BACKGROUND: WHERE DO WE STAND?
- History of immune ataxias: milestones
Immune-mediated pathophysiological mechanisms frequently affect the cerebellum, leading to progressive ataxia characterized by dysmetria in both motor and cognitive domains [
1,
2]. The first documentation of patients with immune-mediated cerebellar ataxias (IMCAs) originated from Charcot [
3]. In a well-documented lecture on multiple sclerosis (MS) delivered in 1868, he described the presence of cerebellar ataxia (CA) in patients with MS, now known as the Charcot’s triad (intention tremor, scanning speech, and nystagmus). Another historical milestone was the first report of paraneoplastic cerebellar degeneration (PCD) by Brouwer [
4] in 1919, in which he described the association of CA with ovarian cancer.
In the 1980s, the identification of autoantibodies targeting cerebellar neurons led to a breakthrough in IMCAs. First, the discovery of anti-Yo antibody (Ab), an autoantibody described in a patient with ovarian cancer, demonstrated the immune nature of the insult resulting in PCD [
5]. This discovery was followed by the identification of specific autoantibodies, including anti-Hu, anti-Tr, anti-CV2, anti-Ri, anti-Ma2, and anti-VGCC Abs, often associated with specific types of neoplasms, including breast, uterine, ovarian, small cell lung carcinoma, and Hodgkin’s lymphoma [
6]. Second, the association of otherwise idiopathic CAs with autoantibodies against the cerebellum was also reported in patients without evidence of cancer [
7-
11]. Two main clinical entities have been established since then, based on specific clinical features and types of associated autoantibodies: gluten ataxia (GA) by Hadjivassiliou et al. [
12] and anti-glutamic acid decarboxylase 65 Abs (GAD 65 Abs)-associated cerebellar ataxia (antiGAD ataxia) by Honnorat et al. [
13].
Finally, in 2008, Hadjivassiliou et al. [
14] proposed a new clinical entity: primary autoimmune cerebellar ataxia (PACA). This entity encompasses all ataxic patients who exhibit features of IMCAs but with a serological profile that does not match any of the known main etiologies. Diagnostic criteria have been proposed recently [
15]. Thus, it has become easier to apply this diagnosis for such atypical patients to consider the use of immunotherapy.
- Classification
IMCAs gather diverse etiologies, and the pathophysiology is not restricted to the cerebellum, also targeting cerebellar-related structures/connections in the central nervous system (CNS). The diversity renders the classification of IMCAs challenging. We previously proposed a classification based on 2 criteria: 1) whether the cerebellum is the main target of autoimmunity, and 2) whether autoimmunity is generated by a known trigger or not (
Table 1) [
8-
11].
In MS, for example, the cerebellum is one of many targets within the CNS. Pure CA types, in which the cerebellum or its related structures are the sole target of autoimmunity, include GA, postinfectious cerebellitis (PIC), PCD, opsoclonus myoclonus syndrome (OMS), anti-GAD ataxia, and PACA. Thus, the term IMCAs is reserved for mainly pure or primarily CA, thus excluding more widespread autoimmune damage such as that seen in the context of MS. Although autoimmune limbic encephalitis can sometimes be characterized by CA in addition to cognitive/behavioral deficits and seizures, these conditions should not be categorized as IMCAs.
Pure or primarily CA etiologies are divided into two groups: 1) etiologies in which autoimmunity is triggered by other conditions, such as infection (e.g., PIC), neoplasm (e.g., PCDs), and gluten sensitivity (GA), and 2) etiologies in which autoimmunity is not triggered by any other condition (e.g., anti-GAD ataxia). When autoimmunity is triggered by other conditions, removal of the trigger should be the first line of therapy. This approach is important because categorization based on the autoimmune triggering factors can also offer therapeutic strategies [
9-
11].
In the case of PACA, immune-mediated mechanisms are strongly suspected, but the serological profile of the patients does not match any of the known diseases in IMCAs [
14,
15]. Thus, this group is likely to be heterogeneous and probably includes several pathophysiological immune-mediated mechanisms.
- Prognosis and pathophysiology
In keeping with the diverse etiologies, IMCAs show characteristically diverse clinical courses (
Figure 1) [
10,
11]. These diverse clinical courses result in two problems when considering therapeutic strategies: the need for early intervention and the unclear immune-mediated pathophysiology.
Cerebellar reserve and early intervention
GA, PIC and the subacute type of anti-GAD ataxia show good response to immunotherapy (full or partial recovery and stabilization) (
Figure 1). The good prognosis is in part due to the benign transient autoimmune insult, which is also seen in other autoimmune diseases. Alternatively, the reversibility can also be attributed to specific cerebellar capacity for compensation and restoration following the immune insult [
16,
17]. We termed this capacity the cerebellar reserve [
16,
17]. Various forms of synaptic plasticity and redundant mossy fiber-mediated inputs constitute the cerebellar reserve [
16,
17]. Notably, the extent of progression of the pathophysiological mechanisms determines the degree of reversibility, suggesting the existence of a threshold. Although various immunotherapies can arrest the progression of the damage under this threshold, such intervention may not be accompanied by clinical improvement. Thus, immunotherapies should be introduced during the period when the cerebellar reserve is above this threshold. Such a threshold can be understood as the limit required to obtain a sufficient level of activity in the cerebellar circuitry.
Autoimmune mechanisms and response to treatment
The chronic subtypes of anti-GAD ataxia and PCD show poor prognosis (relapse and gradual or rapid progression in
Figure 1), which reflects poor responsiveness to immunotherapy. The development of various autoimmune diseases can be attributed to common elementary processes [
18], including autoimmunity triggered by deficits in immune tolerance or molecular mimicry, increased permeability of the blood-brain barrier or blood-nerve barrier, pathogenic actions of T cells (Th1/17 cells and CD8 T cells) or autoantibodies, and exacerbation of local neural inflammation (
Figure 2). However, unlike the advances in the field of MS, little is known about therapeutic autoimmune targets in IMCAs.
In conclusion, the clinical entity of IMCAs has been established during the past three decades. To avoid missing therapeutic opportunities at the early stage, we should carefully promote IMCAs in the differential diagnosis of all ataxias. We also need more research directed toward the identification of new autoimmune pathogenesis-based therapies.
GENERAL ISSUES ON DIAGNOSIS AND IMMUNOTHERAPIES
- Prevalence
Hadjivassiliou et al. [
19] (2017) investigated the prevalence of IMCAs on 1,500 UK patients with progressive ataxia. The authors reported that 30% of the patients had familial/genetic ataxia, although some did not show evident family history, and 9% of the patients had a cerebellar variant of multiple systemic atrophy (prevalence out of total progressive ataxic cases). Apart from the above, 25% had definite IMCAs; 20% had GA, and 2% had PCD, while 2% had anti-GAD ataxia, 1% had PIC, and <1% had OMS. Interestingly, 19% of the patients were classified as idiopathic sporadic ataxia. This category probably included a large number of patients with PACA [
14,
15].
Among other groups, Gebus et al. [
20] (2017) reported a smaller number of IMCAs patients (2 with PCD and 1 patient with PIC among 80 patients). In a systematic study of 684 South Korean patients with progressive ataxia, only 21 had IMCAs; 14 patients with PCD, 3 patients with PIC, and 4 patients other immune causes [
21]. The discrepancy might be attributed to a number of different factors: The high prevalence of IMCAs in Hadjivassiliou’s study [
19] was due to a large number of patients with GA. GA was first characterized and reported by this group and there is therefore the potential for referral bias. Another important factor is the serological test utilized to make a diagnosis of GA (please see section on GA). More recently, the prevalence of GA in Asian countries has been found to be higher than was previously estimated [
22,
23]. For example, patients with GA have recently been reported in Japan [
23]. A survey in China also showed that the prevalence rates of positive-anti-transglutaminase 6 (TG6) IgA were 35% in 100 sporadic CA patients and 9.8% in 51 healthy controls (
p = 0.001) [
23]. Another factor is the so called “idiopathic” unknown etiology diagnosis. As discussed above, we have argued that this group includes a large number of patients with PACA [
8-
11]. The newly established diagnostic criteria on PACA [
15] will be helpful in resolving this issue.
Further studies are needed to examine the prevalence of IMCAs or GA in progressive CAs. However, early diagnosis and intervention is critical during a period when the cerebellar reserve is preserved. In this regard, clinicians should always consider the possibility of IMCAs in the differential diagnosis of patients developing CAs.
- Clue to diagnosis
Patients with IMCAs exhibit common clinical manifestations in spite of divergent etiology [
10,
11]. The onset of CA is usually acute/subacute or sometimes chronic/insidious. The presence of other autoimmune disorders is common in some patients. The main clinical feature is gait ataxia, which influences standing and steady walking and is associated with mild limb incoordination and dysmetria, oculomotor ataxia, and scanning speech. In keeping with these symptoms, MRI characteristically shows vermian atrophy, which can be a clue for the differential diagnosis from degenerative or genetic CAs (
Figure 3A). However, it should be acknowledged that MRI can be normal. As the pathology progresses, atrophy becomes more evident, preferentially affecting the vermis. CSF studies sometimes show the presence of pleocytosis or oligoclonal bands. There may be a context of systemic impaired immunity (affecting skin, joints or other organs).
- Autoantibodies
As discussed earlier in the section on the history of IMCAs, identification of autoantibodies is important in the diagnosis (
Table 2) [
10,
11]. Immunohistochemistry may reveal the binding of autoantibodies toward cerebellar components; therefore, if available, it can add further evidence of the diagnosis. Autoantibodies in IMCAs can be divided into two categories: 1) autoantibodies suggestive of specific etiologies, and 2) nonspecific autoantibodies found in other neurological and systemic conditions, including CAs, which provide only possible autoimmune pathophysiology. The former types of autoantibodies include antigliadin and TG6 Abs in GA, and anti-Yo, Hu, CV2, Ri, and Ma2 Abs for PCD [
5,
6,
8,
10,
11,
24]. The nonspecific autoantibodies are also subdivided into two subcategories: autoantibodies that are assumed to have pathogenic roles and the nonpathogenic autoantibodies (i.e., a diagnostic marker).
The majority of the pathogenic autoantibodies include autoantibodies toward ion channels or ion channel-related proteins (Ca++ or K+ channel), glutamate receptors, or GABA synthesis enzyme (GAD), some of which have been confirmed as pathogenic in both in vitro and in vivo preparations [
24-
34]. Impairments of neural excitability and synaptic transmissions can occur throughout the nervous system, such as limbic encephalitis [
24-
34]. We previously proposed a nomenclature based on the pathogenic role of the autoantibody, i.e., the pathogenic Ab to be coined to the name of the clinical entity (e.g., anti-GAD ataxia) [
10,
11].
Additionally, some autoantibodies are reported to be associated with CAs in only a few patients. These autoantibodies are classified into a category of not-well-characterized autoantibodies. The significance of these autoantibodies remains to be determined, but their presence can be a clue to the autoimmune etiology of the ataxia.
- Principles of treatment
When autoimmunity is caused by another trigger, priority should be given to the treatment of the underlying condition: for example, gluten-free diet (GFD) in GA and surgical excision or chemotherapy and radiotherapy of the neoplasm in PCD (
Figure 4) [
7-
11]. For other IMCAs, immediate immunotherapy is recommended (
Figure 4) [
7-
11].
Depending on the rapidity of progression, induction immunotherapy can be used to arrest progression of CA, followed by maintenance immunotherapy to prevent relapse [
9-
11]. Various immunotherapies have been used, ranging from intravenous immunoglobulins (IVIg), glucocorticosteroids, mycophenolate, plasmapheresis, and rituximab, either alone or in combination, and the selection is often based on the etiology and presentation [
9-
11]. These therapies should be introduced during the period when the cerebellar reserve is still preserved.
To establish a therapeutic rationale, the following two issues need further consideration. First, to date, there are no large-scale randomized studies comparing therapeutic strategies for each IMCA [
9-
11]. Second, the effects of each therapeutic intervention should be quantified with more sensitive biomarkers if possible at an early stage of the disease when MRI shows no atrophy. Recent studies highlighted the potential usefulness of MR spectroscopy as a sensitive biological marker of the response to treatment in CA patients [
10,
11]. The ratio of N-acetylaspartate (NAA)/creatine (Cr) area is decreased in patients with CAs (
Figure 3B) relative to the control and increases in those who respond to immunotherapy and exhibit clinical improvement [
10,
11].
PARTICULAR ISSUES IN COMMON ETIOLOGIES
This section focuses on main etiologies in IMCAs (
Table 3).
- Gluten ataxia
GA is defined as sporadic CA associated with gluten sensitivity [
12]. The autoimmunity is triggered by the sensitivity to gluten found in wheat, rye and barley. Gluten, a family of protein containing grains, is composed of gliadin and glutenin.
Diagnosis and treatment
GA affects individuals in their 40–50s and exhibits either chronic or insidious onset [
7-
11]. Gluten-sensitive enteropathy (coeliac disease) is seen in approximately half of the patients [
7-
11]. The human leukocyte antigen (HLA) type DQ2 is detected in 70% of the patients, and association with other autoimmune diseases, such as thyroiditis, type 1 diabetes mellitus, and pernicious anemia, is common [
7-
11]. Patients present with gait ataxia and variable degree of limb ataxia, scanning speech [
7-
11] and, consistently, MRI shows vermian involvement [
35]. GA can be associated with CSF abnormalities, with oligoclonal bands in up to 50% of patients [
7-
11]. The association of sensorimotor axonal neuropathy and focal myoclonus, rarely palatal tremor, is sometimes observed [
7-
11].
Diagnosis is based on the presence of serological evidence of gluten sensitivity, which is assessed using autoantibody testing. Native anti-gliadin Ab (AGA) remains to be the most reliable test for the diagnosis [
10,
11]. However, the cutoff level for the diagnosis of GA should be noted [
36]. In GA patients without enteropathy, the main autoimmune response occurs within the CNS, not the periphery, resulting in low levels of serum AGA [
10]. The calibration process used in certain commercially available AGA kits is based on the use of serum from patients with celiac disease (CD) and, therefore, the cutoff level is too high for the detection of GA [
10]. Anti-transglutaminase 2 (TG2) Ab, found in patients with CD, is negative in 53% of GA patients free of enteropathy [
36]. However, anti-TG6 Ab is positive in 72% of GA patients on the basis of AGA [
37]. Since TG6 is primarily expressed in the CNS, anti-TG6 Ab could be an important and possibly more specific biomarker [
10,
11].
GFD is the first line of therapy for GA since it can eliminate antigens that can trigger immune-mediated mechanisms, similar to the strategy used in CD [
7-
11]. One large-scale study based on 43 patients with GA showed significant improvements in CA and a decrease in AGA in patients who adhered to GFD, compared with those who did not [
38]. Other reports described the effectiveness of IVIg in nonresponders to GFD [
39]. However, it is now considered that the lack of response to GFD is due to either poor adherence to GFD or hypersensitivity to gluten, where a small amount of gluten present in commercially available gluten-free food or due to cross-contamination can cause strong autoimmune reactions perpetuating the cerebellar damage [
9-
11]. Persistently high levels of AGA are present in the above two groups of patients [
40,
41]. On MR spectroscopy, patients on strict GFD show an increase in the ratio of NAA/Cr area in the cerebellar vermis, whereas no such increase is seen in patients on GFD with persistently positive antibodies and those who do not adhere to GFD [
42]. Taken together, in patients with persistently high AGA titers or no improvement in MR spectroscopy, further dietetic review by an expert dietitian should be considered before switching to immunotherapy [
10,
11]. When CAs cannot be controlled with GFD, maintenance therapy with IVIg or immunosuppressants (e.g., mycophenolate mofetil, cyclosporin, and cyclophosphamide) is recommended [
9-
11].
Autoimmune pathophysiological mechanisms
Loss of immune tolerance is considered the main mechanism, as in coeliac disease [
43]. Digested gliadin is deamidated and cross-linked by TG2, leading to the creation of an immunostimulatory epitope for HLA-DQ2 or HLA-DQ8 on antigen-presenting cells. These epitopes are presented to CD4+ T cells, from which cytokines are released to facilitate the production of antibodies against gliadin and TG2 (TG2-gliadin complex) (
Figure 5).
The neural damage appears to be due to, at least in part, the actions of autoantibodies against tissue transglutaminase, a calcium-dependent enzyme located in intra- and extracellular spaces. Both anti-TG2 and anti-TG6 Abs recognize and cross-react with neurons. Intraventricular administration of anti-TG2 or anti-TG6 Abs elicits ataxia in mice [
44]. TG2 activity is upregulated through calcium binding, and anti-TG2 Abs bind preferentially to the calcium-activated enzyme conformation [
45,
46]. Taken together, antibody-induced deficits of TG2 in its enzymatically active conformation can cause cellular dysfunction. The role of TG6 in cerebellar function is also supported by the observation that mutations in the TGM6 gene can cause CA (spinocerebellar ataxia type 35) [
47].
Although the mechanisms of cell-mediated brain damage remain uncertain, the above immunological studies add support to the therapeutic rationale of gluten avoidance in food and reductions of anti-TG2/6 and AGAs.
Myoclonic ataxia with refractory celiac disease
CA associated with cortical myoclonus is a rare subtype of gluten sensitivity-related neurological disorder [
10,
11,
48]. Although the myoclonus is of cortical origin, hyperexcitability of the cerebral cortex is elicited by cerebellar dysfunction [
10,
11,
48,
49]. Characteristically, this subtype shows resistance to GFD. The neurological refractoriness is associated with residual enteropathy, which is detected by repeat duodenal biopsies [
9-
11]. All such patients require GFD plus immunosuppression, usually with mycophenolate and in some instances cladribine, in addition to anti-epileptic drugs for the control of the myoclonus [
9-
11]. The prognosis remains poor [
10,
11].
- Postinfectious cerebellitis
PIC is defined as cerebellar inflammation induced by immune-mediated mechanisms triggered by viral or bacterial infection [
7,
8,
10,
11,
50]. Some groups have termed this type of autoimmune ataxia as para-infectious cerebellitis or acute cerebellar ataxia [
10,
11,
50,
51]. However, when the cerebellar inflammation is caused by direct invasion of viral or bacterial microorganisms, it is called acute cerebellitis [
10,
11,
52].
Diagnosis and treatment
PIC affects mostly young children after an episode of infection, usually viral infection, most commonly varicella [
7,
8,
10,
50-
52]. Other viruses, such as Epstein-Barr virus, Coxsackie virus, influenza A and B virus, parainfluenza virus, rubeola virus (measles), rubulavirus (mumps), and rubella virus (German measles), and bacteria, such as
Corynebacterium diphtheriae (diphtheria),
Bordetella pertussis (whooping cough),
Salmonella typhi (typhoid fever),
Legionella (Legionnaires disease),
Leptospira (leptospirosis), and
Mycoplasma (mycoplasmosis) can be involved [
7,
8,
10,
50,
52].
Borrelia (Lyme disease) is also implicated. One large-scale study based on 73 patients [
53] showed that 60% of the patients were between 2 and 4 years of age, and 25% of these patients had varicella, 52% had other viral infections, and 3% developed PIC after immunization. The mean latency between infection and the onset of CAs was 9.9 ± 7.9 days. However, some patients (19%) showed no definite preceding infection.
Patients develop mainly acute-onset afebrile gait ataxia, meningeal signs, high intracranial pressure with or without extracerebellar manifestations, such as temporary clouding of consciousness, seizures, altered mental status (e.g., extreme irritability), or extracerebellar focal signs [
53]. The presence of these clinical features is often suggestive of direct infective etiology rather than immune-mediated mechanisms (PIC) [
52]. Additionally, mild behavioral changes, such as mild irritability, hyperactivity, moodiness and whining, are sometimes noted by parents and often correlate with the severity of CA [
52]. These manifestations could reflect cerebellar cognitive and emotional disorders. CSF studies show pleocytosis in most of the patients (roughly equal numbers of granulocytes and lymphocytes in approximately 75% of the patients, and sometimes lymphocyte predominance in the remaining cases) and high CSF/serum IgG index in half of the children. Oligoclonal bands are sometimes present [
53]. MRI is usually negative, with no signs of atrophy or abnormal intensity areas [
53]. Nevertheless, adult patients can develop cerebellar atrophy after several months.
Since PIC is self-limiting, close observation of the patient is recommended [
10,
11]. One large-scale study reported full recovery of CAs in 72% of 60 pediatric patients within two months [
53]. Only when the CAs persist or progress are various combinations of immunotherapeutic agents available for use [
10,
11,
50].
Autoimmune pathophysiological mechanisms
Autoimmune mechanisms remain obscure. For example, it is not certain whether molecular mimicry, observed in GuillainBarré syndrome (GBS, a typical infection-induced neuroimmune disease), also operates in PIC. Although the lack of cerebellar atrophy and self-limiting clinical course suggests immunemediated cerebellar functional disorders, the exact mechanisms remain unclear. A recent study [
54] described the unique actions of anti-GluRδ Ab, which is associated with PIC in some patients [
26]. In one experimental study, injection of an antibody toward the putative ligand-binding site of GluRδ2 resulted in endocytosis of AMPA receptors and attenuated synaptic transmission and ultimately led to the development of ataxic behavior in mice [
54]. Interestingly, the endocytosis of AMPA receptors, which has a transient nature, could explain the reversibility in PIC. Thus, further studies are needed to understand how autoimmune reactions transiently disturb cerebellar function.
- Opsoclonus myoclonus syndrome
OMS has divergent categories, including paraneoplastic, postinfectious, and idiopathic [
10,
11,
55-
57]. It affects mainly children and is especially associated with neuroblastoma [
10,
11,
55-
57].
Diagnosis and treatment
Opsoclonus is characterized by involuntary repetitive, random and rapid eye movements in both horizontal and vertical directions. Action myoclonus is characterized by irregular jerky movements primarily in upper limbs [
10,
11]. Since OMS is usually associated with CA, it is known also as opsoclonus myoclonus ataxia syndrome [
10,
11]. In keeping with the diverse etiology, some differences are observed in the clinical profiles of patients with OMS. The presence of subacute vertigo at onset, together with CSF abnormalities (high leukocyte count and protein content), is more common (half of the patients) in idiopathic OMS than in paraneoplastic OMS. Encephalopathy is more common in paraneoplastic OMS, although it is also found in at least a few cases of idiopathic OMS [
55]. Anti-Ri Ab, an onconeural Ab, is positive in paraneoplastic patients, mainly those with breast cancer [
55]. Onconeural Ab and PET scan should be performed to confirm any association with neoplasm [
7,
10,
11].
In cases of paraneoplastic OMS, any associated neoplasm should be removed first, if possible, followed by a combination of immunotherapies (corticosteroids, IVIg, plasmapheresis, immunosuppressants, and rituximab) [
10,
11,
55-
57]. Postinfectious OMS and some idiopathic OMS are self-limiting. Thus, if symptoms persist, immunotherapy should be introduced [
10,
11,
55-
57]. Several studies showed good prognosis in idiopathic OMS compared with paraneoplastic OMS [
55-
57]. For example, when a good response to immunotherapy was defined as recovery of the modified Rankin Score to ≤ 2 [
55], a good response was reported in 39% of patients with paraneoplastic OMS and in 84% of those with idiopathic OMS. Relapse occurred in 24% of patients with paraneoplastic OMS and in 7% of those with idiopathic OMS [
55].
Autoimmune pathophysiological mechanisms
Autoimmune pathogenesis is based on the presence of CD19+ B cells and gamma-delta T cell subsets in the CSF [
58] as well as perivascular collection of lymphocytes in autopsy cases [
59]. It should be noted that unlike PCD, neither cell loss nor extensive T cell infiltration is characteristically evident in paraneoplastic OMS [
60,
61], which suggests immune-mediated functional disorders in the cerebellar motor controls.
Characteristically in OMS, the autoimmune-induced dysfunction appears to occur not only within the cerebellum but also in inhibitory controls from the cerebellar cortex on cerebellar nuclei neurons. It is assumed that the net inhibitory output from Purkinje cells (PCs) in the cerebellar cortex is attenuated in both opsoclonus and action myoclonus. In ocular motor controls, this elicits hyperactivation of the fastigial nucleus neurons, which in turn facilitates the activities of excitatory burst neurons, leading to a chaotic cascade of events [
62]. In the cerebro-cerebellar loop, the pathway of the dentato-thalamo-cortical tract is hyperactivated, in which even subthreshold inputs can easily activate the motor cortex, generating the action myoclonus in limb movements. Consistently, action myoclonus is of cortical origin [
48,
49], and a low correlation between the cerebellum and motor cortex is observed in functional MRI [
63].
In conclusion, there is no explanation on how the autoimmune processes, typically triggered by diverse pathologies, such as neoplasm and infection, commonly produce a unique functional deficit, decrease the inhibitory output from the cerebellar cortex and generate hyperexcitability of their targets.
- Paraneoplastic cerebellar degeneration
PCD comprises cerebellar degeneration caused by neoplasmtriggered autoimmunity [
5-
8,
24,
64]. The presence of characteristic onconeural autoantibodies is not only diagnostic for PCD but also provides a clue to the specific type of neoplasm underlying PCD (
Table 2). Although aggressive chemotherapies and immunotherapies have been attempted, the prognosis remains poor unless the underlying cancer can be completely and rapidly eradicated.
Diagnosis and treatment
PCD shows acute or subacute onset and is sometimes preceded by prodromal clinical symptoms such as nausea, vomiting, and dizziness, resembling viral infection-related disease [
8,
24,
64]. Subsequently, patients show gait ataxia, which is followed by pancerebellar involvement [
8,
24,
64]. CSF studies show changes consistent with inflammation, including moderate lymphocytic pleocytosis, high protein concentrations, high IgG index, and oligoclonal bands [
8,
24,
64]. Identification of the characteristic autoantibodies helps in the establishment of diagnosis, although seronegative cases have been described [
6]. Since subacute CAs are listed as classical symptoms defined in the 2004 diagnostic criteria [
65], a definite diagnosis is based on 1) confirmation of the presence of cancer, which develops within 5 years of the diagnosis of CA, or 2) appearance of well-characterized onconeural Abs (anti-Yo, anti-Hu, anti-CV2, anti-Ri, anti-MA2) [
65]. CA is the first manifestation of neoplasm in up to 70% of the patients with PCD [
64], highlighting the need for systemic evaluation for malignant tumors following the identification of onconeuronal autoantibodies.
Following confirmation of the diagnosis, anti-neoplasm therapies should be introduced immediately to prevent metastasis and remove antigen(s) driving the autoimmunity [
8,
24,
64]. The selection of cancer treatment, such as surgery, radiotherapy and/or chemotherapy, is based on the type of the neoplasm [
8,
24,
64]. Regardless of the selected cancer treatment modality, various types of immunotherapies (single or combinations of corticosteroids, IVIg, plasmapheresis, immunosuppressants, and rituximab) should be administered.
In spite of these treatment protocols, the prognosis is poor [
64,
66-
68]. The response to immunotherapy is generally unsatisfactory [
64,
66-
68], although therapeutic benefits have been reported in some case reports [
10], and better prognosis was noted in patients with anti-Tr Ab [
69]. The study of Peterson et al. [
70] on 22 patients with anti-Yo Ab reported therapeutic improvement in less than 10% of the trials (one in four trials of plasmapheresis, one in 17 trials of corticosteroids, and none in four trials of cyclophosphamide). Furthermore, Rojas et al. [
71] also reported no benefits from the combinations of immunotherapies in their study of 23 patients with PCDs.
There have been three long-term studies on survival time [
69,
72,
73]. Candler et al. [
72] examined the clinical courses of 63 patients with paraneoplastic neurological syndromes, including 13 patients with PCDs. Three of the 13 (23%) patients with PCDs died before the final follow-up, and the mean survival time from onset of neurological symptoms was 42 months (95% confidence interval: 32–52). Keime-Guibert et al. [
73] reported a long-term follow-up study of 16 patients with PCDs. When the clinical benefits were determined as a more than one point change in the Rankin Score, none of their patients met this criteria (three patients showed stabilization). The median survival time from the first therapy was 10.2 months (range 2–38). Fourteen of the 16 (88%) patients died during the follow-up. The cause of death was cancer-related in 9 patients, neurological-related in 3 patients, and unknown in the remaining patients. Shams’ili et al. [
69] examined the relationship between effectiveness of treatment and associated antibodies in their study of 50 patients with PCDs. The median survival times from diagnosis were 113 months in patients positive for anti-Tr Ab, > 69 months in those positive for anti-Ri Ab, 13 months in those positive for anti-Yo Ab, and 7 months in those positive for anti-Hu Ab. It should be noted that even patients with anti-Tr Ab suffered severe disability or dependence. Taken together, the survival time is generally short once the diagnosis is established, excluding a few patients with anti-Tr Ab.
The poor prognosis could be attributed to the remnant tumor tissue and metastasis, which could cause persistent activation of the immune system and strong autoimmune response. Currently, clinical trials aimed at stopping the progressive autoimmune insults are being conducted. In one such trial, the combination of IVIg or plasmapheresis with cyclophosphamide was reported to be effective in a subgroup of patients [
64].
Autoimmune pathophysiological mechanisms
Pathological studies have identified various inflammatory changes in the early stages of PCD, including perivascular cuffing by lymphocytes, infiltration of lymphocytes within the PC layer, and microglial activation [
74,
75]. The infiltrates consist of both of T and B cells, plasma cells, microglia/macrophage lineages, and reactive Bergmann cell gliosis [
76]. A similar infiltration, albeit of lower magnitude, was found in the dentate nucleus and the pons [
77,
78]. Disease progression is associated with massive loss of PC without inflammation (inflammatory cells or deposits of immunoglobulins and complements) [
76,
79,
80]. Since the loss of PC is often very rapid, such a burnout stage can be observed in some patients at an earlier stage [
76,
79].
Yo (Cdr2) and Hu onconeural antigens are intracellularly located in tumor and neuronal cells [
81]. This favors molecular mimicry mechanisms. First, whether autoantibodies that target intracellular antigens Cdr and Hu proteins play pathogenic roles in the development of the inflammatory cerebellar lesions is still a matter of debate today. Previous studies showed that anti-Yo Ab toward Cdr2 was internalized and induced cell death through apoptosis [
82,
83]. In this regard, Cdr2 encodes a leucine zipper motif and interacts with another leucine zipper motif on c-Myc, the nuclear transcription factor, to regulate nuclear levels of c-Myc and c-Myc-associated signaling activities. Cdr2-specific anti-Yo Ab inhibits the interaction of Cdr2 with c-Myc, possibly allowing excess c-Myc to enter the nucleus and impair cell cycle signaling involved in cell apoptosis. The binding of anti-Yo Ab to Cdr2 can also disrupt interaction with the mortality factor-like proteins MRGX and/or NFkb, leading to changes in transcriptional activity [
83]. Another study showed that administration of anti-Hu and Yo induced the expression of adhesion molecules and more intense cell differentiation rather than cell death [
84]. Furthermore, another histochemical study showed a lack of both IgG deposits and B cell infiltration [
85]. A study of passive transfer of onconeural Abs that targeted intracellular antigens and immunization using protein or DNA failed to induce ataxic symptoms in animals [
86-
88]. Thus, many researchers believe that antibody-mediated mechanisms are not the main autoimmune mechanisms underlying PCDs associated with anti-Yo and anti-Hu Abs [
64,
85].
Patients with anti-Yo (Cdr2) Ab or anti-Hu Ab show accumulation of Cdr2- or Hu-specific T cells in blood and CSF, respectively [
89-
91]. Notably, the lymphocytes found in tumors, peripheral blood, and affected nervous tissue are thought to originate from the same clone population [
81,
92]. The infiltrated lymphocytes in the cerebellum are mainly CD3+ and CD8+ T cells [
78,
92], and only a few are CD4+ lymphocytes [
93]. In general, CD4+ and CD8+ T cell interaction is necessary to induce inflammation, where CD4+ T cells exhibit a Th1 phenotype and secrete IFN-γ and TNF-α [
94]. These studies suggest that PCDs are mediated mainly by a CD8+ T-cell immune response toward an autoantigen recognized by onconeural Ab [
64,
81,
95,
96]. The cytotoxic lymphocytes appear to be aggressive toward cells presenting Cdr2 or Hu antigens and possibly induce cell death through one or more of the following mechanisms; 1) activation of TNF and FasL receptors, which triggers the apoptotic cell death cascade, and 2) excretion of granules that destroy cells through the actions of granzyme-B, perforin, and T-cell restricted intracellular antigen-1 (
Figure 6) [
81].
Despite the above evidence suggesting the involvement of cell-mediated mechanisms in the development of PCDs, the transfer of lymphocytes from patients with anti-Yo Ab did not elicit any PC damage [
97,
98]. In addition, although all small cell lung carcinoma expresses the Hu antigen, and a clinically useful tumor immune response can be detected in up to 20% of the patients, the paraneoplastic neurologic syndrome is extremely rare, suggesting the robust tolerance to neural onconeural antigens
in vivo [
99]. Based on this notion, Blachère et al. [
100] argued that CD8+ T cells that target intracellular antigens, in conjunction with CD4+ T cells, can induce tumor autoimmunity, but not neurologic autoimmunity. They proposed that 1) CD8+ T cells can only activate the peripheral immune response and 2) CD8+ T cells can break CNS tolerance to induce PCDs if they interact with both CD4+ T cells and autoantibodies [
100]. This interaction between cell- and autoantibody-mediated mechanisms might provide the rationale for the design of new therapies that can target T and B cells using tacrolimus and rituximab, respectively [
100].
The following mechanisms can also underlie the break of CNS tolerance [
101]: 1) structural modification of certain intracellular proteins, which stimulates the immune system, 2) death of cancer cells and release of intracellular antigens with subsequent exposure to the immune system, 3) persistence of the above autoimmune responses to neoplasm causes secretion of cytokines that elicit vascular hyperpermeability and infiltration of immune cells, and 4) simultaneous downregulation of regulatory T cells.
- Anti-GAD ataxia
GAD is an enzyme that catalyzes the conversion of glutamate to GABA [
102]. GAD has two isoforms: GAD65 and GAD67. The autoantibodies against the smaller isozyme are associated with CA [
102]. Serum and CSF titers of anti-GAD65 Ab are higher in anti-GAD ataxia, at usually more than 10,000 U/mL (or 10 to 100-fold higher), compared to those of patients with type 1 diabetes mellitus (T1DM) [
13,
102]. The triggering factor of this autoimmunity is not clear at present. Anti-GAD ataxia can be associated with other types of IMCAs, such as PCD and GA [
8]. While autoimmunity toward GAD65 affects the entire CNS, the cerebellum is one of the most vulnerable areas [
102].
Diagnosis and treatment
The condition mostly affects women in their 60s, with subacute or chronic/insidious time course [
8,
102,
103]. Some patients show evidence of other autoimmune conditions, such as T1DM, autoimmune thyroid diseases, and pernicious anemia [
8,
102,
103]. Anti-GAD ataxia is sometimes associated with extracerebellar symptoms, including epilepsy, ophthalmoplegia and Stiff-Person syndrome [
8,
102,
103]. The clinical course can be either subacute or chronic [
103]. CSF studies sometimes show oligoclonal bands [
8,
102,
103], whereas MRI shows normal or atrophic changes depending on the duration of illness [
8,
102,
103]. The diagnosis is confirmed by the presence of high titers of intrathecally produced anti-GAD Ab [
8,
102,
103].
The serum level of anti-GAD antibodies in the context of the diagnosis of anti-GAD ataxia merits further consideration. There is currently no consensus on the threshold used for “low titer” and “high titer” levels of anti-GAD antibodies. In cases of anti-GAD Ab more than 2,000 U/mL (high titer), one can safely consider anti-GAD ataxia. However, titers in the range of 100–200 U/mL in an appropriate clinical/biological/neuroimaging context may be observed in immune ataxias. Some patients with CA and anti-GAD Ab less than 100 U/mL have been reported to show benefits to immunotherapies [
102]. Anti-GAD Ab less than 100 U/mL may raise a suspicion of autoimmune diathesis and require therapy. It is also important to keep in mind that various laboratories use different techniques for the assays.
The aim of any induction therapy should be to minimize progression of CA [
9-
11,
102]. Thus, various immunotherapies, alone or in combination (more common), are used until remission [
9-
11,
102]. Maintenance therapies are usually added to prevent any relapse [
9-
11,
102]. Both the induction and maintenance therapies include corticosteroids, IVIg, immunosuppressants, plasmapheresis, and rituximab, either alone or in combination [
9-
11,
102]. There appear to be no significant differences in the therapeutic benefits among the above immunotherapies [
9-
11,
102]. The prognosis is better in the subacute type than in the chronic type of anti-GAD ataxia [
103].
Autoimmune pathophysiological mechanisms
The significance of anti-GAD65 has been a matter of debate [
102,
104]. Some researchers have argued that anti-GAD65 Ab have no pathogenic roles in the development of CAs based on the following reasons [
105,
106]: 1) anti-GAD65 Ab plays a pathogenic role in T1DM and various neurological conditions, such as Stiff-Person syndrome, 2) GAD65 is intracellularly located together with GABA transported with the cytosolic phase of GABA-containing vesicles, implying that autoantibodies do not have access to GAD65, and 3) lack of correlation between anti-GAD65 Ab titers and clinical symptoms [
107]. However, recent physiological studies have provided substantial evidence for the pathogenic role of anti-GAD Ab [
102,
104]. It should be acknowledged that these studies did not rule out the secondary involvement of cell-mediated mechanisms.
Epitope-specific actions
Studies using slice-tissue preparations showed that the addition of CSF IgGs from patients with anti-GAD ataxia to the medium resulted in presynaptic inhibition of GABAergic synapses between basket cells and PCs [
108,
109]. Furthermore, in
in vivo preparations, application of CSF IgGs resulted in impairment of cerebellar-mediated modulations, as confirmed by various types of tasks [
110-
112]. Importantly, these pathogenic actions of IgGs from patients with anti-GAD ataxia were elicited by the binding of GA65 with anti-GAD65 Ab itself in an epitope-specific fashion: 1) these actions were abolished after the absorption of antiGAD65 Ab using recombinant GAD65 [
113] and anti-GAD65 Ab elicited no actions in slices from GAD65 knockout mice where inhibitory transmissions were mediated compensatorily by GAD67 [
112], 2) the b78 monoclonal Ab with epitope-specificity in CAs has pathogenic actions, whereas another monoclonal Ab has no such actions in T1DM [
111,
112], and 3) differences in neurological phenotype might also be related to epitope specificity [
111,
112]. Notably, low titer anti-GAD Ab has no pathogenic actions [
102,
104].
Internalization and dissociation of GAD65 with vesicles
Anti-GAD65 Ab is internalized, presumably during exocytosis or endocytosis [
112,
114]. Since exposure of GAD65 is assumed to be during exocytosis, anti-GAD65 could access GAD65 [
102,
104], although the exact mechanisms are still not clear. Anti-GAD65 Ab is reported to impair the association of GAD65 with vesicles, resulting in deficits in GABA packaging into vesicles and shuttling of vesicles to the release sites (
Figure 7) [
112].
From functional disorders to cell death
A decrease in GABA release attenuates the spill-over GABA-induced presynaptic inhibition on glutamate release from neighboring parallel fibers, resulting in a major imbalance between GABA and glutamate and excitotoxicity (
Figure 8) [
115]. Consistently, one autopsy study showed the complete loss of PCs [
116].
Clinical correlation
Previous studies showed that some patients respond well to immunotherapies and that the clinical improvement in CA correlates well with the fall in Ab titers [
9-
11], suggesting that Ab titers better reflect functional disorders rather than cell death. The peak CA might be influenced by secondary cell-mediated mechanisms.
Taken together, the above experimental studies clearly show that anti-GAD65 Ab could serve as a trigger of CA (
Figure 7). These studies also provide a generalized principle regarding antibody-induced neurological diseases; autoantibodies are polyclonal, and a cocktail of these polyclonal autoantibodies determines the clinical profile, depending on the proportions of pathogenic epitopes. In addition, pathomechanisms underlying the excitotoxicity could be generalized in other pathologies, such as ischemic diseases and degenerative CAs (
Figure 8).
- Primary autoimmune cerebellar ataxia
Although the clinical profile spectrum of PACA meets the features of IMCAs (e.g., subacute onset, autoimmune disease history, predominant gait ataxia, association of autoantibodies, and good benefits from immunotherapies), a subgroup of patients present with features that do not meet the criteria of the above common subtypes [
8,
10,
11,
14,
15,
117,
118]. Thus, the spectrum of PACA serves as an umbrella that covers heterogeneous etiologies [
8,
10,
11,
14,
15,
117,
118]. Consistently, various types of less characterized autoantibodies are associated in some patients (
Table 2), and immunohistochemical studies show a variety of staining patterns (60% of the patients) [
14]. Despite the heterogeneity, patients with PACA sometimes show HLA type DQ2, which is observed in certain autoimmune diseases (CD, GA, T1DM, Stiff-Person syndrome, autoimmune thyroid disease and autoimmune polyendocrine syndromes), but not in healthy subjects [
14,
15].
Many types of immunotherapies have been used in PACA. A small number of studies have reported the effectiveness of IVIg, prednisolone, plasmapheresis or rituximab in 9/19 patients at the early stage of PACA [
119]. Progression to wheel-chair dependence was faster in patients with neuronal nuclear and/or cytoplasmic antibody than those positive for plasma membrane protein antibody [
14], suggesting the need for the selection of therapeutic strategy based on the underlying autoimmune pathoetiology. In this regard, MR spectroscopy of the cerebellum could be an important tool in monitoring the response to immunotherapies [
15,
117,
118].