To characterize discriminant human brain antigenic targets in patients with neuropsychiatric systemic lupus erythematosus (NPSLE), using a standardized immunoproteomic approach.
To characterize discriminant human brain antigenic targets in patients with neuropsychiatric systemic lupus erythematosus (NPSLE), using a standardized immunoproteomic approach.
Self-IgG reactivity against normal and injured human brain tissues was studied by Western blotting of sera from 169 subjects, 16 patients with NPSLE, 12 patients with SLE without neuropsychiatric manifestations (non-NPSLE), 32 patients with Sjögren's syndrome with or without central nervous involvement, 82 patients with multiple sclerosis, and 27 healthy subjects. A proteomic approach was then applied to characterize discriminant antigens identified after comparisons of all patterns.
The serum self-IgG reactivity patterns against human brain tissue differed significantly between patients with NPSLE and the control groups. Four normal brain antigenic bands were specifically or preferentially recognized by sera from NPSLE patients (p240, p90, p77, and p24). Protein band p240 was characterized as microtubule-associated protein 2B (MAP-2B), p77 as Hsp70–71, and p24 as triosephosphate isomerase. Protein band p90 was not characterized. In contrast, 1 other protein band (p56, characterized as septin 7) was never recognized by sera from NPSLE patients but was recognized by a majority of sera from non-NPSLE patients.
Our findings show that the immunoproteomic approach is a reliable method for assessing serum self-IgG reactivities against human brain tissue in NPSLE. Our characterization of some of the identified discriminant antigens, such as MAP-2B, triosephosphate isomerase, and septin 7, suggests that the stability of neuronal microtubules might be involved in the pathophysiology of NPSLE.
Systemic lupus erythematosus (SLE) is a chronic autoimmune disease characterized by multisystemic manifestations. Central nervous system (CNS) involvement has been reported to occur in 6–95% of SLE patients, depending on the criteria applied (1, 2), but the exact prevalence is probably between 6% and 12% (3). The features of neuropsychiatric SLE (NPSLE) are extremely diverse and include neurologic and psychiatric syndromes (4). The diagnosis of NPSLE is difficult and remains a challenge because the drugs used for the management of SLE, infections, or other non–SLE-related pathologic conditions may be responsible for the neuropsychiatric manifestations and have to be excluded from the assessment. Moreover, there is no “gold standard” diagnostic test that can definitively confirm NPSLE (5). Since the treatment is obviously dependent on the underlying cause, many authors have underlined the need for new diagnostic tools in NPSLE (5–7).
There are some lines of evidence that the immune system plays a role in NPSLE, since the disease typically occurs in the presence of serologically and clinically active SLE (8), and the concentrations of some chemokines (9), such as interferon-inducible protein 10/CXCL10 and fractalkine, are increased in the cerebrospinal fluid (CSF) of NPSLE patients (10–12). However, while it is well established that autoantibodies can directly damage organs, especially the kidney (13), brain antigenic targets have not been fully identified in NPSLE. Several investigators have sought to identify autoantibodies that could bind directly to neurons and could serve as diagnostic markers in NPSLE (14–19). Recently, sera collected from patients with SLE were tested for the presence of antibodies to microtubule-associated protein 2 (MAP-2) and were detected in NPSLE patients in particular (20).
Nevertheless, further investigation is necessary to determine whether there is an alteration of the immune recognition of brain self proteins in NPSLE (6). Moreover, most of the previous studies were performed with purified self antigens and/or relevant peptides from preselected targets. To avoid a restricted analysis with preselected antigenic targets, we chose to assess the global serum self-IgG reactivity against healthy or injured human brain tissue extracts. In a previous study, we found that the serum self-reactive IgG antibody repertoire against such targets differed between patients with multiple sclerosis (MS), patients with Sjögren's syndrome (SS), and healthy subjects and could help to identify brain antigenic targets with a potentially important diagnostic and pathophysiologic role in MS (21).
In the present study, we used a Western blot method to compare serum self-IgG reactivities against human brain tissue extracts in NPSLE patients and in the following control groups: SLE patients without neuropsychiatric manifestations (non-NPSLE patients), SS patients with and without CNS involvement, MS patients, and healthy subjects. We made no a priori assumptions. Using a proteomic approach, we then characterized the most discriminant brain antigenic targets in NPSLE patients. We confirmed previous results concerning the presence of anti–MAP-2B and anti–triosephosphate isomerase antibodies in NPSLE patients and characterized 2 other discriminant antigenic targets: Hsp70–71 and an unidentified p90 antigenic band. In contrast, 1 other protein band, which was characterized as septin 7, was never recognized by sera from the NPSLE patients but was recognized by a majority of sera from the non-NPSLE patients. This characterization emphasizes the possible role of neuronal microtubules in the pathophysiology of NPSLE.
IgG antibody responses to brain tissue antigens were evaluated in sera obtained from 169 study subjects. The SLE group consisted of 28 patients who had definite SLE according to the 1997 update of the American College of Rheumatology (ACR) criteria for the classification of SLE (22). Among the 28 SLE patients, 16 presented with signs and symptoms that fulfilled the case definitions for NPSLE as proposed by the ACR (4), and 12 had no evidence of neuropsychiatric manifestations (non-NPSLE patients). Sera from the first 7 NPSLE patients were used to characterize the most discriminant brain antigenic targets in NPSLE. Sera from the remaining 9 NPSLE patients were used to confirm these results by looking for the presence of autoantibodies directed against the most discriminant brain antigenic targets that were identified in sera from the first 7 NPSLE patients. SLE patients presenting with CNS involvement secondary to antiphospholipid syndrome were excluded from study.
Among the control groups, we studied 32 patients with primary SS according to the American–European Consensus Group revised criteria (23). Of these 32 SS patients, 26 had CNS involvement and 6 did not. The SS patients with CNS involvement had either chronic myelitis (n = 14), acute myelitis (n = 9), or meningoencephalitis (n = 3). We also studied 82 patients diagnosed as having MS according to the criteria described by McDonald et al (24). In addition, sera obtained from 27 healthy subjects were tested as normal controls.
The presence of anti-DNA antibodies was assessed by either the Farr assay (Amerlex anti-dsDNA radioimmunoassay kit; Trinity Biotech, Bray, County Wicklow, Ireland) or the Crithidia luciliae indirect immunofluorescence test. Antiphospholipid antibodies were evaluated by enzyme-linked immunosorbent assay (Orgentec Diagnostika, Mainz, Germany), with cutoff values of 10 IgG phospholipid units and 7 IgM phospholipid units for IgG and IgM, respectively. The clinical and immunologic characteristics of the NPSLE patients are detailed in Table 1.
|Patient/age/sex||Disease duration, months||Neurologic manifestations||MRI findings||Renal involvement||Anti-DNA antibodies, by Farr assay (IU) or CLIFT (dilution)||Antiphospholipid antibodies||Immunosuppressive drug, daily dose|
|1/59/F||191||Cervical myelitis||Cervical myelitis from C2 to C7||No||8 (Farr)||Absent||CS 10 mg|
|2/18/F||6||Cerebral angiitis||Cerebral angiitis in the cortical and subcortical areas||Yes||9 (Farr)||Absent||CS 15 mg, AZA 100 mg|
|3/30/F||0||Cervical myelitis||Cervical centromedullar lesion from C1 to C3||No||20 (Farr)||Absent||CS 30 mg, AZA 100 mg|
|4/60/F||9||Seizure||Multiple hyperintensities in the periventricular white matter and the basal ganglia||No||8 (Farr)||Absent||AZA 150 mg|
|5/45/F||0||Seizure, optic neuritis||Hyperintensities in the right centrum semiovale||No||9 (Farr)||Present||CS 50 mg|
|6/26/M||5||Quadriparesis||Normal||Yes||22 (Farr)||Present||CS 20 mg|
|7/67/F||0||Stroke||Left frontoparietal hyperintensity (stroke) and multiple hyperintensities in the periventricular white matter and right centrum semiovale||No||51 (Farr)||Absent||None|
|10/28/F||312||Seizure, psychosis||Normal||No||NK||Absent||CS 5 mg|
|11/18/F||12||Seizure||Normal||Yes||1:20 (CLIFT)||Present||CS 5 mg, MMF 2 gm|
|13/58/F||120||Psychosis||Normal||No||1:20 (CLIFT)||Present||CS 5 mg|
|14/52/F||264||Psychosis||Normal||No||1:60 (CLIFT)||Absent||CS 5 mg, AZA 100 mg|
|15/45/F||252||Seizure||Normal||Yes||1:320 (CLIFT)||Present||CS 5 mg, AZA 100 mg|
|16/44/F||96||Myelitis||Cervical myelitis||No||1:40 (CLIFT)||Absent||CS 5 mg|
All study subjects gave their written informed consent. The study was approved by the local ethics committee of the individual institutions.
Brain tissue samples dissected from Brodmann's area 10 of the frontal lobe, were obtained at autopsy from a 48-year-old man with MS and from a 68-year-old man with no history of neurologic disease (Department of Neuropathology, Centre Hospitalier Régional Universitaire de Lille and INSERM U422, Lille, France). The autopsies were performed within the framework of a tissue-collection program that was approved by the local ethics committee. In each case, the postmortem delay was less than 8 hours.
For SDS-PAGE, brain tissue samples were homogenized in Tris buffer containing 5% SDS at a final concentration of 10 mg/ml and then heated at 95°C for 10 minutes. For each well, 80 μl of this lysate was loaded onto a 10–20% gradient polyacrylamide gel, beside a molecular weight marker (Amersham Pharmacia Biotech, Uppsala, Sweden). Just before electrophoresis, the homogenates were reduced with 10 mM dithiothreitol (DTT; Sigma, St. Louis, MO). Electrophoresis was conducted for 14–16 hours in Laemmli buffer at 100V (25).
Brain sample homogenization and 1-dimensional protein separation were performed essentially as described previously (21). Briefly, 100 mg of brain tissue was homogenized in a detergent solution (4% Triton X-100 and 1× antiprotease cocktail [both from Sigma]) before protein precipitation using 3 volumes of ice-cold acetone. The sample was centrifuged at 10,000g for 20 minutes at 4°C. The supernatant was removed, and the pellet was air dried. Before electrophoresis, the pellet was resuspended in 500 μl of sample buffer (7M urea, 2M thiourea, 1% DTT, 4% Triton X-100, 1× antiprotease cocktail, and 2% volume/volume ampholines) (Amersham Pharmacia Biotech). After dissolution, the samples were used for overnight in-gel rehydration. Proteins were separated using MultiPhor II (Amersham Pharmacia Biotech) with anode (10 mM H3PO4) and cathode (10 mM NaOH) buffers.
Prior to the second dimension, the immobilized pH gradient (IPG) strips were equilibrated twice for 30 minutes in 2 ml of equilibration solution I (50 mM Tris HCl, pH 8.8, 8 mM EDTA, 10% weight/volume glycerol, 5% w/v SDS, and 1% w/v DTT) and once for 30 minutes in 2 ml of equilibration solution II (50 mM Tris HCl, pH 8.8, 8 mM EDTA 10% w/v glycerol, 5% w/v SDS, and 150 mM iodoacetamide). Equilibrated IPGs were transferred to 9–16% or 10–20% polyacrylamide gradient gels containing the crosslinker piperazine diacrylamide (2.6% crosslinking monomer concentration; Bio-Rad, Hercules, CA) (26). Gels were polymerized overnight. Electrophoresis was performed for 14–16 hours in a Bio-Rad Protean II xi chamber with current limited to 40 mA per gel. For identification of the antigens, a preparative 2-DE gel was stained with Coomassie brilliant blue G250 (Sigma) and then used for spot cutting and protein sequencing.
For immunostaining, 1-DE or 2-DE gels were blotted onto Hybond-P polyvinylidene difluoride membranes (Amersham Pharmacia Biotech Europe) using a “semidry” protocol (0.8 mA/cm2) (27) and later saturated with 5% nonfat dry milk. Western blotting was performed with total sera diluted 1:100 in Tris buffered saline (100 mM Tris, pH 8.0, NaCl 0.3M) containing 0.5% Tween 20 (w/v) and 5% nonfat dry milk. After overnight incubation at 4°C, IgG reactivities were revealed with an anti-human Fcγ horseradish peroxidase–conjugated antibody (1:10,000 dilution; Sigma). Fluorograms were prepared using an enhanced chemiluminescence kit (Amersham Pharmacia Biotech Europe). Immune profiles were analyzed when 2 independent assays had produced identical patterns.
Image analysis was performed on a Molecular Imager GS-800 calibrated densitometer (Bio-Rad) to localize and compare the IgG immune profile patterns. Superimposition and alignment of the antibody reactivities was performed using Diversity database fingerprinting software version 2.2 for 1-DE and PDQuest software for 2-DE (both from Bio-Rad). We performed comparative analyses using detection parameters that allowed us to consider as significant each band intensity that was >10% of the global background intensity. In order to calibrate and define more accurately the alignment of antibody reactivities, molecular weight marker protein standards (Pharmacia LKB, Uppsala, Sweden) as well as internal reference standards were used. Antigenic bands were numbered according to their molecular weight, preceded by a lower case “p” if found in normal brain tissue and by a capital “P” if found in MS brain tissue.
Excised plugs from Coomassie brilliant blue–stained gels were destained with 200 μl of 50% acetonitrile in 10 mM NH4HCO3 and dried in a SpeedVac concentrator (Telechem, Sunnyvale, CA). Protein was digested overnight at 37°C with sequencing-grade trypsin (5 μg/ml; Promega, Madison, WI) in 50 mM NH4HCO3. The resulting peptides were extracted twice with 25 μl of 50% acetonitrile/0.1% trifluoroacetic acid (TFA). The collected extracts were lyophilized, resuspended in 10 μl of 0.1% TFA, and desalted on ZipTip C-18 microcolumns (Millipore, Bedford, MA). Peptides were eluted directly onto the MALDI target with 5 mg/ml of α-cyano-4-hydroxycinnamic acid. MALDI-TOF mass spectrometry was used to obtain mass fingerprinting of proteins using a Voyager-DE STR mass spectrometer (Applied Biosystems, Framingham, MA). Ions were accelerated at 20 kV and reflected at 21.3 kV. Spectra were acquired in the delayed-extraction reflectron R mode. A total of 100–300 scans were averaged to produce the final spectra. Spectra were externally calibrated using the monoisotopic MH+ ion from 3 peptide standards (trypsin autodigestion products 842.510 daltons, 1045.564 daltons, and 2211.1046 daltons).
The peptide mass fingerprint spectra obtained were analyzed by searching the National Center for Biotechnology Information (NCBI) nonredundant protein database with ProFound version 3.2 (http://prowl.rockefeller.edu/) and were verified using the Mascot search engine (http://www.matrixscience.com). The parameters for each search were varied in order to achieve the best possible results. The standard parameters were as follows: Homo sapiens, 0–250-kd molecular mass (depending on the region where the spot occurred in the gel), and tryptic digest with a maximum of 1 missed cleavage. Peptide masses were stated to be monoisotopic, and methionine residues were assumed to be partially oxidized. The identity of proteins was annotated using the Swiss-Prot and TrEMBL databases.
The data were expressed in binary mode (0 = absence of an antigenic band and 1 = presence of an antigenic band) in order to submit IgG antibody patterns to analysis using either the chi-square test or Fisher's exact test. This approach allowed us to select the most relevant antigens that supported qualitatively different immune recognition. In a second stage, we used linear discriminant analysis (LDA) to balance the discriminating antigens between the populations of individuals, as described in detail previously (21, 28). Using a stepwise logistic regression model and supported by the LDA method, we were able to isolate a subgroup of brain antigens according to their strength of discrimination between the different populations involved in the study.
By associating 2 parameters (for the presence [×1] or absence [×0] of each selected antigen) and the coefficient previously defined by the LDA, a score was calculated for each subject as a representative value of the individual immune profile, using the following formula:
where Ag represents antigen and coef represents coefficient.
Statisticians calculated all the scores blindly. The calculated scores were then represented graphically. A threshold value was determined using a receiver operating characteristic curve, and the sensitivity and specificity of this approach were evaluated. When the number of patients was too small to apply LDA, chi-square and Fisher's exact tests were performed.
As the first step, we evaluated the degree of interindividual changes in serum self-IgG reactivities against brain tissue samples obtained at autopsy from an MS patient and a healthy subject. This analysis was successively performed in sera obtained from the first 7 NPSLE patients, all of the non-NPSLE patients, SS patients with and without CNS involvement, MS patients, and healthy subjects. We found a high degree of heterogeneity in the IgG reactivities within a given group as well as between the different groups with regard to the number and the nature of the antigenic bands recognized by serum IgG. Quite different patterns were also observed when a given serum sample was tested against healthy brain tissue or MS brain tissue. However, despite this high degree of heterogeneity, we observed some conserved sets of IgG reactivities against normal and MS brain tissue within the same group of subjects as well as between the different groups. Figure 1 illustrates both the common and the differing self-IgG reactivities within and between the 6 groups of subjects tested.
Despite the qualitative and quantitative changes in self-IgG patterns, comparative studies were performed between the patients with autoimmune diseases (NPSLE = 7, non-NPSLE = 12, SS with CNS involvement = 26, and SS without CNS involvement = 6), the patients with MS (n = 82), and the healthy subjects (n = 27). The mapping and alignment of 160 Western blot strips obtained with normal brain tissue or with MS brain tissue were then assessed. As illustrated in Figure 2A, LDA enabled us to identify 16 discriminant IgG reactivities against antigenic bands, ranging from 16 kd to 140 kd, consisting of 12 protein bands in normal brain tissue (p140, p112, p105, p90, p85, p70, p66, p55, p53, p50, p32, and p28) and 4 protein bands in MS brain tissue (P42, P39, P36, and P16). The coefficient values determined by LDA for each discriminant antigenic band associated with the presence or the absence of these antigens enabled us to calculate graphic coordinates for each subject. The score assigned to each patient with a systemic autoimmune disease, each patient with MS, and each healthy subject revealed clearly distinctive areas for each group (Figure 2B), with an excellent degree of concordance with the clinical data (κ = 0.931).
Self-IgG reactivity against MS or normal brain tissue was compared between patients with systemic autoimmune diseases according to the presence (7 NPSLE patients and 26 SS patients with CNS involvement) and absence (12 non-NPSLE patients and 6 SS patients without CNS involvement) of neurologic manifestations. Mapping and alignment of the 51 immunoreactivity patterns allowed us to compare the obtained self-IgG patterns between these 2 groups.
As illustrated in Figure 3, LDA allowed the identification of 5 protein bands (2 in normal brain tissue [p113 and p32] and 3 in MS brain tissue [P66, P38, and P11]) that were highly discriminant between the self-IgG patterns of patients with systemic autoimmune diseases according to the presence and absence of neurologic manifestations. Figure 3A shows these 5 antigenic bands and their frequencies in the 2 subgroups of patients. Two antigenic bands, p113 and P38, were recognized only by sera from the subgroup of patients with neurologic manifestations. The P66 band was more frequently recognized by sera from the subgroup of patients with neurologic manifestations. In contrast, antigenic bands p32 and P11 were more frequently recognized by sera from the subgroup of patients without neurologic manifestations.
A global score was calculated for each patient, taking into account the coefficient value assigned by LDA for the 5 selected bands. Graphic extrapolation of the LDA data on a single-axis graph (Figure 3B) shows that patients with systemic autoimmune diseases and neurologic manifestations (NPSLE patients or SS patients with CNS involvement) discriminated very distinctively from those with systemic autoimmune diseases without neurologic manifestations (non-NPSLE patients or SS patients without CNS involvement), with a sensitivity of 96.9% and a specificity of 94.5%. The degree of concordance with the clinical data was excellent (κ = 0.960).
To further evaluate the significance of the patterns obtained with sera from the first 7 NPSLE patients, we compared them with the patterns obtained in sera from patients with SLE but without CNS involvement (non-NPSLE) as well as patients with another systemic autoimmune disease (SS) with CNS involvement. Because of the small number of patients in each group, LDA was not applicable, and chi-square and Fisher's exact tests were performed.
Comparison of self-IgG patterns between NPSLE patients and SS patients with CNS involvement revealed 8 discriminant antigenic bands that were detected only in normal brain tissue (p240, p113, p112, p90, p77, p65, p56, and p24) (Figure 4A). Comparison of self-IgG patterns between NPSLE and non-NPSLE patients revealed 9 discriminant antigenic bands detected in normal brain tissue (p240, p126, p90, p77, p56, p37, and p32) or in MS brain tissue (P66 and P39) (Figure 4B). Antigenic band p24 just missed reaching statistical significance; it was recognized by 3 of 7 patients with NPSLE (42.8%), as compared with 1 of 12 patients with non-NPSLE (8.3%) (P = 0.06). Taken together, the data shown in Figures 4A and B demonstrate that some antigenic bands detected only in normal brain tissue were more frequently found in NPSLE patients (p240, p90, p77, and p24) or were never found in NPSLE patients (p56).
To extend these results, we determined the presence of autoantibodies against these discriminant antigenic bands in sera from 9 additional NPSLE patients (NPSLE patients 8–16 in Table 1). Sera from none of these patients recognized the p56 antigenic band. Conversely, sera from all of these patients recognized at least 2 of the 4 discriminant antigenic bands that had been identified with the sera from the first 7 NPSLE patients (i.e., p240, p90, p77, and p24). Table 2 shows a summary of the discriminant antigens that were recognized by each of the 16 patients with NPSLE.
|NPSLE patient||Neuropsychiatric manifestations||Antigenic bands||No. of antigenic bands recognized|
|5||Seizure, optic neuritis||+||+||+||+||4|
To further characterize discriminant antigens that were more frequently targeted by sera from NPSLE patients (p240, p90, p77, and p24) or were never detected by sera from NPSLE patients (p56), we used a serologic proteomic approach. Identification of discriminant proteins was first performed by comparison of 1-dimensional and 2-dimensional immune patterns. The 7 first NPSLE sera were used to identify antigenic candidates on a proteomic map that was obtained after 2-DE using normal and MS brain tissue. The 2-DE technique followed by immunoblotting assays revealed the presence of multiple antigenic spots (Figures 5A and B). The superimposition of the antigenic spots and the protein spots, which were revealed by a standard colloidal Coomassie brilliant blue–stained 2-DE, enabled us to select the proteins to be used for further in-gel digestion and MALDI-TOF analysis (Figure 5C), as previously described (29), based on matching of peptide mass. This approach allowed us to identify some proteins as potent discriminant antigens using the Swiss-Prot database.
Antigens p240 and p77 were characterized as MAP-2B (MAP2_HUMAN, accession no. P11137) and Hsp70–71 (HSP7C-HUMAN, accession no. P11142), respectively. Antigen p24 was identified as triosephosphate isomerase (TPIS_HUMAN, accession no. P60174). The protein spot corresponding to antigenic band p90 could not be characterized by mass spectrometry despite repeated assays. Antigen p56 was identified as septin 7 (SEPT7_HUMAN, accession no. Q16181).
To identify potentially relevant self antigens in brain tissues that were specifically targeted by the serum self-IgG antibody repertoire in patients with NPSLE, we applied an immunoproteomic approach, which was previously standardized in our laboratory (21, 30, 31). To our knowledge, this is the first study in NPSLE patients to assess, without any a priori assumptions, the serum self-IgG reactivities against human brain tissue. The findings might therefore be expected to answer previous questions about the significance of antibodies found in this disease (6). Proteins from healthy and injured brain tissues were used as targets. Brain tissue obtained at autopsy from a subject with MS was chosen as the injured brain tissue, since tissue from a subject with NPSLE was unavailable when the present study was conducted.
Our analysis of IgG isotype antibodies allowed us to evaluate both the natural self-reactive responses (32) and the T cell–dependent adaptive humoral responses (33). Despite a high degree of heterogeneity of the serum self-IgG responses, we found some conserved sets of IgG reactivities in healthy subjects as well as in patients. Such conserved protein antigens, which are possibly targeted by natural autoantibodies, might reflect a “footprint” of the innate immune system (32, 34). The natural B cell and T cell self-reactive repertoire is now recognized as the determinant for the homeostasis of lymphoid cells and the maintenance of self tolerance (33). Unstable patterns of antibody repertoires, possibly related to the adaptive immune response, have been described in systemic autoimmune diseases such as SLE (35). However, discriminant, stable, self-IgG patterns, probably related to pathogenic events, have been found in organ-specific autoimmune disease such as MS after 1 year of followup (30). To define more precisely the significance of such stable and unstable patterns in relation to the pathologic context, we compared the self-IgG responses of sera from healthy subjects and patients with systemic or organ-specific autoimmune diseases to healthy or injured brain tissues.
In the first step of our analyses, we found reactivities to 16 discriminant antigens when we compared the reactivity patterns of sera from healthy subjects, and patients with MS, SS with or without CNS involvement, and SLE with or without CNS involvement. Eleven of these 16 antigens had already been identified in our previous study (21), in which only MS patients, SS patients with CNS involvement, and healthy subjects were compared. Five new antigenic targets were found in the present study, which included SLE patients and SS patients without CNS involvement. Thus, organ-specific autoimmune diseases such as MS, as well as systemic autoimmune diseases such as SLE, are associated with distinct serum changes in self-IgG antibody repertoires against brain antigens.
We also tested the hypothesis that the presence or absence of CNS involvement could shape the patterns of serum self-IgG. We showed that protein bands targeted by serum self-IgG antibodies allowed systemic autoimmune diseases to be differentiated according to the presence and absence of neurologic symptoms. Some protein bands were exclusively or frequently found in sera from SLE and SS patients with CNS involvement (p113, P38, and P66), whereas other protein bands were preferentially found in sera from SLE and SS patients without CNS involvement (p32 and P11). In patients with NPSLE, some antigenic bands were never recognized (p56) and others were detected often (p240, p90, p77, and p24). This could be related to neuropathogenic or neuroprotective events (36). The lack of detection of some serum IgG antibodies in patients with systemic autoimmune diseases might be related to a defect in regulatory processes and may explain the fewer antigenic bands that were recognized as compared with the number of bands recognized by sera from MS patients and healthy subjects. However, the presence of antibodies restricted to sera from NPSLE patients suggests a possible pathogenic involvement. For example, autoantibodies against N-methyl-D-aspartate receptor have been shown to cause apoptotic death of neurons in vivo and in vitro, as well as cognitive impairment in mice (17, 37). Moreover these antibodies are present in the brain of SLE patients (17, 37).
To try to define more precisely the significance of such changes, we characterized some of the protein bands. First, an antigenic band that was never detected in NPSLE patients (p56) was characterized. This p56 protein band was identified as septin 7. Septins comprise a eukaryotic subfamily of guanine nucleotide-binding proteins and may play a conserved role in cytokinesis, exocytosis, and apoptosis in yeast and mammalian cells (38). Septins have been found to regulate microtubule stability through interaction with the microtubule-binding protein MAP-4 (39). Interestingly, an antibody response to Nedd5, which belongs to the septin family, has been found in patients with NPSLE (40). An IgG antibody response to septin 7 was previously identified by using normal sera and healthy brain tissue (30). The absence of an antibody response to septin 7 in patients with NPSLE might reflect either the loss of pathogenic antibodies linked to altered brain tissue or the absence of regulatory antibodies required for the maintenance of self tolerance or neuroprotection. The latter hypothesis, which involves neuroprotective antibodies (36), has previously been proposed in SLE (41, 42).
Second, 4 antigenic bands were more frequently (p240, p90, p24) or exclusively (p77) detected in normal brain tissue using sera from NPSLE patients. The more frequent presence of some IgG reactivities with sera from NPSLE patients suggests a possible specific pathogenic implication. Previous studies have demonstrated a potential neuropathogenic role of circulating IgG antibodies in animal models and in cell cultures (17). We failed to identify p90 antigen because it remained uncharacterized in mass spectrometry despite repeated assays. Protein bands p240, p77, and p24 were characterized as MAP-2B, Hsp70–71, and triosephosphate isomerase, respectively.
MAP-2B is restricted to neurons. It controls cytoskeletal integrity by stabilizing microtubules and is involved in the elaboration of the neuritic compartments (43). Thus, antibodies against MAP-2B may modulate neuronal plasticity. Triosephosphate isomerase is a highly conserved glycolytic enzyme that is present in all cells and is highly expressed in brain tissues (44). It has been shown that inhibition of this key enzyme affects microtubule stabilization (44) and leads to neuronal death in cultured murine cortical cells (45). A defect in this enzyme activity was found to be associated with neurologic symptoms (46). Thus, it could be hypothesized that anti–triosephosphate isomerase antibodies present in sera from NPSLE patients could lead to a sustained inhibition of triosephosphate isomerase, leading in turn to the neuropsychiatric disorders. Moreover, anti–triosephosphate isomerase antibodies have recently been found in CSF from patients with NPSLE (47). Whatever the pathogenic or regulatory role of anti–MAP-2B and anti–triosephosphate isomerase antibodies, they have recently been identified as good markers of NPSLE (20, 48).
No association between NPSLE and anti–Hsp70–71 antibodies has yet been reported. Elevated expression of a member of the Hsp70 family has been described in SLE (49). Hsp70 overexpression in brain tissues has been also associated with neuroprotective effects after cerebral injury (50). A similar frequency of autoantibodies against Hsp70 was previously found in SLE patients and in healthy subjects (51), whereas we found a higher frequency of such antibodies in NPSLE patients. Anti–Hsp70–71 antibodies have also been identified in CSF from MS patients as well as patients with schizophrenia (52, 53). These latter findings also indicate a possible relationship between CNS involvement and detectable antibodies against Hsp70–71. However, we did not store CSF samples from our NPSLE patients, and we are therefore unable to assess the presence of autoantibodies against the discriminant antigens in CSF or to compare serum and CSF self-IgG reactivities against MS and healthy brain tissues. CSF studies could help in elucidating the pathophysiologic mechanism of these antigenic targets (18), and we plan to collect both sera and CSF from all new patients with NPSLE. For the same reason, we were unable to assess levels of cytokines and chemokines in CSF.
Finally, our study included NPSLE patients with psychiatric manifestations and normal findings on magnetic resonance imaging of the brain, which is a challenging diagnosis in SLE patients. We showed that sera from these patients recognized at least 2 of the discriminant antigenic bands, which suggests the potential diagnostic value of this method in isolated psychiatric manifestations as well.
In conclusion, the immunoproteomic approach appears to be a reliable method by which to study the self-IgG antibody repertoires against brain antigens in patients with NPSLE. Using this approach, we confirmed recent reports by showing that MAP-2B and triosephosphate isomerase are brain antigenic targets in NPSLE (20, 47, 48). In contrast, anti–septin 7 antibodies were never observed in our patients with NPSLE. MAP-2B, triosephosphate isomerase, and septin 7 together are involved in neuronal microtubule stability, suggesting a role of microtubules in the pathophysiology of NPSLE. We also found 2 new potent brain antigenic targets: Hsp70–71 and an unidentified p90 antigenic band. Our approach suggests that the combination of IgG antibody responses against a cluster of antigens may be more determinant than a single response, as previously suggested (21). We plan to corroborate this hypothesis in a broader population using homemade proteochips with discriminant proteins synthesized in vitro. The diagnostic value of these antibodies and their pathophysiologic roles in NPSLE merit further studies. Finally, it would be of interest to further investigate in parallel the self-IgG reactivity against CNS antigens in sera and CSF samples from patients with NPSLE.
Dr. Lefranc had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Lefranc, Launay, Dubucquoi, de Seze, Hatron, P. Vermersch, Prin.
Acquisition of data. Lefranc, Launay, Dubucquoi, de Seze, Dussart, M. Vermersch, Hachulla, Hatron, P. Vermersch, Mouthon.
Analysis and interpretation of data. Lefranc, Launay, Dubucquoi, de Seze, Dussart, P. Vermersch, Prin.
Manuscript preparation. Lefranc, Launay, Dubucquoi, de Seze, Hachulla, Hatron, P. Vermersch, Mouthon, Prin.
Statistical analysis. Lefranc.
We thank Dr. Hervé Drobecq (Institut Biologie de Lille, Lille, France) for performing the mass spectrometry analyses. We are grateful to Nicholas Barton for advice on editing the manuscript.