Proc. Nati. Acad. Sci. USAVol. 83, pp. 2728-2732, April 1986Neurobiology

Graft-induced behavioral recovery in an animal model ofHuntington disease

(neuronal transplantation)

OLE ISACSON*t, STEPHEN B. DUNNETTt, AND ANDERS BJ6RKLUND**Department of Histology, University of Lund, Biskopsgatan 5, S-22362 Lund, Sweden; and *Department of Experimental Psychology, University ofCambridge, Downing Street, Cambridge CB2 3EB, England

Communicated by Tomas Hokfelt, December 4, 1985

ABSTRACT Bilateral ibotenic acid lesions of the antero-medial neostriatum produce neuropathological and behavioralchanges in rats that are characterized by locomotor hyperac-tivity and severe maze learning impairments, which can beviewed as analogous to changes seen in Huntington disease.Grafts of fetal striatal neurons, implanted either into thelesioned striatum or into the denervated globus pallidus,reduced both the learning impairments and the locomotorhyperactivity, probably via different mechanisms. The resultsdemonstrate the capacity of neural implants for functionalneuronal replacement and promotion of functional recoveryafter damage to a maijor telencephalic structure participatingin complex cognitive and motoric behaviors.

Neurodegenerative diseases in man represent neuropath-ological conditions where rather selective and anatomicallyrestricted neuronal damage results in characteristic and oftensevere behavioral deterioration. There is accumulating evi-dence from experimental work in rodents that intracerebralimplants of fetal aminergic neurons at least to some degreecan substitute, both anatomically and functionally, for lostintrinsic pathways in the brain. This is based, in particular, onresults obtained with grafts of mesencephalic dopamineneurons in rats with lesions that mimic the dopamine defi-ciency in patients with Parkinson disease (1, 2) and withgrafts of basal forebrain cholinergic neurons in rats withlesions that reproduce the cholinergic deficit in patients withdementia of the Alzheimer type (3-5). In these types ofexperimental lesions the primary damage is confined to suchtypes of neurons, operating with acetylcholine or a mono-amine as their transmitter, which normally act as tonicregulatory or level-setting systems in the brain. Hence, it isconceivable that the functional effects of grafted dopamin-ergic and cholinergic neurons could be mediated via arelatively nonspecific tonic release of the transmitter intotheir local environment.

Huntington chorea is characterized by a massive neuronaldegeneration, particularly in the neostriatum, accompaniedby motor dyskinesias and progressive mental deterioration.In rats, injections of excitotoxic amino acids, such as kainicacid or ibotenic acid (IA), into the caudate-putamen (CPu)are capable of producing neuropathological and neurochemi-cal changes in the basal ganglia that resemble those seen inHuntington disease (6-8). The behavioral changes seen inrats with excitotoxic neostriatal lesions are, by necessity,only of a correlative nature compared to the human disease.Thus, the rats with lesions do not, for example, exhibit anyovert involuntary movements, typically seen in Huntingto-nian patients. However, when the excitotoxic lesion involvesthe anteromedial neostriatum, which is the region mostclosely corresponding to the head of the caudate nucleus in

man, the rats do develop a hyperactivity syndrome combinedwith substantial impairments in learning and memory (9-13),which can be viewed as analogous to the cardinal behavioralfeatures of Huntington disease in humans.The present study explores the potential of the neuronal

replacement paradigm in this rat model. Previous studies(14-16) have shown that grafts of fetal striatal tissue can bemade to survive well in the depth of the lesioned striatum inadult host recipient rats and that they can normalize the rat'slocomotor hyperactivity and the increased local cerebralmetabolic activity seen in several of the striatal outputstructures. Here we report that intrastriatal striatal grafts arecapable also of restoring more complex cognitive behavior inrats with lesions in the striatum.

METHODS

A total of 33 adult-male Sprague-Dawley rats were used.Twenty-five of these received bilateral IA injections into theanteromedial CPu under Equithesin anesthesia (2 x 5 ,ug in0.5 ,ul of 0.1 M sodium phosphate buffer, pH 7.5). Coordi-nates (from bregma) were as follows: A = +0.7 and +1.7, L= ± 2.0, V = 4.5; with the incisor bat at the interaural line.One week later nine of these received bilateral injections ofa fetal striatal cell suspension into the IA-lesionedanteromedial CPu (A = +1.2, L = + 2.3, V = 4.5; incisor barat 0); they constituted the CPu-graft group. Another group of10 rats with IA lesions received identical injections from thesame striatal cell suspensions bilaterally into the principalneostriatal projection area globus pallidus (GP) (A = -0.8, L= ± 3.0, V = 6.0; incisor bar at -2.3). These rats constitutedthe GP-graft group. Each rat received two 3-,u aliquots of acell suspension from the striatal primordia of 14- to 15-day-old rat embryos (crown-rump length, 11-14 mm), as de-scribed (15, 16, 19). Six rats with lesions but without graftsbelonged to the lesion-alone group, and the remaining eightrats served as unoperated controls.Three months after lesion and transplantation surgery the

rats were tested for their impairments in the acquisition of aspatial delayed alternation task and spatial position habits ina conventional elevated T-maze (3, 9, 10), as described in Fig.1 a-c and for their locomotor hyperactivity in automatedphotocell cages (10) (Fig. 1 d and e). Eight of the CPu-graftedrats and nine of the GP-grafted rats survived throughouttesting. The brains of all of these surviving animals werefinally processed for quantitative morphometric analyses ofthe IA lesions and the striatal grafts, using cresyl violet andacetylcholine esterase (21) stained serial cryostat sections.

Abbreviations: IA, ibotenic acid; CPu, caudate-putamen; GP, glo-bus pallidus.tTo whom reprint requests should be addressed.

2728

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. NatL. Acad. Sci. USA 83 (1986) 2729

10 7

0

co

C.

-

0 2-

b

'A -I'.te 0

V~/

c

1 2/3 4/5 1 2/3 4/5 1/2 5/6

Test days

x

0

18-

16-

14-

12-

10-

8-

6-

4-

2

d e

7oteCon+e I~Ggraf grfesteem~~~~~~~..Cont Les CPu GP

graft graft

Cont Les

13/14 17/18

CPu GP

graft graft

FIG. 1. (a-c) T-maze alternation test.

of rats were constituted on the basis of

groups for the recorded side-bias and running

in a T-maze (means over 5 test days) (a, first

test). Pnrorto surgery the rats, maintained

schedule, were habituated to receive food

elevated T-maze (3) during 3 days (three 15-mmnneous alternation testing then started and

goal arms baited (3 pellets per trial and

weeks post-IA lesion the second period

and the rats were rehabituated to the T-maze 15-minbefore receiving 5 days of spontaneous alternation

spontaneous alternation test). The rats

delayed alternation paradigm (9, 10) during

30±5 sec between trials. The average number

for each group is plotted over time (means ±

(c, delayed alternation test). When a rat

correct choices in 30 trials over 3 consecutive

further. Control (e); lesion (0); CPu graft graft (d-e)

Locomotor test. Mean total overnight locomotor (±

recorded in automated photocell cages

different feeding conditions: prior to food 11-weeks

postlesion (d, satiated test) and under food 14-weeks

postlesion (e, food deprived test). Cont,

only group. *, P < 0.05 compared to normal

compared to group with IA lesions

RESULTS AND COMMENTS

Behavioral Impairments of the Rats with

spontaneous alternation bias in the T-maze

delayed alternation acquisition, all rats

maze prior to surgery and observed for

ous alternation (with both arms baited)

(first test shown, Fig. la). The rats were

four experimental groups that were balanced

spontaneous alternation performance.

the surgery, the rats received five further

ous alternation testing before training

nation task (second test shown, Fig.

tendency to alternate had decreased

test, the rats with IA lesions did not differ from thenonlesioned controls on this measure.While all control rats rapidly acquired the delayed alter-

nation task, reaching criterion (27 correct choices in 30 trialsover 3 days) in 7-12 days, none of the rats with IA lesionsreached criterion within 18 days of testing. Indeed, the meanperformance of the lesioned group by the end of the test(between four and five correct alternations) was close to therandom performance in the task (i.e., five correct alterna-tions).

In agreement with previous studies (10, 14, 15, 20) the ratswith IA lesions were significantly hyperactive in the over-night locomotor test (Fig. ld), and this effect was furtheraccentuated when the rats were tested under food deprivation(Fig. le).

Graft-Induced Behavioral Recovery. Both graft groupswere significantly improved in the acquisition of the delayedalternation task, as compared to the rats with IA lesionsalone, when tested 3 months after grafting (Fig. ic). Thegroupsdiffered significantly in the number of correct choices(F = 32.7, P < 0.001) and Newman-Keuls tests indicated thateach groupdiffered from each other (control group >CPu-graft group > GP-graft group > lesion group; all at leastP < 0.05). Three of the eight rats in the CPu-graft groupreached criterion (at days 13, 16, and 17), whereas the otherCPu-grafted rats, although all close to this level of perform-ance, did not reach the criterion level. By contrast, none ofthe rats in the GP-graft group reached criterion, and onlythree of the nine GP-grafted rats reached above 21 of 30correct alternations during the last 3 days of testing.

In the overnight spontaneous locomotor test the lesion-induced hyperactivity was ameliorated in the CPu-graftgroup, both under the satiated (Fig. ld) and food-deprived(Fig. le) conditions, which is consistent with previous studies(14, 15). The rats in the GP-graft group did not differ fromnonlesioned controls in the satiated test but remained onaverage hyperactive in the food-deprived test. In the satiatedtest these group differences were significant (ANOVA F =

3.89, P < 0.05; post hoc Newman-Keuls test: P < 0.01 forlesion versus CPu graft, and P < 0.05 for lesion versus eitherGP graft or control). In the food-deprived test the variabilitywas such that the overall group differences just failed to reachsignificance (F = 2.22, P < 0.1). The individual scatter in thelocomotor activity scores in this test was particularly con-

spicuous in the GP-graft group: four of the nine rats hadactivity scores below the range of the normal control rats, andin the other five the activity scores were above or much abovethis range, a variability that may be due to particular andvariable accidental damage caused by the grafts in this group(see below). If the GP-graft group is excluded from the groupanalysis the differences were significant also in the food-deprived test (F = 3.85, P < 0.05), and the lesion group was

significantly different from both the CPu-graft group and thenonlesioned controls (P < 0.05; Newman-Keuls test) whilethe activity of the CNu-grafted rats was not different from thecontrols.

Lesion and Graft Morphology. The IA lesion involved theentire rostral and medial CNu, where only single scatteredneurons remained, while the ventral and ventrolateral por-

tions of the head of the CPu as well as the tail portion werelargely spared, and the nucleus accumbens was only mini-mally affected (Fig. 2). As a result, the head of the CPu(rostral to the GP, excluding the graft volume) had shrunkenby about 45%, and the volumetric analysis (performed onblind-coded slides) showed that the volume of the residualrostral CPu was similar in all three lesioned groups (Table 1).In most animals the IA lesion extended dorsally into theoverlying neocortex (Fig. 2). The extent of this extrastriataldamage seemed to be similar, however, in all three groups. Inaddition, there was considerable shrinkage of the anterior

Neurobiology: Isacson et al.

ic

2730 Neurobiology: Isacson et al.

A L44 B CPu 19 C GP 31

FIG. 2. Extent of IA lesion and neuronal sparing, L 44, (A) and location of a CPu graft (B) and a GP graft (C) illustrated in six to eight coronallevels through the caudate-putamen (cp) and globus pallidus (gp), from the level of the genu of the corpus callosum (cc) (Top) to the body ofthe GP (Bottom). All figures were drawn from serial sections using a camera lucida technique. The anterior needle tract is indicated by arrowsin the cortex (CX) in A and the lateral ventricles (lv) are black. t, Transplant; na, nucleus accumbens; ca, anterior commissure; s, septum; ic,internal capsule; and 3v, the third ventricle. The gliotic areas of the CPu (with low neuronal density) are indicated by stippling, the transplantsby hatching, while areas of relative neuronal sparing are white.

part of the lateral septum (see Fig. 2) in virtually all animals,probably because of damage to the lateral septal nucleus dueto leakage of the toxin. Again this nonspecific damage had asimilar appearance in all three groups. No so-called remoteneuronal damage, commonly seen after kainic acid lesions-e.g., in hippocampus and mediodorsal thalamus-was de-tectable in the rats with IA lesions with the current stainingmethods.

Surviving striatal grafts were found on both sides in allsurviving 17 animals of the CPu- and GP-graft groups. Boththe size and the location of the grafts were different in the twogroups: The grafts implanted into the lesioned anteromedialCPu were on the average about twice as large as the graftsimplanted into the GP region-i.e., caudal to the lesion(Table 1)-although neuronal density appeared similar in thetwo types of grafts. The CPu grafts were confined to the CPu,and they all extended into the anteromedial sector thatnormally receives a direct projection from the prefrontal

cortex (Fig. 2B). In the GP-graft group (see Fig. 2C) on theaverage about 50% of the graft volume was located in or nearthe GP (i.e., outside the CPu) (Table 1). The grafts extendedto a variable degree in a rostrodorsal direction, along thetrack ofthe injection cannula, into the caudal and dorsal partsof the head of the CPu (Fig. 2C), and only in one case (thelargest grafts in the group) did the grafts extend well into theanteromedial sector. In several ofthe animals in the GP groupthe grafts had caused considerable damage to the GP and/orthe medial portion of the internal capsule, and may also haveinvolved part of the nucleus basalis or its axonal projections.Since GP and the medial portion of the internal capsule carrythe two principal output pathways from the neostriatum-thestriato-pallido-thalamic pathway and the striato-nigral path-way, respectively-we paid particular attention to this acci-dental damage in the analysis of the behavioral performanceof the individual grafted animals.

Correlations with Behavioral Measures. There was no

Proc. Natl. Acad Sci. USA 83 (1986)

Proc. Natl. Acad. Sci. USA 83 (1986) 2731

Table 1. Volume of the lesioned rostral CPu and of thestriatal transplants

CPu-graft GP-graftIA lesion only, group, group,

Volume mm3 mm3 mm3Rostral CPu 27.0 ± 1.2 26.0 ± 1.2 25.9 ± 1.2Transplant

Total 7.1 ± 1.9* 3.8 ± 2.9In CPu 7.0 ± 1.8* 2.2 ± 2.9Outside - 0.1 ± 0.2* 1.6 ± 1.1

Data are means ± SEM of both sides combined.*Significantly different from GP-graft group (P < 0.05; Student's ttest). Remaining rostral CPu volume did not differ between thegroups using one-way analysis of variance (F = 0.21). The volumeof the rostral CPu in normal nonlesioned rats is about 50 mm3 onboth sides combined.

significant correlation between the total graft size and theperformance of the grafted animals in the delayed alternationtest (r = 0.48; P > 0.05), although a tendency existed. Thecorrelation with graft size was, however, significant whenonly the volume ofthe grafts located within the boundaries ofthe CPu was taken into account (r = 0.62; P < 0.01). In Fig.3 the performance in the delayed alternation test for thegrafted animals (mean correct performance over the last 6days of testing) is plotted against the graft volume within theCPu (both sides combined). The performance of the groupwith lesions (without grafts) is given by the triangle (mean ±SEM). Although the two graft placement sites cluster at thetwo ends of a regression line, the diagram suggests that theremay be a critical minimum graft size (in the order of4-6 mm3total, or 2-3 mm3 on each side) for significant behavioralrecovery to occur. Graft tissue located outside the CPu, onthe other hand, appeared to be without effect on the delayedalternation performance. It may be relevant to note that inthree of the rats in the GP-graft group (circled in Fig. 3) thegrafts had produced substantial bilateral damage to themedial part of the internal capsule. One of these was the onlygraft of large size that did not produce any substantialbehavioral improvement in the delayed alternation test. Withthese three rats excluded from the regression analysis thecorrelation between graft size in the CPu and delayedalternation performance is even stronger (r = 0.76; P < 0.01).The overnight activity scores (under food deprivation) was

uncorrelated with graft size (r = -0.16 for total graft volumeand r = -0.39 for graft volume within CPu). However, asnoted above, the variability in the activity scores wasconspicuous in the GP-graft group: four rats were hypoactiveas compared to normal controls (2046 ± 386) (see Fig. le),and five rats were markedly hyperactive (13,286 + 2832).There was a clear-cut difference between these two groupswith respect to the nonspecific damage of the GP: In all fivehyperactive rats the GP was substantially destroyed (>40%)on both sides, and in addition there was damage to the medialinternal capsule on one or both sides. In the four hypoactiveGP rats the GP and the internal capsule were at the most onlymarginally affected (<20% destruction on any one side).

This may be taken to indicate that the GP-graft placementcan be as effective in compensating for the IA-lesion-inducedlocomotor hyperactivity as the CPu-placed grafts providedthat the striatal output pathways are intact and thus that thegraft effect on locomotor hyperactivity might be mediated viathe GP. This is further supported by the observation thatamong the 12 grafted animals without substantial GP damagethere was a high correlation between the overnight activityscores and the proximity of the graft to the host GP (r = 0.74;P < 0.01). By contrast, this parameter did not correlate withthe rats performance in the delayed alternation test (r = 0.42;P > 0.1). In all grafted animals with activity scores within or

9-

8-0

0.

0

7S.

Q

6

r- 5-

lo~c)

3

Om

00

L~~~~~~~~~0~~~~0

00

_0~S

0 1 2 3 4 5 6 7 8 9 10Transplant volume in CPu, mm3

FIG. 3. Individual graft size inside the CPu (both sides combined)is plotted against the number of correct alternations in the delayedalternation task (mean of last 6 days oftesting as described in Fig. 1c)for the 17 grafted rats that survived throughout testing. The CPu-graftgroup is shown by open circles and the GP-graft group by filledcircles. The performance of the group with IA lesions (without graft)is indicated by the open triangle (mean ± SEM). For the three circledpoints, the grafts of the GP group had caused substantial damage tostriatal output structures. The regression analysis showed a signifi-cant correlation between the parameters (r = 0.62; P < 0.01;regression line, shown in figure: y = 5.99 + 0.23x). If the three ratswith nonspecific damage (mentioned above and circled in the figure)were excluded from the analysis, the correlation was even higher (r= 0.76; P < 0.01; y = 6.18 + 0.24x). The morphometry was done asfollows: From each section, the structures were outlined and drawnat x 16 magnification on paper. The structures were cut out, weighed,and calibrated against an area standard. The volumes were calculatedfrom the area figures and the distance between each section.

below the control range the grafts were located within lessthan half a millimeter from the GP.

DISCUSSIONThe results demonstrate that intrastriatal grafts of fetalstriatal tissue are able to substantially ameliorate the mazelearning deficits in animals with striatal IA lesions. Graftsplaced in the anteromedial CPu were significantly moreeffective than those placed in, or in the vicinity of, the GP,although also the GP-grafted rats were significantly improvedrelative to the nongrafted rats with IA lesions. There areseveral possible explanations for this difference in efficacy.

(i) The difference may simply be a graft size effect since thegrafts in the GP group had only about half the size of thosein the CPU group (Table 1). Since both graft groups receivedthe same amount of cells from the same cell suspensions, thisindicates that striatal tissue grafts implanted into the lesionedand gliotic area of the striatum survived and/or grew signif-icantly better than those implanted into the largely intactneostriatal or pallidal tissue.

(ih) The graft effect on the learning deficits may depend onthe proximity to the area of termination of the striatalafferents coming from the anteromedial prefrontal cortex.Performance in the delayed alternation task is characteristi-cally disrupted by lesions of the anteromedial prefrontalcortex, and selective lesions of its projection area in theneostriatum-i.e., the anteromedial part of the head of theCPu-have similar effects (9, 17, 18). It seems likely, there-

Neurobiology: Isacson et al.

Proc. Natl. Acad. Sci. USA 83 (1986)

fore, that the function of the anteromedial neostriatum in thelearning of delayed response tasks normally depends on itsdirect connection with the prefrontal association cortex. Ifso, the graft effect on the performance in the delayedalternation task may critically depend on the ability of thegraft to interact both with the, prefrontal cortical afferents aswell as the normal targets of this neostriatal region. Thisseems consistent with the fact that all CPu grafts were at leastin part located within the prefrontal termination zone, where-as only one of the GP grafts reached so far rostrally.

(iii) The smaller functional effects in the GP-graft groupmay reflect the high incidence ofdamage to the striatal outputstructures (GP, entopeduncular nucleus, and internal cap-sule) caused by the implantation surgery. The post hocmicroscopic analysis suggested that this may indeed havebeen the case for the expression of the graft effect on thelesion-induced-locomotor hyperactivity, whereas the limitedeffects of the GP grafts in the maze learning task could not beexplained on the basis of this factor alone.The present findings that neural implants can significantly

ameliorate the acquisition of complex conditioned behaviorafter destruction of a major telencephalic cortical relaystructure raises a number of interesting questions as to themechanism(s) of action of the striatal grafts. In previousstudies of dopaminergic and cholinergic neurons grafted intothe nigro-striatal, septo-hippocampal, or basal forebrain-cortical systems, functional recovery in sensorimotoric ormaze learning tasks has been obtained by ectopic placementof the graft-i.e., within or in direct contact with thedenervated striatal or cortical targets (1-5). In fact, experi-ments with placement of dopaminergic neurons into thesubstantia nigra region have so far failed to detect anygraft-induced functional effects. This may indicate that,although extensive efferent connections with their normaltargets in the host are necessary for the functional effects ofgrafted aminergic neurons, they may exert at least some oftheir behavioral effects even in the absence of major normalafferent inputs. In the present experiment we tested bothhomotopic placement-i.e., into the site of the lesion-andectopic placement-i.e., into one of the primary projectionareas of the lesioned striatal neurons in the GP-for thestriatal grafts. Interestingly, the effects on the recovery ofmaze learning was significantly better with homotopic place-ments in this model. As discussed above, this would beconsistent with the idea that the functional capacity of thestriatal grafts in conditioned behavior is critically dependenton some of the normal striatal inputs, above all from theprefrontal cortex. On the other hand, since proximity to thedenervated globus pallidus seemed to be a positive factor inthe amelioration of the locomotor deficit, graft-inducedreinnervation of the denervated target may also play animportant role.Although more detailed anatomical studies are needed to

clarify these points, there is some evidence from a parallelanatomical study (22) that fetal striatal neurons implantedinto the striatum with IA lesions are capable of establishing

at least some afferent and efferent connections with the hostbrain. Furthermore, neurochemical studies (16) have dem-onstrated a graft-induced recovery of glutamic acid decar-boxylase in the GP. The extent to which behavioral recoverycorrelates with establishment of specific graft-host connec-tions is an interesting topic for further investigation.

1. Dunnett, S. B., Bjorklund, A., Gage, F. H. & Stenevi, U.(1985) in Neural Grafting in the Mammalian CNS, eds.Bjorklund, A. & Stenevi, U. (Elsevier, Amsterdam), pp.451-469.

2. Freed, W. J., Olson, L., Ko, G. N., Morihisa, J. M., Niehoff,D., Stromberg, I., Kuhar, M., Hoffer, B. J. & Wyatt, R. J.(1985) in Neural Grafting in the Mammalian CNS, eds.Bjorklund, A. & Stenevi, U. (Elsevier, Amsterdam), pp.471-489.

3. Dunnett, S. B., Low, W. C., Iversen, S. D., Stenevi, U. &Bjorklund, A. (1982) Brain Res. 251, 335-348.

4. Low, W. C., Lewis, P. R., Bunch, S. T., Dunnett, S. B.,Thomas, S. R., Iversen, S. D., Stenevi, U. & Bjorklund, A.(1982) Nature (London) 300, 260-262.

5. Dunnett, S. B., Toniolo, G., Fine, A., Ryan, C. N.,Bjorklund, A. & Iversen, S. D. (1985) Neuroscience 16,787-797.

6. Coyle, J. T., Schwarcz, R., Bennett, J. P. & Campocharo, P.(1977) Prog. Neuro-Psychopharmacol. 1, 13-30.

7. Marsden, C. D. (1979) in Advances in Neurology, eds. Chase,T. N., Wexler, N. S. & Barbeau, A. (Raven, New York), Vol.23, pp. 567-576.

8. McGeer, E. G., McGeer, P. L., Hattori, T. & Vincent, S. R.(1979) in Advances in Neurology, eds. Chase, T. N., Wexler,N. S. & Barbeau, A. (Raven, New York), Vol. 23, pp. 577-591.

9. Divac, I., Markowitsch, H. J. & Pritzel, M. (1978) Brain Res.151, 523-532.

10. Dunnett, S. B. & Iversen, S, D. (1981) Behav. Brain Res. 2,189-209.

11. Dunnett, S. B. & Iversen, S. D. (1980) Behav. Brain Res. 1,497-506.

12. Pisa, M., Sanberg, P. R. & Fibiger, H. C. (1981) Exp. Neurol.74, 633-653.

13. Pisa, M., Sanberg, P. R. & Fibiger, H. C. (1980) Physiol.Behav. 24, 11-19.

14. Deckel, A. W., Robinson, R. G., Coyle, J. T. & Sanberg,P. R. (1983) Eur. J. Pharmacol. 93, 287-288.

15. Isacson, O., Brundin, P., Kelly, P. A. T., Gage, F. H. &Bjorklund, A. (1984) Nature (London) 311, 458-460.

16. Isacson, O., Brundin, P., Gage, F. H. & Bjorklund, A. (1985)Neuroscience 16, 799-817.

17. Rosvold, H. E. & Szwarcbart, M. K. (1964) in The FrontalGranular Cortex and Behavior, eds. Warren, J. M. & Akert,K. (McGraw-Hill, New York), pp. 1-15.

18. Oberg, R. G. E. & Divac, I. (1979) in The Neostriatum, eds.Divar, I. & Oberg, R. G. E. (Pergamon, Oxford), pp. 291-313.

19. Bjorklund, A., Stenevi, U., Schmidt, R. H., Dunnett, S. B. &Gage, F. H. (1983) Acta Physiol. Scand. Suppl. 522, 1-7,

20. Mason, S. T. & Fibiger, H. C. (1979) Neuropharmacology 18,403-407.

21. Koelle, G. B. (1954) J. Comp. Neurol. 100, 211-235.22. Pritzel, M., Isacson, O., Brundin, P., Wiklund, L. &

Bjorklund, A. (1985) Neurosci. Lett., Suppl. 22, S531.

2732 Neurobiology: Isacson et al.