen (30%) patients had a previous diagnosis of MG with acetylcholine receptor antibodies. The thymoma was locally invasive or metastatic in 22 (51%) patients and antedated the diagnosis of the encephalitis in 16 (37%). In 7 (44%) of them, the encephalitis coincided with oncological progression or relapse of the thymoma (Table 1). For the other patients in which the encephalitis preceded the tumor diagnosis, the thymoma was detected in the first tumor screening. The presence of MG or aggressive thymoma (invasive or metastatic) did not correlate with the type of encephalitis or associated neuronal antibodies. Long-term follow-up was available for 39 patients (median, 4 years; interquartile range: 10–68 months). Antibodies against neuronal surface antigens were found in 40 (93%) of the 43 patients: GABAAR in 15 patients, followed by AMPAR in 13, CASPR2 in 4, LGI1 in 3, GlyR in 3, and against unknown surface antigens in 2. Concurrent neuronal antibodies were detected in 16 (40%) of 40 patients (Table 1). Two patients with AMPAR antibodies also had GABABR (1 of them along with CASPR2) antibodies; 3 patients, 2 with CASPR2 and 1 with GABAAR antibodies, also had LGI1 antibodies. Concurrent antibodies against intracellular antigens were found in 13 (30%) patients and included CRMP5 in 9 (21%), GAD in 3 (7%), and Hu in 1 (2%) (Table 1). Onconeural and GAD antibodies were more commonly associated with AMPAR antibodies as they occurred in 7/13 (54%) patients but in only 6/30 (20%) without AMPAR antibodies (p = 0.037).

Table 1

Clinical and Immunologic Features According to Neuronal Surface Antibody Type

Clinical Profiles of Thymoma-Associated Paraneoplastic Encephalitis

The most common clinical presentation was encephalitis with multiple T2/FLAIR hyperintense lesions in 23 (53%) patients; 15 of them had GABAAR antibodies, 5 AMPAR antibodies, and 1 an unknown neuropil antibody. In 2 patients, no antibody was found, but CSF was not available in 1 of them (Figure 1). Encephalitis with multiple T2/FLAIR hyperintense lesions and GABAAR antibodies was more prevalent in Japanese patients (11/18, 61%) than in patients of other ethnicities (4/25, 16%; p = 0.003). Eleven (73%) of the 15 patients with GABAAR antibodies had prominent seizures that were the only symptom in 1 and evolved to refractory status epilepticus in 4. Thirteen (87%) of them had cognitive impairment and 11 (73%) developed behavioral changes. CSF studies were normal in 13 (87%) patients. The other 2 patients had mild clinical manifestations, although they had multiple lesions in the brain MRI. The brain MRI showed multiple non–contrast-enhancing bilateral T2/FLAIR hyperintense lesions involving cortico-subcortical areas of both cerebral hemispheres (Figure 2, A and B). Follow-up MRI in 9 (60%) patients showed the appearance of new lesions and improvement of others without a clear correlation with the use of immunotherapy. Although 10 (67%) patients developed 1 or more relapses, 13 had good functional outcome (modified Rankin Scale [mRS] score ≤2), including 1 who had full clinical and radiologic recovery without immunotherapy.

Figure 1

Antibody Distribution According to the Type of Encephalitis

Results are expressed as percentage of each antibody. AE = autoimmune encephalitis; AMPAR = amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; CASPR2 = contactin-associated protein-like 2; FLAIR = fluid-attenuated inversion recovery; GABAAR = gamma-aminobutyric acid receptor A; GlyR = glycine receptor; LGI1 = leucine-rich glioma inactivated 1; PERM = progressive encephalomyelitis with rigidity and myoclonus; PNH = peripheral nerve hyperexcitability.

Figure 2

Basal Ganglia Involvement in Anti-AMPAR Encephalitis

(A and B) Brain MRI of a patient with anti-GABAAR encephalitis. Axial fluid-attenuated inversion recovery images show bilateral hyperintense lesions in temporal lobes (yellow arrows). No lesions are observed in the basal ganglia. (C and D) Brain MRI of a patient with anti-AMPAR encephalitis shows hyperintense lesions in the striatum (blue arrows) and another lesion in the temporal lobe (yellow arrow). AMPAR = amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; GABAAR = gamma-aminobutyric acid receptor A.

Another 5 patients with encephalitis and multiple T2/FLAIR hyperintense lesions had AMPAR antibodies (Table 1). Unlike patients with GABAAR antibodies, CSF studies showed inflammatory signs in all (mean white blood cell count: 56; range: 7 to 164), and the T2/FLAIR hyperintense lesions involved the basal ganglia in 4 (80%) patients, a feature only seen in 2 (13%) patients with anti-GABAAR encephalitis (Figure 1B). Three of the 5 patients had a good response to immunotherapy, and despite a relapse in 2, the mRS score was ≤2 at the last visit.

Seven (16%) of 43 patients developed encephalitis with concomitant PNH (Table 1). Four of them had serum and CSF CASPR2 antibodies, along with low serum titers of LGI1 antibodies in 2, 2 had LGI1 antibodies only, and 1 had an antibody against an unknown antigen. In contrast with the other clinical groups, all of these patients reported prominent sleep disorders. CSF studies were normal in all (only mildly increased protein concentration in 1). The 2 with LGI1 antibodies had hyponatremia and bilateral medial temporal lobe T2/FLAIR hyperintensities, whereas the 4 patients with CASPR2 antibodies had a normal brain MRI. Two of the patients with concurrent CASPR2 and LGI antibodies and another 2 with isolated LGI1 antibodies had an unfavorable clinical course, and 3 of them died despite receiving immunotherapy. The fourth patient in this group was a 68-year-old man who consulted us for subacute anxiety, nocturnal hyperhidrosis, severe insomnia, pain in the legs, intermittent diarrhea, and weight loss over 3 months. He had a 15-year history of MG treated with low dose of steroids and mycophenolate. Brain MRI was normal, and the EMG confirmed the presence of neuromyotonia. Serum and CSF studies revealed CRMP5 antibodies concurrent with an unknown antibody that strongly reacted with the membrane of live neurons, making it a good candidate to precipitate the target antigen (see the section “Immunologic Studies”). An MRI of the chest showed an anterior mediastinal mass consistent with thymoma that was surgically removed. He was further treated with steroids and IV immunoglogulin with good response.

Thirteen (30%) of the 43 patients developed other types of encephalitis: (1) 6 patients had clinical and MRI features of limbic encephalitis with AMPAR antibodies in 4 (1 patient with concurrent CASPR2 and GABABR antibodies), LGI1 antibodies in 1, and absence of antibodies in serum and CSF in the other patient (Figure 1). Compared with other types of encephalitis, the response to immunotherapy was worse. One patient died, and the rest of the patients had mild to moderate cognitive and behavioral sequelae despite immunotherapy. (2) Four patients presented with clinical features of PERM associated with GlyR antibodies in 3 and AMPAR antibodies in 1. (3) Three patients developed clinical manifestations suggestive of encephalitis (confusion, short-term memory loss, and fluctuating level of consciousness), but the brain MRI was normal; all 3 had AMPAR antibodies, with concurrent GABABR antibodies in 1.

The functional outcome of patients according to the mRS score at the last follow-up is represented in Figure 3. Patients with anti-GABAAR encephalitis (13/15, 87%) and those with PERM with GlyR antibodies (3/3, 100%) were more likely to have good outcome than patients with other antibody-associated encephalitis (9/21, 43%; p = 0.019). In patients with AMPAR antibodies, the presence of CRMP5 antibodies associated with poor outcome. Three of the 4 patients who did not respond to immunotherapy, but none of the 6 who responded, had concurrent CRMP5 antibodies (p = 0.033).

Figure 3

Clinical Outcome According to the Type of Encephalitis

Good: modified Rankin Scale score ≤2; poor: modified Rankin Scale score >2.

Immunologic Studies

Immunoprecipitation of granular cells of the cerebellum incubated with serum of the indicated patients with encephalitis and PNH revealed 16 unique peptides encompassing 22% of the sequence of rat mGluR3 protein. A CBA using HEK293 cells transfected with mGluR3 (Figure e-1, links.lww.com/NXI/A517) identified 5 additional patients: 2 with encephalitis and GABAAR or AMPAR antibodies and 3 with isolated neuromuscular disorders. Immunoabsorption studies showed that samples preabsorbed with HEK293-mGluR3 cells abolished the reactivity with mGluR3 expressed in HEK293 cells but did not abrogate the reactivity with rat brain sections, indicating that the neuropil reactivity was due to another unknown antibody.

We next compared the antibody repertoire in thymoma patients with and without AE (Table 2). GABAAR and AMPAR antibodies were only detected in patients with AE, whereas LGI1 and CASPR2 antibodies were also positive in patients with isolated PNH. Onconeural antibodies, mainly CRMP5, were significantly more common in the encephalitis group (10/43, 23%) but they were also detected in 2 (5%) of 39 thymoma patients without encephalitis (p = 0.028) (Table 2).

Table 2

Frequency of Neuronal Antibodies in Patients With Thymoma With and Without Encephalitis

Discussion

This retrospective observational study shows that the AE of patients with thymoma[8] is not a single entity, but a heterogeneous group of encephalitides with different clinical-radiologic features and outcomes that vary according to the associated neuronal surface antibodies. Patients with thymoma with AE shared 3 clinical characteristics independently of the immunologic subgroup: (1) up to 30% had a previous diagnosis of MG, (2) the thymoma was locally invasive or metastatic in 51% and preceded the diagnosis of encephalitis in 37%, and (3) multiple coexisting neuronal antibodies were detected in 40% of patients, whereas only 7% were neuronal antibody negative.

Although the first reports on patients with AE associated with thymoma indicated that they had limbic encephalitis, subsequent single case reports showed that some of them had multiple T2/FLAIR MRI abnormalities beyond the temporal lobes.[8,26-28] Our study confirms these observations and shows that the MRI findings frequently suggested the antibody that was going to be detected; for example, most patients with multiple T2/FLAIR lesions had GABAAR antibodies and less frequently AMPAR, whereas most cases with limbic encephalitis had AMPAR followed by LGI1 antibodies. In addition, patients with AE and PNH had normal MRI and CASPR2 antibodies.

We found that the most common AE in patients with thymoma was associated with GABAAR antibodies. The association of anti-GABAAR encephalitis with thymoma was already recognized in the initial report of this disease and several subsequent case reports from Japan.[22,23,29] Moreover, in a series of 17 patients with anti-GABAAR encephalitis, the 3 patients with thymoma were also Japanese.[20] Thus, our study confirms the frequent association of anti-GABAAR encephalitis in Japanese patients with thymoma, as 11 (73%) of 15 patients with this disorder were Japanese. The reason for this association is unclear but probably indicates a particular genetic susceptibility. For example, the human leukocyte antigen (HLA) class II DPB1*0501 allele is associated with susceptibility to the opticospinal form of multiple sclerosis, which is prevalent in Japan but unusual in Western countries.[30]

The anti-GABAAR encephalitis of patients in the current study is otherwise similar to that of patients without thymoma.[20,31] However, we identified cases with thymoma that developed mild clinical manifestations and spontaneous remission of the brain MRI lesions without immunotherapy. These unusual benign courses should not prevent early onset of treatment as patients may potentially have severe relapses.[22]

AMPAR antibodies were the second most frequent autoantibodies in this series, and the brain MRI findings were heterogeneous. The most common MRI finding was the presence of multiple FLAIR/T2 hyperintense lesions that, unlike anti-GABAAR encephalitis, frequently involved the basal ganglia; less frequently, patients had medial temporal lobe FLAIR changes typical of limbic encephalitis. Three patients had normal brain MRI, a feature previously described in cases with or without thymoma.[19,32] The relative higher frequency (38%) of multifocal T2/FLAIR lesions in patients with anti-AMPAR encephalitis was not observed in previous series of patients with this disease but without thymoma (28% combining 2 previous series[19,33]).

The clinical outcome in patients with anti-AMPAR encephalitis was worse than that of anti-GABAAR encephalitis. This is likely related to the frequent presence of concurrent paraneoplastic antibodies and associated mechanisms, particularly CRMP5 autoimmunity.[34,35] In a series of 22 patients with anti-AMPAR encephalitis not focused on thymoma, a shorter survival was observed in patients with concurrent onconeural antibodies compared with those harboring isolated AMPAR antibodies.[19]

Although previous studies have shown that patients with encephalitis, PNH, and CASPR2 or LGI1 antibodies usually respond to immunotherapy,[36,37] we found that this was the group with the worse outcome. There are preliminary data suggesting that the outcome is worse if patients have thymoma.[36] In a series of 29 patients with Morvan syndrome, the clinical outcome was substantially better in the 17 patients without thymoma.[13] In contrast, the presence of thymoma did not affect the prognosis of patients with other antibody-related syndromes; for example, the 3 patients with PERM and GlyR antibodies made an excellent recovery similar to that observed in idiopathic cases and in 6 of 7 previously reported patients with thymoma.[14,24,38]

Using the serum of a patient with encephalitis, PNH, and an unknown neuropil antibody, we precipitated mGluR3. The CBA identified mGluR3 antibodies in 5 additional patients with and without AE suggesting that the antibodies are biomarkers of the underlying thymoma, rather than the AE, as occur for antibodies against Netrin-1 receptors.[39] Moreover, the persistence of serum immunoreactivity with the rat brain after immunoabsorption with mGluR3 suggests the presence of additional (not yet characterized) antibodies, emphasizing the frequency of multiple autoantibodies in patients with encephalitis and thymoma.

A limitation of this study is that it is retrospective and may suggest a referral bias for testing some antibodies that are not readily available in commercial antibody panels, such as GABAAR antibodies. However, this potential bias does not explain the association with the Japanese ethnicity observed here because samples from Japan were not sent to our laboratory to specifically confirm the presence of GABAAR antibodies. Brain MRIs were not centrally reviewed, and the data were obtained from radiologic reports preventing an accurate analysis of potentially distinctive MRI features of encephalitis associated with different antibodies.

Our findings have several clinical implications: (1) most patients with thymoma and AE have neuronal surface autoantibodies that can occur alone or with concurrent intracellular antibodies. (2) Symptoms may occur without CSF pleocytosis, which is common in anti-GABAAR encephalitis, or with normal brain MRI findings as in some patients with AMPAR or CASPR2 antibodies. (3) Characterization of these antibodies is important because they associate with different syndromes and prognosis. (4) Patients whose serum and CSF are negative for autoantibodies included in clinical or commercial panels, or cases in which the antibody identified does not match with the clinical-radiologic syndrome should have a more comprehensive antibody evaluation. Future studies should include HLA genotyping in anti-GABAAR encephalitis to determine any genetic predisposition among Japanese patients with thymoma.