Physiology & Behavior, Vol. 47, pp. 471-476. ©Pergamon Press plc, 1990. Printed in the U.S.A.
0031-9384/90 $3.00 + .00
Mediodorsal Thalamic Lesions Impair "Reference" and "Working" Memory in Rats K A T H R Y N A. S T O K E S 1 A N D P H I L L I P J. B E S T 2
Department o f Psychology, University o f Virginia, Charlottesville, VA 22903 R e c e i v e d 12 June 1989
STOKES, K. A. AND P. J. BEST. Mediodorsal thalamic lesions impair "reference" and "working" memory in rats. PHYSIOL BEHAV 47(3) 471--476, 1990. --The present experiment evaluated the ability of rats with lesions in the thalarnic mediodorsal nucleus (MD) to perform a task which differentiates between "reference" and "working" memory. In this task, only four of the arms of an 8-arm radial maze were baited, and animals were to restrict their entries to arms which were baited and to avoid never-baited arms. Despite several postoperative acquisition trials, rats with MD lesions did not acquire the task to a degree comparable to control subjects. Subjects with lesions continued to enter never-baited arms (reference memory errors) and to reenter baited arms (working memory errors). Given the lack of specificity in the behavioural impairment, the reference-working memory distinction seems to be an inappropriate one for characterizing the MD lesion deficit in rats. This deficit may involve an inability to use environmental stimuli to distinguish among arms of the maze, or an alteration in motor mechanisms. Mediodorsal thalamus
Reference memory
Working memory
Radial maze
Lesion study
sentational" memory. This treatment produces specific diencephalic degeneration, including the ventral MD area. It has become critical, then, to address specifically the findings of Kolb et al. (5). These authors used a version of the radial maze task in which only four of the eight arms were baited with food. This task was developed and utilized by Olton and colleagues to distinguish between "reference" and " w o r k i n g " memory (2,7). Reference memory refers to acquired information that can be used to generate rules for behaviour across situations or trials, and at different points in time. Working memory applies to information useful for a limited amount of time or for only the present situation or the current trial. Efficient performance of the radial maze task in which only four arms are baited requires both entry to those arms that are baited, and avoidance of those arms that are never baited. Learning to avoid never-baited arms, then, represents acquisition of a reference memory rule; avoiding reentry of already-entered arms represents the operation of working memory for that trial. Whether or not this division of memory capacities is a valuable theoretical distinction, the task provides a useful framework to examine the different types of errors made by animals on the 8-arm radial maze. The animals must discriminate among baited and unbaited arms, and among visited and unvisited arms for optimum solution. Use of this task will enable us to determine whether we can replicate the results of Kolb et al. (5) in this laboratory. Further, if a deficit is found, its nature may be elaborated with respect to the type of information that is retained or lost in animals with MD lesions.
RECENT experiments indicate that the mediodorsal nucleus of the thalamus (MD) may be involved in memory processes in humans. Damage to the mediodorsal nucleus in humans by infarct or injury produces an amnesic syndrome [e.g., (1, 13, 14)]. Further, the memory disorder found in patients with Korsakoff's disease has been related to damage in the diencephalon, particularly the MD nucleus (18). In the rat, however, disruption of spatial memory has not always occurred following MD lesions. For example, Kolb and colleagues (5) found MD-lesioned rats to be unimpaired in postoperative acquisition of the Morris water maze task and of a radial arm maze task. Similarly, Rasmussen and colleagues (10) found no deficits on spatial or nonspatial variants of a radial maze task in animals with MD lesions. These animals did, however, show poor acquisition of an egocentric alternation task. Further, Kessler, Markowitsch and Otto (4) found only subtle effects of MD lesion on radial maze acquisition. A deficit appeared only when a delay was inserted between the fourth and fifth choices. In contrast, we have recently found MD-lesioned rats to exhibit impaired radial maze performance, despite extensive preoperative training (16). This deficit occurred not only when visual stimuli in the extramaze environment were limited, but also when extramaze cues were abundant (17). Further, in both situations, the performance of MD-lesioned subjects was characterized by the development of rigid response patterns. In support of these findings, Mair (6) found that rats given thiamine deficient diets, especially when combined with injection of the thiamine antagonist, pyrithiamine, were impaired on tasks considered to constitute "repre-
lRequests for reprints should be addressed to Kathryn A. Stokes at her current address: Department of Psychology, University of Western Ontario, London, Ontario, N6A 5C2, Canada. 2Current address: Department of Psychology, University of New Orleans, Lakefront, New Orleans, LA 70148.
471
472
STOKES AND BEST
METHOD
Subjects A total of 21 male Long-Evans hooded rats were used for this experiment. At the time of surgery, they weighed 300-350 g. Animals were housed individually in standard laboratory cages in a vivarium in which lights were on between 0800 and 2200 hr. Food and water were available ad lib until the beginning of the experiment. At this time, all animals were placed on a restrictedfeeding schedule to reduce body weight to 85% of free-feeding weight. Behavioural tests were conducted in the latter portion of the light cycle and were followed by access to food. Water remained available to the subjects at all times. Eleven of the subjects underwent lesion of the MD nucleus, while the remaining 10 served as sham-operated controls.
Apparatus The radial 8-arm maze used in this experiment has been described previously (14). Briefly, the maze was constructed of wood and painted a fiat gray. The center portion was 25 cm wide, and each of the 8 arms was 50 cm long and 9 cm wide, with Plexiglas walls, 5 cm high, angling outward from each arm. Metal food cups were placed at the end of each arm to hold the reinforcer. Confinement to the central area was accomplished using a remotely operated system of doors. The maze was situated within an equipment storage room that contained numerous visual cues (tables, cabinets, equipment and the experimenter).
Surgery Animals in the present experiment received chemical lesions of the MD nucleus, using ibotenic acid, an excitotoxin. Ibotenate selectively destroys cell bodies and dendrites, but leaves myelinated fibres of passage intact (12). Thus, it provides an advantage over conventional electrolytic procedures in terms of lesion specificity. All subjects were anesthetized with Chloropent anesthesia (3.0 mg/kg, IP). Using a glass micropipette attached to a stereotaxic instrument, 0.3 I~1 of ibotenate (5 Ixg per ~1 phosphate-buffered saline) was microinjected into four sites within the MD of 7 animals. Bilateral injections were made at the following coordinates, taken from the atlas of Paxinos and Watson (9): - 2.2 and - 2 . 8 mm posterior to Bregma, ---0.5 mm lateral to the midline and - 5 . 2 mm below the surface of the brain. Injections were made over a 3-min period, and the micropipette was left in place for a further 3 min to allow for diffusion of the ibotenate. The micropipette was lowered to the appropriate depth in shamlesioned subjects (n = 10), but no injection of ibotenate was made. The pipette was left in place for 1 min at each site in sham-lesioned subjects. Gelfoam covered the drilling site, and the wound was sutured closed. Animals receiving the ibotenate were carefully monitored during recovery from anesthesia and for the subsequent two weeks, to ensure full recovery of feeding and drinking after surgery. An additional four animals were given electrolytic lesion of MD, for comparison purposes. The coordinates and procedure used for these lesions have been detailed previously (16,17).
Handling and Adaptation During the second postoperative week, the subjects were placed on the food deprivation schedule. On each day of deprivation, the animals were handled, weighed, and given access to food. On Days 5 and 6 of deprivation, they were adapted to the
reinforcer in the home cage. They received pieces of "Froot Loops" cereal, placed in food cups similar to those found on the maze. The animals were introduced to the radial maze on Day 7. After being placed on the central platform, they explored the maze for 10 min and retrieved pieces of cereal strewn around and in the food cups of only those arms to contain reinforcement during training. A second session of exploration occurred the following day.
Training Subsequent to the second adaptation session, animals were trained on the radial maze, with only four of the eight arms baited. The arms to be baited for each subject were determined prior to pretraining. The configuration of baited arms was different for each animal within each group, but configurations were matched between the groups and were constant across days. Further, no more than 2 adjacent arms were baited, and the pattern of baited arms was never such that the task could be solved using an obvious response pattern (e.g., turning at 90 or 135 degree angles). One trial was administered per day, 6 to 7 days a week, for a total of 60 trials. Daily trials were terminated upon the consumption of the fourth piece of cereal. One subject with MD lesion exceeded a 10-minute period to complete one trial. The trial was terminated at that point and was excluded from analysis. Training continued until 60 trials were administered, regardless of level of performance at that time.
Histology Upon completion of testing, the subjects were sacrificed with an overdose of anesthetic and perfused intracardially with normal saline followed by 10% phosphate-buffered formalin. The brains were extracted, soaked in 30% sucrose formalin, then frozen 40-ix sections were made through the lesioned area and stained with cresyl violet. Microscopic examination of the brains were made, in order to verify the extent of toxin-induced cell damage. Reconstructions were made of the lesion extent and any damage outside the MD area was noted.
Data Analysis Note was taken of the total number of errors made, as well as of the type of error committed. Reentries into previously baited arms were counted as working memory-correct (WM-C) errors, while entries into arms that were never baited were considered reference memory errors. Further, a separate record was taken of reentries into never-baited arms. Note that such an error represents both a working memory and reference memory error. These errors were termed working memory-incorrect (WM-I) errors, but were also counted in the total number of reference memory errors. These measures were analyzed parametrically with analysis of variance for repeated measures techniques. RESULTS
Generally, MD-lesioned animals were deficient in acquisition of both the reference and working memory portions of the task. Their performance improved on some measures, but never approximated that of sham control animals, despite the administration of 60 postoperative trials. The performance of electrolytically lesioned animals was comparable to that of chemically lesioned animals, so data for all lesioned subjects were pooled in the analyses.
Reference Memory Errors For all animals, reference errors (entries into never-baited
MD LESIONS AND MEMORY TYPE
REFERENCE ERRORS
473
WORKING MEMORY CORRECT
WORKING MEMORY INCORRECT
2
2
5 3-
o4 ~3 1
. . . . . . . . . . . . .
2
4
6
8
0
10 12
BLOCKS OF 5 TRIALS
4
6
8
10 12
4
6
8
10 12
BLOCKS OF 5 TRIALS
FIG. 1. Numberof errors made by sham-operatedrats and those with MD lesions, presented according to error type, during acquisition trials. Regardless of error type (reference, working-corrector working-incorrect), control subjects improvedrapidly, making few to no errors at the end of testing. MD-lesioned subjects showed some improvement in reference memory errors, but none in working memory errors, and always made many more errors than did control subjects.
arms) composed the greatest proportion of the total number of errors made (see Fig. 1). Sham-lesioned animals exhibited a significant reduction in reference errors over trials, F(5,40)-18.0, p
0031-9384/90 $3.00 + .00
Mediodorsal Thalamic Lesions Impair "Reference" and "Working" Memory in Rats K A T H R Y N A. S T O K E S 1 A N D P H I L L I P J. B E S T 2
Department o f Psychology, University o f Virginia, Charlottesville, VA 22903 R e c e i v e d 12 June 1989
STOKES, K. A. AND P. J. BEST. Mediodorsal thalamic lesions impair "reference" and "working" memory in rats. PHYSIOL BEHAV 47(3) 471--476, 1990. --The present experiment evaluated the ability of rats with lesions in the thalarnic mediodorsal nucleus (MD) to perform a task which differentiates between "reference" and "working" memory. In this task, only four of the arms of an 8-arm radial maze were baited, and animals were to restrict their entries to arms which were baited and to avoid never-baited arms. Despite several postoperative acquisition trials, rats with MD lesions did not acquire the task to a degree comparable to control subjects. Subjects with lesions continued to enter never-baited arms (reference memory errors) and to reenter baited arms (working memory errors). Given the lack of specificity in the behavioural impairment, the reference-working memory distinction seems to be an inappropriate one for characterizing the MD lesion deficit in rats. This deficit may involve an inability to use environmental stimuli to distinguish among arms of the maze, or an alteration in motor mechanisms. Mediodorsal thalamus
Reference memory
Working memory
Radial maze
Lesion study
sentational" memory. This treatment produces specific diencephalic degeneration, including the ventral MD area. It has become critical, then, to address specifically the findings of Kolb et al. (5). These authors used a version of the radial maze task in which only four of the eight arms were baited with food. This task was developed and utilized by Olton and colleagues to distinguish between "reference" and " w o r k i n g " memory (2,7). Reference memory refers to acquired information that can be used to generate rules for behaviour across situations or trials, and at different points in time. Working memory applies to information useful for a limited amount of time or for only the present situation or the current trial. Efficient performance of the radial maze task in which only four arms are baited requires both entry to those arms that are baited, and avoidance of those arms that are never baited. Learning to avoid never-baited arms, then, represents acquisition of a reference memory rule; avoiding reentry of already-entered arms represents the operation of working memory for that trial. Whether or not this division of memory capacities is a valuable theoretical distinction, the task provides a useful framework to examine the different types of errors made by animals on the 8-arm radial maze. The animals must discriminate among baited and unbaited arms, and among visited and unvisited arms for optimum solution. Use of this task will enable us to determine whether we can replicate the results of Kolb et al. (5) in this laboratory. Further, if a deficit is found, its nature may be elaborated with respect to the type of information that is retained or lost in animals with MD lesions.
RECENT experiments indicate that the mediodorsal nucleus of the thalamus (MD) may be involved in memory processes in humans. Damage to the mediodorsal nucleus in humans by infarct or injury produces an amnesic syndrome [e.g., (1, 13, 14)]. Further, the memory disorder found in patients with Korsakoff's disease has been related to damage in the diencephalon, particularly the MD nucleus (18). In the rat, however, disruption of spatial memory has not always occurred following MD lesions. For example, Kolb and colleagues (5) found MD-lesioned rats to be unimpaired in postoperative acquisition of the Morris water maze task and of a radial arm maze task. Similarly, Rasmussen and colleagues (10) found no deficits on spatial or nonspatial variants of a radial maze task in animals with MD lesions. These animals did, however, show poor acquisition of an egocentric alternation task. Further, Kessler, Markowitsch and Otto (4) found only subtle effects of MD lesion on radial maze acquisition. A deficit appeared only when a delay was inserted between the fourth and fifth choices. In contrast, we have recently found MD-lesioned rats to exhibit impaired radial maze performance, despite extensive preoperative training (16). This deficit occurred not only when visual stimuli in the extramaze environment were limited, but also when extramaze cues were abundant (17). Further, in both situations, the performance of MD-lesioned subjects was characterized by the development of rigid response patterns. In support of these findings, Mair (6) found that rats given thiamine deficient diets, especially when combined with injection of the thiamine antagonist, pyrithiamine, were impaired on tasks considered to constitute "repre-
lRequests for reprints should be addressed to Kathryn A. Stokes at her current address: Department of Psychology, University of Western Ontario, London, Ontario, N6A 5C2, Canada. 2Current address: Department of Psychology, University of New Orleans, Lakefront, New Orleans, LA 70148.
471
472
STOKES AND BEST
METHOD
Subjects A total of 21 male Long-Evans hooded rats were used for this experiment. At the time of surgery, they weighed 300-350 g. Animals were housed individually in standard laboratory cages in a vivarium in which lights were on between 0800 and 2200 hr. Food and water were available ad lib until the beginning of the experiment. At this time, all animals were placed on a restrictedfeeding schedule to reduce body weight to 85% of free-feeding weight. Behavioural tests were conducted in the latter portion of the light cycle and were followed by access to food. Water remained available to the subjects at all times. Eleven of the subjects underwent lesion of the MD nucleus, while the remaining 10 served as sham-operated controls.
Apparatus The radial 8-arm maze used in this experiment has been described previously (14). Briefly, the maze was constructed of wood and painted a fiat gray. The center portion was 25 cm wide, and each of the 8 arms was 50 cm long and 9 cm wide, with Plexiglas walls, 5 cm high, angling outward from each arm. Metal food cups were placed at the end of each arm to hold the reinforcer. Confinement to the central area was accomplished using a remotely operated system of doors. The maze was situated within an equipment storage room that contained numerous visual cues (tables, cabinets, equipment and the experimenter).
Surgery Animals in the present experiment received chemical lesions of the MD nucleus, using ibotenic acid, an excitotoxin. Ibotenate selectively destroys cell bodies and dendrites, but leaves myelinated fibres of passage intact (12). Thus, it provides an advantage over conventional electrolytic procedures in terms of lesion specificity. All subjects were anesthetized with Chloropent anesthesia (3.0 mg/kg, IP). Using a glass micropipette attached to a stereotaxic instrument, 0.3 I~1 of ibotenate (5 Ixg per ~1 phosphate-buffered saline) was microinjected into four sites within the MD of 7 animals. Bilateral injections were made at the following coordinates, taken from the atlas of Paxinos and Watson (9): - 2.2 and - 2 . 8 mm posterior to Bregma, ---0.5 mm lateral to the midline and - 5 . 2 mm below the surface of the brain. Injections were made over a 3-min period, and the micropipette was left in place for a further 3 min to allow for diffusion of the ibotenate. The micropipette was lowered to the appropriate depth in shamlesioned subjects (n = 10), but no injection of ibotenate was made. The pipette was left in place for 1 min at each site in sham-lesioned subjects. Gelfoam covered the drilling site, and the wound was sutured closed. Animals receiving the ibotenate were carefully monitored during recovery from anesthesia and for the subsequent two weeks, to ensure full recovery of feeding and drinking after surgery. An additional four animals were given electrolytic lesion of MD, for comparison purposes. The coordinates and procedure used for these lesions have been detailed previously (16,17).
Handling and Adaptation During the second postoperative week, the subjects were placed on the food deprivation schedule. On each day of deprivation, the animals were handled, weighed, and given access to food. On Days 5 and 6 of deprivation, they were adapted to the
reinforcer in the home cage. They received pieces of "Froot Loops" cereal, placed in food cups similar to those found on the maze. The animals were introduced to the radial maze on Day 7. After being placed on the central platform, they explored the maze for 10 min and retrieved pieces of cereal strewn around and in the food cups of only those arms to contain reinforcement during training. A second session of exploration occurred the following day.
Training Subsequent to the second adaptation session, animals were trained on the radial maze, with only four of the eight arms baited. The arms to be baited for each subject were determined prior to pretraining. The configuration of baited arms was different for each animal within each group, but configurations were matched between the groups and were constant across days. Further, no more than 2 adjacent arms were baited, and the pattern of baited arms was never such that the task could be solved using an obvious response pattern (e.g., turning at 90 or 135 degree angles). One trial was administered per day, 6 to 7 days a week, for a total of 60 trials. Daily trials were terminated upon the consumption of the fourth piece of cereal. One subject with MD lesion exceeded a 10-minute period to complete one trial. The trial was terminated at that point and was excluded from analysis. Training continued until 60 trials were administered, regardless of level of performance at that time.
Histology Upon completion of testing, the subjects were sacrificed with an overdose of anesthetic and perfused intracardially with normal saline followed by 10% phosphate-buffered formalin. The brains were extracted, soaked in 30% sucrose formalin, then frozen 40-ix sections were made through the lesioned area and stained with cresyl violet. Microscopic examination of the brains were made, in order to verify the extent of toxin-induced cell damage. Reconstructions were made of the lesion extent and any damage outside the MD area was noted.
Data Analysis Note was taken of the total number of errors made, as well as of the type of error committed. Reentries into previously baited arms were counted as working memory-correct (WM-C) errors, while entries into arms that were never baited were considered reference memory errors. Further, a separate record was taken of reentries into never-baited arms. Note that such an error represents both a working memory and reference memory error. These errors were termed working memory-incorrect (WM-I) errors, but were also counted in the total number of reference memory errors. These measures were analyzed parametrically with analysis of variance for repeated measures techniques. RESULTS
Generally, MD-lesioned animals were deficient in acquisition of both the reference and working memory portions of the task. Their performance improved on some measures, but never approximated that of sham control animals, despite the administration of 60 postoperative trials. The performance of electrolytically lesioned animals was comparable to that of chemically lesioned animals, so data for all lesioned subjects were pooled in the analyses.
Reference Memory Errors For all animals, reference errors (entries into never-baited
MD LESIONS AND MEMORY TYPE
REFERENCE ERRORS
473
WORKING MEMORY CORRECT
WORKING MEMORY INCORRECT
2
2
5 3-
o4 ~3 1
. . . . . . . . . . . . .
2
4
6
8
0
10 12
BLOCKS OF 5 TRIALS
4
6
8
10 12
4
6
8
10 12
BLOCKS OF 5 TRIALS
FIG. 1. Numberof errors made by sham-operatedrats and those with MD lesions, presented according to error type, during acquisition trials. Regardless of error type (reference, working-corrector working-incorrect), control subjects improvedrapidly, making few to no errors at the end of testing. MD-lesioned subjects showed some improvement in reference memory errors, but none in working memory errors, and always made many more errors than did control subjects.
arms) composed the greatest proportion of the total number of errors made (see Fig. 1). Sham-lesioned animals exhibited a significant reduction in reference errors over trials, F(5,40)-18.0, p
