Moving towards a cure: blocking pathogenic antibodies in systemic lupus erythematosus

Abstract

Systemic lupus erythematosus (SLE) is characterized by the presence of autoantibodies that can mediate tissue damage in multiple organs. The underlying aetiology of SLE autoantibodies remains unknown, and treatments aimed at eliminating B cells, or limiting their function, have demonstrated limited therapeutic benefit. Thus, the current therapies for SLE are based on the concept of nonspecific immunosuppression and consist of nonsteroidal anti-inflammatory drugs (NSAIDS), corticosteroids, anti-malarials and cytotoxic drugs, all of which have serious adverse side effects including organ damage. The major auto-specificity in SLE is double-stranded (ds) DNA. Many anti-dsDNA antibodies cross-react with non-DNA antigens that may be the direct targets for their pathogenic activity. Studying anti-dsDNA antibodies present in SLE patients and in animal models of lupus, we have identified a subset of anti-dsDNA antibodies which is pathogenic in the brain as well as in the kidney. We have recently demonstrated that specific peptides, or small molecules, can protect target organs from antibody-mediated damage. Thus, it might be possible to treat the aspects of autoimmune disease without inducing major immunosuppression and ensuing infectious complications.

Introduction

Systemic lupus erythematosus (SLE) is a chronic autoimmune disease whose symptoms include arthritis, immunologic abnormalities, blood disorders, serositis, malar rashes, renal damage, skin rashes and neurological disorders, as listed in Table 1 [1–3]. The pathogenicity of anti-dsDNA antibodies in renal disease in SLE has long been appreciated, and the molecular mechanisms of their pathogenicity remain a subject of intense investigation. Recently, it has been shown that anti-dsDNA antibodies can contribute to tissue damage by multiple mechanisms, such as the activation of the complement cascade and ensuing recruitment of inflammatory cells to the site of immune complex deposition and the direct activation of inflammatory cells through Fc receptor engagement by immune complexes. More recently, it has been shown that DNA-containing immune complexes engage toll-like receptor (TLR) 7 or 9 in dendritic cells, which initiate an inflammatory cascade that may contribute to the accelerated atherosclerosis observed in SLE patients [4]. Lupus patients are known to harbour autoantibodies of multiple specificities, such as anti-phospholipid, anti-Ro and anti-dsDNA antibodies [5]. We have identified a subset of anti-dsDNA antibodies that exhibit renal pathogenicity and, also, cross-react with the N-methyl-d-aspartate receptor (NMDAR) on neurons [6, 7]. We have demonstrated that this subset of anti-dsDNA antibodies has both nephrotoxic and neurotoxic potential. Because it is well established that not all anti-dsDNA antibodies are pathogenic, we reasoned that therapeutic targeting of anti-dsDNA antibodies with known pathogenicity might be sufficient to ameliorate disease activity and prevent irreversible damage in organs commonly affected in SLE, such as the kidneys and the brain. Our results suggest that blocking a subset of anti-dsDNA antibodies can lessen disease burden in animal models and that it may not be necessary to eliminate all DNA reactivity to provide clinical benefit to the patient.

Table 1.

Diagnostic criteria for the diagnosis of systemic lupus erythematosus

Criterion	Definition
1. Malar rash	Fixed erythema, flator raised, over the malar eminences, tending to spare the nasolabial folds
2. Discoid rash	Erythematous circular raised patches with adherent keratotic scaling and follicular plugging; atrophic scarring may occur in older lesions
3. Photosensitivity	Skin rash as a results to unusual reaction to sunlight or exposure to ultraviolet light, by patient history
4. Oral ulcers	Oral or nasopharyngeal ulcers, usually painless, observed by physician
5. Arthritis	Nonerosive arthritis of two or more peripheral joints, characterized by tenderness, swelling or effusion
6. Serositis	Pleuritis or pericarditis, documented by ECG or rub or evidence of effusion
7. Renal disorder	Persistent proteinuria >0.5 g day−1, or >3+ (if quantification not performed), or cellular casts may be red cell, haemoglobin, granular, tubular or mixed
8. Neurologic disorder	Seizures or psychosis without other causes, such as the absence of offending drugs or known metabolic derangement; e.g. uremia, ketoacidosis or electrolyte imbalance
9. Haematologic disorder	Hemolytic anaemia: with reticulocytosis Leukopenia: <4000 mm−3 (total on two or more occassions) Lymphopenia: <1500 mm−3 Thrombocytopenia: less 100 000 mm−3 in the absence of off ending drugs
10. Immunologic disorder	Anti-dsDNA: antibody to native DNA in abnormal titre Anti-Sm: presence of antibody to Sm nuclear antigen Positive finding of anti-phospholipid antibodies
11. Antinuclear antibodies	An abnormal titre of anti-nuclear antibody (ANA) by immunofluorescence or an equivalent assay at any point in time and in the absence of drugs known to induce ANAs

This classification is based on 11 criteria. For the purpose of identifying patients in clinical studies, if any four or more of these criteria, well documented, are present at any time in a patient’s history, the diagnosis is likely to be systemic lupus erythematosus. Specificity is ~95%; sensitivity is ~75%; ANA, antinuclear antibodies; dsDNA, double-strand DNA; ECG, electrocardiography. [1,3].

A subset of anti-dsDNA antibodies cross-reacts with the NMDAR

The development of anti-dsDNA antibodies is not completely understood. However, it is widely appreciated that protein antigens, together with cognate T-cell help, activate B cells that mature in germinal centres and undergo class switch recombination and somatic mutation of immunoglobulin variable region genes to produce second generation, high affinity IgG antibodies [8]. Because anti-dsDNA antibodies often display extensive somatic mutation, we asked whether they might also bind to protein antigens and whether such protein antigens might elicit the activation of dsDNA-reactive B cells. We therefore searched for peptide antigens that were recognized by an anti-dsDNA monoclonal antibody, R4A, with known renal pathogenicity. We screened the R4A antibody for binding to peptides present in a phage peptide display library which harboured random, 10 amino acid inserts encoded by random 30-base pair inserts. We identified a number of phages bound by the R4A antibody. An analysis of the peptide sequences in antibody-bound phage demonstrated that R4A recognized a five amino acid consensus sequence D/E W D/E Y S/G. The peptide could inhibit DNA binding, demonstrating that it occupied the same or a nearby binding site. Furthermore, R4A bound pentapeptides whether they were composed of l- or d-amino acids. A search of protein databases revealed the consensus sequence to be present in some bacterial antigens, supporting the hypothesis that anti-dsDNA antibodies may be triggered by cross-reactive antigens that exist in pathogens. Surprisingly, this same peptide was also found in the NR2A and NR2B subunits of the glutamatergic NMDAR [7]. Indeed, the peptide mapped to an exposed extracellular region of the NMDA receptor and the R4A antibody was able to bind to the extracellular domains of both the NR2A and NR2B subunits. Moreover, the antibody bound the NMDAR, presumably in its native conformation, on neuronal membranes in cell cultures. This observation was of great interest, because it coincided with the increasing appreciation that many patients with SLE display central nervous system(CNS)manifestations [9–11]. The CNS symptoms, including cognitive impairment and mood disorder, are amongst the most frequent symptoms of neuropsychiatric lupus (see Table 2). Both cognitive function and mood stability require intact function of the NMDAR [12]; thus, we hypothesize that anti-dsDNA antibody cross-reactivity might catalyse at least some of the CNS manifestations of disease [13].

Table 2.

Neuropsychiatric syndromes observed in systemic lupus erythematosus

NPSLE associated with central nervous system
1. Aseptic meningitis	Meningeal irritation, pleocytosis in cerebrospinal fluid
2. Cerebrovascular disease	Stroke, transient ischaemic attack, cerebral venous sinus thrombosis
3. Demyelinating syndrome	Acute or relapsing demyelinating encephalomyelitis
4. Headache	Migraine and benign intracranial hypertension
5. Movement disorders	Chorea
6. Myelopathy	Transverse myelitis
7. Seizure disorders	Spells of brief, involuntary, paroxysmal alterations
8. Acute confusional state	Delirium
9. Anxiety disorder	Psychiatric disorder
10. Cognitive dysfunction	Dementia, disorders in memory, attention, conceptual reasoning, and cognitive flexibility
11. Mood disorder	Emotional imbalance with depressive, manic or mixed features
12. Psychosis	Psychiatric disorder
NPSLE associated with peripheral nervous system
1. Guillain-Barré syndrome	Acute inflammatory demyelinating polyradiculoneuropathy, affecting spinal roots, peripheral and occasionally cranial nerves
2. Autonomic disorder	Impairment in sympathetic or parasympathetic nervous systems
3. Mononeuropathy	Peripheral sensorimotor neuropathy, most often affects longer nerves first
4. Myasthenia gravis	Autoimmune disorder caused by circulating antibodies that block acetylcholine receptors at post-synaptic neuromuscular junctions
5. Cranial neuropathy	Most often involve cranial nerves II–VII, with ocular manifestations
6. Plexopathy	Disorder in brachial orlumbosacral plexus
7. Polyneuropathy	Myelinopathy, owing to loss of myelin, or neuronopathy, owing to loss of peripheral neurons

Notes: Adapted from The American College of Rheumatology [9] and Garcia-Cavazos [10].

Neurotoxic mechanisms of anti-dsDNA and anti-NMDAR antibodies

We have investigated the functional consequences of anti-dsDNA antibody binding to the NMDAR, both in vivo and ex vivo (Fig. 1). Our initial approach was to inject the R4A antibody directly into the hippocampus of mice and measured the effects on neurons [7]. Exposure to R4A caused neuronal death, as measured by TUNEL and caspase reactivity, which occurred even when Fab fragments of the antibody were injected, demonstrating that there was no requirement for complement or Fc receptors (on Fc receptor-bearing cells) in the brain. Moreover, damage could be prevented by systemic administration ofMK-801, an NMDAR antagonist that modulates receptor activity, providing further confirmation that the mechanism of R4A-induced neuronal death was through the modulation of NMDAR activity [7].

Fig. 1.

Mechanisms of neurotoxicity of R4A, an anti-dsDNA, anti-NMDAR antibody. (a) R4A displays strong binding to NMDAR-expressing neurons, as shown by the whole-brain mount (left, scale, 1 mm)and the high-magnification view (top right; so, stratum oriens; sp, stratum pyramidale; sr, stratum radiatum; scale, 25 μm); whereas the control antibody, IgG2b, shows null binding (bottom right). (b) Electrophysiological studies in ex vivo slices from the hippocampus reveal that R4A, at low concentrations (10–50 μg mL⁻¹), increases the activity of the receptor, measured as field excitatory post-synaptic potentials (NMDAR fEPSP), when paired with synaptic stimulation (“Stim”). (c) Imaging studies of R4A at high concentrations(100–200 μg mL⁻¹). The left two panels show imaged mitochondria (green dots, scale 10 μm) in the stratum pyramidale of a slice at the onset (T0) and 40 min (T40) after exposure to R4A and NMDA. The weaker signal at T40 indicates mitochondrial dysfunction. The right panel shows TUNEL-positive hippocampal cells (brown, scale 25 μm) after in vivo injection of R4A.

We used the ex vivo hippocampal slice preparation (Fig. 1b) to assess the effects of the anti-dsDNA, anti-NMDAR antibody on neuronal function [14]. TheR4A antibody alone did not alter synaptic activity, but when administered together with agonists of the NMDAR, such as glutamate or NMDA itself, R4A enhanced the synaptic activity mediated by NMDAR. This effect was observed at antibody levels as low as 10–15 μg mL⁻¹, which are concentrations that are present in clinical settings. Because we observed the synergistic activity of glutamatergic agonist and the autoantibody, we hypothesized that R4A bound preferentially to the open configuration of the receptor and suggested that the autoantibody could alter synaptic transmission only in activated neurons. The NMDAR antagonist, MK-801, fixes the receptor in the open configuration, whilst blocking the transit of calcium through the receptor pore [15]. Therefore, we studied R4A binding to hippocampal slices treated with MK-801 [14]. As hypothesized, the antibody preferentially bound to MK-801-treated tissue with the NMDAR in the open configuration. Further studies of hippocampal slices showed that at high levels (100–200 μg mL⁻¹), the R4A antibody altered mitochondrial membrane permeability, a cellular event that initiates excitotoxic death [16] (Fig. 1c). Interestingly, the concentration of antibody needed to induce neuronal death was about 10 times greater than the concentration needed to alter synaptic responses [14]. This result is especially intriguing, as it provides a potential insight into the clinical observation that cognitive impairments in SLE patients are sometimes transient and sometimes permanent. It may be that the local CNS antibody concentration determines whether neuronal dysfunction will be reversible or not. Notably, the concentration of anti-NMDAR antibody measured in cerebrospinal fluid (CSF) of patients with neuropsychiatric lupus spans the range of concentrations, from low levels that affect synaptic function only to high levels that mediate cell death [14].

Modelling CNS lupus in mice carrying anti-dsDNA and anti-NMDAR antibodies

When mice are immunized with the DWEYS peptide in a multimeric configuration, they develop high titres of cross-reactive antibodies and display glomerular immunoglobulin deposition and proteinuria. Despite the presence of NMDAR reactivity in the serum of these mice, there is no detectable binding to or damage in neurons of the CNS [17]. We found that a breach in the integrity of the blood-brain barrier (BBB) was required for antibody to access brain tissue and affect neuronal function and viability. It is now apparent that there are multiple potential insults that can weaken the BBB [18]. Infection and stress are two common triggers by which the BBB can be breached. Patients with SLE are highly susceptible to infections by virtue of both the intrinsic immune suppression of the disease and the immunosuppressive medications used to treat it. As they also experience disease-related and nondisease-related stress, both of these conditions may represent clinically relevant triggers for the breach of the BBB. We chose to examine the ability of systemic administration of bacterial lipopolysaccharide (LPS) to trigger a breach of the BBB and thereby facilitate transit of antibody into brain parenchyma in mice harbouring high titres of anti-NMDAR antibody [17]. Histological analysis showed the binding of antibody to hippocampal neurons following LPS administration. Several days later, a loss of neurons was detectable, which was not progressive. The cessation of new damage was owing to the re-establishment of barrier integrity and the consequent decrease in available antibody in brain tissue. Structural damage was also visible by magnetic resonance spectroscopy of the mouse brain and was again highly localized to the hippocampus. Mice with hippocampal neuron loss displayed impairments in tasks measuring the flexible use of memory. The targeting of hippocampal neurons and the consequent memory deficit were observed whether the serum anti-NMDAR antibodies were of mouse or human origin [17,19].

To mimic the biological effects of stress, epinephrine was employed as the agent to breach the BBB in mice that had been immunized to produce anti-dsDNA, anti-NMDAR antibody [20]. Exposure to epinephrine led to the penetration of anti-NMDA antibody into the amygdala, a lower brain region with high concentration of the NMDAR. There was an ensuing loss of neurons in the amygdala and a consequent impairment of the mice in the performance of a Pavlovian fear-conditioning paradigm [20]. Under these conditions, there was no apparent structural or functional damage to the hippocampus. This observation supports the notion that agents that compromise the BBB exhibit regional specificity. It also provides an explanation for the development of distinct cognitive and behavioural alterations depending on the brain region in which the BBB is breached. Furthermore, these studies suggest that multiple manifestations of neuropsychiatric lupus might be attributable to the same antibody accessing different regions of the brain.

Foetal brains are vulnerable to anti-dsDNA and anti-NMDAR antibodies

There is a small but astonishingly consistent literature reporting an increased frequency of learning disabilities in the children of mothers with SLE [21–23]. Whilst not many children of lupus fathers have been studied, because of the much lower incidence of lupus in men, there are no data to support an increased risk for developing a learning disability in the children of fathers with SLE. We, therefore, reasoned that this observation might reflect the effects of in utero exposure to maternal antibody. It is known that maternal antibody crosses the placenta beginning at approximately the second trimester of pregnancy. It is also known that the full integrity of the BBB is achieved at around the time of birth. Thus, there is a considerable interval during which maternal antibodies are present in the foetal circulation and can access the developing brain.

To study whether anti-NMDAR antibodies in the mother might cause learning disability in the off-spring, we immunized female mice with amultimeric form of the DWEYS peptide, allowed them to become pregnant and analysed the offspring during foetal development and post-partum [24]. The foetal brains exposed to anti-NMDAR antibody displayed both increased apoptotic neurons and excessive mitotic neurons, including the presence of ectopic mitosis, by the 15th day of gestation (E15). The foetal brains also displayed a thin cortical plate. These anatomical changes were reflected in functional deficits after birth. During the first weeks of life, the offspring exposed in utero to anti-NMDAR antibody exhibited a transient delay in acquiring certain reflexes. As adults, these mice displayed impairments in tasks that are critically dependent on the cerebral cortex, although they were normal on a broad range of other behaviours, including grooming, social behaviours, motor skills, balance, navigation and memory function and fear conditioning. Specifically, they performed abnormally in tasks that assessed the recognition of novel objects and the spatial arrangement of objects. Further, they had a significant impairment in the extinction of fear responses. The associated histopathology of the animals exposed to high titres of anti-NMDAR antibodies in utero showed that they had a thinning of the cerebral cortex and that the cytoarchitectonics of the cortex was disorganized. These histopathological effects and cognitive phenotypes were dependent on anti-NMDAR antibody concentration, because mice exposed in utero to low titres of the toxic antibody had normal cortical histopathology and behaved normally. It is likely that some of the abnormal cortical changes might result from the toxic antibody affecting the normal process of radial neuronal migration; a phenomenon thought to be influenced by the activation of the NMDAR [25]. Orderly, radial cortical migration occurs early and persistently throughout gestation, whilst tangential cortical migration of neurons (particularly gamma amino butyric acid positive interneurons) occurs late in gestation and even post-natally [26]. Moreover, migration from the medial, lateral and caudal ganglionic eminences to the emerging amygdala and hippocampus occurs late in gestation, and, for the hippocampus, post-natally [27, 28]. We have demonstrated that access of the toxic anti-NMDAR antibody occurs during gestation; the antibody is not transferred to brains of newborn pups through lactation (Kowal and Diamond, unpublished data). This phenotype is not strictly analogous to the children of mothers with SLE, who exhibit defective reading and math skills [22, 23, 29]. It is, however, comparable as the affected mice display isolated impairments in cortically dependent behaviours. Thus, it is plausible that maternal anti-NMDAR antibody might contribute to the learning disabilities present in the children of mothers with SLE.

Prevention of antibody-mediated neuronal damage through antibody blockade

It is known that NMDAR antagonists can protect neurons from excessive activation of the receptor and from glutamate toxicity. Because we showed that anti-NMDAR antibodies, whether passively administered or actively produced by immunization, were neurotoxic, we explored whether NMDAR antagonists might protect against their pathogenicity. We treated mice with an FDA-approved noncompetitive NMDAR antagonist, memantine, after they had been immunized with a multimeric form of the DWEYS peptide that elicited high titres of anti-NMDAR antibody. Following systemic administration of epinephrine to cause a breach in the BBB, the potentially vulnerable neurons in the amygdala survived the exposure to anti-NMDAR antibody. This observation again confirmed that antibody-mediated neurotoxicity functioned via an NMDAR-activated pathway.

As memantine is increasingly used in patients with Alzheimer’s disease, there is a growing literature on the toxic effects of persistent NMDAR antagonism, and an equally rich literature on the dependence of normal cognitive processes on optimal activation of the NMDAR [30, 31]. Therefore, we tested whether an NMDAR-independent strategy to block the antigen-binding site of the antibody, which avoided interfering with NMDAR function, might be a better approach to neuroprotection. The DWEYS peptide, composed of d-amino acids (termed d-peptide), can bind to the cross-reactive antibodies [6]. To explore the therapeutic potential of soluble d-peptide, we first needed to determine that it did not self-aggregate and so would not form immune complexes when in the presence of antibody. Indeed, the DWEYS peptide does not form aggregates. We could then administer it to mice and be confident that, being so small, it was unlikely to be immunogenic and would serve as a nidus for the formation of immune complexes. Administration of soluble peptide prevented the R4A monoclonal anti-dsDNA, anti-NMDAR cross-reactive antibody from depositing in glomeruli. We then demonstrated that the administration of d-peptide could also protect neurons from the toxic effects of antibody exposure and was as effective as memantine [20]. Thus, soluble d-peptide could be both nephroprotective and neuroprotective. Of potential clinical importance, as a neuroprotective agent, the d-peptide would not interfere with appropriate NMDAR function [24]. The d-peptide was also given intravenously to gestating mice and protected the developing foetal brain from the toxic effects of exposure to the anti-NMDAR antibody. Foetal brains, extracted from dams treated with d-peptide, displayed normal cortical plate architecture, despite exposure to antibody.

The incidence of cross-reactive antibodies in SLE patient is high

Our investigations in animal models of SLE described previously are clinically relevant. We, and others, have shown that approximately 40% of lupus patients (range of 20–80%) have circulating anti-dsDNA, anti-NMDAR cross-reactive antibodies [32–34]. In several studies of lupus sera collected from patients in Norway, Canada, Japan and several US locations (Washington, DC, TX, New York City), most of whom had minimal-to-modest disease activity, approximately one-third to one-half of patients displayed anti-DWEYS peptide reactivity [34–37]. Essentially, all patients with anti-DWEYS peptide reactivity have anti-dsDNA reactivity. Thus, we have evidence that anti-dsDNA antibodies with this fine specificity are present in a large and diverse patient population.

Because of the link between anti-dsDNA antibody titres and the development of lupus nephritis, several investigators have asked whether elevated serum titres of anti-dsDNA and anti-NMDAR antibodies correlate with the presence of neuropsychiatric lupus symptoms. Whilst some reports have supported such a correlation [38–40], others have not [37]. Studies of antibody titres in CSF, however, have demonstrated a correlation confirming that such antibodies are not neurotoxic until they access brain tissue through a breach in the BBB [38, 39, 41, 42]. Indeed, anti-NMDAR antibody titres in the CSF are elevated in individuals with CNS manifestations of disease and decline as symptoms regress [38,39].

The therapeutic potential of the DWEYS peptide has been supported by studies of human lupus, ex vivo. It was found that in 9 of 10 lupus patients, with anti-DNA antibodies and lupus nephritis, the DWEYS peptide was able to inhibit DNA binding [43]. In five patients, 80%ormore of the DNA binding was inhibited by the DWEYS peptide. The DWEYS peptide has certain features that are potentially beneficial to therapeutic application. Presumably because of its small size, it is nonimmunogenic and it also appears to be nontoxic in mice. Peptides, however, have relatively short half-lives and cannot be given orally. Small molecules, with the protective activity of memantine and MK-801, that might mimic the peptide and occupy the antigen-binding site of the antibody, would be highly desirable. We have recently identified a small molecule that can prevent antibody from binding kidney glomeruli ex vivo and can protect neurons from toxicity after exposure to anti-NMDAR antibodies (Fig. 2; unpublished results).

Fig. 2.

Rational design of drugs for protection against autoantibodies. Left panel, antibody binds target autoantigen (DNA) initiating disease pathology and cross-reacts with a peptide that can prevent binding to target autoantigen. Right panel, antibody may cross-react with a small molecule with more favourable therapeutic characteristics. Either peptide or small molecule may function as atherapeutic agent.

The therapeutic implications of these studies apply to all antibody-mediated pathology

We have identified a new antigenic specificity, the NMDAR, of a subset of anti-DNA antibodies. We have been able to show that antibodies with this specificity are present in the CSF and so can contribute to the symptoms that are observed in NPSLE, following a compromise of the BBB. In a mouse model, we have demonstrated how these antibodies alter neuronal function and cause neurotoxicity. We have used antigenic specificity to develop a strategy to neutralize the antibodies. Whilst our screening of a peptide library yielded a sequence from a biological target antigen, the NMDAR, peptide that blocks the antigen-binding site of a pathogenic antibody, may be therapeutic, even if the peptide is only a structural mimetope of the target antigen and not an actual antigenic epitope. Therapeutic efficacy will depend not only on half-life but also on its remaining amonomer in solution so as not to be immunogenic and form immune complexes, and on the absence of unforeseen toxicities. Even as we continue to seek cures to autoimmune diseases, this therapeutic approach can spare the damaging effects of autoantibodies on target organs, including the kidneys, the brain or other organs. This approach can protect against both transient and irreversible effects of autoantibodies.