Peripheral nervous system involvement accompanies central nervous system involvement in anti-glial fibrillary acidic protein (GFAP) antibody-related disease

Glial fibrillary acidic protein (GFAP) is the predominant neurofilament in the astrocytes of the central nervous system (CNS), responsible for the cytoskeleton structure [1]. GFAP has been identified as an antigenic target of an inflammatory CNS disease, called autoimmune anti-GFAP astrocytopathy [2]. GFAP is also expressed in the peripheral nervous system (PNS) in non-myelinating immature Schwann cells and in satellite cells of dorsal root ganglia (DRG). It is likely, however, that CNS and PNS GFAP are not perfectly identical, since a monoclonal GFAP antibody was found to selectively recognize GFAP in astrocytes and not in PNS cells [3‐5]. Furthermore, some studies have shown the expression of this protein in Schwann cells after axonal damage, and a poor Schwann cell proliferation is observed after nerve injury when the GFAP expression is lost [6, 7]. GFAP mutations are also associated with demyelinating neuropathies in Alexander disease [8]. All these data suggest that GFAP could be a target in some PNS autoimmune diseases. Six patients with PNS involvement and autoimmune anti-GFAP astrocytopathy have been reported, and PNS symptoms were the initial manifestation for five of them [9]. Interestingly, a T-cell predominant perivascular inflammatory infiltrate and demyelinating features were observed in the nerve biopsies of these patients [9]. However, the clinical presentation seems to be heterogeneous, since polyradiculoneuropathy [9], radiculoneuritis [10], axonal sensorimotor neuropathy [11], axonal predominant motor neuropathy [12], and acute motor axonal neuropathy (AMAN) with positive GM1 IgG antibodies [13] have been reported. Despite these few reports, the phenotype of PNS involvement in autoimmune anti-GFAP astrocytopathy remains unclear, while the central manifestations have been subsequently characterized by several cohorts [14‐17]. The aim of this study was to describe PNS involvement in a cohort of patients with cerebrospinal fluid (CSF) GFAP antibodies (GFAP-Abs).

We report herein the largest cohort of patients with PNS involvement associated with CSF GFAP-Abs. The following data were instructive: (1) a prevalence of around one-fifth of PNS involvement, indicating that it is not a rare feature; (2) the clinical presentation of PNS involvement was heterogeneous; (3) cranial nerve involvement and radiculitis were the most common presentations; (4) bladder dysfunction was frequent; (5) PNS involvement was rarely the first manifestation; (6) all patients except one had associated CNS involvement; (7) all patients had pleocytosis and elevated protein in CSF; (8) onset was acute or subacute in most cases, and the prognosis seemed favorable.

The peripheral clinical picture in patients with GFAP-Abs is heterogeneous and dominated by cranial nerve involvement and radiculopathy, while inflammatory demyelinating polyradiculoneuropathy, sensory neuronopathy, or length-dependent axonal polyneuropathy were more rarely found. Radiculopathy was found to especially affect the lower limbs and cause proximal motor weakness with frequent root gadolinium enhancement.

These findings are partially consistent with a recently published cohort of patients with CSF GFAP-Abs and PNS involvement, which included 12 patients in whom the PNS involvement was electrophysiologically confirmed [12]. These patients mainly had lower limbs motor weakness with reduced compound muscle action potentials (7/12) on ENMG leading the authors to conclude to predominant motor axonal neuropathy. However, F waves were prolonged in 8/12 patients, questioning the possibility of radiculopathy rather than axonal motor neuropathy [12]. The present findings are also partially consistent with the first published case series of patients with GFAP-Abs and PNS involvement in which acute/subacute inflammatory demyelinating polyradiculoneuropathy was diagnosed in three of six patients [9]. This could be explained by the diagnostic criteria used in the above-mentioned study, as the diagnosis of acute/subacute inflammatory demyelinating polyradiculoneuropathy relied only on prolonged or absent F waves in two of these three cases, which may have instead been considered as cases of radiculopathy or polyradiculopathy in the current study. However, the third case had undoubtful inflammatory demyelinating polyradiculoneuropathy with electrophysiological arguments for peripheral nerve demyelination with prolonged distal latencies and slowed conduction velocities, and segmental demyelination on nerve biopsy [9]. In the cohort presented herein, only one patient fulfilled the criteria of inflammatory demyelinating polyradiculoneuropathy, characterized by a proximal motor weakness of the lower limbs, sensory ataxia, and clear demyelinating abnormalities on ENMG [18]. A direct role of the GFAP-Abs could be suspected in the cases of demyelinating polyradiculoneuropathy, because GFAP is expressed by immature Schwann cells, and demyelinating features were observed in the nerve biopsies of patients with GFAP-Abs expressing this phenotype [6, 9]. The clinicians should, therefore, consider testing CSF GFAP-Abs in cases of atypical CIDP-like diseases associated with CNS symptoms and CSF pleocytosis.

Interestingly, one patient herein had sensory neuronopathy. To our knowledge, it is the first case of sensory neuronopathy associated with GFAP-Abs (although previously reported by Gravier-Dumonceau et al.) [14]. We suggest a direct causal link between the identified CSF GFAP-Abs and such phenotype for the following reasons: (1) GFAP was found to be expressed in the DRG, and a patient’s CSF was able to stain the DRG by immunofluorescence [4, 9], (2) autoimmune processes are frequent in sensory neuronopathies, and several other antibodies have already been described in this syndrome such as Hu, CV2/CRMP5, FGFR3, or AGO antibodies [20‐23], (3) the patient presented herein had a lung adenocarcinoma found on tumor screening, suggesting a possible immune paraneoplastic origin [23]. The case reported herein thus suggests that CSF GFAP-Abs should be added to the screening tests of acute/subacute sensory neuronopathy, particularly when a CNS symptom is associated.

Regarding cranial nerves, nerve VI was the most frequently affected. Cranial nerve involvement could also be due to an inflammatory process as suggested by the cranial nerve gadolinium enhancement observed in around one-third of the patients herein. However, nerve VI lesions may not always rely on an autoimmune inflammatory mechanism. Indeed, this cranial nerve is known to be highly sensitive to increased intracranial pressure, and intracranial hypertension is described in patients with CSF GFAP-Abs [14, 24]. Moreover, papilledema, a common feature of intracranial hypertension, was found in four of the patients with nerve VI involvement described herein (40%). None of these patients benefited from CSF pressure measurement. While a direct link between GFAP-Abs and PNS involvement can be suggested in patients with sensory neuronopathy or demyelinating polyradiculoneuropathy, it is more difficult to explain such a link in patients with isolated cranial nerve involvement or radiculitis. Indeed, in such cases, the PNS involvement could be due to a contiguous consequence of the inflammation of brainstem and spinal cord leptomeninges where the roots originate.

Another important finding is the possibility that PNS involvement might be the first manifestation of the disease, prior to CNS symptoms. However, this situation was rare in the present cohort, occurring in 3/25 patients (12%) with PNS involvement. In contrast, five of six patients in a previously published cohort of patients with GFAP-Abs and PNS involvement had inaugural peripheral nerve symptoms [9]. The onset was acute or subacute in all except one of the patients presented herein, which is in line with the previous cohort [9].

The comparison of patients with CSF GFAP-Abs and PNS involvement to those with isolated CNS involvement showed that patients with PNS involvement had more bladder dysfunction probably due to cauda equina impairment. Areflexia tended to be more frequent in the patients with PNS involvement, but the difference was not statistically significant. It could be explained by a lack of data in the group with isolated CNS involvement, where the frequency of areflexia is expected to be low. Interestingly, the type of CNS involvement, especially the frequency of myelitis, was similar in both groups. Finally, the demographic characteristics of the patients with PNS involvement herein were concordant with those of previously published cohorts of GFAP-Abs patients, in which the median age at onset ranged from 44 to 54 and affected between 43 and 68% of female patients [16‐18, 25, 26].

We believe that CSF analysis is a decisive tool in the diagnosis of suspected GFAP-Abs-related disease, as pleocytosis and elevated protein level were identified in all the patients herein. In line with previous data, we consider that the absence of pleocytosis and high protein level should lead clinicians to rule out the diagnosis of GFAP-Abs-related disease [14]. Also, root or/and leptomeningeal contrast enhancement was frequently present in patients with abnormal ENMG (81%). Such a radiological pattern should lead to consider the diagnosis [12].

In the present cohort, the 7/25 patients with PNS involvement and CSF GFAP-Abs who were concomitantly tested in the serum all had a negative CBA test. It might be considered unusual for patients with PNS disease to be serum negative and CSF positive regarding Abs detection. However, the pathogenicity of GFAP-Abs is not established, and the patients herein mainly had involvement of proximal parts of the PNS (roots and cranial nerves), that pass through the CSF and leptomeninges. In the previous cohorts, 4/6 and 17/21 patients with PNS involvement had GFAP-Abs detected in serum, but all were concomitantly positive for GFAP-Abs in the CSF when available (19 patients) [9, 12]. Of note, serum herein was tested only by CBA, which is the more specific test but for which the exact sensitivity is unknown. In light of these heterogenous data, it is difficult to draw conclusions about the clinical relevance of GFAP-Abs in the serum of patients with PNS involvement.

Three patients of the present cohort diagnosed with radiculopathy were found to have positive CSF tests for infectious diseases, known to cause root injuries (VZV infection and Lyme disease). We concede that these findings may represent important confounding factors. However, we point out that these infectious agents might have played a role in triggering GFAP-Abs-mediated autoimmunity, in addition to their own pathogenicity. Indeed, in the present study, the clinicians in charge of the patients concluded to a triggering role of infections for a GFAP-Abs-related disease in two of these patients. In a previous study by our group, infectious prodromal symptoms were found in 82% of cases, suggesting a triggering role of infections in this autoimmune disease [14]. A severe axonal sensorimotor neuropathy associated with GFAP-Abs encephalitis and Epstein–Barr virus (EBV) DNA in the CSF has been reported, which the authors hypothesized to be a possible trigger for the autoimmune response [11]. Finally, it is well established that infections can trigger some other neurological autoimmune diseases, sometimes by molecular mimicry between antigens, as in Guillain–Barré syndrome following Campylobacter Jejuni or EBV/CMV infections [27], or in NMDAR-Abs encephalitis after herpes simplex virus infection [28, 29]. Regarding the patient with positive VZV PCR in the CSF, he was immunocompetent, did not develop a rash, and presented an undoubtful positive CBA test for GFAP-Abs. The only other patient with a possible alternative cause of PNS involvement was the patient with a history of type 2 diabetes mellitus. However, this patient first developed a central meningitis with fever and headache, and then developed binocular diplopia, proprioceptive ataxia with lower limb apallesthesia, proximal lower limb weakness, and radiculalgia suggestive of PNS involvement. Moreover, the CSF analysis revealed a meningitis and oligoclonal bands. Type 2 diabetes mellitus is not an etiology for sensory neuronopathy. Taken together, these data are not in favor of a PNS involvement due to the type 2 diabetes.

The prognosis of patients with GFAP-Abs and PNS involvement seems to be favorable [12]. In the present cohort, the last follow-up mRS was similar in both patients with or without PNS involvement, suggesting that the PNS involvement does not worsen the prognosis. Moreover, improvement of peripheral signs and ENMG was frequent during follow-up. However, severe sequelae might also persist over months in rare cases of severe neuropathy, as reported in the other studies [11, 12, 18]. An early diagnosis, allowing an early immunosuppressive treatment, is probably an important factor for PNS prognosis.

The present study has some limitations. First, it is a retrospective study, but detailed clinical data were provided by the clinicians in charge of the patients using a structured questionnaire. Then, GFAP-Abs testing was performed in patients recruited by the two French reference centers for CNS inflammatory diseases who were mainly patients presenting with signs of encephalitis or myelitis. Patients presenting with isolated suspected inflammatory PNS involvement in the same time window were mainly not tested for GFAP-Abs, causing a possible selection bias and underestimation of GFAP-Abs in these patients.

In conclusion, this retrospective cohort study confirmed that PNS involvement is not a rare feature in GFAP-Abs-related disease. Patients typically present with acute or subacute cranial nerve injury and/or lower limb radiculopathy. These manifestations are mainly associated with CNS symptoms but may, in rare cases, present initially as isolated. Bladder dysfunction is an important clue suggesting a PNS involvement. In future prospective studies, screening GFAP-Abs in patients with inflammatory PNS disease of unknown etiology, associated or not with CNS symptoms or meningitis, may be useful to better define the spectrum of GFAP-Abs-related disease phenotypes.