The role of microglia and the TLR4 pathway in neuronal apoptosis and vasospasm after subarachnoid hemorrhage

Aneurysmal subarachnoid hemorrhage (aSAH) is a devastating disease that affects 30,000 Americans each year [1, 2]. A total of 30 to 40% of these patients will have delayed cerebral ischemia (DCI) from vasospasm, anywhere from 4 to 14 days after ictus, resulting in increased morbidity and mortality [3]. Few studies have addressed the molecular mechanisms that lead to DCI. The inciting event is the release of oxyhemoglobin from red blood cell lysis in the subarachnoid space [4, 5]. The heme that is released from hemoglobin results in a significant cerebral inflammatory response [6]. To dissipate the heme burden, the resident cerebral macrophage or microglia can engulf heme, as demonstrated in mouse intracerebral hemorrhage models, and degrade it via heme oxygenase [7, 8]. Furthermore, in primary macrophage culture, heme has been shown to be a specific agonist for the Toll-like receptor 4 (TLR4) [9].

In both SAH patients and mouse models, TLR4 expression is up-regulated in the brain [10‐12]. However, little is known about the signal transduction that occurs downstream of TLR4 to cause DCI or what cells mediate DCI. That is, after stimulation of TLR4, the MyD88 (myeloid differentiation primary response gene) pathway can facilitate further action by phosphorylation and activation of IRAK4 (IL-1 receptor associated kinase 4), which results in NF-kB-dependent inflammation [13]. Alternatively, TLR4 signaling can be transduced with the TRIF (toll receptor associated activator of interferon) pathway [13]. TRIF signaling results in a delayed NF-kB activation, similar to MyD88, and facilitates apoptosis [14].

The understanding of the immune cells involved in DCI is in its infancy. Neutrophils may play a role in DCI as systemic depletion resulted in improved cognitive performance and decreased vasospasm; however, their role in cerebral inflammation after SAH is likely indirect [15]. Neutrophils are not endogenous to the brain, nor do they have a published role in aneurysm formation. Resident macrophages of the brain, on the other hand, are first responders to the deluge of heme after SAH, and are necessary for aneurysm formation [9, 16]. As of yet, no causal relationship has been elucidated between macrophages and DCI. In this study, we show the critical role that microglia play in facilitating vasospasm and neuronal apoptosis. Furthermore, we establish a temporal role for the TLR4 pathway in the induction of apoptosis and vasospasm using both in vivo and in vitro models.

To determine the optimal time to visualize vasospasm, we performed a time course of subarachnoid hemorrhage from PODs 1 through 15 (Figure 1). The dimensions of the middle cerebral artery were analyzed at the level of the hippocampus for consistency, as shown in Figure 1C. We found that the difference between sham and wild type (WT) SAH vasospasm was maximal on days 3 and 10. There was no difference between sham and WT SAH on POD 5, indicating resolution of vasospasm by Day 5, followed by delayed vasospasm beginning on Day 7 and plateauing on Day 10 (Figure 1). To determine if neuronal cell death correlated with vasospasm in our model, we quantified TUNEL positive cells from days 1 through 15 in the dentate gyrus of the hippocampus. Similarly, cell death also had a bimodal distribution, peaking on PODs 7 and 15 (Figure 2).

We focused on PODs 7 and 15 because neuronal cell death was maximal at these two time points, and it correlated well with human vasospasm peaking on post-bleed Day 7, with continued risk of vasospasm through post-bleed Day 14 [21]. We then set out to determine the role of the TLR4 signaling cascade as it relates to vasospasm and neuronal apoptosis on PODs 7 and 15 (Figure 3). On PODs 7 and 15, vasospasm in the WT SAH was significantly increased compared to the TLR4−/− SAH (Figure 3A). Intracerebroventricular injection of LPS, a known TLR4 agonist, showed similar degrees of vasospasm to the WT SAH on both PODs 7 and 15 [22]. Furthermore, when comparing SAH in MyD88−/− to TRIF−/−, vasospasm was significantly greater in the TRIF−/− on POD 7, while the opposite was true at POD 15. Of note, when comparing maximal vasospasm on PODs 7 and 15, there was no difference between WT SAH and TRIF−/− SAH on POD 7, and WT SAH and MyD88−/− SAH on POD 15. Minimal vasospasm was seen in the sham and TLR4−/− SAH on both days.

Additionally, we measured neuronal apoptosis in these groups, as described in Figure 2. Interestingly, TRIF−/− SAH had a statistically equivalent neural apoptotic burden to WT SAH on POD 7, while on POD 15, there was no difference between MyD88−/− SAH and WT SAH. In terms of minimal apoptotic burden, there was no difference between the number of apoptotic neurons quantified in sham, TLR4−/− SAH, and MyD88 −/− SAH on POD 7. At POD 15, there was no difference between the number of apoptotic neurons quantified in sham, TLR4−/− SAH and TRIF−/− SAH (Figure 3B). Of note, the LPS injected mice, demonstrated significantly less neuronal apoptosis at POD 7, compared to WT SAH; however at POD 15 there was no difference between these groups. With the understanding that the TLR4 pathway may play a role in vasospasm and neuronal apoptosis, we wanted to identify what type of cell was expressing TLR4. Microglia, astrocytes and neurons were examined for TLR4 expression with representative images shown in Figure 4A-C and quantification of TLR4 co-localization in Figure 4D. Based on these results, we found that the majority of TLR4 is expressed in microglia at both POD 7 and 15 after SAH.

To further elucidate the relationship between microglia and vasospasm, we modified an in vitro assay [18] where we incubated neonatal primary microglial (PMG) culture with hemin for 24 hours to simulate the in vivo environment after subarachnoid hemorrhage. After 24 hours, the supernatant from the cultures was taken and incubated with axial sections of wild type mouse aortic rings for 5 minutes and vasospasm was measured. PMG from all genotypes, except the TLR4−/− mice, were able to induce vasospasm and secrete TNF-α (Figure 5A, B). Based on these results, the TLR4 receptor is necessary for microglia to secrete some factor into the media that causes vasospasm in mouse aortic slice culture. To exclude endotoxin contamination in our hemin preparation, endotoxin was measured by ELISA and found to be <0.001 ng/μl (data not shown).

The in vitro data suggest that the microglial TLR4 receptor is necessary for vasospasm, as well as the in vivo data which indicate that the TLR4 receptor is necessary for vasospasm. To resolve whether microglia are necessary for vasospasm in vivo, we depleted microglia in vivo via intraventricular injection of clodronate liposomes into WT SAH. Immunohistochemistry for microglia shows virtually complete depletion with clodronate compared to PBS liposomes (Figure 6A-C). Additionally, when vasospasm was measured at PODs 7 and 15, the depletion of microglia with clodronate resulted in significant amelioration of vasospasm at both time points (Figure 6D).

Finally, we determined whether microglial depletion effected neuronal apoptosis in Figure 7, and found a similar bimodal theme. At POD 7, the depletion of microglia resulted in a significant decrease in neuronal apoptosis (Figure 7A, C); while at POD 15, microglial depletion had no effect (Figure 7B, D).

In this study, we have elucidated possible roles for microglia and the TLR4 pathway with respect to vasospasm and neuronal apoptosis. Many have suggested a role for TLR4 in SAH and more recently a modulatory role for peroxisome proliferator-activated receptor gamma (PPAR-γ) via the TLR4 pathway [10‐12, 23]. However, our study is unique in that we attempted to ascertain the roles of different signaling pathways downstream of TLR4 activation and followed the mice out to 15 days to better parallel the human condition where delayed cerebral ischemia can occur up to 21 days from ictus [24].

It is interesting that two phases of vasospasm were observed, one beginning on POD 3, resolving by POD 5, and then another phase beginning on POD 7 and plateauing by Day 10 (Figure 1). This is important to note because it is similar to the human condition in that vasospasm is bimodal, with an early ictal phase followed by a delayed phase; it is the delayed vasospasm that is associated with increased neuronal cell death or delayed cerebral ischemia (DCI) [3, 21, 25].

Furthermore, our in vivo immunohistochemistry data suggest that TLR4 is necessary for neuronal apoptosis. It is interesting to note that neuronal apoptosis seen in WT SAH and TRIF−/− SAH was significantly greater than that seen in LPS on POD 7, despite equivalent degrees of vasospasm (Figure 3). On POD 7, one can infer that some product from red blood cell breakdown is able to induce increased cell-death in the dentate gyrus through the TLR4-MyD88 pathway and downstream signal transduction systems that LPS cannot; however, an insufficient amount of LPS could also be the case. Because there was no difference between neuronal apoptosis and vasospasm in WT SAH and TRIF−/− SAH at POD 7, we conclude that some portion of the neuronal damage and vasospasm occurring on POD 7 (early SAH) is TLR4-MyD88-dependent. Likewise, because there was no difference between neuronal apoptosis and vasospasm observed in WT SAH, MyD88−/− SAH and LPS on POD15, neuronal damage and vasospasm on POD 15 (late SAH) are largely TLR4-TRIF-dependent. Based on our results in Figure 3, we see that vasospasm is, for the most part, directly correlated with neuronal apoptosis. Given that vasospasm is either constant or increasing with time, it is expected that the apoptotic burden will not change or decrease only slightly as resolution of the insult occurs. The exception is the TRIF−/− SAH where vasospasm decreases with time and, therefore, we believe that the clearance of the apoptotic burden results in virtually no apoptotic cells by Day 15. Furthermore, the temporal relationship between TLR4-MyD88 activation and TLR4-TRIF activation is not unprecedented. Expression of NF-kB is seen in two phases after LPS stimulation of TLR4 [14]. MyD88 is responsible for the early phase and TRIF, the later phase, similar to our results. Because the role for microglia in DCI and SAH is largely unknown with respect to TLR4, we performed immunohistochemistry to verify that microglia do indeed express a majority of TLR4 in vivo after SAH at both POD 7 and 15. Although astrocytes and neurons express less TLR4, their role in DCI and cerebral inflammation may still be significant (Figure 4). To verify the relationship between the TLR4 pathway and microglia with respect to vasospasm, we performed in vitro vasospasm assays that confirmed a necessary role for the TLR4 receptor (Figure 5). While other groups have implicated microglia and heme in the pathogenesis of intracerebral hemorrhage and shown that TLR4−/− was protective, the downstream mediators of TLR4 were never examined with respect to cytokine production or vasospasm [26].

To elucidate the role of microglia in vivo, with respect to vasospasm and neuronal apoptosis, we depleted microglia using clodronate liposomes and showed that in both early and late phases of SAH, microglia are necessary for vasospasm. Also of note, neuronal apoptosis was abrogated by microglial depletion in early SAH only (Figure 7). With respect to a role for microglia in SAH, to our knowledge, we are the first to effectively deplete this specific cell type in the brain and demonstrate a functional response. Others have noted proliferation and activation of microglia after hemorrhagic stroke, specifically SAH and intracerebral hemorrhage [26, 27]. Interestingly, in hemorrhagic stroke, microglia seem to be detrimental by some accounts, whereas in neonatal stroke and neurodegenerative diseases they have a more beneficial role [26, 28, 29]. The caveat is that effects of microglial depletion were only examined at one time point after depletion. Perhaps, if other time points after microglial depletion had been studied, the role of microglia in the pathogenesis of these diseases would also change with time.

Taken together, our model suggests that there could be different phases of SAH. The early phase of SAH, where neuronal apoptosis is largely TLR4-MyD88-dependent and microglial-dependent, followed by a late phase that is characterized by a TLR4-TRIF dependent, microglial-independent neuronal apoptosis. Furthermore, vasospasm is characterized by an early and late phase response that depends on MyD88 and TRIF, respectively. Finally, microglia seem to be both necessary and sufficient to cause vasospasm in both the early and late phases of SAH, based on our in vitro and in vivo models (Figures 5 and 6).

These findings may explain why therapies tailored to aSAH patients have failed for the most part. If these data can be translated to the SAH patient population, novel immunotherapies could be conceived that target different components of Toll like receptor signaling and microglia at different points in the patient’s hospital course to alleviate the cerebral inflammatory burden and improve outcomes.