Cytokines such as IL-1, TNF and IL-6 are key regulators of inflammation at sites of infection or tissue damage. Following AIS, it seems reasonable to hypothesise that cytokines measured in the plasma arise from cells associated with inflammatory activity in the brain, but this is not certain. Therefore, in addition to evaluating the relationships between plasma cytokines and the clinical impact of stroke, we have attempted to identify the possible contribution of peripheral blood leukocytes to the plasma cytokine pool. We also wanted to know whether leukocyte activation, apparent in terms of increased numbers in the circulation [6
], was associated with an increased potential to produce cytokines. In this study our primary focus has been the IL-1 system, but we also examined other cytokines, their receptors, and neopterin, as molecules that have previously been associated with inflammatory cell activation.
In accord with its localised role, and with previous studies [3
], there was no evidence of an increase in plasma IL-1β concentration. We also found no change in plasma sIL-1-RII concentration. However, the elevation in plasma IL-1ra concentration, at a much earlier stage after stroke onset than previously reported [10
], suggests early activation of anti-inflammatory components directed at regulating IL-1. Leptin was evaluated because it appears in some cases to mediate neuroimmune responses dependent on IL-1 [23
] and, although high leptin levels were recently found to be associated with stroke [24
], the present data support neither a clear change in leptin concentration following stroke, nor any relationship with outcome.
Our findings with respect to TNF-α are consistent with one previous study [8
], but differ from those showing that plasma or serum TNF-α concentrations after stroke are increased and correlate with cerebral infarct volume [12
]. The role of TNF-α in stroke remains controversial, and indeed a potentially beneficial role in ischaemic tolerance has been suggested [25
]. However, we found sTNF-RI was increased at 5–7 d and the peak concentration in the first week correlated with infarct volume and outcome (mRS and BI) at 3 months and 1 year. This is consistent with earlier observations of elevated sTNF-RI concentration in stroke patients with nosocomial infection [26
]. A more robust relationship between sTNF-RI and stroke severity or outcome may reflect a longer half-life of receptors compared to TNF-α. The possibility that TNF-α and its receptors are activated in concert is supported by the strong correlations between their peak concentrations and suggests their release is proportionate to each other. The fact that we used controls matched for atherosclerosis, and that elevated serum sTNF-RI and sTNF-RII concentrations are increased in carotid atherosclerosis [27
] may explain why our data are not as clearly differentiated in respect of TNF-α as those where healthy controls were used.
Atherosclerosis may also be a confounding factor with IL-10 analysis, since patient plasma IL-10 concentrations were similar to those of controls and reduced concentrations of plasma IL-10 have been noted in studies where plasma of patients with stroke has been compared to that of healthy controls [15
]. In support of this explanation, higher plasma concentrations of IL-10 have been detected in plasma of patients with stable, versus unstable coronary syndromes [28
] and, experimentally, IL-10 deficiency results in increased atherosclerosis in apolipoprotein E-deficient mice [29
Activation of cells of monocytoid lineage is associated with increased production of neopterin in plasma or urine. Although we found urinary neopterin to be increased following stroke, and this is consistent with a previous report that found neopterin to be increased during the first week after stroke [30
] we also found it to be higher in control subjects with significant atherosclerosis. The association with atherosclerosis has been reported previously [31
] and it is possible that this influences the strength of the apparent association. Consistent with this interpretation, another study reported neopterin as being higher one year after stroke than at the acute stage, and to be correlated with age [32
CSF is arguably more relevant than plasma to events occurring in the brain. However, lumbar puncture poses a risk of cerebral herniation and it is difficult to obtain CSF at repeated time points. Plasma, or serum markers therefore have potentially greater practical value and have been evaluated in most studies. It may be that plasma cytokines do reflect brain cytokines, particularly after an acute cerebral event, but the validity of this has not been well established, and blood monocytes in particular are potential sources of plasma cytokines. This is particularly the case where serum, rather than plasma, has been collected and there is potential for monocytes to release cytokines during clotting, particularly if endotoxin is present in collection tubes [33
]. The ability of blood cells to produce cytokines is potentially relevant, not only because they may be responsible for producing the cytokines while in the circulation but because these cells have the potential to enter the CNS after stroke and their activation state may reflect systemic responsiveness of the immune system after stroke. Our determination of cytokine gene activation in the blood of patients, and of spontaneous, whole blood cytokine induction, indicates peripheral blood cells are not the source of cytokines after ischaemic stroke. This supports the hypothesis that cytokine production is upregulated within the brain, possibly by microglia, in response to cerebral ischaemia, and these cytokines enter the peripheral circulation via reabsorption from the CSF, or via afferent lymphatics. While cells in other peripheral tissues, such as those associated with concomitant peripheral atherosclerosis, might also be expected to be contributors to a pre-existing inflammatory state, it seems unlikely that they would be more readily activated than blood cells following a stroke.
In contrast to the increased plasma cytokines, presumed to originate from the brain, we report, for the first time, reduction of cytokine production by endotoxin-stimulated whole blood from patients relative to controls. This was seen within hours of ischaemic stroke and persisted during the first week. These data differ from a previous report, where LPS-induced more IL-6 and TNF in blood from patients with stroke, compared to controls [19
]. However, they are consistent with a number of studies examining cytokine production by whole blood from trauma patients [35
]. Reasons for conflict with the earlier study [19
] are not clear, although in that study: (1) cytokines were induced for only 4 h by LPS, which might be relevant if downregulation is relatively delayed; (2) data were not corrected for monocyte count; (3) non-normally distributed data were analysed with parametric statistical tests and (4) controls were not matched for degree of atherosclerosis. Suppression of endotoxin-stimulated cytokine production, and the inverse relationships with stroke severity and outcome, may relate to the extent of CNS inflammation and a physiological attempt to downregulate cytokine activation, although it should be noted that the correlation with infarct volume was relatively weak. The mechanisms are therefore unclear although, following trauma, a suppressive factor has been shown to be present in serum and this was apparently not transforming growth factor (TGF)-β, IL-4 or IL-10 [36
]. IL-10 seems unlikely to have contributed to the relative reduction in cytokine production observed in the present study because its production pattern was similar to those of IL-1β, IL-6 and TNF-α. It seems unlikely that reduced capacity to produce cytokines is due to depletion in vivo, since there was no evidence for increased IL-1β or TNF-α in the plasma, and mRNA for all three cytokines had not been induced at baseline.
We have previously reported elevation in plasma cortisol concentration within 12 hours of ischaemic stroke [6
] and the correlation between plasma cortisol and suppression of cytokine production, suggests this might be a potential explanation. These observations are consistent with experimental data describing a systemic anti-inflammatory response, including reduced endotoxin stimulated whole blood TNF-α and increased IL-10 production, after infusion of IL-1β into the brain, through stimulation of the hypothalamo-pituitary-adrenal axis and the sympathetic nervous system [37
]. However, evidence from experimental ischaemia suggests that activation of the sympathetic nervous system, but not corticosteroid receptors, is more important in suppression of whole blood TNF production [38
]. Both catecholamines and corticosteroids suppress endotoxin-induced cytokine production by whole blood from healthy volunteers [39
]. Such regulatory responses, and increases in inhibitory cytokines such as IL-1ra, may be responsible for increased vulnerability of stroke patients to infections [38
]. This might contribute toward the increased mortality in patients with greatest inhibition of systemic cytokine induction.
Particular strengths of the present study are the very early recruitment of patients after onset of symptoms of AIS (median < 5 h), and further assessments within the first 24 h of admission. In studies where later sampling time points have been used it is difficult to distinguish the effects of the stroke itself from subsequent infections or other events. We also recognise that cytokine responses seen at 3 months and 1 year reflect the fact that many patients with the most severe strokes did not survive to later time points. Control subjects were individually matched for degree of atherosclerosis, in addition to age and sex, because each of these variables are known to influence peripheral inflammatory markers, including IL-6 and neopterin [43
]. The rationale for carotid artery stenosis and ABPI thresholds for 'significant atherosclerosis' used in the present study was guided by the available evidence, as discussed previously [6
]. Possible criticisms of the present study might include our relatively small sample size and multiple statistical testing. We may have missed some associations through the former, and may have identified spurious associations through the latter. However, we clearly stated the primary questions at the study outset and believe that we are reporting data that should provide an important stimulus for further study.