There are 3 million cases of chicken pox each year in the United States and 90% of this infection occurs in children 1–14 years of age [1]. The sequelae of varicella infection in children includes a number of severe complications such as purpura fulminans and thromboembolism [2, 3]. In these conditions, autoantibodies to a variety of phospholipids and protein related to coagulation were reported and the related syndrome was referred to as The Varicella-Autoantibody Syndrome [3]. Encephalitis is another of sequelae of varicella in children that accounts for 20% of hospital admissions due to varicella [4]. A cerebellar syndrome associated with post-infectious and post-vaccinial states that usually begins abruptly 5 to 6 days following the subsidence of a rash, is manifest primarily as ataxia. This syndrome is common following varicella but also appears in measles, rubella and Epstein Barr virus (EBV) infections [1, 5–7]. The course is usually mild and most children recover, but it can progress in some children to more severe disease as manifested by vomiting, dehydration and drowsiness. Although central nervous system complications usually develop within a week after eruption, they may be delayed for up to 21 days.

The pathogenesis of post-varicella cerebellar ataxia is unknown. A number of autoantibodies have been described in other post-viral conditions, including antineuronal antibodies in a case of post-infectious cerebellar ataxia following EBV infection [6], antibodies to Purkinje cells in children with opsoclonus myoclonus associated with neuroblastoma or a prodromal viral illness [5], and antibodies to centrioles in 2 children with ataxia and other central nervous system manifestations following Mycoplasma pneumoniae infection [8]. In a recent study of children with post-varicella ataxia, we found antibodies to cerebral and cerebellar tissue [9].

A family of proteins located at the centrosome that react with human autoantibodies include pericentrin, ninein, enolase, PCM 1 and a newer autoantigen, Mob1 [10, 11]. These components are part of the pericentriolar material (PCM) that surrounds the centrioles and mediates functions such as microtubule organization and recruitment of proteins to the centrosome. Pericentrin in particular has been shown to be an important structural protein within the PCM [12] and exist as particles in the cytoskeleton. These particles transit to the centrosome via the motor protein dynein [12].

Antibodies directed against centrosomes have been described in a number of conditions including arthritis, Raynaud's phenomenon and systemic sclerosis [11, 13–18]. In addition, 'naturally' occurring antibodies to centrosomes have been described in laboratory rabbits [19]. The specific antigenic target of many of these autoantibodies remains undefined but with the identification of centrosome components and the availability of the cDNAs and respective recombinant proteins, systematic studies of autoantibody specificity can now be determined with more precision.

Most children with post-infectious ataxia have a mild course but some are more severely affected. If the mechanisms that lead to the development of post-infectious cerebellar ataxia are more completely understood, then a rational approach to treatment could be considered. We investigated the possibility that autoantibodies to specific centrosome proteins are associated with a subset of children with post-varicella acute cerebellar ataxia.

Eleven of 12 sera from children with post-varicella ataxia had IgG antibodies that reacted with the centrosome of formalin-fixed HeLa cells in an IIF assay (Figure 1). We observed that commercially prepared substrates, such as HEp-2 cells, were not sensitive for the detection of centrosome autoantibodies but that the formalin-fixed HeLa cells are a more reliable and sensitive substrate. When the sera were tested for IgG antibodies by immunoblotting (Figure 2), 10/12 reacted with at least one of the five centrosome proteins: ninein, PCM 1, γγ-enolase, Mob-1, or pericentrin (Table 1). Thus, one serum had autoantibodies to a centrosome antigen that was not included in the proteins listed above. Specifically, 5/12 (42%) reacted with pericentrin (Figure 2), 3/12 (25%) with ninein, 6/12 (50%) with PCM 1, 3/12 (25%) with Mob-1, and 3/11 (27%) with γγ-enolase. By comparison, 3/5 siblings with varicella infections but no ataxia, 1/1 with post-EBV ataxia, 2/2 with ADEM, 1/2 with neuroblastoma and ataxia, and 2/2 with cerebellitis had antibodies directed against at least one of the centrosome proteins. None of the 12 sera from other disease groups, including 20 children with juvenile chronic arthritis, or 20 normal controls reacted with pericentrin, although they did react with other recombinant centrosome proteins. When convalescent sera were tested, the frequency of anti-centrosome antibodies as detected by IIF and anti-pericentrin as detected by immunoblotting decreased to 4/12 (33%) of children with post-varicella ataxia, and none of the other children sustained detectable reactivity at two weeks follow-up. The number of sera from children with ataxia that was not associated with varicella is admittedly small and validation of our data would require multi-center studies or access to large neurological or infectious diseases serum banks.

The demographics, clinical profile and outcome of our patients with post-varicella ataxia has been reported elsewhere [9]. We are aware of only one report of virus or viral antigen identification in varicella-associated cerebellar ataxia [25]. A number of autoantibodies have been reported in other post-infectious cerebellar syndromes. Ito and his colleagues reported IgG and IgM anti-neuronal antibodies that reacted with both nuclear and cytoplasmic elements of Purkinje cells in acute cerebellar ataxia following infection [6]. In this study, the IgM reactivity gradually decreased and the ataxia gradually improved. Sera from 9 children with opsoclonus-myoclonus syndrome (3 associated with neuroblastoma and six with a prodromal illness) had IgM and IgG binding to the cytoplasm of cerebellar Purkinje cells and to some axons in white matter. Western blot analysis showed a distinctive pattern of binding to several neural proteins, including a 210 kDa antigen identified as the high molecular weight subunit of neurofilament [7]. Last, two children with central nervous system manifestations (including ataxia) following Mycoplasma pneumoniae infection were reported to develop anti-centriole antibodies [8].

Autoantibodies to centrosomes have also been associated with systemic sclerosis and other conditions [10, 11, 26–28]. These observations suggest that autoantibodies to centrosomes as detected by IIF are not disease specific. Studies of specific antibodies directed against pericentrin showed that they are found in 27% of systemic sclerosis sera [17]. Although ataxia or central nervous system involvement is unusual in systemic sclerosis, these patients, as well as patients with an autoimmune paraneoplastic syndrome are known to have autonomic nervous system dysfunction that is associated with antibodies directed against antigens in the myenteric complex [29–31].

Since unique reactivity with pericentrin was noted, we proceeded to determine the distribution of pericentrin in cerebellum tissue by IIF. We observed that pericentrin is expressed in cerebellum as evidenced by intense staining of centrosomes and less intense staining of neuronal axons (Figures 3,4). Cells stained by human sera and rabbit antibodies to pericentrin were found throughout the cerebellum including those in the nuclear and granular layers. This observation is consistent with evidence that, in addition to it being a component of centrosomes, pericentrin exists as particles in the cytoplasm [12]. We observed that some cells, particularly certain Purkinje cells, contain numerous particles and more than one centrosome (Figure 3) and strong reactivity was also found in adjacent cells that were consistent with the location and appearance of Bergman glial cells (Figure 4). These observations are consistent with earlier observations that Purkinje cells are polyploidy [32, 33]. The observation that children with ataxia produce antibodies to centrosome proteins, including pericentrin, and that Purkinje cells are polyploid raises the question of the vulnerability of these cells to autoimmune attack. Although we have no direct evidence to substantiate this hypothesis, in paraneoplastic cerebellar degeneration where ataxia is a dominant clinical feature, anti-Purkinje cell antibodies (APCA or anti-Yo) are believed to be responsible for the Purkinje cell loss that accompanies the disease [34, 35]. This is supported by observations that removal of the autoantibodies from the circulation of a paraneoplastic cerebellar degeneration patient provided improvement of cerebellar signs [36].

Of potential relevance to varicella-related ataxia, it was observed that newly assembled viruses disrupt microtubule organization, as well as centrosome duplication and function [37, 38]. This provides insight into mechanisms by which varicella infection might lead to altered proteins and neural function, and the subsequent production of autoantibodies. Obviously, additional studies to delineate differential expression of pericentrin and its interaction with viral components in various cells and tissues are required to substantiate this idea. Of interest, recent evidence suggests that the centrosome may be an important locus for MHC class I antigen processing and that targeting antigens to the centrosome enhances the immune response [39].

A recent study [3] reported the presence of anti-phospholipid (APL) antibodies in children with acute varicella infections. In this study we found 75% of the children with post-varicella ataxia, 50% of children with varicella but no ataxia, and none of the children with post-EBV ataxia or cerebellitis had elevated APL as detected by ELISA. All positive sera reacted with β2-GP1 + CL, three reacted with CL alone, and individual sera reacted with phosphatidyl inositol (post-varicella ataxia), phosphatidyl serine (varicella without ataxia) and sphingomyelin (varicella without ataxia). These results confirmed that children with varicella infections have a high frequency of antibodies to phospholipids. In the previous study these antibodies were found to be transiently associated with infection and were not predictive of thrombotic complications [3]. In the clinical setting of assessing the risk of thrombosis, a functional clotting assay may have more clinical relevance. In our study the most common APL was bound β2-GP1 + CL. However, these antibodies or other the other PL antibodies showed no correlation with ataxia. The specificity of APL for varicella infections was indicated by the observation that patients with cerebellitis and post-EBV ataxia did not have APL.

This is the first report to document that children with post-varicella ataxia develop autoantibodies to a variety of centrosome components. The reactivity of sera from children with post-infectious varicella ataxia appears to be unique in that they react with pericentrin. The origin of centrosome autoantibodies as a sequelae of varicella infection and the potential pathogenic role of antibodies to pericentrin awaits further study.