Proc. Natl. Acad. Sci. USAVol. 93, pp. 6355-6360, June 1996Neurobiology

Long-term functional recovery from age-induced spatial memoryimpairments by nerve growth factor gene transfer to the ratbasal forebrain

(experimental gene therapy/immortalized neural stem cells/learning and memory/aging/neurotrophins)

ALBERTO MARTINEZ-SERRANO*t, WALTER FISCHER*, STINE SODERSTROMt, TED EBENDALt,AND ANDERS BJ6RKLUND**Department of Physiology and Neuroscience, Lund University, Solvegatan 17, S-223 62-Lund, Sweden; and tDepartment of Developmental Neuroscience,Biomedical Center, Box 587, Uppsala University, S-751 23-Uppsala, Sweden

Communicated by Tomas Hokfelt, Karolinska Institutet, Stockholm, Sweden, February 6, 1996 (received for review September 22, 1995)

ABSTRACT Nerve growth factor (NGF) stimulates func-tional recovery from cognitive impairments associated withaging, either when administered as a purified protein or bymeans of gene transfer to the basal forebrain. Because genetransfer procedures need to be tested in long-term experi-mental paradigms to assess their in vivo efficiency, we haveused ex vivo experimental gene therapy to provide localdelivery of NGF to the aged rat brain over a period of 2.5months by transplanting immortalized central nervous sys-tem-derived neural stem cells genetically engineered to secreteNGF. By grafting them at two independent locations in thebasal forebrain, medial septum and nucleus basalis magno-cellularis, we show that functional recovery as assessed in theMorris water maze can be achieved by neurotrophic stimu-lation of any of these cholinergic cell groups. Moreover, thecholinergic neurons in the grafted regions showed a hyper-trophic response resulting in a reversal of the age-associatedatrophy seen in the learning-impaired aged control rats.Long-term expression of the transgene lead to an increasedNGF tissue content (as determined by NGF-ELISA) in thetransplanted regions up to at least 10 weeks after grafting. Weconclude that the gene transfer procedure used here is effi-cient to provide the brain with a long-lasting local supply ofexogenous NGF, induces long-term functional recovery ofcognitive functions, and that independent trophic stimulationof the medial septum or nucleus basalis magnocellularis hassimilar consequences at the behavioral level.

Aging in rodents is associated with a progressive decline inlearning abilities, memory storage and use of spatial informa-tion (1-4), and associated with a hypofunction of the cholin-ergic forebrain system, with the neurons in the medial septum(MS) and nucleus basalis magnocellularis (NBM) region beingmost affected and displaying a marked degenerative atrophy(2, 4-11). Memory impairments correlate with cholinergicatrophy in forebrain nuclei (7, 10, 12, 13), and both can bereversed by the exogenous administration of nerve growthfactor (NGF) or neurotrophin-3 (NT-3) (6, 9, 14-22). Con-versely, administration of anti-NGF antibodies induces mem-ory impairments and cholinergic neuron atrophy in the adultbrain (23, 24). The similarities between the age-associatedchanges in the cholinergic system seen in rodents and theneurodegeneration that occurs in normal aging and in Alz-heimer disease in humans (25-30) have raised the possibilitythat exogenous supply of NGF to the aged brain could beeffective as a clinical therapy for the amelioration of cognitiveimpairments and atrophic neuronal changes. However, theidentification of optimal routes for long-term administration

of NGF or other therapeutic proteins to the central nervoussystem remains a major challenge (31-33). Continuous intra-cerebral infusion of NGF delivered from minipumps is oflimited duration and requires a permanent cannula implanta-tion, a procedure that causes chronic trauma and inflamma-tion. NGF can be made available to the brain after conjugationto other proteins that can be recognized at the blood brainbarrier and transported to the brain side (34, 35); this ap-proach, though proven useful, results in a nonlocal NGFsupply, generally in a low dose range and of very limitedtime-duration (35).

Cell-based gene transfer represents an interesting alterna-tive to these procedures for long-term intracerebral delivery oftherapeutically active proteins in that it may circumvent thedrawbacks associated with chronic intracerebral infusions orinjections. Fibroblastic cell lines (encapsulated in some cases),primary fibroblasts, and central nervous system-derived neuralprogenitors have been used to deliver NGF to the rodent orprimate brain, and have been shown to exert neurotrophiceffects on lesioned cholinergic neurons of the basal forebrain(36-41) and, in rodents, recovery from age-related cognitiveimpairments (14, 19). For any gene transfer procedure, how-ever, it is important to demonstrate that the functional effectsare long-lasting and that it allows for stable in vivo transgeneexpression that leads to increased levels of the transgenicprotein in the brain. The present study was undertaken toexamine these two essential aspects of NGF gene transfer tothe aged rat brain after transplantation of a clonal neural stemcell line genetically engineered to synthesize and secrete NGF(14, 36). We report here that this ex vivo gene transferapproach can be used to ameliorate behavioral deficits in ratsfor extended time periods, up to 2.5 months, that it is useful inanimals of advanced age, and that it provides the target regionwith an increased amount of NGF protein (as determined byELISA). Furthermore, we show that local, independent stim-ulation of cholinergic neurons in the MS and NBM is sufficientto produce a sustained improvement in the ability of theanimals to perform the task.

MATERIALS AND METHODSGene Transfer. Ex vivo gene transfer of NGF was performed

by transplanting control (clone D11) or NGF-secreting (cloneE8) conditionally immortalized neural progenitors as de-scribed (36). A brain-derived neurotrophic factor (BDNF)-producing clone (C7) engineered in the same way to secretehBDNF (h, human) (42) was used in one group of animals to

Abbreviations: MS, medial septum; NBM, nucleus basalis magnocel-lularis; PS, postsurgery; NGF, nerve growth factor; RT-PCR, reversetranscription-PCR; BDNF, brain-derived neurotrophic factor.tTo whom reprint requests should be addressed. e-mail: abto@biogen.wblab.lu.se.

6355

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

Dow

nloa

ded

by g

uest

on

July

11,

202

0

6356 Neurobiology: Martinez-Serrano et al.

control for specificity of the neurotrophic effects. A detailedcharacterization of these neural progenitor cell lines in vitroand in vivo after transplantation can be found elsewhere (14,36,42,43). Cells were cultured in DMEM containing 10% fetalbovine serum, 2 mM glutamine, and 10,000 units per ml of bothpenicillin and streptomycin. For transplantation, cell suspen-sions in Hanks' balanced salt solution (GIBCO) at 100,000cells per IlI were prepared by trypsinization and used within 3hr. The cells were labeled for 72 hr before grafting with[3H]thymidine [10 ACi/ml (1 Ci = 37 GBq), Amersham].Animal Groups and Surgical Procedures. The animals used

in this study were 18 young (3 months old) and 54 aged femaleSprague-Dawley rats that were housed and treated followinginstitutional guidelines. Before grafting, 10 young rats and 90aged animals (24 months old) were tested in the Morris watermaze (15). Ten aged rats were classified as nonimpaired in thetask, and 35 aged impaired animals were further subdividedinto the following groups: Control (receiving a graft of Dllcells into the septum, n = 6, or NBM, n = 5), NGF cells in MS(grafted with E8 cells bilaterally into the MS, n = 9), NGF cellsin NBM (grafted with E8 cells bilaterally into the NBM region,n = 7), and BDNF cells in the NBM (grafted with C7 cellsbilaterally into the NBM region, n = 8). Coordinates for thedifferent transplants were as follows: NBM: (TB +5.0) AP =+0.2, ML = ±3.4, V(d) = -7.0 and AP = + 1.0, ML = +2.6,V(d) = -7.3; MS: (TB -2.0) AP = +0.5, ML = ±0.6, V(d)= -7.0 and -6.7. Four deposits (two on each side) were placedin either the MS or in the NBM region; a total of 400,000 cells(100,000 cells in 1 Al per deposit) were injected in each rat.Because the animals grafted with control cells in the MS orNBM regions did not differ during behavioral testing, they aretreated as a single combined control-graft group. The agedanimals that were used to demonstrate long-term expression ofthe transgene [13 rats for NGF ELISA measurements and 4rats for reverse transcription-PCR (RT-PCR) amplification ofthe retroviral transcript] were either rats that did not fit thecriteria for inclusion into the nonimpaired or impaired groups,or animals with a swimming speed lower than 0.11 m/s (theonly parameter used to exclude animals from the test). Eightyoung rats bilaterally grafted with control- or NGF-cells in theNBM in different hemispheres were used for NGF-ELISAdetermination from tissue samples. All animals were sacrificed10 weeks after grafting once the behavioral experiment wascompleted; out of 54 transplanted aged rats, only four rats diedduring the experiment (from 25 to 27-28 months old).Long-Term Behavioral Test. Animals were tested for spatial

learning and memory in the Morris water maze test (44) andwere given daily blocks of eight trials with a cut-off time of 60 s.The tests were arranged as follows (see Fig. 1): pretest, usedto allocate the animals to the different experimental groupsduring a 5-day testing schedule; test 1, starting 8 days post-surgery (PS) consisting of 5 days of testing; test 2, performed7 weeks after grafting with a duration of 3 days; test 3, startingthe day after the finish of test 2 (day 53 PS) with a duration of16 days (3 days of testing, 10 days of rest, and another 3 daysof testing); visible platform test was performed on the day aftertest 3 and lasted 1 day (day 71 PS).The submerged platform was located in the southwestern

quadrant of the pool during the pretest, test 1, and test 2, and itwas changed to the northeastern quadrant for test 3 to study theability of the rats to relearn the task. Spatial probe trials (removedplatform) were given at the end of day 5 in the pretest and test1, and at the end of days 3 and 16 during test 3.Analyses of Transgene Expression in Vivo. Ten weeks after

grafting (day 72 PS), the animals were killed under halothaneanaesthesia, and their brains were quickly removed and frozenunder crushed dry ice; the selected regions were dissected atsubzero temperatures. RT-PCR amplification of the NGF-coding retroviral transcript was performed on total RNAextracted from these pieces of grafted tissue (14, 36). NGF

ELISA assay was applied in parallel tissue samples followingdescribed procedures (45) using a monoclonal anti mouse -3NGFantibody 27/21 as a capture antibody and a sandwich 27/21-f3-galactosidase-conjugate antibody (Boehringer Mannheim).

Histology. Animals that participated in the behavior exper-iment (together with the young control subjects) were perfusedat the end of the experiment with buffered 4% paraformal-dehyde and their brains were sectioned at 40 ,um thickness;[3H]thymidine autoradiography, immunohistochemical stain-ing for the low-affinity Neurotrophin receptor (p75NTR), andstereological assessment of cell numbers and volumes in theMS and NBM regions were performed following standardprocedures (14, 36, 46).

RESULTSLong-Term Recovery from Age-Associated Learning and

Memory Impairments. Before grafting, the aged animals weretested in parallel with a group ofyoung (3 months old) rats andclassified as nonimpaired or learning impaired on the basis ofpreviously defined criteria (ref. 15; escape latency for thenonimpaired rats was within the mean ± two standard devi-ations of the young group during the last 2 days of the test)(Fig. 1). Starting on day 8 PS, the animals were tested for 5 daysin the water maze for their spatial navigation performance;during this week (test 1), overall, the NGF-grafted rats per-formed better than the control-grafted animals (Fig. 1). BothNGF-grafted groups were not different from the nonimpairedrats, whereas the control-grafted animals remained impaired.The animals transplanted in the MS were significantly betterthan the controls at the beginning of the test, and the NBM-grafted rats showed a nonsignificant trend to reduced escapelatency in this initial test. Eight weeks after grafting, during test2 (starting on day 50 PS) and regardless of the location of thetransplant (MS or NBM), the NGF-grafted animals showed afurther reduction in the time to find the platform and wereindistinguishable from the nonimpaired animals, and signifi-cantly different from the control-graft group.

In the subsequent test 3, the platform was moved to a newlocation to study the rats ability to relearn the task; after thischange, all groups showed an increase in the escape latencyparameter (repeated measures ANOVA, F1,30 = 37.1, P =0.0001, effect of day, day 3 of test 2 versus day 1 of test 3),indicating that the rats were using a spatial strategy to find theplatform at the end of test 2. In test 3, all groups reduced theirescape latency scores to the level seen at the end of test 2 (P =0.0003, F1,30 = 17, for repeated measures), showing that theycould learn the reversal task using new spatial extra maze cues.The animals were also tested for their ability to remember

the location of the hidden platform in the spatial probe trials(removed-platform), performed at the end of the pretest, test1 and twice during test 3 (Table 1): Both groups of NGF-transplanted rats had a score as low as the control-graftedanimals in the pretest, but improved their spatial acuity(number of crossings over the former platform site, indicatingfocused search) significantly during test 3 in two different trials(days 3 and 16 of test 3), demonstrating that the NGF-graftedgroups had improved their ability to remember the platformlocation. In the first day of test 3, when the platform locationwas changed, the nonimpaired and both NGF-grafted groupsshowed during the first swim a marked focused search, result-ing in a higher number of crossings over the former platformlocation (Table 1). In the spatial probe trials during test 3(analyzed as the average of the trials at days 3 and 16), thenonimpaired rats and both NGF-groups, but not the impairedcontrol animals, displayed significant focus over the newplatform site, indicating that the improved spatial learning wasmaintained for at least 10 weeks in the rats receiving NGF-cells.The same results were obtained when analyzing swimdistances, and, consistent with this, the swimming speeds did

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

July

11,

202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 6357

(housing)

AGE---AGE

months

GRAFT TEST 2

PRE-TEST TEST 1 | TEST 3 sacrificeIE ~ Im--odayo8 5053 6672

23 24 25 28 27 2n

C.)

1 2 3 4 5 1 2 3 4 5 123

Days of Testing

FIG. 1. Spatial memory testing. (Upper) Time cciment and subjects age during the testing; the indicto the start of each test. (Lower) Nonimpaired rats ([with control-cells (0), and two groups of animals gcells (e) in the MS (upper plots) or NBM region Itested for spatial navigation in the water maze. Theescape latency scores to find the platform during ti(pretest) and two postgrafting tests, performed durin1) or eighth week (test 2) after surgery. Durinhnonimpaired group was significantly different (Pother groups during days 2, 4, and 5. Young animaldid not differ from the nonimpaired group at arpretest (the performance reached by the young gro10 ± 1.3 s, is shown for comparison as a dotted linthe control-group was still different from the noninthe NGF-groups improved their performance to thwere no longer different from the nonimpaired groiindicated by (t), the NGF groups were significantlycontrol-graft group and not from the nonimpaired (ANOVA, post hoc Fisher pairwise least significant

not differ among groups during the experimeithe visible platform test, there were no diffeilatency among groups (F3,30 = 1.01, P =

ANOVA). Thus, the improved task performErats receiving grafts of NGF-secreting cellascribed to differences in motor performanceAn additional group of animals was graft

region with a similar cell line engineered to(42); this BDNF-graft group did not show anyany parameter during the testing, being in allthe control group (data not shown).MS Versus NBM Location of the NGF-

Grafts. Throughout testing, no clear-cut dfound between the animals transplanted in theregion in either escape latency or the number (the former platform site. During test 3, the nunin the spatial probe trial showed that both NGIwere clearly improved compared with the contand they were not different from the nonin(Table 1). However, the NGF-grafted grouldifferent from each other in the escape lat

during the relearning (test 3, after changing thiplatform). Thus, the NBM-graft group perfori

control (0)

NGF in MS (-)non-impaired (a)

Table 1. Platform crossings in spatial probe trials

Pre test Test 1 Test 3a Test 3bNonimpairedc 3.8 0.3* 5.1 ± 0.4* 3.6 ± 0.7* 4.2 ± 0.5*Control MS and NBM 2.2 ± 0.4t 3.1 ± 0.7t 1.1 ± 0.4t 1.5 ± 0.3tNGF-MS 1.6 ± 0.3t 2.6 ± 0.5t 3.2 ± 0.6* 4.1 ± 0.5*NGF-NBM 1.4 ± 0.5t 5.4 ± 0.9* 3.8 ± 1.0* 3.0 ± 0.6*

Different from control-graft (*) or nonimpaired (t), P < 0.05,one-way ANOVA, Fisher PLSD post hoc test. Values are given asmean ± SEM.aNumber of crossings over the southwestern location (used as platformsite in previous tests), once the platform was shifted to northeasternfor test 3.bAverage of crossings during the 3rd and 16th days of test 3.CNontransplanted animals.

the control group, whereas the MS-graft group was indistin-guishable from the nonimpaired rats (data not shown).Long-Term in Vivo Expression of the Transgene. NGF

protein was quantified in tissue pieces dissected from animalscontrol (0) grafted in the NBM with control- or NGF-cells and comparedNGF in NBM .W with other brain regions; the hippocampus and cortex were

non-impard (a) used as high-NGF content controls and ventral mesencephalonas low NGF-content tissue. The amount of immune-detectableNGF protein in an ELISA assay is shown in Table 2, both foryoung (4 weeks after grafting) and aged animals (10 weeksafter grafting). The NGF levels in the NGF-grafted regionswere clearly higher than in the control grafted-tissue (increas-

ated days PS refer ing 1.55- and 2.64-fold in young and aged rats, respectively, P =

]), animals grafted 0.01 in both cases). Compared with the NGF content in the)rafted with NGF- cortex (100% at each age), the NGF values recorded in the(lower plots) were NGF-grafted NBM amounted to 84% and 122% in the youngdiagram shows the and aged recipients, respectively, whereas the values in thehe pregraft session control-grafted NBM were only 54% and 46%, respectively.ng the second (test Consistent with the NGF-ELISA data, RT-PCR amplificationg the pretest, the of the retroviral transcript in total RNA samples from grafted

s tested in parallel NBM tissue revealed clear expression of the NGF transgene inIy day during the four aged animals analyzed at 10 weeks after transplantationup on the 5th day, (data not shown).e). After grafting, Histological Analyses. [3H]Thymidine autoradiography re-npaired group and vealed surviving grafts in all grafted animals (Fig. 2D and H):ie extent that they The grafted cells were found scattered around the implanta-up (*); on the days tion sites, covering a distance of about 1 mm from the sites ofdifferent from the ...'(P < 0.05, one-way injection, with a glia-like morphology in Nissl-stained sections.tdifference). The morphological appearance, migration, and integration of

the control- and NGF-cells were similar to those describednt. Moreover, in earlier at 10 weeks after transplantation in young animals andrences in escape at 4 weeks after transplantation in aged rats (14, 36).0.404, one-way NGF-producing grafts induced a hypertrophic response ofance seen in the the p75NTR.positive neurons in each target region (163% ands could not be 182% of control-graft values in the MS and NBM, respectively;D. seen only when the NGF graft was placed in these regions; foured in the NBM to seven randomly selected animals per group; Fig. 2 and Tablesecrete BDNF 3). By contrast, grafts of BDNF-producing cells had no effect.improvement in In the NBM, the NGF-induced local trophic response lead tocases similar to a normalization of cell volumes in the aged animals compared

-Producing Celllifferences were

MS or the NBM)f crossings over

nber of crossingsF-grafted groupstrol-grafted rats,npaired animalsps were slightlyency parametere location of themed similarly to

Table 2. NGF ELISA determination on tissue samples fromtransplanted adult or aged animals at 4 or 10 weeks after grafting

NGF content (pg/g wet tissue)Region Young (n = 8) Aged (n = 13)

Ventral mesencephalon 189 ± 24 203 ± 49Hippocampus 1929 ± 338 1503 ± 120Cortex 709 ± 55 690 ± 79NBM, control-graft 382 ± 57* 320 ± 59*NBM, NGF-graft 592 ± 86t 845 ± 178t*Different from cortex, P < 0.05 (one-tailed, student t test).tNGF-grafted NBM is different from control-grafted NBM (P =0.01) and not from cortex (P > 0.4).

Neurobiology: Marti'nez-Serrano et aL

Dow

nloa

ded

by g

uest

on

July

11,

202

0

6358 Neurobiology: Martinez-Serrano et al.

.A .. .:* . . ...... ; c. >... .- .. .....

B.- *zt * -S2

........... 2 F...... ... ,,,,,.j jF,,,wkij...F J....

' ,,-4,,;-,',;- '.:- ?,

*: A. . StSe'-_.. .S* x S ' ' ... ::: SS::': ,, ., .' ....... ',,

:. ';. L .

3!f

= Ss 0p a.?- -; ,<,, >s eefl Mil*s ?: '? ';

B C*

.,:..... . . .... a.

s .....

.........................

.... ... ,;R... ...... . ..*.. - a:^s:::

!

;-

\

F

A'I'L| ...

FIG. 2. Reversal of cholinergic neuron atrophy in aged rats by NGF-producing grafts. Overview of p75NTR immunostained cholinergic neuronsin the NBM (A-D) and MS (E-H) in nonimpaired (A and E) or memory-impaired animals receiving either control- (B and F) or NGF-secreting(C and G) grafts. The stars denote the original placement of one of the two deposits in each case, dorsal and lateral to the NBM nucleus (A-D)and the arrowheads point to the midline (E-H). (D and H) Darkfield microphotographs of [3H]thymidine autoradiograms showing the NGF-graftsinto the NBM (D) and MS (H). Individually labeled grafted cells (identified by grain clusters, white arrows) have integrated into the host tissuewithin a region extending about 1 mm from the injection site. The stars denote portions of the original implantation site. (Bar in H: A-C, 400 jm;E and F, 200 jim, D and H, or 100 jim.)

with both the young and the nonimpaired aged subjects (Fig.2 A-C). The MS graft group exhibited hypertrophy whencompared to the young controls (illustrated in Fig. 2 E-G;Table 3). p75NTR -positive neuron cell counts did not differbetween any of the groups (Table 3).

DISCUSSIONThe experimental evidence described here demonstrates thelong-term efficiency of an NGF-secreting neural stem cell linefor NGF gene transfer to the aged rat brain. The resultsprovide evidence for functional recovery of memory-impairedanimals after supplying NGF to the relevant cholinergic neu-ronal populations in MS and NBM, accompanied by a reversalof their atrophic state. In vivo transgene expression was

demonstrated by RT-PCR at 10 weeks after gene transfer;moreover, the NGF content in the transplanted regions wasincreased as determined by ELISA, thus substantiating thesustained in vivo production of the transgene protein productfrom the transplanted engineered cells.We and others (14, 19) have previously reported ameliora-

tive effects of short-term NGF gene transfer on age-inducedmemory impairments in rodents in experiments limited to a

3-week test schedule as previously used in NGF infusionexperiments (6, 9). In fact, the present study is the first toreport sustained effects of any type of exogenous NGF ad-ministration in aged rats beyond 1 month, showing thatNGF-producing neural stem cells implanted into either of thetwo principal cholinergic basal forebrain cell groups, MS andNBM, are effective in inducing long-lasting improvement in

Table 3. Morphometric analyses of the p75NTR-positive neurons

MSa NBMb

n Number Volume Number Volume

Youngc 4 22242 ± 4065 6050 ± 177 6998 ± 393 9024 ± 152Nonimpairedc 5 19885 ± 1838 6333 ± 602 5706 ± 573 9808 ± 1124Control MS and NBM 8 19251 ± 1566 5163 ± 213* 5748 ± 560 6294 ± 162*NGF-MS 7 17389 ± 1723 8396 ± 157*t - 6200 ± 200 6476 ± 363*NGF-NBM 4 18525 ± 2931 5213 ± 103 5869 ± 556 11444 ± 1725tBDNF-NBM 5 16446 ± 921 5250 ± 118 5971 ± 708 6104 ± 287*

Cell number and neuronal volume (tm3) in the MS and NBM regions, ten weeks after grafting control-, NGF-,or BDNF-cells. Values are mean ± SEM.aThe MS was analyzed bilaterally as a single structure.bNBM figures from each hemisphere were averaged to obtain cell counts and volumes in each animal.cNongrafted groups.*Different from young and nonimpaired groups.tDifferent from control-graft group. P < 0.05, one-way ANOVA, Fisher PLSDpost hoc test. Cell numbers werenot different among groups. Control-MS and control-NBM were not different from each other and thuscombined in a single control-MS and NBM group.

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

July

11,

202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 6359

spatial learning (both escape latency and platform crossings)in the Morris water-maze task for at least 10 weeks after cellimplantation. On the platform crossings measure, the mostrobust index of spatial acuity, both NGF-grafted groups weresignificantly improved compared with the control group, withscores as high as the nonimpaired animals. Consistent withprevious observations using osmotic minipump infusions ofBDNF (16), the BDNF-producing cells did not induce anyimprovement of the rats' performance in this task.The extent of behavioral improvement found in the present

experiment is quite similar to that observed after continuousinfusions of high doses ofNGF (6 ,ug per day for 4 weeks) (refs.16 and 17; see also ref. 18), and in previous studies on theshort-term effects of NGF gene transfer in aged rats (14, 19).Both the acquisition of place navigation (shown as a reductionin escape latency or swimming distance) and the formation/retention of spatial memory (evaluated as platform crossingsin the spatial probe trials) were improved. In the presentexperiment, the long-term stability of the recovery in spatiallearning and memory was observed in a 10-week testingparadigm. In particular, the use of ex vivo gene transferprocedures allowed for the first time to study the effect of localsustained NGF-delivery on the rats' ability to relearn the taskin a long-term experimental design. In the platform reversaltest (test 3), the NGF-grafted animals could relearn and retainthe new platform location similarly to the nonimpaired rats,whereas the control animals were impaired on the acquisitionof this new platform site. These observations imply that thespatial learning abilities of the NGF-grafted rats were im-proved in a sustained manner, rendering the animals able tomake proper use of spatial information to adopt a new spatialsearch strategy.Improved spatial learning has previously been observed with

NGF-secreting stem cells grafted either into the NBM alone orinto both the NBM and MS (14); NGF-secreting fibroblastshave been shown to induce similar effects when transplantedin the NBM region of learning-impaired aged rats (19). In ourprevious study, we could not elucidate the relative importanceof the MS and NBM regions in mediating the NGF-effect.Previous infusion experiments are equally inconclusive in thisrespect, because intraventricularly administered NGF is widelydistributed in the brain and may affect multiple targets,including the basal forebrain, striatum, hippocampus, andcortex (47, 48). In the present work, we confirm that long-termstimulation of the NBM region results in improved perfor-mance in the water maze, but we also show that NGF supplyconfined to the septal area results in an equal or even greaterimprovement compared with NBM transplants. Interestingly,in young animals, widespread cholinergic deficits induced bythe cholinergic immunotoxin 192 IgG-saporin are required toimpair task performance in the water maze test, whereas focalMS or NBM lesions have little or no effect, suggesting thateach of the two cholinergic nuclei is able by itself to ensureperformance in the task (49-51).

Consistent with the pattern of behavioral improvement, thehistological analyses revealed a trophic response in the basalforebrain cholinergic neurons only in those animals and re-gions that received an NGF-producing graft, such that thep75NTR_positive neuron volume was restored or even increasedabove that found in the young controls. No indication ofincreased cell numbers was obtained in any of the groups,which is in agreement with our previous short-term genetransfer experiment (ref. 14; see ref. 19), as well as with severalNGF infusion experiments (9, 16, 17). The fact that a localizedhost trophic response, as seen here, is linked to significantspatial memory improvement provides strong evidence infavor of the interpretation that functional stimulation of eitherof two principal cholinergic forebrain projection systems, i.e.,MS or NBM, is sufficient for restoration of spatial memoryperformance in aged rats, and moreover that local application

of the neurotrophin at the cell body level is sufficient to inducethis effect.The estimated NGF secretion rate of the present NGF grafts

(100 ng per day on each side; see ref. 36) is much lower thanthe amount of NGF necessary to induce similar effects afterintraventricular infusion (6 jig per day). This indicates that thecellular gene transfer system is a considerably more efficientdelivery system for neurotrophic proteins than pump infusioninto the cerebrospinal fluid. In this context, it should be notedthat the in vitro NGF secretion rate of the E8 stem cell cloneused here is 10-fold higher than that of the transducedprimary fibroblasts used by Chen and Gage (19), which mayexplain why more pronounced cellular and behavioral effectsare obtained with the NGF-secreting neural stem cells.An important aspect addressed in our study relates to the

effectiveness of the present ex vivo gene transfer approach inproviding the brain with a significant amount of transgenicprotein, which is a critical point in the validation of any genetransfer protocol. Previous reports (14, 36) have indicated thatgene expression from the (monocistronic) retroviral vector inthe NGF-stem cell line could be detected in vivo as long as 10weeks in adult animals and 4 weeks in aged subjects, asmeasured by RT-PCR or as NGF-like neurotrophic activity(14). Here we have used NGF-ELISA determinations on tissuesamples dissected from transplanted young or aged rat brainsto demonstrate the presence of NGF protein in the graftedNBM region at 4 and 10 weeks after transplantation of thetransduced cells, as well as RT-PCR determinations to dem-onstrate the expression of the retroviral NGF message. Theactual quantification of NGF protein at the target regionprovides a direct demonstration that the cell line used in ourstudies continues to express the transgene, and that this resultsin an increased NGF tissue content in the transplanted area.The NGF levels obtained in vivo are high enough to exert awide-spread neurotrophic effect within the target nuclei. In-deed, recent evidence indicates that the continued presence ofelevated NGF levels is required to maintain NGF-inducedcholinergic neuronal hypertrophy over time. Thus, Kordoweret al. (52) have reported that the NGF-induced cellular re-sponse dissipates within 3 weeks following cessation of exog-enous NGF supply.The results reported here demonstrate the efficiency of the

present ex vivo gene transfer strategy as a localized deliverysystem for neurotrophic factors to the brain. When comparedwith other protein delivery systems to the brain, the currentapproach shows clear advantages over injections, infusions, orother cellular vehicles, exemplified by fibroblasts: Immortal-ized neural stem cells provide the brain with an even buttargeted source of transgenic protein for long periods of time,resulting in an increased content of the transferred trophicfactor. Taken together, available data on the molecular, cel-lular, and behavioral levels provide a solid basis for the use ofneurotrophin-secreting neural stem cell lines as a powerfulexperimental therapeutic approach to counteract neurodegen-erative changes and promote functional recovery in the dis-eased or injured central nervous system.

The authors wish to thank Sten Nilsson for excellent animal care,Birgit Haraldsson, AnnaKarin Olden, Gertrude Stridsberg, KerstinFogelstrom, Alicja Flasch, Ulla Jarl, and Cristina Ciornei for experttechnical assistance, and Soledad Conte for data processing. This workwas funded by grants from the Human Frontier Science Program(A.B.), the Swedish Medical Research Council (Grant 04X-3874 toA.B. and Grant 19X-11632 to A.M.-S.), the Swedish Natural ScienceResearch Council (Grant B-AA/BU 04024-317 to T.E.), and theSegerfalk, Ake Wiberg and Kock foundations. A.M.-S. was partiallysupported by the Human Frontier Science Program and the SwedishMedical Research Council.

1. Fibiger, H. C. (1991) Trends Neurosci. 14, 220-223.

Neurobiology: Martinez-Serrano et al.

Dow

nloa

ded

by g

uest

on

July

11,

202

0

6360 Neurobiology: Martinez-Serrano et al.

2. Dunnett, S. B. & Fibiger, H. C. (1993) Prog. Brain Res. 98,153-420.

3. Finch, C. E. (1993) Trends Neurosci. 16, 104-110.4. Gallagher, M. G. & Colombo, P. J. (1995) Curr. Opin. Neurobiol.

5, 161-168.5. Biegon, A., Greenberger, V. & Segal, M. (1986) Neurobiol. Aging

7, 215-217.6. Fischer, W., Wictorin, K., Bjorklund, A., Williams, L. R., Varon,

S. & Gage, F. H. (1987) Nature (London) 329, 65-68.7. Fischer, W., Gage, F. H. & Bjorklund, A. (1989) Eur. J. Neurosci.

1, 34-45.8. Altavista, M. C., Bentivoglio, A. A. R., Crociani, P., Rossi, P. &

Albanese, A. (1988) Brain Res. 455, 177-181.9. Fischer, W., Bjorklund, A., Chen, K. S. & Gage, F. H. (1991)

J. Neurosci. 11, 1889-1906.10. Fischer, W., Chen, K. S., Gage, F. H. & Bjorklund, A. (1991)

Neurobiol. Aging 13, 9-23.11. Smith, M. L. & Booze, R. M. (1995) Neuroscience 67, 679-688.12. Koh, S., Chang, P., Collier, T. J. & Loy, T. (1989) Brain Res. 498,

397-404.13. Gallagher, M., Burwell, R. D., Kodsi, M. H., McKinney, M.,

Southerland, S., Vella-Roundttree, L. & Lewis, M. H. (1990)Neurobiol. Aging 11, 507-514.

14. Martinez-Serrano, A., Fischer, W. & Bjorklund, A. (1995) Neu-ron 15, 473-484.

15. Gage, F. H., Kelly, P. A. T. & Bjorklund, A. (1984) J. Neurosci.4, 2856-2865.

16. Fischer, W., Sirevaag, A., Wiegand, S. J., Lindsay, R. M. &Bjorklund, A. (1994) Proc. Natl. Acad. Sci. USA 91, 8607-8611.

17. Fischer, W. (1994) Neurochem. Int. 25, 47-52.18. Markowska, A. L., Koliatsos, V. E., Breckler, S. J., Price, D. L. &

Olton, D. S. (1994) J. Neurosci. 14, 4815-4824.19. Chen, K. S. & Gage, F. H. (1995) J. Neurosci. 15, 2819-2825.20. Rylett, R. J., Goddard, S., Schmidt, B. M. & Williams, L. R.

(1993) J. Neurosci. 13, 3956-3963.21. Lindsay, R. M., Wiegand, S. J., Altar, C. A. & DiStefano, P. S.

(1994) Trends Neurosci. 17, 182-190.22. Rylett, R. J. & Williams, L. R. (1994) Trend Neurosci. 17, 486-

490.23. Nitta, A., Murase, K., Furukawa, Y., Hayashi, K., Hasewaga, T.

& Nabeshima, T. (1993) Neuroscience 57, 495-499.24. Van der Zee, C. E. E. M., Lourenssen, S., Stanisz, J. & Diamond,

J. (1995) Eur. J. Neurosci. 7, 160-168.25. Bartus, R. T., Dean, R. L., Beer, B. & Lippa, A. S. (1982) Science

217, 408-157.26. Coyle, J. T., Price, D. L. & DeLong, M. R. (1983) Science 219,

1184-1190.27. Whitehouse, P. J., Price, D. L., Struble, R. G., Clark, A. W.,

Coyle, J. T. & DeLong, M. R. (1982) Science 215, 1237-1239.28. Mufson, E. J., Bothwell, M. & Kordower, J. H. (1989) Exp.

Neurol. 105, 21-232.29. Wilcock, G. K., Eseri, M. M., Bowen, D. M. & Smith, C. T.

(1982) J. Neurol. Sci. 57, 407-157.

30. Strada, O., Vyas, S., Hirsch, E. C., Ruberg, M., Brice, A., Agid,Y. & Javy-Agid, F. (1992) Proc. Natl. Acad. Sci. USA 89,9549-9553.

31. Longo, F. M., Holtzman, D. M., Grimes, M. L. & Mobley, W. C.(1993) inNeurotrophic Factors, eds. Loughlin, S.E. & Fallon, J. H.(Academic, London), pp. 209-256.

32. Jelsma, T. N. & Aguayo, A. J. (1994) Curr. Opin. Neurobiol. 4,717-725.

33. Bjorklund, A. (1993) Nature (London) 362, 154-155.34. Friden, P. M., Walus, L. R., Watson, P., Doctrow, S. R.,

Kozarich, J. W., Backman, C., Begman, H., Hoffer, B., Bloom, F.& Granholm, A. C. (1993) Science 259, 373-377.

35. Kordower, J. H., Charles, V., Bayer, R., Bartus, R. T., Putney, S.,Walus, L. R. & Friden, P. M. (1994) Proc. Natl. Acad. Sci. USA91, 9077-9080.

36. Martinez-Serrano, A., Lundberg, C., Horellou, P., Fischer, W.,Bentlage, C., Campbell, K., McKay, R. D. G., Mallet, J. &Bjorklund, A. (1995) J. Neurosci. 15, 5668-5680.

37. Rosenberg, M. B., Friedmann, T., Robertson, R. C., Tuszynsky,M., Wolf, J. A., Breakefield, X. 0. & Gage, F. H. (1988) Science242, 1575-1578.

38. Stromberg, I., Wetmore, C. J., Ebendal, T., Ernfors, P. & Olson,L. (1990) J. Neurosci. Res. 25, 405-411.

39. Kawaja, M. D., Rosenberg, M. B., Yoshida, K. & Gage, F. H.(1992) J. Neurosci. 12, 2849-2864.

40. Dekker, A. J., Fagan, A. M., Gage, F. H. & Thal, L. J. (1994)Brain Res. 639, 149-155.

41. Emerich, D. F., Hammang, J. P., Baetge, E. E. & Winn, S. R.(1994) Exp. Neurol. 130, 115-150.

42. Martinez-Serrano, A., Hantzopoulos, P. A. & Bjorklund, A.(1996) Eur. J. Neurosci. 8, 727-735.

43. Renfranz, P. J,. Cunningham, M. G. & McKay, R. D. G. (1991)Cell 66, 713-729.

44. Morris, R. G. M. (1981) Lear. Mot. 12, 239-260.45. Soderstrom, S., Hallbook, F., Ibafiez, C. F., Persson, H. &

Ebendal, T. (1990) J. Neurosci. Res. 27, 665-677.46. Gundersen, H. J. G., Brendtsen, T. F., Korbo, L., Marcussen, N.,

Moller, A., Nielsen, K., Nyengaard, J. R., Pakkenberg, B., So-rensen, F. B., Vesterby, A. & West, M. J. (1988) Acta Pathol.Microbiol. Immunol. Scand. 96, 379-394.

47. Yan, Q., Matheson, C., Sun, J., Radeke, M. J., Feinstein, S. C. &Miller, J. A. (1994) Exp. Neurol. 127, 23-36.

48. Anderson, K. D., Alderson, R. F., Altar, C. A., DiStefano, P. S.,Corcoran, T. L., Lindsay, R. M. & Wiegand, S. J. (1995) J. Comp.Neurol. 357, 296-317.

49. Nilsson, 0. G., Leanza, G., Rosenblad, C., Lappi, D. A., Wiley,R. G. & Bjorklund, A. (1992) NeuroReport 3, 1005-1008.

50. Torres, E. M., Perry, T. A., Blokland, A., Wilkinson, L. S., Wiley,R. G., Lappi, D. A. & Dunnett, S. B. (1994) Neuroscience 63,95-122.

51. Berger-Sweeney, Heckers, S., Mesulam, M.-M., Wiley, R. G.,Lappi, D. A. & Sharma, M. (1994) J. Neurosci. 14, 4507-4519.

52. Kordower, J. H., Chen, E.-Y., Mufson, E. J., Winn, S. R. &Emerich, D. F. (1996) Neuroscience, in press.

Proc. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

July

11,

202

0