White-matter injury—along with neuronal loss—is an integral part of the pathologic processes that accompany cerebrovascular diseases. Poor white-matter recovery significantly affects the long-term outcome after an acute stroke [1
], impairs sensorimotor function and causes profound neurobehavioral and cognitive impairment [2
]. The main function of myelin, which is to increase the speed with which electrical impulses propagate along the myelinated fiber, is well known. Other functions of myelinating cells, including structural, metabolic and energetic support of the axon, are far less studied or have been established recently [3
Despite its importance for normal brain functioning and recovery after stroke, demyelination has rarely been investigated in the animal models of brain ischemia. The studies on the animal middle cerebral artery occlusion (MCAO) model [4
] studies showed that myelin destruction begins from the first day and even as early as hours after stroke. The mechanisms of demyelination [5
] include intramyelinic edema that manifests as myelin pallor, further vacuole formation and sheath destruction followed by myelin debris clearance by microglia and macrophages. The main part of the destroyed myelin is removed by the 7–10th day after the stroke [4
]. In addition, oligodendrocytes, which are extremely sensitive to ischemia, die, which complicates the remyelination of remaining axons [5
]. Molecular mechanisms underlying oligodendrocyte death include glutamate excitotoxicity, mitochondrial dysfunction, oxidative stress and proinflammatory factors [10
The brain is inherently more vulnerable to ischemia than any other organ and suffers significant damage after only a few min of complete cerebral blood flow interruption. The localization, severity of lesions, and reparative response are significantly dependent on the animal model of cerebral ischemia. The most widely used and standardized models of focal ischemia, such as the (MCAO) model, are characterized by damage to the striatum and cortex [4
]. In contrast with this type of models, a common feature of the global cerebral ischemia (GCI) model is primary damage to the hippocampus [13
]. These models are relatively rare, less standardized and used in several variants. There are fundamentally different GCI models of hypoxia, in which only inhibition of systemic circulation occurs and models of complete interruption of cerebral blood flow that simulates cardiac arrest in humans. Among the GCI models, only a few provide complete interruption of blood flow followed by reperfusion. A simple closure of the common carotid arteries completely interrupts the blood flow to the brain only in gerbils, which, unlike mice or rats, have an incomplete a Circle of Willis [20
]. The model of GCI with complete interruption of the brain blood flow in a mouse or rat requires more comprehensive techniques.
The three-vessel model of GCI [16
] used in this study provides complete interruption of the cerebral blood flow, followed by reperfusion that mimics cardiac arrest in humans and emulates possible injuries of the hippocampus after it. Changes in the hippocampus observed in rats after GCI are similar to lesions in the hippocampus in humans undergoing anoxia [21
]. Thus, MRI studies demonstrate a decrease in the volume of the hippocampus in both hemispheres in patients after anoxia during cardiac arrest or an ischemic episode [22
]. These results are confirmed by post-mortem histological studies of brain sections of patients after anoxia, which reveal a limited bilateral lesion of the hippocampus, resulting in a complete loss of pyramidal neurons of the CA1 field [24
], extensive loss of pyramidal neurons of the CA3 field and partial damage to the CA2 field [25
In humans, damage to the hippocampus caused by anoxia can lead to severe anterograde amnesia, which confirms the role of the hippocampus as a critical structure for memory function [21
]. The pyramidal neurons of the CA1 field of hippocampus are extremely vulnerable to ischemic conditions. Synaptic terminals are the primary and early target in the development of damage to postischemic neural cells [26
]. Disruption of ion transport and depolarization of presynaptic membranes leads to excessive release of glutamate from synaptic terminals into the intercellular space, which leads to metabolic disruption and triggers cell death in neurons of the CA1 field [15
Despite the importance of damage to the hippocampus, such simulations of damage in cardiac arrest in humans and the timing of these lesions are poorly documented. Most studies report neuronal loss in the CA1 field of the hippocampus, enhancement of neurogenesis, astrogliosis and inflammation [16
]. Demyelination in the three-vessel model of GCI has not yet been described. The study aimed to assess demyelination in the hippocampal layers and closely related neuronal loss, neurodegeneration, neuroinflammation, astrogliosis and oligodendrogenesis in the hippocampus on the three-vessel model of GCI. Although the study focused more on the assessment of damage to the hippocampus, signs of neurodegeneration were assessed in the neocortex and thalamus adjacent to the hippocampus.
Our results for the first time have demonstrated myelin and oligodendrocyte loss of the hippocampus on the rat model of GCI, which implies complete interruption of cerebral blood flow and simulates cardiac arrest in humans. We investigated the dynamics of neuronal loss, neurodegeneration, astrogliosis, inflammation, utilization of disrupted myelin by microglia/macrophages and specific rod-shaped changes of microglial morphology.
To the best of our knowledge, demyelination has not been studied in a rat model of GCI to date. Earlier Lee et al. [33
] in the gerbil model of GCI found myelin loss in the hippocampus, which started from 4 days after ischemia. Our results show that myelin and oligodendrocyte loss was observed 10 days after GCI and intensified by the 30th day in the hippocampal layers adjacent to the pyramidal neurons in the CA1 field (SO and SR), as well as the SP layer itself. Unlike focal ischemia, when there is an interruption of the blood supply to a limited area, global cerebral ischemia leads to a temporary cessation of blood supply to the entire brain. However, short-term cessation of blood flow leads to serious damage to only certain brain regions. A distinctive feature of the global ischemia-hypoxia model in rodents is severe damage to the hippocampal pyramidal neurons in the CA1 field [13
], which is confirmed by the results of our previous [17
] and present studies. In addition, a number of studies have found damage to the neurons of the cerebellum, cortex, striatum and thalamus in global ischemia-hypoxia [36
]. Our study also found neurodegeneration in the neocortex and thalamus, slight on the 10th day after ischemia and increasing by the 30th day. The hippocampal subfields respond differently to GCI: pyramidal neurons in the CA3 field are relatively resistant to transient global cerebral ischemia, while neurons in the CA1 field are more susceptible to hypoxia. The death of neurons in this area is called the “delayed hippocampal neuronal death” [13
], noticeable 4–7 days after GCI in gerbils [33
]. Lee et al. [33
] found an almost complete loss of hippocampal pyramidal neurons in the CA1 field in gerbils as early as 4 days after GCI. Our results demonstrated more delayed neuronal death in the rat model of GCI: we found the death of approximately half of neurons on the 10th day and more than 90% of neurons on the 30th day after GCI the CA1 field, whereas the CA2 and CA3 fields were less affected. The higher sensitivity to ischemia of the CA1 pyramidal neurons in gerbils compared to rats may be associated with increased NO synthesis [41
] or metabolic features of GABAergic pyramidal neurons in the hippocampus, which is associated with the propensity of this species to epilepsy [42
Apparently, pyramidal neurons of the CA1 field die mainly by the mechanism of glutamate-dependent excitotoxicity [44
], since they receive a huge number of exciting glutamate inputs, which is dozens of times more than the number of inhibitory GABA inputs [45
]. Main inputs are represented by afferents from the entorhinal cortex as part of alvear and performant pathways [46
], Schaffer collaterals and numerous inputs from GABAergic interneurons within the hippocampus, among which up to 12 subtypes are distinguished [47
]. The hippocampal network is very complex and is still the subject of morphologic studies [47
]. One pyramidal neuron of the CA1 field can receive more than 30 thousand inputs in different layers of the hippocampus, including the SO, SP, SR and SL [45
]. Among these numerous inputs the long axons from entorhinal cortex within alvear and perforant are myelinated, located in all layers of the hippocampus except the SE and having contacts with the basal and apical dendrites of the CA1 pyramidal neurons [49
]. In addition, the SO and SP contain the myelinated axons of the CA1 pyramidal neurons that send fibers to the subiculum, entorhinal cortex and other extrinsic brain structures [50
]. The Schaffer collaterals within the SL and their branches in the SR, which contacted with apical dendrites of the CA1 pyramidal neurons, are largely unmyelinated or weakly myelinated [51
According to our results, on Day 10 after GCI significant myelin and oligodendrocyte loss was found only in the SO and SP. By the 30th day, myelin loss in these layers intensified and affected the SR, the number of oligodendrocytes decreased in the SR and SL. It is likely that earlier changes (Day 10) were found to be associated with axonal myelin destruction of dead pyramidal neurons of the CA1 field, while later changes (30 days) also affect the demyelination of afferents from alveus and the perforant pathway in the SO, SP and SR layers.
These assumptions are confirmed by phagocytosis of myelin and the time course of the detected morphologic changes in glial cells in the above layers of the hippocampus. In addition to the expected activation and increase in the number of astrocytes and microglia in these layers, which is consistent with numerous studies [52
], we observed specific morphologic changes in these cells in the SR on the 10th day after GCI, associated with the orientation and extension of their bodies parallel to myelinated fibers and perpendicular to the layer of pyramidal neurons.
A morphology similar to that which we observed in the SR 10 days after ischemia–an elongated rod-shaped nucleus and a body with short processes, is characteristic of rod-shaped microglia [55
]. Rod-shaped microglia have been described in the cortex in rats with diffuse brain injury [56
], in the hippocampus in older adults [58
] and present in experimental models of optic nerve degeneration and in some slowly developing neurodegenerative diseases progressing to dementia [55
]. Individual cells or chains from rod microglia are adjacent to neuronal processes. It was shown that postischemic changes in the SR layer of the hippocampus are limited to the region of synaptic terminals (clumping or dispersion of the synaptic vesicle pools and damage to synaptic membranes) and synaptic terminals are the primary and early target in the development of damage to the postischemic neurons [14
]. It is suggested that a close relationship with the processes of neurons reflects the participation of Rod microglia in synaptic stripping and neuronal circuitry reorganization [56
]. The study of microglia-synaptic interactions showed that there is not complete destruction of the synapses, but selective partial phagocytosis or trogocytosis, of presynaptic structures [60
]. The cell bodies of rod microglia, which we observed in the SR 10 days after GCI, spread along myelinated axons and made contact with them by their short processes. We suggest that together with the utilization of myelin and other cellular components, the rod microglia in this layer “disconnects” the excitatory inputs of pyramidal neurons from the afferents of the alvear and performant pathways to prevent neuronal loss due to glutamate excitotoxicity. With continued neuroinflammation on the 30th day after GCI, an accumulation of activated microglia, mainly with large round cell bodies and short processes, was observed in the SR and the phagocytosis occurs more actively. This corresponds to a further decrease in the MBP- positive area in the SR at Day 30 after GCI.
An interesting observation was the appearance of different types of contacts of rod-shaped microglia with myelinated axons including contacts by the bipolar ends of microglial bodies, by the microglial lateral processes and close grasping contact of microglial bodies with myelin fibers. Recent studies reveal that microglia and astrocytes play a major role in network formation during early development, in adulthood and neurodegeneration by directly pruning redundant synapses [61
]. Microglia are also able to act as a sensor of synaptic activity and play a key role in the homeostatic regulation of neural excitation [62
]. We assume that the bipolar contacts of rod-microglia bodies with myelinated fibers that we found, as well as the contacts of the lateral processes in the nodes of Ranvier, act as sensors for searching for synapses with excessive excitation and for pruning of such a contact. Similar pruning of superfluous excitatory synapses is shown during neurodevelopment [65
]. Verification of these assumptions, as well as the functional significance of the third type of contacts, which is a close grasping contact of microglial bodies with myelin fibers, is the subject of further research.
According to our results, in parallel with ongoing myelin and oligodendrocyte loss on the 30th day after ischemia, remyelination processes begin, as evidenced by an increase in the number of OPCs in the hippocampus. A significant increase in the number of OPCs in the SO, SP and SR layers of the hippocampus, the most subjected to demyelination after GCI, indicates the beginning of the regenerative process of remyelination. The number of OPCs in these layers significantly increases by the 10th day and multiplicatively increases on the 30th day after GCI. NG2+ glia is recognized as a separate glial cell population giving rise to oligodendrocytes [66
]. In response to damage, NG2+ glial cells are not only able to proliferate and migrate to lesions, but also differentiate into oligodendrocytes, forming new myelin sheaths that wrap around damaged axons and lead to their functional recovery [67
]. In addition, the results of genomic and epigenetic studies have shown that reactive NG2 glia can also differentiate into GFAP-labeled astrocytes and DCX-expressing immature neurons after traumatic injury [69
4. Materials and Methods
4.1. Animals and Housing
The study was performed on adult male Wistar rats weighing 250–300 g (n = 29; 15 rats survived after surgery) obtained from the vivarium of the E.D. Goldberg Institute of Pharmacology and Regenerative Medicine, Tomsk, Russia. Experiments were carried out in accordance with the rules adopted by the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes. The study was approved on 22 March 2012 by the Animal Care and Use Committee at the E.D. Goldberg Institute of Pharmacology and Regenerative Medicine (protocol #22032012). Animals were housed in groups of seven animals per cage with about 300 cm2 per animal under standard conditions (12/12-h light/dark cycle, temperature of 22 ± 2 °C, humidity of 60%). Standard rodent chow (PK-120-1, Laboratorsnab, Ltd., Moscow, Russia) and water were provided ad libitum.
4.2. Experimental Design
Animals were randomly divided into two groups: sham-operated animals (n
= 5) and animals, which underwent GCI (n
= 10). Acute global cerebral ischemia was induced according to the new three-vessel model [16
The surgery was performed as described previously [16
]. Briefly, rats were anesthetized with chloral hydrate in a dose of 450 mg/kg intraperitoneally (Sigma-Aldrich Chemical Co., St. Louis, MO, USA) and placed on a homeothermic blanket (Temperature Control Unit HB 101/2, Letica Scientific Instruments, Barcelona, Spain) in a supine position. The body temperature was maintained at 37 °C. Cerebral blood flow was completely interrupted for 7 min by ligation of the major arteries supplying the brain (truncus brachiocephalicus
, arteria subclavia sinistra
, and arteria carotis communis sinistra
). Access to arteria carotis communis sinistra
was implemented through the ventral surface of the neck, while truncus brachiocephalicus
and arteria subclavia sinistra
were reached through the first intercostal space, bypassing the pleural cavity to avoid pneumothorax. The animals were intubated through the oral cavity. The same surgery was performed in sham-operated rats, but without ligation of blood vessels.
At Days 11 and 31 after surgery, neurological deficit was evaluated with the Stroke-index McGraw scale [72
Animals were euthanized at Days 11 and 31 after surgery by transcardial perfusion with 4% paraformaldehyde under ether anesthesia. The brains were removed, fixed overnight with 4% paraformaldehyde solution, cryoprotected in sucrose phosphate buffer (24 h in 10% and 24 h in 20% solutions, respectively) at 4 °C, frozen in liquid nitrogen and stored at −80 °C for further immunohistochemical studies.
4.3. Immunochemistry and Microscopy
Coronal brain sections with 10-µm thickness were prepared using an HM525 cryostat (Thermo Fisher Scientific, Walldorf, Germany). Brain locations for immunohistochemical analysis were defined from −2.64 mm to −3.60 mm from bregma according to a rat brain atlas [74
The following primary antibodies were used:
rabbit-anti-NeuN (ABN78, Merck Millipore, Bedford, MA, USA) for detection of mature neurons;
rabbit-anti-GFAP (ab7260, Abcam, Cambridge, MA, USA) for detection of astrocytes;
rabbit-anti-Iba1 (019-19741, Wako Pure Chemical, Osaka, Japan) for detection of microglia/macrophages;
rabbit-anti-NG2 (AB5320, Merck Millipore, Bedford, MA, USA) for detection of oligodendrocyte precursors (OPCs);
mouse-anti-CNPase (MAB326, Merck Millipore, Bedford, MA, USA) for detection of myelinating oligodendrocytes;
goat-anti-MBP (C-16, sc-13914, Santa Cruz Biotechnology, CA, USA) for myelin detection.
Additionally, double immunofluorescence was used for investigation of interaction of microglia and astrocytes with myelinated fibers. To assess neurodegeneration Fluoro-Jade C (FJC) staining was performed by combining with NeuN immunofluorescence according to manufacturer’s protocol (Fluoro-Jade® C, Merck Millipore, Bedford, MA, USA). Primary antibodies were visualized by the following secondary antibodies: donkey-anti-goat AlexaFluor 594 (705–585–147, Jackson ImmunoResearch Laboratories, West Grove, PA, USA), donkey-anti-rabbit Alexa Flour 594 (711–585–152, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and donkey-anti-rabbit Alexa Fluor 488, (711–545–152, Jackson ImmunoResearch Laboratories, West Grove, PA, USA).
Brain sections were incubated with primary antibodies carried out overnight at 4 °C, with secondary antibodies–for 3 h at RT. Afterwards, stained brain sections were covered with antifade mounting medium Vectashield with DAPI (40,6-diamidino-2-phenylindole).
For each animal at least 4 microphotographs of the whole hippocampus of both (the left and right) hemispheres were obtained. Hippocampus general plans were photographed using an Axio Imager.Z2 (Carl Zeiss, Oberkochen, Germany) microscope (objective lens Plan-Apochromat 20×) and AxioVision 4.8 (Carl Zeiss) software with a MozaiX program module. 3D micrographs of separate groups of cells were obtained using Axio Imager Z1 (Carl Zeiss, Oberkochen, Germany) microscope (objective lens EC Plan-Neofluar 40× oil) with 3D ApoTome imaging system as a set of Z-stacks (45–50 sections per image), followed by the creation of a maximum intensity projection (MIP) image and 3D cell models.
4.4. Image Processing
Calculation of NeuN+, GFAP+, Iba1+, NG2+, CNP+ cells was carried out using ImageJ software (National Institutes of Health, Bethesda, MD, USA). To quantify the severity of the ischemic lesion, cell count change in CA1 field of the hippocampus was evaluated by visual calculation of cells in the stratum oriens (SO), stratum pyramidale (SP), substratum radiatum (SR), substratum lacunosum (SL) and substratum eumoleculare (SE). Double-labeled FJC+\NeuN+ cells were calculated in the SP layer. The cells were counted within regions of interest (ROIs) according to the scheme presented in Figure 7
. Three ROIs pre section, from 4 to 5 sections of the left and right hemispheres for each animal were analyzed (ROI: SO, SL, SE–500 × 250; SP–500 × 100; SR–500 × 500). Calculated number of cells per ROI were normalized to 1 mm2
Myelin content in the hippocampal layers was assessed in ROIs similar to those for calculation of NeuN+, GFAP+, Iba1+, NG2+, CNP+ cells in the layers SO, SP, SR, SL and SE near the CA1 field (Figure 6
). Assessment was performed using the Otsu thresholding method in the ImageJ (National Institutes of Health, Bethesda, MD, USA) implementation as a percent of MBP-positive area [35
4.5. Statistical Analysis
Statistical analysis was performed using the Statistica 10.0 software (StatSoft, Inc., Tulsa, OK, USA). Mean values and standard errors of mean (SEM) for each type of labeled cells were calculated in the hippocampal layers around the CA1 fields: SO, SP, SR, SL and SE. Quantitative histology data were compared between the control, 10-day ischemic and 30-day ischemic groups using a two-way repeated measures analysis of variance (ANOVA) followed by post hoc least significant difference (LSD) tests for individual layers of the hippocampus. The statistical significance for all the analyses was less than 0.05.