Inflammation in delayed ischemia and functional outcomes after subarachnoid hemorrhage

Patients with acute SAH admitted to the Neuroscience Intensive Care Unit at the Memorial Herman Hospital-Texas Medical Center, Houston, TX, between July 2013 and March 2015 were enrolled in a prospective observational study. Institutional review board approval was obtained (HSC-MS-12-0637). Inclusion criteria were (a) age > 18 years and (b) spontaneous aneurysmal SAH diagnosed by computed tomography (CT) within 24 h of ictus or by xanthochromia in cerebrospinal fluid if the CT was not diagnostic. Exclusion criteria were (a) non-aneurysmal SAH due to trauma, arteriovenous malformation, or mycotic aneurysms and (b) comorbidities that could significantly affect baseline inflammation including autoimmune diseases, suspicion of infection, history of malignancy, and current pregnancy. Written informed consent was obtained from the patient or surrogate.

Clinical data collected included age, sex, medical and social history, neurological status at admission based on the World Federation of Neurosurgical Societies scale (WFNS grade) [17], and loss of consciousness at ictus. Admission CT images were adjudicated for the presence of intraventricular hemorrhage, hydrocephalus, and cerebral infarction [18, 19]. All CT scans were independently evaluated by a study neurologist for the amount and location of blood as categorized by the modified Fisher score (mFS) [20, 21].

We adhered to the proposed definition of DCI [22] as “The occurrence of focal neurological impairment (such as hemiparesis, aphasia, apraxia, hemianopia, or neglect), or a decrease of at least 2 points on the Glasgow Coma Scale (either on the total score or on one of its individual components [eye, motor on either side, verbal]). This should last for at least 1 hour, is not apparent immediately after aneurysm occlusion, and cannot be attributed to other causes by means of clinical assessment, CT or MRI scanning of the brain, and appropriate laboratory studies.” DCI was diagnosed only after rigorous exclusion of other possible causes. Delayed cerebral ischemia was adjudicated through consensus of at least two attending neurointensivists in weekly research meetings. In the rare instance when no consensus can be achieved, the principal investigator made the final determination. Characteristics of the CT scan including IVH, modified Fisher scale, hydrocephalus, and cerebral infarction were graded by the PI of the study blinded to other aspects of the clinical course. Specifically, any blood in the ventricles was graded as IVH. Functional outcome was assessed by the 0–6 modified Rankin Scale (mRS). The mRS at discharge was assessed by the attending neurointensivist taking care of the patient. All physicians are certified to grade the mRS. The 3-month mRS was obtained via a follow-up phone call by a trained research personnel using a standardized questionnaire. Unfavorable outcome at 3 months after discharge was defined as a mRS of 3–6. Subjects were treated according to standard guidelines.

Serum samples were collected at four predetermined time points after admission: < 24 h (T₁), between 24 and 48 h (T₂), 3–5 days (T₃), and 6–8 days (T₄). Forty-one serum cytokine concentrations in picograms per milliliter were measured following the manufacturer’s protocol using a MAGPIX magnetic bead-based ELISA 41-plex assay (EMD Millipore, Billerica, MA): sCD40L, EGF, Eotaxin (CCL11), FGF-2, Flt-3 ligand (FLT3L), Fractalkine (CX3CL1), G-CSF (CSF3), GM-CSF (CSF2), GRO (CXCL1), IFN-α2, IFN-γ, IL-1α, IL-1β, IL-1ra, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12 (p40), IL-12 (p70), IL-13, IL-15, IL-17A, IP-10, MCP-1 (CCL2), MCP-3 (CCL7), MDC (CCL22), MIP-1α, MIP-1β, PDGF-AA, PDGF-AB/BB, CCL5, TGF-α, TNF-α, TNF-β, and VEGF. They were analyzed based on previously published protocols [14] (see Additional file 3 under Cytokine analysis).

Study subjects’ characteristics were described using unpaired Student’s t tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables. The non-parametric Mann-Whitney U test (p <  0.10) was used to identify candidate cytokines in each time period that differentiated DCI and functional outcomes (Additional file 3: Table S1). Using these candidate cytokines, multivariate logistic regression analyses were performed to find cytokines independently associated with DCI and unfavorable functional outcome after adjusting for clinically relevant variables including age, sex, mFS (as a radiographic surrogate for injury severity), and WFNS grade (as a clinical surrogate for injury severity), respectively. The results of the multivariate logistic regression analysis were reported as an odds ratio (OR) at a 95% confidence interval (CI). A p-value of ≤ 0.05 was considered statistically significant (two-tailed). All statistical analyses were performed using SPSS 17.0 (SPSS Inc., Chicago, IL).

During the study period, 151 patients with an acute aneurysmal SAH were admitted to our unit. One hundred twelve patients met the inclusion criteria and consented to participate in the study. Since the baseline inflammatory status (and cytokine levels) can vary subject-to-subject, to preserve homogeneity, we prioritized the inclusion of consecutive subjects from whom the maximum number of samples were available from T₁ to T₄. Of the 60 subjects included, 52 subjects had 3 or more samples (a total of 189 samples) and only 8 subjects had less than 3 samples (a total of 14 samples). We analyzed 203 samples in total. The differences in demographics between subjects who were included in the study versus subjects who were not included are shown (Additional file 3: Table S2). The mean age of the subjects was 52.2 ± 13.0 years (range 23–83 years) and 46 (76.7%) were women. Seventeen subjects had their aneurysm clipped, 40 had embolization with coils, and 3 subjects were untreated. Fourteen subjects (23.3%) developed DCI on an average of 6.2 days after admission. Sixteen (26.7%) subjects were reported to have poor functional outcome (mRS ranging from 3 to 6 score at 3 months).

There were no significant differences in clinical variables between the DCI and no-DCI group. Subjects with poor functional outcomes were older and had higher WFNS grade and LOC at ictus, higher mFS grade, and hydrocephalus on admission CT compared to those with good functional outcomes (Table 1). In the multivariable model, age and high-grade WFNS were independent predictors of poor functional outcome (Additional file 3: Table S3).

At T₁, T₂, T₃, and T₄, the levels of 2, 4, 3, and 13 cytokines, respectively, were significantly different across the DCI and no-DCI group. At the same time points, the levels of 8, 7, 5, and 11 cytokines, respectively, were significantly different across the good and poor outcome groups (Additional file 3: Table S1). Multivariable logistic regression analyses reveal that higher PDGF-ABBB at T₂, higher CCL5 at T₂ and T₃, lower IP-10 at T₁, and lower MIP-1α at T₄ was independently associated with DCI after adjusting for age, sex, mFS, and WFNS. In addition, IL-6 at T₂ and MCP-1 at T₄ were independently associated with poor mRS at 3 months after adjusting for age, sex, and mFS but not WFNS grade (Table 2).

We plotted the trends of median values of six cytokines (PDGF-ABBB, RANES, IP-10, MIP-1α, IL-6, and MCP-1) that were independently associated with either DCI or poor outcome. Median levels of PDGF-ABBB (particularly at T₂) and CCL5 (particularly at T₂ and T₃) were consistently higher, and median levels of IP-10 (particularly at T₁ and T₂) and MIP-1α (particularly at T₄) were consistently lower in the DCI group compared to the no-DCI group (Fig. 1a–d). Median levels of IL-6 (particularly at T₂) and MCP-1 (particularly at T₄) were consistently higher in subjects with poor functional outcome than those with good functional outcome (Fig. 1e, f).

We identified distinct cytokine associations in the DCI and functional outcome group from T₁ to T₄. In the DCI network at T₁ and T₂, we observed a cluster including VEGFA, EGF, and FGF2 and another cluster consisting of IL9, IL15, and IFNA2 in the DCI cohort. Another small cluster including the cytokines PDGF-ABBB and CCL5 was common to both groups and the cytokines were elevated and independently associated with DCI. This cluster was observed from T₂ to T₄ as well (Fig. 2 and Additional file 1: Figure S1).

For functional outcome networks, IL-6 and MCP-1 (CCL2) were highly correlated (which were independently associated with poor outcomes) and extend to other cytokines. IL-6 was closely associated with a group of cytokines (including CCL2, TNF-α, IL-8, IL-10, IL-1Ra, and IP-10) which formed a unique cluster in poor outcome group from T₁ to T₄ (Fig. 3 and Additional file 2: Figure S2).

PDGF-ABBB (at T₂), CCL5 (at T₂ and T₃), IP-10 (at T₁ and T₂), and MIP-1α (at T₄) were independently associated with DCI. Platelet-released PDGF is a mitogenic growth factor linked to peripheral thrombosis, and increased PDGF levels have been reported in ischemic brain injury and brain puncture wounds [23]. CCL5 is a member of the CC-chemokine family produced by a variety of cells, including T lymphocytes, platelets, endothelial cells, smooth muscle cells, and glial cells. It can interact with chemokine receptors that recruit leukocytes to inflammatory sites, helping migration of leukocytes across the endothelium contributing to the pathogenic process of arterial injury and atherosclerosis [24, 25]. T-cell recruitment and activation in the injured brain parenchyma could be the prelude to secondary ischemic injury [26, 27] and have been shown to migrate across the vessel wall in response to brain injury [28]. T cell-directed chemokines (including CCL5) are particularly important in stroke prediction including DCI [29]. Interestingly, IP-10 and MIP-1α were lower in DCI compared to the no-DCI group. Both cytokines have shown to be involved in both pro-inflammatory and regulatory mechanisms. IP-10 signaling, with the combination of CCL5, is a known promoter of T cell recruitment and activation [30]. Two known pathways involving IP-10, the restraining of IP-10 production by CCL5 [31] and the inhibition of regulatory T cell recruitment by IP-10 [30], have been previously implicated in controlling inflammatory response and in neuroprotection after ischemic insult [26, 27]. MIP-1α is primarily involved in inflammation and homeostasis [32] and plays an important role in both acute and chronic inflammation, primarily in the role of recruiting pro-inflammatory cells. It has been implicated in the development of ischemic brain injury [15]; however, they also promote healing by inducing inflammatory responses against external injuries [32].

A small cluster (including platelet-released PDGF-ABBB and CCL5) was observed in the DCI and no-DCI group. The levels of most cytokines in this cluster were significantly different across the DCI/no-DCI group at T₂ and T₃ suggesting increased coagulatory and inflammatory activity prior to DCI (which typically occurs after T₃). Previous studies have shown that early activation of the coagulation cascade and impairment of the fibrinolytic cascade can lead to DCI after SAH [33] and the levels of platelet-activating factors increase within 4 days after SAH [34]. Elevated concentrations of the von Willebrand factor (the primary initiator of coagulation) have been shown to be associated with DCI as well [35, 36]. Furthermore, other studies have shown that overactive inhibition of fibrinolysis is associated with DCI [37]. Another cluster including IFNg, MIP1A, IL17a, IL12p70, and VEGFa was observed in DCI. VEGF is involved in neurogenesis and in the inhibition of apoptosis [38], promoting neuroprotection. Also, vascular CD4+ TH1 cells induce the release of IL12 and IFNg [39] which are mediators of tissue damage.

In addition to cytokine clusters, we observed quantitative differences in the networks as a whole. At T₂ (typically a period prior to when DCI occurs), the DCI and no-DCI group had a similar number of cytokines and significant correlations (19 and 13 vs 24 and 32). At T₄ (a period where DCI typically occurs), the DCI group had a higher number of cytokines and correlations compared to no-DCI (22 and 14 vs 12 and 10) suggesting an overall increase in systematic inflammatory mechanisms. Additionally, the concentrations of 13 cytokines differed significantly across the DCI and no-DCI groups at T₄ (more so than in any other time period) suggesting a peak in systemic inflammation at T₄. Overall, there is a simultaneous activation of angiogenesis, inflammation, and thrombotic mechanisms during the course of DCI [9].

IL-6 (at T₂) and MCP-1 (at T₄) were independently associated with functional outcomes confirming the importance of the association between early inflammation and clinical outcomes. Our findings are consistent with previous studies where IL-6 and MCP-1 were shown to be sequentially expressed in CSF after SAH [40, 41]. MCP-1 is a CC chemokine family member and can regulate leukocyte trafficking by affecting their chemotactic activity, activating inflammatory cells, and modulating interactions between leukocytes and endothelial cells [42]. Previous studies have shown that activated mononuclear leukocytes can be found surrounding the major cerebral arteries in the subarachnoid space after SAH [43], and MCP-1 has been found to be significantly expressed in the major cerebral arteries in experimental vasospasm [7]. Our findings confirm that early rise in serum IL-6 and MCP-1 are associated with poor clinical status and consequently poor clinical outcomes. This finding further emphasizes the importance of “early brain injury” (defined as the injury process occurring < 72 h after aneurysm rupture) and its impact on functional outcomes [44, 45].

In the poor functional outcome group, a unique cluster including the cytokines IL1R1, TNFa, IL8, IL6, CCL2, CCL11, IP10, and CSF3 (Fig. 3) was observed at T₁ to T₄. This suggests the presence of inflammatory mechanisms mediated by IL1a, TNFa, and IL6. IL1 has been shown to upregulate the pro-inflammatory IL6. It is released by macrophages and microglia via an endogenous DAMP-mediated inflammatory response triggered by heme released from lysed blood cells [46]. IL1R1 is a IL1 receptor antagonist which blocks the actions of IL1 and has been identified as a therapeutic candidate in ischemic diseases [47]. In pre-clinical models, administration of IL1RA has been shown to reduce blood-brain barrier disruption and mitigation of cell death, and, clinically, it has also shown to reduce inflammation as well [48]. We observed a high correlation between TNFa and IL8 in MRS ≥ 4 group in comparison to MRS ≤ 3 (0.80 vs 0.33 and T₁). This strong correlation persisted through T₄ indicating the mediation of inflammation via persistent activation of the NF-kB-dependent pathway. It is likely that inflammation-triggered oxidative stress increases TNFa, which in turn increases IL8 expression [49]. TNFa activates vascular Rac-1 promoting vasoconstriction, and targeting TNFa has shown to mitigate neuronal injury [50, 51]. Network analysis shows strong correlation between TNFa, IL8, and IL1R1 suggesting that IL1- and NF-kB-mediated responses are associated. Additionally, a strong correlation between IL-6 and MCP-1 was observed in the poor outcome group. It is purported that IL-6 may induce MCP-1 expression via signal transducer and activator of transcription 3 after SAH [15]. Overall, network analysis suggests simultaneous activation of pro-inflammatory and anti-inflammatory mechanisms sustained through the acute stages of SAH which consequently leads to poor functional outcomes. This analysis also shows that cytokines that are not significantly elevated across groups can still play an important role in the underlying inflammatory mechanisms.