DEMA found in the perihematoma zone is one ofthe most important factors causing secondary braininjury after intracerebral hemorrhage (ICH). The

mechanisms involved in brain edema formation follow-ing ICH still have to be fully elucidated. Investigationshave indicated that mass effect, ischemia, neurotoxicity,and blood-brain barrier disruption may all be involved inedema formation.19,30,35 Recent studies have indicated thatclot retraction and plasma proteins induce brain edema asearly as one hour after ICH.29,30 Thrombin and the coagu-lation cascade also appear to play a major role in early(24-hour) edema formation following ICH, with edemaformation being markedly reduced by thrombin inhibi-tors.10–13 In contrast, injection of red blood cells (RBCs)into brain fails to induce edema formation by 24 hours.12,34

However, the possible role of RBCs in the induction ofbrain edema needs to be evaluated carefully. There is evi-dence that hemoglobin can cause brain injury. Reportshave shown that hemoglobin can inhibit Na+/K+ adenosinetriphosphatase (ATPase) activity,24 generate the toxic hy-

droxyl radical,20 stimulate peroxidation of central nervoussystem lipids,5 and cause neuronal death.8,22

In this study, we have reexamined the roles of RBCsand hemoglobin in brain edema formation following ICHby examining the effects of autologous packed RBCs,lysed RBCs, and rat hemoglobin on brain water contentand by comparing the time course of edema forma-tion after thrombin injection to that found previouslywith ICH.

Materials and MethodsAnimal Preparation

The protocol for these animal studies was approved by the Uni-versity Committee on the Use and Care of Animals. One hundredseventeen adult male Sprague–Dawley rats, each weighing between300 and 450 g, were allowed food and water ad libitum before andafter surgery. The rats were anesthetized with an intraperitoneal in-jection of pentobarbital (40 mg/kg). After anesthesia was achieved,a polyethylene catheter (PE-50) was inserted into the right femoral

J. Neurosurg. / Volume 89 / December, 1998

J Neurosurg 89:991–996, 1998

Erythrocytes and delayed brain edema formation followingintracerebral hemorrhage in rats

GUOHUA XI, M.D., RICHARD F. KEEP, PH.D., AND JULIAN T. HOFF, M.D.Department of Surgery (Neurosurgery), University of Michigan, Ann Arbor, Michigan

Object. The mechanisms of brain edema formation following spontaneous intracerebral hemorrhage (ICH) arenot well understood. In previous studies, no significant edema formation has been found 24 hours after infusion ofpacked red blood cells (RBCs) into the brain of a rat or pig; however, there is evidence that hemoglobin can be neu-rotoxic. In this study, the authors reexamined the role of RBCs and hemoglobin in edema formation after ICH.

Methods. The experiments involved infusion of whole blood, packed RBCs, lysed RBCs, rat hemoglobin, orthrombin into the right basal ganglia of Sprague–Dawley rats. The animals were killed at different time points andbrain water and ion contents were measured. The results showed that lysed autologous erythrocytes, but not packederythrocytes, produced marked brain edema 24 hours after infusion and that this edema formation could be mim-icked by hemoglobin infusion. Although infusion of packed RBCs did not produce dramatic brain edema during thefirst 2 days, it did induce a marked increase in brain water content 3 days postinfusion. Edema formation followingthrombin infusion peaked at 24 to 48 hours. This is earlier than the peak in edema formation that follows ICH, sug-gesting that there is a delayed, nonthrombin-mediated, edemogenic component of ICH.

Conclusions. These results demonstrate that RBCs play a potentially important role in delayed edema develop-ment after ICH and that RBC lysis and hemoglobin toxicity may be useful targets for therapeutic intervention.

KEY WORDS • cerebral hemorrhage • cerebral edema • erythrocyte • hemoglobin •thrombin • rat

E

991

artery for continuous blood pressure monitoring. Arterial blood wasobtained for analysis of blood gas levels, arterial pH, hematocritvalue, blood glucose concentration, and as a source of erythrocytesfor intracerebral injection. Body temperature was maintained at37.5˚C by using a feedback-controlled heating pad.

Intracerebral InfusionBefore intracerebral infusion, the rat was positioned in a stereo-

tactic frame and the scalp was incised along the sagittal midline byusing a sterile technique. A cranial burr hole (1 mm) was drillednear the right coronal suture 4 mm lateral to the midline. A 26-gauge needle was inserted stereotactically into the right basal gan-glia (coordinates: 0.2 mm anterior, 5.5 mm ventral, and 4 mm later-al to the bregma). Whole blood, packed RBCs, lysed RBCs, or othersolutions were infused using a microinfusion pump into the rightbasal ganglia. After infusion, the needle was removed and the skinincisions were closed with sutures. The animal was allowed torecover.

Experimental GroupsThis study was divided into five parts. In the first four parts, the

roles of whole blood, lysed RBCs, hemoglobin, and packed eryth-rocytes in brain edema formation were assessed. In the fifth part, weexamined the time course of brain edema formation following infu-sion of 5 U rat thrombin.

Part 1. In this portion of the study five groups of rats (five ani-mals in each group) were used. Whole autologous blood (100 �l)was injected into the right caudate nucleus of each rat. The animalswere killed at 1, 2, 3, 5, or 7 days.

Part 2. Three groups (five rats each) were used in this portion ofthe experiment. The first group underwent a sham operation thatconsisted of needle insertion without infusion. The rats in the sec-ond group were each infused with 30 �l of autologous packedRBCs. Packed RBCs (hematocrit level 87 � 1%) were obtained bycentrifuging unclotted blood. The plasma and buffy coat were dis-carded. The RBCs were washed with five volumes of saline threetimes. A 30-�l infusion of lysed autologous RBCs was used for therats in the third group. The washed RBCs were lysed by freezing thecells in liquid nitrogen for 5 minutes followed by thawing at 37˚C.The rats were killed 24 hours after RBC infusion.

Part 3. Three groups of animals were used in this part of thestudy. The first group (six rats) received 20 �l saline. The secondand third groups (six rats each) received 20 �l rat hemoglobin at aconcentration of either 150 mg/ml or 300 mg/ml. The normal con-centration of hemoglobin in RBCs is approximately 300 mg/ml.Because native hemoglobin is readily oxidized in air, this hemoglo-bin is predominantly methemoglobin, as stipulated by the manufac-turer in its 1997 catalogue. All the animals were decapitated 24hours after infusion.

Part 4. In five groups of rats (five or six animals in each group),50 �l of packed RBCs were injected into the right basal ganglia.

The animals were decapitated at 1, 2, 3, 5, or 7 days to assess edemaformation.

Part 5. In this portion of the study, six groups of rats (five or sixrats each) were studied. Each rat received an infusion of 60 �l of ratthrombin (5 U). The rats were killed at 4, 12, 24, 48, 72, or 168hours.

Brain Water, Sodium, and Potassium ContentsAfter deep pentobarbital anesthesia (80 mg/kg injected peri-

toneally) had been induced, the rats were killed by decapitation. Thebrains were removed immediately and, 4 mm from the frontal pole,a 3-mm-thick coronal brain slice was cut. The slice was divided intofour samples, ipsilateral and contralateral basal ganglia and ipsilat-eral and contralateral cortex. Tissue samples were weighed using anelectronic analytical balance calculated to the nearest 0.1 mg toobtain the wet weight (WW). The tissue was dried in a gravityoven set at 100˚C for more than 24 hours to determine the dryweight (DW). Tissue water contents (%) were calculated as follows:(WW � DW)/WW � 100.

The dehydrated brain samples were digested in 1 ml of 1 N nitricacid for 1 week. The sodium and potassium ion contents in thissolution were measured using flame photometry. Ion contents wereexpressed in microequivalents per gram of dehydrated brain tissue(�Eq/g DW).

Statistical AnalysisData from different animal groups were analyzed using analysis

of variance with a Scheffé F-test post hoc test, except for the re-sponse to different doses of hemoglobin, in which case a Dunnett’spost hoc test was used. Differences were considered significant atprobability values less than 0.05. All values in the text are given asthe mean � standard error of the mean (SEM).

Sources of Supplies and EquipmentThe animals, which were obtained from Charles River Labor-

atories (Portage, MI), were positioned in a stereotactic framepurchased from Kopf Instruments (Tujunga, CA). The microinfu-sion pump used in the experiments was manufactured by Har-vard Apparatus, Inc. (South Natick, MA). The rat hemoglobin andthrombin were obtained from Sigma Chemical Co. (St. Louis, MO).The electronic analytical balance (model AE 100) used to weigh tis-sue samples was obtained from Mettler Instrument Co. (Highstown,NJ) and the flame photometry system (model IL 943) from Instru-mentation Laboratory, Inc. (Lexington, MA).

ResultsPhysiological parameters were recorded just before in-

fusion of the whole blood, RBCs, hemoglobin, thrombin,

G. Xi, R. F. Keep, and J. T. Hoff

992 J. Neurosurg. / Volume 89 / December, 1998

TABLE 1Physiological parameters in 117 rats before intracerebral infusion*

Experiment Parts

Variable 1 2 3 4 5

MABP (mm Hg) 112 � 10 110 � 12 107 � 8 108 � 11 109 � 7pH 7.43 � 0.02 7.43 � 0.04 7.43 � 0.02 7.43 � 0.02 7.44 � 0.02PaO2 (mm Hg) 80.0 � 4.4 82.1 � 3.4 81.9 � 4.7 82.1 � 5.1 85.3 � 6.5PaCO2 (mm Hg) 43.2 � 4.5 45.8 � 3.8 45.7 � 3.0 46.4 � 3.0 46.3 � 3.2hematocrit (%) 41.4 � 2.6 41.2 � 1.4 41.1 � 1.2 40.9 � 1.0 40.1 � 1.5glucose (mg%) 122.2 � 20 134 � 20 136 � 13 131 � 16 127 � 13

* Values are expressed as the mean � standard deviation. Abbreviation: MABP = mean arterial blood pressure.

or saline (Table 1). There were no significant differencesamong groups used in each part of the experiment. Themean values for levels of blood gases, blood pH, meanarterial blood pressure, hematocrit, and blood glucosewere within normal ranges.

Time Course of Edema Formation Following Whole BloodInfusion

Figure 1A shows the time course of brain water contentat 1, 2, 3, 5, and 7 days after a 100-�l whole blood infu-sion. Although marked brain edema was observed on the1st day after whole blood infusion, the peak increase inbrain water content in the ipsilateral cortex and basal gan-glia was at 3 days.

Brain Water Content Following Lysed and Packed RBCInfusion

Lysed RBCs, but not packed RBCs, produced markedbrain edema in the ipsilateral basal ganglia and cortex 24hours after infusion (Fig. 2). Brain water content in thelysed RBC group was increased significantly in the ip-silateral basal ganglia compared with the packed RBCgroup (83.2 � 0.5% and 78.4 � 0.3%, respectively). Theedema formation after lysed RBC infusion was associatedwith an ipsilateral accumulation of sodium and loss of po-tassium (Fig. 3).

Brain Water Content Following Hemoglobin InfusionIntracerebral infusion of rat hemoglobin produced a

dose-dependent increase in brain water content at 24 hours(Fig. 4). The increase in brain water content was associat-ed with an accumulation of brain sodium and loss of brainpotassium (Table 2).

J. Neurosurg. / Volume 89 / December, 1998

Erythrocytes and delayed edema after intracerebral hemorrhage

993

FIG. 1. Graphs showing brain water content at 1, 2, 3, 5, and 7days after infusion of whole blood or packed RBCs. A: Afterinfusion of 100 �l whole blood. B: After infusion of 50 �lpacked RBCs. Values are expressed as the mean � SEM in five orsix rats. #p � 0.01.

FIG. 2. Bar graph showing brain water content 24 hours aftersham operation or infusion of 30 �l of either packed or lysedRBCs. Values are expressed as the mean � SEM in five rats. *p �0.05 compared with the sham-operated and packed RBC groups.#p � 0.01 compared with the sham-operated and packed RBCgroups.

FIG. 3. Bar graph depicting brain sodium (A) and potassium ion(B) contents 24 hours after sham operation or a 30-�l infusion ofeither packed or lysed RBCs. Values are expressed as the mean �SEM. *p � 0.05 compared with the sham-operated and packedRBC groups. #p � 0.01 compared with the sham-operated andpacked RBC groups.

Time Course of Edema Formation Following Packed RBCInfusion

Figure 1B shows the time course of brain water contentat 1, 2, 3, 5, and 7 days after injection of 50 �l packedRBCs. On the 1st day after infusion, no marked brainedema was observed. There was a peak increase in brainwater content in the ipsilateral cortex and basal ganglia 3days after RBC infusion. At 5 to 7 days, ipsilateral brainwater content returned to the level of the contralateral tis-sue. Figure 5 shows the time course of sodium accumula-tion (Fig. 5A) and potassium loss (Fig. 5B) after infusionof packed RBCs. Again, the peak change in both sodiumand potassium was on the 3rd day.

Time Course of Edema Formation Following ThrombinInfusion

Figure 6 shows the time course of brain water contentafter infusion of rat thrombin (5 U). Edema formationstarted as early as 4 hours after thrombin infusion. Thepeak in edema was at 24 to 48 hours. Although water con-tent began to decrease after 48 hours, at 7 days it was stillhigher than that on the contralateral side.

DiscussionThis study demonstrates that intracerebral infusion of

lysed RBCs results in marked brain edema formation by24 hours. This edema formation appears to be mediated byhemoglobin because intracerebral infusion of rat hemo-globin at concentrations found in RBCs also resulted inmarked increases in brain water content. Studies also indi-cate that hemoglobin may have other deleterious effectson the brain. Intracortical hemoglobin injection in ratsproduced chronic focal spike activity, cavity lesions, andgliosis at injection sites.23 Hemoglobin inhibits Na+/K+

ATPase activity in brain homogenates24 and induces de-polarization in hippocampal CA1 neurons.36 Regan andPanter22 found brief exposures (1–2 hours) to hemoglobinwere not toxic, but exposure of neuronal cell cultures tohemoglobin for 1 day produced concentration-dependentneuronal death.

Although we found that lysed RBCs and hemoglobininduced edema formation, this did not necessarily implythat the RBCs in an intracerebral hematoma contributeto edema formation. In an intracerebral hematoma, RBClysis and hemoglobin release might occur gradually overa period of time and the brain hemoglobin concentrationsmight not reach toxic levels. Therefore, we examined thetime course of brain water content following injectionof packed RBCs into the caudate nucleus. As previously,we found no edema formation 24 hours postinjection.12,34

However, there was marked edema after 3 days, whichresolved by 5 days. This delayed edema formation is inaccordance with studies in which delayed RBC lysis and

G. Xi, R. F. Keep, and J. T. Hoff

994 J. Neurosurg. / Volume 89 / December, 1998

FIG. 4. Bar graph depicting water content at 24 hours after infu-sion of 20 �l hemoglobin (Hb) or saline. Values are expressed asthe mean � SEM. *p � 0.05 compared with 150-mg/ml hemoglo-bin group. #p � 0.01 compared with the saline group.

TABLE 2Brain tissue sodium and potassium ion contents 24 hours after a

20-µl hemoglobin or saline infusion*

Sodium Ion Potassium IonGroup & Region (�Eq/g DW) (�Eq/g DW)

ipsilat hemispheresaline 186 � 2 425 � 4hemoglobin (150 mg/ml) 240 � 6 423 � 3hemoglobin (300 mg/ml) 362 � 46† 339 � 29†

contralat hemispheresaline 177 � 2 432 � 2hemoglobin (150 mg/ml) 188 � 2 446 � 4hemoglobin (300 mg/ml) 183 � 3 444 � 7

* Values are expressed as the mean � SEM in six rats.† p � 0.05 compared with the saline and hemoglobin (150 mg/ml)

groups.

FIG. 5. Graphs showing brain sodium (A) and potassium (B) ioncontents 1, 2, 3, 5, and 7 days after infusion of 50 �l packed RBCs.Values are expressed as the mean � SEM in five or six rats. *p �0.05 compared with the contralateral side. #p � 0.01 comparedwith the contralateral side.

hemoglobin release have been found after cerebral hem-orrhage. Hemoglobin concentrations reach their peak onthe 2nd day after blood injection into the subarachnoidspace in the dog and then gradually disappear.15 Hemo-globin release from lysis of RBCs in human intracranialhemorrhage increases during the first few days.31 By usinghistochemical methods, hemoglobin and heme can be ob-served in the perihematoma zone 24 hours after wholeblood injection in the rabbit.9 The reason for this delayedRBC lysis appears to be either depletion of intracellularenergy reserves7 or activation of a complement system andformation of membrane attack complex.21

A comparison of edema formation produced by ICHand thrombin infusion also suggests there may be delayededema formation from RBC lysis and hemoglobin releasein ICH. Data from this laboratory35 and from others4,18,26,27

indicate the amount of edema found in the perihematomazone after ICH reaches a peak between Day 3 and Day7. Our present results also show that the peak of edemaformation after whole blood infusion is on Day 3 (Fig.1A). In contrast, edema formation after thrombin injec-tion peaks at 1 to 2 days. Delayed edema formation fromRBCs, peaking at 3 days, might explain the differencebetween the ICH and thrombin data.

The deleterious effects of hemoglobin may derive fromthe hemoglobin itself or its breakdown products. For ex-ample, Gutteridge5 found that hemoglobin activates lipidperoxidation through two different phases. The first phaseis induced by the hemoglobin itself, which can be inhibit-ed by haptoglobin. Haptoglobins are glycoproteins thatform stable complexes with hemoglobin. The secondphase is stimulated by iron, one of the hemoglobin break-down products, and is inhibited by the iron chelator des-ferrioxamine, and an iron-binding protein, transferrin.

The adverse effects of hemoglobin vary with its chemi-cal form. Oxyhemoglobin is a spasminogen that has beenimplicated in cerebral vasospasm.14 In ICH, however,Bradley2 only found oxyhemoglobin in the hematoma forthe first few hours following hemorrhage. Thus, it is un-likely that oxyhemoglobin plays a role in ICH-inducededema formation, a suggestion supported by the fact thatICH does not produce marked reductions in cerebralblood flow in the rat35 and intracerebral infusion of essen-tially methemoglobin can mimic the effects of RBCs onedema formation.

The heme from hemoglobin is broken down by hemeoxygenase in the brain into iron, carbon monoxide, andbiliverdin.16,17 Carbon monoxide is a free radical that maycause tissue damage analogous to nitric oxide–mediateddamage.28 Iron can also stimulate the formation of freeradicals leading to neuronal damage. Ferrous and ferriciron are able to react with lipid hydroperoxides to producealkoxy and peroxy radicals and cause brain damage.25

Cortical injection of iron causes focal epileptiform par-oxysmal discharges6 and neuronal damage.33 Andersonand Means1 found iron salts inhibit spinal cord Na+/K+

ATPase in vivo, an effect blocked by antioxidants, where-as Willmore and Rubin32 found that subpial FeCl2 injectioninduces focal brain edema and malonaldehyde formation.Roles for iron, free radicals, and lipid peroxidation in he-moglobin-induced brain injury are supported by findingsthat the inhibition of brain Na+/K+ ATPase activity byhemoglobin can be blocked by desferrioxamine mesylate,

an iron chelator,24 whereas hemoglobin-induced toxicityin neuronal cell cultures is blocked by the 21-aminosteroidU74500A, the antioxidant Trolox, and desferrioxamine.22

In conclusion, edema formation after ICH appears toinvolve several phases. These include a very early phase(first several hours) involving hydrostatic pressure andclot retraction,3,30 a second phase (1st day) involving theclotting cascade and thrombin production,10–12 and a thirdphase (approximately Day 3 in the rat) involving RBClysis and hemoglobin-induced toxicity. Because of thedelay in onset, this third phase may be more amenable totherapeutic intervention either by altering RBC lysis orlimiting hemoglobin-induced toxicity.

References

1. Anderson DK, Means ED: Lipid peroxidation in spinal cord:FeCl2 induction and protection with antioxidants. NeurochemPathol 1:249–264, 1983

2. Bradley WG Jr: MR appearance of hemorrhage in the brain.Radiology 189:15–26, 1993

3. Brott T, Broderick J, Barsan W, et al: Hyper-acute clot retrac-tion in spontaneous intracerebral hemorrhage. Stroke 23:141,1992 (Abstract)

4. Enzmann DR, Britt RH, Lyons BE, et al: Natural history ofexperimental intracerebral hemorrhage: sonography, computedtomography and neuropathology. AJNR 2:517–526, 1981

5. Gutteridge JMC: The antioxidant activity of haptoglobin to-wards haemoglobin-stimulated lipid peroxidation. BiochimBiophys Acta 917:219–223, 1987

6. Hammond EJ, Ramsay RE, Villarreal HJ, et al: Effects of intra-cortical injection of blood and blood components on the elec-trocorticogram. Epilepsia 21:3–14, 1980

7. Kase CS, Caplan LR: Intracerebral Hemorrhage. Boston:Butterworth-Heinemann, 1994

8. Koenig ML, Meyerhoff JL: Neurotoxicity resulting from pro-longed exposure to hemoglobin. Soc Neurosci Abstr 23:1935,1997 (Abstract)

9. Koeppen AH, Dickson AC, McEvoy JA: The cellular reactionsto experimental intracerebral hemorrhage. J Neurol Sci 134(Suppl):102–112, 1995

10. Lee KR, Betz AL, Keep RF, et al: Intracerebral infusion ofthrombin as a cause of brain edema. J Neurosurg 83:1045–1050, 1995

11. Lee KR, Betz AL, Kim S, et al: The role of the coagulation cas-

J. Neurosurg. / Volume 89 / December, 1998

Erythrocytes and delayed edema after intracerebral hemorrhage

995

FIG. 6. Graph displaying water content 4, 12, 24, 48, 72, and168 hours after infusion of rat thrombin (5 U). Values are ex-pressed as the mean � SEM in five or six rats. *p � 0.001 com-pared with the contralateral side.

cade in brain edema formation after intracerebral hemorrhage.Acta Neurochir 138:396–401, 1996

12. Lee KR, Colon GP, Betz AL, et al: Edema from intracerebralhemorrhage: the role of thrombin. J Neurosurg 84:91–96,1996

13. Lee KR, Kawai N, Kim S, et al: Mechanisms of edema forma-tion after intracerebral hemorrhage: effects of thrombin on cere-bral blood flow, blood-brain barrier permeability, and cell sur-vival in a rat model. J Neurosurg 86:272–278, 1997

14. Macdonald RL, Weir BKA: A review of hemoglobin and thepathogenesis of cerebral vasospasm. Stroke 22:971–982, 1991

15. Marlet JM, Barreto Fonseca JP: Experimental determination oftime of intracranial hemorrhage by spectrophotometric analysisof cerebrospinal fluid. J Forensic Sci 27:880–888, 1982

16. Matz P, Turner C, Weinstein PR, et al: Heme-oxygenase-1induction in glia throughout rat brain following experimentalsubarachnoid hemorrhage. Brain Res 713:211–222, 1996

17. Matz PG, Weinstein PR, Sharp FR: Heme oxygenase-1 and heatshock protein 70 induction in glia and neurons throughout ratbrain after experimental intracerebral hemorrhage. Neurosur-gery 40:152–162, 1997

18. Mun-Bryce S, Kroh FO, White J, et al: Brain lactate and pH dis-sociation in edema: 1H and 31P-NMR in collagenase-inducedhemorrhage in rats. Am J Physiol 265:R697–R702, 1993

19. Nath FP, Jenkins A, Mendelow AD, et al: Early hemodynamicchanges in experimental intracerebral hemorrhage. J Neuro-surg 65:697–703, 1986

20. Ohyagi Y, Yamada T, Goto I: Hemoglobin as a novel pro-tein developmentally regulated in neurons. Brain Res 635:323–327, 1994

21. Park CC, Shin ML, Simard JM: The complement membraneattack complex and the bystander effect in cerebral vasospasm.J Neurosurg 87:294–333, 1997

22. Regan RF, Panter SS: Neurotoxicity of hemoglobin in corticalcell culture. Neurosci Lett 153:219–222, 1993

23. Rosen AD, Frumin NV: Focal epileptogenesis after intracorticalhemoglobin injection. Exp Neurol 66:277–284, 1979

24. Sadrzadeh SMH, Anderson DK, Panter SS, et al: Hemoglobinpotentiates central nervous system damage. J Clin Invest 79:662–664, 1987

25. Siesjö BK, Agardh CD, Bengtsson F: Free radicals and braindamage. Cerebrovasc Brain Metab Rev 1:165–211, 1989

26. Suzuki R, Tomita H, Ito U, et al: Late expression of vascular

endothelial growth factor in the reactive astrocytes in experi-mental intracerebral hemorrhage in cats. J Cereb Blood FlowMetab 17 (Suppl 1):S416, 1997 (Abstract)

27. Tomita H, Ito U, Ohno K, et al: Chronological changes in brainedema induced by experimental intracerebral hematoma in cats.Acta Neurochir Suppl 60:558–560, 1994

28. Verma A, Hirsch DJ, Glatt CE, et al: Carbon monoxide: a puta-tive neural messenger. Science 259:381–384, 1993

29. Wagner KR, Xi G, Hua Y, et al: Clot retraction contributes toearly perihematomal edema volumes following intracerebralhemorrhage. Soc Neurosci Abstr 23:1921, 1997 (Abstract)

30. Wagner KR, Xi G, Hua Y, et al: Lobar intracerebral hemor-rhage model in pigs. Rapid edema development in perihema-tomal white matter. Stroke 27:490–497, 1996

31. Wahlgren NG, Lindquist C: Haem derivatives in the cerebro-spinal fluid after intracranial haemorrhage. Eur Neurol 26:216–221, 1987

32. Willmore LJ, Rubin JJ: Formation of malonaldehyde and focalbrain edema induced by subpial injection of FeCl2 into rat iso-cortex. Brain Res 246:113–119, 1982

33. Willmore LJ, Sypert GW, Munson JB, et al: Chronic focalepileptiform discharges induced by injection of FeCl2 into ratand cat cortex. Science 200:1501–1503, 1978

34. Xi G, Wagner KR, Hua Y, et al: Plasma but not packed redblood cells infused into pig cerebral white matter induces rapid-ly developing, marked edema. Soc Neurosci Abstr 22:1424,1996 (Abstract)

35. Yang GY, Betz AL, Chenevert TL, et al: Experimental intra-cerebral hemorrhage: relationship between brain edema, bloodflow, and blood-brain barrier permeability in rats. J Neurosurg81:93–102, 1994

36. Yip S, Ip JKH, Sastry BR: Electrophysiological action of hemo-globin on rat hippocampal CA1 pyramidal neurons. Brain Res713:134–142, 1996

Manuscript received February 20, 1998.Accepted in final form July 20, 1998.This study was supported by Grant No. NS-17760 from the

National Institutes of Health.Address reprint requests to: Guohua Xi, M.D., R5605 Kresge I,

University of Michigan, Ann Arbor, Michigan 48109–0532. email:guohuaxi@umich.edu.

G. Xi, R. F. Keep, and J. T. Hoff

996 J. Neurosurg. / Volume 89 / December, 1998