Neurologic autoimmunity and immune checkpoint inhibitors

CNS manifestations

CNS manifestations, observed in 31 patients (clinical cohort, 13; laboratory cohort, 18; table 2), included encephalopathy with or without seizures (n = 21), cerebellar ataxia (n = 11), hyperkinetic movement disorders (n = 3), optic neuropathy (n = 4), myelopathy (n = 3), or combinations thereof. Only 11/31 patients with CNS manifestations had neuroendocrine tumors, while the rest had malignancies not typically associated with paraneoplastic autoimmunity. Encephalitis manifestations were either localized (e.g., limbic or basal ganglia restricted) or diffuse. Concomitant posterior reversible encephalopathy syndrome (PRES) and fasciculations developed in an LGI1 immunoglobulin G (IgG) seropositive patient. Two patients with CSF positivity for GFAPα IgG had, as anticipated, a steroid-responsive meningoencephalitis.²¹ In a patient with high GAD65 IgG levels, limbic encephalitis appeared after new onset of type 1 diabetes.

Among 20 patients with available CSF data, 19 (95%) had abnormalities: lymphocytic pleocytosis, 14 (median cell count, 25; range, 8–76); CSF-restricted oligoclonal bands, 9.

MRI abnormalities, noted in 11/18 (61%) with available information, included localized T2 hyperintensities (e.g., mesiotemporal, basal ganglia, spinal cord), parenchymal/meningeal enhancement, or focal atrophy (e.g., cerebellum). MRI findings were consistent with the patients' clinical syndromes (figure, A–D). ¹⁸F-Fluorodeoxyglucose PET revealed diffuse or focal areas of hypometabolism or hypermetabolism in 2 patients with limbic encephalitis; 2 additional patients (both lacking CSF biomarkers for Alzheimer disease) had diffuse and frontotemporal hypometabolism, respectively (table 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure Brain neuroimaging and nerve and muscle biopsies from patients with immune checkpoint inhibitor–related neurologic autoimmunity

(A–D) Brain MRI. (A) Axial image shows basal ganglia fluid-attenuated inversion recovery (FLAIR) hyperintensities, affecting the putamen and caudate head bilaterally, in a patient with phosphodiesterase 10A antibodies and hyperkinetic movements. (B) Sagittal T1-weighted image shows atrophy of the superior cerebellar vermis in a patient with seronegative autoimmune ataxia. (C) Axial imaging shows bilateral mesial temporal pole FLAIR hyperintensities, more prominent on the left, in a patient with limbic encephalitis and autoantibodies as yet unidentified molecularly. (D) ¹⁸F-Fluorodeoxyglucose PET axial images show hypometabolism of the temporal poles in a patient with limbic encephalitis, ataxia, optic neuritis, and antineuronal nuclear antibody 1 antibodies. (E) Nerve biopsy of the left sural nerve of a patient with ICI-related vasculitic neuropathy demonstrates a large perivascular epineurial inflammatory collection with infiltration and disruption of the blood vessel wall: hematoxylin & eosin (H&E) (E.a) and CD45 (E.b) stain. (F) Muscle biopsy of a patient with ICI-related myopathy: H&E-stained section (F.a) shows regenerating (arrowhead) and necrotic fibers invaded by macrophages (arrows). Macrophages appear red on acid phosphatase-stained section (F.b).

Neuromuscular manifestations

Thirty-four patients had neuromuscular symptoms. Peripheral neuropathy was most common (n = 22; clinical cohort, 19; laboratory cohort, 3), in isolation (n = 14) or accompanied by myopathy (n = 4) or CNS manifestations (n = 4). Isolated neuralgic amyotrophy developed in 4 patients (unilateral brachial plexopathy, 2; bilateral phrenic nerve palsy presenting with acute dyspnea, 2). Five patients had a subacute length-dependent polyradiculoneuropathy (both axonal and demyelinating, 4; pure axonal, 1). Two patients had diffuse acute inflammatory demyelinating polyneuropathy with facial nerve involvement. Two patients had biopsy-proven vasculitic neuropathy (figure, E) and improved minimally or not at all after ICI discontinuation or immunosuppression/immunomodulation. Other neuropathy phenotypes included isolated cranial neuropathy (excluding optic neuritis; n = 5) or other nonspecified neuropathies (n = 4).

Three patients had clinical generalized myasthenia gravis (MG): clinical cohort, 1; laboratory cohort, 2. In one of these patients, the disease was known before ICI treatment initiation and this was the only patient with MG who tested positive for muscle AChR IgG.

Myopathy developed in 13 patients (clinical cohort, 11; laboratory cohort, 2) and predominantly involved bulbar muscles (n = 5), proximal limb muscles (n = 5), or was confined to the extraocular muscles (n = 1). The 2 remaining patients had diffuse myopathy causing diaphragmatic paralysis with respiratory failure in one. Serum creatine kinase was elevated in 8/12 (67%) with available data (median 680 U/L [range 72–13,000]). Troponin was elevated in 8 patients, with decreased ejection fraction by echocardiography in 4. Muscle biopsy, available for 6 patients, contained multifocal areas with necrotic and regenerating fibers (n = 4; figure, F) or mild inflammation (n = 2).

Other clinical manifestations

Three patients (clinical cohort) had ICI-related hypophysitis before or coinciding with onset of neurologic manifestations; 17 patients (45%) had other preexisting or coinciding nonneurologic autoimmunity (e.g., thyroiditis, myocarditis, pneumonitis, hepatitis, colitis). Two patients had isolated autoimmune retinopathy: (1) bilateral diffuse uveal melanocytic proliferation (BDUMP) in a 60-year-old man with esophageal junction adenocarcinoma treated successfully with IV immunoglobulin (IVIg), high-dose steroids, and plasmapheresis; and (2) preexisting melanoma-associated retinopathy in a 75-year-old woman who worsened after pembrolizumab and improved following IVIg and bevacizumab.

Autoantibody profile

Of 31 patients with CNS involvement, 24 patients (77%) had detectable neural-specific autoantibodies. Neural antibodies with known antigenic specificities were seen in 16/24 (67%) (serum, 4; CSF, 6; both, 6): ANNA1, 3; CRMP5, 3; amphiphysin, 2; GFAP, 2; NIF, 2; PDE10A, 2; SOX1, 1; LGI1, 1; GAD65, 1; NMDAR, 1; with or without coexisting low-titer serum VGKC, 3; P/Q type VGCC, 2; and N type VGCC, 2. In 9 patients (38%), the neural-specific antibodies detected were of yet unidentified molecular specificities (serum, 2; CSF, 6; both, 1), including one with coexisting SOX1 and P/Q-type VGCC autoantibodies. In the unbiased clinical cohort, the frequency of any neural-specific autoantibody among patients with CNS involvement was 54% (7/13 patients; 1/7 had an isolated UNA [CSF of a patient with limbic encephalitis]).

Of 22 patients with peripheral nerve involvement, 9 were positive for the following: ANNA1, 3; LGI1, 1; UNA, 1; muscle AChR, 3; and STR, 3. Of these patients, 7 had concomitant CNS manifestations (all ANNA1 or LGI1-IgG positive) or myopathy (muscle AChR or STR autoantibody positive). One patient (muscle AChR IgG seropositive) had isolated oculomotor neuropathy per the referring provider that could represent an ocular myopathy and an UNA-seropositive patient with limited information had unspecified neuropathy (both patients were from the laboratory cohort).

All patients with MG or myopathy (n = 16) were tested for STR autoantibodies and 9 were positive; 5 of them had coexisting muscle AChR. All seropositive patients had myopathy except for one who had preexisting MG before ICI treatment. In the unbiased clinical cohort, 6/11 (55%) patients with myopathy were seropositive for muscle AChR and/or STR autoantibodies; 4 muscle AChR binding autoantibodies showed modulation. None of 8 patients with myopathy who were tested for SRP or HMGCR autoantibodies was positive.

A “classical” paraneoplastic association (i.e., clinical, oncologic, and autoantibody profiles typical of what is classically recognized as paraneoplastic) was observed in 6 patients: 5 of them had small-cell lung carcinoma (SCLC) and IgGs specific for amphiphysin and P/Q-type VGCC with preexisting cerebellar ataxia (n = 1); SOX1, P/Q-type VGCC, and UNA (n = 1) with limbic encephalitis; ANNA1 and CRMP5 (n = 1) with preexisting cerebellar ataxia, myelitis, and peripheral neuropathy (video 1); CRMP5 (n = 1) with a longitudinally extensive myelopathy; and ANNA1 and STR (n = 1) with limbic encephalitis, cerebellar ataxia, and cranial neuropathy. The sixth patient had Merkel cell carcinoma with NIF IgG and CRMP5 IgG and encephalitis, ataxia, and vision changes.

Among 9 patients with CNS manifestations and neuroendocrine tumors (without preexisting autoimmunity), only 1 patient was antibody negative, 4 had SCLC-predictive antibody specificities (ANNA1, CRMP5, NIF, P/Q-type VGCC, SOX1), 1 had GAD65 IgG, 1 had NMDAR IgG, and 2 had UNAs only, without other well-described neural antibodies. SCLC-predictive antibody profiles were seen in 2 patients with melanoma and CNS manifestations (NIF IgG and ANNA1) and 1 with non-SCLC (amphiphysin IgG).

Disease course and outcome

Thirty-four patients in the clinical cohort discontinued ICI therapy before or at neurologic symptom onset (87%). Thirty-one patients (79%) received immunosuppression/immunomodulation for their neurologic syndrome: oral corticosteroids, 23; high-dose IV steroids, 18; IVIg, 11; and plasmapheresis, 8. The 5 patients who continued their ICI therapy after neurologic symptom onset had lower adverse event severity grade at onset (median, 2 [range, 2–3]) compared to patients who discontinued ICI (median, 3 [range, 2–4]; p = 0.02; Wilcoxon rank sum test). Their clinical phenotypes were as follows: (1) a 60-year-old man with BDUMP and esophageal junction adenocarcinoma who received monthly IVIg (0.4 g/kg) in concomitance with pembrolizumab for 17 months with improvement of visual symptoms but cancer progression; (2) a 62-year-old man with metastatic melanoma and bilateral hearing loss after one cycle of pembrolizumab who was treated with prednisone 70 mg/d for 7 days with minimal improvement (required hearing aid 4 years later); pembrolizumab was continued for 15 months with cancer remission; (3) a 69-year-old man with melanoma and severe headache and loss of taste and hearing after 5 cycles of nivolumab and ipilimumab, treated with oral prednisone (60 mg daily for 1 week, then slow taper) with symptom resolution; nivolumab was continued for 10 months with cancer remission; (4) a 75-year-old man with recurrent melanoma and brachial plexopathy with right hemidiaphragm palsy after one cycle of nivolumab; no immunosuppression was used but there was mild spontaneous improvement; nivolumab was continued for 8 months with cancer remission; (5) a 63-year-old woman with small cell lung cancer and CRMP5 IgG progressive myelopathy after 3 cycles of atezolizumab treated with IVMP and cyclophosphamide with improvement of neurologic symptoms; an additional dose of atezolizumab was given and the patient is in cancer remission (patient 11 in table 2).

After a median follow-up of 10 months (range, 0.5–46), 14/39 patients (36%) showed unfavorable neurologic outcomes (residual severity grade ≥3) including 2 deaths from neurologic complications; 13 patients (33%) had favorable neurologic outcomes (severity grade = 1). Three patients relapsed after immunosuppressive therapy was tapered or discontinued (encephalitis, 1; CNS demyelination, 1; and proximal myopathy, 1). The patients with unfavorable outcome had encephalitis/ataxia, 8; vasculitic neuropathy, 2; bulbar myopathy, 2; generalized myopathy with diaphragmatic involvement, 1; or generalized MG, 1.

Of all the variables investigated by univariate logistic regression analysis (table 3), the following were found to be significantly associated with unfavorable outcome: onset age ≥70 years, female sex, CNS involvement, lung cancer, higher severity grade at onset, shorter follow-up, and absence of preceding or accompanying non-neurologic ICI-related autoimmunity. By multivariate analysis, only onset age ≥70 years remained independently associated with poor outcome (p = 0.01; table 3).

Preexisting neurologic autoimmunity

Five patients had evidence of neurologic autoimmunity before ICI therapy began. Three had clinically recognized paraneoplastic syndromes affecting the CNS (cerebellar ataxia with amphiphysin IgG, P/Q-type VGCC IgG, and SCLC; encephalomyelitis, cerebellar ataxia, and peripheral neuropathy with ANNA1 IgG and CRMP5 IgGs and SCLC; and limbic encephalitis with UNA IgG and melanoma). All of these patients had irreversible neurologic worsening after ICI initiation and did not improve following immunosuppressive therapy (table 2; video 1). One patient with muscle AChR IgG–positive MG worsened after ICI, requiring bilevel positive airway pressure ventilation despite immunosuppressive/immunomodulatory therapy. The single patient who improved with immunosuppression/immunomodulatory therapy had autoimmune retinopathy and had initially worsened on starting ICI therapy.