Erythropoietin Protects from Post-Traumatic Edema in the Rat Brain

Introduction

The pathophysiology of traumatic brain injury (TBI) is very complex—it induces excitotoxic damage, alterations in calcium homeostasis, free radical induced damage, inflammation, and mitochondrial dysfunction (Enriquez and Bullock, 2004). Various mediators are released, resulting in the occurrence of diffuse brain edema. The latter plays a major role in the worsening of the neurological status through the elevation of intracranial pressure and a concomitant impairment of brain perfusion (Unterberg et al, 2004). While numerous compounds have shown neuroprotective effects in experimental models of TBI, there is no pharmacological agent specifically aimed at blocking the progression of diffuse brain edema in the clinical setting, and therapy applied to patients with diffuse brain edema is still purely symptomatic (Narayan et al, 2002).

Among the agents with neurological tropism, erythropoietin (Epo) has been shown to confer neuroprotection and to reduce neuronal death in many in vivo and in vitro models of central nervous system (CNS) injuries (Bernaudin et al, 1999; Brines et al, 2000; Gorio et al, 2002; Grasso et al, 2002; Sakanaka et al, 1998; Siren et al, 2001). Also, improved neurological outcome and reduced lesion size were obtained in ischemic stroke patients treated with recombinant human Epo (rhEpo) (Ehrenreich et al, 2002). The mechanisms by which rhEpo exerts neuroprotective action are, however, incompletely understood. The overall expression of Epo receptors by neurons, astrocytes, and capillary endothelial cells elicited by hypoxia, ischemia (Bernaudin et al, 2000), or by nervous tissue injury (Grasso et al, 2005) supports the hypothesis of a well-organized Epo and Epo receptors network involved in the setting of an injury. While literature on the versatile properties of rhEpo in treating CNS disorders has flourished (Brines and Cerami, 2005; Grasso et al, 2004; Maiese et al, 2004), the effects of rhEpo after TBI have been sparsely studied. In a model of focal TBI induced by a blow delivered to the cortex, mice receiving rhEpo before or after TBI developed less necrosis at 10 days after injury than controls (Brines et al, 2000). In mice subjected to focal TBI, injection of rhEpo at 1 and 24 h after TBI induced a significant reduction of inflammation and of neuronal apoptosis, along with improved functional recovery (Yatsiv et al, 2005). Rats treated with rhEpo daily for 14 days after a focal TBI had an improvement in the performance of spatial memory tasks and increased neurogenesis in the dental gyrus (Lu et al, 2005). Since rhEpo could limit the production of tissue-injuring molecules (glutamate, radical oxygen species) (Morishita et al, 1997; Sakanaka et al, 1998) that are involved in the occurrence of diffuse brain edema, we hypothesized that rhEpo could reduce the development of post-traumatic brain edema. In a model of TBI resulting in early diffuse brain edema, rats receiving rhEpo after TBI were compared with non-treated TBI rats and to sham-operated rats. We used diffusion-weighted MR imaging to monitor the course of brain edema over 6 h after injury, and a gravimetric technique to measure brain water content (BWC) 6 h after the insult. We also tested T₁ mapping as a possible approach to quantify regional BWC after TBI.

Materials and methods

Magnetic Resonance Imaging Measurements

In a first series of experiments, MRI was performed at 7 T, in a horizontal bore magnet (Magnex Scientific Ltd, Oxford, UK) equipped with a 200 mT/m gradient system (12 cm inner bore diameter) and an SMIS console (SMIS Ltd, Guildford, UK). The rat was lying prone, its head secured via ear bars. A 79 mm internal diameter quadrature volume coil (Rapid Biomedical GmbH, Würzburg, Germany) was used for excitation and for detection. After radiofrequency coil matching and tuning, the magnetic field homogeneity was adjusted to obtain a half-height linewidth for water smaller than 0.3 part per million in the brain. A T₁-weighted axial scout image was acquired to locate slices of interest underneath the metallic disc area. Five adjacent coronal slices (1.5 mm thickness) were then acquired in a multiple slice mode for diffusion weighted imaging, and one slice (1.5 mm thickness) co-localized with the central coronal image for T₁ mapping.

Diffusion weighted images were acquired hourly from H₁ to H₆ using a T₂-weighted spin-echo sequence (repetition time = 2000 ms; echo time = 80 ms; field of view = 30 × 30 mm²; acquisition matrix = 128 × 66; 2 averages) (Figure 1). In addition to a first image acquired without diffusion weighting, three images were acquired in the presence of diffusion sensitizing gradients applied in each of the three main directions (x, y, z) of the gradient system, respectively. Diffusion sensitization was obtained by applying the diffusion gradients symmetrically about the inversion RF pulse. Parameters of the diffusion sensitization were the following: δ = 9 ms, Δ = 20 ms, b = 500 s/mm². Acquisition of the set of diffusion-weighted images lasted about 15 mins. T₁ maps were also obtained hourly from H₁ to H₆, using an inversion recovery FLASH sequence (repetition time = 10 s; echo time = 3.2 ms; field of view = 30 × 30 mm²; matrix = 64 × 64; 4 averages) with 12 inversion times (T_I) ranging between 100 and 6000 ms. The inversion pulse was specially nonselective to minimize in-flow contributions to the signal. Acquisition of each T₁ map lasted approximately 5 mins.

OPEN IN VIEWER

Figure 1

Representative T₂-weighted (left) and diffusion-weighted (right) images, with the two regions of interest delineated in each hemisphere: neocortex and caudoputamen.

Apparent diffusion coefficient (ADC) and T₁ maps were calculated on a Sunsparc workstation (Sun Microsystems Inc., Mountain View, CA, USA). The ADC maps were generated pixelwise by averaging the ADC values obtained for each of the three directions as follows:

A D C i = − ln ⁡ ( S i / S 0 ) / b i

(1)

with S₀ being the T₂-weighted image intensity obtained without diffusion weighting, S_i the image intensity obtained in the presence of the diffusion gradient in the spatial direction i, and b_i the gradient factor in that direction, taking into account the diffusion and the imaging gradients. The T₁ values were obtained by fitting the following model function to the data:

S ( T I ) = S 0 { 1 − a exp ⁡ ( − T I / T 1 ) }

(2)

with S₀, a and T₁ being the parameters estimated. Four regions of interest were manually drawn on the purely T₂-weighted image and copied onto to the ADC images and T₁ maps: right and left parietal neocortex, right and left caudoputamen.

Gravimetric Measurements

In a second series of experiments, BWC was determined 6 h after TBI (or equivalent) in three groups of rats: TBI-saline (n = 12), TBI-rhEpo (n = 12), and sham-operated (n = 12). After the completion of a similar experimental protocol, rats were killed by an intravenous injection of a KCl solution. The brain was rapidly removed and placed in a rat brain matrix previously cooled in ice. Two, 2-mm thick, coronal slices were cut through the impact lesion. In each hemisphere, samples (20 to 40 mg each) were taken from the neocortex (n = 4) and from the caudoputamen (n = 3) and subjected to gravimetric analysis. The gravimetric technique determines the tissue specific gravity (SG), a parameter that is known to correlate with BWC (Marmarou et al, 1982). The technique uses a layered kerosene–bromobenzene linear column. After calibration of the column with 7.5 μL drops of anhydrous K₂SO₄ solutions with known SGs (1.050, 1.045, 1.040, 1.035, 1.030, and 1.025), brain samples were deposited in the column, and their SG noted 2 mins later. Samples containing gas bubbles or blood were excluded from the analysis. The BWC was then derived according to the equation applied to grey matter (Fatouros and Marmarou, 1999):

H 2 O ( % ) = ( 462.6 / S G ) − 362.6

(3)

with H₂O being the BWC fraction and SG the mean value of the SG measurements obtained from the neocortex or the caudoputamen.

Results

Magnetic Resonance Imaging Experiments

Physiological data are shown in Table 1. Temperature (H₁), arterial pH (H₆), and PaO₂ (H₁ and H₆) decreased slightly over time in the three groups, with no intergroup differences. PaCO₂ remained unchanged throughout the experiments. Significant interaction between groups and serial measurements was found for mean arterial blood pressure (F = 3.6; P = 0.01) and for the hemoglobin content (F = 3.1; P = 0.02). This was due, respectively, to mean arterial blood pressure increase at H₆ in the TBI-rhEpo and in the sham-operated groups, and to hemoglobin decrease at H₆ in the TBI-rhEpo group. The two TBI groups (TBI-saline and TBI-rhEpo) had comparable data at any time of the experiment. No MRI evidence for intracerebral hemorrhage was found at the site of impact in these two groups.

OPEN IN VIEWER

Table 1

Physiological data in the three groups of rats having undergone the MRI experiments: TBI with saline (n = 12), TBI with rhEpo (n = 12), and sham-operated (n = 12)

Variables	Time	TBI-saline	TBI-rhEpo	Sham-operated
MABP (mm Hg)	H 0	117 ± 28	98 ± 10	106 ± 14
	H 1	107 ± 16	103 ± 16	118 ± 17
	H 6	111 ± 19	118 ± 15*	136 ± 21*, †
PaO2(mm Hg)	H 0	216 ± 59	211 ± 70	210 ± 46
	H 1	181 ± 35*	193 ± 37*	181 ± 28*
	H 6	183 ± 45*	177 ± 48*	174 ± 32*
PaCO2(mm Hg)	H 0	33 ± 5	35 ± 4	30 ± 2
	H 1	31 ± 7	34 ± 3	33 ± 6
	H 6	34 ± 4	35 ± 3	36 ± 4
Arterial pH	H 0	7.42 ± 0.05	7.43 ± 0.05	7.44 ± 0.04
	H 1	7.42 ± 0.07	7.44 ± 0.04	7.40 ± 0.06
	H 6	7.35 ± 0.06*	7.36 ± 0.03*	7.37 ± 0.03*
Hemoglobin (g/L)	H 0	130 ± 17	128 ± 11	117 ± 18
	H 1	115 ± 19	123 ± 13	119 ± 19
	H 6	118 ± 19	104 ± 21*	115 ± 13
Temperature (°C)	H 0	37.0 ± 0.4	36.6 ± 0.6	36.9 ± 0.2
	H 1	36.2 ± 0.8*	35.8 ± 1.0*	36.3 ± 0.6*
	H 6	36.7 ± 0.3	36.7 ± 0.3	36.6 ± 0.5

Data are expressed as mean ± s.d. H₀corresponds to the time of the TBI impact (or equivalent time in sham-operated rats). MABP, mean arterial blood pressure.

*

P< 0.05 versus H₀(Scheffé F-test);

†

P< 0.05 versus TBI-saline (Scheffé F-test)

Apparent diffusion coefficient measurements in the neocortex and in the caudoputamen were performed on all rats from the three groups. No significant ADC changes over time were found within each group of rats in the neocortex or in the caudoputamen (intragroup analysis). Apparent diffusion coefficient values in the neocortex were significantly lower in the TBI-saline group during the entire experiment: overall mean values of 0.61 × 10⁻³ ± 0.13 × 10⁻³ versus 0.73 × 10⁻³ ± 0.11 × 10⁻³ (TBI-rhEpo) and 0.71 × 10⁻³ ± 0.09 × 10⁻³ mm²/sec (sham-operated), respectively (P < 0.01). Similar differences were found for the ADC values in the caudoputamen: 0.60 × 10⁻³ ± 0.19 × 10⁻³ versus 0.74 × 10⁻³ ± 0.10 × 10⁻³ (TBI-rhEpo) and 0.79 ± 0.14 × 10⁻³ mm²/sec (sham-operated), respectively (P < 0.01). Intergroup analysis indicated significant differences between the TBI-saline group and the two other groups at the different measurement times of the experiment (Figures 2A and 2B). No differences in ADC were found between the TBI-rhEpo and the sham-operated groups.

OPEN IN VIEWER

Figure 2

Time course of ADC values in neocortex (A) and in caudoputamen (B) in the three groups of rats: TBI with saline (n = 12), TBI with rhEpo (n = 12), and sham-operated (n = 12) (MRI experiments). Traumatic brain injury impact was delivered on H₀ for TBI-saline and TBI-rhEpo groups. Data are expressed as mean ± s.d. *P < 0.05 TBI-saline versus TBI-rhEpo; ^†P < 0.05 TBI-saline versus sham-operated.

T₁ maps were obtained in 35 rats (one rat of the sham-operated group was excluded because of insufficient image quality). No significant T₁ changes over time were observed within each group of rats, in the neocortex as well as in the caudoputamen. T₁ values during the entire experiment were significantly longer in the TBI-saline group than in the TBI-rhEpo group in neocortex and in caudoputamen: 1821 ± 62 versus 1782 ± 85 and 1745 ± 119 versus 1676 ± 81 ms, respectively (P < 0.01) (Figure 3). No differences in T₁ values were found between the TBI-rhEpo group and the sham-operated group. However, there were minor differences between the three groups when compared at the different measurement times of the experiment (intergroup analysis) (data not shown).

OPEN IN VIEWER

Figure 3

Quantitative T₁ mean values in neocortex and in caudoputamen during the entire experiment in the three groups of rats: TBI with saline (n = 12), TBI with rhEpo (n = 12), and sham-operated (n = 11) (MRI experiments). Data are expressed as mean ± s.d. *P < 0.05 TBI-saline versus TBI-rhEpo; ^†P < 0.05 TBI-saline versus sham-operated.

Gravimetric Experiments

Specific gravity measurements were performed at H₆ on 36 rats, 12 in the TBI-saline group, 12 in the TBI-rhEpo group, and 12 in the sham-operated group. No significant differences in the physiological parameters of these three groups were found between the two series of experiments, MRI versus gravimetric (data not shown). Brain water content measurements in neocortex and in caudoputamen were significantly lower in the TBI-rhEpo group than in the TBI-saline group: 78.2% ± 0.3% versus 79.1% ± 0.4% and 78.0% ± 0.3% versus 78.6% ± 0.5%, respectively (P < 0.01) (Figure 4). No differences in BWC were found between the TBI-rhEpo group and the sham-operated group.

OPEN IN VIEWER

Figure 4

Brain water content in neocortex and in caudoputamen at H₆ in three groups of rats: TBI with saline (n = 12), TBI with rhEpo (n = 12), and sham-operated (n = 12) (gravimetric experiments). Traumatic brain injury impact was delivered on H₀ for TBI-saline and TBI-rhEpo groups. Data are expressed as mean ± s.d., *P < 0.05 TBI-saline versus TBI-rhEpo. ^†P < 0.05 TBI-saline versus sham-operated.

Discussion

This study shows that the administration, postinjury, of rhEpo can significantly alter the development of TBI-induced brain edema in rats. The effect of rhEpo is early and lasts at least 6 h, as shown by serial MR measurements of the ADC and of tissue T₁, and by BWC measurements at H₆. These parameters show consistent results in favor of a strong action of post-TBI administration of rhEpo against TBI-induced brain edema.

To investigate the effects of rhEpo following diffuse TBI, we used the rodent impact-acceleration model (Marmarou et al, 1994). In pilot experiments on spontaneously breathing rats (n = 42), we observed, after TBI, a rate of skull fractures of 5% and a mortality rate of 45% at animal awakening, in good agreement with the initial report (Marmarou et al, 1994). To prevent mortality, we maintained animals under anesthesia and mechanical ventilation during the course of the experiments. Impact-acceleration model does not induce subdural- and intraparenchymal hemorrhages, in contrast to the focal TBI models such as the fluid percussion model and the controlled cortical impact model. In addition, early diffuse injury of neurons, axons and microvasculature, astrocytic swelling in cortex, and massive diffuse axonal swelling in corpus callosum, thalamus, internal capsule, and brain stem were reported using impact-acceleration model (Foda and Marmarou, 1994). Diffuse brain edema, elevated intracranial pressure and brain hypoperfusion were reported as early as the first hour postinjury in anesthetized and ventilated rats (Barzo et al, 1997; Ito et al, 1996). Given these observations, we performed a longitudinal study of TBI-induced brain edema starting 1 h after trauma, with the neocortex and the caudoputamen as the sites of edema measurement.

We used diffusion-weighted MR imaging and T₁ mapping over 6 h after TBI. Diffusion-weighted imaging is a well-documented technique to study changes in ADC values in various CNS insults and to indicate the nature of brain edema (Huisman, 2003). The reduction in brain ADC is thought to represent mainly cellular (cytotoxic) edema, while an ADC increase indicates mainly vasogenic edema resulting from blood–brain barrier (BBB) disruption. In the fluid percussion model (focal TBI), a significant decrease in tissue ADC was reported 1 to 2 h post-trauma (Albensi et al, 2000; Assaf et al, 1999). In the rodent impact-acceleration model (diffuse TBI), ADC measurements were less conclusive. An ADC increase up to 4 h post-TBI was reported along with an increased BBB permeability (Cernak et al, 2004). Others found no significant post-TBI ADC changes in normotensive and normo-ventilated rats (Ito et al, 1996). A biphasic response to injury was found with a transient ADC increase lasting less than 1 h after TBI (vasogenic edema), followed immediately by a prolonged and marked reduction in brain ADC (cellular edema) (Barzo et al, 1997). Although we cannot exclude a transient post-TBI increase (< 1 h) in the permeability of the BBB, the lower ADC measurements found from H₁ on in the TBI-saline group compared with the sham-operated group are in line with the latter findings, suggesting cellular predominance of the post-TBI edema. Our ADC measurements thus indicate that the postinjury treatment with rhEpo can significantly reduce the cellular contribution to brain edema. Postinjury rhEpo treatment did not affect ADC of brain water in rats subjected to a stereotactic resection of the frontal lobe (Matchett et al, 2006). Whether the lack of response is due to the vasogenic nature of this model with BBB breakdown or to the limited area of damaged tissue is unclear.

To our knowledge, this study is the first to use T₁ mapping to monitor the evolution of post-traumatic brain edema. T₁ mapping has been successfully used to quantify BWC, and a linear relationship was obtained between R₁ (= 1/T₁) and 1/BWC following stroke (Fatouros et al, 1991; Lin et al, 2000). In a model of focal brain ischemia in rat, a significant increase in ischemic tissue T₁ was reported, together with an ipsilateral decrease in brain ADC and BWC increase as measured with the gravimetric technique (Barbier et al, 2005). Our results are consistent with those findings, showing an increase in the tissue T₁ in the TBI-saline group. They are also in line with the ADC results by showing a reduction in the tissue T₁ values in rhEpo-treated TBI rats. This indicates a possible benefit of rhEpo on the occurrence of brain edema. The T₁ results were however less clearcut than those of the ADC measurements (Figure 3). This could be due to brain water accumulation being more reduced after diffuse TBI than after occlusion of a terminal brain artery in rats: 0.5% to 1% in this TBI model (Cernak et al, 2004) versus 2% to 4% in models of ischemia (Barbier et al, 2005; Lin et al, 2000). Similarly, moderate brain water accumulation was reported in diffusely head-injured patients presenting brain swelling and concomitant elevated intracranial pressure (Marmarou et al, 2006).

We found strong differences in the ADC and BWC values between rhEpo-treated and saline-treated TBI rats, providing evidence for rhEpo affecting brain water accumulation after TBI. In both experiments, physiological parameters of the two groups of TBI rats (rhEpo versus saline) were comparable and maintained within normal ranges. The 0.5% to 1% rhEpo-induced reduction in BWC at H₆ may significantly improve brain perfusion in closed head injury with low compliance (Frei et al, 1973). Although we could not address whether the drug had a persistent effect, these results are in line with those obtained in rhEpo-treated mice showing attenuation of cell apoptosis and tissue inflammation, and improvement of neurological outcome 3 to 14 days after focal TBI (Yatsiv et al, 2005). In the present study, ADC differences between treated and non-treated groups were found as early as the first hour post-TBI. Actually, rhEpo can rapidly cross the BBB since a significant increase in the concentration of Epo in cerebrospinal fluid has been found as soon as 30 mins after intraperitoneal administration of rhEpo at 5000 UI/kg (Brines et al, 2000). A peak serum level of 2000 ng/mL was obtained 15 mins after intravenous administration of a similar dose of rhEpo (Gorio et al, 2005). Despite that the exact mechanisms by which rhEpo mediates its effects across the BBB are uncertain, these findings might account for immediate changes in biochemical pathways involved in injured brain tissue after intravenous injection of rhEpo.

At present, we can only formulate conjectures to explain our main findings. While apoptotic and necrotic pathways after TBI and antiapoptotic effects of rhEpo in models of ischemia and of TBI have been well identified (Raghupathi, 2004; Siren et al, 2001; Yatsiv et al, 2005), such mechanisms need several hours after the insult to be effective. Similarly, neuroinflammatory response is another target for rhEpo (Brines et al, 2000; Yatsiv et al, 2005). Tissue accumulation of leukocytes, activation of glial cells, and/or cytokine production equally occur hours after TBI (Enriquez and Bullock, 2004). We should thus consider processes occurring within minutes after TBI, such as the release, by injured cells, of glutamate, lactate, proton, potassium, calcium, nitric oxide (NO), arachidonic acid, free oxygen radicals, histamine, and kinins (Unterberg et al, 2004). In response to the early release of glutamate into the extracellular space, glutamate uptake by astrocytes is enhanced through sodium-dependant transporters, leading to astrocyte swelling and intracellular calcium accumulation (Chen and Swanson, 2003). This ionic imbalance with a dramatic intracellular influx of sodium, calcium and/or, chloride may explain early glial swelling after injury (Kimelberg, 2004). Rapid and massive intracellular accumulation of free calcium was found within minutes after TBI (Nilsson et al, 1993). Therefore, the biological basis of an immediate beneficial effect of rhEpo could involve modulation of the glutamate release, restoration of ionic imbalance, and/or alteration in free oxygen radicals production, including NO. In cultured neurons subjected to chemical ischemia, activation of Epo receptors inhibited the exocytosis of glutamate (Kawakami et al, 2001). Pretreated neurons with rhEpo prevented glutamate-induced neuronal damage (Bernaudin et al, 1999; Morishita et al, 1997). In animals subjected to brain ischemia, post-treatment with rhEpo significantly reduced the NO toxicity to neurons, inhibited its tissue production, and reduced the formation of brain edema (Calapai et al, 2000; Sakanaka et al, 1998). Binding of Epo to its neuronal receptor triggers the activation of several Epo signal-regulated kinases and protein kinases (Jak2 and NF-κB), which in turn modulates damage because of the generation of glutamate, free radicals, and NO in cultured neurons (Digicaylioglu and Lipton, 2001). We are investigating whether the post-TBI administration of rhEpo results in an early activation (< 1 h after treatment) of such molecular pathways. There are yet concordant results to support an early benefit of post-traumatic treatment with rhEpo on the formation of brain edema.

Epoietin alfa was administered at a dose of 5000 UI/kg because this dose had been proven effective in various models of CNS injuries (Brines et al, 2000; Gorio et al, 2002; Siren et al, 2001; Yatsiv et al, 2005). Although a dose–response study of rhEpo has not yet been conducted in TBI, benefit increased with the dose of rhEpo in models of spinal cord injury (Gorio et al, 2002). The rhEpo dose used here is already higher than that currently recommended in the clinical setting, and concerns may be raised about the safety of rhEpo when administered for neuroprotection (Grasso et al, 2004). The development of fully neuroprotective analogs of rhEpo without any erythropoietic action should thus be of particular interest (Leist et al, 2004). If such analogs of rhEpo also appear beneficial in treating post-traumatic edema, their optimal therapeutic window and dose will need to be determined before using them in the clinical setting.

In conclusion, our findings add to the accumulating evidence of beneficial effects of Epo in various CNS insults by showing a significant reduction in brain edema after experimental TBI. Given the current lack of a specific pharmacological treatment aimed at blocking the development of brain edema in head-injured patients, these results offer new perspectives in the treatment of TBI and should prompt mechanistic as well as outcome studies on the potential use of rhEpo in TBI-induced brain edema.