Interactive effects of age and estrogen on cognitionand pyramidal neurons in monkey prefrontal cortexJiandong Hao*, Peter R. Rapp*†, William G. M. Janssen*, Wendy Lou‡, Bill L. Lasley§¶, Patrick R. Hof*†�,and John H. Morrison*†**

*Fishberg Department of Neuroscience and Kastor Neurobiology of Aging Laboratories, †Department of Geriatrics and Adult Development,and �Computational Neurobiology and Imaging Center, Mount Sinai School of Medicine, New York, NY 10029; ‡Department of PublicHealth Sciences, University of Toronto, Toronto, ON, Canada M5T 3M7; and §Center for Health and the Environment and¶California National Primate Research Center, University of California, Davis, CA 95616

Communicated by Bruce S. McEwen, The Rockefeller University, New York, NY, May 21, 2007 (received for review April 11, 2007)

We previously reported that long-term cyclic estrogen (E) treat-ment reverses age-related impairment of cognitive function me-diated by the dorsolateral prefrontal cortex (dlPFC) in ovariecto-mized (OVX) female rhesus monkeys, and that E induces acorresponding increase in spine density in layer III dlPFC pyramidalneurons. We have now investigated the effects of the same Etreatment in young adult females. In contrast to the results foraged monkeys, E treatment failed to enhance dlPFC-dependenttask performance relative to vehicle control values (group youngOVX�Veh) but nonetheless led to a robust increase in spinedensity. This response was accompanied by a decline in dendriticlength, however, such that the total number of spines per neuronwas equivalent in young OVX�Veh and OVX�E groups. Robusteffects of chronological age, independent of ovarian hormonestatus, were also observed, comprising significant age-relateddeclines in dendritic length and spine density, with a preferentialdecrease in small spines in the aged groups. Notably, the spineeffects were partially reversed by cyclic E administration, althoughyoung OVX�Veh monkeys still had a higher complement of smallspines than did aged E treated monkeys. In summary, layer IIIpyramidal neurons in the dlPFC are sensitive to ovarian hormonestatus in both young and aged monkeys, but these effects are notentirely equivalent across age groups. The results also suggest thatthe cognitive benefit of E treatment in aged monkeys is mediatedby enabling synaptic plasticity through a cyclical increase in small,highly plastic dendritic spines in the primate dlPFC.

aging � estradiol � hormone � neocortex � plasticity

Numerous studies have demonstrated that E affects synapticstructure and function in multiple brain regions importantfor memory and cognition (1). In the hippocampus of youngfemale rats, 17�-estradiol [the dominant estrogen (E) in rats,monkeys, and humans] increases the density of dendritic spinesand axospinous synapses on CA1 pyramidal cells (2, 3). Theseeffects are N-methyl-D-aspartate (NMDA) receptor-dependent(4), and E both directly increases NMDA receptor levels andfacilitates NMDA receptor-mediated responses in CA1 pyrami-dal neurons (5, 6). E also affects GABAergic (7) and cholinergic(8) systems in the CA1 field of the young female rat hippocam-pus. The data from young adult monkeys are generally consistentwith these observations, which is particularly relevant to humansgiven the reported similarities in cyclicity and menopause be-tween women and female rhesus monkeys (9, 10). Of particularnote, E increases CA1 spine number in ovariectomized (OVX)young African green (11) and rhesus monkeys (12). Estrogenadministration also increases spine number in layer I of dorso-lateral prefrontal cortex (dlPFC) of young rhesus monkeys (13)and enhances cholinergic and monoaminergic inputs to thisregion (14, 15), demonstrating that potential targets for ovarianhormone influences on cognitive function extend beyond thehippocampus in primates.

Given the vulnerability of many of these same systems to aging,together with the marked decline in circulating ovarian hor-mones at menopause in women, a number of laboratories haveinvestigated potential influences of endocrine senescence oncognitive and neurobiological aging in animal models. Studiescomparing young and aged OVX rats have demonstrated thatthe hippocampal response to E in aged female rats is attenuatedas compared with young females with respect to both CA1synapse number (16) and cognitive function (17), and that theefficacy of intervention declines further when the delay betweenOVX and initiation of E administration is extended for severalmonths (18, 19). However, the emerging story in aged non-human primates differs in important ways from the rodentfindings (20). First, the available evidence suggests that thecognitive benefit of E in aged OVX monkeys is substantiallygreater than in young adults, particularly on tests emphasizingdlPFC integrity (21). A previous study, for example, reportedthat performance on the dlPFC-dependent delayed response(DR) is not enhanced by continuous E treatment in OVX youngmonkeys (22). In contrast, the same group observed a reliablecognitive benefit of intervention in middle-aged monkeys (23).Extending the latter results, we recently reported that long-termcyclic E substantially improves DR performance in aged OVXmonkeys, yielding task accuracy comparable to levels observedin young adults (24). The same E-treated monkeys also displayeda corresponding increase in spine density, together with a subtleshift toward smaller spines on dlPFC neurons (i.e., area 46) (25),establishing a potential neurobiological substrate for the cogni-tive benefit of E.

The present study extended our analysis of estrogen influenceson the cognitive and morphological outcome of aging to youngadult monkeys treated equivalently. The results revealed bothage and treatment effects, suggesting that whereas the ageddlPFC remains responsive, the specific pattern is qualitativelyand quantitatively different in young monkeys. Overall, thefindings suggest that whereas young adults without E can sustainexcellent cognitive function against a background of dynamicspine plasticity, the ‘‘double hit’’ of age and E deficiency leadsto morphologic alterations that significantly compromise thefunctional integrity of dlPFC in aged female rhesus monkeys.

ResultsEstrogen Levels. The same cyclical estradiol treatment that wasshown previously to elevate serum levels to a physiological range,

Author contributions: J.H., P.R.R., P.R.H., and J.H.M. designed research; J.H., W.G.M.J., andB.L.L. performed research; J.H., P.R.R., W.L., and P.R.H. analyzed data; and J.H., P.R.R., andJ.H.M. wrote the paper.

The authors declare no conflict of interest.

Abbreviations: dlPFC, dorsolateral prefrontal cortex; E, estrogen; OVX, ovariectomized;Veh, vehicle; DR, delayed response.

**To whom correspondence may be addressed at: Department of Neuroscience, Box 1065,Mount Sinai School of Medicine, New York, NY 10029. E-mail: john.morrison@mssm.edu.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0704757104 PNAS � July 3, 2007 � vol. 104 � no. 27 � 11465–11470

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

June

6, 2

021

consistent with the preovulatory peak in intact monkeys (24),was used in the young E treated animals. Serum samples takenat the time of perfusion (24 h after the final injection) demon-strated serum estradiol levels averaging 112 pg/ml in the treat-ment group and 24 pg/ml in the Veh group (P � 0.01). Excludingan unexplained, outlying value in the Veh group, serum estradiollevels among controls averaged 15 pg/ml.

DR Performance. We previously reported that cyclic E treatmentsubstantially improves DR performance in aged OVX monkeys(24), establishing a valuable basis for comparison with thecurrent findings from young subjects, tested under identicalconditions. Although the number of trials required to reach thecriterion level of accuracy with a 1-s delay was substantiallyhigher in the aged Veh group, task acquisition at both the 0- and1-s delay failed to differ reliably across age and treatmentconditions (all P values � 0.5, data not shown). Notably, allsubjects scored at comparable, high levels of accuracy (90%correct or better) before testing with longer retention intervals.Performance on the delay portion of the task, imposing succes-sively more challenging memory demands, is illustrated in Fig. 1.Accuracy declined in all groups across delays of 5–60 s, con-firming that the procedure effectively taxed memory (maineffect of delay; P � 0.0001). Accuracy differed reliably acrossconditions (main group effect; P � 0.05), and there was nointeraction between group and delay (P � 0.5). Subsequentbetween group contrasts demonstrated that the aged OVX�Vehgroup scored less accurately than monkeys in each of the otherconditions (all P values � 0.03). Performance across the re-maining groups, by comparison, was statistically equivalent (allP values � 0.9). These results confirm our earlier conclusion thatcyclic E treatment substantially reverses DR impairment in theaged monkey, demonstrating for the first time that accuracy isrestored to levels comparable to young OVX subjects. In addi-tion, the findings indicate that, in striking contrast to the effectsobserved in aged subjects, spatiotemporal memory measured bythe DR task appears insensitive to OVX and E treatment inyoung adult monkeys.

Effects of Long-Term Cyclic E Treatment on Dendritic Arbor of Area 46Layer III Pyramidal Neurons. We previously reported that long-term cyclic E treatment had no effect on total length andbranching pattern of dendrites in layer III pyramidal neurons ofarea 46 in aged female rhesus monkeys (25). In contrast, asillustrated in Fig. 2, a numerically substantial treatment effect

was apparent in young female rhesus monkeys, with basaldendritic length and branch number in the young OVX�Vehgroup exceeding that of young OVX�E (mean � SEM, totalbasal dendritic length 2,148.3 � 72.5 �m and branch number of21.5 � 0.7 in OVX�Veh compared with total basal dendriticlength of 1,832.8 � 81.5 �m and branch number of 18.5 � 1.1 inyoung OVX�E; P � 0.04 for total basal dendritic length, P �0.02 for basal branch number). A similar trend observed amongapical dendrite parameters failed to reach significance. Shollanalysis showed that the relative increase in basal dendritic arborin young OVX�Veh compared with E treated adults occurredmainly 60–120 �m from the soma (data not shown).

Comparison of the young groups with our previous data fromaged animals revealed clear effects of aging on dendritic arborof layer III pyramidal neurons in area 46. Apical and basaldendritic length and branch number (Fig. 2) were decreased inthe aged monkeys, and these effects were especially pronouncedin comparisons between the young OVX�Veh (total dendriticlength 1,911.3 � 112.2 �m for apical and 2,148.3 � 72.5 �m forbasal; total branch number, 18.8 � 1.3 for apical and 21.5 � 0.7for basal) and aged OVX�Veh groups (total dendritic length1,603.4 � 107.9 �m for apical and 1,875.0 � 107.2 �m for basal;total branch number, 15.4 � 0.6 for apical and 19.1 � 0.9 for

Fig. 1. Performance on DR of young and aged OVX�Vehicle (Veh) andOVX�E monkeys. Unlike aged monkeys, young OVX�Veh and young OVX�Erhesus monkeys performed equally well on DR tasks. In addition, the perfor-mance of the aged OVX�E group was equivalent to both young treatmentgroups, with the aged OVX�Veh performing significantly worse than all threeother age and treatment groups. Error bars represent SEM.

Fig. 2. Quantitative analysis of aging and treatment effects on dendriticarbor of layer III pyramidal neurons in area 46. Significant age-related den-dritic arbor reductions were manifested between young and aged OVX�Vehgroups on apical (A and B) and basal (C and D) dendritic length and branchnumber. An age-related decrease in apical branch number also occurs be-tween young and aged OVX�E groups (B). Young OVX�Veh monkeys dem-onstrated increased dendritic length and higher branch number when com-pared with young OVX�E monkeys, although significance was limited to basaldendrites (C and D). Error bars represent SEM.

11466 � www.pnas.org�cgi�doi�10.1073�pnas.0704757104 Hao et al.

Dow

nloa

ded

by g

uest

on

June

6, 2

021

basal). Additional, marginally significant age-related decreasesin apical branch number were observed between the youngOVX�E and aged OVX�E groups (P � 0.04). Sholl analysisdemonstrated that the aging effects occurred mainly 60–150 �mfrom the soma in the apical dendritic arbor (data not shown).

Effects of Long-Term Cyclic Treatment with 17�-Estradiol on DendriticSpine Density. As previously observed in aged OVX female rhesusmonkeys (25), young OVX female rhesus monkeys also displaya profound treatment effect on spine density, with E treatmentincreasing apical (from 1.06 � 0.04/�m to 1.31 � 0.04/�m) andbasal dendritic spine density (from 1.03 � 0.03/�m to 1.25 �0.04/�m) across all dendritic orders except the primary dendrites(Fig. 3, P � 0.001 for both apical and basal dendrites). Com-parisons across age groups in the same treatment conditions (i.e.,young versus aged Veh groups, young versus aged E groups)revealed highly reliable declines in spine density in both apicaland basal dendrites (Fig. 3, P � 0.001). Importantly, E treatmentpartially restored spine density among aged subjects, resulting inlevels comparable to vehicle-treated young adults, but substan-tially lower than in young monkeys provided E.

We were intrigued by the opposing effect of treatment inyoung animals on dendritic length and spine density. Notably,the 18.3% decrease in spine density in young OVX�Vehanimals, relative to the OVX�E group, was coincident with a15.3% increase in dendritic length and as a consequence, spinenumber per neuron was roughly equivalent across these groups.These findings suggest that total dendritic spine representationis held constant under conditions of low and high circulating Ein the young adult dlPFC, potentially accounting for the lack oftreatment effect on DR performance. In contrast, dendriticlength appears insensitive to estrogen treatment in aged mon-keys, and fails to counteract the robust spine proliferationinduced by hormone administration.

Effects of Long-Term Cyclic 17�-Estradiol Treatment on DendriticSpine Morphology in Area 46 Layer III Pyramidal Neurons. Currentevidence indicates that spine size (e.g., spine head diameter,head volume and surface area) is proportional to synapse size(26) and related to glutamate receptor representation (27, 28).In addition, small spines appear especially motile and plastic (29,30), potentially comprising a specialized subpopulation, distin-guished by its capacity for growth and expansion in relation tonew learning (29). We previously investigated the effects of E onspine head diameter (Hd), head length (Hl), and neck length (Nl)in aged OVX rhesus monkeys and demonstrated that E led to

both an increase in spine density and a shift toward smallerspines (25). The current analysis determined whether there wasan overall E-induced shift in the spine size distribution, or adisproportionate increase in the incidence of spines with small orlarge heads in young OVX monkeys. In addition, we performedanalyses comparing all four groups to reveal any age effects onspine size. The results for these spine parameters are presentedin Fig. 4, displayed as cumulative percentage against increasingsize. The effect of age on spine head diameter (Fig. 4A) wasdramatic and highly significant (P � 0.0001), with aged animalsdisplaying far fewer small spines than young monkeys. Hormonetreatment exerted an opposing effect; i.e., E administrationinduced a leftward shift in the size distribution, toward smallspines. The latter effect, however, failed to fully reverse the resultattributable to age. Although E administration shifted size to theleft in both age groups, this effect was more pronounced in agedmonkeys (P � 0.005) than in young subjects (P � 0.02).

A further, quartile analysis confirmed that E induced a shifttoward smaller dendritic spines in both young and aged monkeys(P � 0.001, Fig. 5). Independent of hormone treatment condi-tion, it is also noteworthy that whereas spines in the 25thpercentile were nearly twice as prevalent in young monkeys,large spines (75th percentile) clearly predominated in the agedgroup. Perhaps most striking, small spines in the young OVX�Egroup were nearly 3-fold more frequent than among agedOVX�Veh monkeys, reinforcing the heightened vulnerability ofaged animals without E. Head length, by comparison, was similaracross groups, without discernible treatment or age effects(Fig. 4B).

Neck length measurements also revealed prominent effects ofage, together with a more subtle influence of hormone status(Fig. 4C). Aged monkeys displayed a strong shift to the left incumulative percentage, demonstrating a preponderance of shortspines, and corresponding loss of long spines, as compared withyoung monkeys (P � 0.0001). E treatment had no effect on necklength in aged monkeys and therefore failed to account for the

Fig. 3. Quantitative analysis of aging and treatment effects on spine densityof layer III pyramidal neurons in area 46. Significant treatment effects are seenin both young and aged spine densities in apical (A) (average from 2nd to 4thorder) and basal (B) dendrites (P � 0.001 for both young and aged groups).Significant age-related decreases are evident in spine densities on both apical(A) and basal (B) dendrites (P � 0.001 for young and aged OVX�Veh andOVX�E groups). Error bars represent SEM.

Fig. 4. Cumulative frequency analysis of aging and treatment effects onspine size: Hd, Hl, and Nl. (A) Significant aged related increases of Hd wereobserved between the aged (OVX�Veh and OVX�E) vs. young (OVX�Vehand OVX�E) groups (P � 0.0001). Although a highly significant treatmenteffect was observed in the aged group (P � 0.005), i.e., Hd decreased followingtreatment, a marginally significant difference was observed in young animals(P � 0.02). (B) No aging and treatment effects on Hl were observed. (C) Overall,aging significantly reduced Nl (P � 0.0001). Although no treatment effects onNl were observed within the aged group, a marginally significant increase inNl was observed in the young OVX�E compared with young OVX�Veh group(P � 0.015).

Hao et al. PNAS � July 3, 2007 � vol. 104 � no. 27 � 11467

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

June

6, 2

021

observed age-related shift toward shorter spines. In contrast, Etreatment induced a marginally significant increase in longspines in young animals (P � 0.015).

DiscussionOur results offer a framework for understanding the cognitiveresiliency of young female rhesus monkeys under conditions offluctuating ovarian hormone levels, and the contrasting vulner-ability of aged female rhesus monkeys with chronically low E. Inthe morphological analysis, we focus on three reflections ofneuronal and synaptic structure and plasticity: dendritic arbor,spine number, and spine size. All three indices display substan-tial age effects that could impact prefrontal cortex function, butthey vary in the degree to which E treatment counteracts theinfluence of aging. In addition, the data on dendritic arbor andspine density from young animals suggest that the effect oftreatment on one neuronal attribute can be neutralized by itsopposing influence on another structural parameter. Focusingexclusively on layer III pyramidal neurons in area 46 of themonkey prefrontal cortex, the present results provide substan-tial, novel insight concerning the effects of neuroendocrine agingon a class of neurons that directly mediate high level cognitiveprocesses such as working memory (31) and that support criti-cally important corticocortical interactions with other associa-tion areas (32, 33).

Aging affects all three parameters investigated: (i) dendriticlength and number of branches decreases with age; (ii) spines arelost throughout the dendritic tree; (iii) age-related spine loss isparticularly pronounced among long spines with small heads thatare thought to disproportionately participate in dynamic synapticplasticity. Age-associated decreases in dendritic length andbranch number are insensitive to E treatment. Furthermore, inyoung animals, the absence of E is associated with increaseddendritic length and branch number, particularly in the basaltree. In contrast to dendritic arbor, E increases spine density inboth young and aged OVX monkeys. Thus, the combined effectsof aging and E deficiency comprise a double hit in agedOVX�Veh monkeys. The E-induced recovery in aged monkeys

is partial, returning spine density to the level of Veh-treatedyoung adults, but still substantially less than observed in youngmonkeys given E. Importantly, the E-induced increase in spinedensity in young monkeys occurs alongside a concurrent de-crease in dendritic length, leading to an equalization of totalspine number per neuron with and without E. As with spinedensity, aging and E treatment also exert opposing influences onspine size, with the effect of age predominating. Specifically,aging is associated with a shift in the distribution of spine size inwhich long neck, small head spines are lost or do not form. Eadministration only partially reverses this effect, and the fullextent of vulnerability was particularly striking in comparisonsbetween young OVX�E monkeys and aged monkeys that re-ceived vehicle injections; here, the young adult group displayed3-fold more spines with small heads than aged subjects. Thepresent results reveal a complex interplay of age and E effects ondendritic architecture that provides a potential substrate forovarian hormone influences on cognitive function mediated byprefrontal cortex. We previously reported that cyclic E treat-ment profoundly benefits DR performance in aged femalemonkeys, restoring accuracy to levels statistically equivalent withintact young adult rhesus monkeys (24). Here we extended thosefindings to conduct a parallel assessment in OVX young adults.The key results were that E treatment had no effect on DRperformance in young monkeys, and in contrast, that cyclical Eadministration fully reversed the marked impairment observedin vehicle-treated aged subjects. These findings are consistentwith earlier reports demonstrating little or no cognitive effect ofE in young monkeys (21, 34), and importantly, they are the firstto document a corresponding morphologic resiliency in theyoung primate prefrontal cortex. First, increased dendritic arborin young untreated monkeys largely counterbalances E-inducedincreases in spine density, as gauged by the total number ofspines per neuron. Second, and perhaps more important, therelative abundance of small spines in both treated or untreatedyoung monkeys exceeded that of treated aged monkeys, suggest-ing that formation and retention of small spines is less dependenton E in young animals and remains robust in the absence of E.

Spine size may be a critically important index of the capacityfor synaptic plasticity in the aged brain. Available evidencesuggests that dynamic functional and structural plasticity arepreferentially mediated by small, thin spines with a relativelylong neck, and that active spinogenesis is selective for thismorphological class (29, 30). Rodent studies also suggest thatthese spines exhibit expansion, synaptic stabilization and retrac-tion (29, 30, 35), although the prevalence of these eventsthroughout different neocortical areas is controversial (36).Nonetheless, there is reason to suspect that small ‘‘learningspines’’ (29) may play a particularly significant role under testingconditions, like DR, that place substantial demands on workingmemory, i.e., the capacity for rapidly updating acquired infor-mation (37). Thus, the dramatic loss of these spines in agedmonkeys lacking E treatment may be a critically importantcontributor to the robust deficits they exhibit on dlPFC-dependent cognitive tasks. Viewed in relation to the behavioraland morphologic results from young adults, our proposal is thatthe representation of small, plastic spines in aged OVX monkeyswithout E may fall below a critical threshold needed to sustainnormal learning mediated by the dlPFC. By this perspective, theinfluence of endocrine decline on cognitive aging is best under-stood as the cumulative outcome of declining synaptic plasticityand the reduced capacity for structural reorganization enabledby circulating E.

It remains to be determined whether learning-related spineplasticity and turnover are dependent on the specific timing ofand schedule of E treatment in monkeys. In the present exper-iments, young and aged animals were treated cyclically, with anintramuscular injection of estradiol every 21 days throughout

Fig. 5. Quartile analysis of Hd. (A) Randomly sampled spines (850; 170 in eachof five animals) were selected from each of the four groups to determine therelative contribution of each group to the smallest 25th percentile (�0.4 �m)and the largest 25th percentile (�0.6 �m). Proportions of spines below andabove the two quartiles were generated for each group. (B) The smaller (Hd �0.4 �m) spine counts in the young animals were almost double those of theaged animals (P � 0.001). (C) The larger (Hd � 0.6 �m) spine counts were higherin the aged animals (P � 0.001). Overall, E treatment induces greater numbersof smaller spines in both young and aged groups (P � 0.001).

11468 � www.pnas.org�cgi�doi�10.1073�pnas.0704757104 Hao et al.

Dow

nloa

ded

by g

uest

on

June

6, 2

021

behavioral testing, and monkeys were killed 24 h after the lastinjection. The cyclical nature of E-induced spine formation hasbeen demonstrated in rodent hippocampus (2), and it may bethat cyclical E is critical for the proliferation of plastic spines thatparticipate in encoding new learning. Furthermore, our OVXand subsequent treatment occurs in the peri-menopausal stagefor the aged rhesus monkeys such that the cyclical exposure andresultant spinogenesis would not have been interrupted by anextended menopausal phase. The time of initiation of E treat-ment relative to menopause has become a major focus ofresearch in women (38, 39), consistent with emerging findings inrats (40). The present study speaks to this issue indirectly,demonstrating that early intervention in aged primates cansignificantly benefit the structural and functional integrity of akey neocortical region that is vulnerable to aging. The dissectionof age and neuroendocrine influences on the prefrontal cortexin primates will continue to be a high priority for research aimedat promoting optimally healthy cognitive aging.

Materials and MethodsAnimals and Treatment. All experiments were conducted in com-pliance with the National Institutes of Health Guidelines for theCare and Use of Experimental Animals approved by the Insti-tutional Animal Care and Use Committee at the University ofCalifornia, Davis.

Sixteen young female Rhesus monkeys (Macaca mulatta; meanage � SEM, 10 years � 3 months) were used in the behavioralphase of this study, and 12 of the 16 were used in the morphologicanalysis to demonstrate the treatment effects of long-termcyclical E treatment on layer III pyramidal cells in dlPFC (i.e.,area 46) of young female rhesus monkeys. The protocols wereidentical to those used previously on aged female rhesus mon-keys (24, 25), and data from all four age/treatment groups areused for new analyses in this report to reveal age and treatmenteffects. The young monkeys received bilateral OVX and wereassigned to age-matched OVX�Veh and OVX�E treatmentgroups, as was done previously with the aged monkeys. After anaverage post-OVX interval of 29 weeks, OVX�E monkeysreceived estradiol cypionate (100 �g/ml of sterile peanut oil, i.m.;Amersham Pharmacia, Peapack, NJ) in a single injection everythree weeks. OVX�Veh age-matched monkeys were providedan equivalent volume of vehicle injection according to the sameschedule. Treatment extended over �2 years of behavioral

testing. E and Veh injections were coded and administered in ablinded fashion until all experiments were completed. Serumestradiol values were measured at perfusion, 24 h after the finalinjection.

DR Testing. The DR testing was equivalent to the aged animalstested in ref. 24. Subjects watched from behind a transparentscreen while one of the wells of the apparatus was baited with afood reward, and both wells were then covered. During initialtraining, the screen was raised immediately, allowing monkeys todisplace the cover and retrieve the reward if the baited locationwas chosen. Testing continued until animals met a criterion of90% correct or better across nine consecutive blocks of 10 trials.In subsequent testing, a 1-s retention interval was imposedbetween baiting and the opportunity to respond. Training witha 1-s delay continued to the same performance criterion (�90%correct over 90 trials), and the demands of testing were thenmade progressively more challenging by imposing successivelylonger retention intervals of 5, 10, 15, 30, and 60 s. Each delaywas tested for a total of 90 trials (30 trials/daily session; intertrialinterval � 20 s).

Quantitative Analyses of Dendritic Arbor, Spine Density and SpineMorphology. Animals were perfused 24 h after the last E or Vehtreatment, and dissected into multiple standardized blocks, oneof which contained all of area 46. Sections (400-�m-thick)spaced 2.6 mm apart throughout the prefrontal cortex werecollected for intracellular injections and quantitative analysisaccording to methods described in ref. 25. Layer III pyramidalneurons in Brodmann area 46 within dlPFC were targeted foranalysis. Briefly, dendritic arbor from Lucifer Yellow filledpyramidal neurons [5–17 per monkey, total of 267 neurons (Fig.6)] were reconstructed in 3D, using Neurolucida morphometrysoftware (MicroBrightField, Williston, VT). Dendrograms and3-D Sholl analyses were prepared for each neuron with Neuro-Explorer. Dendritic arbor was expressed in terms of totaldendritic length and total branch number.

Printed datasets of dendritic tree maps were used to allowidentification of dendritic segments chosen in a systematic-random way for spine analysis. A clear acetate sheet, containingconcentric rings, was placed over the working maps and orientedsuch that the cell soma and innermost rings were aligned overeach other. Points of intersection between dendritic branches

Fig. 6. Representative layer III pyramidal neuron of area 46. (A and B) LY-loaded neuron at intermediate magnification (B) was reconstructed by usingNeurolucida and a 3D rendering dataset of the traced neuron (A) generated by NeuroExplorer was used for dendritic arbor measurements and to obtain thesystematic sampling for spine density analysis (see text for details). (C) Deconvolved confocal z-stacks from the LY-loaded neuron for 3D spine density analysis.(D) Single spines were analyzed for Hd, Hl, and Nl. (Scale bars: A and B, 50 �m; C, 5 �m; D, 0.5 �m.)

Hao et al. PNAS � July 3, 2007 � vol. 104 � no. 27 � 11469

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

June

6, 2

021

and the circles (60 �m apart) were used to identify thosesegments to be used for microscopic spine analysis. Confocalz-stacks of identified intersections were captured on a Zeiss LSM410 (Carl Zeiss, Oberkochen, Germany). An average of 12 stackswere captured per neuron (total 189 neurons). The confocalstacks were deconvolved with AutoDeblur (Media Cybernetics,Silver Spring, MD) and imported to Neurolucida for 3D spinedensity, and to VIAS (41) for spine morphology analysis. Allspine measurements were done manually (total 15,000 spines)from the z-stacks. Spine Nl was measured as the distance frompoint of attachment of the dendrite to the beginning of spinehead. Hl was the distance across the spine in continuity with thespine neck. Hd was taken at the largest diameter perpendicularto the head/neck axis.

Statistical Analysis. Group differences in serum estradiol levels inyoung monkeys were analyzed with a nonparametric (Mann–Whitney) test. Differences in behavioral performance across thefour groups on DR were assessed by using repeated measuresANOVA and Fisher’s probable least-squares difference. Toassess possible differences in various morphometric parametersbetween groups, ANOVA, or, when appropriate, the nonpara-metric Kruskal–Wallis test, was applied. The values are shown asmeans � SEM, calculated based on one aggregate (e.g., average)

per animal. The two-way mixed model for repeated measureswas used to test the treatment effect on dendritic length andbranch number across apical and basal dendritic trees, and onspine density across basal and apical trees. For spine morphologymeasurements, the overall distribution difference betweengroups was examined by using the Kolmogorov–Smirnov test.The �2 test was used for the quartile analysis of Hd. The overallstatistical significance level was set at 0.05. For the pairwisecomparisons of age and treatment effects, the P value thresholdwas corrected for multiple comparison (Bonferroni), and set atP � 0.0125. While adopting this conservative criterion, P valuesbetween the uncorrected 0.05 level and 0.0125 were consideredindicative of potentially important observations, characterizedas ‘‘marginally significant.’’ Statistical analyses were performedby using the software packages StatView and SAS (Version 9).

We thank Athena Wang, Carine Hamo, Ginelle Andrews, Susan Fink,Anne Canfield, Don Canfield, Mary Roberts, Sania Fong, DeborahKent, Harry Arwell, Sona Santos, and Heather McKay for technicalassistance; Douglas Ehlenberger and Alfredo Rodriguez for softwaredevelopment; and Drs. Murat Yildirim, Chet Sherwood, Anne Rocher,Jason Radley, Yong Tang, Christina Weaver, and Doron Kabaso forcritical discussion. This work was supported by National Institutes ofHealth Grants AG16765, AG06647, AG10606, MH58911, MH60734,and RR16754.

1. McEwen B (2002) Recent Prog Horm Res 57:357–384.2. Woolley CS, McEwen BS (1992) J Neurosci 12:2549–2554.3. Gould E, Woolley CS, Frankfurt M, McEwen BS (1990) J Neurosci 10:1286–

1291.4. Woolley CS, McEwen BS (1994) J Neurosci 14:7680–7687.5. Gazzaley AH, Weiland NG, McEwen BS, Morrison JH (1996) J Neurosci

16:6830–6838.6. Woolley CS, Weiland NG, McEwen BS, Schwartzkroin PA (1997) J Neurosci

17:1848–1859.7. Murphy DD, Cole NB, Greenberger V, Segal M (1998) J Neurosci 18:2550–

2559.8. Gibbs RB (2002) Horm Behav 42:245–257.9. Downs JL, Urbanski HF (2006) Biol Reprod 75:539–546.

10. Gilardi KV, Shideler SE, Valverde CR, Roberts JA, Lasley BL (1997) BiolReprod 57:335–340.

11. Leranth C, Shanabrough M, Redmond DE, Jr (2002) J Comp Neurol 447:34–42.12. Hao J, Janssen WG, Tang Y, Roberts JA, McKay H, Lasley B, Allen PB,

Greengard P, Rapp PR, Kordower JH, et al. (2003) J Comp Neurol 465:540–550.13. Tang Y, Janssen WG, Hao J, Roberts JA, McKay H, Lasley B, Allen PB,

Greengard P, Rapp PR, Kordower JH, et al. (2004) Cereb Cortex 14:215–223.14. Kritzer MF, Kohama SG (1999) J Comp Neurol 409:438–451.15. Tinkler GP, Tobin JR, Voytko ML (2004) J Comp Neurol 469:507–521.16. Adams MM, Shah RA, Janssen WG, Morrison JH (2001) Proc Natl Acad Sci

USA 98:8071–8076.17. Savonenko AV, Markowska AL (2003) Neuroscience 119:821–830.18. Gibbs RB (2000) Neurobiol Aging 21:107–116.19. Adams MM, Oung T, Morrison JH, Gore AC (2001) Exp Neurol 170:345–356.20. Morrison JH, Brinton RD, Schmidt PJ, Gore AC (2006) J Neurosci 26:10332–

10348.

21. Lacreuse A (2006) Neuroscience 138:859–867.22. Voytko ML (2000) Behav Neurosci 114:1078–1087.23. Tinkler GP, Voytko ML (2005) Prog Neuropsychopharmacol Biol Psychiatry

29:423–431.24. Rapp PR, Morrison JH, Roberts JA (2003) J Neurosci 23:5708–5714.25. Hao J, Rapp PR, Leffler AE, Leffler SR, Janssen WG, Lou W, McKay H,

Roberts JA, Wearne SL, Hof PR, Morrison JH (2006) J Neurosci 26:2571–2578.26. Harris KM, Stevens JK (1989) J Neurosci 9:2982–2997.27. Takumi Y, Ramirez-Leon V, Laake P, Rinvik E, Ottersen OP (1999) Nat

Neurosci 2:618–624.28. Kharazia VN, Weinberg RJ (1999) J Comp Neurol 412:292–302.29. Kasai H, Matsuzaki M, Noguchi J, Yasumatsu N, Nakahara H (2003) Trends

Neurosci 26:360–368.30. Holtmaat AJ, Trachtenberg JT, Wilbrecht L, Shepherd GM, Zhang X, Knott

GW, Svoboda K (2005) Neuron 45:279–291.31. Goldman-Rakic PS (1999) Ann NY Acad Sci 868:13–26.32. Hof PR, Morrison JH (1995) J Comp Neurol 352:161–186.33. Goldman-Rakic PS (1988) Annu Rev Neurosci 11:137–156.34. Voytko ML (2002) Behav Neurosci 116:187–197.35. Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, Welker E, Svoboda

K (2002) Nature 420:788–794.36. Zuo Y, Lin A, Chang P, Gan WB (2005) Neuron 46:181–189.37. Miller EK (2000) Nat Rev Neurosci 1:59–65.38. Sherwin BB (2006) Neuroscience 138:1021–1026.39. Maki PM (2006) Neuroscience 138:1027–1030.40. Gibbs RB, Gabor R (2003) J Neurosci Res 74:637–643.41. Rodriguez A, Ehlenberger D, Kelliher K, Einstein M, Henderson SC, Morrison

JH, Hof PR, Wearne SL (2003) Methods 30:94–105.

11470 � www.pnas.org�cgi�doi�10.1073�pnas.0704757104 Hao et al.

Dow

nloa

ded

by g

uest

on

June

6, 2

021