**12. Deciphering a potentially disease‐related IgG ITS**

#### **12.1. MS‐specific targets? A puzzling problem**

From an historical point of view, the quest for an Ig target in MS has followed the methodol‐ ogy successfully used to solve the paradigmatic problem of subacute sclerosing panence‐ phalitis (SSPE), where it was demonstrated that the majority of intrathecal IgG was virus‐ specific. Unfortunately, MS CSF demonstrated weak and inconsistent interaction with CNS antigens in numerous earlier studies. It was thereafter proposed that IgG production in MS may be due to nonsense antibodies [71]. These over‐hasty conclusions should now be miti‐ gated.

CSF Igs constitute a private repertoire that may harbor toxic properties [216]. Functional electrophysiological modifications of neurotransmission after CSF application were observed long ago. Patterns of demyelinating lesions containing Ig and complement deposition in and around macrophages are typical of pattern 2 observed in MS [148, 217, 218]. The application of CSF to cultured cells (rat cerebellar granule neurons) significantly labels the axonal surface with IgM dots [118] and purified antibodies against myelin/oligodendrocyte glycoprotein (MOG) from MS serum bind to intact myelin in rat [219]. Various recombinant antibodies (rAb) synthesized from CD138+ CSF B‐cells of MS patients stain most of the glial components [220, 221] and many specific antibodies have been shown to react against cultured oligodendrocytes or human CNS tissue [222]. In animal models, anti‐MOG antibodies purified from human MS serum strongly enhance lesions without increasing inflammation [219, 223]. In a model of focal implantation of hybridoma cells secreting Ig against myelin in rat spinal cord, demyelination occurred around the implantation site [224]. Moreover, the implantation of anti‐GalC hybrid‐ oma cells may inhibit myelination in a complement‐dependent pathway both in vivo and in vitro [225, 226]. These latter models elegantly combine focal IgG synthesis with focal anatom‐ ical lesions and are reminiscent of the infiltration by plasma cells inside non‐remyelinated plaques.

In conclusion, the hypothesis that CSF Ig could play an active role in brain dysfunction, both at functional and at anatomical levels, is largely justified.

#### **12.2. Does the microheterogeneity of ITS translate into disease heterogeneity?**

inflammatory disorders [194, 213]. An experimental demonstration was given by a series of MS patients who developed CSF IgG against TT after tetanus toxoid vaccination [196]. According to this assumption, the recruitment of intrathecal antibody‐secreting cells should be progressive and the complete MRZ pattern might take years or even decades to achieve. This was observed in clinical studies where the mean age at spinal tap and disease duration correlated with the number of elevated antibody indices [39]. Half of the patients having fewer than 5 years of disease duration had no elevated AI (MRZ), whereas 70% of those with more than 10 years duration had AI ≥2 [39]. A similar trend was observed in a study based on 17

From an historical point of view, the quest for an Ig target in MS has followed the methodol‐ ogy successfully used to solve the paradigmatic problem of subacute sclerosing panence‐ phalitis (SSPE), where it was demonstrated that the majority of intrathecal IgG was virus‐ specific. Unfortunately, MS CSF demonstrated weak and inconsistent interaction with CNS antigens in numerous earlier studies. It was thereafter proposed that IgG production in MS may be due to nonsense antibodies [71]. These over‐hasty conclusions should now be miti‐

CSF Igs constitute a private repertoire that may harbor toxic properties [216]. Functional electrophysiological modifications of neurotransmission after CSF application were observed long ago. Patterns of demyelinating lesions containing Ig and complement deposition in and around macrophages are typical of pattern 2 observed in MS [148, 217, 218]. The application of CSF to cultured cells (rat cerebellar granule neurons) significantly labels the axonal surface with IgM dots [118] and purified antibodies against myelin/oligodendrocyte glycoprotein (MOG) from MS serum bind to intact myelin in rat [219]. Various recombinant antibodies (rAb) synthesized from CD138+ CSF B‐cells of MS patients stain most of the glial components [220, 221] and many specific antibodies have been shown to react against cultured oligodendrocytes or human CNS tissue [222]. In animal models, anti‐MOG antibodies purified from human MS serum strongly enhance lesions without increasing inflammation [219, 223]. In a model of focal implantation of hybridoma cells secreting Ig against myelin in rat spinal cord, demyelination occurred around the implantation site [224]. Moreover, the implantation of anti‐GalC hybrid‐ oma cells may inhibit myelination in a complement‐dependent pathway both in vivo and in vitro [225, 226]. These latter models elegantly combine focal IgG synthesis with focal anatom‐ ical lesions and are reminiscent of the infiltration by plasma cells inside non‐remyelinated

In conclusion, the hypothesis that CSF Ig could play an active role in brain dysfunction, both

at functional and at anatomical levels, is largely justified.

**12. Deciphering a potentially disease‐related IgG ITS**

**12.1. MS‐specific targets? A puzzling problem**

antigens [197].

72 Trending Topics in Multiple Sclerosis

gated.

plaques.

Although the bulk of ITS targets unknown antigens, some of the antibody clones are thought to play an active biological role that could be either deleterious or beneficial. For example, occasional cases of MS may display anti‐N‐methyl‐D‐aspartate receptor (NMDAR) (<1%) or anti‐MOG (12%) antibodies, which are associated with cognitive and demyelinating features [227]. However, until now and unlike the anti‐aquaporin4 (AQP4) observed in neuromyelitis optica, none of the hundred or so identified antibodies has definitely been associated with the clinical (i.e., epilepsy, amnesia) or pathological distinctive features of MS (i.e., ability to remyelinate). The pathogenicity of these antibodies may be characterized in vitro by the extent of the demyelination or axonal loss obtained in myelinated cultures [216], by specific cell losses such as oligodendrocyte progenitors [228] or by functional alteration of synaptic transmission [89, 228]. For example, IgGs directed against MOG or sulfatide produce a complete myelin loss whereas the axons are preserved. On the other hand, IgG against neurofascine produces both myelin loss and loss of 15–40% of axons [216]. Therefore, at the global level, subsets of patients harbor variable toxicity against myelin or axons owing to one or more different antibody specificities against the targets. In the seminal work of Elliott et al. [216], a third of patients' sera were able to destroy myelin and only 6% targeted axons, but this was probably massively underestimated owing to the limited number of targets examined, fewer than 10. Heteroge‐ neity is also present at the clonal level since different IgG clones targeting the same antigens may display different pathogenicity ranging from null to high. This clonal microheterogeneity is driven by the ability of each clone to fix complement, and, for example, among the many MOG‐specific clones generated in EAE, only some of them also recognize the myelin sheath, and their pathogenicity is driven by complement fixation [187]. On the other end of the Ig spectrum, anti‐MBP IgMs are associated with a more benign course. Interestingly, these Ig target a shared epitope antigen between MBP and the extracellular loop of CD64, which is one of the IgG Fc receptors responsible for the immunologic properties of macrophages [229].

Microarrays using a small subset of autoantigens discriminated immune signatures in the sera of RR, SP, and PP MS patients [84]. Interestingly, immune signatures based on 13 IgG and one IgM also discriminated pathologic patterns I and II. These results assign a main role to Ig patterns in the pathophysiology of axonal loss, and replication in the CSF may be even more informative since paired serum and CSF samples display different patterns [93]. The large‐ scale collection of targeted antigens would help to define the functional state of the intrathecal immune system. Moreover, the availability of techniques able to profile ITS IgG diversity and pathogenicity might help to decipher the crucial problem of predicting the severity of MS.

#### **12.3. Putative target antigens—what do OCBs target inside the brain?**

The targets are often considered to be largely unknown [230] and studies have yielded contradictory results. CSF antibodies were often considered to be nonspecific owing to their lack of specificity against the three major myelin proteins, but this does not preclude any other specificity and studies devoid of a priori may reveal unknown specificities against other membrane proteins, lipids, or glycolipids. The failure to find a major antigen for intra‐BBB‐ synthesized Ig may not relate to a nonsense antibody production but instead may reflect the molecular complexity of the CNS and the presumed antigenic target [71]. By analogy with infectious diseases where the antigenic response is driven by pathogen antigenicity, most strategies have concentrated upon the identification of relevant antigenic targets. Paradoxi‐ cally, from these efforts emerged the unexpected problem of the immense multiplicity and variability of targets.

#### *12.3.1. A priori methods*

In the last 30 years, the paradigmatic representation came by analogy with EAE models triggered by a single antigen vaccination, in line with Koch's postulates. Therefore, it was initially thought that ITS was directed against a (single) putative antigen. Unfortunately, intensive investigations of a wide range of potential targets were based on this misleading conception and failed to demonstrate any definitive association apart from more than 100 protein targets [231–233] and roughly 300 epitopes (6% being non‐peptidic) [234]. Unlike in many other autoimmune disorders (e.g., myasthenia gravis), no study has succeeded in demonstrating any single antibody specificity for MS but few candidate epitopes (i.e., GAGA4) are still awaiting confirmation of their specificity by replication studies and the availability of commercial kits. Similar results have been obtained in SLE where large‐scale clusters of antibodies capture significant information regarding diagnosis and prognosis, whereas limited information is obtained from routinely measured antibodies [235].

#### *12.3.2. Non‐a priori methods*

#### *12.3.2.1. Phage display*

The phage display random peptide library was used to try to characterize unknown epitopes. Phagotopes identified from a CSF test pool of 10 MS patients were not recognized by the CSF of 55 other MS patients [76], which was interpreted as the demonstration that isolated phagotopes are patient‐ but not disease‐specific epitopes [76]. In a study using a similar design, 6 clones were selected from a CSF of 4 CIS, but panning to 187 MS patients found 0–22% positivity depending on the clone and MS type [236]. In the seminal work by Yu et al. [72], one dominant peptide was isolated from each of five MS CSF. Each peptide was patient‐specific, and in a given CSF, multiple OCBs (but not all) reacted against the peptide under study, confirming a restricted clonal IgG expression in CSF [72]. The specificity of CSF Ig against peptides was 10‐fold higher than from serum, suggesting a local affinity maturation of antibodies.

#### *12.3.2.2. Generation of recombinant antibodies (rAbs) from clonally expanded plasma cells from CSF*

The major advantages of this technique are as follows: (a) absence of a priori knowledge of the putative targets, (b) revelation of the most representative targeted peptides, and (c) precise identification of the candidate protein. Its main drawbacks are that it samples very few clones from each CSF. The generated rAbs clones bound to neuronal or astrocyte targets but myelin array failed to identify any precise target in most of them [237]. Interestingly, the aligned protein candidates identified among the multiple CSF analysis from the different groups did not overlap, thus suggesting a broad range of specificities [72, 238].

#### *12.3.2.3. CSF analysis with protein microarrays*

molecular complexity of the CNS and the presumed antigenic target [71]. By analogy with infectious diseases where the antigenic response is driven by pathogen antigenicity, most strategies have concentrated upon the identification of relevant antigenic targets. Paradoxi‐ cally, from these efforts emerged the unexpected problem of the immense multiplicity and

In the last 30 years, the paradigmatic representation came by analogy with EAE models triggered by a single antigen vaccination, in line with Koch's postulates. Therefore, it was initially thought that ITS was directed against a (single) putative antigen. Unfortunately, intensive investigations of a wide range of potential targets were based on this misleading conception and failed to demonstrate any definitive association apart from more than 100 protein targets [231–233] and roughly 300 epitopes (6% being non‐peptidic) [234]. Unlike in many other autoimmune disorders (e.g., myasthenia gravis), no study has succeeded in demonstrating any single antibody specificity for MS but few candidate epitopes (i.e., GAGA4) are still awaiting confirmation of their specificity by replication studies and the availability of commercial kits. Similar results have been obtained in SLE where large‐scale clusters of antibodies capture significant information regarding diagnosis and prognosis, whereas

The phage display random peptide library was used to try to characterize unknown epitopes. Phagotopes identified from a CSF test pool of 10 MS patients were not recognized by the CSF of 55 other MS patients [76], which was interpreted as the demonstration that isolated phagotopes are patient‐ but not disease‐specific epitopes [76]. In a study using a similar design, 6 clones were selected from a CSF of 4 CIS, but panning to 187 MS patients found 0–22% positivity depending on the clone and MS type [236]. In the seminal work by Yu et al. [72], one dominant peptide was isolated from each of five MS CSF. Each peptide was patient‐specific, and in a given CSF, multiple OCBs (but not all) reacted against the peptide under study, confirming a restricted clonal IgG expression in CSF [72]. The specificity of CSF Ig against peptides was 10‐fold higher than from serum, suggesting a local affinity maturation of

*12.3.2.2. Generation of recombinant antibodies (rAbs) from clonally expanded plasma cells from CSF*

The major advantages of this technique are as follows: (a) absence of a priori knowledge of the putative targets, (b) revelation of the most representative targeted peptides, and (c) precise identification of the candidate protein. Its main drawbacks are that it samples very few clones from each CSF. The generated rAbs clones bound to neuronal or astrocyte targets but myelin array failed to identify any precise target in most of them [237]. Interestingly, the aligned

limited information is obtained from routinely measured antibodies [235].

variability of targets.

74 Trending Topics in Multiple Sclerosis

*12.3.1. A priori methods*

*12.3.2. Non‐a priori methods*

*12.3.2.1. Phage display*

antibodies.

Basically, protein microarray chips are high‐throughput parallelized enzyme‐linked immu‐ nosorbent assay (ELISA) enabling extensive analysis of the breadth of antibody response. Sera from 17 RR‐MS patients were compared to those from 15 controls on a protein microarray printed with more than 3101 candidate antigens that represented multiple antigen families [231]. Although numerous candidate proteins were identified, no clue was obtained about individual pattern variability. A larger microarray displaying 37,000 tagged proteins was created in which patterns of immunoreactivity differed between individual patients [232]. The main drawback of this screening method is the biased antibody profiling approach, since the technique does not account for posttranslational modifications of proteins and non‐proteic targets.

#### **12.4. Screening new targets associated with chronic CNS inflammation**

Comparison of CSF Ig recognition patterns against normal or MS brain tissue lysates revealed a profile of 16 bands highly predictive of MS, suggesting that new antigens are specifically expressed in inflamed brain [239]. Chronic inflammation was associated with the aberrant expression of carbohydrate markers, which were associated with a high level of specific antibodies in CSF [240].

Development of microarrays targeting myelin proteome or brain lipids allowed the confirma‐ tion of a diverse polyclonal reaction against these non‐protidic targets both in humans and in EAE [84, 113, 241, 242]. Unconventional lipid targeting may be common since the proportion of CSF B‐cells targeting lipids was as high as 27% in a study based on rAb [243] and injection of anti‐sulfatid monoclonal antibody causes more severe EAE [241]. Unfortunately, only a limited number of potential targets were printed on these pioneering microarrays and only a minute number of rAb clones were obtained from a single patient.

Sugar moieties are an integral part of normal glycoproteins and glycolipids. Glycosylation occurs in the intracellular compartment where the first added sugar moieties are further masked in the mature‐released glycoprotein. Such "cryptic" carbohydrate structures are potentially immunogenic as demonstrated in human immunodeficiency virus (HIV)‐1 and cancer models [240]. Finally, an antigen microarray platform was constructed to display a wide range of autoantigens (proteins of the myelin sheath, liposomes of varying lipid composition, and cryptic glycan epitopes) and confirmed that CSF antibodies directed against various non‐ proteinic structures are common in EAE and MS [240]. Although simultaneously targeting a limited number of structures, these pioneering studies opened perspectives for subsequent work based on larger microarrays [244, 245], thereby acknowledging the necessity to include various natural targets such as peptides, lipids, and glycans of both human and foreign origin and their posttranslational modifications. Interestingly, microbe‐derived polysaccharides are targeted by the host immune response and may share cross‐reactivity with normal human glycan structures [245]. In‐depth screening of foreign antigens including viral particles might be fruitful for deciphering a potential complex network of cross‐reactivity (**Figure 6**).

**Figure 6. Overview of antigen microarray principles**. Interesting antigens are selected from libraries of autoantigen peptides or foreign peptides (i.e., EBV proteins), but post‐translationaly modified proteins, lipids, or glycans are suita‐ ble. Dense arrays of antigens are spotted on prepared glass slides. Patient samples (serum, CSF) are dropped on the chip, relevant antibodies hybridize to target antigens, and fluorescent‐labeled antibodies are applied for detection. The specific antibody signature results in a pattern informative of disease (MS vs other) and stages (RR vs SP) and reveals the early dynamics of the antibody response [42, 93]. Future chips may include a large variety of antigens including atypical (i.e., lipids, glycans, oxidized) and foreign antigens (i.e., viral proteins).

#### **12.5. Posttranslational modifications**

About 200 posttranslational modifications for proteins and sharing common properties have been described to date: PTMs target protein hotspots, mostly exposed on the protein surface, they may occur in physiologic life or in reaction to pathology, and protein modifications are irreversible. Although PTM‐targeting antibodies occurring in vivo have been rarely described until now in autoimmune disorders, they are being increasingly recognized and all of them might exist since most PTMs are targeted by commercially available monoclonal antibodies raised for research purposes. "Normal" antigens may be masked by PTM and may generate new self‐antigens. A growing body of evidence suggests that PTM may play a central role in some inflammatory pathologies such as rheumatoid arthritis [246], and in inflammatory bowel diseases, type I diabetes, and atherosclerosis. Tolerance to self‐antigens may be broken when self‐antigens undergo protein modifications, whereby they are recognized by antibodies and T‐cells as foreign antigens. This kind of molecular (self‐)mimicry might trigger inflammation. Therefore, PTM may also be key players for specific autoantibody recognition in MS. This major hypothesis could account for the failure to uncover specific targets of the intrathecally synthesized antibodies screened using native protein microarrays. Screening of libraries of synthetically posttranslational‐modified peptides mimicking antigenic epitopes may allow the characterization of new targets.

glycan structures [245]. In‐depth screening of foreign antigens including viral particles might

**Figure 6. Overview of antigen microarray principles**. Interesting antigens are selected from libraries of autoantigen peptides or foreign peptides (i.e., EBV proteins), but post‐translationaly modified proteins, lipids, or glycans are suita‐ ble. Dense arrays of antigens are spotted on prepared glass slides. Patient samples (serum, CSF) are dropped on the chip, relevant antibodies hybridize to target antigens, and fluorescent‐labeled antibodies are applied for detection. The specific antibody signature results in a pattern informative of disease (MS vs other) and stages (RR vs SP) and reveals the early dynamics of the antibody response [42, 93]. Future chips may include a large variety of antigens including

About 200 posttranslational modifications for proteins and sharing common properties have been described to date: PTMs target protein hotspots, mostly exposed on the protein surface, they may occur in physiologic life or in reaction to pathology, and protein modifications are irreversible. Although PTM‐targeting antibodies occurring in vivo have been rarely described until now in autoimmune disorders, they are being increasingly recognized and all of them might exist since most PTMs are targeted by commercially available monoclonal antibodies raised for research purposes. "Normal" antigens may be masked by PTM and may generate new self‐antigens. A growing body of evidence suggests that PTM may play a central role in some inflammatory pathologies such as rheumatoid arthritis [246], and in inflammatory bowel diseases, type I diabetes, and atherosclerosis. Tolerance to self‐antigens may be broken when self‐antigens undergo protein modifications, whereby they are recognized by antibodies and T‐cells as foreign antigens. This kind of molecular (self‐)mimicry might trigger inflammation. Therefore, PTM may also be key players for specific autoantibody recognition in MS. This

atypical (i.e., lipids, glycans, oxidized) and foreign antigens (i.e., viral proteins).

**12.5. Posttranslational modifications**

76 Trending Topics in Multiple Sclerosis

be fruitful for deciphering a potential complex network of cross‐reactivity (**Figure 6**).

Citrullination is increased during EAE and MS and the targeted proteins are catabolized faster. Antibodies in rheumatoid arthritis mainly target citrullinated peptides. Methylation of arginine and lysine residues occurs in EAE brains and these PTMs are targeted in lupus erythematosus. Acetylation of intracellular proteins increases as MS evolves and acetylated MBP produces an immunodominant T‐cell epitope in mice that may trigger EAE [247].

Enzymatic glycosylation of protein is a common PTM and polymorphism of N‐glycan‐ branching enzyme (MGAT5) is associated with the risk of MS and spontaneous EAE. The MOG fragment (aa30‐50) bears an N‐glycosylated asparagine at position 31 that is responsible for its antigenic properties and this PTM peptide is mimicked by the synthetic peptide CSF114(Glc) [248]. Another interesting candidate is the glucose‐based antigen GAGA4 [Glc(α1,4)Glc(α)] [249]. Both peptides are associated with specific serum antibodies in many MS patients.

Oxidization generates a large range of by‐products since reactive oxygen species (ROS) are able to oxidize nearly all the biological components according to their oxidation potential as follows:


Considering the wide range of possibilities to modify self‐antigens by oxidation, this process is an attractive candidate to initiate an "oxidative PTM intolerance" [255].

#### **12.6. Designing microarrays of PTM proteins for multiplex screening with antibodies from CSF of MS patients**

Most of the PTM described above are induced by enzymatic catalysis or a chemical reaction occurring under stringent conditions, and their biologic synthesis is not facilitated by the large range of parameters involved. The question is whether it is possible to print large microarrays of antigens able to express all or most of the possible combinations of PTM. Technical consid‐ erations and successful experimentation on a small scale suggest that various PTMs may be successfully applied on immobilized proteins [255, 256] and that the development of com‐ mercial PTM protein microarrays holds much promise for the future.
