Increased Production of Tumor Necrosis Factor α, and Not of Interferon γ, Preceding Disease Activity in Patients With Multiple Sclerosis

Objective  To study whether tumor necrosis factor (TNF) α or interferon (IFN) γ production by stimulated white blood cells precedes or accompanies clinical and magnetic resonance imaging signs of disease activity in patients with multiple sclerosis.

Design  Prospective study with a follow-up of 9 months.

Setting  Patients visiting an outpatient university clinic.

Patients  The 30 Amsterdam-based patients (28 completing all evaluations) participating in a multicenter, randomized, placebo-controlled, double-blind trial of a chimeric anti-CD4 antibody in the treatment of active relapsing-remitting and secondary progressive multiple sclerosis. Patients in both treatment arms were included, because for these patients anti-CD4 treatment in this study did not affect TNF-α and IFN-γ production and did not reduce signs of disease activity on magnetic resonance imaging.

Main Outcome Measure  Distribution of classes of TNF-α and IFN-γ production (expressed as z scores) in patients with or without clinical or magnetic resonance imaging signs of disease activity.

Results  One month preceding exacerbations of multiple sclerosis, there was a shift toward higher z scores of TNF-α production (P<.05), but not of IFN-γ production. There was no statistically significant relationship between IFN-γ and TNF-α production and magnetic resonance imaging markers of multiple sclerosis activity.

Conclusion  The production of TNF-α, and not of IFN-γ, is significantly higher in patients with multiple sclerosis before exacerbations than in patients with stable disease. Although present, this relationship is too weak to use TNF-α production as a surrogate marker of disease activity in multiple sclerosis.

ALTHOUGH THERE is still no generally accepted model of the cause of multiple sclerosis (MS), it is clear that inflammatory events in the central nervous system of patients with MS play a role in the pathogenesis of this disease. In active MS plaques, a perivascular infiltration of CD4-positive T-helper lymphocytes and macrophages is present.^1-3 As an analogy to the animal model of MS, experimental allergic encephalomyelitis, it seems likely that cells of the T-helper 1 type, which produce the (pro)-inflammatory cytokines tumor necrosis factor (TNF) α, TNF-β (or lymphotoxin), interferon (IFN) γ, and interleukin (IL) 1, are involved in the initiation of inflammation in MS, whereas T-helper 2 cells, which produce the anti-inflammatory cytokines IL-4, IL-10, and transforming growth factor β, appear in the recovery phase.^4,5 Recent research established the presence of T-helper 1 cytokines and, to a lesser extent, T-helper 2 cytokines in MS plaques, although they were also present (generally at lower levels) in the brains of patients with other inflammatory and even noninflammatory central nervous system diseases.^6,7

Experiments in patients with MS, which showed that peripheral blood mononuclear cells express higher levels of TNF-α and TNF-β messenger RNA in patients with active disease than in those with stable disease and normal controls,^8,9 are also in line with the hypothesis that T-helper 1 cytokines could be relevant to the pathogenesis of MS. Furthermore, increased production of TNF-α, IFN-γ, and IL-1 precedes clinical exacerbations in MS in experiments that used a whole-blood mitogen stimulation assay^10-12 or lipopolysaccharide or 12-O-tetradecanoylphorbol 13-acetate stimulation of monocytes and macrophages.¹³ Various investigators could correlate disease activity to TNF-α levels in cerebrospinal fluid of patients with MS.^11,14-16 However, others did not find this correlation^17,18 or could not detect TNF-α in the cerebrospinal fluid of patients with MS.^19-21

Other experiments found evidence of an association between high levels of messenger RNA coding for the T-helper 2 cytokine IL-10 in peripheral blood mononuclear cells of patients with MS who had stable disease.⁸ Accordingly, other investigators found that stimulated myelin protein–specific T-cell clones of patients with MS produced more IL-10 and transforming growth factor β in the recovery phase than during exacerbations,²² or demonstrated lower levels of serum IL-10 in patients with MS than in healthy controls.²³

The majority of the studies mentioned above were cross sectional and do not give insight into the temporal relationship between cytokine production and disease activity. Drawing conclusions from these studies is also hampered by the fact that cytokine production varies between individuals, irrespective of the presence and activity of disease.²⁴ Most likely this variation is influenced by genetic determinants.²⁵ The few studies that provide longitudinal data investigated the relationship between cytokine production and clinical measures of disease activity. However, it has become clear during past years that serial magnetic resonance (MR) imaging detects many active lesions in the central nervous system of patients with MS that do not give rise to clinical symptoms,^26-28 and is therefore presumably a better marker for disease activity than clinical markers. So far, only 1 prospective and longitudinal study has aimed at correlating cytokine production to both clinical and MR imaging measures of disease activity.²⁹ This study included few patients and did not find a statistically significant association between serum levels of TNF-α and TNF receptor and MR imaging activity.

The present study included 30 patients with MS, who were studied monthly for 9 months in the context of a clinical trial, and investigated the relationship between TNF-α and IFN-γ production by peripheral blood mononuclear cells and clinical and MR imaging measures of MS activity in a prospective and longitudinal manner.

This study included 30 patients who were enrolled in Amsterdam, the Netherlands, in a multicenter, randomized, double-blind, placebo-controlled phase 2 trial of anti-CD4 antibodies in the treatment of MS.³⁰ Patients from both treatment arms were included in this study because TNF-α and IFN-γ production were not significantly affected by treatment with anti-CD4 in these patients,^31,32 and because the anti-CD4 treatment did not result in a statistically significant reduction in the number of active MR images during 9 months (primary outcome measure).³⁰

Inclusion and exclusion criteria, as well as the treatment protocol of this trial, are given elsewhere.³³ In short, patients with an active form of relapsing-remitting or secondary progressive, clinically definite MS³⁴ were included (either 2 relapses within the previous 12 months, 1 of which was in the last 6 months but not within 2 months of study entry, or deterioration on the Kurtzke Expanded Disability Status Scale [EDSS]³⁵ of at least 1 point within the last 18 months). They had to be between 18 and 55 years of age and to have an EDSS between 3.0 and 7.0 (inclusive) at study entry. We excluded patients who had received treatment with immunomodulatory drugs in the 2 years preceding study entry or treatment with corticosteroids within 4 weeks before study entry. All patients gave their informed consent for all study procedures.

We recorded the frequency of relapses and methylprednisolone interventions monthly for 9 months. Clinical examination (expressed as an EDSS total score) and MR imaging were performed 1 month before and immediately before the start of treatment (baseline) and were repeated at monthly intervals for 9 months. The imaging protocol was based on guidelines prepared by the MRI Steering Committee of the Commission of the European Community–funded European Concerted Action on MS,³⁶ and is described elsewhere.³⁰ Briefly, 5-mm axial slices were obtained through the brain by means of dual-echo, long–repetition time (T₂-weighted) and short–repetition time/short–echo time (T₁-weighted) spin-echo images, with the use of a 0.6-T imager (Technicare, Solon, Ohio). Repositioning between consecutive images was achieved by using anatomic landmarks identified from coronal and sagittal pilot scans. Starting 5 to 10 minutes after intravenous injection of gadopentetate dimeglumine dimethylenetriaminepentaacetatogadolinate, 0.1 mmol/kg, T₁-weighted spin-echo images were obtained. If a patient experienced a relapse, 1000 mg of methylprednisolone per day for 3 consecutive days could be given intravenously. If this was within 1 week of the next scheduled MR imaging, the imaging date was brought forward and performed immediately before corticosteroid treatment was started (because of the transient suppressive effect of corticosteroids on gadopentetate enhancement); the following monthly MR imaging was performed at the usual time. If corticosteroids were given more than 1 week before the next scheduled MR imaging, the latter was performed at the usual time. Two experienced readers (F.B. and Ivan Moseley, PhD, London, England), who were blinded to the randomization code and the clinical and laboratory findings, defined in conference the number of active lesions on each examination by comparing each image with the previous one. Active lesions were defined as lesions that showed new gadopentetate enhancement on T₁-weighted images, and as enlarging or new lesions on T₂-weighted images that were not seen on gadopentetate-enhanced T₁-weighted images.

Blood samples for TNF-α and IFN-γ stimulation assays were taken monthly for 9 months, on the same day that MR images were made. Assessment of TNF-α and IFN-γ production was performed on blood obtained by venous puncture with the use of a specimen tube system (Venoject, Terumo Europe NV, Leuven, Belgium) that contained endotoxin-free heparin obtained from the Free University Hospital Pharmacy, Amsterdam.

Immediately after collection, 200 µL of whole blood was stimulated at 37°C and 5% carbon dioxide atmosphere in 2 mL of RPMI medium with HEPES, to which penicillin, streptomycin, and glutamine were added, with phytohemagglutinin, 4.5 µg/mL, and lipopolysaccharide, 22.5 µg/mL (both supplied by Medgenix, Fleurus, Belgium) in accordance with a protocol described previously.³⁷ Stimulation was carried out for 24 hours to measure TNF-α production and 48 hours to measure IFN-γ production. This protocol appeared to result in such high TNF-α production that variability between samples was lost. Therefore, we did not use these data in our calculations. We changed the method of TNF-α assay after the initial months of the trial. All assays were done in the new manner and stimulated with phytohemagglutinin-lipopolysaccharide for only 2 hours. The resulting supernatants were stored at −80°C. The TNF-α and IFN-γ levels were assessed in the supernatant by means of enzyme-linked immunosorbent assay kits (Easia and Screening Line, Medgenix, Fleurus, Belgium) according to the manufacturer's instructions.

Statistical analysis was done with the SPSS/PC+ package (Version 4.0; SPSS Inc, Chicago, Ill). We related the distribution of TNF-α and IFN-γ levels to the presence of clinical relapses and active lesions on MR images. Levels of TNF-α and IFN-γ were analyzed after transformation into z scores for individual patients. This was done because there are major differences in baseline levels of TNF-α and IFN-γ production between individuals. Transformation into z scores enabled us to relate deviations from an individual's baseline TNF-α or IFN-γ production to clinical or MR imaging measures of disease activity. For instance, a z score of +1 represents a cytokine production that is 1 SD above a certain individual's mean production. We created 5 classes of z scores (z<−2, −2≤z<−1, −1≤z<1, 1≤z<2, and z≥2) and tested for differences of the distribution of z scores over these classes between patients with active and nonactive disease (Mann-Whitney test), using both clinical (exacerbations) and MR imaging measures of disease activity. We analyzed TNF-α and IFN-γ production in relation to measures of disease activity (exacerbations and MR imaging activity), not only on the same day, but also 1 month preceding assessment of disease activity and, as a control, 1 month after assessment of disease activity.

Our group consisted of 30 patients (16 female and 14 male) and included 17 patients with a relapsing-remitting form of MS and 13 patients with secondary progressive MS. Fifteen patients each were included in the placebo group and the anti-CD4 group. Two placebo-treated patients withdrew from the trial after their first infusion and were unavailable for follow-up. Further characteristics of the patients are given in Table 1 and Table 2.

Except for the 2 dropouts, there were no missing clinical or MR imaging observations; the total percentage of missing TNF-α and IFN-γ assays was 4.1% and 4.3%, respectively (because of incidental reasons).

The individual mean TNF-α and IFN-γ production in our patients varied widely. The mean individual cytokine production ranged from 428 to 3630 pg/mL for TNF-α and from 50 to 924 U/mL for IFN-γ. We found no relationship between individual mean levels of production of any of these cytokines and the number of clinical exacerbations, the change in the EDSS, or the cumulative number of active lesions on MR images during the entire duration of this study (data not shown).

As shown in Table 3, there was a statistically significant shift toward higher z scores in patients 1 month preceding a clinical disease exacerbation compared with clinically stable patients (P<.05). This rise in TNF-α production expressed as z scores was short-lived: when measured during clinical signs of disease activity, the production was comparable with that in clinically stable patients. As a control, we also calculated z scores for both groups 1 month after disease activity and found no statistically significant differences.

From these data, we calculated the probability that an individual patient with MS will experience an exacerbation, given a certain TNF-α production, expressed as a z score, 1 month earlier. Although higher z scores were associated with a higher probability of suffering a clinical exacerbation 1 month later, this probability was still only 25% in patients who produced at least 2 SDs above their baseline (Table 4).

When we performed the same calculations as described above for TNF-α on the IFN-γ data, we did not find any statistically significant differences between patients with active and stable disease at any of the 3 time points (Table 5). We also did not find any statistically significant differences between z scores of TNF-α and IFN-γ production in patients with or without active lesions on MR images at any of the 3 time points (Table 6).

We longitudinally studied the relationship between the production of the inflammatory cytokines TNF-α and IFN-γ and clinical and MR imaging–based MS disease activity. This study, because of the longitudinal design, allowed the assessment of the relationship between individual levels of cytokine production and both present and future disease activity, as well as the relationship between cytokine production at each time point and the mean baseline production for that individual.

Using various time intervals, we made 2 comparisons for each cytokine: (1) relative production vs presence or absence of clinical exacerbation and (2) relative production (expressed as the distribution of z scores, to indicate deviations from an individual's baseline to correct for differences in production between individuals) vs presence or absence of MR imaging activity. We chose not to use absolute cytokine production because of the wide variability between individuals, as exemplified in the "Results" section.

We found a significant shift toward higher z scores for TNF-α production in patients 1 month preceding a clinical exacerbation, compared with clinically stable patients (P<.05). As can be derived from Table 3, z scores were elevated with at least 1 SD above an individual's mean score in 50% of patients 1 month preceding an exacerbation, whereas this was found in only 18% of patients who were clinically stable 1 month later. The z scores for TNF-α production were not higher in patients 1 month before or during MR imaging signs of disease activity.

For IFN-γ production–derived z scores, we did not find any significant relationship with clinical or MR imaging signs of disease activity. We think that our results provide reliable evidence that clinical exacerbations of MS are often preceded by an increase in TNF-α production.

Our results are in line with most other studies that investigated a possible association of an increase of TNF-α production and MS activity. What is also clear, however, is that the association between TNF-α production and clinical and MR imaging measures of disease, although statistically significant, is rather weak. Only 50% of clinical exacerbations were preceded by a rise of TNF-α production of at least 1 SD above an individual's mean. Inversely, only 25% of patients who produced at least 2 SDs above their mean had developed clinical signs of disease exacerbation 1 month later. This can probably be explained in part by the fact that not all peaks of TNF-α production were noted with the use of monthly blood samples.

Our findings regarding IFN-γ suggest there is not a strong relationship between IFN-γ production and MS activity. However, it is possible that short pulses of increased IFN-γ production are sufficient to lead to changes in disease activity, and many of them could be missed by monthly sampling. Results of several other studies in MS and experimental allergic encephalomyelitis favor a role of IFN-γ in these diseases, although the exact role of IFN-γ is not yet clear. Administration of IFN-γ to patients with MS increases disease activity.³⁸ Treatment of patients with MS with IFN-β, which has been shown to favorably affect the course of MS,³⁹ resulted in an initial increase in IFN-γ production.⁴⁰ In the experimental allergic encephalomyelitis model, treatment with IFN-γ or anti–IFN-γ has been shown to have unpredictable effects.^41,42 These observations point to the fact that the role of IFN-γ in autoimmune disease of the central nervous system may be very complex.

Our results lead to the conclusion that TNF-α is only 1 of more factors that determine whether there will be disease activity in the near future in an individual patient. We suggest that future studies on the relationship between cytokine production and MS activity will have to use a panel of several T-helper 1 and 2 cytokines, circulating cytokine receptors, adhesion molecules, and possibly also other as yet unknown factors to be able to adequately explain all changes in disease activity. Following from this line of reasoning, we think it is unlikely that TNF-α assays can be effectively used to monitor disease activity in individual patients with MS, or can be used as a surrogate marker of disease activity in future MS treatment trials.

The data obtained in this study do not support our assumption that measures of disease activity on MR imaging would show a closer association with cytokine production than clinical measures would. We can only speculate that changes in cytokine production preceding isolated disease activity on MR imaging in the absence of a clinical exacerbation are of a lower magnitude and therefore cannot be identified among the many fluctuations that are caused by the other factors that seem to be involved.

Accepted for publication November 7, 1997.

We thank Medgenix (Fleurus, Belgium) for supplying the TNF-α and IFN-γ enzyme-linked immunosorbent assay kits. We also thank Ton Schweigmann for performing the MR imaging.

Reprints: C. H. Polman, MD, Department of Neurology, Free University Hospital, PO Box 7057, 1007 MB Amsterdam, the Netherlands.