Background
Over 30,000 Americans will fall victim to an aneurysmal subarachnoid hemorrhage (SAH) this year; nearly half of these patients will die within 6 months [
1,
2]. Of those SAH survivors, approximately 50 % will develop severe cognitive and functional deficits [
3,
4]. Although the majority of research in SAH has focused on the treatment of vasospasm, only nimodipine has been shown to improve outcome [
5,
6]. Multiple clinical trials have demonstrated that even when vasospasm was effectively treated, morbidity and mortality were not ameliorated [
7‐
9]. These studies suggest that neurological injury can be vasospasm-independent. Further research aimed at mitigating this heme-induced cerebral inflammatory response is required [
10].
In SAH, the heme from blood spilled into the subarachnoid space is metabolized by heme oxygenase (HO), generating excess free iron [
11]. This iron is hypothesized to enable cell membrane damage via free radicals [
12] with studies showing a causal relationship between unbound iron and brain injury following SAH [
13]. Deferoxamine (DFX), an iron-chelating agent, has been shown to be neuroprotective in various hemorrhagic models via several mechanisms [
12,
14]. Previous studies using a rat model of SAH showed decreased brain edema, oxidative stress, and neuronal apoptosis after DFX treatment [
15,
16]. However, none of these studies assess the necessity of heme oxygenase-1 (HO-1 or Hmox1) in DFX neuroprotection nor has the efficacy of intracerebroventricular administration been fully examined. Our lab previously showed microglia to be critical in red blood cell-induced neuroinflammation [
10], and most recently, we found microglial HO-1 to be neuroprotective after SAH in a mouse model [
17]. We hypothesized that microglial/macrophage HO-1 is critical for DFX neuroprotection and that intracerebroventricular administration would provide superior neuroprotection in a mouse model of SAH.
We undertook the current set of experiments to first compare the effects of systemic versus intracerebroventricular injection of DFX on neuronal damage, vasospasm, pro-inflammatory and oxidative biomarkers, and immune cell populations in a mouse model of SAH; second, to see if microglial/macrophage HO-1 is sufficient for DFX neuroprotection; and third, to compare the effects of cell-specific HO-1 knockouts on DFX neuroprotection and cognitive outcome. Our study provides a platform for the potential translation of DFX treatment into the SAH patient population.
Methods
All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Beth Israel Deaconess Medical Center (BIDMC). The facility is accredited by the Association for Assessment and Accreditation of Lab Animal Care and fully complied with all Federal, State and Local Law. Animals were housed at BIDMC and fed a standard rodent diet ad libitum with 24-h access to either water and/or hydrogel while kept on a 12-h light/12-h dark cycle. All surgical manipulations were performed under general anesthesia with ketamine (10 mg/kg) and xylazine (4 mg/kg), and buprenorphine (50 μg/kg) was systemically administered. All mice used were male on a C57BL/6 background (The Jackson Laboratory). Cell-specific HO-1 knockout in myeloid cells (
LyzM
Cre
:
Hmox1
fl/fl
) and astrocytes and neurons (
Nes
Cre
:
Hmox1
fl/fl
) was achieved as previously described by our lab [
17]. All mice had similar fur color and were approximately the same size and weight. The average mouse weight was 25 g (0.025 kg) with a weight range of 24 g (0.024 kg) to 27 g (0.027 kg). Wild-type (WT) mice were randomly assigned between the following four treatment groups equally: WT subarachnoid hemorrhage (SAH) sham + intraperitoneal (IP) normal saline (NS) + intracerebroventricular (ICV) NS (SAH sham + vehicle), WT SAH + IP NS + ICV NS (SAH + vehicle), WT SAH + IP deferoxamine (DFX) + ICV NS (SAH + IP DFX), and WT SAH + IP NS + ICV DFX (SAH + ICV DFX). Lab personnel performing surgical procedures were not the same as those performing cognitive assays to allow for appropriate blinding.
SAH
The method used to induce SAH has been previously tested and validated in a mouse model [
18]. After the mice were anesthetized with xylazine (10 mg/kg) and ketamine (12 mg/kg), SAH was performed as previously described by our lab using a standard stereotaxic instrument set-up (KOPF Instruments, Tujunga, CA, USA) [
17]. To open the skin overlying the anterior skull, a midline incision was performed. Then, a burr hole was drilled into the anterior skull, 4.5 mm anterior to the bregma. Sixty microliters of autologous blood from a C57BL/6 wild-type blood donor mouse was injected over a 10-s period with a 27-gauge needle at a 40° caudal angle into the drilled burr hole. The needle was left in place for 5 min to prevent backflow of blood.
Intracerebroventricular injection of deferoxamine
After the mice were anesthetized, intracerebroventricular injection was performed using a standard stereotaxic instrument set-up (KOPF Instruments, Tujunga, CA, USA). One burr hole was drilled 0.22 mm posterior to the bregma, 1 mm lateral, and 2.25 mm in depth to enter the ventricle. On SAH POD1, 24 h after the induced SAH, a single, non-repeated injection of 8 mg/kg of DFX was administered using pre-measured capillaries. Dosing was chosen based on dose-tolerance data generated by our lab (Table
1).
Table 1
Intracerebroventricular deferoxamine dose-tolerance chart
0.8 mg/kg (~0.02 mg per mouse) | 3 | Well tolerated |
8 mg/kga (~0.2 mg per mouse) | 3 | Well tolerated |
80 mg/kg (~2 mg per mouse) | 2 | Immediately died |
Intraperitoneal injection of deferoxamine
Starting on SAH POD1, 24-h after the induced SAH, the mice were given systemic injections of 200 mg/kg of DFX every morning, 30 min prior to cognitive test, until euthanization on SAH POD7. Dosage was chosen based on a previous publication testing systemic deferoxamine treatment in another mouse model of hemorrhagic stroke [
19].
TUNEL
All in vivo imaging was taken on SAH sham or SAH post-operative day (POD) 7 due to our lab’s previous publication showing SAH POD7 to have the most significant hippocampal cell damage [
10]. Brain sections and HT-22 cells were stained with terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL; Roche Life Science, Indianapolis, IN, USA). Slides were covered using Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, CA, USA) for nuclear counterstaining. HT-22 cells were counterstained with Hoechst 33258 (Sigma-Aldrich, Natick, MA, USA). Lab personnel interpreting TUNEL stains were not aware of the groups to which they were assigned.
H&E
Coronal brain sections were stained with hematoxylin and eosin (H&E) (Poly Scientific R&D Corp., Bay Shore, NY, USA). The middle cerebral artery (MCA) lumen radius/wall thickness ratio was quantified using ImageJ software (NIH). Lab personnel interpreting H&E stains were not aware of the groups to which they were assigned.
Confocal imaging
Eight-micrometer coronal brain sections from each experimental group were post-fixed and permeabilized, followed by being blocked with 10 % donkey serum. The sections were then stained with the following primary antibodies: goat anti-Iba1 (1:500) and rabbit anti-HO-1 (1:500) to identify HO-1 expression and HO-1 co-localization to microglia. The sections were then stained with the following secondary antibodies: donkey anti-goat 488 (1:250) and donkey anti-rabbit 594 (1:250). The sections were then counterstained with nuclei marker DAPI and sealed. The sections were viewed and images were taken on a Zeiss confocal microscope. Layers of images from the most anterior to the most posterior portion were taken, and a Z-stack image was created. ImageJ software was used for co-localization cell counting as well as co-localization heat maps. Co-localization percentage per high-powered field for each experimental group’s confocal Z-stacked images was obtained by dividing the total number of yellow-positive by the total number of DAPI-positive cells. Co-localization histograms were generated using the coloc-2 feature of ImageJ software. In brief, the program combines the green HO-1 single channel-1 saturation pixels with the red Iba-1/microglia single channel-2 saturation pixels to calculate the level at which they overlap. The x-axis represents channel-1 pixel intensity, while the y-axis represents channel-2 pixel intensity. Blue represents the lowest population frequency possible while yellow represents the highest. The lower the slope for each heat map, the more HO-1 staining there is per Iba-1/microglia staining.
ELISA
Whole brain lysates were equally loaded onto a 96-well plate and cerebral IL-6 concentration was determined using ELISA Max Biolegend kit per manufacturer’s instructions. Media from microglia-neuron trans-well experiments were equally loaded onto a 96-well plate and culture TNF-α concentration was determined using ELISA Max Biolegend kit per manufacturer’s instructions (Biolegend, San Diego, CA, USA).
Flow cytometry
All flow cytometry acquisition was performed on a FACSCalibur (BD Biosciences, San Jose, CA, USA), and analysis was completed using FlowJo software (FlowJo, LLC, Ashland, OR, USA). Cells were isolated from whole brain or blood and re-suspended in FACS buffer (1 % bovine albumin, 2 mM ethylenediaminetetraacetic acid (EDTA), and 0.05 % NaN
3 in phosphate-buffered saline (PBS)). To block unspecific sites, the cells were first stained with CD16/32 Trustain (1:100; Biolegend, San Diego, CA, USA). The cells were then washed with FACS buffer and stained with the following fluorescent-tagged antibodies: PE-Gr-1 and PeCy7-CD11b (1:100; Biolegend, San Diego, CA, USA). To identify the myeloid cell populations in the blood and brain, CD45
+ cells were gated off of a CD45/SSC-H dot plot. Then, using a Gr-1/CD11b dot plot, macrophages were identified as CD11b
hi/Gr-1
lo while neutrophils were CD11b
hiGr-1
hi (Additional file
1: Figure S1). To measure the total mitochondrial superoxide anion production, whole brain cell lysates were incubated with MitoSOX red mitochondrial superoxide indicator per manufacturer’s instructions (5 μM; Life Technologies/ThermoFisher, Cambridge, MA, USA) and cells positive in the FL-2 channel were reported. Appropriate unstained controls for each channel were used to determine stained cell populations.
Primary microglia and neuronal HT-22 cells—trans-wells
Microglia cells were isolated from the brains of neonatal mice using the Neural Tissue Dissociation Kit (P) (Miltenyi Biotec, Cambridge, MA, USA). The resulting mixed glia culture, containing astrocytes and microglia, was cultivated in media containing macrophage-colony stimulating factor (M-CSF). After 1 week of cultivation, the microglia were collected and grown on 3-μm cell culture inserts (EMD Millipore, Billerica, MA, USA). Murine hippocampal neuronal HT-22 cells were grown on six-well plates in normal media without M-CSF. For trans-well assays, the inserts with microglia were placed on top of the HT-22 neuron wells; microglia and HT-22 neuronal cells shared media during experiments. A total of 200 μl of whole blood was collected from the submandibular vein of a wild-type donor mouse and placed into 10 ml of PBS to keep the blood from clotting. The whole blood in PBS was then centrifuged at 2000 rpm for 5 min. The plasma, leukocyte, and platelet layers were aspirated, leaving an erythrocyte pellet. The erythrocytes were washed in an additional 10 ml of PBS and centrifuged at 2000 rpm for 5 min. The supernatant was aspirated and the erythrocyte pellet was re-suspended in 2 ml of PBS; 100 μl of this RBC suspension in PBS was added to the top microglia-containing chamber of the microglia-neuron trans-well; after 1-h of blood exposure, DFX (100 μM) or vehicle was added to the microglia chamber, and an additional hour of incubation was completed.
Western blot
Primary microglial cell lysates were equally loaded onto a polyacrylamide gel and transferred to an immune-blot PVDF membrane (Bio-Rad, Hercules, CA, USA). Membranes were blocked with 5 % milk and stained with rabbit HO-1 antibody (1:1000, Abcam, Cambridge, MA, USA) and mouse vinculin antibody (1:1000, Sigma-Aldrich, Natick, MA, USA).
Morris water maze
Cognitive performance was assessed using the Morris water maze as previously described, with minor modifications made by our lab [
17]. In brief, after 7 days of acquisition, SAH sham or SAH was performed. On the day of SAH sham or SAH surgical procedure, the mice were not tested on the Morris water maze. The mice resumed testing on the Morris water maze on POD1. The mice to be treated with DFX were then either given daily IP DFX injections starting on POD1 and ending on POD7 or a one-time ICV injection of DFX on POD1. Additionally, mice that received daily IP DFX injections also got a one-time ICV injection of NS on POD1, and the mice that received a one-time ICV DFX injection also got daily IP NS injections from POD1 to POD7. SAH sham or SAH mice that were not treated with DFX, each received daily IP NS from POD1 to POD7 and a one-time injection of ICV NS on POD1 (Table
2). Spatial memory testing, measured by time to reach goal platform and consisting of 1-trial in the morning per animal per day, was started on SAH POD1 and continued for 7-days. On SAH POD4, the goal platform was moved to the opposite side of the maze (spatial reversal), while visual cue locations were unchanged. An investigator blinded to treatment groups performed the maze procedures.
Table 2
Wild-type experimental mouse groups
WT SAH sham + IP NS + ICV NS | SAH sham + vehicle | Mice received subarachnoid hemorrhage (SAH), sham surgical procedure, daily intraperoteneal (IP) injections of normal saline (NS) starting on post-operative day (POD) 1 and ending on POD7, and a one-time intracerebroventricular (ICV) injection of NS on POD1. |
WT SAH + IP NS + ICV NS | SAH + vehicle | Mice received SAH surgical procedure, daily IP injections of NS starting on POD1 and ending on POD7, and a one-time ICV injection of NS on POD1. |
WT SAH + IP DFX + ICV NS | SAH + IP DFX | Mice received SAH surgical procedure, daily IP injections of deferoxamine (DFX) starting on POD1 and ending on POD7, and a one-time ICV injection of NS on POD1. |
WT SAH + IP NS + ICV DFX | SAH + ICV DFX | Mice received SAH surgical procedure, daily IP injections of NS starting on POD1 and ending on POD7, and a one-time ICV injection of DFX on POD1. |
Statistical analysis
Multiple experimental groups were compared using repeated-measures two-way ANOVA with Bonferroni’s post hoc test for in vivo TUNEL staining and HO-1/Iba1 co-localization of multiple brain regions, and the results are presented as the mean ± SD (GraphPad Prism). Morris water maze data was presented as the mean ± SEM (GraphPad Prism) and analyzed using repeated-measures two-way ANOVA with Bonferroni’s post hoc test. For all other statistical comparisons, multiple experimental groups were compared using one-way ANOVA with Bonferroni’s post hoc test, and the results are presented as the mean ± SD (GraphPad Prism). Differences were considered significant at P < 0.05.
Study approval
All procedures involving animals were approved by the IACUC and the Radiation Safety Office (RSO) of Beth Israel Deaconess Medical Center.
Discussion
In a mouse model of SAH, we found that DFX exerted neuroprotective effects by non-canonical mechanisms. (1) DFX improved cognitive outcomes and reduced cerebral damage, independent of vasospasm. (2) ICV DFX was the most neuroprotective. (3) ICV DFX decreased neuroinflammatory markers. (4) Microglial HO-1 is sufficient for DFX neuroprotection in an in vitro model of blood-induced inflammation. (5) ICV DFX neuroprotection and cognitive improvement is dependent on microglial/macrophage HO-1.
The iron-chelating agent, DFX, has been tested for therapeutic use in many animal neurological disease models including Huntington’s disease [
20], traumatic brain injury [
21‐
23], cerebral ischemia [
24‐
26], and hemorrhagic stroke among others. DFX has been extensively studied in animal models of intraventricular hemorrhage (IVH) and intracerebral hemorrhage (ICH). In IVH animal models, DFX reduced ventricular enlargement, brain damage, and markers of post-hemorrhagic chronic hydrocephalus [
27‐
31]. In ICH animal models, DFX has been shown to reduce brain damage [
32‐
37], decrease neuroinflammation [
34,
38‐
40], and improve cognitive outcome [
38,
41]. Further, DFX was shown to reduce DNA damage [
37,
40], oxidative stress [
32,
33,
38], neuronal hemoglobin expression [
42], and autophagy markers [
43], following ICH. Additionally, in a germinal matrix hemorrhage model of neonatal rats, DFX reduced brain damage, ventricular dilation, and improved cognitive outcome [
44].
Although DFX treatment has also been studied in SAH animal models, experiments using DFX treatment specifically in a mouse model of SAH are lacking. In rat models of SAH, DFX has been shown to decrease overall mortality, edema, oxidative stress, and neuronal death [
15,
16]. Additionally, DFX treatment after SAH has been shown to decrease cortical apoptotic markers [
16] and reduce markers of brainstem damage in rats [
45], as well as reduce lipid peroxidation markers and improve sodium-potassium ATPase activity in guinea pigs [
46]. Our study looked to further elucidate the mechanisms of DFX-induced neuroprotection in a mouse model of SAH.
In our current study, administration of DFX
after induction of SAH in our mouse model was effective in reducing the cerebral inflammatory response. DFX administered via two different routes, reduced cortical and hippocampal damage after SAH on POD7; the greatest reduction was seen with ICV DFX, followed by IP DFX (Fig.
1b,
c). Further, both IP and ICV DFX treatment improved cognitive outcome during a later phase of SAH POD7, but only ICV DFX treated mice showed early improvement on POD4 and 5. Although DFX administration effectively reduced brain damage (Fig.
1b,
c) and improved cognitive outcome (Fig.
1d) after SAH, it had no effect on vasospasm (Fig.
1f,
g). These results were interesting for two reasons. First, the IP dose of DFX is 25-fold greater than the ICV dose, and yet the ICV dose showed more neuroprotection and earlier cognitive improvement. A potential explanation is that ICV administration allows for proximity to the heme burden, while IP-administered DFX has to effectively cross the blood brain barrier. Second, DFX provided cerebral protection and improved cognition after SAH independent of any effect on vasospasm, similar to previous studies that showed DFX treatment had no effect on vascular response after SAH [
47,
48]. Vasospasm-independent cerebral protection provided by DFX is not surprising when one considers recent clinical trials that effectively treated vasospasm but did not improve morbidity or mortality after SAH [
7,
8].
DFX has been previously shown to have anti-inflammatory effects in hemorrhagic stroke [
14]. We investigated whether these anti-inflammatory effects might be partly mediated by changes in immune cell populations, cerebral IL-6 concentration, or mitochondrial superoxide anion production. The SAH + vehicle group showed an upward trend in the cerebral microglial/macrophage cell population as compared to the SAH sham + vehicle group, but DFX did not significantly reduce this increase. Cerebral neutrophils and hematogenous populations of macrophages and neutrophils remained unchanged with DFX administration (Fig.
2a,
b). Further, SAH caused a significant increase in both cerebral IL-6 concentrations and mitochondrial superoxide production as compared to sham; in the ICV DFX-treated group, cerebral IL-6 concentrations were reduced and a trend towards decreased reduction in mitochondrial superoxide production was present (Fig.
2c–
e). These results indicate that ICV DFX may partially exert protective effects not by changing the total number of microglia/macrophage cells, but instead via modulation of the pro-inflammatory mechanics of these cells, while systemic DFX injection does not.
We investigated whether DFX neuroprotection was dependent on microglia using a microglia-neuron trans-well assay. These assays revealed that DFX offered protection against red blood cell (RBC) induced neuronal damage, even when DFX was added to microglia
after RBC exposure had already begun (Fig.
3a,
b). When we repeated these trans-well assays with HO-1
−/− microglia, DFX offered no neuroprotection (Fig.
3c,
d). Additionally, we found that DFX treatment increased the protein expression of HO-1 in primary microglial culture (Fig.
3e,
f). These results suggest that microglial HO-1 is critical to the mechanism of DFX neuroprotection, possibly, in part, by facilitating the increased expression of microglial HO-1. Our lab has previously shown that administration of carbon monoxide (CO) rescues the neuronal injury seen in co-cultures with HO-1
−/− microglia [
17]. This indicates that the neuroprotective product of heme breakdown via microglial HO-1 in the context of microglia-neuron co-cultures is CO. Since this current work shows microglial HO-1 to be increased following DFX treatment, it is likely that increased CO production may be involved. Additionally, DFX would chelate the excess iron from heme breakdown, potentially leading to synergistic benefits produced by DFX administration: increased CO protection and decreased iron toxicity.
When we looked at confocal images of all of our experimental groups, we found that SAH markedly increased the co-localization of microglia and HO-1 compared to sham. Further, we found that ICV DFX treatment affected a significant increase in HO-1 expression within microglia while IP DFX did not (Fig.
4a–
c). Because these results demonstrated that ICV DFX most effectively increased microglial HO-1 expression in vivo, we next sought to ascertain the necessity of microglial HO-1 for ICV DFX neuroprotection. Mice lacking myeloid HO-1 (
LyzM
Cre
:Hmox1
fl/fl
) and mice lacking neuronal and astrocyte HO-1 (
Nes
Cre
:Hmox1
fl/fl
) were exposed to SAH and then treated with ICV DFX. Interestingly, ICV DFX treatment after SAH protected the
Nes
Cre
:Hmox1
fl/fl
mice similarly to the
Hmox1
fl/fl
. Conversely,
LyzM
Cre
:Hmox1
fl/fl
mice showed significantly more neuronal damage and cognitive dysfunction compared to
Hmox1
fl/fl
mice (Fig.
5c–
e). This, together with the in vitro data, supported the hypothesis that myeloid HO-1, but not astrocyte or neuronal HO-1, was critical for ICV DFX to reduce brain damage and improve cognition after SAH.
It is still not completely clear why the lack of myeloid HO-1 would suppress DFX neuroprotection in vivo, but we speculate that the DFX mediated increase in myeloid HO-1 cannot occur in the myeloid HO-1 knockout (
LyzM
Cre
:Hmox1
fl/fl
) mice. Without the increased myeloid HO-1 expression, the subsequent increase in neuroprotective CO production would be absent. In our previous work, we showed that the neuronal damage and cognitive dysfunction seen in
LyzM
Cre
:Hmox1
fl/fl
mice could be saved by administering external CO, showing that CO was the neuroprotective byproduct of myeloid HO-1 heme breakdown [
17]. Since our current in vitro data suggests that microglial HO-1 is crucial for DFX neuroprotection, it is possible that increased myeloid HO-1 expression in vivo, due to DFX administration, could cause increased CO production and thus result in better protection. Further experimentation would be necessary to test this theory.
We chose the anterior circulation model [
18] over the endovascular perforation model and acknowledge that there are limitations; however, we felt that the strengths of the anterior circulation model outweighed the weaknesses. In the anterior circulation model, the increase in intracranial pressure is less severe. Additionally, blood entering the subarachnoid space of a mouse in this method would be that of a donor mouse. On the other hand, the endovascular perforation method better approximates the intracranial pressure crises that can occur in SAH patients.
However, the anterior circulation model has a number of advantages over the endovascular perforation method. First, the amount of blood (60 μl) injected into each mouse, and the resultant increased intracranial pressure is consistent between mice. Because of this, we feel that results obtained using the anterior circulation method are better reproduced. Further, the lower mortality seen with this model is helpful when performing experiments on conditional knockouts, as well as dual injection procedures required for SAH and intracerebroventricular DFX. We have used the anterior circulation method in our past research and believe it to be suitable for our current research interests as well.
In stroke patients, DFX reduced serum markers of oxidative stress and increased antioxidant species [
49], while in ICH patients, phase-I testing revealed DFX to be safe and well tolerated [
50]. Currently, promising clinical trials investigating the use of DFX for ICH are underway [
51]. Our research provides a platform for linear translation of DFX treatment into the human SAH population. Our data show that intracerebroventricular DFX yields the greatest neuroprotection via a mechanism that is dependent on microglial HO-1 and possibly a protective microglial polarization. Given the fact that high-grade SAH patients will have an external ventriculostomy drain (EVD) placed at admission, a feasibility study for the use of intracerebroventricular DFX in these patients should be explored in the future. Furthermore, monitoring patient HO-1 expression during DFX treatment for hemorrhagic stroke may help clinicians identify patients that are more likely to respond to treatment.