Post-stroke inflammation—target or tool for therapy?

The most extensively studied cytokine in experimental stroke is the proinflammatory and immune regulatory cytokine TNF. It exists in a secreted form (solTNF) and a transmembrane form (tmTNF), which is also involved in reverse signaling [87]. solTNF is derived from tmTNF, which is cleaved by the protease ADAM-17, also known as TNF-alpha converting enzyme (TACE) [14]. tmTNF and solTNF signals are transmitted through two distinct receptors, TNFR1 and TNFR2, that differ significantly both in cellular expression and downstream effects. Although solTNF binds both receptors with high affinity, it preferentially binds to TNFR1 (dissociation constant [K_d] 20 pM) versus TNFR2 ([K_d] ~ 400 pM), owing to a 30-fold faster dissociation rate from TNFR2 than from TNFR1 [69]. This has given rise to a ligand-passing hypothesis, stating that solTNF binding to TNFR2 is quickly passed to TNFR1. Binding by TNFRs to tmTNF or even TNF antagonists can induce reverse signaling through tmTNF, leading to cell activation, cytokine suppression, or apoptosis of the tmTNF-bearing cell (reviewed in [49]). While TNFR1 is expressed on virtually all cells, TNFR2 expression is restricted to cells of the immune system, glial cells, and endothelial cells. TNF’s proinflammatory effects are likely mediated through solTNF–TNFR1 signaling, leading to activation of two major, well-understood pathways. One leads to the induction of anti-apoptotic genes, mainly through activation of the transcription factor nuclear factor-kappa B (NF-κB). The second signaling pathway results in activation of cellular suicide programs, including the prototype of programmed cell death, apoptosis, but also the execution of programmed necrosis (necroptosis) [179]. Under physiological conditions, TNF does not induce cell death unless transcription, translation, or specifically the NF-κB pathway are blocked. Unlike TNFR1, TNFR2 is not associated with induction of apoptosis but preferentially promotes cell growth, and regeneration through NF-κB activation. TNFR1 can be activated by binding of either solTNF or tmTNF, whereas TNFR2 is only fully activated by tmTNF [68, 69]. A further level of complexity is added by the proteolytic cleavage of the extracellular domains of both TNFR1 and TNFR2 [182], which is increased upon TNFR activation (reviewed by [1]). The soluble TNFR1 and solTNFR2 ectodomains that are shedded can bind to TNF, albeit with low affinity, and can thus act as natural inhibitors of TNF.

Lymphotoxin-alpha (LTα), another TNFR agonist with important roles in immune regulation, also binds TNFR1 and TNFR2 and mainly mediates NF-κB-mediated signaling [134].

The low baseline levels of TNF in the CNS under physiological conditions play an important role in neuronal function, by modulating glutamatergic synaptic transmission and plasticity [164]. Furthermore, TNF regulates neuronal networks involved in cognition and behavior [9], thereby attributing importance to TNF both in the healthy and stroke-lesioned CNS. Multiple checks are in place to finetune TNF’s production and activity, including regulation of TNF gene expression at transcriptional and translational levels, and the regulated shedding of TNF [117] and its receptors [135].

A particular role of TNF/TNFR1 in the etiopathogenesis of stroke is suggested by genome-wide association studies that found a polymorphism in the TNF gene that increases the susceptibility for stroke [178]. After pMCAO, TNF is acutely and significantly upregulated, peaks at 12–24 h (Fig. 2a), and remains elevated for days (Fig. 1d), making TNF a key player both in acute and chronic ischemia and in post-ischemic neuronal plasticity (reviewed by [91]). TNF is primarily produced by microglia in the early phase after experimental stroke and sustained by macrophages at later time points [20, 32, 92, 94], although other cell types like ependymal, astroglial and neuronal cells have also been reported to produce TNF during ischemic conditions (reviewed by [91]).

The use of genetically modified mice has been invaluable for establishing the role of TNF in the pathogenesis of ischemic stroke. Conventional TNF-knock out (KO) mice [92] and conditional TNF-KO mice with ablation of TNF in myeloid cells, including microglia [31] have larger infarcts and worse behavioral deficits than control mice after pMCAO. This suggests a neuroprotective role of microglial-derived TNF in ischemic stroke, an effect which appears to be mediated via TNFR1 [92, 170]. Interestingly, mice with a loss of TACE-mediated cleavage preventing shedding of solTNF (and thus expressing only tmTNF) develop smaller infarcts than their littermates [104], suggesting that removal of solTNF but preservation of tmTNF is neuroprotective in ischemic stroke.

Finally, a polymorphism in the LTα gene (LTA) has been linked to increased susceptibility for stroke [178], suggesting that also LTα plays a role in the etiopathogenesis of stroke. However, LTα levels appear to remain relatively constant in the acute phase after pMCAO in mice (Fig. 2a, Lambertsen et al., unpublished data), suggesting that brain-derived LTα has no major role in the inflammatory response post-stroke.

The currently used FDA- and EMA-approved anti-TNF therapeutics block both solTNF and tmTNF (Table 2). These therapeutics appear to relieve fatigue and symptoms of depression that can be associated with chronic inflammatory diseases [177]. Despite reports of improved neurological outcome in patients with stroke or traumatic brain injury who are treated with perispinal etanercept [172, 174] (Table 1), none of the currently used anti-TNF therapeutics have so far been approved as a neuroprotective strategy in combination with tissue plasminogen activator treatment. This may be because targeting both solTNF and tmTNF can predispose patients to both cardiovascular and demyelinating diseases [151], which is in line with the finding that a single nucleotide polymorphism in the TNFR1 gene (TNFRSF1A) that mimics the effect of anti-TNF therapeutics, is a risk factor for developing multiple sclerosis [67]. In combination with the observation that not only TNF-KO mice but also TNF-R1 KO mice develop larger infarcts than wild-type mice [92, 170], this calls for precaution in using the currently approved anti-TNF therapeutics and emphasizes the need for more specific anti-TNF therapeutics.

There has been little preclinical testing of therapeutics that exclusively target solTNF (XPro1595, XEN345, and possibly R1antTNF) (Tables 1, 2 and Fig. 3a) and leave signaling via tmTNF–TNFR1/2 intact. A comparative study of a single i.v. dose of XPro1595 (a dominant-negative solTNF inhibitor) or etanercept, administered at a dose of 10 mg/kg, 30 min after pMCAO, showed that both compounds affected the inflammatory response and improved motor functions and motor learning skills compared to vehicle 1 and 5 days after pMCAO, but had no effect on infarct volume [30]. This indicates that targeting solTNF alone may be efficient for the treatment of post-stroke inflammation. Similarly, recent findings showed that topical, but not systemic administration, of XPro1595 can rescue tissue at risk after experimental spinal cord injury, while etanercept had no effect [129], suggesting that topical administration of XPro1595 can inhibit solTNF present locally in the CNS. Clearly, more studies are needed to clarify whether XPro1595 is able to rescue tissue at risk in the peri-infarct. However, given the prevalence of post-stroke infections in humans, leaving tmTNF signaling intact may decrease the risk of infections.

While it seems relevant to retain the neuroprotective TNFR1 signaling in the acute phase after stroke, TNFR1 also plays a role in sustaining chronic inflammation in mouse models of multiple sclerosis and TNFR2 is important for remyelination [18]. Although more studies are clearly required to clarify the role of neuronal TNFR1 signaling in the acute phase post-stroke, it is possible that TNFR1-specific antagonists [R1antTNF, DMS5540, TROS (TNF receptor one silencer), ATROSAB (antagonistic TNF receptor one-specific antibody)] (Table 2) that preserve TNFR2 signaling, will be important in improving neuronal and synaptic remodeling in the chronic phase of stroke.

Due to their large size, many biologic TNF inhibitors do not cross the BBB and must be modified to enable BBB penetration and access to the brain parenchyma. One such drug is cTfRMAb-TNFR (Table 2), which ferries TNFR across the BBB using the transferrin receptor (TfR) [197]. In a preclinical study, i.v. injection of cTfRMAb-TNFR was compared to etanercept in a tMCAO model and when administered 90 min after occlusion resulted in reduced infarct volumes and reduced neural deficit 1 and 7 days post-stroke, whereas etanercept had no effect [167](Table 1). Despite the fact that both cTfRMAb-TNFR and etanercept are TNFR2 fusion proteins, the authors ascribed the beneficial effect of cTfRMAb-TNFR to the modification of this protein to allow it to be transported across the BBB [15].

In another preclinical study, sTNF-α R1 (solTNFR1) (Table 2) administered by intracortical infusion for 1 week after photothrombotic stroke was found to preserve deprivation-induced axonal plasticity in the cerebral cortex post-stroke [98] (Table 1). This effect was ascribed to sTNF-α R1 competing for solTNF with TNFR1 receptors, supporting the hypothesis that ablating solTNF is beneficial in ischemic stroke. This is in line with a preclinical study showing that intra-arterial injection of solTNFR1-overexpressing dendritic cells 6 h after tMCAO reduces infarct size and inflammation 3 days post-stroke [186] (Table 1).

The IL-1 family comprises 11 members (IL-1α, IL-1β, IL-1 receptor antagonist (IL-1Ra), IL-18, IL-33, IL-36α, IL-36β, IL-36γ, IL36-Ra, IL-37, and IL-38), forming a network of proinflammatory cytokines that regulate innate immune cells and function as key players in inflammation (review by [43]). Despite structural and functional similarities and evidence of a common ancestry [143], so far only IL-1α, IL-1β, and IL-1Ra have been studied extensively in ischemic stroke.

Both IL-1α and IL-1β are expressed and translated as precursor (pro) proteins. ProIL-1α is biologically active, but it lacks the signal peptide that allows it to leave the cell [143]. IL-1α is a ‘dual-function’ cytokine with both nuclear and cytoplasmic functions, but danger signals from necrotic cells can promote the secretion of IL-1α [48], causing neutrophil recruitment and exacerbation of inflammation [24]. Apoptosis causes IL-1α to translocate to the nucleus, where it binds to chromatin, a mechanism which is known to restrain inflammation [34]. IL-1α is considered to be an early danger signal that modulates a wide range of inflammatory reactions through the interleukin-1 receptor type 1 (IL-1R1) [48, 143]. Following injury, the proteolytic cleavage of IL-1α occurs through the actions of calpain, and possibly inflammasomes [194]. Membrane-bound, unprocessed IL-1α acts in a paracrine fashion on IL-1R expressing cells [42] to modulate angiogenesis, cell proliferation, senescence, apoptosis, and migration, and cytokine production ([149] and review by [43]).

In contrast to proIL-1α, proIL-1β is a biologically inactive protein, and both proIL-1β and mature IL-1β appear extracellularly [143], indicating that processing can take place after secretion. ProIL-1β is cleaved by caspase-1 (or IL-1 converting enzyme) [143], which gets activated by the assembly of the inflammasome, a process triggered in turn by damage-associated molecular pattern signals [72]. ProIL-1β can also be cleaved by neutrophil serine proteases such as proteinase 3 and elastase [123].

The natural regulator of IL-1 is IL-1Ra, which is found in two structural variants, secreted (s)IL-1Ra and intracellular IL-1Ra (icIL-1Ra), that both target the IL-1R1 [6]. The icIL-1Ra isoform is less explored but believed to exert multiple functions inside the cell [6], such as modulating the effect of IL-1α and/or acting as regulator of proIL-1β [102]. IL-1Ra is expressed by monocytes/macrophages, neutrophils [105], microglia [33], and other cells [42].

IL-1α/β induce their biological effects through IL-1R1, which is expressed in low numbers (< 100) on nearly all cells in the brain [42]. Binding of IL-1 to IL-1R1 allows the binding of the interleukin-1 receptor accessory protein (IL-1RAcP, IL-1R3), which is a key component of the receptor/agonist signaling complex [6, 143]. Recruitment and binding of IL-1RAcP converts the low-affinity binding between IL-1R1 and IL-1 to a high-affinity binding allowing further signal transduction [65]. IL-1 signaling is complex but potent with < 10 receptors/cell required to be occupied before a full response is triggered [166]. This means that IL-1Ra needs to be present in 100–1,000-fold molar excess to control its biological properties [42].

IL-1R2 shares structural characteristics with IL-1R1, but it lacks the cytoplasmic domain that allows signal transduction. IL-1R2 binds IL-1 as a decoy receptor [42, 143]. IL-1R2 is expressed by the same cells as IL-1R1 but is particularly abundant on monocytes, and neutrophils [42, 45]. IL-1R2 binds IL-1α in the cytosol, preventing its interaction with IL-1R1 when released from necrotic cells [196]. All the IL-1Rs are also found in a soluble form [90].

IL-1 is a major player in stroke pathology (Fig. 1g, h). As for the TNF gene, a polymorphism in the IL-1A gene has been associated with an increased susceptibility for ischemic stroke [199] whereas a polymorphism in the IL-1B gene has been associated with lower stroke risk [13], although this is still controversial [193, 199]. Polymorphisms in the IL1RN gene do not affect the risk for stroke [199], but increased plasma IL-1α combined with a polymorphism in the IL1RN gene increases the risk of post-stroke infection [10].

So far, focus has been on understanding the role of IL-1β in experimental stroke models, however data suggests that also IL-1α, which is significantly upregulated in mice 6–24 h after pMCAO (Fig. 2b) [33] and 7 days after tMCAO [149], plays an important role in stroke-induced neuroinflammation [33, 171]. Following experimental stroke in rodents, IL-1α was shown to be expressed by platelets and microglia [33, 40]. The presence of platelet-derived IL-1α acutely (6 h) after experimental stroke [33] supports findings that IL-1α drives neurovascular inflammation and facilitates neutrophil infiltration into the ischemic brain [171]. At 24 h after pMCAO, microglia are the key producers of IL-1, with approximately 50% of the IL-1α producing microglia co-expressing IL-1Ra and 17% co-expressing IL-1β, demonstrating that IL-1β and IL-1α are largely produced by segregated populations of microglia in the ischemic brain [33]. It is, therefore, likely that IL-1α in platelets in addition to few IL-1α/β producing microglia impacts the balance between IL-1/IL-1Ra early after stroke onset [33]. Findings that IL-1α and IL-1Ra are co-expressed in microglia support the view that icIL-1Ra can regulate the action of intracellular IL-1α [113].

IL-1β is constitutively expressed in the CNS [42] where it exerts neurotrophic factor-like activity [161] or regulates both the expression and activity of ion channels [181]. IL-1β is upregulated acutely after ischemic stroke (Fig. 1)[32, 33, 37] and peaks at 12-24 h (Fig. 2b) primarily in microglia and macrophages [32, 37], and later in astroglial-like cells [183].

IL-1 has been shown to aggravate stroke pathology (Table 1) as demonstrated by findings in transgenic mice overexpressing a dominant-negative form of caspase-1 in neurons [54], caspase-1 KO mice [73], and IL-1α/β KO mice [17], which all show reduced infarct volumes after experimental stroke. Additional support comes from early studies demonstrating that administration of recombinant IL-1β exacerbated damage [99] as does intracerebroventricular (i.c.v.) delivery of an IL-1Ra antiserum [101]. Systemic administration of IL-1β just before tMCAO worsened outcome in rodents through neutrophil- and platelet-dependent mechanisms reducing reperfusion [109].

In addition, IL-1Ra is an acute phase protein [55] that blocks the action of IL-1. Administration of IL-1Ra reduced ischemic brain damage after both tMCAO and pMCAO in rats [59, 137] and mice [175] (Table 1) and IL-1Ra-overexpressing mice show reduced infarct volumes, whereas IL-1Ra KO mice display increased infarct volumes compared to littermate mice after pMCAO [33].

IL-1Ra is the only therapeutic agent directed against IL-1-induced inflammation (Fig. 3b) that has been tested in randomized clinical trials in ischemic stroke (Table 1). In pre-clinical stroke models, recombinant (r)IL-1Ra is protective after central [137] and peripheral [59] administration and, similar to i.c.v. injection of anti-IL-1β antibody (Ab) [191] or IL-1Ra, was shown to reduce infarct volumes after MCAO in rats [99, 137] and pMCAO in mice [121].

Although IL-1Ra can reach the brain after systemic administration in the rat [66] and modulates long-term functional recovery after experimental stroke [62], its use in stroke patients has proven challenging. Pharmacokinetic studies have shown that rIL-1Ra crosses the BBB slowly [71] and has a very short half-life in the circulation [64], and thus it is difficult to achieve therapeutic IL-1Ra concentrations in the brain [57].

The first randomized, double-blind, placebo-controlled trial using i.v. injected recombinant human (rh)IL-1Ra in acute stroke patients (given within the first 6 h of stroke onset) showed a reduction in neutrophil count, plasma CRP, and IL-6 compared to placebo, and exploratory efficacy analysis indicated that patients receiving rhIL-1Ra had minimal to no disability three months after stroke [51]. Recently, the SCIL-STROKE (subcutaneous interleukin-1 receptor antagonist in ischemic stroke) phase II trial, using subcutaneous (s.c.) injections of IL-1Ra in combination with i.v. thrombolysis, showed reduced plasma IL-6 levels, whereas neurological recovery three months after stroke was unaffected [159]. Exploratory efficacy analysis suggested that the expected beneficial effect of IL-1Ra on clinical outcome by reducing inflammation might have been counteracted by a negative effect, which could represent an interaction with alteplase [159].

Another potent proinflammatory cytokine with pleiotropic functions is IL-6, which is expressed on many cell types, including monocytes, neurons and glial cells (Fig. 1j, k)[52, 70]. The pleiotropism of IL-6 may be explained by IL-6 eliciting fundamentally different cellular responses depending on whether the classic or the trans-signaling pathway is activated [152]. This depends on the IL-6 receptor system that consists of the IL-6 receptor (IL-6R) as well as soluble IL-6R (sIL-6R) and glycoprotein 130 (gp130), which due to its cytoplasmic domain is responsible for the signal transduction. Soluble IL-6R is formed by cleavage from the IL-6R by TACE/ADAM17 [141] or by translation of different IL-6R mRNA splice variants [103].

In classic signaling, IL-6 binds to and forms a complex with membrane-bound IL-6R, which then recruits gp130. Trans-signaling occurs when IL-6 binds sIL-6R, which then binds to membrane-anchored gp130 [141]. Unlike IL-6R, which is expressed by neurons, microglia, neutrophils, monocytes, hepatocytes and CD4⁺ T cells and thus limits classic signaling to only a few tissues [58], gp130 is ubiquitously expressed in the body (reviewed by [145]), increasing the spectrum of IL-6 target cells. Trans-signaling is normally tightly regulated [185] and can be counteracted by a soluble form of gp130 (sgp130), which is generated by alternative splicing of gp130 mRNA and is present in serum [85]. Once IL-6 is released into the blood it can bind sIL-6R but also sgp130 [150], which immediately interferes with IL-6 trans-signaling [58]. As sgp130 levels are much higher than sIL-6R, trans-signaling does not occur under physiological conditions.

Classic IL-6 signaling is believed to be anti-inflammatory and protective [185], while trans-signaling is responsible for the pro-inflammatory effects mediated by IL-6 [147, 152].

IL-6 is expressed in the normal CNS, where it influences neuronal homeostasis by acting as a neurotrophic factor via the classical signaling pathway (reviewed by [147]). Ischemic stroke in mice and rats leads to a significant increase in the levels of IL-6 from 6 to 12 h (Fig. 1 and 2c), and in both IL-6R and gp130 from 3 days [3, 70]. IL-6 has been shown to be neuroprotective in experimental stroke [192] although this is still debated [29]. In human stroke, IL-6 serum levels increase within the first 24 h and have been shown to correlate significantly with infarct size and survival [11, 157]. A similar correlation has not been observed for sIL-6R [46, 70]. While studies of IL-6 expression in the ischemic brain post-mortem are sparse, one study showed that IL-6 levels were elevated in the infarct already in the acute phase after stroke and remained elevated at later time points [126]. Supporting the neuroprotective effect of brain-derived IL-6 are findings showing a positive effect of IL-6 on post-stroke neurogenesis, leading to long-term functional recovery [111].

Similar to patients treated with nonspecific TNF antagonists, non-neurological patients treated with IL-6 inhibitors are at increased risk of infections (reviewed in [169]). Clinical stroke studies show that sIL-6R correlates with the degree of leukocyte infiltration [85] and that sIL-6R neutralizing antibodies are beneficial [146]. In comparison, anti-IL-6R antibodies target both the membrane-bound form of IL-6R and sIL-6R, and therefore, affect classical and trans-signaling equally (Fig. 3c and Table 2).

If classical IL-6 signaling is protective and trans-signaling detrimental, selective neutralization of the potential, detrimental trans-signaling is possible by administration of the chimeric protein sgp130Fc (Fig. 3c and Table 2). Sgp130Fc is a fusion protein that contains the extracellular domain of human gp130 and the Fc-fragment of human IgG1. This allows sgp130Fc to bind to the IL-6/solIL-6R complex, but not to sIL-6R alone [86], whereby spg130Fc blocks trans-signaling [52] (Fig. 2c). Such specific inhibition of the trans-signaling pathway using, i.e. sgp130, which does not compromise classic signaling, could be a promising therapeutic tool in future stroke research.

IL-10 is a pleiotropic anti-inflammatory cytokine mainly produced by type-2 helper T cells, which in turn regulate inflammatory reactions. IL-10 binds to IL-10 receptors (IL-10R) to reduce inflammation and limiting apoptosis [148]. In the CNS, astrocytes, neurons, and microglia have been reported to produce IL-10 [114, 188].

A meta-analysis investigating the association of IL10 gene polymorphism with the risk of ischemic stroke showed no overall significant association between IL-10 and the risk of ischemic stroke, but an association was found with large vessel disease and small vessel disease [89], suggesting that some subtypes of ischemic stroke are associated with IL10 gene polymorphisms.

In experimental stroke, IL-10 mRNA and protein and IL-10R mRNA levels are increased, with IL-10 noted in microglia and IL-10R on astrocytes in the peri-infarct area [126, 132]. In transgenic mice overexpressing IL-10, infarct volumes were reduced, and apoptosis decreased 4 days after pMCAO [38]. Furthermore, low IL-10 levels were associated with poor stroke outcome and a delayed, exacerbated inflammatory response after pMCAO that was ameliorated by IL-10 administration after pMCAO [132] (Table 1). Therapeutic administration of IL-10 has been shown to be neuroprotective in experimental stroke and to limit post-stroke inflammation [96, 97, 130, 139, 160, 165] (Table 1),

Low plasma IL-10 levels in patients with subcortical or lacunar stroke are associated with neurological worsening within the first 48 h [180], attributing IL-10 a role in the acute neuroinflammatory response after stroke. This is in line with findings by Protti et al. showing that patients with low IL-10 levels deteriorated neurologically within the first 3 days post-stroke [136]. Stroke patients are prone to infection due to stroke-induced immunodepression, however, and increased serum IL-10 levels have been identified as an independent predictor of post-stroke infection [22, 187]. Women have poorer recovery after ischemic stroke than men, even after controlling for age and stroke severity [19, 80]. This may be partly due to the increased IL-10 levels 24 h post-stroke and an associated higher incidence of post-stroke urinary tract infection and poorer overall outcomes in women have been suggested to be a contributing factor [35]. Overall, these studies indicate that an excessive IL-10 response can lead to post-stroke immunosuppression and worsen neurological outcome, suggesting that IL-10 therapeutics should be given with caution. Future studies should be aimed at differentiating between central and peripheral IL-10 effects post-stroke.