Regulatory T Cells as Master Switches in Immune Homeostasis

The immune system has evolved several regulatory mechanisms to prevent autoreactivity to self-antigens and control excessive immune reaction after tissue injury. The presence of regulatory T cells (Treg) suppressing actively (self-reactive) immunity is one of the key mechanisms of preserving immune homeostasis and limiting inflammatory collateral damage [1]. Depletion of CD25 + CD4+ Treg naturally arising in the immune system induces autoimmune diseases, and reconstitution of this cell population prevents disease development [2]. A lack of Treg has been shown to be a primary cause of autoimmune diseases in humans [3]. In addition to maintaining autotolerance, Treg are also involved in control of immune homeostasis in a broad range of immunological conditions, including those against autologous tumor cells [4], allergens [5], and organ transplantation [6]. Although Treg might be simplistically described as immunosuppressive T cells, several phenotypically and functionally distinct Treg subpopulations have been defined [7]. Emerging evidence arises for substantial phenotypic differences between circulating and tissue-infiltrated Treg cells as well as for functional specialization of tissue-localized Treg cells in mice [8], for example, the Treg cells in muscle that promote muscle regeneration [9]. Therefore, it is essential to develop a better understanding of the biology of the site-specific function of Treg cells for control of tissue and immune homeostasis [10]. Despite the large variability of Treg function and phenotype, the best investigated population of Treg are CD4 + CD25 + Foxp3+ naturally occurring Treg [11]. This is neither a homogenous nor static cell population but consists of several specific subsets determined by their origin (thymic or peripherally induced), is resident in most non-lymphoid tissue as site-specific Treg subsets, and is determined by their recent antigen-encounter or activation status [12]. However, virtually all suppressive Treg subsets express the common transcription factor Foxp3 which is directly linked to their immunosuppressive function. Foxp3, an X-chromosome-linked factor that controls Treg cell development and function, is generally thought to positively control Treg cells, as Foxp3 expression is sufficient to determine the immunosuppressive function of conventional T cells [13]. However, at the same time, Foxp3 expression levels—in addition to other factors and effector functions of Treg—determine the “quality” of Foxp3+ cells with regard to the immunosuppressive properties [14]. However, virtually all previous analyses leading to the concept of phenotypic and qualitative differences in the Treg population have been performed by investigating secondary immunological organs such as lymph nodes or spleens. Therefore, this important information about the detailed phenotypic differences of Treg subpopulations—i.e., thymic or peripherally induced Treg, tissue-specific Treg subsets, and subsets characterized by expression of transcription factors and epitope markers [12]—and their specific functional role is missing in the context of acute brain injury.

Regulatory T Cells in Experimental Stroke

Cerebral Treg Invasion

While it is still uncertain at which specific site Treg act as immunomodulators after stroke, several interfaces of Treg-brain interaction can be assumed, such as the endothelium, impact on microglial activation, or direct cell-cell interaction with neurons—which will be discussed later in more detail. In contrast, the kinetics and magnitude of Treg infiltration into the ischemic brain has been characterized in detail in several studies using models of experimental middle cerebral artery occlusion (MCAO), with the limitation that none of the previous studies has systematically characterized Treg subpopulations but focused mainly on natural CD25 + Foxp3+ Treg. One of the first studies systematically analyzing brain leukocyte invasion by flow cytometry of brain homogenates in a model of transient mechanical vascular occlusion (TMVO) using intravascular filaments was performed by Gelderblom et al. [15]. In this study, only a very low number of CD25 + Foxp3+ Treg were observed within the first week after transient ischemia. In contrast, using the distal, permanent MCAO occlusion model [16], we detected substantial T cell and Treg counts in the ischemic hemisphere, with Foxp3+ Treg constituting approx. 20 % of all CD4+ T-helper cells [17]. These findings are consistent with several reports demonstrating that distal permanent occlusion induces a significantly stronger neuroinflammatory reaction with manifold higher T cell counts in the ischemic hemisphere than in proximal transient occlusion models [18, 19]. Stubbe et al. detected after large hemispheric lesions only negligible amounts of Foxp3+ Treg early after stroke, but substantial Treg invasion occurred 14 and 30 days after MCAO, indicating probably delayed kinetics for the adaptive immune response in this TMVO model [20].

Treg Effects on Peripheral Immunity After Brain Injury

Despite their delayed recruitment into the ischemic brain, Treg have been found to influence stroke outcomes already within the first days after ischemia, arguing that effector sites outside the brain parenchyma might be critically involved. In accordance with this assumption, delayed deletion of Treg by pharmacological or genetic depletion at 3 days after MCAO did not affect stroke outcome [20, 21]. A potential extra-cerebral target of Treg in post-stroke immunity might be leukocytes in the peripheral immune system. A consistent finding in the subacute phase after extensive experimental stroke is that cellular immunosuppression and splenic atrophy are accompanied by a relative expansion of Treg in the spleen and blood [22, 23]. Interestingly, a recent report suggested that the adoptive transfer of Treg reduced the systemic inflammatory reaction in the early phase after stroke. On the other hand, Treg transfer also ameliorated the extent of immunosuppression (i.e., lymphopenia) in the subacute phase after brain injury [24]. This finding of a dualistic role of Treg in the peripheral immune system after stroke—abrogating early immune (over-)activation and subacute immunosuppression—is in accordance with the concept that the initial immune over-activation might result in an exhausted immune phenotype and even induce subsequent leukocyte cell death by currently still unclear mechanisms [25]. The potent homeostatic function of Treg might help to suppress the initial systemic inflammation, thereby attenuating the subsequent immune disturbances, such as activation-induced lymphopenia and exhaustion of antigen-presenting cells.

An additional—probably non-immunological—function has been proposed for Treg within the brain vasculature to contribute to coagulation and thrombosis of the microvasculature. It has been demonstrated that Treg might play an important role in the development of secondary microthrombosis TMVO models, which was termed thromboinflammation [26]. This concept of thromboinflammatory infarct progression proposes an interaction of Treg with platelets and the activated endothelium as an independent mechanism of Treg function [27, 28]. This is of prime clinical importance, taking into account the rapidly increasing number of mechanical recanalizations—modeled by the TMVO animal models—and the large number of insufficient reperfusions despite successful recanalization in stroke patients [28, 29].

Cerebral Targets of Treg-Dependent Mechanisms After Stroke

The most prominent immune mechanisms of Treg function in vivo are the secretion of anti-inflammatory cytokines (IL-10, tumor growth factor β (TGF-β), expression of transmembrane immunosuppressant molecules (CTLA-4, CD39, PD), consumption of vital cytokines (IL-2), and the secretion of cytolytic molecules such as granzymes and perforin [30]. Several reports have demonstrated that IL-10 is a critical cytokine-modulating post-stroke neuroinflammation in experimental brain ischemia [31–34]. The main sources of cerebral IL-10 are regulatory lymphocytes as well as microglia/macrophages. Therapeutic approaches to increase lymphocyte-derived IL-10 production [21, 33, 35, 36] or by therapeutic IL-10 administration [31, 34, 37] have resulted in improved stroke outcome. In addition, alternative Treg-related mechanisms have been found to act in ischemic brains, such as the expression of TGF-β and IL-35 [38] as well as a role for Treg in the co-inhibitory PD-1/PD-L1 pathway. It has been shown that the immune regulatory function of PD-1 limits ischemic brain lesions and blood–brain barrier damage [39]. In contrast, PD-L1 appears to have a detrimental effect in stroke [40]. This at first sight contradictory function of receptor and ligand has been attributed to two potential mechanisms: First, PD-Ls might “inhibit-the inhibitors” as PD-1 is also expressed on immunosuppressive cells; secondly, alternative receptors for PD-Ls such as CD80 might have under conditions of acute inflammation after brain ischemia an activatory net effect on T cell activation. This can be explained by the bidirectional PD-L—CD80 interaction and interference with other suppressive signals such as CD80/CD28 on effector T cells [40, 41]. The beneficial effect of Treg at later stages of ischemic stroke was frequently related to the reduced invasion of pro-inflammatory T cells [17, 42], suppression of effector T cell proliferation [21], and reduced cytokine production [38, 42, 43]. Interestingly, others and we have also detected reduced microglia/monocyte activation in the presence of Treg [17], or priming toward an M2-like microglial phenotype [38] and reduced microglial TNF-α production under the influence of Treg [17]. These findings suggested that Treg cells have an impact on the resident immune cells, most likely via soluble mediators such as IL-10 because the effect on microglia could be detected at early time points after stroke even before invasion of substantial numbers of Treg into the injured brain. In contrast, studies reporting an exacerbation of stroke by Treg-targeted therapies during the very early stage [26, 44] found an association of increased lesion volumes with amplified cerebral immune cell accumulation [44]. These effects were most likely independent of the immunological function of Treg but rather attributable to the extracerebral function of Treg in secondary microthrombosis as discussed above [26].

Regulatory T Cells in Stroke Patients

Only few studies analyzed Treg cells in human stroke patients, and these studies were understandably limited to the analysis of blood samples from patients with different stroke entities. Hug et al. have found that Treg function is preserved in the subacute phase after stroke in contrast to the dysfunction of effector cell populations such as circulating monocytes or helper T cells [45] in the context of post-stroke immunosuppression. An opposing study detected an impaired Treg suppressive function in female but not male stroke patients, proposing gender-specific effects in post-stroke immunomodulation [46]. These discrepant findings can be attributed to differing patient characteristics regarding comorbidities and stroke severity. While the latter study has detected a robust increase of Treg cell counts after stroke in accordance with rodent experiments, others have shown the opposite. Li et al. detected a significant reduction of circulating Treg in stroke patients [47]. Overall, clinical data is supporting the experimental finding of substantial peripheral immunomodulation after stroke including the Treg population. However, specific changes might depend on stroke entity, severity, and patient characteristics, and further studies to better characterize the involvement of Treg cells in human stroke are urgently needed.

Treg Depletion in Experimental Brain Ischemia

Depletion of the Treg population is a common experimental paradigm to investigate the functional role of Treg in experimental stroke. Antibody-mediated cell depletion using anti-CD25 antibodies and genetic depletion using transgenic mice with a diphtheria toxin receptor (DTR) transgene under the control of the Foxp3 promoter have been used for efficient Treg depletion. Using Treg-depletion paradigms, about half of the experiments performed revealed an increase in infarct volume16, 19, 28, while the other half did not detect any effect on stroke outcome [21, 26, 48] and one study even observed a reduction of infarct size in Treg-deficient mice [21, 26, 48]. This irreproducibility of findings in-between several studies from independent laboratories led to an intense debate on the biological role of Treg in stroke [49–51]. Notably, inconsistencies cannot be attributed to the specific depletion approach, since transgenic mouse models have been used on both sides of the “efficacy spectrum.” Moreover, each of the three studies using genetic depletion of Treg used a different inducible Foxp3-KO mouse line [21, 26, 48]. Each of these transgenic Foxp3-deletion models are derived from different transgenic constructs and differ in their depletion efficacy. Moreover, previous reports have demonstrated a substantial dysregulation of peripheral immune homeostasis upon Treg depletion [52]. Treg depletion induces rapid expansion of lymphocytes and dysbalance in peripheral immune homeostasis in the first week after depletion. Hence, small methodological differences in the depletion paradigm itself or by priming of lymphocytes during this vulnerable phase by environmental factors, such as housing conditions, the specific microbiota conditions, or experimental stressors, might have an unforeseen impact on immunological outcome [53, 54]. The immunological effects of Treg depletion on the neuroinflammatory response after brain ischemia have been investigated only in a fraction of the published reports. All three studies detecting an exacerbation of stroke outcome after Treg depletion also observed an associated increase in neuroinflammatory biomarkers [17, 21, 38]. The most robust findings included an increase in cerebral leukocyte invasion and an increase in pro-inflammatory cytokine secretion. Interestingly, a previous study investigating the role of Treg in secondary microthrombosis and stroke infarct progression has detected a detrimental role of Treg in mediating post-stroke neurodegeneration which was independent of their immunological function and alterations in post-stroke neuroinflammation [26]. Accordingly, “wannabe Treg” without immunomodulatory function were still able to exert their impact on stroke progression without affecting neuroinflammatory markers, supporting the hypothesis of non-immunological functions of Treg in stroke models with potential endothelial injury, vascular inflammation, and microthrombosis.

Therapeutic Enhancement of Treg Function in Brain Ischemia

Despite the discrepancies on Treg function in Treg-depletion paradigms, several reports have tested Treg expansion approaches for stroke therapy. These studies have investigated very different methods to increase Treg numbers and/or function. The most intuitive approach might be the adoptive cell transfer of purified Treg to recipient animals to increase circulating Treg counts. The second most widely used approach was the administration of a CD28 superagonist (CD28SA), which induces in vivo expansion of Treg and amplification of their suppressive function. A pathomechanistically distinctive approach used in previous studies was the paradigm of “mucosal immunization” by administration of cerebrovascular antigens, such as myelin oligodendrocyte glycoprotein, or selectins, which should result in the expansion of autoantigen-specific Treg [43, 55]. Out of 16 experiments reported until December 2015 in 14 independent studies, 13 found an improvement in stroke outcome [21, 24, 35, 38, 39, 42, 43, 55–57], while two described an increase in infarct volume when using two independent Treg-targeted therapies [26, 44] and one study did not detect an effect at all [58] (Table 1). This discrepancy cannot be easily explained by the therapeutic paradigm used, since the methods of adoptive Treg transfer into wild-type mice and the CD28SA treatment utilized in the two studies showing an exacerbation of lesion volume were also tested in three or more other experiments that showed an improved outcome (Table 1). It appears that stroke severity or other factors directly eminent to the specific stroke model might predict the net biologic effect of Treg: The majority of the 13 experiments detecting an overall benefit of Treg boosting was performed in stroke models with small- to moderate-sized lesion volumes. Surprisingly, however, the two studies [35, 44] in which the identical CD28SA has been used in very similar stroke models reported opposing results. Na et al. described improved stroke outcome following treatment with the CD28SA [35], while Schuhmann et al. observed a significant deterioration of stroke outcome [44].

Table 1 Studies investigating Treg-targeted therapeutic approaches (by treatment and outcome)
Full size table

We have performed a meta-analysis to better estimate the efficacy of Treg-targeted therapeutic approaches in all studies published until December 2015 using therapeutic paradigms of Treg targeting in experimental stroke models. All studies with the aim to therapeutically modulate Treg (defined at least by CD25 or Foxp3 expression in CD4+ T-helper cells) numbers and/or Treg function in models of experimental brain ischemia were included in the analysis. The studies were clustered by the respective treatment paradigm: mucosal immunization [43, 57], adoptive Treg transfer [24, 26, 39, 42, 56], including intracerebroventricular Treg injection [58], CD28SA therapy [35, 44, 56], and others (i.e., HDAC inhibition [21] and mTOR inhibition [38]). Odds ratios for effect size estimation were calculated and the forest plot illustrated using RevMan software (version 5.3). The overall effect size estimation revealed an odds ratio favoring each subgroup of interventions and for all studies in total (Fig. 1a). However, these results have to be interpreted with caution due to several critical caveats: Funnel plot analysis displayed substantial asymmetry, suggesting publication bias and lack of reporting neutral or negative results (Fig. 1b). Moreover, a large heterogeneity between the included studies has to be taken into account regarding study design, experimental model, and species (rat and mice). In addition, most studies included in this meta-analysis did not adhere to the guidelines on data reporting according to the Stroke Therapy Academic Industry Roundtable (STAIR) recommendations. Therefore, further studies investigating Treg function in acute brain injury are urgently needed to increase the robustness of effect size estimation. Also, the publication of negative results should be encouraged to avoid a publication bias in this emerging research field. We have recently reported the feasibility and value of performing multicenter preclinical studies (pRCTs) adopting key elements of clinical randomized controlled trials (RCT) in such situations where reports from individual laboratories are discrepant, irreproducible, or simply not comparable due to their vast heterogeneity [19, 59]. Performing such a multicenter pRCT using a unified and highly standardized study protocol across laboratories might resolve the uncertainty about the role of Treg-targeted therapies in experimental stroke and guide further translational research in this field.

Fig. 1
figure 1

Meta-analysis for Treg-targeted therapies in experimental brain ischemia. a The meta-analysis, depicted as a forest plot, includes all published studies until 30th of November 2015 investigating Treg (at least characterized by CD25− and/or Foxp3-expressing CD4+ cells) in rodent stroke models. Odds ratios were calculated using inverse variance in a fixed effects model. CI confidence interval. b Publication bias was estimated by illustration of the odds ratio (OR, x-axis) and the logOR standard error (SE, y-axis) in a funnel plot

Full size image

Reasons for Discrepant Treg Function on Stroke Outcome

The Role of Infarct Size in Post-stroke Neuroinflammation

Evaluating all published reports on Treg function in experimental stroke suggests that lesion size might be a systematic confounder independent of reperfusion. It has been consistently reported that post-stroke peripheral immunosuppression occurs in stroke patients and animal stroke models only after extensive brain tissue injury [23, 60–62]. While mice and humans with substantial brain lesions develop lymphopenia and changes in monocyte subpopulation, small brain lesions induce only a minor immunomodulation, but no immunosuppressive syndrome [23, 63]. Surprisingly, the impact of peripheral immunosuppression after substantial brain injuries on the neuroinflammatory reaction has to our knowledge not been systematically investigated until now, but it is plausible that peripheral immune alterations might also affect central neuroinflammation. Indeed, a previous study performing a face-to-face comparison of the extent of cerebral leukocyte invasion, microglial activation, and cytokine secretion in three common models of brain ischemia of differing lesion size detected crucial differences among models with a manifold stronger inflammatory reaction in small permanent ischemia models than in extensive hemispheric lesions using TMVO models [18]. This discrepancy in the extent of the neuroinflammatory response between stroke models has also been confirmed in a recent multicenter experimental stroke trial [19]. The impact of brain-invading leukocytes on the local neuroinflammatory milieu became particularly evident in a previous study where inhibition of cerebral leukocyte invasion by antibodies against the alpha4-integrin (Anti-CD49d) resulted also in reduced microglial cell activation [64]. Therefore, it is plausible that Treg has an inferior role in stroke models, with only minor bystander inflammation. This comparably low neuroinflammatory response might be due to the secondary immunosuppression after extensive brain lesions or due to currently unknown factors (such as differences in skull trepanation, surgical wound, pain, and food intake) also in other stroke models, which have not been systematically characterized in respect to their inflammatory component. Moreover, the extent of secondary neuroinflammation and contribution to outcome has until now not been characterized in human stroke subtypes. These studies are urgently needed to get a better understanding of the specific inflammatory milieu and its role in stroke outcome depending on lesion size and location.

Secondary Microthrombosis in TMVO Stroke Models

The most commonly used acute brain ischemia models in experimental stroke research are transient mechanical vascular occlusion (TMVO) models with potential endothelial damage and induction of secondary microthrombosis [27, 65, 66]. Notably, such an experimental stroke model was used in both studies detecting a deleterious role of Treg in stroke [26, 44]. A decisive feature of this model is the occurrence of delayed neuronal damage due to secondary microthrombosis [65]. In contrast, secondary microthrombosis was not observed in animal models with gradual reperfusion [66]. Furthermore, the rate of secondary microthrombosis after recanalization in stroke patients is currently still unclear; however, microthrombosis might substantially contribute to pathophysiology particularly after mechanical recanalization of proximal arteries. Several potential mechanisms leading to microthrombosis and thromboinflammation have been suggested: (1) the extent of endothelial activation induced by the prompt reperfusion, (2) endothelial damage and thereby activation of the contact-dependent coagulation system by the intravascular manipulation, and (3) the occlusion time itself and thereby potential endothelial ischemic injury [27, 28, 67]. These methodological differences might have a major impact on the presence of thromboinflammation and secondary microthrombosis. Considering that microthrombosis is at least partially an inflammation-independent mechanism, the occurrence or absence of such a confounder to a cerebrovascular disease model has justifiably a critical impact on outcome. Moreover, the occurrence of secondary microthrombosis will shift the importance of local tissue-specific pathophysiological cascades toward the role of intravascular events, hence underestimating the contribution of inflammation in tissue homeostasis in cases where no microthrombosis/thromboinflammation occurs. The cascade of endothelial activation, thromboinflammation, and microvascular dysfunction in stroke models and potentially also in patients with pronounced endothelial damage not only leads to secondary ischemic infarct progression but will also alter the neuroinflammatory response to brain ischemia as such.

Milieu-Dependent Treg Function

Moreover, the above-stated differences between different experimental models as well as potentially between clinical stroke subtypes—including the occurrence of secondary microthrombosis, induction of peripheral immune alterations, and the extent of local neuroinflammation—might have a direct impact on Treg function. It is known from other disease paradigms that Treg function in vivo might depend on the specific immunologic milieu in order to preserve tissue homeostasis. In cancer immunology, this phenomenon has been termed the “Janus-faced function of Treg” [68, 69]. The concept that Treg adapt their suppressive function to a particular inflammatory milieu is based on functional studies of transcription factors, prominently the members of interferon regulatory factor (IRF) among others. It has been demonstrated that several transcription factors such as IRF, BLIMP-1, and GATA-3 are associated with Foxp3 regulation and contribute in a tissue-specific manner to Treg activation and function [70]. Therefore, the same genetic program in Treg might on one side contribute to tissue homeostasis (e.g., preventing an inflammatory collateral damage in the brain parenchyma after stroke) and at the same time has a detrimental function at a different site associated with the same disease (e.g., promoting thromboinflammation within the cerebral microvasculature). This hypothesis is well in line with the recent finding that Treg effector function is tightly controlled and differentially regulated by metabolic cues such as extracellular ATP and hypoxia [71], which evidently differs between post-stroke vasculature and the brain parenchyma. Therefore, it is conceivable that differences in oxygen partial pressure, glucose supply, and consequently ATP concentrations will differentially affect Treg function within the different compartments in the post-ischemic brain.

Particularly, the currently prevalent perception of post-stroke neuroinflammation as too much of a bad thing should be revisited. Several previous reports in primary autoimmune diseases as well as acute brain injury models have established the concept of “protective autoimmunity” (see Schwartz and Raposo [72] for a review). This concept ascribes secondary inflammation as a generally physiological and protective mechanism in which excessive immune activation as well as immunosuppression might be deleterious. Hence, in certain situations of post-stroke neuroinflammation—determined by the complex interplay of model-dependent features, kinetics, and the tissue-specific milieu—both Treg depletion as well as expansion might have a negative effect on stroke outcome [73–75]. Therefore, the immunological properties of the used stroke model, the targeted mechanisms of Treg function and potentially differing tissue-specific Treg functions need to be carefully considered before ascribing Treg in a potentially oversimplified view as good or bad immune cells after stroke.