Differential Effects of Global Ischemia on Delayed Matching- and Non-Matching-to-Position Tasks in...

NEUROBIOLOGY OF LEARNING AND MEMORY 67, 228–247 (1997)ARTICLE NO. NL963758

Differential Effects of Global Ischemia on Delayed Matching- andNon-Matching-to-Position Tasks in the Water Maze and Skinner Box

ALAN NELSON, PETER SOWINSKI, AND HELEN HODGES1

Department of Psychology, Institute of Psychiatry, De Crespigny Park, Denmark Hill, London SE5 8AF, United Kingdom

the CA3 field to 5% loss in one striatal area. No corticaldamage was seen. The extent of CA1 cell loss correlatedIn order to assess effects of global ischemia in tasks of

spatial learning and working memory, male Wistar rats modestly with water maze acquisition (Phase 3) and work-ing memory scores, but not with trials to criterion in thewere subjected to four vessel occlusion (4 VO) for periods

of 5, 10, and 20 min and compared with sham-operated Skinner box task. There were significant correlations be-tween different measures both within and between watercontrols over four test phases, from 6 to 54 weeks after

surgery. Rats were assessed on acquisition in the water maze tasks, but not Skinner box tasks, suggesting thatthe two types of procedure engaged different cognitive pro-maze, a task that is sensitive to ischemic impairments,

before testing in Skinner box and water maze working cesses. The results indicate that the intrahippocampaldamage induced by 4 VO impaired tasks which requiredmemory tasks, which both require the short-term storage

of information, but make different demands on spatial in- processing of allocentric spatial information, but did notimpair the storage of limited spatial information in work-formation processing. Phases 1 and 3 assessed spatial

learning in a standard water maze procedure (12 and 10 ing memory. q 1997 Academic Press

training days, 2 trials/day with a 10-min intertrial inter-val: ITI). Phase 2 involved training and testing in delayed

INTRODUCTIONnon-matching-to-position task in the Skinner box, withdelays of 2–10 s between the information and choicestages. Phase 4 examined working memory in a water Interruption of cerebral blood flow, as occurs inmaze delayed matching-to-position task with 4 trials/day, heart attack, bypass surgery, or coronary artery oc-an ITI of 30 s, and a novel platform position on each day. clusion, results in a high incidence (up to 48%) ofIschemic rats showed duration-related impairments in memory deficits, in the absence of general cognitivewater maze acquisition and working memory, but not in or neurological impairment (Cummins, Tomiyasu,the less spatially demanding Skinner box task. Since wa- Read, & Benson, 1984; Graham, 1992; Roine, Ka-ter maze acquisition deficits were seen both before and

jaste & Kaste, 1992; Volpe & Petito, 1985). Hippo-after testing in the Skinner box the lack of effect cannotcampal cells, particularly cells in the hilar andbe attributed to time or to prior training. Ischemic deficitsCA1 fields, are highly vulnerable to effects of isch-were more marked in Phase 3 than in Phase 1 of acquisi-emia (Hsu & Buzsaki, 1993; Freund & Magloczky,tion, suggesting that impairment may be progressive. His-

tological assessment showed that cell loss was largely con- 1993; Matsuyama, Tsuchiyama, Matsumoto, Naka-fined to the hippocampal CA1 field and was linearly re- mura, & Sugita, 1993). Postmortem and neuroimag-lated to duration of occlusion. At the maximal level of ing studies of patients with memory loss followingloss (5.7 mm before the interaural line) the 20-min group ischemic episodes have suggested that bilateralshowed 90% loss, the 10-min group 60% loss, and the 5- damage to the CA1 field may underlie impairments,min group, which did not differ from controls, less than which may involve both retrograde and anterograde10% loss. Only the 20-min group showed significant dam- memory loss (Petito, Feldman, Pulsinelli, & Plum,age beyond the CA1 field, ranging from 30–40% loss in

1987; Zola-Morgan, Squire, & Amaral, 1986; Kart-sounis, Rudge, & Stevens, 1995). Animal models of

1 This work was supported by a British Heart Foundation stu- global ischemia induced by transient bilateral occlu-dentship awarded to A.N. and by the Wellcome Trust. Address sion of the common carotid arteries with hypotensioncorrespondence and reprint requests to Alan Nelson at Depart-

(two vessel occlusion: 2 VO) or cauterization of thement of Psychology, Institute of Psychiatry, De Crespigny Park,Denmark Hill, London SE5 8AF UK. Fax: 0171 708 3497. vertebral arteries combined with occlusion of the ca-

2281074-7427/97 $25.00Copyright q 1997 by Academic PressAll rights of reproduction in any form reserved.

AID NLM 3758 / 6v0b$$$181 04-19-97 04:11:35 nlma AP: NLM

229ISCHEMIA DISRUPTS SPATIAL BUT NOT NONSPATIAL TASKS

rotids (four vessel occlusion: 4 VO) have been used to largely employed tasks similar to those used for in-vestigating effects of hippocampal lesions, particu-identify regional brain damage and cognitive deficits

associated with defined durations of occlusion larly tasks involving spatial learning and spatialworking memory. However, findings have been vari-(Hodges, Sinden, Meldrum, & Gray, 1994; Nunn &

Hodges, 1995; Schmidt-Kastner & Freund, 1991). able. Some workers have reported deficits in the ra-dial maze (both in working memory errors of reentryWith short durations (up to 15 min of 4 VO) cell loss

is largely confined to the CA1 and hilar regions, but and reference memory errors of entry into arms thatare never rewarded) (Davis, Tribuna, Pulsinelli, &with longer durations CA1 cell loss increases lin-

early, and damage is more extensive within the hip- Volpe, 1986; Davis and Volpe, 1990; Kiyota, Miya-moto, & Nagaoka, 1991; Volpe, Pulsinelli, Tri-pocampus and extends to cortex and striatum

(Nunn, LePeillet, Netto, Hodges, Gray, & Meldrum, buna, & Davis, 1984; Gionet, Thomas, Warner,Goodlett, Wasserman, & West, 1991), but others1994; Pulsinelli, Brierley, & Plum, 1982).

Lesion studies have suggested that spatial work- have found no impairment in radial maze placelearning (Nunn, Le Peillet, Netto, Sowinski, Hodges,ing memory tasks are peculiarly sensitive to effects

of hippocampal damage. However, a long-standing Meldrum, & Gray, 1991). Place learning in the watermaze, a task argued to require allocentric spatialcontroversy has centered around whether this re-

flects intermediate-term storage (Olton, Becker, & information processing that is reliably impaired byhippocampal lesions (Eichenbaum, Stewart, & Mor-Handleman, 1979; Rawlins, 1985) or spatial ‘‘cogni-

tive mapping’’ (O’Keefe and Nadel, 1978) functions ris, 1991; Morris, Garrud, Rawlins, & O’Keefe, 1982)has similarly been found in some laboratories to beof the hippocampus. One approach to resolving this

issue is to compare the performance of animals with significantly impaired after global ischemia (Jas-pers, Block, Heim, & Sontag, 1990; Netto, Hodges,hippocampal damage in both nonspatial working

memory tasks (e.g., delayed object or goal box recog- Sinden, LePeillet, Kershaw, Sowinski, Meldrum, &Gray, 1993; Nunn et al., 1994; Olsen, Scheel-Kruger,nition) and more spatially demanding tasks (e.g., T-

maze alteration, radial maze arm reentries). In the Moller, & Jensen, 1994), whereas others workers re-port that it is not disrupted (Kiyota et al., 1991;rat evidence for hippocampal involvement in non-

spatial working memory tasks is mixed; Raffaele and Gionet et al., 1991), impaired only in complex tasks(Auer, Jensen, & Whishaw, 1989), or impaired tran-Olton (1988), and Mumby, Wood, and Pinel (1992)

report impairments in delayed object recognition siently relative to prolonged deficits in T-maze alter-nation (Hagan and Beaughard, 1990). Despite thisafter hippocampal damage, whereas Aggleton, Hunt,

and Rawlins (1986), Kelsey and Vargas (1993), and variability, which may reflect differences in the ex-tent of ischemic damage, and in the amount of pre-Rothblat and Cromer (1991) found that rats with

damage to the hippocampus or its inputs discrimi- and postoperative training (Davis and Volpe, 1990;Netto et al., 1993), these findings combine to suggestnated successfully between objects, but performed

poorly in T-maze alternation or spatial variants of that ischemic impairments are frequently shown intasks that require use of allocentric spatial cues,the delayed recognition tasks. Primate studies of

hippocampal damage have relied heavily on delayed whether of learning or working memory.Effects of ischemia on working memory in whichobject recognition (DNMTS) tasks that emphasise

storage of nonspatial information in working mem- there is no spatial component have been less widelyinvestigated than in spatial tasks, and there haveory and are also sensitive to amnesia in humans

(Zola-Morgan, Squire, Amaral, & Suzuki, 1989). been few direct comparisons of the same animalsin different types of working memory task. PrimateHowever, in parallel studies Murray, Davidson, Gaf-

fan, Olton, and Suomi, (1989) and Markowska, Ol- studies (Bachevalier & Mishkin, 1989; Zola-Morgan,Squire, Rempel, Clower, & Amaral; 1992) haveton, Murray, and Gaffan, (1989) have shown that

both monkeys and rats with fornix transection were found delay-dependent deficits in DNMTS tasks fol-lowing bilateral occlusion of the posterior cerebralprofoundly impaired in a T-maze runway spatial

working memory task, but not in conditional dis- artery supplying the hippocampus or 15 min ofglobal ischemia induced by a pediatric cuff and hypo-crimination tasks that required place-reward learn-

ing. Thus, in both rats and monkeys, spatial infor- tension. Wood, Mumby, Pinel, and Phillips (1993)used a task analogous to DNMTS in primate models,mation processing rather than working memory

storage appears to be a critical target for hippocam- in which rats subjected to 20 min of 2 VO with hypo-tension, were tested on delayed object recognition,pal damage, although there is some evidence for im-

pairment in nonspatial working memory. using two objects, one familiar, one novel, drawnfrom a large pool for quasi trial-unique presentationStudies of the effects of global ischemia have


230 NELSON, SOWINSKI, AND HODGES

and placed randomly in left or right locations in a of 4 VO surgery and 800 g at the end of behavioraltesting. Rats were housed 5 to a cage and maintainedrunway. Ischemic rats were impaired both in acqui-

sition of this task and with retention intervals of on a 14 h light/10 h dark cycle (lights on 7 AM–9 PM).They were fed ad libitum, except during training and4–300 s, more so than animals with hippocampal

lesions assessed in the same task (Mumby et al., assessment of DNMTP (see below), when weight wasreduced to 80–85% of free feeding weight, and ani-1992). Performance correlated with the number of

surviving CA1 cells. These findings suggested that mals fed for 1 h each day after training.substantial ischemic disruption of storage in work-

Ischemia Surgerying memory occurs, irrespective of the spatial de-mands of the task. Transient global ischemia was induced by the four

The present experiments aimed to compare the vessel occlusion (4 VO) method of Pulsinelli et al.effects of different durations of global ischemia (5, (1982), in which the vertebral arteries were electro-10, and 15 min of 4 VO) on two tasks of working coagulated through the alar foramenae on the firstmemory: delayed matching-to-position in the water cervical vertebrae, under 2% halothane anesthesiamaze and delayed non-matching-to-position in the (in 70% nitrous oxide and 30% oxygen), and silasticSkinner box. Rats were also assessed in standard ties were inserted around the carotids and broughtwater maze place learning before both of the working to the surface. Twenty-four hours later the ties werememory tasks, in order to verify the presence of an tightened and clamped for 5, 10, or 20 min. Rats thatischemic deficit. The two working memory tasks failed to lose righting reflex within 2 min were notwere chosen to highlight different cognitive de- included in the experiments. Head and body temper-mands. Matching-to-position in the water maze, in atures were recorded once a minute and maintainedwhich rats are given ca. 4 trials a day, with a differ- at 377 { 17C using a heating pad and a lamp. Atent platform position on each day and a short in- the end of the specified period the ties were rapidlytertrial interval (ITI), features rapid allocentric spa- removed and wounds sealed with Michel clips.tial learning, in which normal rats may show one- Twelve controls were subject to an identical proce-trial learning (Hodges et al., 1996). DNMTP in the dure, except that the ties were not tightened. WeSkinner box requires rats to select the lever not pre- had previously established that such sham-operatedsented at the information stage, when both levers animals do not differ from unoperated controls inare presented in the choice stage. Although this re- water maze performance (Nunn et al., 1994).quires lengthy training (ca. 80 sessions of 40 trials) For postoperative care rats were housed in theit permits accurate control over the interval between recovery room for 3–4 weeks. They were first housedthe information and choice stages, which is not possi- singly for ca. 5 days for individual monitoring andble in the maze procedures previously used to test given wet mash and sc injections of Duphalyte orspatial and nonspatial working memory in ischemic saline for fluid replacement if necessary. They wererats, so that once the procedure has been mastered, weighed daily, food and water intake was noted, andthe temporal parameters of working memory can ex- excretion, urination, and general appearance wereamined with precision. Moreover, since there were checked. For the first week rats were tested dailyonly two lever positions, the DNMTP task provided for neurological deficits (e.g., speed of righting reflex,a method of assessing working memory in a less spa- gait, balance, paw placement, and grasping), as parttially demanding situation than in the water maze, of postoperative care. Few deficits were found 2–3so that the use of these two procedures in effect rep- days after ischemia, no animals were excluded forresents assessment of ischemic animals in spatial neurological impairment 7 days after surgery, andand nonspatial working memory tasks (for discus- transient postoperative neurological deficits did notsions see see Nunn & Hodges, 1995; Hodges, 1996). correlate with initial (Phase 1) water maze swim

speed. Rats were then rehoused in groups of 4–5MATERIALS AND METHODSand were returned to the animal house after 2–3

Experiments were conducted in accordance with weeks to await behavioral testing, which commencedthe terms of the UK Animals (Scientific Procedures) ca. 4 weeks after ischemia.Act, 1986, and followed local and national guidelinesof good practice. Behavioral Test Procedures and Apparatus

Animals Rats were tested in four phases from 4–6 to 54weeks after ischemia surgery (see Table 1). Phase 1Forty-two male Wistar rats were used (Charles

River Ltd.) with a mean weight of 280 g at the time (6 to 9 weeks after the start of surgery) assessed



TABLE 1Time Course of the Experiments

Weeks

1–5 Ischemia surgery and postoperative recovery6–9 Phase 1: water maze acquisition (12 days of training with 2 trials/day, followed

by a probe trial12–48 Phase 2: delayed non-matching-to-position in the Skinner box (120 acquisition

sessions of 40 trials, 16 sessions with delays of 2–10 sec)50–52 Phase 3: water maze acquisition (10 days of training with 2 trials/day, plus a probe

trial)54 Phase 4: water maze working memory (4 days, 4 trials/day with a different platform

position on each day).

spatial learning in the water maze for 12 days (2 placed in the water facing the maze wall and allowedto swim until they found the platform, where theytrials/day with a 10-min ITI, followed by a proberemained for 10 s before removal either to a holdingtrial with the platform removed to test memory forcage between trials or to their home cage after beingthe platform location. In Phase 2 (12 to 48 weekstoweled dry. A rat which failed to find the platformafter surgery) rats were trained in a non-matching-within 1 min was guided there by the experimenter,to-position task in the Skinner box, and, after reach-and the maximum latency of 60 s was scored. Theing a criterion of 80% correct choices, were testedswim path was recorded by an HVS image analyzingwith delays of 2–12 s inserted between retraction ofsystem (VP 112, HVS Image Ltd., Hampton, UK)the response lever and presentation of two levers forand latency to reach the platform, swim path length,choice. Phase 3 (50–52 weeks after surgery) retestedpercentage of time spent in each quadrant oracquisition in the water maze (2 trials/day for 10annulus, time (in 0.1-s units) in the counter sur-days, followed by a Probe Trial). Phase 4 (54 weeksrounding the platform position, and heading angleafter surgery) assessed working memory in the wa-(a measure of initial divergence from the direct pathter maze (delayed matching-to-position) using 4 tri-to the platform) were automatically computed. Swimals/day for 4 days, with a 30-s ITI and a differentspeed was measured by dividing distance (swim pathplatform position on each day. Four groups werelength) by latency.compared: Sham-operated controls (CON: n Å 12)

For acquisition rats were given 2 trials/day, withand rats subjected to 4 VO for 5 min (ISC-5: n Å 10),two different start points, one close and one distant10 min (ISC-10: n Å 10), and 20 min (ISC-20: n Åto the platform, and a 10 min ITI. A Probe Trial was10). The total number of animals declined from 42conducted 24 h after the last training trial with thein Phase 1 to 41 in Phase 2 and 37 in Phases 3 andplatform removed, to assess memory for the correct4. This was because one rat in the ISC-20 group withplatform location by the heading angle, the percent-a bitten foot was excluded in Phase 2 and four ratsage of time spent in the training quadrant, and timewith respiratory infection (two controls and two inspent in the counter area.the ISC-5 group) were excluded in Phase 3.

For delayed matching-to-position in the waterWater maze acquisition and working memory maze rats were given 4 trials/day and a 30-s ITI,

(Phases 1, 3, and 4). Rats were trained to find a with a different platform position on each of 4 days,hidden platform (9 cm diameter) submerged ca. 3 and the four start points used in a different ordercm below the surface in a water maze (2 m diameter, each day. The platform was placed in nonstandard0.5 m high, with 25 cm depth of water, clouded with positions (e.g., in the center or near the edge of themilk). The maze was divided conceptually into four pool). The first trial, when the platform was reachedquadrants (Quads 1–4) and three annuli (A, B, and by chance, constituted the ‘‘information’’ stage, andC, with C outermost). The platform was centered the second and subsequent trials required matchingin one of the quadrants within Annulus B and was to the novel position for that day. Working memorysurrounded by an area twice the size of the platform was assessed by the difference in latency betweenin diameter, termed the counter area, which was Trials 1 and 2 (in absolute terms and as percentageused as a measure of search accuracy during the of savings from Trial 1 to 2), and the slope of acquisi-Probe Trial. Start positions were designated as from tion over trials provided a measure of speed of learn-

ing (Netto et al., 1993).N, S, W, or E. At the start of all trials rats were



Delayed non-matching-to-position in the Skinner fused brains were embedded in wax. Coronal 7-mmsections were cut by a Leitz rotary microtome (Modelbox (Phase 2). MAGAZINE TRAINING: Rats were de-

prived to 80–85% of their free feeding weight and 1512) and serial sections through the hippocampalformation were mounted onto slides, stained withfirst accustomed to take pellets (25 mg, Camden In-

struments) from the food tray in the Skinner box, cresyl violet, and counterstained with luxol fast blue.Under light microscopy sections were examined atwhich was illuminated with the flap taped open, and

then learned to push the flap to reach the food. They four levels: 10.7, 9.7, 5.7, and 3.7 mm anterior to theinteraural line, with level 1 being the most anterior.were then shaped to press the lever in daily sessions

of 10 min. Magazine training was completed when The cortex was examined at all four levels, the stria-tum at levels 1 and 2, and fields CA1–4 of the hippo-the rats made 50 presses in three consecutive ses-

sions. campus at levels 3 and 4. Cell loss was graded bilat-erally on a 6-point scale (where 0 Å 0–10% loss, 1NON-MATCHING-TO-POSITION TRAINING: A single

‘‘information’’ lever was presented and the rats re- Å 10–30% loss, 2 Å 30–50% loss, 3 Å 50–70% loss,4 Å 70–90% loss, and 5 Å 90–100% loss) by twoceived a pellet when they pressed it. Four seconds

later both levers were presented (the choice stage) independent observers, who did not know the experi-mental condition of the animals (Netto et al., 1993;and a press on the other lever was rewarded. Both

levers were retracted after any press, or after 5 s, Nunn et al., 1994), and agreement between observa-tions was ca. 80%.and there was a 20-s delay before presentation of

the next information lever. Rewards for pressing theData Analysisinformation lever served to motivate the animals,

and the requirement to nose poke to retrieve food Water maze results were analyzed by split plotfrom the magazine between the levers prevented the analyses of variance (GENSTAT-V PC, Rothamsted,animals from using their body position to denote the UK) with Groups as the between- and Days and Tri-correct lever. Each daily training session consisted als as the within-subjects factors. Linear trends forof 40 information and choice trials, and training was Groups (duration of occlusion) and Days were ex-continued until the rats made 80% correct choices tracted by orthogonal polynomials. Groups were(32/40) over 3 consecutive sessions. Groups were compared by Newman–Keuls multiple comparisons.compared by the number of sessions to criterion (cri- If significant interactions between trends of Groupsterion—3 sessions) and by the percentage of correct and Days occurred, indicating differences in rates ofchoices over blocks of 10 sessions. Training in all learning. Groups were compared (Newman–Keulsanimals was completed in 120 sessions, sessions to procedure) using trend contrasts. Separate ANOVAcriterion ranged from 30 to 118. were used for measures of latency, distance, heading

DELAYED NON-MATCHING-TO-POSITION: All ani- angle, and time spent in counter areas. Percentagesmals were run for 5 days with a 10-s delay between of time in each quadrant and annulus were subjectinformation and choice lever presentation (the pro- to angular transformation (arcsine square root) andgrammed delay was 7 s, since total delay was in- analyzed as additional factors. For quantitative his-creased by the 3–4 s required by the apparatus). tology, cell loss grades were first compared in isch-Animals were then trained with delays of 2, 4, 6, 8, emic groups and controls by one-way Kruskal–Wal-and 10 s (i.e., actual delays of ca. 6, 8, 10, 12, and lis ANOVAs. Since these produced equivalent sig-14s), for 5 sessions/week for 4 weeks. Only one delay nificance levels to those obtained by one-waywas presented in each session, but a different ran- ANOVAs, results for the latter have been reported,dom order of delays was presented to each animal followed by Newman–Keuls comparisons of means,across sessions, so that by the end of training each since standard ANOVAs permitted trend analysisanimal had received four session at each delay. The of the effects of duration of occlusion. Correlationspercentage of correct responses at each delay were between cell loss grades and behavioral measurescompared across groups. were maze by Spearman’s rank correlation coeffi-

cient, corrected for ties (Seigel, 1956).Histology

RESULTSAt the end of behavioral testing the rats were per-

Phase 1: Acquisition in the Water Maze (4–6 Weeksfused transcardially for 5 min under deep terminalafter Ischemia)pentobarbital anesthesia with 400 ml of FAM (form-

aldehyde:40%, w/v, glacial acetic acid:methanol, Latency. Groups showed marginal overall differ-ences in latency to find the platform [F(3, 38)Å 2.78,made up in proportions of 1:1:8) (BDH, UK). Per-



Annulus search. Groups differed in the percent-ages of time spent in the Annuli [F(3, 38) Å 5.75, pÅ .002] and these differences were linearly relatedto duration of occlusion [F(1, 38) Å 9.06, p Å .005for the linear trend of Groups; F(2, 1824) Å 72.61, põ .001 for the interaction of Annuli with the lineartrend of Groups]. The highest proportion of time wasspent in C and the lowest in A (põ .01 in all compari-sons). The ISC-20 group spent a lower percentage oftime in the innermost Annulus A and Annulus Bcontaining the platform than the other groups anda greater proportion of time circling round the poolwalls in C (p õ .01 in all comparisons: see Fig. 2b).The ISC-5 and ISC-10 groups also showed increasedthigmotaxis in C relative to controls (p õ .05).

Probe Trial. Groups performed comparably onFIG. 1. Phase 1, 6–9 weeks after surgery: Mean time takento find the platform over 12 days of training by sham-operated the Probe Trial; there were no differences for head-controls (CON) and three groups subjected to 4 VO for 5 min ing angle or percentages of time spent searching in(ISC-5), 10 min (ISC-10), and 20 min (ISC-20). Only the ISC-20 the appropriate pool sectors. Although there weregroup was significantly impaired relative to controls and to the

substantial differences between quadrants [F(3, 114)other ischemic groups. SED bar shows twice the standard errorÅ 25.69, p õ .001], with a higher proportion of timefor the difference of means for the Groups 1 Days interaction.spent in Quad 4 than the other sectors (p õ .01), allgroups showed a similar degree of preference for thetraining quadrant, so that there were no interactionspÅ .05], which were significantly related to durationbetween Groups and Quadrants. The high level ofof occlusion [Flin(1, 38) Å 6.95, p õ .015: see Fig.thigmotaxis seen during training in ischemic ani-1]. Although all ischemic groups took longer thanmals was not evident in the Probe Trial.controls, by Newman–Keuls comparisons only the

ISC-20 group differed significantly (p õ .05). All Phase 2: Delayed Non-Matching-to-Position in thegroups showed substantial decreases in latency over Skinner Box (12–48 Weeks after Ischemia)Days [Flin(1, 418) Å 213.62, p õ .001], but trend

Acquisition. For acquisition the mean number ofcontrast comparisons showed that the decrease wascorrect responses in each group were compared overmore shallow in the ISC-20 group than all the other120 sessions (40 trials/session) averaged over blocksgroups (p õ .05: see Fig. 1). Results for swim pathof 10 sessions. Groups were also compared for thelengths were similar to those for latency, and therenumber of sessions taken to reach the criterion of 3were no differences between Groups for speed ofsuccessive sessions at 80% correct responding. Thereswimming [F(3, 38) Å 0.13, NS], so that results forwas no significant difference between groups in thelatency are not likely to reflect motor or motivationalnumber of errors over training sessions. All groupsdifferences. All groups showed a similar accuracy ofshowed a highly significant linear increase in correctheading angle.choices over blocks [F(1, 418) Å 1938, p õ .001],which leveled off as performance reached criterion,Quadrant search. Percentages of time spent inyielding a quadratic trend [F(1, 418) Å 287.29, p õthe Quadrants differed substantially [F(3, 2736 Å.001]. Rates of learning were equivalent in all377.41, p õ .001]. The training quadrant attractedgroups, so that there were no interactions betweenfar more exploration than the other sectors (p õ .01groups and trends of blocks.in all comparisons), followed by Quad 3 adjacent to

Groups also showed no significant differences inthe Experimenter, which differed from Quads 1 andthe number of sessions needed to reach criterion.2 (p õ .01). However, Groups differed in the propor-Controls took the greatest number of sessions, aver-tion of time spent in the various Quadrants [F(9,aging 82, as opposed to 59, 60, and 74 in the ISC-2736) Å 5.23 for the interaction of Quadrants with20, ISC-5, and ISC-10 groups, respectively, but thisGroups]. Differences were most marked for thedifference was not significant.Training Quadrant (Quad 4), where the ISC-20

group spent a smaller proportion of time than all the Delayed responding. Groups were compared forpercentage of correct responses at delays of 2–10 sother groups (p õ .05: see Fig. 2a).



FIG. 2. Phase 1, 6–9 weeks after surgery: Search strategies. Mean percentages of time ({ SD) spent in the quadrants (a) and annuli(b) averaged over 12 days of training by sham-operated controls (CON) and three groups subjected to 4 VO for 5 min (ISC-5), 10 min(ISC-10), and 20 min (ISC-20). The ISC-20 group spent less time in the training quadrant (4) than controls or the other ischemic groups.This group spent a lower percentage of time in the inner areas of the pool (A, B) and all ischemic groups spent more time circling roundthe wall in C. SED bars show twice the standard errors for difference of means for the groups 1 sector interactions. Difference fromcontrols: wp õ .05; wwp õ .01.

between retraction of the response lever and presen- improve, while at the longer delays it deterioratedover sessions, leading to interactions between ses-tation of choice levers. There was a marked decrease

in accuracy with increasing delays, resulting in a sions and the the linear [F(3, 555) Å 6.78, p õ .001]and quadratic [F(3, 555) Å 4.62, p Å .003] trends ofrobust effect of Delay [F(4, 148) Å 70.46, p õ .001]

with a strong linear component [F(1, 149) Å 277.37, Delay. There was no vestige of difference betweenthe performance of control and ischemic Groups, orp õ .001: see Fig. 3]. There was also a difference

between testing sessions. This did not reflect a uni- between the different ischemic Groups, and no evi-dence for increasing impairment with increasing du-form practice effect of improvement over sessions,

as at the shorter intervals performance tended to ration of occlusion. As Fig. 3 shows, the slopes of



were substantial interactions between the lineartrends of Days and of Groups [F(1, 297) Å 10.51, põ .001 and F(1, 297) Å 7.78, p Å .006, respectively].These results indicate that there were significantduration-related effects on the rates of learning overthe training period, as illustrated for Latency by Fig.4. Slopes of decrease in latency over days differedsignificantly (p õ .05) in the ISC-10 and -20 groups,relative to controls.

Quadrant search. All rats spent more time in thetraining quadrant (1) than in the other sectors [F(3,1980) Å 209.79, põ .001; põ .01 in all comparison],but there was also marked preference for the formertraining quadrant (4) which attracted more explora-tion (põ .01) than Quads 2 and 3. There were inter-actions between the linear trends of Groups and both

FIG. 3. Phase 2, 12–48 weeks after surgery: Delayed non- Days [F(1, 297) Å 6.27, p õ .015] and Quadrantsmatching-to-position. Mean performance of controls (CON) and[F(3, 1980) Å 5.64, p õ .001], showing that durationthree groups subjected to 4 VO for 5 min (ISC-5), 10 min (ISC-of ischemia affected distribution of search in the10), and 20 min (ISC-20). Animals were trained to a criterion of

80% choice accuracy before interpolation of delays of 2–10 s be- quadrants. The ISC-5 group performed comparablytween retraction of the information lever and presentation of the to controls. However, the ISC-10 and ISC-20 groupsresponse levers. All groups showed decreasing accuracy with inspent a lower proportion of time in the trainingcreasing delays, but the control group was nonsignificantly less

quadrant than controls (p õ .05), and the ISC-20accurate than the ischemic groups at all delays. SED bar showsgroups also spent significantly more time than con-twice the standard error for the difference of means for the Groups

1 Delay interaction. (Data adapted from Behavioural Brain Re- trols in Quad 2 (see Fig. 5a).search 65, Nunn, J. A., & and Hodges, H. Cognitive deficits in-

Annulus search. In contrast to initial acquisi-duced by global cerebral ischemia, pp. 1–31, copyright 1994, withkind permission of Elsevier Science BV, Amsterdam Publishing tion, the greatest percentages of time were nowDivision, Sara Burgerhartstraat 25, 1055 KV, Amsterdam, TheNetherlands.)

delay-dependent decrease in accuracy were verysimilar in all groups. In contrast to results for thewater maze, the ISC-20 group showed the most accu-rate performance of all groups, but did not differfrom controls.

Phase 3: Retraining in the Water Maze (50–52Weeks after Ischemia)

Lack of effects of ischemia on delayed non-match-ing-to-position performance might have resultedfrom the effects of time or prior training in the watermaze. Rats were therefore reassessed on water mazeacquisition with the platform in a different quad-rant, but using the same procedure as that in Phase1, to see whether a spatial learning deficit was still

FIG. 4. Phase 3, 50–52 weeks after surgery: Mean time takenevident in this task, before assessment in the waterto find the platform over 10 days of training by sham-operatedmaze working memory task.controls (CON) and three groups subjected to 4 VO for 5 min(ISC-5), 10 min (ISC-10), and 20 min (ISC-20). There were noLatency, distance, and heading angle. Thereoverall differences between Groups, but both the ISC-10 and ISC-were no significant differences overall between 20 groups showed significantly retarded rates of learning relative

Groups for Latency, Distance, and Heading Angle to controls. SED bar shows twice the standard error for the differ-ence of means for the Groups 1 Days interaction.measures. However, for Latency and Distance there



FIG. 5. Phase 3, 50–52 weeks after surgery: Search strategies. Mean percentages of time ({ SD) spent in the quadrants (a) andannuli (b) averaged over 10 days of training by sham-operated controls (CON) and three groups subjected to 4 VO for 5 min (ISC-5), 10min (ISC-10), and 20 min (ISC-20). Both the ISC-10 and ISC-20 groups spent less time in the training quadrant (1) than controls. Thesetwo groups also spent lower percentages of time in the inner area of the pool (A) and the ISC-20 group spent more time circling in C.SED bars show twice the standard errors for the difference of means for the Groups 1 sector interactions. Difference from controls: wpõ .05; wwp õ .01.

spent in A, rather than C (p õ .01, in comparison 5 groups (p õ .05 for the ISC-10 group; p õ .01 forthe ISC-20 group: see Fig. 5b). The ISC-20 groupwith B and C). However, thigmotaxis was again seen

in ischemic groups, with severity related to duration also displayed more thigmotaxis than the ISC-10of occlusion [F(2, 1320) Å 28.88, p õ .001 for the group (p õ .01).interaction of Annuli with the linear trend of

The Probe Trial. There were no significantGroups]. The ISC-10 and ISC-20 Groups spent lessGroup differences or duration-related trends fortime searching in the center of the pool (p õ .01)

and more time circling in C than the control or ISC- heading angle and distribution of search time in the



3.59, p õ .025 for the difference between Groupsand F(1, 33) Å 7.62, p õ .01 for the linear trend ofduration]. Comparisons of means (see Fig. 6b)showed that all three ischemic groups differed fromcontrols on both measures (p õ .05 in all compari-sons).

Comparison of Phases 1 and 3

Groups were compared over 10 days of training inPhases 1 and 3, omitting animals that did not com-plete both phases. Specifically, interactions betweengroups and phases were examined to see if therewere changes in the relative performance of the isch-emic groups over the two phases. There were noGroups1 Phase interactions for latency and headingangle, but this interaction emerged strongly for dis-tance swum [F(3, 660) Å 6.93, p õ .001]. Further-more, for latency, massive interactions betweenPhases and the linear trend of Groups [F(3, 660) Å20.76, p õ .001] and the linear trends of Groups 1Days [F(1, 660) Å 18.46, p õ .001] indicated thatduration of occlusion had more marked effects onPhase 3 than Phase 1 and that Groups differed in therates of improvement over days in the two phases.

Phase 4: Working Memory in the Water Maze

On Trial 1 the platform was found by chance, andworking memory was assessed by three furthermatching-to-position trials, using a different plat-

FIG. 6. Phase 3, 50–52 weeks after surgery: Probe trial. form position on each of the 4 days of testing.Means ({ SD) for (a) percentages of time spent in the trainingquadrant and outer annulus C; and (b) for time spent in the Latency. Time taken to find the platformtraining counter and number of platform position crossings in dropped dramatically over trials and leveled off assham-operated controls (CON) and three groups subjected to 4 the position was learned, leading to substantial lin-VO for 5 min (ISC-5), 10 min (ISC-10), and 20 min (ISC-20).

ear [F(1, 396) Å 366.91, p õ .001] and quadraticIschemic groups did not spend a significantly reduced percentage[F(1, 396) Å 52.61, p õ .001] trends in the mainof time in the former training quadrant than controls and did not

show increased thigmotaxis. However, all ischemic groups spent effect of Trials [F(3, 396)Å 139.89, põ .001]. Groupsless time searching in the precise former location of the platform differed [F(3, 33) Å 3.18, p õ .05] and showed(Counter 1) and swam over this position less frequently than marked duration-related impairment [F(1, 33) Åcontrols. SED bars show twice the standard error for difference

8.74, p õ .006), with longer durations showing thein means for the main effect of Groups. Difference from controls:most retardation over trials [F(1, 396) Å 8.44, p õwp õ .05..005 for the interaction of the linear trends of Trialsand Groups]. Within-trial comparisons of Groups(see Fig. 7) indicated that there were no differencesquadrants and annuli, and all groups showed compa-

rable preference for the training quadrant [F(3, 99) in Trial 1, but that all three ischemic groups tooklonger to find the platform than controls on Trial 2Å 11.47, p õ .001 for the difference between quad-

rants: see Fig. 6a]. However, for measures of recall (p õ .05 for the ISC-10 group; p õ .01 for the ISC-5 and ISC-20 groups). On Trial 3 both the ISC-10 (pof the precise former position of the platform

(Counter 1 and Crossings), effects of ischemia were õ .05) and ISC-20 (p õ .01) groups were retardedrelative to controls; the ISC-20 groups also differedsignificant and related to duration of occlusion

[Counter 1: F(3, 33)Å 3.41, põ .05 for the difference significantly from the other ischemic groups (põ .05cf. the ISC-10 group; p õ .01 cf. the control and ISC-between Groups and F(1, 33) Å 5.81, p õ .025 for

the linear trend of duration; Crossings 1: F(3, 33) Å 5 groups). By Trial 4 only the ISC-20 group was



CA1 field, ranging from 30–40% in the CA3 field toca. 5% at level 2 in the striatum. CA1 cell loss rangedfrom 90 to 100% in the ISC-20 group and from 50 to60% in the ISC-10 group, but was less than 10% inthe ISC-5 group, indicating a significant linear in-crease with duration of occlusion. No cortical cell losswas evident at any of the four levels in any group.

Although striatal damage was minimal, differ-ences between Groups at level 2 were significant[F(3, 37) Å 5.50, p õ .005] and showed a duration-related linear trend [F(1, 37) Å 12.14, p Å .001]. TheISC-20 group differed from all the other groups (põ .01). Differences between Groups for cell loss infields CA1 and CA3 were highly significant at bothlevels 3 and 4. CA1 loss at level 3 was most extensiveand the substantial Group differences [F(3, 37) Å

FIG. 7. Phase 4, 54 weeks after surgery: Working memory in 36.52, p õ .001] showed a marked linear duration-the water maze. Latency to find the platform over Trials 1–4,related trend [F(1, 37) Å 103.04, põ .001]. The ISC-averaged over 4 days of testing, in sham-operated controls (CON)20 group displayed significantly more damage thanand three groups subjected to 4 VO for 5 min (ISC-5), 10 min

(ISC-10), and 20 min (ISC-20). Groups did not differ on Trial 1, all the other groups (p õ .01 in comparison with thewhen the platform was found by chance. Controls rapidly learned control and ISC-5 groups, põ .05 in comparison withthe position, reaching asymptotic levels by Trial 3. Ischemic the ISC-10 group). The ISC-10 group also differedgroups showed duration-related impairment, with the ISC-20

significantly (p õ .01) from the control and ISC-5group taking significantly longer than controls on all trials. SEDgroups. Group differences for CA1 cell loss werebar show twice the standard error for the difference in means for

the Groups 1 Trials interaction. (Data reproduced from Cognitive equally marked at level 4 [F(3, 37)Å 38.76, põ .001,Brain Research 3, Hodges, H. Maze procedures, pp. 167–181, with a strong duration-related linear trend [F(1, 37)Copyright 1996, with kind permission of Elsevier Science BV, Å 106.12, p õ .001], though damage was less exten-address as in Fig. 3.)

sive than at level 3. The ISC-20 group differed fromcontrols and the other ischemic groups (põ .01), andthe ISC-10 group also had substantially (p õ .01)impaired (põ .01 cf. the controls and ISC-10 groups;greater loss than the ISC-5 and control groups,p õ .05 cf. the ISC-5 group). Results for Distancewhich did not differ. Figure 9 highlights the extentwere comparable to those for latency [F(3, 33)Å 3.81,to which this variable CA1 cell loss at level 3 wasp õ .02 for the main effect of Groups; F(1, 33) Åevident across groups. With respect to CA3 cell loss,7.43, p Å .01 for the linear trend of duration], andGroup differences and the linear duration-relatedthere was no difference between Groups for speedtrends of Group were attributable to damage in theof swimming, so that latency effects appear to beISC-20 group, but not the other groups for both levelunconfounded by motor or motivational changes.3 [F(3, 37) Å 13,12, põ .001 for the main effect; F(1,

Heading angle. Groups differed substantially in 37) Å 31.64, põ .001, for the linear trend of Groups]accuracy of heading angle [F(3, 33) Å 13.23, p õ .001] and level 4 [F(3, 37) Å 9.32, p õ .001, for the mainand impaired accuracy (i.e., increased size of angle) effect; F(1, 37) Å 23.65, põ .001, for the linear trendwas also strongly related to duration of occlusion, as of Groups]. At both levels The ISC-20 group showedshown by the substantial linear trend of Groups [F(1, more CA3 cell loss than the other groups (p õ .0133) Å 25.29, p õ .01]. All the ischemic groups were in all comparisons), but the ISC-5 and ISC-10 groupssubstantially less accurate than controls (p õ .01 in did not differ from each other or from controls.all comparisons). Accuracy of heading angle differedmarginally over Trials [F(3, 396) Å 2.63, p Å .05], but

Correlations between Cell Loss and Behavioralshowed no systematic improvement.Scores

Ischemic Cell Loss There were no significant correlations between theextent of cell loss and behavioral scores for initialCoronal sections from control and ischemic groups

were graded for cell loss at four levels (10.7, 9.7, 5.7, water maze acquisition (Phase 1), despite linear du-ration-related trends in the data. Trials to criterionand 3.7 mm anterior to the IAL). Figure 8 shows that

only the ISC-20 group sustained damage beyond the in learning the non-matching-to-position task



FIG. 8. Mean cell loss grades ({ SD) in groups subjected to 4 VO for 5, 10 and 20 min (ISC-5, ISC-10, ISC-20, respectively) andcontrols (CON). There was no cortical (CTX) cell loss in any group. The ISC-10 group showed a minimal level of striatal (STR) damagedamage (5% cell loss) at levels 1 and 2. Maximal damage was seen in the CA1 sector at both at level 3 (averaging 85% in the ISC-20group), and at level 4 (averaging 75% in the ISC-20 group). Loss was linearly related to duration of occlusion, and only the ISC-10 andISC-20 groups showed significant differences from controls. The ISC-20 group also showed minor damage in the CA3 field (20–40% cellloss) at levels 3 and 4. SED bars show twice the standard errors for the difference in means for the main effect of Groups, wheresignificant. Difference from controls: **p õ .01.

(Phase 2) also showed no relationship with the ex- between water maze measures, there were no corre-lations between water maze parameters and trialstent of CA1 cell loss. However, latency scores for the

second period of training in the water maze (Phase to criterion on the DNMTP task, so that acquisitionof this task did not correlate either with CA1 damage3) and for the water maze working memory task

(Phase 4) showed a moderate (p õ .05) association or with water maze measures.with the extent of CA1 cell loss at level 3 and alsoat level 4 for acquisition (see Table 2A). Parameters DISCUSSIONof water maze acquisition and working memoryshowed strong internal relationships, as exemplified The results confirmed our previous findings (Netto

et al., 1993; Nunn et al., 1994; Hodges et al., in press;in Table 2B. Rats that found the platform rapidlyspent longer in the training quadrant, resulting in Nelson et al., in press), and those from other labora-

tories (Jaspers et al., 1990; Olsen et al., 1994), thatsubstantial negative correlations between these twoindices of learning. Moreover rats that performed rats subjected to global ischemia show impairment

in spatial learning in the water maze. Deficits werewell or poorly on one phase of water maze testingtended to perform similarly on others, resulting in seen more clearly in inappropriate search patterns

than in measures of latency or heading angle. How-strong positive correlations between the two phasesof acquisition and to a lesser extent, between acqui- ever, animals subjected to 20 min of 4 VO showed

increased latency in Phase 1, both the ISC-10 andsition latencies and overall latencies in the workingmemory task (see Table 2C). However, the correla- ISC-20 groups showed retarded rates of learning in

Phase 3, and all three ischemic groups showed im-tion between latency scores on Trial 2 of the workingmemory task and CA1 cell loss was not significant, paired recall of the platform position on Probe Trial

measures in Phase 3. These findings indicated thatand this trial also did not correlate with either Phase1 or Phase 3 of acquisition, so that the correlations cognitive deficits were present before both of the

working memory tasks. Since Phases 1 and 3 wereof mean latency over Trials 2–4 of working memorywith CA1 cell loss and acquisition may have reflected spaced 42 weeks apart, spatial learning was as-

sessed at two widely distributed intervals. The datarapid spatial learning, rather than working memoryper se. In contrast to the consistent relationships suggested that ischemic rats were more impaired



TABLE 2Correlations (rs) between Behavioral Measures

CA1 CA1Measure (level 3a) (level 4a)

(A) Correlations between behavioral measures and CA1 cell lossPhase 1, water maze acquisitionb: latency .137 .225Phase 2, delayed non-matching-to-position, trials to criterion .150 .260Phase 3, water maze acquisitionb latency .334* .436*Phase 4, water maze working memoryc: latency .384* .258

(B) Correlations between measures within behavioral testsPhase 1

Acquisition, latency vs. % time in training quadrant 0.889**Percentage of time in training quadrant, acquisition vs. probe .698**

Phase 3Acquisition, latency vs. % time in training quadrant 0.702**Percentage of time in training quadrant, acquisition vs. probe .407*

Phase 4Latency vs. heading angle .325*

(C) Correlations between measures on different testsPhase 1 vs. Phase 3, Latency .653**Phase 1 vs. Phase 3, % Time in training quadrant .618**Phase 1 vs. Phase 4, Latency .323*Phase 3 vs. Phase 4, Latency .341*Phase 1 vs. Phase 2, latency vs. trials to criterion .127Phase 3 vs. Phase 2, latency vs. trials to criterion .149Phase 4 vs. Phase 2, latency vs. trials to criterion .294

a Level 3 Å 5.7 mm, level 4 Å 3.7 mm anterior to the interaural line.b Acquisition scores used the means over the last 4 days of training.c Working memory scores used the means over Trials 2–4.* p õ .05, **p õ .01.

relative to controls on Phase 3 than on Phase 1, de- Working memory/rapid spatial learning in the wa-ter maze was more severely compromised in isch-spite prior experience. The ISC-10 group, for exam-

ple, showed no significant impairment on the first emic rats than acquisition in the standard proce-dure. All ischemic groups failed to match to the plat-phase, apart from increased thigmotaxis, whereas

in Phase 3, this group showed a retarded rate of form position as efficiently as controls and the ISC-20 group was highly impaired throughout. This sub-acquisition, spent less time in the training quadrant

and inner annuli than controls, and was impaired stantial impairment in spatial working memory inthe water maze contrasted with lack of deficit seenon the Probe Trial. Interactions between Phases 1

and 3 indicated that deficits were more strongly re- either in acquisition of DNMTP in the Skinner box,or in delay-dependent decreases in accuracy, relativelated to duration of occlusion in Phase 3 than Phase

1. Thus, marginal deficits in spatial learning appear to controls. Lack of impairment in DNMTP also con-trasts with the findings of Wood et al. (1993) in ro-to have increased over time, particularly in the ISC-

10 group. This evidence for progressive impairment dents, and Zola-Morgan et al. (1992) in monkeys, ofsubstantial delay-dependent impairments in ani-is consistent with suggestions that ischemic brain

damage is not static, but that neuronal degeneration mals subjected to global ischemia in nonspatial de-layed object recognition (DNMTS) tasks. Severaland reorganisation persists for some time after an

ischemic insult (Onadera, Aoki, Yae, & Kogure, task-related factors might account for these discrep-ancies in working memory performance; for exam-1990; Nakano, Kogure, & Fujikura, 1990). Although

death of pyramidal cells largely occurs within the ple, tasks may differ in spatial requirements, in sus-ceptibility to interference, in relational informationfirst week (Simon, Griffiths, Evans, Swan, & Mel-

drum, 1984) GABAergic interneurons, for example, processing, and in procedural demands. The vari-ability of ischemic lesions would also be likely toare still declining in the CA1 sector some 6–8 weeks

after global ischemia in gerbils (Fukuda, Nakano, affect performance.The foremost factor that might account for dis-Yoshiya, & Hashimoto, 1993).



crepancies in performance of working memory tasks DNMTP results and those of Wood et al., since weused only two similar stimuli (levers), while Wood etis that the water maze taxes allocentric spatial infor-

mation processing to a greater extent than recall of al. (1992, 1993) used a large number of trial-uniqueobjects, so that according to the interference theory,two lever positions in the Skinner box. There is a

considerable amount of evidence from lesion studies working memory deficits should have been greaterin the Skinner box than the DNMTS task. Similarly,that rats with hippocampal damage perform nor-

mally in tasks that require nonspatial working mem- working memory DNMTS performance was notfound to be impaired in animals with medial septalory, such as object or goal box recognition (Aggleton

et al., 1986; Rothblat & Cromer, 1991; Kelsey & Var- lesions when only two (i.e., interference prone) ob-jects were used (Kelsey & Vargas, 1993), whereasgas, 1993), but are significantly impaired in spatial

working memory tasks such as T- or Y-maze alterna- disruption was seen after hippocampal lesions(Mumby et al., 1992), when a large number of trial-tion. These results have led Eichenbaum Otto and

Cohen (1994) to propose that functions of the hippo- unique (i.e., discriminable) stimuli were used. Al-though the number and complexity of stimuli in de-campus are dissociable, with the CA fields and den-

tate gyrus involved in processing relational and spa- layed recognition tasks may affect the amount ofinterference, as Rawlins et al. (1993) suggest, thesetial information and the parahippocampal regions

storing information in working memory. Recent le- features may influence working memory perfor-mance in other ways. For example, tasks with trial-sion studies indicate that animals with perirhinal

cortex damage show pronounced delay-dependent unique stimuli may put heavy demands on relationalinformation processing, because the variety of stim-working memory deficits, which are more severe

than those in animals with hippocampal damage uli retards learning of the recognition rule (e.g.,‘‘choose the novel object’’), when all objects are rela-(Zola-Morgan, Squire, Amaral, & Suzuki, 1989;

Wiig & Bilkey, 1994; Mumby & Pinel, 1994). On this tively unfamiliar. Thus relational complexity, ratherthan spatial or interference factors, might accountview, the predominantly intrahippocampal cell loss

induced by global ischemia would be expected to infor deficits in ischemic rats in the DNMTS but notthe DNMTP procedure.volve spatial/relational deficits, but not impairment

of working memory per se, in line with the present Procedural differences may also have contributedto our failure to show working memory deficits in thefindings. Our results do, however, conflict with find-

ings of deficits in Skinner box DNMTP in animals Skinner box, as opposed to the water maze. Workingmemory in the water maze is very rapidly learned;with hippocampal or fornix lesions (Aggleton,

Hunt, & Shaw, 1990; Aggleton, Keith, & Sahgal, controls essentially showed one-trial learning. In theSkinner box controls required a mean of 82 sessions1991; Aggleton, Keith, Hunt, & Rawlins, 1992) and,

as noted, impaired object recognition in DNMTS of 40 trials to reach 80% accuracy. Slow learning ofthe procedural aspects of the Skinner box task maytasks in both rats and monkeys with ischemic hippo-

campal damage, indicative of working memory defi- therefore have masked early deficits in positionlearning that were exposed in the water maze. Previ-cits in both lever position and nonspatial working

memory tasks. Thus, features of the tasks other than ous findings (Netto et al., 1993) have shown thatischemic rats are not impaired in recall of spatialspatial requirements may interact with lesion or

ischemic hippocampal damage. information from long-term memory. Thus repeatedexposure to the Skinner box may have thoroughlyThe DNMTS tasks used in primate studies and by

Wood et al. (1992, 1993) may have been more sensi- familiarized the animals with the lever positions. Itis also possible that animals may have avoided usetive to ischemic hippocampal damage than DNMTP

in the Skinner box by being more prone to interfer- of working memory in the Skinner box by adoptinga mediating strategy such as a fixed position relativeence and/or by making heavier demands on ‘‘rela-

tional’’ information processing, which Eichenbaum to the lever to be chosen, which was not overcomeby the requirement to press the sample lever andet al. (1994) ascribe to the hippocampus. Rawlins,

Lyford, Seferiades, Deacon, and Cassaday (1993) retrieve a pellet. Mediating strategies, however,would result in comparable performance across de-found that rats with FF lesions and hippocampal

ablation were highly sensitive to interference in non- lays. The strong delay-related trends of impairmentin all the animals is suggestive of loss of informationspatial goal box recognition tasks, showing impair-

ments with simple and similar goal boxes, but not from working memory, rather than non-mnemonicstrategies.with a large set of quasi trial-unique boxes, which

were easy to discriminate. These findings, however, Correlations between cell loss and behavioralscores support the findings that ischemic damagedo not resolve the discrepancy between the present


FIG. 9. Nissl stained (Cresyl Violet/Luxol Fast Blue) coronal sections of the dorsal CA1 field (arrows) at c. 5.7 mm anterior to theinter-aural line, taken from a non-ischemic control rat (A) and rats subjected to global ischemia (4 VO) for 5 min (B), 10 min (C) and20 min (D). Increasing durations of ischemia resulted in progressively greater loss of CA1 pyramidal cells from 10% after 5 min to 65%after 10 min and almost total loss after 20 min, in comparison with controls. In control brains (A) rounded cells with well-defined nucleiwere densely packed. At low (B) and intermediate (C) levels of damage gaps appeared in the cell body layer, together with distortedand calcified cell remnants (arrowheads). With maximal damage (D) the CA1 cell layer was almost totally destroyed, and only a fewisolated nuclear cells were visible (arrowheads). cc: corpus callosum. Magnification 1 200.

242

04-19-97 04:11:35 nlma AP: NLM


FIG. 9—Continued

was associated with impairment in spatial water tween water maze acquisition scores and the extentof CA1 cell loss for Phase 1, there were modest corre-maze learning and working memory tasks, but not

in the less spatially demanding Skinner box task. lations between indices of water maze performanceand CA1 cell loss for Phases 2 and 3, when ischemicAlthough there were no significant correlations be-



deficits were most pronounced and which were closer significant correlations persisted. Thus, correlationsmay only be robust if the sample includes animalsto the time of histological assessment. These correla-

tions were obtained using a six-point grading scale, with substantial CA1 damage.Factors other than the extent of CA1 cell loss arewhich clusters the data to a greater extent than use

of cell counts and may therefore be more subject to also likely to contribute to behavioral impairments.Jaspers et al. (1990), for example, reported waterType 2 errors. Relationships would be more clearly

assessed by quantitative image analysis methods maze deficits in the absence of any visible CA1 cellloss in animals that underwent 24 min of 2 VO with-(Olsen et al., 1994) that permit accurate counts of

viable cells throughout the hippocampus. In contrast out hypotension, suggesting that functional changesalone may be sufficient to induce cognitive impair-to water maze scores, acquisition of the DNMTP

task, which was unimpaired in ischemic rats, was ment. Global ischemia induces widespread changesthroughout the hippocampus, including upregula-not related to the extent of CA1 cell loss. Findings

of correlations between and within the water maze tion of heatshock proteins, trophic factors, and recep-tor binding, and also changes in calcium channeltests, but not between water maze measures and

trials to criterion in the DNMTP task, support the conductances, neurotransmitter levels and secondmessenger systems (including increases in cytosolicsuggestion that the cognitive requirements of Skin-

ner box and water maze tasks were not closely re- inositol phosphate levels). These amount to substan-tial functional reorganization (Onadera, Sato, & Ko-lated. The lack of a correlation between DNMTP per-

formance and hippocampal damage contrasts with gure, 1987; Onadera, Araki, & Kogure, 1989; Onad-era & Kogure 1989; Schmidt-Kastner & Freund,the findings of Wood et al. (1993), who report a sig-

nificant relationship between the number of surviv- 1991; Siekluca, Heim, Sontag, & Osborne, 1991).Thus, cognitive deficits may relate to more subtleing neurons at six levels in the CA1 field and object

recognition scores (trials to criterion and perfor- neural changes than the death of CA1 cells. In addi-tion there is loss of nonpyramidal cell types, notablymance at a 30-s delay interval) in the DNMTS proce-

dure. However, Wood et al. included controls so that somatostatin-positive cells in the hilus and GABAer-gic interneurons. These changes might also contrib-these correlations need to be reexamined without

this bias. ute to cognitive dysfunction, although they wouldnot be detected by conventional Nissl cell stains.Findings of a modest relationship between CA1

cell loss and spatial learning and working memory Furthermore, at longer durations of occlusion, wherebehavioral correlations with CA1 cell loss appear toin the water maze are in agreement with results

from a variety of studies using spatial tasks, includ- be more robust, extrahippocampal damage is alsolikely to occur, which may contribute to cognitiveing the split stem T-maze (Volpe, Davis, Towle, &

Dunlap, 1992), the radial maze (Kiyota et al., 1991) impairment. Nunn and Jarrard (1994), for example,used silver staining to detect loss of cells in the cin-and water maze learning set task (Rod, Whishaw, &

Auer, 1990), and standard water maze acquisition gulate cortex following 15 min of 4 VO, an areashown by lesion studies to contribute to spatial(Olsen et al., 1994), and replicate some of our own

previous results (Nelson, Lebessi, Sowinski, & learning in the water maze (Marston, Everitt, &Robbins, 1993).Hodges, in press). However, a number of these stud-

ies also included nonischemic controls in the correla- In conclusion, the results suggested that ischemicrats showed duration-related impairment in watertions, making the results difficult to interpret. Fur-

thermore, results from our laboratory to date have maze acquisition and working memory tasks, butnot in the less spatially demanding working memorybeen mixed. Nunn et al. (1994) used periods of 5–

30 min of 4 VO occlusion to obtain a spread of values DNMTP task. The extent of CA1 cell loss correlatedwith water maze deficits, but not with trials to crite-in which cell loss was related to duration of occlusion

by a strong linear trend, but no significant relation- rion in DNMTP. These results are consistent withthe suggestion that intrahippocampal damage in-ship between the extent of CA1 cell loss and several

water maze scores was seen. Nunn et al. did not duced by 4 VO disrupted the processing of spatialinformation, but did not impair storage of informa-include animals with maximal (grade 5) cell loss in

the correlations to increase the variability of scores. tion in working memory. Therefore, the workingmemory tasks that are most likely to detect cognitiveNelson et al. (in press) therefore examined correla-

tions between CA1 cell loss and water maze acquisi- deficits following global ischemia are those involvinga substantial spatial component, such as the T-maze,tion in animals with and without maximal Grade 5

damage and found that the relationship was weaker radial maze, and water radial maze. Interestingly,even apparently spatial tasks may not be disruptedin animals with submaximal damage, though some



retinin-containing neurons in the rat hippocampus after isch-following global ischemia if place learning can beemia. Neuroscience, 56, 581–596.achieved by use of associative cues rather than allo-

Fukuda, T., Nakano, S., Yoshiya, I., & Hashimoto, P. H. (1993).centric spatial information. We have found that isch-Persistent degenerative state of non-pyramidal neurons inemic rats were not impaired in Jarrard’s ‘‘place’’ and the CA1 region of the gerbil hippocampus following transient

‘‘cue’’ tasks (Jarrard, 1993) in the radial maze, where forebrain ischemia. Neuroscience, 53, 23–38.intramaze cues provided intentional (arm inserts) or Gionet, T. X., Thomas, J. D., Warner, D. S., Goodlett, C. R., Was-inadvertent (odors, nicks, and scratches) local land- serman, E. A., & West, J. R. (1991). Forebrain ischemia in-

duces selective behavioral impairments associated with hip-marks for associative place learning (Nunn et al.,pocampal injury in rats. Stroke, 22, 1040–1047.1991; Nunn and Hodges, 1994). However, robust

Graham, D. I. (1992). Hypoxia and vascular disorders. In J.deficits were seen in the water radial maze, used toHulme Adams & L. W. Duchen (Eds.), Greenfield’s neuropa-eliminate local cues such as odors (Nelson et al., inthology (pp. 153–286). London: Edward Arnold.

press; Hodges, 1996). These studies further suggestHagan, B. J., & Beaughard, M. (1990). The effects of forebrain

that ischemic rats show marked sensitivity to spe- ischemia on spatial learning. Behavioural Brain Research,cific spatial requirements of cognitive tasks. 41, 151–160.

Hodges, H., Sowinski, P., Fleming, P., Kershaw, T. R., Sinden,J. D., Meldrum, B. S., & Gray, J. A. (1993). Foetal CA1 graftsREFERENCESselectively alleviate ischemic deficits in spatial learning andworking memory. Journal of Cerebral Blood Flow and Metab-

Aggleton, J. P., Hunt, P. R., & Rawlins, J. N. P. (1986). Effects of olism, 13(Suppl. 1), S51.hippocampal lesions upon spatial and non-spatial tests of

Hodges, H., Sinden, J., Meldrum, B. S., & Gray, J. A. (1994). Cere-working memory. Behavioural Brain Research, 19, 133–146.

bral transplantation in animal models of ischaemia. In S. B.Aggleton, J. P., Hunt, P. R., & Shaw, C. (1990). The effects of Dunnett & A. Bjorklund (Eds.), Functional neural trans-

mamillary body and combined amygdalar-fornix lesions on plantation (pp. 347–385). New York: Raven Press.tests of delayed non-matching-to-sample in the rat. Behav-

Hodges, H. (1996). Maze Procedures: The radial maze and waterioural Brain Research, 40, 145–157.

maze compared. Cognitive Brain Research, 3, 167–181.Aggleton, J. P., Keith, A. B., & Sahgal, A. (1991). Both fornix and Hodges, H., Sowinski, P., Fleming, P., Kershaw, T. R., Sinden,

anterior thalamic, but not mamillary, lesions disrupt delayed J. D., Meldrum, B. S., & Gray, J. A. (1996). Contrasting ef-non-matching-to-position memory in rats. Behavioural Brain fects of foetal CA1 and CA3 hippocampal grafts on deficitsResearch, 44, 151–161. in spatial learning and working memory induced by global

Aggleton, J. P., Keith, A. B., Rawlins, J. N. P., Hunt, P. R., & Sah- cerebral ischaemia in rats. Neuroscience, 72, 959–988.gal, A. (1992). Removal of the hippocampus and transection Hsu, M., & Buzsaki, G. (1993). Vulnerability of mossy fiber targetsof the fornix produce comparable deficits on delayed non- in the rat hippocampus to forebrain ischemia. Journal of Neu-matching-to-position by rats. Behavioural Brain Research, roscience, 13, 3964–3979.52, 61–71.

Jarrard, L. E. (1993). Review: On the role of the hippocampusAuer, R. N., Jensen, M. L., & Whishaw, I. Q. (1989). Neurobehav- in learning and memory in the rat. Behavioral and Neural

ioral deficit due to ischemic brain damage limited to half of Biology, 60, 9–26.the CA1 sector of the hippocampus. Journal of Neuroscience,

Jaspers, R. M. A., Block, F., Heim, C., & Sontag, K-H. (1990).9, 1641–1647.Spatial learning is affected by transient occlusion of common

Bachevalier, J., & Mishkin, M. (1989). Mnemonic and neuropath- carotid arteries (2 VO): Comparison of behavioural and histo-ological effects of occluding the posterior cerebral artery in pathological changes after ‘2 VO’ and ‘four vessel occlusion’macaca mulatta. Neuropsychologia, 27, 83–105. in rats. Neuroscience Letters, 117, 149–153.

Cummins, J. L., Tomiyasu, U., Read, S., & Benson, D. F. (1984). Kartsounis, L. D., Rudge, P., & Stevens, J. M. (1995). BilateralAmnesia with hippocampal lesions after cardiopulmonary ar- CA1 and CA2 fields of the hippocampus are sufficient to causerest. Neurology, 34, 679–681. a severe amnesic syndrome in humans. Journal of Neurology,

Davis, H. P., Tribuna, J., Pulsinelli, W. A., & Volpe, B. T. (1986). Neurosurgery and Psychiatry, 59, 95–98.Reference and working memory of rats following ischemic Kelsey, J. E., & Vargas, H. (1993). Medial septal lesions disrupthippocampal damage. Physiology and Behavior, 37, 387–392. spatial but not nonspatial working memory in rats. Behav-

Davis, H. P., & Volpe, B. T. (1990). Memory performance after ioral Neuroscience, 107, 565–574.ischemic or neurotoxin damage of the hippocampus. In L. R. Kiyota, Y., Miyamoto, M., & Nagaoka, A. (1991). RelationshipSquire & E. Lindenlaub (Eds.), The biology of memory (pp. between brain damage and memory impairment in rats ex-477–507), Symposia Medica Hoechst 23. Stuttgart, New posed to transient forebrain ischemia. Brain Research, 538,York: Schattauer Verlag. 295–302.

Eichenbaum, H., Stewart, C., & Morris, R. G. M. (1991). Hippo- Markowska, A. L., Olton, D. S., Murray, E. A., & Gaffan, D.campal representation in place learning. Journal of Neurosci- (1989). A comparative analysis of the role of the fornix andence, 10, 3531–3542. cingulate cortex in memory: Rats. Experimental Brain Re-

Eichenbaum, H., Otto, T., & Cohen, N. J. (1995). Two functional search, 74, 187–201.components of the hippocampal memory system. Behavioral Matsuyama, T., Tsuchiyama, M., Nakamura, H., Matsumoto,and Brain Sciences, 17, 449–518. M., & Sugita, M. (1993). Hilar somatostatin neurons are more

vulnerable to an ischemic event than CA1 pyramidal neu-Freund, T. F., & Magloczky, Z. (1993). Early degeneration of cal-



rons. Journal of Cerebral Blood Flow and Metabolism, 13, pus, space and memory. Behavioral and Brain Sciences, 2,313–365.299–234.

Morris, R. G. M., Garrud, P., Rawlins, J. N. P., & O’Keefe, J. Onodera, H., Sato, G., & Kogure, K. (1987). Quantitative autora-(1982). Place navigation is impaired in rats with hippocampal diographic analysis of muscarinic cholinergic and adenosinelesions. Nature, 297, 681–683. A1 binding sites after transient forebrain ischemia in the

gerbil. Brain Research, 415, 309–322.Moser, E., Moser, M-B., & Andersen, P. (1993). Spatial learningimpairment parallels the magnitude of dorsal hippocampal Onodera, H., Araki, T., & Kogure, K. (1989). Protein Kinase Clesions, but is hardly present following ventral lesions, Jour- activity in the rat hippocampus after forebrain ischemia: au-nal of Neuroscience, 13, 3916–3925. toradiographic analysis by [3H]phorbal 12,13-dibutyrate.

Brain Res. 481, 1–7.Mumby, D. G., Wood, E. R., & Pinel, J. P. J. (1992). Object recog-nition memory is only mildly impaired in rats with lesions Onodera, H., & Kogure, K. (1989). Mapping second messengerof the hippocampus and amygdala. Psychobiology, 20, 18– systems in the rat hippocampus after transient forebrain27. ischemia: in vitro [3H]forskolin and [3H]inositol, 4,5,-triphos-

phate binding. Brain Research, 487, 343–349.Mumby, D. G., & Pinel, J. P. J. (1994). Rhinal cortex lesions andobject recognition in rats. Behavioral Neuroscience, 108, 11– Onodera, H., Aoki, H., Yae, T., & Kogure, K. (1990). Post-ischemic18. synaptic plasticity in the rat hippocampus after long-term

survival: A histochemical and autoradiographic study. Neuro-Murray, E. A., Davidson, M., Gaffan, D., Olton, D. S., & Suomi,science, 38, 125–136.S. (1989). Effects of fornix transection and cingulate cortical

ablation on spatial memory in rhesus monkeys. Experimental Petito, C. K., Feldman, E., Pulsinelli, W. A., & Plum, F. (1987).Brain Research, 74, 173–186. Delayed neuronal hippocampal damage in humans following

cardiorespiratory arrest. Neurology, 37, 1281–1286.Nakano, S., Kogure, K., & Fujikura, H. (1990). Ischemia-inducedslowly progressive damage in the rat brain. Neuroscience, 38, Pulsinelli, W. A., Brierley, M. D., & Plum, F. (1982). Temporal115–124. profile of neuronal damage in a model of transient forebrain

ischemia. Annals of Neurology, 11, 491–498.Nelson, A. J., & Hodges, H. (1995). Global cerebral ischaemiadisrupts learning and working memory in spatial but not in Raffaele, K. C., & Olton, D. S. (1988). Hippocampal and amygda-nonspatial tasks. Journal of Cerebral Blood Flow Metabo- loid involvement in working memory for nonspatial stimuli.lism, 15(Suppl. 1), S196. Behavioral Neuroscience, 102, 349–355.

Nelson, A., Lebessi, A., Sowinski, P., & Hodges, H. (in press) Rawlins, J. N. P. (1985). Associations across time: The hippocam-Comparison of effects of global ischaemia on spatial learning pus as a temporary memory store. Behavioral and Brain Sci-in the standard and radial water maze: Relationship of hippo- ences, 8, 479–496.campal damage to performance. Behavioural Brain Research. Rawlins, J. N. P., Lyford, G. L., Seferiades, A., Deacon, R. M. J., &

Netto, C. A., Hodges, H., Sinden, J. D., Le Peillet, E., Kershaw, Cassaday, H. J. (1993). Critical determinants of nonspatialT., Sowinski, P., Meldrum, B. S., & Gray, J. A. (1993). Effects working memory deficits in rats with conventional lesionsof foetal hippocampal field grafts on ischaemic-induced defi- of the hippocampus or fornix. Behavioral Neuroscience, 107,cits in spatial navigation in the water maze. Neuroscience, 420–433.54, 69–92. Rod, M. R., Whishaw, I. Q., & Auer, R. N. (1990). The relationship

Nunn, J. A., Le Peillet, E., Netto, C. A., Sowinski, P., Hodges, H., of structural ischemic brain damage to neurobehaviouralMeldrum, B. S., & Gray, J. A. (1991). CA1 cell loss produces deficit: The effect of postischemic MK-801. Canadian Journaldeficits in the water maze but not in the radial maze. Society of Psychology, 44, 196–209.for Neuroscience Abstracts, 17, 108. Roine, R. O., Kajaste, S., & Kaste, M. (1992). Neuropsychological

Nunn, J. A., LePeillet, E., Netto, C. A., Hodges, H., Gray, J. A., & sequelae of cardiac arrest. Journal of the American MedicalMeldrum, B. S. (1994). Global ischaemia: Hippocampal pa- Association, 269, 237–242.thology and spatial deficits in the water maze. Behavioural Rothblat, L. A., & Kromer, L. F. (1991). Object recognition mem-Brain Research, 62, 41–54. ory in the rat: The role of the hippocampus. Behavioural

Nunn, J. A., & Hodges, H. (1994). Cognitive deficits induced by Brain Research, 42, 25–32.global cerebral ischaemia: Relationship to brain damage and Scheller, M. S., Grafe, M. R., Zornow, M. H., & Fleischer, J. H.reversal by transplants. Behavioural Brain Research, 65, 1– (1992). Effects of ischemia duration on neurological outcome.31. CA1 histopathology and nonmatching to sample learning in

Nunn, J. A., & Jarrard, L. E. (1994). Silver impregnation reveals monkeys. Stroke, 23, 1471–1478.neuronal damage in cingulate cortex following 4 VO isch- Schmidt-Kastner, R., & Freund, T. F. (1991). Selective vulnerabil-aemia in the rat. NeuroReport, 5, 2363–2365. ity of the hippocampus in brain ischaemia. Neuroscience, 40,

Olsen, G. M., Scheel-Kruger, J., Moller, A., & Jensen, L. H. (1994). 599–636.Relation of spatial learning of rats in the Morris water maze Siekluca, M., Heim, C., Sontag, K-H., & Osborne, N. N. (1991).task to the number of viable CA1 neurons following four ves- Transient occlusion of rat carotid arteries increases forma-sel occlusion. Behavioral Neuroscience, 108, 681–690. tion of inositol phosphate: Evidence for specific effect on a-1-

O’Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive receptors. Neurochemistry International, 18, 175–189.map. Oxford: Clarendon Press. Simon, R. P., Griffiths, T., Evans, M. C., Swan, J. H., & Meldrum,

O’Keefe, J. (1983). Spatial memory within and without the hippo- B. S. (1984). Calcium overload in selectively vulnerable neu-campal system. In W. Seifert (Ed.), Neurobiology of the hippo- rons of the hippocampus during and after ischemia: An elec-campus (pp. 375–403). London: Academic Press. tron microscopy study in the rat. Journal of Cerebral Blood

Flow and Metabolism, 4, 850–852.Olton, D. S., Becker, J. T., & Handelman, G. E. (1979). Hippocam-



Volpe, B. T., Pulsinelli, W. A., Tribuna, J., & Davis, H. P. (1984). Wood, E. R., Bussey, T. J., & Phillips, A. G. (1993). A glycine an-tagonist 7-chlorokynurenic acid attenuates ischemia-inducedBehavioral performance of rats following transient forebrain

ischemia. Stroke, 15, 558–562. learning deficits. NeuroReport, 4, 151–154.Volpe, B. T., & Petito, C. K. (1985). Dementia with bilateral me- Zola-Morgan, S., Squire, L. R., & Amaral, D. G. (1986). Human

dial temporal lobe ischemia. Neurology, 35, 1793–1797. amnesia and the medial temporal region: Enduring memoryimpairment following a bilateral lesion limited to field CA1Volpe, B. T., Davis, H. P., Towle, A., & Dunlap, W. P. (1992). Lossof the hippocampus. Journal of Neuroscience, 6, 2950–2967.of hippocampal CA1 neurons correlates with memory impair-

ment in rats with ischemic or neurotoxin lesions. Behavioral Zola-Morgan, S., Squire, L. R., Amaral, D. G., & Suzuki, W.Neuroscience, 106, 457–464. (1989). Lesions of perirhinal and parahippocampal cortex

that spare the amygdala and hippocampal formation produceWiig, K. A., & Bilkey, D. K. (1994). Perirhinal cortex lesions insevere memory impairment. Journal of Neuroscience, 9,rats disrupt performance in a spatial DNMS task. NeuroRe-4355–4370.port, 5, 1405–1408.

Wood, E. R., Mumby, D. G., Pinel, J. P. J., & Phillips, A. G. (1993). Zola-Morgan, S., Squire, L. R., Rempel, N. L., Clower, R. P., &Amaral, D. G. (1992). Enduring memory impairment in mon-Impaired object recognition memory in rats following isch-

emia-induced damage to the hippocampus. Behavioral Neu- keys after ischemic damage to the hippocampus. Journal ofNeuroscience, 12, 2582–2596.roscience, 107, 51–62.


Differential Effects of Global Ischemia on Delayed Matching- and Non-Matching-to-Position Tasks in...

Documents

Transcript of Differential Effects of Global Ischemia on Delayed Matching- and Non-Matching-to-Position Tasks in...