Central nervous system (CNS) involvement has been reported to occur in 14–75% of all systemic lupus erythematosus (SLE) patients [1–3]. The large differences regarding the frequency depend on the diagnostic criteria applied. CNS lupus can occur at any time during the course of SLE and its symptoms are extremely diverse. The features of this condition may include seizures, stroke, depression, psychoses and disordered mentation. Dementia, a common state in the population in general, is occasionally also reported in SLE patients [4]. Neuropsychiatric involvement in SLE (NPSLE) has been shown to predict a high frequency of flares and is considered a major cause of longstanding functional impairment as well as a cause of mortality [5]. CNS lupus in recent decades has been treated with cytotoxic drugs that improve the disease outcome [6, 7]. However, acquisition of valuable treatment remedies increases the need for early recognition of CNS manifestations in lupus and continuing evaluation of local (i.e. intrathecal) responses to the medication.

Due to the multiple pathogenic mechanisms causing manifestations of CNS lupus, there is no single confirmatory diagnostic test available. Several clinical, laboratory, and radiographic test findings are reported to be abnormal in some but not all patients with CNS lupus. Magnetic resonance imaging (MRI) of the brain has been shown valuable in detecting even minor lesions caused by CNS lupus and correlated to CNS manifestations in SLE [8]. Pleocytosis and elevated protein levels are found in some but not all patients with CNS lupus [9, 10]. Elevated concentrations of IgG in the cerebrospinal fluid (CSF), of the IgG–albumin ratio and of the IgG index, and the presence of oligoclonal bands have all been described with varying frequencies in patients with NPSLE [11–13]. A few studies have demonstrated elevated IL-6 levels in the CSF from patients with CNS lupus [14–18]. Some other reports have described increased levels of IL-1 [14], of IL-8 [16] and of interferon gamma [19] in CSF from patients with CNS lupus. All these biochemical indices are indirect measures of brain inflammation. In contrast to a multitude of studies analyzing local inflammatory response in NPSLE, the evaluation of neuronal damage and the formation of toxic metabolic products such as β-amyloid protein (Aβ42) have not been assessed at all in this condition.

Protein tau is a microtubuli-associated protein that promotes assembly and stability of microtubuli [20]. Elevated CSF-tau levels are found in neurodegenerative disorders such as Alzheimer disease (AD) and in acute CNS disorders such as stroke [21], and they reflect neuronal and axonal degeneration and damage. In contrast, decreased CSF-Aβ42 levels reflect either increased deposition of this molecule in senile plaques and cerebral blood vessels (e.g. commonly occurring in AD) and/or decreased synthesis of β-amyloid precursor protein (APP) [22], a large transmembrane protein. Numerous studies have found increased CSF-tau levels, decreased CSF-Aβ42 levels but variable levels of CSF-APP in AD [22–25].

Transforming growth factor beta (TGF-β), produced by both glial cells and neuronal cells within the brain [26], is a multifunctional cytokine of importance in many physiological processes, including vascular development, immune responses and fibrosis [27–29]. TGF-β in AD is present at increased levels in the CSF [30], and has also been detected in senile plaques.

The aim of the present study was to measure the levels of brain-specific proteins directly or indirectly related to dementia disorders (tau, Aβ42, APP and TGF-β) in the CSF of SLE patients with and without CNS involvement, and to assess these data prospectively with respect to immunosuppressive treatment. Our results indicate that the SLE patients studied display significantly decreased levels of Aβ42 and APP, irrespective of their CNS engagement. Prospective evaluation of a subgroup of these patients indicates that combined anti-inflammatory/immunosuppressive treatment further downregulated expression of these molecules. Despite this clear down-regulation of Aβ42, none of the SLE patients included in the study fulfilled the criteria of dementia of any type.

One hundred and fourteen patients met the inclusion criteria for SLE diagnosis. Thirty-six patients were found to have NPSLE in accordance with the criteria presented in Materials and methods, seven patients were found to have phospholipid antibody syndrome and the remaining 71 SLE patients were considered cerebrally healthy.

Mild pleocytosis was seen in patients with CNS lupus (10 × 10⁶ ± 7 × 10⁶ cells/l) compared with cerebrally healthy SLE patients (2 × 10⁶ ± 0.5 × 10⁶ cells/l) (not significant). As previously validated [39], we found an increased number of oligoclonal bands in the CSF of the CNS lupus group (2.2 ± 0.4) as compared with SLE patients without CNS involvement (0.4 ± 0.2) (P ≤ 0.0005). The mean level of the CSF/serum albumin ratio was not increased in patients with NPSLE (5.5 ± 0.4 mg/dl) as compared with cerebrally healthy SLE subjects (5.2 ± 0.4 mg/dl) (not significant). There were no significant differences regarding levels of serum antibodies specific for native DNA, or regarding complement levels (C3 and C4) between SLE patients with or without CNS involvement.

Intrathecal Aβ42 levels were decreased in all the SLE patients compared with healthy controls (mean ± standard error of the mean, 574 ± 17 pg/ml versus 831 ± 25 pg/ml; P < 0.0001) (Fig. 1a). Patients with NPSLE had slightly but not significantly decreased Aβ42 levels as compared with patients without CNS lupus (541 ± 34 pg/ml versus 583 ± 18 pg/ml, not significant).

A second lumbar puncture was performed in nine patients who met the criteria of NPSLE and were successfully treated with cyclophosphamide. The CSF levels of Aβ42 were further decreased on the second occasion as compared with the first (702 ± 45 pg/ml versus 621 ± 45 pg/ml, P ≤ 0.05). In contrast, CSF levels of tau were not affected by treatment (275 ± 77 pg/ml versus 273 ± 61 pg/ml, not significant). To investigate whether the decreased Aβ42 levels in SLE patients are due to a diminished production or to its local deposition in the brain, we analyzed the intrathecal levels of APP. We found decreased CSF-APP levels (Fig. 1b) in SLE patients compared with healthy controls (4834 ± 314 pg/nl versus 7010 ± 714 pg/nl, P < 0.001), clearly indicating decreased production of this precursor molecule.

There were no statistically significant differences regarding CSF-tau levels between healthy controls and patients with NPSLE (173 ± 16 pg/ml versus 237 ± 35 pg/ml). However, upon stratification of all the SLE patients into two groups, according to the presence or absence of MRI findings, we found a statistical difference regarding CSF-tau levels in SLE patients with MRI pathology compared with those without (261 ± 26 pg/ml versus 164 ± 15 pg/ml, P < 0.01) (Fig. 1c). Next we assessed the possible relationship between decreased Aβ42 and increased tau levels in the whole patient population. Our results show clearly that these two molecules, related to neuronal toxicity and damage, display a significant relationship (r = 0.48, P < 0.0001).

We finally decided to assess CSF levels of TGF-β, another amyloidogenic protein with anti-inflammatory properties. In the CNS lupus group, intrathecal levels of TGF-β were significantly increased compared with those of SLE patients without overt CNS disease (mean ± standard error of the mean, 54.3 ± 8.4 pg/ml versus 31.6 ± 4.8 pg/ml; P ≤ 0.01), as seen in Fig. 1d.

Patients with SLE present with a wide array of neuropsychiatric features, although evidence for dementia in these patients is scarce.

The precise pathogenic mechanisms of neuropsychiatric manifestations are still the subject of intense investigations, but autoantibody and cytokine-mediated neural dysfunction, intracranial angiopathy and coagulopathy have all been implicated. Although there is no diagnostic golden standard for CNS lupus, there is a wide selection of non-invasive tests that are of value in the assessment and monitoring of the individual patients. Autopsy studies of brains from SLE patients have revealed vasculopathy [40] as a pathogenic event either directly responsible for clinical neuropsychiatric events or, alternatively, by altering blood–brain barrier permeability and facilitating the access of pathogenic antibodies from the circulation into an immunologically privileged site that is normally protected from aberrant immune responses. The extracellular fluid of the brain is in direct contact with the CSF, and biochemical processes in the brain can be reflected therein. Analysis of degradation products from neuronal cells found in CSF would be of value to improve the clinical evaluation of NPSLE and to study the ongoing brain parenchyme status in living subjects with NPSLE.

In the current study, we found significantly decreased levels of Aβ42 in the CSF of SLE patients compared with healthy control subjects. Low levels of CSF-Aβ42 may be either due to their accumulation and deposition (e.g. as diffuse plaques, as in the case of AD) and/or to a decreased production of APP by neuronal cells. Our results showed significantly decreased levels of APP in SLE patients included in the study compared with neurologically healthy subjects, indicating decreased production rather than increased tissue deposition. What would be the cause of the decreased APP production in SLE patients? At first glance, there were no differences in the CSF-APP levels with respect to the medication (corticosteroids, immunosuppressive drugs). However, the variability of the patient material as well as differences in duration and intensity of the treatment throw doubt on this conclusion. Interestingly, a subpopulation of nine patients was followed prospectively after immunosuppressive treatment. In this group such a treatment resulted in significantly decreased Aβ42 levels. This finding indicates that treatment with alkylating agents such as cyclophosphamide might affect either the production of APP or the activity of secretases. Indeed, it has been recently demonstrated that anti-inflammatory drugs may directly modulate γ-secretase activity and thereby decrease Aβ42 production [41].

We showed that SLE patients with MRI findings compared with those without MRI findings had significantly higher CSF-tau levels. These increases reflect the neuronal damage of the brain in SLE patients. Indeed, our recent study [32] supports this conclusion since another neuronal protein, neurofilament protein, was also increased.

Importantly, a significant relationship was noted between the occurrence of tau and of Aβ42. This finding, together with the correlation between tau and MRI-verified brain damage, indicates that Aβ42 might have exerted its toxicity locally, leading to sequels. Finally, as previously found in AD [30] and in low-pressure hydrocephalus-induced dementia (unpublished observation), levels of TGF-β in the CSF of NPSLE patients were clearly increased. Such an increase may be very well related to the disease process itself since, at least in the case of patients with low-pressure hydrocephalus, a shunt operation will simultaneously lead to decreased levels of TGF-β and increased mentation (unpublished observation).

Altogether, our results show that there is increased production of TGF-β and an ongoing destruction of brain parenchyma (manifested as increased tau levels) in the brains of SLE patients. We suggest that levels of toxic Aβ42 and its precursor APP reflect efficient cytotoxic therapy, as shown in a limited subgroup analysis. Notably, while our findings shed some light on the pathogenesis of cerebral lupus, none of the findings reported in the present study may be used as a diagnostic parameter, since many other brain diseases show similar patterns.

Our findings indicate that intrathecal levels of tau, a marker of neuronal degeneration, are clearly increased in SLE patients with MRI-verifiable brain lesions, both as compared with CSF findings of healthy controls as well as of lupus patients without MRI-verified brain pathology. This finding, together with increased intrathecal levels of TGF-β, indicates an ongoing destructive parenchymatous process in the SLE brain.

In addition, the present study provides evidence of decreased intrathecal levels of Aβ42 in SLE patients irrespective of CNS engagement. We interpret this finding not as a sign of early dementia, but rather as a sign of decreased APP synthesis, a precursor of Aβ42, as verified in the present study. We suggest that this decrease may be an outcome of an intense immunosuppressive treatment to which the majority of SLE patients included in the study are being subjected.