Intracerebral Pro-inammatory Cytokine Increase in Surgically Evacuated Intracerebral Hemorrhage - an Observational Microdialysis Study

Background: Treatment options for spontaneous intracerebral hemorrhage (ICH) are limited. A possible inammatory response in the brain tissue surrounding an ICH may exacerbate the initial injury and could be a target for treatment. Methods: In this observational study, ten patients needing surgical evacuation of supratentorial ICH received two cerebral microdialysis (MD) catheters; one in the perihemorrhagic zone (PHZ), and one in non-eloquent cortex (SNX) remote from the ICH. The microdialysate was analysed for energy metabolites (including lactate/pyruvate ratio (LPR) and glucose) and for inammatory mediators using a multiplex immunoassay of 27 cytokines and chemokines at 6-10 hours, 20-26 hours and 44-50 hours after surgery. Results: Deranged metabolic outcome, supportive of the hypothesis that inammation plays a role in secondary brain injury following ICH. In this study, we sampled interstitial uid using paired MD catheters, one in the perihemorrhagic zone (PHZ) and one in seemingly normal and non-eloquent cortex (SNX), to investigate changes in inammatory mediators in the acute phase after surgical evacuation of ICH. Following analysis of routine low-molecular weight metabolites (glucose, lactate, pyruvate, glycerol, glutamate), the remaining microdialysate was investigated by a Multiplex immunoassay. We hypothesized a difference in pro- and anti-inammatory cytokine and chemokine expression in the PHZ compared to SNX, indicative for distinct inammatory proles in these brain areas.

shows an acute hypometabolic and hypoperfusion state (7)(8)(9) including mitochondrial dysfunction and metabolic failure,(10, 11) and this region may be particularly vulnerable to secondary injury. Furthermore, in experimental ICH studies an in ammatory response has been observed that includes rapid activation of microglia, (12)(13)(14)(15) followed by the in ltration of in ammatory cells over hours and days.(16) This in ammatory response may result in breakdown of the blood-brain barrier exacerbating the vasogenic oedema and tissue damage.(16) In animal ICH models, neurologic outcome can be improved by attenuating the in ammatory process. (17)(18)(19)(20)(21)(22)(23) There are no available pharmacological treatment options with proven clinical bene t targeting the secondary injury following ICH. (24) Clinical studies using early glucocorticoid administration in an attempt to attenuate the ICH-induced in ammation resulted in an impaired clinical outcome. (25) These results emphasize the complex contribution of in ammation to post-ICH outcome and since the in ammatory response is necessary as part of the healing process, it is essential to understand the dynamics and complex interaction of in ammatory mediators in brain tissue following ICH.
Cerebral microdialysis (MD) is a useful clinical tool used to monitor the microenvironment of the brain tissue in the neurocritical care patient. MD has been extensively used in human traumatic brain injury (TBI) and subarachnoid hemorrhage (SAH) although only rarely in clinical ICH research. (11,(26)(27)(28)(29) In recent years, and with the introduction of catheters with a larger pore size, MD has enabled sampling of in ammatory mediators from the interstitial uid of brain tissue in humans (30,31) and in animal models. (32) The cytokine expression has been explored in SAH and TBI patients, (31,(33)(34)(35)(36)(37)(38) demonstrating cerebral production of cytokines,(36) but no such studies exist in ICH. Furthermore, several studies have explored the concentration of various pro-in ammatory mediators in plasma samples of ICH patients, showing a correlation between levels of IL-6, (39)(40)(41)(42) TNF-α (41,42) and poor outcome, supportive of the hypothesis that in ammation plays a role in secondary brain injury following ICH.
In this study, we sampled interstitial uid using paired MD catheters, one in the perihemorrhagic zone (PHZ) and one in seemingly normal and non-eloquent cortex (SNX), to investigate changes in in ammatory mediators in the acute phase after surgical evacuation of ICH. Following analysis of routine low-molecular weight metabolites (glucose, lactate, pyruvate, glycerol, glutamate), the remaining microdialysate was investigated by a Multiplex immunoassay. We hypothesized a difference in pro-and anti-in ammatory cytokine and chemokine expression in the PHZ compared to SNX, indicative for distinct in ammatory pro les in these brain areas.

Patients
Adult patients >18 years old and requiring emergent surgery for ICH using an open craniotomy, at the Department of Neurosurgery, University Hospital Linköping, Sweden, during the period 2016-2018 were prospectively recruited to the study. Exclusion criteria included severe coagulation disorders and a known source of bleeding, such as an aneurysm or an arteriovenous malformation. The ICH was evacuated via a standard craniotomy and by microneurosurgical evacuation. Intracranial pressure (ICP)-monitoring was achieved using either Neurovent-P parenchymal pressure monitoring device (Raumedic AG, Helmbrechts, Germany) or Bactiseal external ventricular drainage (EVD) catheter (DePuy Synthes, Raynham, USA). One microdialysis (MD) catheter (CMA-71 Brain Catheter, M-Dialysis, Solna, Sweden) was inserted via the craniotomy into the perihemorhagic zone (PHZ) aiming for within 1 cm of the evacuated ICH, and one catheter was inserted ipsilateral to the hemorrhage in seemingly normal and non-eloquent cortex (SNX). MD catheters were inserted at a 45 degree angle aiming for the microdialysis membrane to include the cortex-white matter junction. As it is essential for the interpretation of MD sampling to know the location of the membrane(43, 44) a post-operative CT-scan was performed to verify MD catheter placement.
Three of the patients were included in a previous publication from our group evaluating energy metabolic disturbances using MD following ICH surgery. (29) Patients were treated according to a standardized neurocritical care protocol to avoid secondary insults. This included intubation and mechanical ventilation if the patient was unconscious (Glasgow Coma Scale motor score (GCS-M) ≤5), aiming for normoventilation, normovolemia, and normothermia. ICP was maintained at ≤ 20 mmHg and cerebral perfusion pressure (CPP) > 60 mmHg with the use of volume substitution and inotropic drugs such as norepinephrine or dobutamine, when needed.

Metabolite analysis
Interstitial levels of glucose, lactate, pyruvate, glycerol and glutamate in the MD samples were analysed bedside using an ISCUS Flex® analyser (M Dialysis AB, Solna, Sweden). The lower limit of detection (LLOD) was 1.0 μmol/L for glutamate, 0.1 mmol/L for glucose and lactate, 10 μmol/L for pyruvate and 0.22 mg/mL for glycerol. No sample preparation was needed. Metabolite concentrations in the MD samples were analysed by the enzymatic methods using the ISCUS Flex® analyser, immediately after sample collection. The sample volume required for the different metabolites was 0.5 µL for glucose, 0.2 µL for lactate, 0.5 µL for pyruvate, 0.5 µL for glycerol, 1 µL for glutamate and 0.5 µL for urea. Following the analysis of these metabolites, approximately 30 µL of dialysate remained that were used in the Multiplex immunoanalysis.
Multiplex immunoassay MD samples from three different time-periods were analyzed for in ammatory mediators using the MSD (Meso Scale Discovery, Rockville, USA) MULTI-SPOT Assay System V-PLEX Human Proin ammatory Panel 1, Cytokine Panel 1 and Chemokine Panel 1(cat #K15210D). The proteins in each panel were detected by immunoassays. Antibodies with electrochemiluminescent labels bind the proteins, the MSD instrument applies a voltage to emit light from these labels and measures the intensity of the light, providing a quantitative measure of each protein concentration.
Pooling of three MD vials, after analysis of the routine analytes, was needed to achieve a su cient sample volume of 25 µl per well enabling the analysis and thus a 6-hour time resolution was achieved.

Assay Protocol
After blocking using the blocker H for one hour with shaking for one hour, plates were washed three times with 150 μL Wash Buffer (PBS + 0.05% Tween 20). Thereafter, 25 μL sample or calibrator were added. The plates were sealed and incubated at room temperature with shaking for two hours. The plates were washed three times with 150 μL Wash Buffer, after which 25 μL of antibody solution was added to each well. The plates were sealed and incubated at room temperature with shaking for two hours, and subsequently washed three times with 150 μL Wash Buffer. To each well, 150 μL of 2X Read Buffer T was added, and the plates were analyzed with a MESO QuickPlex SQ 120 instrument.
Data distribution was assessed using Shapiro-Wilks' normality test. Normally distributed data are presented as mean and standard deviation (SD). Non-normally distributed data are presented as median and range, median and interquartile range or median and individual values.
Low-molecular weight metabolite data were analysed using linear mixed model (MML) method using patient number as subject level and catheter location as xed effect.
Multivariate analysis (MVA) was performed by overviewing the data using principal component analysis (PCA) and thereafter tting orthogonal projections to latent structures discriminant analysis (OPLS-DA) model to the data. Critical outliers were investigated with Hotelling's T2, a multivariate generalization of con dence interval, in the PCA. Any data point outside the resulting >99% con dence interval was removed from analysis. Moderate outliers were investigated using distance to model X (DModX) (50). Model validity was investigated with a cross-validated ANOVA (CV-ANOVA) and a p-value < 0.05 considered signi cant. R2 and Q2 are presented for the OPLS-DA model along with number of latent variables. Scaling to unit variance and mean centering was employed. Variables with a │p(corr)│> 0.4 and VIP > 1 were considered signi cant. (50) Univariate analysis was performed using paired t-test for normally distributed data and paired Wilcoxon signed rank test for non-normally distributed data.
In the cytokine data analysis, values below lower limit of detection (LLOD) of the assay were substituted with the exact value given for LLOD by the manufacturer.

Results
Ten patients (8 males, 2 females; Table 1) underwent surgery with evacuation of intracerebral hemorrhage (ICH) and placement of dual microdialysis (MD) catheters ( Figure 1 a-b) and were thus included in the study. Median age was 64 (51-71) years; median time to surgery 27.8 (6-82) hours and median MD sampling time was 95 (50-148) hours. Mean distance from MD catheter to evacuated ICH was 6 (±5.0) mm for the PHZ catheter and 24.5 (±7.6) mm for the SNX catheter. Mean volume of ICH was 77 (±25.8) mL (Table 1). The PHZ catheter of patient 10 was not placed according to protocol as it was found to be located >10 mm from the hematoma cavity. However, since it was on CT control located in tissue affected by the ICH and the surgical approach, it was included in the analysis. The SNX catheter of patient 7 was placed in parenchyma in which an ischemic infarction developed post-surgery, thus an area not representing seemingly normal cortex, and the data from this MD catheter was excluded from further analysis.
Metabolites MD-Glucose levels were signi cantly lower in PHZ compared to SNX (Figure 2a; p < 0.05), however, in both locations MD-glucose levels were consistently above the 0.2 mmol/L critical level,(47) suggesting adequate substrate delivery. Nevertheless, the lactate pyruvate ratio (LPR) was signi cantly higher in PHZ compared to SNX (Figure 2b; p<.05) indicative of a metabolic crisis in the PHZ. (29) Although the LPR elevation in the PHZ was particularly evident early following surgery, it persisted beyond 48 hours. MDglutamate and MD-glycerol levels were also signi cantly higher in PHZ compared to SNX ( gure 2c-d; p < 0.05).

Cytokines and chemokines
All cytokines were in detection range in the microdialysate although the expression of each cytokine was highly variable with the highest and lowest levels differing by a 4-fold dynamic range. TNF-β was recovered in less than 2% of the samples and was therefore excluded from analysis. In total 1512 multiplex analysis results were obtained.

Univariate analysis
Univariate analysis of cytokines and chemokines revealed a signi cantly higher level of IL-2, IL-8 and IL-1α at 20-26 hours post-surgery whereas IL-6 and IL-4 levels were increased at 44-50 hours in the PHZ when compared to the SNX (Figure 3; p < 0.05). Several cytokines, including most anti-in ammatory ones, displayed a similar pattern in the PHZ and the SNX (Supplemental gure I). Chemokines generally displayed a pattern of higher levels in the SNX when compared to PHZ, however, only TARC levels were statistically different (Supplemental gure II).

Multivariate analysis
A multivariate analysis with unsupervised principal component analysis (PCA) illustrated clustering in the data, showing some group separation (Figure 4a). Hotelling's T2 analysis showed two critical outliers (>99% T2) which were subsequently removed. Neither global nor pair wise tests results were changed by removal of these outliers. Score scatter plot for PCA showed inherent group separation in the data ( Figure   4a). The corresponding loading scatter plot showed PHZ data clustering largely due to metabolites and pro-in ammatory cytokines (Figure 4b). The pro-in ammatory and metabolite variables clustered near time point 2 (corresponding to 20-26 hours post-surgery) as opposed to time points 1 and 3 (corresponding to 4-10 and 46-50 hours post-surgery, respectively).
A supervised OPLS-DA model was then tted (OPLS-DA (1+0+0) R2 0.179, Q2 0.23; p<.001) with MD catheter location as class (Figure 4c). The OPLS-DA model was used to evaluate which metabolites and cytokines discriminate between the two MD catheter locations. To determine which variables contributed most strongly to the model │p(corr)│ > 0.4 and VIP > 1 was used, resulting in four metabolites and eight cytokines ( Table 2) showing the strongest contribution to the difference between the PHZ and SNX. Notably, low molecular weight metabolites suggestive of a metabolic crisis and tissue injury such as lactate, LPR, pyruvate and glutamate, were signi cantly higher in PHZ when compared to the SNX. Furthermore, the pro-in ammatory cytokines TNF-α, IL-6, IFN-γ, IL-1β, IL-8, and IL-2 were higher in PHZ compared to SNX, as were the anti-in ammatory cytokines IL-13 and IL-4, and growth factor VEGF-A.

Discussion
We used an approach of placing paired microdialysis catheters, one in the vicintiy of the intracerbral hemorrhage (ICH) into the perihemorrhagic zone (PHZ) and one into non-injured brain tissue in a region not affected by the surgical approach or the ICH-the seemingly normal cortex (SNX). Our key ndings were that both univariate and multivariate analysis showed a higher expression of predominantly proin ammatory cytokines in the PHZ when compared to the SNX area. These differences were most evident at 20-26 hours, although some pro-in ammatory cytokines peaked at the later 44-50 hours post-surgery interval. Pro-in ammatory cytokines IL-8, TNF-α, IL-2, IL-1β, IL-6 and IFN-γ, were signi cantly higher in PHZ compared to SNX. Increased pro-in ammatory cytokine expression in perihemorrhagic tissue was previously observed in animal models including expression of IL-1β, TNF-α, and MIP-1α (18, 51-54), although there are no previous studies of cytokine expression in brain tissue of ICH patients. Our present ndings support the presence of a pro-in ammatory environment in the tissue surrounding an ICH and provides information on the temporal pro le of the in ammatory cascades. VEGF-A, which mediates increased permeability of the blood-brain-barrier (BBB)(55, 56), was distinctly higher in PHZ compared to SNX, which may thus potentially aggravate vasogenic edema and cause further secondary brain injury.
Analysis of low-molecular weight metabolites showed a metabolic crisis in the perihemorrhagic tissue, persisting beyond the initial 48 hours post-surgery, consistent with a previous study by our group. (29) The initial high level of MD-glutamate in PHZ suggests an ongoing neuronal death immediately following surgery and a subsequent plausible cell membrane degradation could contribute to the increase in MDglycerol seen over time.
There is also a bene cial role for the in ammatory response in ICH, in particular the M2 phenotype of microglia, involved in clearing debris and repair of tissue following ICH (12-15, 18-20, 23, 57-63). IL-13 and IL-4 are known to be anti-in ammatory or have a modulating effect on other pro-in ammatory mediators(64) and in the present study were increased in the PHZ. IL-4 has been shown to convert M1 phenotype microglia toward an M2 phenotype,(65, 66) further illustrating that tissue reaction to ICH is complex and the balance of pro-and anti-in ammatory mediators may shift over time.
Cytokines can have both a dual and con icting role. To classify them as either anti-or pro-in ammatory may be overly simplistic, as the immune modulators constitue an intricate network of interconnected paracrine and autocrine molecules whith both positive and negative feed-back loops inherent to the system.(67) Therefore, it is important to consider both the context, the timing and the target of any given cytokine in order to understand and interpret its action as either a driver or a moderator of the immune response.(68) In particular IL-6, TGF-β and IFN-γ can be either pro-or anti-in ammatory depending on timing and target.(68) In similarity to our present study, both anti-and proin ammatory cytokines are released simultaneously in the tissue following TBI.(69, 70) One known limitation of intracerebral microdialysis is the lack of control levels, and interpretation of absolute levels must be made with caution. Our approach enables evaluation of the temporal pro le of the in ammatory cascades but also a comparison to a relatively uninjured brain region. While the SNX is in a region exposed by the surgical approach, it is not injured by the surgery. While we cannot exclude an in ammatory response also in the SNX, our data argue for more marked changes occurring in the PHZ. Furthermore there are known challenges with interpretation of microdialysis data with regards to relative recovery, which is dependent on several factors, (47,71) and which was not determined in the present study. However, relative recovery can be assumed to be similar between the paired catheters, thus differences between these should not be affected by changes in recovery.
Here, we did not measure in ammatory mediators in blood plasma or cerebrospinal uid. Previous studies have shown elevated serum levels of IL-6 and TNF-α in ICH patients, (39)(40)(41)(42) suggesting some mediators of the in ammatory response can be monitored also in blood. Cytokines generally have a paracrine target, however, and as such the cytokine expression in brain tissue can be expected to contribute to ongoing in ammatory response without any measurable corresponding increase in plasma concentration. In addition, a previous study of 22 TBI patients showed the concentration of cytokines to be several times higher in cerebral interstitial uid, measured by microdialysis, than in blood.(36) The same study showed recovery of 42 cytokines from the interstitial uid, and demonstrated distinct temporal pro les in 16 of those 42. Similarly, in the present study, TNF-α, IL-8, IL-1β and IL-6 were found to peak within the rst two days of monitoring,(36) although there was no comparison of peri-contusional and more normal tissue. Of note, while TBI and ICH may share some characteristics, they are also two distinct disease entities with unique pathophysiology. Thus, the data in this study are highly novel since cytokine expression in ICH patients has rarely been measured, and only as a part of a mixed cohort of neurointensive care patients including both SAH and TBI patients. Due to the unique set-up of this present study we were able to compare the perihemorrhagic expression of cytokines to those in control brain tissue. This control tissue, while not directly targeted by the surgical approach nor in uenced by the ICH, it should not be regarded as entirely uninjured. However, the distinct differences between the two catheters argues against against a marked tissue reaction and cytokine release from insertion of the microdialysis catheter per se. Due to ethical reasons the insertion of microdialysis catheters in ICH patients who do not undergo surgical evacuation is not possible, therefore there are no control patients in our cohort without craniotomy. The surgical approach, performed using microsurgical technique to minimize the trauma to the surrounding tissue, may have contributed to the release of in ammatory mediators.
In this present study we do not report on outcome of the tissue surrounding MD catheters. Therefore, we cannot say that the demonstrated increase in pro-in ammatory cytokine expression in PHZ is detrimental to tissue outcome, however, previous animal studies show improved neurological function and tissue outcome following augmentation of in ammatory responses. (51,(74)(75)(76) The low number of measured time points (three) limit the time resolution and preclude the discovery of cytokines possibly peaking later than 50 hours as has previously been demonstrated for IL-12p70 and IL-10 in TBI patients.(36) It is plausible that some protein expression may have been elevated beyond the studied 50 hours. Nevertheless, our results show that the most signi cant changes occurred already at 20-26 hours after surgery suggesting an early in ammatory peak following ICH. It is plausible that antiin ammatory cytokines may also have an increasingly important role in the PHZ over a longer period of time, thus beyond the scope of our study design. Furthermore, we do not know the concentration of cytokines prior to the insertion of the MD catheters, thus any peak prior to MD insertion would also evade detection.
Multivariate analysis of complex dataset is excellent at discerning correlations within the data material and displaying such correlations in understandable two-dimensional graphs. The unsupervised methods, such as principal component analysis (PCA), show inherent correlations in the data material. The supervised models, such as orthogonal projection to latent structures discriminant analysis (OPLS-DA) attempts to de ne which variables contribute to a difference between speci ed classes of data. The OPLS-DA utilizes cross validated (CV)-cycles to validate the model and although many samples are omitted in each CV-cycle, strong groupings may persist and cause the estimate of the predictive power to be overestimated. A possible limitation of the supervised model could be low R 2 and Q 2 values, particularly if the model was to be used for class prediction. In this present study the OPLS-DA model was employed to discern overall correlations and differences in the data. A highly signi cant model was developed, but the R 2 and Q 2 point to a rather large amount of noise in the data. Small sample size is also problematic, but the model is robust enough for exploratory analysis, which was the purpose in this case. Biomarker discovery for diagnostics naturally requires a higher R2 and Q2, but this was not the aim in this study. Instead multivariate analysis (MVA) was employed to describe the overall correlation patterns in the data. Further studies are required to verify the ndings in this present study.

Conclusions
This study shows an increased expression of pro-in ammatory cytokines occuring in the brain tissue surrounding an intracerebral hemorrhage 20-26 hours following surgical evacuation. This may constitute a target for future therapies aiming to reduce potential secondary brain injury. Microdialysis monitoring of the cytokine expression in interstitial uid may also be a method of monitoring future therapies aimed at reducing the secondary injury caused by in ammation following ICH. Abbreviations ICH = intracerebral haemorrhage; ICP = intracranial pressure; IFN = interferon; IL = interleukin; IP-10 = interferon-gamma induced protein 10; LPR = lactate pyruvate ratio; MCP = monocyte chemoattractant protein; MD = microdialysis; MIP = macrophage in ammatory protein; PHZ = perihemorrhagic zone; SNX = seemingly normal cortex; TNF = tumor necrosis factor; VEGF-A = vascular endothelial growth factor A; VIP = variable of importance in projection.

Declarations
Ethics approval and consent to participate The regional ethical committee in Linköping, Sweden approved the study protocol (decision number 2014/236-31). The study was carried out in accordance with relevant guidelines and regulations, including the World Medical Association (WMA) Declaration of Helsinki. Since the included ICH patients could not themselves consent to the study, a written informed consent was obtained from the patient's closest relative.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This study was supported by local hospital grants from Region Östergötland, and by grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALFagreement (Lio-925101).

Author contributions
LT and NM devised the study. AG and KR analysed the in ammatory mediators. LT and NM drafted the manuscript and subsequently all authors edited and revised the manuscript.     Microdialysis (MD)-glucose (A) levels were signi cantly lower in the perihemorrhagic zone (PHZ) when compared to those in seemingly normal cortex (SNX) (p < 0.05) although consistently above critical (0.2 mmol/L) and warning levels (0.8 mmol/L) in both locations. (47) The LPR (B), however, was pathologically elevated in the PHZ indicating a metabolic crisis in the brain tissue. In contrast, the SNX LPR normalized within the rst hours after surgery and thereafter remained within the normal range (B).
MD-glutamate levels decreased with time in the PHZ but was signi cantly higher than in the SNX (C; p < 0.05) during the initial 48 hours. In addition, MD-glycerol levels were persistently higher in the PHZ when compared to the SNX (D; p < 0.05). Abbreviations: LPR = lactate pyruvate ratio. Data are presented as mean and standard error of the mean (SEM) for clarity. to the group separation between PHZ and SNX in the supervised OPLS-DA model (highlighted with grey background in this gure) did not reach statistically signi cant differences when explored using univariate statistical methods. In contrast, IL-1α was signi cantly higher in PHZ at 20-26 hours after surgery (p < 0.05; Wilcoxon signed rank) in univariate analysis, but did not contribute to the multivariate model.

Figure 4
Multivariate data analysis showed group separation between perihemorrhagic zone (PHZ) and seemingly  Table 2). The majority of the anti-in ammatory cytokines and chemokines did not contribute signi cantly to the model, thus suggesting that their concentrations were similar in the PHZ and the SNX.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. supplemental gure1.tif supplemental gure2.tif