Background
Systemic lupus erythematosus (SLE) is an autoimmune disease that affects many organs including the joints, kidneys and brain [
1]. Some of the symptoms include arthritis, rashes, seizures, and psychoses. One of the characteristics of lupus is the detection of autoantibodies to numerous different antigens in the body [
2,
3]. The brain is one of the affected organs, causing neuropsychiatric manifestations in 31% to 70% of lupus patients, resulting in cognitive impairment and psychoses [
4‐
7]. We have hypothesized that there are brain reactive autoantibodies that bind to integral membrane proteins of the brain and this interaction is partly responsible for some of the neuropsychiatric manifestations seen in lupus [
8,
9]. These BRAA can enter the brain through increased permeability of the blood–brain barrier as lupus progresses or are produced in the brain once antibody producing cells enter the brain [
10].
Our model of lupus is the MRL/lpr mice (Jackson Lab, Bar Harbor, ME). These mice develop lupus symptoms after 2 months of age and have 50% mortality at about 5–6 months of age. The MRL/lpr have manifestations similar to humans including rashes, swollen joints and neurobehavioral manifestations [
11,
12]. A mutant of the
fas gene, the
lpr gene, is thought to help accelerate lupus-like symptoms in these mice. Because of the similarity to human lupus, the MRL/lpr mouse is an excellent model of SLE and has been used by many other researchers as their model of choice [
11,
12].
The manifestations of lupus resemble the manifestations of other diseases, making accurate diagnosis difficult. Physicians use a set of 11 different criteria and patients must satisfy 4 out of 11 to be diagnosed as having lupus [
13]. Antinuclear antibodies and anti-DNA autoantibodies have been used as some of the markers for the diagnosis of lupus [
14]. However, these markers are not specific for lupus. Therefore, being able to correctly diagnose and even predict the onset of lupus and its CNS manifestations is of high importance due to the current inability to do so [
14].
We have multiple goals in this report. The first goal is to diagnose lupus, and CNS lupus, using sera, in a reliable and rapid manner. We tested the idea that we could do this using immunosignaturing [
15]. There is mounting evidence that this technique may be useful to diagnose other CNS diseases such as Alzheimer’s [
16,
17]. Our second goal was to predict lupus onset, and specific CNS manifestations, pre-symptomatically. There are low concentrations of autoantibodies present in the sera even before clinical signs of lupus. If some autoantibodies predict the onset of lupus, and specific CNS manifestations, detection using immunosignaturing is possible. Identification of potential predictive peptides for specific CNS manifestations would be unique to this study. We and others have used the forced swim test as a measure of depressive like behavior in the MRL/lpr model [
1,
2,
18]. In the current study we utilized this test to indicate CNS dysfunction, however, it should be noted that this test is only one measure and therefore does not represent all CNS dysfunction. It is expected that other peptide subsets generated using our microarray techniques will probably correlate with other measures of CNS dysfunction.
Our third goal was to obtain preliminary information on the utility of this technique for future studies in characterizing the antigenic targets of potentially pathogenic brain-reactive autoantibodies. The random-sequence peptide microarray was used to identify peptides reactive to antibodies against BRAA. Peptide sequences were analyzed using the Guitope computer program [
19] to identify potential protein targets. As an initial test, we created five monoclonal BRAA from an MRL/lpr mouse with behavioral dysfunction to identify likely targets of these monoclonal BRAA.
The latter is important because determining the identity of BRAA targets will allow us to better understand their potential functional effects, and whether they may be mediating neuropsychiatric manifestations. These might also provide new therapeutic targets. For example, one group of researchers has found an autoantibody that reacts with double-stranded DNA and the NMDA receptor [
20]. This autoantibody resulted in cognitive deficits in their murine model, suggesting that this NMDA receptor autoantibody could be responsible for CNS manifestations [
21]. Another researcher found an autoantibody that is cross-reactive with the dynamin-1 protein that also altered the behavioral performance of their autoimmune murine model in comparison to controls [
22]. These research findings are, however, relatively random. Our techniques should provide a reliable method for identifying such potentially pathogenic BRAA and their targets.
Overall, our microarray technology should be more accurate in diagnosing and predicting lupus and its CNS manifestations, and allow for the optimum identification of potentially pathogenic BRAA and their target antigens. We were using a well validated animal model in this study in order to see if future studies with humans are warranted. We believe that our findings warrant human studies.
Discussion
Our goal was to characterize antibodies in three categories: 1) Diagnostic, 2) Predictive, and 3) Pathogenic. Diagnostic (auto)antibodies may be reliably used for diagnosing a specific disease. Predictive (auto)antibodies can predict the future onset of that disease long before it occurs. Finally, pathogenic (auto)antibodies are responsible for pathogenic mechanisms in the disease causing symptomatology. There is probably a good overlap between these 3 categories, but they can also be different. Thus, diagnostic antibodies need not be predictive (although they can be), and vice versa. Likewise, pathogenic antibodies need not be predictive, although they will almost certainly be diagnostic.
Our first and second goals were to create a detection kit that could diagnose and predict lupus and specific CNS manifestations, using a unique high-throughput microarray technology. This technology has been used in other studies to determine binding patterns specific to that disease, such as Alzheimer’s [
16], and we expect that for each disease there will be a different binding pattern that could allow us to distinguish one illness from another ([
32,
33] and unpublished data). In the future, it would be good to test our microarray technology on other autoimmune diseases to ensure that our tests can distinguish one autoimmune disease from another. Because of the various CNS manifestations and the idea that certain autoantibodies are partly responsible for each manifestation, differences in binding patterns may allow us to distinguish between them in lupus patients.
To begin, we ran the same protocols in two different studies because the first study identified possible diagnostic peptides of lupus and its CNS manifestations and the second study further suggested which peptides may indeed be diagnostic. At the end of Study 2, we identified 58 potential diagnostic peptides of lupus. Of more interest was trying to identify possible diagnostic peptides for specific CNS manifestations. Most MRL/lpr mice will eventually develop lupus, but CNS manifestations differ from one mouse to another [
1]. Looking specifically at the forced swim test, we identified 39 possible diagnostic peptides of this altered behavior.
Since we were interested in predicting lupus, in Study 1, we compared the 4 month MRL/mp to the C3H/HeJ since we made the assumption that the MRL/mp would not start developing lupus until 9–12 months of age and therefore at 4 month of age should be pre-symptomatic but having low levels of autoantibodies that are predictive. We assumed that this was also the case for the 1.5 month MRL/lpr used in Study 2 since they would not develop any symptoms until after 2 months of age. Comparison of both studies identified 18 potential predictive peptides of lupus. For the forced swim test, we identified 96 possible predictive peptides of this altered behavior. To further narrow down the true number of predictive peptides, we would need to run another study and administer a comparison of the peptide sets. All these diagnostic and predictive peptide sets only give us indications of which peptides might in fact be diagnostic and predictive, but further studies must be done to support these statements. We would need to run additional test groups to validate our findings. In addition, these studies and techniques give us useful methods and directions to pursue such supportive studies.
When selecting the peptides, we chose only the peptides that were the highest binders; however, it would be interesting in future studies to look at what the lower binding peptides might be telling us. Also, for example, when we selected peptides that may be diagnostic of altered behavior in the forced swim test, we only selected the 39 overlapping peptides between Studies 1 and 2 since diagnostic peptides should reappear from one study to another, however, the non-overlapping peptides were also high binders and therefore are still of interest to us because some of the autoantibodies that are binding these peptides may be diagnostic of another CNS manifestation that we did not investigate.
Due to current inaccurate means of diagnosing lupus and no methods to predict lupus, this microarray technology could help provide proper treatments, improve patient care and add needed therapies. We tested this technology in our mouse model first because this allowed us to not only look at lupus but, of more interest, determine if our technology could distinuguish one CNS manifestation from another. Plus, we wanted to be able to use these techniques in future studies to verify pathogenic BRAA, by injecting the appropriate BRAA into mice (which could not be done in human studies). Another benefit of doing this study in mice first was that we could obtain significant results with fewer samples as in the case of Study 1. A limited set of mice were used in the first study, but a larger number were used in the second study. The purpose of the first study was simply to identify some potential diagnostic peptides, so a large number of mice were not needed. In addition, the statistics were performed on measures that we were very certain would show differences (based on many previous reports in the literature [
34,
35], including our own [
2]) and, in fact, they did. The behavioral test and immunological assessments confirmed that the mice we were using were indeed developing murine lupus as expected. Also, with the mice, we could identify which will develop lupus and CNS manifestations, whereas in human studies this would be difficult and we would need a tremendously larger study to get the same information.
Finally, we have shown in previous studies [
36] that human sera can react with murine brain antigen. Thus, we might use murine brain antigens as a diagnostic tool for human CNS-lupus. We found that our microarrays could identify antibody reactivity that demonstrated predictability of specific CNS manifestation in our mouse model. It therefore seems likely that human lupus and its CNS manifestations would also demonstrate a similar trend, although we predict different peptides would be found. Larger trials using human sera samples may determine whether this is true. Expanded human trials would allow us to determine how much personal variability or environmental influences exist in the lupus signature. Mice studies are constrained to clonal animals in identical environments with identical histories. Humans impose much more natural immunological variability. Immunosignaturing has identified both predictive and diagnostic peptides of autoimmune diseases such as diabetes [
37], and therefore should be applicable to lupus.
Our third goal was to use the peptide sequences and the Guitope computer analysis program to determine possible natural protein matches, particularly for characterizing the brain antigens which might be mediating CNS manifestations [
28]. When looking at the possible diagnostic peptides of lupus, some matches were very interesting since autoantibodies to some of these proteins have been detected in lupus patients [
38]. Autoantibodies to the 60S ribosomal protein L12, which is important in protein synthesis, has been detected in 3-28% of lupus patients [
38]. Even though this protein is not the 60S ribosomal subunit we detected, it is possible that autoantibodies to our 60S ribosomal protein L22-like 1 can be affecting protein synthesis. Autoantibodies to 40S ribosomal protein S10, which is also important in protein synthesis, have been detected in 11-40% of lupus patients [
38]. The histone H3-like centromeric protein A, is like the H3 nucleosome which is important for packaging the DNA in the cell, and autoantibodies to this protein have been detected in 50-90% of lupus patients [
38,
39]. These anti-histones antibodies are thought to play a role in lupus nephritis, which is one of the manifestations of lupus [
38]. The exact mechanism that is occurring and how these autoantibodies are altering body function is not known, but these results help to suggest that these proteins are being affected during disease activity.
When looking at the possible predictive peptides of lupus proteins of interest include C1q tumor necrosis factor-related protein 6 and alpha-actinin-2. The collagen-like region of C1q protein is believed to play a role in lupus nephritis and autoantibodies to this protein occur in about 30-50% of lupus patients [
38]. Anti-alpha-actinin-2 antibodies have been detected in patients with lupus nephritis [
40]. It was interesting that these researchers detected anti-alpha-actinin-2 antibodies even before lupus nephritis was present. Autoantibodies to histone H3-like centromeric protein A and 60S ribosomal protein L22 were in common with the diagnostic peptides of lupus, so these autoantibodies may be present early on as biomarkers and remain throughout the disease process. The metabotropic glutamate receptor 4 was one of the possible natural protein matches for the potential diagnostic peptides of CNS manifestations (altered behavior in the forced swim test). Since researchers found that using an agonist to this receptor helped to decrease the float time in the forced swim test, this glutamate receptor may play a role as an anti-depressant in CNS-SLE [
41].
To further identify potentially affected protein in CNS-SLE we used MRL/lpr #2 from Study 2 to create five monoclonal BRAA. Their possible protein matches are listed in Table
4 and contain some interesting molecules. For example, one possible target for D9 was the D (1B) dopamine receptor, which is expressed in the limbic system and plays a role in neurotransmission; therefore any dysregulation of this receptor would likely result in some neurological deficit, such as affecting memory [
42,
43]. Other interesting matches include the gamma-aminobutyric acid receptor subunit rho-1, which may play a role in synaptic plasticity in the amygdala (an area of the brain important in emotional dysfunction) [
44]; the leucine-rich repeat-containing protein 4C (MAb G10), also known as Netrin-G1 ligand (NGL-1) [
45], which may be associated with schizophrenia, and schizophreniform-like behavior is seen in CNS-SLE; the GRB2-associated-binding protein 2 (MAb G4) may be involved in susceptibility to Alzheimer disease [
46]; and synaptojanin-2 (MAb F9), which is important in the secretion of vesicles in the synapse and any disruption would affect normal brain functioning [
47].
Future experiments, including the use of affinity chromatography where the monoclonal BRAA would be immobilized to a column and then used to isolate its corresponding antigen from mouse brain homogenate, should be employed. One could then identify the antigen using mass spectrometry. These interesting experiments are beyond the scope of the current study, however, we have shown that our microarray technology allows for further characterization of monoclonal BRAA. As a strong test of the autoantibody hypothesis, confirming the role of BRAA in CNS-SLE and the pathogenicity of specific BRAA, it would be necessary to inject these monoclonal BRAA in control mice and see if we can replicate the behavioral dysfunctions.
BRAA are not the only mechanism for triggering CNS involvement in SLE. As seen in Study1, when we split the 4 month MRL/lpr into two groups based on their anti-DNA autoantibody levels, it was noted that this breakdown was the same as for the forced swim test grouping. This suggests factors other than BRAA contribute to the behavioral manifestations in the forced swim test. However, higher anti-DNA autoantibody levels cannot distinguish between several autoimmune diseases so the hope is that the identified diagnostic and prognostic peptides will be more specific to SLE, and more importantly, the different CNS manifestations [
48]. Of course, this will need to be verified in future translational studies. A further indication that anti-DNA autoantibody levels is not the best diagnostic marker was seen in our sucrose preference test since the regrouping of the 4 M MRL/lpr into low and high consumers were not the same as the regrouping for the forced swim test or the anti-DNA autoantibody levels (data not included due to inconsistencies across the two studies).
Genetics is a contributing factor to the development of autoimmune diseases such as SLE [
49,
50]. The major histocompatibility complex in humans (HLA), TNFα, TNFβ, deficiency of complement components and IL10 are just a few examples of genes associated with SLE [
49,
51]. Alterations in these genes can influence disease onset and progression. The involvement of genetics on the development of lupus-like disease in a mouse model, MBN2, revealed two loci on different chromosomes that suppressed the autoimmune phenotype in male mice [
23]. Therefore the influence of such genetic elements on the production of autoantibodies that promoted the different banding patterns discovered using our microarrays need additional investigation. These studies would create further understanding into the mechanisms promoting the pathogenesis of SLE.
One very important asset of using our chip in predicting and diagnosing lupus and its CNS manifestations is its affordability, since this chip is created to be used for any disease. There will be no need to develop a specialized chip just to detect lupus.
Competing interests
The authors declare that they have no competing interests.
Authors’ contribution
SW helped design, collect and analyze the data presented in this paper. PS and SH helped design the different experiments and also with the analysis of the data. SW, PS and SH have been involved in the preparation of this manuscript and writing the revisions. All authors approve the final version of this manuscript.