0306-4522/90$3.00+ 0.00 Pergamon Press plc 0 1990IBRO
Neuroscience Vol. 34, No. 2, pp. 293-298, 1990 Printed in Great Britain
HIPPOCAMPAL MOSSY FIBERS AND RADIAL-MAZE LEARNING IN THE MOUSE: A CORRELATION WITH SPATIAL WORKING MEMORY BUT NOT WITH NON-SPATIAL REFERENCE MEMORY H. SCHWEGLER,*~ W. E. CRUSIO$ and I. BRUST$ *Zentrum der Morphologic, Universitlt Frankfurt, Theodor-Stern-Kai 7, 6000 Frankfurt/M 70, F.R.G. SInstitut fiir Humangenetik und Anthropologie, Universitlt Heidelberg, Im Neuenheimer Feld 328, 6900 Heidelberg, F.R.G. Abstract--One hundred and eight male mice from nine different inbred strains were tested for two aspects of learning in an eight-arm radial maze. In the first experimental arrangement of the maze, measuring spatial working memory, clear strain differences were found on the fifth day of training. Furthermore, this type of learning showed a high positive correlation with the size of the intra- and infrapyramidal hippocampal mossy fiber terminal field as revealed with Timm’s staining. In the second experiment, in which non-spatial reference memory was tested, significant strain differences were found for the learning variables, but there were no significant covariations with the sizes of the intra- and infrapyramidal mossy fiber terminal fields. These results, combined with previous data, suggest that heritable variations of the hippocampal intraand infrapyramidal mossy fiber brojection influence processes determining spatial learning capabilities in mice.
The study of covariations between naturallyoccurring structural variations of the hippocampus and behavior is a powerful approach to the investigation of hippocampal functioning. Employing this strategy, we previously demonstrated the existence of strong correlations between the size of the hippocampal intra- and infrapyramidal mossy fiber (iip-MF) terminal field and a number of mouse behaviors.
Positive correlations have been found with Y-maze discrimination learning,15 simple radial-maze learning,6 habituation and exploration in an openfield,5.7 and spatial learning in a water maze.24 Negative correlations have been obtained for two-way active-avoidance learning in the shuttle-box25 and activity-related learning in a water maze.24 Based on these findings we hypothesized that the iip-MF are involved in the modulation of processes related to spatial learning,” working memory,2’ timing,22 and/or memory consolidation.28 Radial mazes have been used successfully to study a number of these different aspects of learning and memory (see e.g. Refs 3, 18 and 21). Depending on the experimental set-up of the maze, spatial learning, tTo whom correspondence should be addressed. Abbreviations: ER, number of errors; GRD, grooming duration; GRF, bout-frequency of grooming; iip-MF, intra- and infrapyramidal mossy fibers; NE, number of novel arms entered in the first eight choices; RM, reference memory; RS, running speed; RT, running time from entering the central platform until all rewards were eaten; sp-MF, suprapyramidal mossy fibers; WM, working memory.
non-spatial learning, working memory2’ (WM, storing trial-specific information), reference memory2’ (RM, concerning information relevant over all trials), or chaining responses can be tested. In a series of experiments we attempt to evaluate the role of the hippocampal mossy fibers in the regulation and/or modulation of different possible components of radial-maze learning using genetically-defined mice. The size of the iip-MF terminal field varies between inbred mouse strains and such variation is genetically dependent.4,” In a previous experiment using an unconfined radial maze with all arms reinforced we found a very strong positive correlation between learning performance and the size of the iip-MF terminal field.6 Since in that experimental set-up animals predominantly used kinesthetic strategies, i.e. entering an arm adjacent to the preceding one, it is not yet clear whether the subjects used a spatial WM or a non-spatial RM strategy to solve that particular radial-maze problem. Thus, in the first experiment presented here, we modified the maze for testing spatial WM using a confinement procedure known to disrupt chaining responses and kinesthetic strategies.’ Olton et ~1.~’ have shown that hippocampal lesions impair the performance of rats in this task. In the second experiment, a non-spatial RM task was used in order to test the involvement of the iip-MF in this type of learning. This test was modified after a task developed by Olton,” who also reported that the performance of fornix-lesioned rats was not impaired in this task. 293
H. SCHWEGLERet al.
294 EXPERIMENTAL PROCEDURES
Animals Six male mice aged 12-16 weeks were taken from each one of the highly inbred strains BA, C57BL/6J, C57BR/cdJ, BALB/cJ, DBA/2J, C3H/HeJ, CPB-K, NZB/BINJ 27 and from the partially inbred strain NMRI for each one of the two experiments. Animals were bred in the mouse colony o f the Institute o f Human Genetics and Anthropology (Heidelberg) and were weaned at the age o f 21 days. Six males of the same strain were housed together in plastic breeding cages with metal covers and a bedding o f wood shavings. Cages were cleaned once a week~ Food pellets and tap water were always available. Subjects were maintained in a mouse room (22__+T'C) with a 12L:I2D schedule; lights were turned on at 0600 h. Behavioral testing The radial maze used in the spatial WM task was similar to that described previously. 6 The central part measured 22 cm in diameter. The arms (25 cm long, 6 cm high, 6 cm wide) were closed and were made of transparent plexiglas. At the end of each arm some food pellets were deposited behind a perforated wall in order to prevent the mice from smelling the presence or absence o f the food reward. All arms were baited by placing a small food pellet (approximately 10 mg) behind a low barrier preventing the animal from seeing whether a specific arm was still baited or not. The maze was always oriented in space in the same way. Several extra-maze cues were provided close to the arms. A confinement procedure was used utilizing transparent guillotine doors at the entrance of each arm, These doors were lowered and kept closed for 5s after the animal had returned to the center box. This procedure is known to disrupt chaining responses and kinesthetic strategies in rats? Twenty-four hours prior to training the mice were deprived of food, but not of water, and received a 10-min habituation trial with fresh food at the end o f each arm and free access to all arms. During training body weight was kept at 80-90% of the pre-test body weight. Animals were trained on five consecutive days, one trial per day. Only data from the last day of training were used for the subsequent analyses. Trials were terminated after 15 min or after the animal had eaten all eight rewards, whichever came first.
An arm entrance was scored if an animal entered an arm with all four paws. An error was noted if an animal entered an arm previously visited or if it did not eat the reward. Thus, to perform well in this task, an animal continuously had to store (in its WM) information about which arms had already been visited during a particular trial and which arms had not. The following behavioral measures were recorded: number o f errors (ER), number o f novel arms entered in the first eight choices (NE), running time from entering the central platform until all rewards were eaten (RT), running speed in cm/s (RS), grooming frequency (GRF), grooming duration (GRD) and defecation (number o f fecal boll deposited). Additionally, the sequence o f the arms entered was registered. In contrast to the spatial WM task, for the non-spatial RM task the arms consisted of opaque polyvinyl chloride, covered with transparent plexiglas. In this way, animals were prevented from using spatial, extra-maze cues to solve this maze problem. Furthermore, perforated aluminium plates fixed on the floor with adhesive tape were used here instead of guillotine doors, Subjects could easily open the doors, but, because this took a small amount o f time, this procedure was expected to disrupt chaining responses and kinesthetic strategies. A door was left open if the mouse had eaten the reward from that arm, otherwise, it was closed again. Prior to training, two habituation trials of 10min each were provided on two consecutive days: on the first, the animal was allowed to move freely between all arms that had been reinforced with a fresh food pellet; in the second trial the arms were closed by the doors as described above. After having been opened, the doors were not closed again during the habituation trial. Neither extra-maze nor any further explicit intra-maze cues were provided, in both experiments only behavioral scores of the last day of training were used for the subsequent analysis.
Histology and morphometry The brains o f the mice from experiment I were processed for Timm's staining procedure) This staining visualizes the terminal fields o f the hippocampat projections in the form of colored bands and patches (Fig. I). Because o f its high content of zinc, the hippocampal mossy fiber projection with its various terminal fields (hitus, suprapyramidal, intra- and infrapyramidal) is stained particularly clearly and its size CA1
Fig. 1. Diagram of a Timm-stained cross-section of the hippocampus. The hippocampal subregion CA3-CA4 (the area of morphometry) is indicated by black, stippled and hatched areas. Black areas: suprapyramidal (SP), intra- and infrapyramidal (IIP) and hilar (CA4) mossy fiber terminal fields, originating from the dentate gyrus. Stippled areas: strata oriens (OR) and radiatum (RD), the terminal fields o f intrinsic hippocampal projections. Hatched areas: stratum lacunosum-moleculare (LM), receiving afferents from entorhinal cortex. CA1, subregion of the hippocampus without mossy fibers; FI, fimbria hippocampi; FD, fascia dentata; OL and ML, outer and middle molecular layers o f the fascia dentata (receiving entorhinal projections); SG, supragranular layer; GC, granular cells.
Hippocampal
mossy
fibers and radial-maze
Table 1. Means & S.E.M. of the behavioral variables measured in the spatial working-memory task (n = 6 per strain) Strain BA BALB/c CPB-K C3H/He C57BL/6 C57BRicd DBA/2’ NMRI NZB/BlN
ER
NE
5.2kO.5 3.3 & 0.6 4.8 k 0.5 1.3 + 0.5 1.7 f 0.6 3.0 + 0.7 5.7 : 0.8 4.2 k 1.0 9.3 + 2.0
5.8kO.4 6.5 + 0.2 6.3 f 0.2 7.3 f 0.2 7.0 + 0.3 6.7 + 0.2 5.8 7 0.2 6.5 k 0.2 5.5 * 0.2
RT 261 +26 134k 16 264 f 45 133 f 18 178 & 35 183 + 59 108 + 7 113+9 322 k 102
RS 3.2 f 0.3 5.5kO.7 3.6 k 0.8 4.5 f 0.5 3.7 f 0.5 4.9 + 0.9 7.7 IO.3 6.6 k 0.6 3.8 k 0.4
can be measured with high rehabi1ity.‘4 The hippocampal variables used for the correlative study, measured in five defined horizontal 40-pm sections taken randomly from either the right or the left hippocampus at the midseptotemporal level, were the mean absolute size of the CA3/CA4 and the mean relative sizes of the suprapyramidal (sp-MF) and the iip-MF terminal fields expressed as a percentage of CA3/CA4. The iip-MF projection includes all darklystained bands and patches which appear clearly separated from the suprapyramidal mossy fiber layer. Morphometrical analysis was done manually by measuring outlines and distribution of bands and patches on drawings made with the aid of a projection microscope (magnification x 160) on a graphics tablet connected to a desk computer (for details see Ref. 25). The hippocampal measures taken from the animals used in the spatial task were also used to calculate correlations with the behavioral data from the non-spatial task.
295
learning
preference for the 90 and 135” angles (Table 2). The measured hippocampal variables also revealed, as expected, strain differences (Table 3). Large variation was found for the size of the iip-MF terminal field (x2 = 48.91, d.f. = 8; P < O.OOl), C3H having the largest, NZB the smallest projection. Strain differences of the other variables measured were also significant (CA3/CA4: x2 = 15.89, d.f. = 8; P < 0.05; sp-MF: x2 = 27.98, d.f. = 8; P < O.OOl), thus providing the opportunity for a correlative analysis. We found significant morpho-behavioral correlations only between the iip-MF and the two learning variables ER (r, = -0.92, d.f. = 7; P < 0.001) and NE (r, = 0.89, d.f. = 7; P < 0.01). None of the other hippocampal variables exhibited covariations with behavior. As might be expected, ER and NE were strongly correlated (rs = -0.99, d.f. = 7; P < 0.001) as were RS and RT (rs = -0.78, d.f. = 7; P < 0.05). No correlations appeared between the measures of activity and those of learning performance. Non-spatial reference-memory
Table 4 shows the results of the second experiment. Again, all animals, with the exception of one BA individual, succeeded in finding the eight food pellets within 15 min. There are no substantial changes in the following results if the data for this particular BA mouse are omitted. As revealed by Kruskal-Wallis one-way analyses of variance, we found significant strain differences for the two learning variables ER (x2 = 16.31, d.f. = 8; P < 0.05) and NE (x2 = 21.27, d.f. = 8; P < O.Ol), but not for the measures of activity RT and RS. Seven of the mouse strains reached fairly high scores for both learning variables whereas strains BA and NZB did not show any increase in learning performance over trials. Calculating Spearman rank correlation coefficients between the strain means of the hippocampal measures taken from the animals trained in the spatial WM task and the means of the behavioral variables measured in the non-spatial RM task, we found no significant covariation between the iip-MF and behavior. However, the size of the sp-MF terminal field correlated significantly with ER (r, = -0.69, d.f. = 7; P < 0.05), but not with NE. The correlation between RS and the size of CA3/CA4 was almost significant (r, = -0.65, d.f. = 7; P < 0.06). Again, ER and NE correlated with each other (r, = -0.77, d.f. = 7; P < 0.05) as did RT and RS (r, = -0.67, d.f. = 7; P < 0.05). No correlation appeared between the measures of
Statisticalanalysis Group differences were evaluated by means of the nonparametric Kruskal-Wallis one-way analysis of variance.26 Correlations between variables were estimated by calculating Spearman’s rank correlation coefficient, corrected for tiesz6 between strain means. RESULTS
Spatial working-memory task
Strain means of the behavioral scores of day 5 are shown in Table 1. It can be noted that all animals managed to find the eight rewards within the 15-min time limit. Non-parametric analyses of variance (Kruskal-Wallis test) revealed significant strain differences for the two learning variables ER (x2 = 30.32, d.f. = 8; P < 0.001) and NE (1’ = 28.30, d.f. = 8; P < 0.001) and for the measures of activity RT (x2 = 27.90, d.f. = 8; P < 0.001) and RS (x2 = 19.70, d.f. = 8; P < 0.05). Strain C3H learned best, whereas NZB had the poorest performance. Despite its medium performance, strain DBA/2 had the lowest mean RT because of its very high RS. Determination of the interchoice angles between successively chosen arms in the last trial showed a
Table 2. Occurrence of various angles between successively chosen arms of the radial maze on the fifth day of training (mean percentage f S.E.M.)* Task Spatial WM Non-spatial RM
task
0”
45”
90”
135”
180”
l&O 1*1
14&2 21&-3
34 + 3 31*2
35 f 3 32 k 3
15+2 16&-2
*Data presented are the results of all mice (n = 54 for each experiment) pooled over strains, because similar distributions of interchoice-angle frequencies were found for all strains.
H.
296 Table
Regio inferior (x lO’pm*)
expressed
as percentage
mossv fibers* 8.25 9.04 8.45 9.54 8.84 9.02 9.82 7.38 8.14
DISCUSSION
Our experiments reveal the existence of strain differences for spatial WM as well as for non-spatial RM abilities. While the learning scores of the mouse strains in the spatial WM task are more or less uniformly distributed, variation in the non-spatial RM task is mainly determined by two strains: only BA and NZB mice perform rather poorly in the latter task. Mizumori et al.” reported that mice failed to learn the spatial task, although, in addition to the usual extra-maze stimuli, they provided each arm with its own unique extra-maze cue. That result is at variance with both the present and previous findingsb,13 Differences between strains most probably explain why the CD mice used by Mizumori et aLi did not learn a simple radial-maze task, whereas other mouse strains have been shown to perform well in the same or in an even more difficult task 2.10.13.16 Also in contrast to our present results, Reinstein aL2” reported that C57BL/6 mice performed the spatial task very poorly. A probable explanation for this discrepancy is the different testing procedure employed by these authors, in which trials lasted 5 min only. According to our data, many mice will not succeed in finding all eight rewards within et
Table 4. Means & S.E.M. of the behavioral variables measured in the non-spatial reference-memory task (n = 6 her strain)
BA BALB/c CPB-K C3H/He C57BL/6 C57BRjcd DBAj2 NMRI NZBjBiN
8.2 + 3.5 * 2.7f 2.8 + 2.5 + 2.5 + 2.2 * 4.2 & 9.7 +
NE 1.7 1.3 1.1 1.0 0.8 0.9 0.7 2.4 2.7
4.7kO.6 6.3 + 0.6 7.0 k 0.3 6.5 & 0.3 6.5 &OS 7.0 * 0.3 7.0 k 0.3 6.8 + 0.3 5.0 & 0.4
RS
RT 414, 173 + 296 + 360 + 240 It 227 i 204 & 230 i 421 +
113 26 82 59 47 51 27 73 93
& 0.17 * 0.31 * 0.18 i 0.38 & 0.22 & 0.29 f 0.46 f 0.20 2 0.32
Intra- and infrapyramidal mossy
fibers*
1.71 _t 0.14 I.83 t 0.08 0.89 + 0.07 4.03 * 0.25 3.14 + 0.08 2.32 & 0.13 i.lS;tO.O9 I .61 i: 0.08 0.86 F 0.08
of the regio inferior.
activity and those of learning performance. Comparing the results from both tasks we found no significant correlation between the learning performances in the spatial WM and the non-spatial RM tasks @R/S-ER/NS: Y,= 0.34, d.f. = 7, ns.; NE/S-NE/NS: r, = 0.23, d.f. = 7, n.s.).
ER
variables (n = 6 per strain)
Suprapyramidal
684 & 39 692 h 24 705 i 26 742 i: 36 668 1.19 685+31 694 + 22 642,36 849 rt 44
BA BALE/c CPB-K C3H/He C57BLi6 C57BRjcd DBA/Z NMRt NZB,‘Bl N
Strain
et al.
3. Means & S.E.M. of hippocampal
Strain
*Areas
SCHWEGLEK
3.1 4.3 3.0 2.0 3.3 3.2 3.2 4.6 3.1
+ 0.6 + 0.5 + 0.7 & 0.3 + 0.8 + 0.5 + 0.3 & 1.2 ;f: 0.6
this time period, which probably interferes with acquisition. Apparently, this interference is not compensated for by the longer (three-week) training period applied by Reinstein et ~1.‘~It might be noted, however, that the BALB/c and C57BR/cd strains did perform well in their procedure. An alternative explanation for the divergent findings might be the existence of substrain differences. Heimrich et u(.‘~ have demonstrated significant differences between C3H substrains with regard to both behavior and hippocampal anatomy and such differences also occur between different C57BL/6 substrains (unpublished observations). Ammassari-Teule and Caprioli’ compared the mouse strains DBA and C57BL/6 in an elevated radial maze without extra-maze cues, a non-spatial task. The latter strain showed a superior learning performance. However, the authors ascribed this better performance to a higher activity and a larger number of solving strategies of C57BL/6 mice. In our experiment, activity appears to be unimportant for learning performance in either task. No significant correlation has been found here between the variables related to learning and the variables related to activity. In our previous study,’ we supposed that most animals mastered the simple radial-maze task, without confinement procedure, by using kinesthetic strategies. This conclusion was sustained by the fact that in the unconfined radial maze most animals showed a tendency to subsequently visit adjacent arms. Bolhuis et al.’ demonstrated that rats do not use this strategy in a confined maze but instead show a slight preference for the 90 and 135” interchoice angles. Similarly, the mice in our present experiments did not show any preference for visiting an adjacent arm, but displayed a more or less pronounced preference for the 90 and 135” angles. From these findings we may conclude that, as expected, the confinement procedure used in the spatial WM task induced our mice to use extra-maze cues for solving the problem. Also, the doors used in the non-spatial RM task forced subjects to orient themselves on intra-maze cues instead of displaying chaining responses, confirming our expectations. There is no evidence whatsoever for the use of kinesthetic strategies by any strain in the present tasks.
Hippocampal mossy fibers and radial-maze learning The non-significant correlations between the learning performances in the spatial WM and non-spatial RM tasks indicate that these tasks test different capabilities. The learning scores of mice in the task used by Crusio et ~1.~ (unconfined maze) correlate with those obtained in the spatial WM task (with confinement procedure, rS = 0.77, d.f. = 6; P < 0.05) but not with those obtained in the non-spatial RM task (rS = -0.05, d.f. = 6, ns.). It appears that in the unconfined radial maze either spatial abilities have also been tested or the use of kinesthetic and spatial strategies is at least partially based on the same mechanisms. The latter explanation seems plausible in the light of the fact that, in maintaining a course (a spatial activity) animals often use kinesthetic input as a source of directional information.’ As expected, a strong positive correlation has been revealed between the learning variables (ER and NE) in the spatial WM task and the size of the iip-MF terminal field, whereas other hippocampal variables did not covary with behavior in this task. This result is in agreement with our previous experiments, where we also found positive correlations between the size of the iip-MF terminal field and spatial aspects of learning and behavior.5-7,24 Furthermore, our data are also in agreement with most lesion studies, which have demonstrated in rats2’ and mice2’ a decrease of learning capacity in the spatial WM radial maze task following fimbria-fornix lesions. In contrast, no correlation between the iip-MF projection and non-spatial RM in the radial maze was revealed. This suggests that the iip-MF projection is not involved in this aspect of learning. However, we did find a significant positive correlation between the suprapyramidal mossy fiber projection and ER in the non-spatial RM task, but it should be noted that the correlation between the sp-MF and NE did not reach significance. The relationship between the sp-MF and non-spatial RM in the radial maze seems to be fortuitous or very weak at best. Olton” studied fornix-lesioned rats in a similar task and his results also indicate that the hippocampus is not involved in
297
this type of learning since no decrease of learning capacity was observed in lesioned animals. CONCLUSION
Our results might be interpreted in two different ways. The first one is that the hippocampus is involved in the modulation and/or regulation of the processing of spatial, but not of non-spatial information, as has been hypothesized by O’Keefe and Nadel.” However, the two experiments presented here not only differed in the spatial or non-spatial nature of the cues offered to the subjects, but also in the type of memory that had to be used in these tasks. For a good performance in the spatial task, it is necessary for the subject to remember exactly which arms have already been visited within a particular session: a WM task (cf. Ref. 21). In the non-spatial task, animals had to learn that a closed door indicated the presence, an open door the absence of a food reward: an RM2’ task. Thus, the different correlations found in the present experiments might also be interpreted to indicate that the hippocampus is involved in WM, but not in RM, as hypothesized by Olton et al.” However, in previous experiments we obtained positive correlations of the size of the iip-MF with spatial RM in a water maze24 and with habituation and exploration in an open-field’,’ (which also involves RM). Thus, the results of our experiments employing a number of different tasks suggest that the hippocampal iip-MF have an important function in the regulation and/or modulation of spatial RM and WM, but not in non-spatial RM. In the light of the hippocampal working memory theory, 21 it will be interesting to see whether the iip-MF also covary with non-spatial WM abilities. Further experiments addressing the latter problem are now in progress. wish to thank Prof. F. Vogel (Heidelberg) for reading the manuscript, Gary G. Mueller (Frankfurt) for correcting the English, and Prof. W. Busel-
Acknowledgements-We
maier (Heidelberg)
for providing
mice and testing facilities.
REFERENCES
I. Ammassari-Teule M. and Caprioli A. (1985) Spatial learning and memory, maze running strategies and cholinergic mechanisms in two inbred strains of mice. Behav. Brain Res. 17, 9-16. 2. Bernstein D., Olton D. S., Ingram D. K., Wailer S. B., Reynolds M. A. and London E. D. (1985) Radial maze performance in young and aged mice: neurochemical correlates. Pharmac. Biochem. Behau. 22, 301-307. 3. Bolhuis J. J., Bijlsma S. and Ansmink P. (1986) Exponential decay of spatial memory of rats in a radial-maze. Behav. neural Biol. 46, 115-I22. 4. Crusio W. E., Genthner-Grimm G. and Schwegler H. (1986) A quantitative-genetic analysis of hippocampal variation in the mouse. J. Neurogenet. 3, 203-214. 5. Crusio W. E. and Schwegler H. (1987) Hippocampal mossy fiber distribution covaries with open-field habituation in the mouse. Behau. Brain Res. 26, 153-158. 6. Crusio W. E., Schwegler H. and Lipp H.-P. (1987) Radial-maze performance and structural variation of the hippocampus in mice: a correlation with mossy fibre distribution. Brain Res. 425, 182-185. I. Crusio W. E., Schwegler H. and van Abeelen J. H. F. (1989) Behavioral responses to novelty and structural variation of the hippocampus in mice. II. Multivariate genetic analysis. Behav. Brain Res. 32, 81-88. 8. Danscher G. and Zimmer J. (1978) An improved Timm sulfide silver method for light and electron microscopic localization of heavy metals in biological tissues. Histochemistry 55, 2740. 9. Gallistel C. R. (1989) Animal cognition: the representation of space, time and number. A. Rev. Psychol. 40, 1555189. 10. Goldowitz D. and Koch J. (1986) Performance of normal and neurological mutant mice on radial arm maze and active avoidance tasks. Behao. neural Biol. 46, 216-226.
298
H. SCHWEGLER et ul.
11. Heimrich B., Schwegler H. and Crusio W. E. (1985) Hippocampal variation between the inbred mouse strains C3H/HeJ and DBA/2: a quantitative-genetic analysis. J. Neurogenef. 2, 389401. 12. Heimrich B., Schwegler H., Crusio W. E. and Buselmaier W. (1988) Substrain divergence in C3H inbred mice. Be&,. Gener. 18, 671674. 13. Idrobo F., Nandy K., Mostofsky D. I., Blatt L. and Nandy L. (1987) Dietary restriction: effects on radial-maze learning and lipofuscin pigment deposition in the hippocampal and frontal cortex. Arch. Gerontol. Geriutr. 6, 355-362. 14. Lipp H.-P. and Schwegler H. (1983) Hippocampal mossy fibers and avoidance learning. In Genetics of the Brain (ed. Lieblich I.), pp. 325-364. Elsevier Biomedical, Amsterdam. 15. Lipp H.-P., Schwegler H. and Leisinger-Trigona M.-C. (1986) lnfrapyramidal mossy fibers and Y-maze discrimination learning of mice: a positive correlation. Experienfia 42, 676. 16. Mishima N., Higashitani F., Teraoka K. and Yoshioka R. (1986) Sex differences in appetitive learning of mice. Physiol. Behav. 37, 263-268. 17. Mizumori S. J. Y., Rosenzweig M. R. and Kermisch M. G. (1982) Failure of mice to demonstrate spatial memory in the radial-maze. Behav. neural Biol. 35, 33-45. 18. Nadel L. and MacDonald L. (1980) Hippocampus: cognitive map or working memory? Behav. neural Biol. 29,405409. 19. O’Keefe J. and Nadel L. (1978) The Hippocampus as a Cognitive Map. Clarendon Press, Oxford. 20. Olton D. S. (1977) Spatial memory. Sci. Am. 236, 82-98. 21. Olton D. S., Handelmann G. E. and Walker J. A. (1979) Hippocampus, space, and memory. Behav. Brain Sci. 2, 313-365. across time: the hippocampus as a temporary memory store. Behav. Bruin Sci. 22. Rawlins J. N. P. (1985) Associations 8, 479496. 23. Reinstein D. K., DeBoissiere T., Robinson N. and Wurtman R. J. (1983) Radial maze performance in three strains of mice: role of the fimbriaifornix. Brain Res. 263, 172-176. learning in the mouse correlates with 24. Schwegler H., Crusio W. E., Lipp H.-P. and Heimrich B. (1988) Water-maze variation in hippocampal morphology. Behan. Gene/. 18, 153-165. 25. Schwegler H. and Lipp H.-P. (1983) Hereditary covariation of neuronal circuitry and behavior: correlation between the proportions of hippocampal synaptic fields in the regio inferior and two-way avoidance in mice and rats. Behav. Brain Res. 7, l-38. New York. 26. Siegel S. (1956) Nonparametric Statisrics ,for the Behavioral Sciences. McGraw~Hill, nomenclature for inbred strains of mice: eight listing. Cancer Res. 45, 9455977. 27. Staats J. (1985) Standardized memory indexing theory. Behac. Neurosci. 100, 147-154. 28. Teyler T. J. and DiScenna P. (1986) The hippocampal (Accepted 6 September 1989)
Neuroscience Vol. 34, No. 2, pp. 293-298, 1990 Printed in Great Britain
HIPPOCAMPAL MOSSY FIBERS AND RADIAL-MAZE LEARNING IN THE MOUSE: A CORRELATION WITH SPATIAL WORKING MEMORY BUT NOT WITH NON-SPATIAL REFERENCE MEMORY H. SCHWEGLER,*~ W. E. CRUSIO$ and I. BRUST$ *Zentrum der Morphologic, Universitlt Frankfurt, Theodor-Stern-Kai 7, 6000 Frankfurt/M 70, F.R.G. SInstitut fiir Humangenetik und Anthropologie, Universitlt Heidelberg, Im Neuenheimer Feld 328, 6900 Heidelberg, F.R.G. Abstract--One hundred and eight male mice from nine different inbred strains were tested for two aspects of learning in an eight-arm radial maze. In the first experimental arrangement of the maze, measuring spatial working memory, clear strain differences were found on the fifth day of training. Furthermore, this type of learning showed a high positive correlation with the size of the intra- and infrapyramidal hippocampal mossy fiber terminal field as revealed with Timm’s staining. In the second experiment, in which non-spatial reference memory was tested, significant strain differences were found for the learning variables, but there were no significant covariations with the sizes of the intra- and infrapyramidal mossy fiber terminal fields. These results, combined with previous data, suggest that heritable variations of the hippocampal intraand infrapyramidal mossy fiber brojection influence processes determining spatial learning capabilities in mice.
The study of covariations between naturallyoccurring structural variations of the hippocampus and behavior is a powerful approach to the investigation of hippocampal functioning. Employing this strategy, we previously demonstrated the existence of strong correlations between the size of the hippocampal intra- and infrapyramidal mossy fiber (iip-MF) terminal field and a number of mouse behaviors.
Positive correlations have been found with Y-maze discrimination learning,15 simple radial-maze learning,6 habituation and exploration in an openfield,5.7 and spatial learning in a water maze.24 Negative correlations have been obtained for two-way active-avoidance learning in the shuttle-box25 and activity-related learning in a water maze.24 Based on these findings we hypothesized that the iip-MF are involved in the modulation of processes related to spatial learning,” working memory,2’ timing,22 and/or memory consolidation.28 Radial mazes have been used successfully to study a number of these different aspects of learning and memory (see e.g. Refs 3, 18 and 21). Depending on the experimental set-up of the maze, spatial learning, tTo whom correspondence should be addressed. Abbreviations: ER, number of errors; GRD, grooming duration; GRF, bout-frequency of grooming; iip-MF, intra- and infrapyramidal mossy fibers; NE, number of novel arms entered in the first eight choices; RM, reference memory; RS, running speed; RT, running time from entering the central platform until all rewards were eaten; sp-MF, suprapyramidal mossy fibers; WM, working memory.
non-spatial learning, working memory2’ (WM, storing trial-specific information), reference memory2’ (RM, concerning information relevant over all trials), or chaining responses can be tested. In a series of experiments we attempt to evaluate the role of the hippocampal mossy fibers in the regulation and/or modulation of different possible components of radial-maze learning using genetically-defined mice. The size of the iip-MF terminal field varies between inbred mouse strains and such variation is genetically dependent.4,” In a previous experiment using an unconfined radial maze with all arms reinforced we found a very strong positive correlation between learning performance and the size of the iip-MF terminal field.6 Since in that experimental set-up animals predominantly used kinesthetic strategies, i.e. entering an arm adjacent to the preceding one, it is not yet clear whether the subjects used a spatial WM or a non-spatial RM strategy to solve that particular radial-maze problem. Thus, in the first experiment presented here, we modified the maze for testing spatial WM using a confinement procedure known to disrupt chaining responses and kinesthetic strategies.’ Olton et ~1.~’ have shown that hippocampal lesions impair the performance of rats in this task. In the second experiment, a non-spatial RM task was used in order to test the involvement of the iip-MF in this type of learning. This test was modified after a task developed by Olton,” who also reported that the performance of fornix-lesioned rats was not impaired in this task. 293
H. SCHWEGLERet al.
294 EXPERIMENTAL PROCEDURES
Animals Six male mice aged 12-16 weeks were taken from each one of the highly inbred strains BA, C57BL/6J, C57BR/cdJ, BALB/cJ, DBA/2J, C3H/HeJ, CPB-K, NZB/BINJ 27 and from the partially inbred strain NMRI for each one of the two experiments. Animals were bred in the mouse colony o f the Institute o f Human Genetics and Anthropology (Heidelberg) and were weaned at the age o f 21 days. Six males of the same strain were housed together in plastic breeding cages with metal covers and a bedding o f wood shavings. Cages were cleaned once a week~ Food pellets and tap water were always available. Subjects were maintained in a mouse room (22__+T'C) with a 12L:I2D schedule; lights were turned on at 0600 h. Behavioral testing The radial maze used in the spatial WM task was similar to that described previously. 6 The central part measured 22 cm in diameter. The arms (25 cm long, 6 cm high, 6 cm wide) were closed and were made of transparent plexiglas. At the end of each arm some food pellets were deposited behind a perforated wall in order to prevent the mice from smelling the presence or absence o f the food reward. All arms were baited by placing a small food pellet (approximately 10 mg) behind a low barrier preventing the animal from seeing whether a specific arm was still baited or not. The maze was always oriented in space in the same way. Several extra-maze cues were provided close to the arms. A confinement procedure was used utilizing transparent guillotine doors at the entrance of each arm, These doors were lowered and kept closed for 5s after the animal had returned to the center box. This procedure is known to disrupt chaining responses and kinesthetic strategies in rats? Twenty-four hours prior to training the mice were deprived of food, but not of water, and received a 10-min habituation trial with fresh food at the end o f each arm and free access to all arms. During training body weight was kept at 80-90% of the pre-test body weight. Animals were trained on five consecutive days, one trial per day. Only data from the last day of training were used for the subsequent analyses. Trials were terminated after 15 min or after the animal had eaten all eight rewards, whichever came first.
An arm entrance was scored if an animal entered an arm with all four paws. An error was noted if an animal entered an arm previously visited or if it did not eat the reward. Thus, to perform well in this task, an animal continuously had to store (in its WM) information about which arms had already been visited during a particular trial and which arms had not. The following behavioral measures were recorded: number o f errors (ER), number o f novel arms entered in the first eight choices (NE), running time from entering the central platform until all rewards were eaten (RT), running speed in cm/s (RS), grooming frequency (GRF), grooming duration (GRD) and defecation (number o f fecal boll deposited). Additionally, the sequence o f the arms entered was registered. In contrast to the spatial WM task, for the non-spatial RM task the arms consisted of opaque polyvinyl chloride, covered with transparent plexiglas. In this way, animals were prevented from using spatial, extra-maze cues to solve this maze problem. Furthermore, perforated aluminium plates fixed on the floor with adhesive tape were used here instead of guillotine doors, Subjects could easily open the doors, but, because this took a small amount o f time, this procedure was expected to disrupt chaining responses and kinesthetic strategies. A door was left open if the mouse had eaten the reward from that arm, otherwise, it was closed again. Prior to training, two habituation trials of 10min each were provided on two consecutive days: on the first, the animal was allowed to move freely between all arms that had been reinforced with a fresh food pellet; in the second trial the arms were closed by the doors as described above. After having been opened, the doors were not closed again during the habituation trial. Neither extra-maze nor any further explicit intra-maze cues were provided, in both experiments only behavioral scores of the last day of training were used for the subsequent analysis.
Histology and morphometry The brains o f the mice from experiment I were processed for Timm's staining procedure) This staining visualizes the terminal fields o f the hippocampat projections in the form of colored bands and patches (Fig. I). Because o f its high content of zinc, the hippocampal mossy fiber projection with its various terminal fields (hitus, suprapyramidal, intra- and infrapyramidal) is stained particularly clearly and its size CA1
Fig. 1. Diagram of a Timm-stained cross-section of the hippocampus. The hippocampal subregion CA3-CA4 (the area of morphometry) is indicated by black, stippled and hatched areas. Black areas: suprapyramidal (SP), intra- and infrapyramidal (IIP) and hilar (CA4) mossy fiber terminal fields, originating from the dentate gyrus. Stippled areas: strata oriens (OR) and radiatum (RD), the terminal fields o f intrinsic hippocampal projections. Hatched areas: stratum lacunosum-moleculare (LM), receiving afferents from entorhinal cortex. CA1, subregion of the hippocampus without mossy fibers; FI, fimbria hippocampi; FD, fascia dentata; OL and ML, outer and middle molecular layers o f the fascia dentata (receiving entorhinal projections); SG, supragranular layer; GC, granular cells.
Hippocampal
mossy
fibers and radial-maze
Table 1. Means & S.E.M. of the behavioral variables measured in the spatial working-memory task (n = 6 per strain) Strain BA BALB/c CPB-K C3H/He C57BL/6 C57BRicd DBA/2’ NMRI NZB/BlN
ER
NE
5.2kO.5 3.3 & 0.6 4.8 k 0.5 1.3 + 0.5 1.7 f 0.6 3.0 + 0.7 5.7 : 0.8 4.2 k 1.0 9.3 + 2.0
5.8kO.4 6.5 + 0.2 6.3 f 0.2 7.3 f 0.2 7.0 + 0.3 6.7 + 0.2 5.8 7 0.2 6.5 k 0.2 5.5 * 0.2
RT 261 +26 134k 16 264 f 45 133 f 18 178 & 35 183 + 59 108 + 7 113+9 322 k 102
RS 3.2 f 0.3 5.5kO.7 3.6 k 0.8 4.5 f 0.5 3.7 f 0.5 4.9 + 0.9 7.7 IO.3 6.6 k 0.6 3.8 k 0.4
can be measured with high rehabi1ity.‘4 The hippocampal variables used for the correlative study, measured in five defined horizontal 40-pm sections taken randomly from either the right or the left hippocampus at the midseptotemporal level, were the mean absolute size of the CA3/CA4 and the mean relative sizes of the suprapyramidal (sp-MF) and the iip-MF terminal fields expressed as a percentage of CA3/CA4. The iip-MF projection includes all darklystained bands and patches which appear clearly separated from the suprapyramidal mossy fiber layer. Morphometrical analysis was done manually by measuring outlines and distribution of bands and patches on drawings made with the aid of a projection microscope (magnification x 160) on a graphics tablet connected to a desk computer (for details see Ref. 25). The hippocampal measures taken from the animals used in the spatial task were also used to calculate correlations with the behavioral data from the non-spatial task.
295
learning
preference for the 90 and 135” angles (Table 2). The measured hippocampal variables also revealed, as expected, strain differences (Table 3). Large variation was found for the size of the iip-MF terminal field (x2 = 48.91, d.f. = 8; P < O.OOl), C3H having the largest, NZB the smallest projection. Strain differences of the other variables measured were also significant (CA3/CA4: x2 = 15.89, d.f. = 8; P < 0.05; sp-MF: x2 = 27.98, d.f. = 8; P < O.OOl), thus providing the opportunity for a correlative analysis. We found significant morpho-behavioral correlations only between the iip-MF and the two learning variables ER (r, = -0.92, d.f. = 7; P < 0.001) and NE (r, = 0.89, d.f. = 7; P < 0.01). None of the other hippocampal variables exhibited covariations with behavior. As might be expected, ER and NE were strongly correlated (rs = -0.99, d.f. = 7; P < 0.001) as were RS and RT (rs = -0.78, d.f. = 7; P < 0.05). No correlations appeared between the measures of activity and those of learning performance. Non-spatial reference-memory
Table 4 shows the results of the second experiment. Again, all animals, with the exception of one BA individual, succeeded in finding the eight food pellets within 15 min. There are no substantial changes in the following results if the data for this particular BA mouse are omitted. As revealed by Kruskal-Wallis one-way analyses of variance, we found significant strain differences for the two learning variables ER (x2 = 16.31, d.f. = 8; P < 0.05) and NE (x2 = 21.27, d.f. = 8; P < O.Ol), but not for the measures of activity RT and RS. Seven of the mouse strains reached fairly high scores for both learning variables whereas strains BA and NZB did not show any increase in learning performance over trials. Calculating Spearman rank correlation coefficients between the strain means of the hippocampal measures taken from the animals trained in the spatial WM task and the means of the behavioral variables measured in the non-spatial RM task, we found no significant covariation between the iip-MF and behavior. However, the size of the sp-MF terminal field correlated significantly with ER (r, = -0.69, d.f. = 7; P < 0.05), but not with NE. The correlation between RS and the size of CA3/CA4 was almost significant (r, = -0.65, d.f. = 7; P < 0.06). Again, ER and NE correlated with each other (r, = -0.77, d.f. = 7; P < 0.05) as did RT and RS (r, = -0.67, d.f. = 7; P < 0.05). No correlation appeared between the measures of
Statisticalanalysis Group differences were evaluated by means of the nonparametric Kruskal-Wallis one-way analysis of variance.26 Correlations between variables were estimated by calculating Spearman’s rank correlation coefficient, corrected for tiesz6 between strain means. RESULTS
Spatial working-memory task
Strain means of the behavioral scores of day 5 are shown in Table 1. It can be noted that all animals managed to find the eight rewards within the 15-min time limit. Non-parametric analyses of variance (Kruskal-Wallis test) revealed significant strain differences for the two learning variables ER (x2 = 30.32, d.f. = 8; P < 0.001) and NE (1’ = 28.30, d.f. = 8; P < 0.001) and for the measures of activity RT (x2 = 27.90, d.f. = 8; P < 0.001) and RS (x2 = 19.70, d.f. = 8; P < 0.05). Strain C3H learned best, whereas NZB had the poorest performance. Despite its medium performance, strain DBA/2 had the lowest mean RT because of its very high RS. Determination of the interchoice angles between successively chosen arms in the last trial showed a
Table 2. Occurrence of various angles between successively chosen arms of the radial maze on the fifth day of training (mean percentage f S.E.M.)* Task Spatial WM Non-spatial RM
task
0”
45”
90”
135”
180”
l&O 1*1
14&2 21&-3
34 + 3 31*2
35 f 3 32 k 3
15+2 16&-2
*Data presented are the results of all mice (n = 54 for each experiment) pooled over strains, because similar distributions of interchoice-angle frequencies were found for all strains.
H.
296 Table
Regio inferior (x lO’pm*)
expressed
as percentage
mossv fibers* 8.25 9.04 8.45 9.54 8.84 9.02 9.82 7.38 8.14
DISCUSSION
Our experiments reveal the existence of strain differences for spatial WM as well as for non-spatial RM abilities. While the learning scores of the mouse strains in the spatial WM task are more or less uniformly distributed, variation in the non-spatial RM task is mainly determined by two strains: only BA and NZB mice perform rather poorly in the latter task. Mizumori et al.” reported that mice failed to learn the spatial task, although, in addition to the usual extra-maze stimuli, they provided each arm with its own unique extra-maze cue. That result is at variance with both the present and previous findingsb,13 Differences between strains most probably explain why the CD mice used by Mizumori et aLi did not learn a simple radial-maze task, whereas other mouse strains have been shown to perform well in the same or in an even more difficult task 2.10.13.16 Also in contrast to our present results, Reinstein aL2” reported that C57BL/6 mice performed the spatial task very poorly. A probable explanation for this discrepancy is the different testing procedure employed by these authors, in which trials lasted 5 min only. According to our data, many mice will not succeed in finding all eight rewards within et
Table 4. Means & S.E.M. of the behavioral variables measured in the non-spatial reference-memory task (n = 6 her strain)
BA BALB/c CPB-K C3H/He C57BL/6 C57BRjcd DBAj2 NMRI NZBjBiN
8.2 + 3.5 * 2.7f 2.8 + 2.5 + 2.5 + 2.2 * 4.2 & 9.7 +
NE 1.7 1.3 1.1 1.0 0.8 0.9 0.7 2.4 2.7
4.7kO.6 6.3 + 0.6 7.0 k 0.3 6.5 & 0.3 6.5 &OS 7.0 * 0.3 7.0 k 0.3 6.8 + 0.3 5.0 & 0.4
RS
RT 414, 173 + 296 + 360 + 240 It 227 i 204 & 230 i 421 +
113 26 82 59 47 51 27 73 93
& 0.17 * 0.31 * 0.18 i 0.38 & 0.22 & 0.29 f 0.46 f 0.20 2 0.32
Intra- and infrapyramidal mossy
fibers*
1.71 _t 0.14 I.83 t 0.08 0.89 + 0.07 4.03 * 0.25 3.14 + 0.08 2.32 & 0.13 i.lS;tO.O9 I .61 i: 0.08 0.86 F 0.08
of the regio inferior.
activity and those of learning performance. Comparing the results from both tasks we found no significant correlation between the learning performances in the spatial WM and the non-spatial RM tasks @R/S-ER/NS: Y,= 0.34, d.f. = 7, ns.; NE/S-NE/NS: r, = 0.23, d.f. = 7, n.s.).
ER
variables (n = 6 per strain)
Suprapyramidal
684 & 39 692 h 24 705 i 26 742 i: 36 668 1.19 685+31 694 + 22 642,36 849 rt 44
BA BALE/c CPB-K C3H/He C57BLi6 C57BRjcd DBA/Z NMRt NZB,‘Bl N
Strain
et al.
3. Means & S.E.M. of hippocampal
Strain
*Areas
SCHWEGLEK
3.1 4.3 3.0 2.0 3.3 3.2 3.2 4.6 3.1
+ 0.6 + 0.5 + 0.7 & 0.3 + 0.8 + 0.5 + 0.3 & 1.2 ;f: 0.6
this time period, which probably interferes with acquisition. Apparently, this interference is not compensated for by the longer (three-week) training period applied by Reinstein et ~1.‘~It might be noted, however, that the BALB/c and C57BR/cd strains did perform well in their procedure. An alternative explanation for the divergent findings might be the existence of substrain differences. Heimrich et u(.‘~ have demonstrated significant differences between C3H substrains with regard to both behavior and hippocampal anatomy and such differences also occur between different C57BL/6 substrains (unpublished observations). Ammassari-Teule and Caprioli’ compared the mouse strains DBA and C57BL/6 in an elevated radial maze without extra-maze cues, a non-spatial task. The latter strain showed a superior learning performance. However, the authors ascribed this better performance to a higher activity and a larger number of solving strategies of C57BL/6 mice. In our experiment, activity appears to be unimportant for learning performance in either task. No significant correlation has been found here between the variables related to learning and the variables related to activity. In our previous study,’ we supposed that most animals mastered the simple radial-maze task, without confinement procedure, by using kinesthetic strategies. This conclusion was sustained by the fact that in the unconfined radial maze most animals showed a tendency to subsequently visit adjacent arms. Bolhuis et al.’ demonstrated that rats do not use this strategy in a confined maze but instead show a slight preference for the 90 and 135” interchoice angles. Similarly, the mice in our present experiments did not show any preference for visiting an adjacent arm, but displayed a more or less pronounced preference for the 90 and 135” angles. From these findings we may conclude that, as expected, the confinement procedure used in the spatial WM task induced our mice to use extra-maze cues for solving the problem. Also, the doors used in the non-spatial RM task forced subjects to orient themselves on intra-maze cues instead of displaying chaining responses, confirming our expectations. There is no evidence whatsoever for the use of kinesthetic strategies by any strain in the present tasks.
Hippocampal mossy fibers and radial-maze learning The non-significant correlations between the learning performances in the spatial WM and non-spatial RM tasks indicate that these tasks test different capabilities. The learning scores of mice in the task used by Crusio et ~1.~ (unconfined maze) correlate with those obtained in the spatial WM task (with confinement procedure, rS = 0.77, d.f. = 6; P < 0.05) but not with those obtained in the non-spatial RM task (rS = -0.05, d.f. = 6, ns.). It appears that in the unconfined radial maze either spatial abilities have also been tested or the use of kinesthetic and spatial strategies is at least partially based on the same mechanisms. The latter explanation seems plausible in the light of the fact that, in maintaining a course (a spatial activity) animals often use kinesthetic input as a source of directional information.’ As expected, a strong positive correlation has been revealed between the learning variables (ER and NE) in the spatial WM task and the size of the iip-MF terminal field, whereas other hippocampal variables did not covary with behavior in this task. This result is in agreement with our previous experiments, where we also found positive correlations between the size of the iip-MF terminal field and spatial aspects of learning and behavior.5-7,24 Furthermore, our data are also in agreement with most lesion studies, which have demonstrated in rats2’ and mice2’ a decrease of learning capacity in the spatial WM radial maze task following fimbria-fornix lesions. In contrast, no correlation between the iip-MF projection and non-spatial RM in the radial maze was revealed. This suggests that the iip-MF projection is not involved in this aspect of learning. However, we did find a significant positive correlation between the suprapyramidal mossy fiber projection and ER in the non-spatial RM task, but it should be noted that the correlation between the sp-MF and NE did not reach significance. The relationship between the sp-MF and non-spatial RM in the radial maze seems to be fortuitous or very weak at best. Olton” studied fornix-lesioned rats in a similar task and his results also indicate that the hippocampus is not involved in
297
this type of learning since no decrease of learning capacity was observed in lesioned animals. CONCLUSION
Our results might be interpreted in two different ways. The first one is that the hippocampus is involved in the modulation and/or regulation of the processing of spatial, but not of non-spatial information, as has been hypothesized by O’Keefe and Nadel.” However, the two experiments presented here not only differed in the spatial or non-spatial nature of the cues offered to the subjects, but also in the type of memory that had to be used in these tasks. For a good performance in the spatial task, it is necessary for the subject to remember exactly which arms have already been visited within a particular session: a WM task (cf. Ref. 21). In the non-spatial task, animals had to learn that a closed door indicated the presence, an open door the absence of a food reward: an RM2’ task. Thus, the different correlations found in the present experiments might also be interpreted to indicate that the hippocampus is involved in WM, but not in RM, as hypothesized by Olton et al.” However, in previous experiments we obtained positive correlations of the size of the iip-MF with spatial RM in a water maze24 and with habituation and exploration in an open-field’,’ (which also involves RM). Thus, the results of our experiments employing a number of different tasks suggest that the hippocampal iip-MF have an important function in the regulation and/or modulation of spatial RM and WM, but not in non-spatial RM. In the light of the hippocampal working memory theory, 21 it will be interesting to see whether the iip-MF also covary with non-spatial WM abilities. Further experiments addressing the latter problem are now in progress. wish to thank Prof. F. Vogel (Heidelberg) for reading the manuscript, Gary G. Mueller (Frankfurt) for correcting the English, and Prof. W. Busel-
Acknowledgements-We
maier (Heidelberg)
for providing
mice and testing facilities.
REFERENCES
I. Ammassari-Teule M. and Caprioli A. (1985) Spatial learning and memory, maze running strategies and cholinergic mechanisms in two inbred strains of mice. Behav. Brain Res. 17, 9-16. 2. Bernstein D., Olton D. S., Ingram D. K., Wailer S. B., Reynolds M. A. and London E. D. (1985) Radial maze performance in young and aged mice: neurochemical correlates. Pharmac. Biochem. Behau. 22, 301-307. 3. Bolhuis J. J., Bijlsma S. and Ansmink P. (1986) Exponential decay of spatial memory of rats in a radial-maze. Behav. neural Biol. 46, 115-I22. 4. Crusio W. E., Genthner-Grimm G. and Schwegler H. (1986) A quantitative-genetic analysis of hippocampal variation in the mouse. J. Neurogenet. 3, 203-214. 5. Crusio W. E. and Schwegler H. (1987) Hippocampal mossy fiber distribution covaries with open-field habituation in the mouse. Behau. Brain Res. 26, 153-158. 6. Crusio W. E., Schwegler H. and Lipp H.-P. (1987) Radial-maze performance and structural variation of the hippocampus in mice: a correlation with mossy fibre distribution. Brain Res. 425, 182-185. I. Crusio W. E., Schwegler H. and van Abeelen J. H. F. (1989) Behavioral responses to novelty and structural variation of the hippocampus in mice. II. Multivariate genetic analysis. Behav. Brain Res. 32, 81-88. 8. Danscher G. and Zimmer J. (1978) An improved Timm sulfide silver method for light and electron microscopic localization of heavy metals in biological tissues. Histochemistry 55, 2740. 9. Gallistel C. R. (1989) Animal cognition: the representation of space, time and number. A. Rev. Psychol. 40, 1555189. 10. Goldowitz D. and Koch J. (1986) Performance of normal and neurological mutant mice on radial arm maze and active avoidance tasks. Behao. neural Biol. 46, 216-226.
298
H. SCHWEGLER et ul.
11. Heimrich B., Schwegler H. and Crusio W. E. (1985) Hippocampal variation between the inbred mouse strains C3H/HeJ and DBA/2: a quantitative-genetic analysis. J. Neurogenef. 2, 389401. 12. Heimrich B., Schwegler H., Crusio W. E. and Buselmaier W. (1988) Substrain divergence in C3H inbred mice. Be&,. Gener. 18, 671674. 13. Idrobo F., Nandy K., Mostofsky D. I., Blatt L. and Nandy L. (1987) Dietary restriction: effects on radial-maze learning and lipofuscin pigment deposition in the hippocampal and frontal cortex. Arch. Gerontol. Geriutr. 6, 355-362. 14. Lipp H.-P. and Schwegler H. (1983) Hippocampal mossy fibers and avoidance learning. In Genetics of the Brain (ed. Lieblich I.), pp. 325-364. Elsevier Biomedical, Amsterdam. 15. Lipp H.-P., Schwegler H. and Leisinger-Trigona M.-C. (1986) lnfrapyramidal mossy fibers and Y-maze discrimination learning of mice: a positive correlation. Experienfia 42, 676. 16. Mishima N., Higashitani F., Teraoka K. and Yoshioka R. (1986) Sex differences in appetitive learning of mice. Physiol. Behav. 37, 263-268. 17. Mizumori S. J. Y., Rosenzweig M. R. and Kermisch M. G. (1982) Failure of mice to demonstrate spatial memory in the radial-maze. Behav. neural Biol. 35, 33-45. 18. Nadel L. and MacDonald L. (1980) Hippocampus: cognitive map or working memory? Behav. neural Biol. 29,405409. 19. O’Keefe J. and Nadel L. (1978) The Hippocampus as a Cognitive Map. Clarendon Press, Oxford. 20. Olton D. S. (1977) Spatial memory. Sci. Am. 236, 82-98. 21. Olton D. S., Handelmann G. E. and Walker J. A. (1979) Hippocampus, space, and memory. Behav. Brain Sci. 2, 313-365. across time: the hippocampus as a temporary memory store. Behav. Bruin Sci. 22. Rawlins J. N. P. (1985) Associations 8, 479496. 23. Reinstein D. K., DeBoissiere T., Robinson N. and Wurtman R. J. (1983) Radial maze performance in three strains of mice: role of the fimbriaifornix. Brain Res. 263, 172-176. learning in the mouse correlates with 24. Schwegler H., Crusio W. E., Lipp H.-P. and Heimrich B. (1988) Water-maze variation in hippocampal morphology. Behan. Gene/. 18, 153-165. 25. Schwegler H. and Lipp H.-P. (1983) Hereditary covariation of neuronal circuitry and behavior: correlation between the proportions of hippocampal synaptic fields in the regio inferior and two-way avoidance in mice and rats. Behav. Brain Res. 7, l-38. New York. 26. Siegel S. (1956) Nonparametric Statisrics ,for the Behavioral Sciences. McGraw~Hill, nomenclature for inbred strains of mice: eight listing. Cancer Res. 45, 9455977. 27. Staats J. (1985) Standardized memory indexing theory. Behac. Neurosci. 100, 147-154. 28. Teyler T. J. and DiScenna P. (1986) The hippocampal (Accepted 6 September 1989)
