The Journal of Neuroscience, February 1995, f5(2): 1162-i 171
Variants of Olfactory Memory and Their Dependencies on the Hippocampal Formation
Ursula Sttiubli,’ To-Tam Le,2 and Gary Lynch*
‘Center for Neural Science, New York University, New York, New York 10003 and *Center for the Neurobiology of Learning and Memory, University of California, Irvine, California 92717
Olfactory memory in control rats and in animals with entor- pal pyramidal cells (e.g., Hjorth-Simonsen, 1972; Witter, 1993). hinal cortex lesions was tested in four paradigms: (1) a known Moreover, physiological activity in rat hippocampus becomes correct odor was present in a group of familiar but nonre- synchronized with that in the olfactory bulb and cortex during warded odors, (2) six known correct odors were simulta- odor sampling(Macrides et al., 1982). Theseobservations have neously present in a maze, (3) correct responses required prompted speculationand experimentation concerningthe pos- the learning of associations between odors and objects, and sible contributions of the several stagesof the olfactory-hip- (4) six odors, each associated with a choice between two pocampal circuit to the encoding and use of memory. Lesions objects, were presented simultaneously. Control rats had no to the lateral entorhinal cortex, which separatethe hippocampus difficulty with the first problem and avoided repeating se- from its primary source of olfactory input, did not detectably lections in the second; this latter behavior resembles that affect the ability of rats to perform odor discriminations learned reported for spatial mazes but, in the present experiments, prior to surgery although they did disrupt the learning of new was not dependent upon memory for the configuration of discriminations (Staubli et al., 1984, 1986).This result suggested pertinent cues. Control animals varied considerably in their that the olfactory cortex and its connections with telencephalic acquisition of odor-object associations with only a subgroup regionsother than hippocampusare the sitesin which odors are learning every set of pairings. These latter animals also per- recognized and assignedsignificance. In accord with this idea, formed well in the fourth task and, as indicated by post hoc long-term potentiation was found to develop in the piriform analyses, developed complex strategies in dealing with the cortex when patterned stimulation of the lateral olfactory tract problem of serial odor-object pairs. Lesioned animals had was usedas a discriminative cue (Roman et al., 1987). no difficulty in selecting correct odors learned prior to sur- Hypothesesabout which aspectsof olfactory memory might gery (problem one) but repeated their choices in problem be encoded in the hippocampus have, for the most part, been two. This latter result suggests that hippocampus contrib- borrowed from studies using nonolfactory cues, for example, utes to the transient memory of prior choices for odors as the capacity of rats to rememberchoices made earlier in a spatial it does for prior choices in spatial mazes. Entorhinal rats maze (Olton et al., 1979) and the increasedfiring of field CA1 were able to form odor-object associations (problem three), neuronswhen an improper sequenceof visual cuesoccurs (Ranck, and a subgroup of the animals periodically succeeded in 1973). The extent to which these results hold for odor cues is doing a long series of such choices (problem four), though unknown and indeed there is controversy about their proper with less frequency than controls. These results indicate that interpretation with regard to memory of visual stimuli (seebe- rats use both long-term memory and transient memory in low). Accordingly, the present experiments tested olfactory dealing with olfactory problems and suggest that the second memory acrossseveral behavioral paradigms with the goal of of these is dependent upon a hippocampal process that providing information needed for evaluation and further de- encodes a type of information other than the relationship velopment ofhypotheses regardingpossible encoding operations between cues. by the different stagesof the olfactory-hippocampal circuit. The [Key words: o/factory memory, paired-associates, cross- first problem involved long-term memory and required rats to modal learning, olfactory-hippocampal circuit, hippocam- selecta known correct cue from a group of simultaneously pres- pus, entorhinal cortex, working memory, cognitive map, spa- ent familiar odors. Lesionsof the lateral entorhinal cortex were tial memory] usedto test if hippocampus contributes to recognition memory in a situation in which the animal is required to reject several Anatomical and physiological studiesindicate that the olfactory possiblechoices. The secondparadigm examined the possibility system provides substantial and direct input to the hippocam- that rats have a transient or “working” memory which allows pus. The lateral entorhinal cortex, the posterior extension of the them to identify odors respondedto earlier in a test session.It olfactory cortex, denselyinnervates the granule cells of the den- hasbeen proposedthat memory of this kind, in contrast to long- tate gyms and also projects monosynaptically to the hippocam- term recognition/significance memory, is dependent upon hip- pocampus,and in particular the mossy fiber synapsesbetween the dentate gyrus and field CA3 (Lynch and Staubli, 1991; Lynch Received Apr. 22, 1994; revised July 27, 1994; accepted July 27, 1994. et al., 1991; Staubli, 1992). It was also anticipated that the We thank Gene Go for excellent technical assistance. outcome of the experiment would be relevant to a controversy Correspondence should be addressed to Dr. Ursula Stiubli, New York Uni- versity, Center for Neural Science, 4 Washington Place, New York, NY 10003. concerning the nature of hippocampal contributions to memory Copyright 0 1995 Society for Neuroscience 0270-6474/95/l 5 1162-10$05.00/O involving spatial cues. Olton and coworkers (1979) showedthat The Journal of Neuroscience, February 1995, 15(2) 1163 rats with hippocampal damage make numerous reentries into B the arms of a radial maze during a given test session while control animals typically avoid earlier choices. They interpreted this as showing that the hippocampus encodes a transient mem- ory of prior selections. O’Keefe and others (e.g., O’Keefe and Nadel, 1978; Rasmussen et al., 1989) argue that reentry errors in hippocampal animals are secondary to the disruption of a spatial map, that is, a stable memory of the configuration of spatial cues. Tests with odor cues should be useful for deter- mining if memory of prior choices requires hippocampus in- dependent of any role it plays in encoding configurations of cues. The third paradigm involved the learning of associations be- C A--a vs b A*a vs d tween odors and objects. Youngentob and colleagues (1990, 1991) have recently shown that rats are adept at associating odors with visual cues in an open field environment. We were interested in testing if animals are able to acquire paired-asso- ciates involving odors and intra-maze objects and the extent to which this is dependent upon hippocampus. The hippocampus is the most obvious site for convergence of olfactory and visual information-as noted, it receives a very large input from the posterior extension of olfactory cortex (e.g., Hjorth-Simonsen, B--avsb LJ E--bvse 1972; Witter, 1993) and is known to be responsive to visual B--a vs b cues (e.g., O’Keefe and Nadel, 1978; O’Keefe and Conway, 1978). Figure I. Illustration ofthe various paradigms used in the study. Shape The dense interconnectivity within hippocampus would seem- and proportions of the 12 arm maze are schematized and do not exactly ingly provide a substrate for associating contiguous odors and correspond to the actual maze (dark arm = start alley). A, Long-term objects. However, O’Keefe and Nadel (1978) have made the recognition memory: choose correct odor (A) from a group of five fa- important point that hippocampus is not needed for stimulus- miliar odors; B, olfactory working memory: sample each of six odors only once within a trial, irrespective of odor rotations and delays in- response learning involving intra-maze cues and specific re- terposed after the third choice; C, Cross-modal paired-associates be- sponses, for example, maze learning in the absence of extramaze, tween odors and objects: associate odor A with object a and odor B spatial cues. From this perspective, it might well be the case with object b (phase l), then odor C with object c and odor D with that an extrahippocampal system encodes and uses intra-maze object d (phase 2), and last odor E with object e and odor F with object f (phase 3); D, serial odor-object associations: solve all six odor-object associations of odors with objects. The fourth paradigm com- associations within a single trial. bined problems two and three: rats were required to make a series of odor choices each of which was followed by an object choice. This provided (1) a test of the rat’s ability to integrate through an opening in the hurdle located at the entrance to each alley. transient memory (which odors were already selected) and long- A fan mounted 25 cm above the center platform served to exhaust the odors. term memory (which object is associated with the odor just Experiment 1: long-term memory for sign$cunce of simultaneously selected), and (2) an opportunity to determine if any deficits in present odors. Rats were trained to discriminate between five odors transient memory of choices caused by hippocampal denerva- ejected simultaneously from different arms of the radial maze (see Fig. tion would impair utilization of long-term memory. 1A). Two sets of four odors were used, the members of the first set [celery (X), rose (Y), garlic (Z), basil (W)] were present in each trial and always marked arms with empty food wells; the fifth odor was randomly Materials and Methods selected in each trial from the second odor set [coconut (A), banana (B), Subjects. Ten male Sprague-Dawley rats, 3 months of age at the begin- chocolate (C) and cheese (D)] and always denoted an arm containing a ning of the study, were used. They were housed individually and trained food reward (cheese bits, banana chips, coconut chips, and chocolate and tested during the dark phase of their 12 hr: 12 hr light-dark cycle. chips, respectively). The five odors were assigned to five different arms Access to food was restricted to maintain the animals at 90% of control in every trial and one of the arms containing no odor served as start body weight, and water was available ad libitum. The same 10 animals alley. The two arms directly adjacent to the right and left of the start were run in each of four experiments during a period of over 12 months. alley were never used. Animals were allowed to make one choice per Behavioral apparatus. The apparatus consisted of a modified elevated trial and were then removed to their home cage for lo-15 min; five radial maze made of painted wood that was kept in a semidark room trials were given per day. Training was conducted for several weeks with minimal lighting. The center platform was 35 cm in diameter with until each animal made three or more correct choices per day across 5 12 radial arms that were 90 cm long, 7 cm wide at the entrance and consecutive days. Five of the animals were randomly picked and sub- extended progressively to a width of 25 cm at the distal end. Each arm jected to bilateral electrolytic lesions of the entorhinal cortex. Testing was fully enclosed by 25 cm high walls except at the entrance where a was resumed one week later during 10 d, using identical procedures and 5 cm high wooden hurdle was located which the rat had to climb over odors as preoperatively. to get access from the center platform. A hole drilled in the floor at the Memory for odors sampled during a testing session: “working mem- distal end of each arm served as hidden food well. Odors were prepared ory” tusk. Six odors were used that were released simultaneously from by enclosing a small amount (- 3 ml) of a given odor concentrate (In- randomly selected and baited arms of the radial maze as shown in the ternational Flavors and Fragrances, Inc.) into a 50 ml Erlenmeyer flask illustration of Figure 1B. The odors included the set of four odors and gassing it with air monitored by a flow-meter at a rate of 0.1-1.0 marking food rewards in the first experiment plus two additional odors, liters/min (depending on the intensity of the odor concentrate used). that is, butter odor (E) with Fruit Loop bits as reward, and walnut odor The odor-saturated air leaving the flask then joined with a clean air (F) with walnut bits as reward. Correct performance consisted in re- stream regulated by a second flow-meter at 4.5-5.5 liters/min, such that sponding to each odor once and collecting the food. Rats received one the final volume combined of clean and odor-saturated air delivered trial per day for several weeks. The actual experiment involved 15 d of per minute was identical for each odor. The odorized air was carried testing with no delay, followed by 15 trials with a delay after the third via Tygon tubing to the maze and ejected towards the center platform choice during which the rat was confined inside a large nontransparent 1164 Staubli et al. - Hippocampus and Variants of Olfactory Memory
~ ~ a s$g&+.$*
# * Banana Butter Coconut Cl-______--____------+ $ . lesion O'- I I I I I I I I I I -5 -4 -3 -2 -1 t 1 2 3 4 5 6 7 8 9 10 Days i:. ,.:G :;;;xi;;::::::::::::::::::::::::::::::::::. ,,,::::: i;;jj;j .:::::::::::::::::::::::::::: :: : : : : : :: :: : : : : :: :: : : : Figure 3. Effects of bilateral entorhinal cortex lesions on recognition ..:::::::::::::::.. memory, that is, the ability to remember and distinguish the correct @ one out of five familiar odors presented simultaneously. The task in- Cheese Chocolate Walnut volved a total of four correct and four incorrect odors, with the same four incorrect odors uresented in every trial while the fifth and correct odor varied randomly across five daily trials. Performance is shown for Figure 2. Illustration of the six three-dimensional objects, each of the last 5 preoperative days and for 10 d following recovery from the which was paired with the odor listed underneath. The objects were lesion. glued onto a circular cardboard of 3.5 cm in diameter, large enough to cover a food well. ventral -6.0, -4.0, -2.0; posterior -0.8, lateral 4.0, ventral -7.0, -5.0, -3.0; posterior -0.8, lateral 5.0, ventral -7.0, -5.0, -3.0; all cylinder in the center area and the maze was either rotated by 90” or at an angle of 10” from the vertical midline and with the toothbar set all six odors were reassigned to different arms. These control procedures at -5.0. Current of 1 mA was passed for 45 set at each of the above interposed a delay of approximately 3-5 min between the first and described nine placements using an insulated stainless steel electrode second set of three choices. that was cut at the tip. The five control animals were prepared identically Cross-modal associations: “paired-associates” between odors and ob- except that no current was passed (sham lesions). jects. Each odor used in the preceding experiment was paired with one Histological processing. After completion of testing all animals were of six three-dimensional object as illustrated in Figure 2. An insert with deeply anesthetized and perfused transcardially with buffered saline two recessed food wells was placed at the end of all arms. Training was solution followed by 10% buffered formaldehyde solution. The brains conducted in three phases that each involved two different odor-object were postfixed in formaldehyde solution for 24 hr and sunk in 10% pairs and lasted several weeks. To obtain a reward, the animal had to buffered sucrose solution overnight. They were then frozen and cut enter both odor-releasing arms and select the correct of two objects. A coronally at 40 Frn starting at the septal end of the hippocampus. Hor- food reward was hidden underneath the correct object, removing the izontal sections were taken from the last third of hippocampus and wrong object uncovered an empty well. The animals encountered the adjacent entorhinal cortex. Every fourth coronal section was stained for same object pair in both arms such that the correct object in one arm acetylcholinesterase, a histochemical procedure which allows determi- was incorrect in the second arm, and vice versa (for an illustration of nation ofthe extent of denervation in hippocampus produced by damage the experimental design, see Fig. 1C). Left-right positions of baited food to the retrohippocampal area (Lynch et al., 1972; Cotman et al., 1973). wells varied randomly within and across five daily trials. Animals were Every fourth horizontal section was stained with cresyl violet to verify removed to their home cage for 15-25 min between trials. A strong spot the extent of actual tissue loss in the area of lesion. All sections were light positioned on the ceiling above the center of the maze served to mounted on microscope slides and examined with a desk projector. enhance visibility of the objects. Training on each set of odor-object pairs was carried out for several weeks. To test if object choice was Results guided by food smells, probe trials were carried out with no food present Long-term memory for significanceof multiple, (the animal received the reward from the experimenter upon displacing the correct object) or with food under the wrong object. simultaneously present odors Serial odor-object associations. The serial version of the paired-as- Olton and colleagues(1979) trained rats in a radial maze in sociate task required that all six odor-object pairs introduced in the which a subsetof arms (spatial locations) was always rewarded preceding experiment were solved sequentially within a single trial. and a second set was not, and found that hippocampal lesions Thus, all six odors were released simultaneously and each odor led to two objects (see Fig. ID). In contrast to the preceding experiment the did not cause the animals to confuse the two groups of arms incorrect object placed with a particular odor varied from trial to trial. (i.e., the experimental animals entered correct locations and Individual odors were released from different arms in each trial. A trial avoided incorrect arms). These authors considered the long- was terminated upon displacement of a wrong object, while repeated term memory of correct versus incorrect positions in the maze entries into odor arms were noted but did not result in termination of as a kind of “reference” memory and concluded from their a trial. Each animal received five daily trials for 18 d with the animal in the home cage for 15-25 min between trials. results that it is not dependent upon hippocampus. The first Lesions. Bilateral entorhinal cortex lesions were performed since (1) study tested if these observations hold true for olfactory mem- our previous studies on the hippocampal role in olfactory memory used ory, that is, are rats able to distinguish the significanceof several entorhinal lesions to disconnect the hippocampus from its olfactory simultaneouslypresent odors and do they retain this ability after input (Staubli et al., 1984, 1986) and (2) studies by others on spatial memory have indicated that damage to the entorhinal cortex results in lesionsto the primary olfactory input to hippocampus?Several deficit patterns similar to those found after extensive damage to the weeks were required to train the animals on this task, which hiDDOCaIDDUS nrooer (Jarrard et al., 1984: Schenk and Morris, 1985; involved different phasesof learning, such as acclimation to the Rasmussen et -al.,&1989). Animals were pretreated with atropine (0.1 testing situation, learning that odors are the relevant cues (i.e., mg) to reduce salivation and anesthetized with a mixture of ketamine ignoring spatial and nonolfactory intramaze cues)and then mas- (50 mg/kg) and Rompun (10 mg/kg) injected intraperitoneally. Bilateral electrolytic ablations of the entorhinal cortex were conducted identical tering a few simple “rules” (leave the current arm, enter the to protocols described earlier (Staubli et al., 1984, 1986). Stereotaxic decision point, follow an odor, find the reward, etc.). As illus- coordinates in reference to lambda were posterior -0.8, lateral 3.0, trated in Figure 3, during the last 5 d before surgery both the The Journal of Neuroscience, February 1995, 15(Z) 1165 intact and control groups were equally successful in picking A B correct odors. In the first five postoperative testing days exper- Lesion 801 81 imental animals confused the two subsets of odors more fre- quently than controls [t(8) = 3.09, p < 0.015, two-tailed t test, 260. days l-5 postsurgery] as well as compared to their own pre- I= . operative performance [t(4) = 6.32 p < 0.003, two-tailed paired E 40. t test, days l-5 pre- vs postsurgery]. This deficit was temporary i! . as evidenced by a recovery of their scores during days 6-10 to levels indistinguishable from preoperative performance [t(4) = 0.69, p -c 0.52, days l-5 pre vs 6-10 post]. ABCDEF Memory for odorssampled during a testing session:“working memory” C D WithDelay Rats in radial mazesquickly learn to avoid arms from which 80 they have already removed the reward. Their memory in this No Delay NoDelay regard is impressive since the animals enter and subsequently 2 60 avoid reenteringmany arms even when a delay of severalhours I= is interposed in the middle of the testing session.Olton and E 40 40- coworkers(1979) suggestthat this type of memory is analogous 8 to “working memory” (in essencea list of items tested during a" 20 20- a several hour period) and argue that its disruption is at least in part responsiblefor the profound impairment in spatial mazes 0 0 :iiiii 23456X 456X produced by hippocampal lesions. A counterhypothesis (e.g., 23456X 456X O’Keefe and Nadel, 1978; Rasmussenet al., 1989) is that the Choice Number Choice Number hippocampusencodes spatial relationshipsusing extramaze vi- Figure 4. Effectsof bilateralentorhinal cortex lesions on memoryfor sual cuesand that lesionsto the structure disrupt the utilization odorsalready responded to earlierin the test session,that is, olfactory of suchinformation thereby producing random behavior in the workingmemory. A, Percentageof total trialsin whichindividual odors maze (i.e., numerous reentries into already sampledarms). In wereselected as first choiceby the 10animals (A. coconut;B, banana; C, chocolate;D, cheese;E, butter; F, walnut).B, Mean and median this view, the effectsof hippocampal lesionson what Olton and (MDN) numberof reentryerrors committed by the controland exper- colleagues(1979) refer to as working memory are secondary to imentalgroup in 15 trials without delay. C, Choiceon whichcontrol a disturbanceof memory involving spatial relationships.In the animalsmade their first reentry errors in trials in which a delay was or presentexperiment we asked(1) if rats possessa working mem- was not interposed after the third odor selection. X denotes the per- centage of trials in which no reentries were made. D same as in C but ory for odors in a situation in which spatial relationships are for the experimental group. irrelevant to performanceand (2) if sucha memory is disrupted by disconnectingthe olfactory efferents to hippocampus. Overall, the animals were quick at learning this task. As il- lustrated in Figure 4A, they randomly picked any of the six > 0.20; for control rats: D = 0.57 (no delay), D = 0.47 (delay); odors as their first choice in a trial (chancelevel, 16.7%), with p < 0.011; however, the entorhinal animals did not perseverate the exception of cheeseodor (D) which was selectedin approx- as they did not repeat their initial choices with a frequency imately a third of total trials (31%), significantly more often greaterthan that expectedfrom chance.These observations pro- than butter-(E) [t(8) = 3.28;p
P 3 6 9
“Ol N 0 II 0123456 0123456 0123456 0123456 012345 i6 60 (I) : 2 1 N 5 7 8 10 a 7o i: 8 340’ i$ 50 i5 l- 30 ‘ij20- E 8 SETS OF PAIRED-ASSOCIATES BO- Figure 5. Acquisition of odor-object associations (paired-associate 0123456 0123456 0123456 0123456 0123456 learning). Five control animals (open bars)and five animals with bilat- Choices eral entorhinal cortex lesions (solid bars) were tested for their ability to B associate six familiar odors with a specific three-dimensional object. I Training was conducted in three phases (I, 2, 3) entailing each a set of ii? 10 1 paired-associates (two different odor-object pairs). Bars denote the ‘CE 8 - OControl amount of correct odor selections made in 25 trials during the last 5 d of a training phase. Stippledhorizontal lines correspond to chance per- 2 6- w Lesion formance (50%) and learning criterion (80%).Letters above bars cor- respond to individual rats. : 4- 5 2- third sets. Figure 5 illustrates performance during the last 5 d = o- of training on each set. Three of the animals in the control group 2 3 4 5 6 I Choice Number learned all odor-object associations (> 80% correct), a fourth 1 animal acquired the first set but failed on the second and third Figure6. Serial odor-object associations involving six simultaneously set, while the fifth animal never surpassed chance level (50%). present odor-object pairs. A, Correct choices within a trial. Percent trials In the entorhinal group, one animal acquired all three sets, a in which intact animals (top)and animals with bilateral entorhinal cortex second animal mastered the first two sets, a third animal learned lesions (bottom)made from 0-6 correct choices. Displacing the incorrect the first set well (> 80% correct) and the second one reasonably object resulted in termination of the trial. The results shown were col- well (72%), and a fourth animal only acquired the first set. In lected over 15 d of five daily trials. B, Working memory errors. Mean number of repeated visits to the same odors by entorhinal and control all, the scores of the two most deficient animals in the entorhinal animals in trials with 2-6 correct choices. The experimental group made group (5,7) and the control group (P,3) were comparable; sim- significantly more errors than controls (p < 0.02, p < 0.0007,p < 0.04 ilarly, the best learners in the entorhinal group (8,lO) were not for choice number 4, 5, and 6, respectively; two-tailed t test). obviously different from the three animals with highest scores in the control group (M,6,9). Probe trials involving a single odor released from two loca- the rat to sample each of six odors within the same trial and to tions indicated that the high scores of the above subset of leam- select the correct object associated with each specific odor. ers were not due to object alternation (i.e., if object #1 is re- Working memory components of the task were minimized by warded to first odor, choose object #2 in second arm). Similarly, not penalizing the rats for approaching already selected odors. tests in which (1) both objects in a given arm covered empty Correct performance thus entailed moving through the maze wells or (2) food was placed under the incorrect object did not using cues (odors) that had no explicit relationships with each interfere with the superior performance of these animals. other and making a series of decisions based on associations in long-term (reference) memory. Performancein a maze requiring serial odor-object The three intact animals (M,6,9) that had learned all odor- associations object relationships in the preceding experiment performed quite This study measured the effect of hippocampal denervation on well on this paradigm; that is, they accomplished error-free runs the utilization of long-term memory in circumstances that re- in 49%, 47%, and 65% of total trials, respectively (Fig. 6A). quire the rat to maintain in parallel a particular mode of search- Similarly, the two entorhinal rats that had performed well in ing, cue approach and sequential decision making. In essence, the previous study (8,lO) made six correct choices in 21% and this paradigm combined experiments 2 and 3, and thus required 30% of total trials, respectively (chance, 2%). The remaining The Journal of Neuroscience, February 1995, M(2) 1167
10% Coconut (A) B 50 First Choice I ul 40 2 1 2 3 4 5 6 1 2 3 4 5 6 c 30 5 15% Banana (B) E3 20 I
iii a ‘0
0 - zl.rlA B C lL D-E F 1 2 3 4 5 6 1 2 3 4 5 6 Odors 5% Chocolate (C) C 25 175 1 Error Rate
B c 150 + 125 20 a 1 2 3 4 5 6 1 2 3 4 5 6 9 ,o(J ...... 5 E 75 ...... a2 50
2 25 I ’ A B C D E clF 1 2 3 4 5 6 123456 lfldOdors 50 1 18% Butter (E)
r = 0.70
1 2 3 4 5 6 1 2 3 4 5 6 co 50 6% Walnut (F)
25 -10 I 20 40 80 80 100 120 140 160 l-l 1 2 3 4 5 6 1 2 3 4 5 6 Normalized Error Rate (Median) Choice Number
Figure 7. Relationship between type of odor selection, time point of selection (choice number), frequency of selection and error rate for three control rats (M, 6, 9) during performance of the serial odor-object association task. A, Frequency at which each odor was selected as first, second, third, fourth, fifth, or last choice in a trial. Data are expressed relative to total number of choices counted for each odor. Mean values are depicted on the left and median values on the right half, and the number above each subgraph indicates the fraction of total responses made to a particular odor that were incorrect (absolute error rate). B, Percentage of total trials at which each odor was first choice. C, Normalized error rate; that is, percentage incorrect responses to each odor expressed relative to the median number of incorrect responses for all odors. D, Relationship between first choice selection and relative (normalized) error rate, that is, success rate. animals in both groups were impaired and finished the majority correct first choices: control 82 f 5% vs lesion 83 f 4%, mean of trials with three or fewer correct sequences. In cases with four z!z SEM). or more correct choices, entorhinal rats committed significantly Performance of the three control animals (M,6,9) that came more reentry errors (working memory errors) than controls (see closest to mastering the serial odor-object problems was ana- Fig. 6B for significance levels). It is noteworthy, however, that lyzed in more detail (Fig. 7). Three of the odor-object relation- the lesioned group made substantially more correct first choice ships were easier for these animals than the remaining three. selections than expected from random performance (50%) and Thus, incorrect choices were made on 3%, 5%, and 6% of the did not differ from the control group (percentage of trials with trials for three odors (D,C,F) and lo%, 15%, and 18% of the 1166 Stlubli et al. - Hippocampus and Variants of Olfactory Memory
trials for the remaining group (A,B,E) (seeFig. 74. Interestingly A enough, the rats generally selectedone of the odors with a low -10 loo- Walnut (F) error rate (walnut, chocolate, cheese)for their first or second choice and tended to avoid thosewith higher error rates(butter, Chocolate (C) banana, coconut) until late in the trial. Comparing first choice 80 - -8 selectionsbetween odors across all trials confirmed that the three odors with a low relative error rate were selectedmore often as % & first choice than the remaining three odors (Fig. 7&C). Indeed, 2 60 - the probability that an odor was selectedas a first choice was ‘K correlated with the probability of successwith that odor (r = I- 0.70; seeFig. 70); this correlation was significant when evalu- $ 40 - ated by Spearman’srank order correlation (r, = 0.82, p < 0.05). The number of occurrencesfor odor selectionswith low error I- rate (odors F,D,C) decreasedlinearly across successivechoice 20 - -2 numbers (Fig. 84 and increasedfor odors with high error rate (Fig. 8B), and in both casesthe number of errors per selection had no obvious relationship to choice number; that is, error rate for a given odor was independent of when, in a seriesof OJ., . , . ‘,. u’. , . , . !o choices, it was selected.Interestingly, the first odor selectedin 0 123456 i a trial biasedsubsequent odor choices, that is, the animals made Choice Number different secondchoices depending on whether walnut- or choc- I3 olate-odor wasthe first choice (walnut- and chocolate-odorwere first choicesin 2/3 of total trials). Specifically, walnut-odor clear- 100 -15 Butter(E) ly predisposedthe animals to choosecheese-odor(D) as second Banana (B) choice over any other odor (T = 3.75; p < 0.0003), while choc- olate-odor as first choice did not causea significant trend for a -12 80 specific secondchoice (seeFig. 8C). Curiously enough,the two % 1 Q odors that were preferred as first choice (walnut- and chocolate- odor) were not favored as secondchoice. $j 60- -9 g ‘K Histological analyses I- ii Examination of horizontal sectionsrevealed that all lesionsbi- 3 40- -6 3 laterally included the lateral entorhinal cortex, the projection s zone of the lateral olfactory tract. In all casesthe damagealso F extended into medial entorhinal cortex and caudalhippocampus 20 -3 and included part of the subicular complex. Examination of 1 coronal sectionsrevealed a denseband of acetylcholinesterase 1 in the entorhinal projection area of the molecular layer and 010 confirmed the massive deafferentation expected from this pat- 0 1 2 3 4 5 6 7 tern of damage(Fig. 9). No correlation was apparent between Choice Number the extent of the lesion in individual animals and the degreeof impairment in any of the behavioral measures. c Discussion 8o Second Choices After: The present results confirm and extend previous findingsshow- 1 ing that rats have stablememories for odors (Staubli et al., 1986, T 1987, 1989). The task used in experiment one was somewhat 60 - more difficult than the two-odor discrimination paradigm em- 2 Walnut a ployed in the earlier work. The rat was confronted with five ‘K choices with the correct cue varying across days; the greater I- difficulty of the task is evidenced in the higher error ratesthan 3 40- Chocolate E t Figure 8. Lackof relationshipbetween numbers of errorsper selection and time point of selection.Bias for secondodor choiceas a function of first odor selection.A, Frequencyof selection,choice number and frequencyof errorsobserved for subgroupof three odorswith high successrat (lowerror rate); data are from the samegroup ofthree control rats(M, 6, 9) shownin Figure7. B, Sameas in A, but for the subgroup of threeodors with higherror rate. C, Threecontrol rats (M, 6, 9) make ABCDEF ABCDEF differentsecond choices depending on whetherwalnut or chocolateodor Choices wasfirst choice. The Journal of Neuroscience, February 1995, 75(2) 1169
Figure 9. Coronalsections of a sham-lesioned(A) andentorhinal-lesioned (B) brain stainedfor AChE activity. The entorhinalcortex lesioncaused significantshrinking of the outer two thirdsof the molecularlayer (m), that is, the dendritic zoneof stratumgranulosum (Sg) which is normally innervatedbv entorhinalcortex afferents.The wideband of dark stainin the shrunkenarea (indicated by arrows) reflectslesion-induced sprouting of AChE-containingterminals into the deafferentedzone. occurred in two-odor discrimination learning. Lesions cutting useful for testing the relationship of the memory systemsem- the primary olfactory input to hippocampus did not eliminate ployed in the spatial versus olfactory paradigm. Denervation of retention or utilization of the long-term memoriesused in the the hippocampus caused a substantial impairment of perfor- task. These results indicate that recognition memoriesas well mance in the olfactory maze. This could not be ascribed to as the systemsneeded for samplingacross odors, withholding perseveration since the animals began to make reentry errors responses,etc., do not depend upon the flow of olfactory infor- only later in the test session,that is, only after there was a good mation through hippocampus.Physiological studiesusing slices statistical chancefor a repeated choice assumingresponses were and chronic animals have now establishedthat the primary made randomly. Experiment one indicated that rats with de- olfactory cortex hasLTP-like plasticity (Junget al., 1990; Kanter nervated hippocampi had no difficulty in discriminating be- and Haberly, 1990) and that patterned stimulation of the lateral tween the odors and in withholding incorrect responses.A major olfactory tract usedas a discriminative cue producesboth leam- difference between experiment one and two was that incorrect ing and LTP (Roman et al., 1987); it is thus reasonableto assume odors were encoded into long-term memory in the first study that the cortex itself provides a site for encoding of recognition and into transient memory in the secondstudy, that is, an odor memories.The extensive projections from the periamygdaloid was incorrect becauseof recent (positive) experiencewith it. It subdivision of olfactory cortex into the amygdala (Ottersen, is thus not unreasonableto assumewith Olton and colleagues 1982) could then be usedby the brain to evaluate the value of (Olton et al., 1979; Olton and Feustle, 1981; Raffaeleand Olton, the cuewith amygdala-striatal connections (Groenewegenet al., 1988) that the hippocampus is the site of transient encodingfor 1980; Kelley et al., 1982) providing for appropriate responses recent decisions(cue-response sequences) and that this function (approach, ignore). (i.e., working memory) is independent of the contributions of Rats were found in the secondstudy to avoid reentering arms hippocampus to long-term memory of spatial relationships. containing odors which had already been responded to earlier The third experiment demonstrated that rats are capableof in the session;the imposition of a delay and a rearrangement acquiring cross-modalpaired-associates between odors and ob- of odor positions caused only a slight disturbance of this be- jects. The first set of pairings was acquired relatively quickly havior. It seemsthen that the animals have a memory system while the remaining two proved to be more difficult; indeed, for recent cue-responsesequences. It was apparent that long- two of the control rats exhibited poor performance even with term memory of the configuration of the odors was not involved extensive training. This variation acrossproblems may reflect in this type of memory since the positions of the odors varied differences in the difficulty of the odor-odor or object-object acrossdays-presumably the animals encodedeach of the six discriminations. Interestingly, the entorhinal group showed a odors and their responsesto them into transient memory. The very similar pattern of acquisition in that the secondand third olfactory memory involved in avoiding repetitive selectionsmay sets of odor-object associationsproved to be more difficult to be analogousto the transient (4-8 hr) memory used in spatial master (seeFig. 5). Two animals with lesionsperformed well mazeswhere extramaze visual cuesguide behavior. In the latter in the task while a third and fourth at least partially mastered case,the memory decays significantly after 6 hr (Gallagher et the first two sets; the remaining subject showed little signsof al., 1983; Staubli et al., 1994), an observation that could be learning. These differencesdid not correlate with performance 1170 Stiiubli et al. * Hippocampus and Variants of Olfactory Memory
in the “working memory” problems; rats with pronounced def- as seems likely, is used in spatial mazes, then it would also be icits on the earlier task did not appear to be particularly impaired independent of sensory modality. While evidence indicative of in the odor-object association paradigm. If the hippocampus is hippocampal contributions to long-term olfactory memory was not needed for this type of learning, then the question arises as not obtained, it is certainly possible that the paradigms used in to where the requisite cross-modality matching might occur. the studies reported here were not appropriate for detecting such The olfactory cortex does not receive afferents from, or send contributions. Relationships between distant (e.g., extramaze, efferents to, sensory neocortex. Each of the sensory cortices spatial) cues figure prominently in the physiological and behav- projects to the striatal complex, but recent anatomical studies ioral literatures (e.g., O’Keefe and Conway, 1978; Muller et al., indicate that the target areas within that structure are segregated 1987) on hippocampus, but were not essential to correct per- from each other; for example, olfactory tubercle to ventral stria- formance in the present experiments. Accordingly, it would be turn (Heimer, 1978) and visual cortex to dorsal striatum (Nauta, of interest to test if rats are able to quickly form associations 1979). Complex interactions involving frontal cortex and stria- between odors and spatial locations or distant objects and, if turn (Goldman and Nauta, 1977; Groenewegen et al., 1990) so, if memory for such relationships is disrupted by lesions to might potentially provide the bases for cross-modal associa- hippocampus. Studies of this kind would have to be designed tions, but the neuroanatomical description of the relevant cir- using behaviors that do not depend upon the transient memory cuitries is not sufficiently precise to develop this idea into a system as, for example, is the case for performance in radial more specific hypothesis. A perhaps more likely site for olfac- mazes. tory/nonolfactory comparisons is the amygdala. This region re- The idea that hippocampus encodes both transient and long- ceives input from sensory neocortex via the perirhinal cortex term forms of memory finds indirect support in work showing and from olfactory cortex via its periamygdaloid extension (Ot- that the structure contains more than one type of synaptic plas- tersen, 1982; Amaral et al., 1992). It is thus conceivable that ticity. Specifically, short bursts of stimulation delivered repeat- an odor influences the response of the amygdala to a subsequent edly to the Schaffer-commissural pathways induce long-term visual input, and it will be of interest to test if damage to peri- potentiation that is stable for days or even weeks (Staubli and rhinal or periamygdaloid cortex disrupts odor-object associa- Lynch, 1987, 1990); equivalent bursts given in pairs produce a tions. weaker potentiation effect that decays over a period of hours The final experiment showed that rats are able to integrate and is absent at 24 hr (Staubli et al., in press). This latter effect the long-term memory of odor-object associations with the tran- is a not unlikely substrate for the type of transient memory sient memory of choices already made. The three control rats described here, while the much more stable LTP is suggestive that came closest to mastering this task organized their selections of a very long lasting encoding of information. Recent studies according to fixed sequences. One factor that entered into this have also provided evidence that potentiation in the mossy fiber was the probability of making the correct object choice for a synapses involves induction and expression processes that have particular odor; that is, higher probability odors were selected little in common with LTP (Staubli et al., 1990; Zalutsky and before low probability ones (Fig. 7). Presumably then even rel- Nicoll, 1990, 1992; Staubli, 1992). The presence of multiple atively small differences in success rates have a lasting influence plasticities prompts the argument that the hippocampus sup- on the organization of olfactory behavior. Odors already sam- ports multiple forms of memory. Hopefully, it will be possible pled and responded to also influenced subsequent choices; that to develop more specific versions of this general hypothesis is, the rats made a different second selection depending on using variants of the behavioral designs employed in the present whether walnut- or chocolate-odor was the first choice (Fig. 8C). experiments. This may relate to the different foods associated with each odor in that the taste of a food item could influence the processing References of a subsequent odor. Because of its interconnections with gus- Amaral DG, Price JL, Pitkaenen A, Carmichael ST (1992) Anatomical tatory regions in brainstem and with hypothalamus (Burton and organization of the primate amygdaloid complex. In: The amygdala: Benjamin, 1971; Bernard et al., 1993) the amygdala seems a neurobiological aspects of emotion, memory, and mental dysfunction likely site to mediate the influences of reward probability and (Aggleton JP, ed), pp l-66. New York: Wiley-Liss. food tastes on odor selections. Bernard JF, Alden M, Besson JM (1993) The organization of the efferent projections from the pontine parabrachial area to the amyg- Two rats with entorhinal lesionsperformed well above chance daloid complex: a Phaseolus vulguris leucoagglutinin(PHA-L) study in the serial odor-object association problem despite commit- in the rat. J CompNeurol 329:201-229. ting numerousreentry errors, confirming that encodingand use Burton H, BenjaminRM (1971) Centralprojection of the gustatory of associationsbetween odors and objects can still be accom- system.In: Chemicalsenses, handbook of sensoryphysiology, Vol plishedafter cutting the major olfactory input to hippocampus. IV (BeidlerLM, ed),pp 148-164.Berlin: Springer. CotmanCW, Matthews DA, Taylor D, Lynch G (1973) Synaptic However, even though the above subset of lesioned animals rearrangementin the dentategyrus: histochemicalevidence of ad- were able to accomplish long series of correct choices, they justmentsafter lesionsin immarureand adult rats. Proc Nat1Acad succeededin doing so at a reduced rate compared to the three Sci USA 70~3473-3477. control rats that came closeto mastering the final task (seeFig. GallagherM, King RA, YoungNB (1983) Opiateantagonists improve spatialmemory. Science 221:975-976. 6A). It is not unlikely that the poorer performanceof the lesioned GoldmanPS, Nauta WJH (1977) An intricately patternedprefronto- rats reflects their greater tendency to make reentry errors. Such caudateprojection in the rhesusmonkey. J CompNeurol 171:369- errors could have an effect on arousal and other state variables 385. known to affect the quality of performance in difficult tasks. GroenewegenHJ, BeckerNE, LohmanAH (1980) Subcorticalaffer- In summary, the present results suggestthat the entorhinal entsof the nucleusaccumbens septi in the cat, studiedwith retrograde axonal transportof horseradishperoxidase and bisbenzimide.Neu- cortex-hippocampal system encodesor prompts the encoding roscience5: 1903-l 9 16. of transient memoriesconcerned with recent choices;this mem- GroenewegenHJ, BerendseHW, WoltersJG, LohmanAH (1990) The ory system was not dependent upon cue configuration and if, anatomicalrelationship of the prefrontalcortex with the striatopal- The Journal of Neuroscience, February 1995, 15(2) 1171
lidal system, the thalamus and the amygdala: evidence for a parallel Raffaele KC, Olton DS (1988) Hippocampal and amygdaloid involve- organization. Prog Brain Res 8595-l 16. ment in working memory for nonspatial stimuli. Behav Neurosci 102: Heimer L (1978) The olfactory cortex and the ventral striatum. In: 349-355. Limbic mechanisms (Livingston M, Homykiewicz H, eds), pp 95- Ranck JB Jr (1973) Studies on single neurons in dorsal hippocampal 187. New York: Plenum. formation and septum in unrestrained rats. Exp Neuro14 1:46 l-555. Hjorth-Simonsen A (1972) Projections of the lateral part of the en- Rasmussen M, Barnes CA, McNaughton BL (1989) A systematic test torhinal area to the hippocampus and fascia dentata. J Comp Neurol of cognitive mapping, working-memory, and temporal discontiguity 146:219-232. theories of hippocampal function. Psychobiology-171335-348. Jarrard LE, Okaichi H, Steward 0, Goldschmidt RB (1984) On the Roman F, Staubli U, Lvnch G (1987) Evidence for svnautic poten- role of hippocampal connections in the performance of place and cue tiation in a cortical.network during learning. Brain Res4 l&22 l-226. tasks: comparisons with damage to hiDDOCamDUS.__ _ Behav Neurosci Schenk F, Morris RGM (1985) Dissociation between components of 98:946-954. spatial memory in rats after recovery from the effects of retrohip- Jung MW, Larson J, Lynch G (1990) Long-term potentiation ofmono- pocampal lesions. Exp Brain Res 48: 1 l-28. svnaptic EPSPs in rat piriform cortex. Svnapse 6:279-283. Staubli U (1992) A peculiar form of potentiation in mossy fiber syn- Kanter-ED, Haberly LB ‘( 1990) NMDA-dependent induction of long- apses. In: Epilepsv research. SUDD~._-, The dentate avrus-< and its role in term potentiation in afferent and association fiber systems of piriform seizures (Ribak-C; Gall CM, Mody I, eds), pp 15 l-l 57. Amsterdam: cortex in vitro. Brain Res 525:175-179. Elsevier. Kelley AE, Domesick VB, Nauta WJH (1982) The amygdalostriatal Staubli U, Lynch G (1987) Stable hippocampal long-term potentiation connections projection in the rat-an anatomical study by anterograde elicited by ‘theta’ pattern stimulation. Brain Res 435:227-234. and retrograde tracing methods. Neuroscience 7:6 13-650. Staubli U, Lynch G (1990) Stable depression of potentiated synaptic Lynch G, Staubli U (199 1) Possible contributions of long-term po- responses in the hippocampus with l-5 Hz stimulation. Brain Res tentiation to the encoding and organization of memory. Brain Res 13:113-118. Rev 16:204-206. Staubli U, Ivy G, Lynch G (1984) Hippocampal denervation causes Lynch G, Matthews DA, Mosko S, Parks T, Cotman C (1972) Induced rapid forgetting of olfactory information in rats. Proc Nat1 Acad Sci acetylcholinesterase-rich layer in rat dentate gyrus following entor- USA 81:5885-5887. hinal lesions. Brain Res 42:3 1 l-3 18. Staubli U, Fraser D, Kessler M, Lynch G (1986) Studies on retrograde Lynch G, Larson J, Staubli, U, Granger R (199 1) Variants of synaptic and anterograde amnesia of olfactory memory after denervation of potentiation and different types of memory operations in hippocam- the hippocampus by entorhinal cortex lesions. Behav Neural Bio146: pus and related structures. In: Memory: organization and locus of 432-444. - change (Squire LR, Weinberger NM, Lynch G, McGaugh J, eds), pp Staubli U, Fraser F, Faraday R, Lynch G (1987) Olfaction and the 330-363. New York: Oxford UP. “data” memorv svstem in rats. Behav Neurosci 101:757-765. Staubli U, Thibault 0, DiLorenzo M, Lynch G (1989) Antagonism Macrides F, Eichenbaum HB, Forbes WB (1982) Temporal relation- of NMDA receptors impairs acquisition but not retention ofolfactory ship between sniffing and the limbic theta rhythm during odor dis- memory. Behav Neurosci 103:54-60. crimination reversal learning. J Neurosci 2: 1705-l 7 17. Staubli U, Larson J, Lynch G (1990) Mossy fiber potentiation and Muller RU, Kubie J, Ranck JB Jr (1987) Spatial firing patterns of long-term potentiation involve different expression mechanisms. Syn- hippocampal complex-spike cells in a fixed environment. J Neurosci apse 5:333-335. 7:1935-1950. Staubli U, Rogers G, Lynch G (1994) Facilitation of glutamate re- Nauta WJH (1979) A proposed conceptual reorganization of the basal ceptors enhances memory. Proc Nat1 Acad Sci USA 9 I:777-78 1. aanalia and telencenhalon. Neuroscience 4: 1875-l 88 1. Staubli U, Perez Y, Xu F, Rogers G, Ingvar H, Stone-Elander S, Lynch O’Keefe JA, Conway bH (1978) Hippocampal place cells in the freely G (in press) Centrallv active modulators of slutamate (AMPA) re- moving rat: whv thev fire where thev fire. EXD Brain Res 31:573-590. ceptors facilitate the induction of LTP in vivo. Proc Nat1 Acad Sci O’Keefe J, Nadei L (1978) The hippocampus as a cognitive map. USA, in press. London: Clarendon. Witter MP (1993) Organization of the entorhinal-hippocampal sys- Olton DS, Feustle WA (198 1) Hippocampal function required for tem: a review of current anatomical data. Hippocampus 3:33-44. nonspatial working memory. Exp Brain Res 41:380-389. Youngentob SL, Markert LM, Mozell MM, Homung DE (1990) A Olton DS, Becker JT, Handelmann GE (1979) Hippocampus, space method for establishing a five odorant identification confusion matrix and memory. Brain Behav Sci 2:3 13-365. task in rats. Physiol Behav 47:1053-1059. Ottersen OP (1982) Connections of the amygdala of the rat. IV. Cor- Youngentob SL, Markert LM, Hill TW, Matyas EP, Mozell MM (1991) ticoamygdaloid and intraamygdaloid connections as studied with ax- Odor identification in rats: an undate. Phvsiol Behav 49:1293-1296. onal transport of horseradish peroxidase. J Comp Neurol205:30-48. Zalutsky RA, Nicoll R (1990) Comparisons of two forms of long-term Otto T, Eichenbaum H (1992) Complimentary roles of the orbital potentiation in single hippocampai neurons. Science 248: 16 19-1624. prefrontal cortex and the perirhinal-entorhinal cortices in an odor- Zalutsky RA, Nicoll RA (1992) Mossy fiber long-term potentiation guided delayed-nonmatching-to-sample task. Behav Neurosci 106: shows specificity but no apparent cooperativity. Neurosci Lett 138: 762-775. 193-197.