- Email: [email protected]
Maturational phase of hippocampal neurogenesis and cognitive flexibility
Neuroscience Letters 711 (2019) 134414
Contents lists available at ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
Research article
Maturational phase of hippocampal neurogenesis and cognitive flexibility a
b
c
Ryan D. Webler , Sasha Fulton , Tarique D. Perera , Jeremy D. Coplan
d,⁎
T
a
Department of Psychology, University of Minnesota, MN, United States Mount Sinai Medical Center, New York, NY, United States Contemporary Care LLC, United States d Department of Psychiatry & Behavioral Sciences, State University of New York, Downstate Medical Center, United States b c
A R T I C LE I N FO
A B S T R A C T
Keywords: Hippocampus Neurogenesis Cognitive flexibility BrdU-NeuN Doublecortin
Introduction: Pattern separation aids cognitive flexibility by reducing interference between closely related memories. Dentate gyrus (DG) neurogenesis may facilitate pattern separation by blocking memory retrieval via inhibition of non-neurogenic downstream CA3 neurons. We hypothesized that immature adult-born DG neurons would be associated with decreased CA3 activation and increased cognitive flexibility. Method: Two groups of adult male rats were tested either on the place avoidance task (PAT) (unflipped condition) or a subtly altered-PAT (flipped condition). Four weeks prior, the rats were injected with the mitotic marker BrdU. Immature new neurons were detected by the microtubule protein doublecortin (DCX). Cells that took up BrdU and expressed NeuN were identified as relatively more mature neurons. Synaptic activation was determined by c-Fos expression. Adaptation to the flipped versus unflipped condition reflected a measure of cognitive flexibility. Results: CA3 but not DG c-Fos was lower in the flipped versus unflipped condition [p = 0.002]. CA3 c-Fos correlated inversely with flipped task performance and immature (DCX) neurons with primary and secondary but not tertiary dendrites or more mature (BrdU + NeuN) new neurons. CA3 c-Fos was a significant predictor for the flipped versus unflipped condition specifically for DCX versus BrdU-NeuN neurons. Conclusion: Immature new neurons (DCX+) without tertiary dendrites may be preferentially implicated in cognitive flexibility relative to more mature new neurons (BrdU-NeuN). In combination with decreased CA3 activation in the flipped PAT, the functional contribution of these immature DG neurons may involve the inhibition of postsynaptic CA3 neurons containing traces of previously salient conditioned memories.
1. Introduction Adaptive behavior requires cognitive flexibility, the ability to efficiently shift between mental processes [1]. To date, areas of the prefrontal cortex have mostly been implicated in cognitive flexibility [2,3]. Recently however, the putative contribution of the hippocampus has been detailed [4]. The hippocampus plays a key role in pattern separation, the process of distinguishing closely related information [5,6]. Pattern separation serves cognitive flexibility by reducing interference between closely related information sets [7]. A developing literature suggests that newly derived, adult-born dentate gyrus (DG) neurons play a critical role in hippocampal mediated pattern separation [8–10]. These highly excitable neurons are more likely to activate to novel cues [11] and may
reduce interference between closely related information sets by dampening activation patterns corresponding to closely related past experiences, allowing for the encoding of novel information [12]. The developmental stage and mechanism by which new DG neurons interact with mature hippocampal neurons to form a substrate for cognitive flexibility is unclear. We have previously used the active place avoidance task (flipped PAT) to address these questions [13]. The flipped PAT, which ‘flips’ the shock zone 180 degrees from the initially learned position, is particularly sensitive to both hippocampal activation and neurogenesis [14]. In Perera et al. (2013), immature new neurons were identified using the microtubule doublecortin (DCX) and relatively more mature new neurons were identified using the mitotic marker bromodeoxyuridine (BrdU) and the neuronal marker NeuN. Our results suggested that the flipped PAT directly invokes immature new
⁎ Corresponding author at: Department of Psychiatry & Behavioral Sciences, SUNY Downstate Medical Center, 450 Clarkson Ave, Brooklyn, NY, 11203, United States. E-mail address: [email protected] (J.D. Coplan).
https://doi.org/10.1016/j.neulet.2019.134414 Received 20 August 2018; Received in revised form 11 July 2019; Accepted 5 August 2019 Available online 17 August 2019 0304-3940/ © 2019 Elsevier B.V. All rights reserved.
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
original task wherein a previously learned shock zone location was flipped 180 degrees (n = 11), and (2) an unflipped control condition (n = 7). The unflipped control condition implicated retrieval of the original shock location memory; the flipped condition putatively required suppression of this original memory and the encoding of a new shock location memory, invoking cognitive flexibility. The degree to which the animal can “adapt” to its new environment is a task putatively requiring cognitive flexibility, of which a plausible biological substrate should be identifiable.
neurons (DCX labeled) and that combined activation in the DG/CA3 regions may have components that are functionally implicated in cognitive flexibility. However, one important weakness of this study was that hippocampal activation (measured via c-Fos) was collapsed into a single measure combining the DG and CA3 regions. The combining of these distinct sub-regions obscured the significance of a negative correlation between immature DCX neurons and activation in a collapsed DG/CA3 region. Moreover, the correlation between the combined DG/ CA3 activation and a negatively predicted task performance on the flipped PAT was also ambiguous. The present analysis endeavored to identify the role of specific hippocampal sub-regions (DG and CA3) in cognitive flexibility and the developmental stage (immature DCX or relatively mature BrdU-NeuN) at which immature new neurons contribute to cognitive flexibility. In addition, we sought to identify the manner in which immature new DG neurons and CA3 activation interact to facilitate cognitive flexibility. Data from several studies shed light on interactions between these regions and their interactive functional significance. Specifically, these studies suggest that engram specificity, a pre-requisite for cognitive flexibility, requires DG feed-forward inhibition onto CA3 [15–17]. These findings necessitated an approach that distinguishes the DG and CA3 sub-regions. We now hypothesize that CA3 activation would correlate negatively with performance on the task invoking cognitive flexibility (flipped PAT) but not the task invoking recent spatial memory (unflipped PAT). Additionally, consistent with the feedforward inhibitory hypothesis and our previous work implicating DCX neurons in cognitive flexibility, we hypothesized that subjects with high counts of immature, DCX but not BrdU-NeuN marked new neurons would have reduced CA3 c-Fos in the flipped task requiring new memory formation and the putative blockade of a closely related memory (cognitive flexibility).
2.3. Neurogenesis and CA3 activation Cellular activity in the DG and CA3 regions was measured via expression of the early gene precursor, c-Fos [18]. c-Fos has a half-life of 2 h; rats were euthanized within 1 h of task completion on Day 4 to allow for task induced c-Fos activation. Neurogenesis was measured using previously published methods [19]. The microtubule protein double-cortin (DCX) stains DG neurons that have not yet completed the maturational point of tertiary dendrite formation (i.e. immature new neurons) [20]. The rats were injected with BrdU (100 mg/kg, i.p., for 3 days) 4 weeks prior to randomization. BrdU-NeuN co-labelled neurons were identified as relatively more mature new neurons [21,22]. These cells were assumed to be at least 5 weeks old given the timing of BrdU injection. Two independent raters (ICC > 0.90) who were blinded to animal identity, counted DCX expressing immature new neurons, c-Fos expressing synaptically activated mature neurons, and BrdU-NeuN co-labeled mature new neurons in every twentieth section (average of 15 per animal) of the left subgranular zone (SGZ) defined as a 50-mm band at the border of the granule cell layer and hilus. We calculated the density of DCX, BrdU-NeuN, and c-Fos labeled cells by dividing the total number of counted cells by the volume of SGZ in the counted sections (length of SGZ of section = 50 mm width = 40 mm depth/thickness before post-mounting). Of note, c-Fos was only counted in DG neurons that were BrdU negative as that would provide a counter-balanced relationship to CA3, where no new neurons expressing the mitotic marker were anticipated. Each neurohistological count was performed on separate hippocampal sections.
2. Materials and methods All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of SUNY Downstate Medical Center in compliance with NIH and national standards for the ethical treatment and use of vertebrate animals in research. These subjects were previously reported on in Perera et al. (2013). However, the relationship between immature new neurons as reflected by DCX-stained neurons in comparison to more mature new neurons reflected by BrdU-NeuN-stained neurons as predictors of CA3 c-Fos was not directly examined. Additionally, as previously described, hippocampal sub-regions (DG and CA3) were not properly disambiguated. In the current study, we examined, using general linear models which had not previously been employed, whether DCX-stained neurons vis-à-vis BrdU-NeuN-stained neurons were differentially predictive of CA3 c-Fos in the modified (flipped) versus unmodified (unflipped) condition of the PAT.
3. Statistical methods Data was analyzed using general linear modeling (Statistica 12.0) with post-hoc GLMs, generalized linear/non-linear models, and Pearson’s correlations. Due to technical difficulties, we have available 94% of study values and therefore we have had to omit samples, for instance in general linear models with repeated measures analyses. See Appendix Table 1 for total number of samples available for all study variables broken down by condition. For the first analysis, we used DG c-Fos and CA3 c-Fos as repeated measures, the flipped versus unflipped condition as the categorical variable, and individual DCX counts as the predictor variable. The GLM was then set up as a factorial ANOVA where an effect was provided for ROI c-Fos, for the flipped versus unflipped condition, and the interactive effect of ROI c-Fos* condition effect. When introducing DCX counts as a predictor variable, the factorial ANOVA now contained the triple interactive term – condition* ROI c-Fos * DG DCX interaction. This term would allow us to determine whether DG DCX would differentially predict CA3 c-Fos vis-à-vis DG c-Fos as a function of condition and ROI. Post-hoc generalized linear/non-linear models were separately performed for the flipped versus unflipped condition to examine for differential patterns. Finally, post-hoc Pearson’s correlations were performed contrasting DG DCX versus DG c-Fos to CA3 c-Fos in the flipped versus unflipped condition. For the next GLM DG DCX and DG BrdU-NeuN counts were used as repeated measures, condition (flipped versus unflipped) was entered as the categorical variable, and CA3 c-Fos was used as the predictor
2.1. Active place avoidance task In brief, we exposed Long-Evans male rats (N = 18) to the PAT, a paradigm that employs a continuously revolving circular arena to measure spatial segregation. In this task, rats avoid a stationary foot shock segment of the arena by distinguishing a salient, stationary spatial cue linked with a safe location, from a non-salient, rotating cue identified through deposition of rodent olfactory cues. [13]. Conditioned avoidance performance was assessed by: (1) time to first enter the shock-zone; (2) number of times entering into the shock zone; and 3) time to second entry into the shock zone. 2.2. Unflipped vs. Flipped modification of the PAT After undergoing training on the place avoidance task (PAT), rats were randomized to one of two conditions: (1) a ‘flipped’ version of the 2
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Table 1 General Linear Model Analysis of Dependent Repeated Measures Variables CA3 c-Fos and DG c-Fos in a Factorial ANOVA Design (see Fig. 1 and Results).
Group DCX Group*DCX Error R1 R1*Group R1*DCX R1*Group*DCX Error
DCX Error R1 R1*DCX total Error
Degr. Of Freedom
F
p
1 1 1 10 1 1 1 1 10
0.85 0.87 0.87
0.38 0.37 0.37
0.23 6.61 0.64 5.93
0.64 0.03 0.44 0.04
Degr. Of Freedom
F
p
1 8 1 1 8
3.21
0.111
18.94 15.95
0.002 0.004
Fig. 1. Immediate Gene c-Fos Expression in CA3 section of Hippocampus versus the Subgranular Zone (SGZ) of the Dentate Gyrus (DG) in relation to a Flipped versus Unflipped Version of the Place Avoidance Task. Legend: Whereas CA3 c-Fos was reduced in comparison to DG SGZ c-Fos in the flipped condition, there were no differences for c-Fos in the unflipped condition between CA3 and SGZ. See text and Table 1 for numerical results.
DG=Dentate Gyrus; Group = “flipped” versus “unflipped” condition. DCX=doublecortin-stained new neurons in the subgranular zone of the DG. Significant Results are Bolded. Post-hoc GLM: Flipped condition. No significant effects observed for unflipped condition.
revealed that, although the relationship between CA3 c-Fos and BrdUNeuN in the flipped condition was a trend using Pearson’s correlations, when examined non-parametrically, analyses revealed a significant inverse relationship (Spearman r = -0.89, N = 7, p = 0.007) whereas in the flipped condition there was no relationship between the variables (r = 0.22, N = 11, p = 0.50) leading to a significant condition*CA3 cFos interactive effect [Wald Statistic (1) = 4.60; p = 0.03; lower 95% CI = -0.01, upper 95% CI = 0.0005]. Moreover, in the flipped condition, the inverse relation between DG DCX and CA3 c-Fos differed significantly from the DG BrdU-NeuN to CA3 c-Fos relationship (the latter correlation: r = 0.22, N = 11, p = 0.50); F(1,9) = 16.46, p = 0.004]. An Omnibus GLM Analysis using DG DCX and DG BrdUNeun as the Repeated Measure, condition as the categorical variable and CA3 c-Fos as the independent predictor variable in a factorial ANOVA Design, revealed a triple interactive effect [SGZ Marker*condition*CA3 c-Fos interactive effect: F (114) = 4.83, p = 0.045]. In Appendix Fig. 1 we examine the relationship between DG c-Fos and DG DCX compared to the DG c-Fos and DG BrdU-NeuN relationship in the flipped versus non-flipped condition. No significant relationships were noted for the four correlations examined. In summary, CA3 c-Fos was specifically reduced in the flipped condition, and the magnitude of CA3 c-Fos reductions were inversely related to DCX counts in the DG. The latter inverse relation described was specific only to the flipped condition. By contrast, there was an inverse relationship between CA3 c-Fos and DG BrdU-NeuN only in the unflipped condition and that relationship was shown to be specific to the unflipped condition (Table 3).
variable. Variables were formatted into a factorial design so that the triple interactive effect could be evaluated. Finally, the relationship between CA3 c-Fos, DG c-Fos, DG DCX, and DG BrdU-NeuN were examined in relationship to the cognitive performance outcomes in the flipped versus unflipped condition. Because of the unusual distribution seen with the behavioral outcomes (e.g. time to enter cannot be greater than 600 s), a generalized linear/non-linear model was employed, in addition to Pearson’s correlations, to assess the relationships between neurohistological measures and cognitive performance. 4. Results 4.1. CA3 c-Fos as a function of flipped versus unflipped version of the PAT task In a model using the dependent regions-of-interest measures CA3 cFos and DG c-Fos in a factorial design with condition as a categorical variable (flipped versus unflipped) and continuous predictor variable, doublecortin stained new neurons of the subgranular zone of the dental gyrus, there was a main GLM effect for a condition*ROI interactive effect [F(110) = 6.60, p = 0.027] and a condition*ROI*Doublecortin triple interactive effect. The former result is illustrated by the observation that in a post-hoc GLM, in comparison to the “unflipped” condition where no ROI differences were noted, in a post-hoc GLM for the “flipped” condition, CA3 c-Fos was reduced in comparison to DG cFos [F(1,8) = 18.94, p = 0.002] (see Fig. 1, Table 1). As CA3 c-Fos and DG c-Fos were not standardized, using a post-hoc generalized linear/ non-linear model in a factorial design (Appendix Table 3), there was a condition effect [Wald Statistic (1) = 5.55; p = 0.02] indicating reduced CA3 c-Fos in the flipped versus unflipped condition [Lower CL 95.0% = -3.11; Upper CL 95.0% = -0.28]. By contrast, using the identical model, (see Appendix Table 3B) c-Fos of the dentate gyrus showed no significant differences as a function of condition. In the Fig. 2 scatterplot, inspection of the top two panels demonstrated that CA3 c-Fos was inversely related to DG DCX in the flipped condition (r= -0.79; N = 11, p = 0.003) whereas the same relationship in the unflipped condition was a direct non-significant correlation [(r= -0.22, N = 7, p = 0.62); [Condition * CA3 c-Fos interactive effect [Wald Statistic (1) = 5.49; p = 0.02; lower 95% CI = 0.0006, upper 95% CI = 0.007] (Table 2a, Fig. 2). Analysis of the lower two panels
4.2. Doublecortin-Stained Neurons with or without tertiary dendrites The next analysis addressed whether it was the more mature form of the DCX expressing neuron exhibiting tertiary dendrites in comparison to the more immature form of DCX-expressing neurons that only stained for primary and secondary dendrites. Using a GLM an overall condition* immature vs relatively more mature new neurons* CA3 c-Fos interaction was noted [F(111) = 6.32, p = 0.03] (See Fig. 3, Table 4). Post hoc-testing of the flipped condition revealed an effect for maturation of new neurons*c-Fos interactive effect indicating that it was the DCX neurons with primary or secondary dendrites that correlated inversely with CA3, rather than the DCX stained new neurons with tertiary dendrites [F(1,6) = 9.11, p = 0.023] (see Fig. 3). Moreover, when examining DCX neurons without tertiary dendrites, it was only in 3
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Fig. 2. Relationship of Maturational Phase of Adult Born Dentate Gyrus New Neurons to CA3 c-Fos. Legend: CA3 c-Fos was inversely related to DG DCX c-Fos in the flipped condition whereas the same relationship in the unflipped condition was a direct non-significant correlation. Although the relationship between CA3 c-Fos and BrdU-NeuN in the flipped condition was a trend, when examined non-parametrically it revealed a significant inverse relationship whereas in the flipped condition there was no relationship between the variables leading to a significant condition*CA3 interactive effect. See text and Table 2 for numerical results.
the flipped condition that there was a relation with CA3 c-Fos [F(111) = 4.88, p = 0.049]. These post-hoc results were replicated using a generalized linear/non-linear model.
4.3. Functional performance Pearson’s correlations revealed a direct relationship between CA3 cFos and number of entries into the shock zone during flipped condition (r = 0.61, N = 11, p = 0.046). However, this effect was not different from the unflipped condition (see Appendix Table 2 for all correlations 4
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
would differentially contribute to cognitive flexibility compared to those with tertiary dendrites. As hypothesized, both the individual variation of observed differences in CA3 activity and the proliferation of immature new DG neurons had functional implications on cognitive flexibility vis-à-vis task performance. The first measure of task performance, number of shock zone entries, was directly related to CA3 c-Fos in the flipped condition. A second measure, time to first entry, was inversely predicted by CA3 cFos in the flipped but not unflipped condition. The third measure, time to second entry, was positively predicted by total DG DCX (not simply DCX neurons with primary and secondary dendrites as noted in above findings) only in the flipped condition. Of the three significant analyses of neurohistological measures and task performance, the latter two were statistically distinguishable from the unflipped condition. No relationships in the unflipped condition were significant. Thus, low CA3 c-Fos and high DG DCX were predictive of outcomes in the cognitive flexibility condition (flipped task) but not the control condition (unflipped task). Because of the timing of our staining procedure, even the most mature BrdU-NeuN co-labelled new neurons were presumed to be at most 5 weeks of age. The functional dissociation that emerged between DCX and BrdU-NeuN neurons is thus especially striking given the narrow developmental window for transitioning neurohistological marker characterization in populations of new neurons in this sample. Dendritic findings lend additional support to these findings, suggesting that the most developmentally immature new DG neurons may contribute to cognitive flexibility. Previous work by Swan et al. (2014) implicates new neurons over one month old in cognitive flexibility [23]. While Swan and colleagues used mice, the current study used Long-Evans rats. There is data to suggest that new rat neurons mature more quickly than new mice DG neurons, outpacing their development by one or more weeks [24]. Thus, cross-species differences in the rate of new DG cell development may partially explain why new DG neurons older than one month were implicated in Swan et al. (2014) but not in the current study. Additionally, differences in task valence – Swan et al. (2014) used reward learning while the current report used aversive learning – need to be considered. Though DG engrams may be valence independent [25,26], divergent hippocampal afferents instantiate positive and negative value [27,28]. It is unclear whether these afferents differentially implicate immature DG neurons across various stages of development. The vast majority of new rat DG neurons express NeuN by 4 weeks [20,24]. New DG neurons between 4–6 weeks old have properties that confer increased plasticity, including decreased inhibition and higher plasticity in input and output synapses [17,29–32]. Thus, plasticity differences between immature DCX neurons and relatively more mature ≥ 5-week old BrdU-NeuN neurons are counterintuitive. However, we note that for this form of cognitive flexibility (measured via flipped PAT performance), it is the more immature stage of new DCX neurons that accounts for the effect. It is argued – discussed below—that a plausible interaction with mossy cells in the hilar area may account for the early forms of neuroplasticity observed in the current study. Mossy cells are large hilar neurons with a prominent commissural projection from ipsilateral hilar area to the contralateral DG and form the major source of afferents to the contralateral inner molecular layer [33]. Adult born DG cells receive monosynaptic input from a number of cells, including mossy cells [34]. Mossy cells provide new DG neurons with their first glutamatergic and disynaptic GABAergic inputs [35]. The mossy cells also have the capacity, via GABAergic interneurons, to synapse on primary and secondary dendrites of adult born DG cells [36]. Therefore, one mechanism by which immature new neurons with only primary and secondary dendrites may have modulated CA3 activation is through interactions with mossy cells. The ability of these latter cells to provide functionally relevant information in the flipped version of the PAT to the contralateral DG has to be considered. However, because the current study lacks mossy cell data, this
Table 2 Relationship of CA3 c-Fos to Subgranular Zone Neuronal Maturational Phase Markers: Doublecortin versus BrdU-NeuN Generalized Linear/Nonlinear Model. a. Dependent Variable : Total Doublecortin – Top Panel Effect
Degr. Of Freedom
Wald Stat
p
Condition CA3 c-Fos Condition*CA3 c-Fos
1 1 1
0.83 11.75 5.49
0.3630 0.0006 0.0191
b. Dependent Variable: BrdU-NeuN – Bottom Panel Effect
Degr. Of Freedom
Wald Stat
p
Condition CA3 c-Fos Condition*CA3 c-Fos
1 1 1
6.48 1.04 4.60
0.011 0.307 0.032
Table 3 General Linear Model Analysis Using DG DCX and DG BrdU-Neun as the Repeated Measures, Condition as the Categorical Variable and CA3 C-Fos as the Independent Predictor Varaible in a Factorial ANOVA Design.
Group CA3 c-Fos Group*CA3 c-Fos Error R1*Group R1*CA3 c-Fos R1*Group*CA3 c-Fos Error
Degr. Of Freedom
F
p
1 1 1 14 1 1 1 14
0.00 8.01 1.69
0.986 0.013 0.215
1.09 6.07 4.83
0.315 0.027 0.045
between functional performance on PAT and neurohistological markers). CA3 c-Fos inversely predicted time to entry into the shock zone in the flipped condition (r = -0.69, N = 11, p = 0.017) whereas in the unflipped condition this effect was not significant [(r = 0.07, N = 7, p = 0.86; condition* CA3 c-Fos interactive effect Wald statistic (df = 1) = 5.03, p = 0.02, 95%ile lower CI = 0.0023, higher CI = 0.035] (see Table 5a). For DG DCX, a direct relationship to second time to enter was observed in the flipped condition (r = 0.61, N = 11, p = 0.046) whereas in the unflipped condition there was no relationship [(r = -0.26, N = 7, p = 0.56); condition* DG DCX interactive effect – Wald Statistic (df = 1) = 4.95, p = 0.03, 95%ile lower CI = -0.0014, higher CI = -0.00009] (See Table 5b). Of note, DCX neurons with or without tertiary dendrites did not predict PAT performance (see Appendix Table 2). DG BrdU without NeuN or BrdU c-Fos was not predictive of PAT outcomes. Thus, only CA3 c-Fos and DG DCX total were predictive of cognitive outcomes specifically in the flipped condition. The directionality of these effects is consistent with our hypotheses – both high DG DCX and low CA3 c-Fos were predictive of better cognitive outcomes specifically in the flipped condition (Figs. 4 and 5). 5. Discussion This study examined the effect of individual variation in the number of developmentally distinct adult-born populations of new neurons DCX compared to BrdU-NeuN staining neurons - and influence on tasks invoking cognitive flexibility and recent spatial memory. Unexpectedly, the proliferation of especially immature DCX neurons with only primary and secondary but not tertiary dendrites were inversely correlated with CA3 activity in the condition implicating cognitive flexibility (flipped condition) but not the condition implicating recent spatial memory (unflipped condition). Although we expected DCX neurons to be linked to decreased CA3 activation and cognitive flexibility, we did not anticipate that DCX neurons with primary and secondary dendrites 5
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Fig. 3. The Relationship between Doublecortin-Stained Neurons with or without Tertiary Dendrites and CA3 c-Fos as a function of Experimental Condition. Legend: Post hoc-testing of the flipped condition revealed an effect for maturation of new neurons*c-Fos interactive effect indicating that it was the DCX neurons with primary or secondary dendrites that correlated inversely with CA3, rather than the DCX stained new neurons with tertiary dendrites in the flipped condition. Moreover when examining DCX neurons without tertiary dendrites, it was only in the flipped condition that there was a relation with CA3 c-Fos.
conflictive learning as studies using more precise, optogenetics techniques have done [17]. The time-course of our study also bears mention. As previously stated, because of the length of our study, the oldest new neurons were ≥ 5-weeks old. Thus, we were unable to compare the functioning of DCX neurons and still relatively immature BrdU-NeuN neurons against truly mature new DG neurons. The lack of supplementary morphological data must also be acknowledged. The use of only BrdU-NeuN and
explanation remains speculative. Our findings should be interpreted with caution in light of several limitations. First, we did not use a naïve group of animals that followed the same experimental paradigm but stayed in their home cage on Day 4. Such a design would have allowed us to compare correlations without assuming a putative hippocampal mechanism serving cognitive flexibility. Additionally, we did not examine the particular subpopulations of CA3 neurons (e.g. GABAergic or pyramidal) engaged during 6
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Table 4 Relationship of Doublecortin-Stained Neurons with or without Tertiary Dendrites in Relationship to CA3 c-Fos. a. Repeated Measures (R1) is Immature DCX Neurons versus Mature DCX (tertiary dendrites)
R1*condition R1*CA3 c-Fos R1* condition * CA3 c-Fos Error
Degr. Of Freedom
F
p
1 1 1 11
5.84 2.04 6.32
0.03 0.18 0.03
b. Post-hoc: Only flipped condition Effect
Degr. Of Freedom
F
p
c-Fos (CA) Error R1*c-Fos (CA) Error
1 6 1 6
9.61
0.021
9.11
0.023
Fig. 4. Low CA3 c-Fos predicts less errors on “Flipped” Active Avoidance Paradigm. Legend: There was a direct relationship between CA3 c-Fos and number of entries into the shock zone during flipped condition. However, this effect was not different from the unflipped condition.
c. Post-hoc Immature Neurons only
Condition CA3 c-Fos Condition * CA3 c-Fos Error
Degr. Of Freedom
F
1 1 1 11
2.51 4.22 4.88
0.142 0.064 0.049
Table 5 Generalized Linear/Non-Linear Models. a. Relationship between CA3 c-Fos and time to enter shock zone in the flipped versus unflipped condition
Condition CA3 c-Fos Condition*CA3 c-Fos
Degr. Of Freedom
Wald Stat
1 1 1
2.54 5.26 5.03
0.11 0.02 0.02
b. Relationship between DG DCX and Time 2 to Enter Shock Zone.
Condition DG DCX Condition*DG DCX
Degr. Of Freedom
Wald Stat
1 1 1
5.33 5.41 4.95
Fig. 5. Higher Rates of Dentate Gyrus Doublecortin Predict Delayed Time (Time 2) to Entrance to Shock Zone on the “Flipped” Active Avoidance Paradigm. Legend: For DG DCX, a direct relationship to second time to enter was observed in the flipped condition (r = 0.61, N = 11, p = 0.046) whereas in the unflipped condition there was no relationship [(r = -0.26, N = 7, p = 0.56); condition* DG DCX interactive effect – Wald Statistic (df = 1) = 4.95, p = 0.03, 95%ile lower CI = -0.0014, higher CI = -0.00009] (See Table 5b).
0.02 0.02 0.03
Significant results bolded.
DCX to distinguish developmental differences between DG new cell populations is potentially problematic. Though BrdU and NeuN are frequently used to mark mature adult born DG neurons, and DCX is used to mark immature new DG neurons, some studies find that these putative differentiators co-express in neurons ranging from 1 to 4 weeks old, suggesting that molecular expression and morphological features should be used to mark the developmental stage of DG neurons [37]. However, recent work highlights a clearer distinction between DCX and NeuN expression and suggests that these measures may accurately differentiate immature from mature new DG neurons [38]. Though no recorded new neurons co-expressed both NeuN and DCX in this study, the use of additional morphological features to more clearly delineate maturational phase would have strengthened this analysis. Our unexpected finding that immature new neurons with primary and secondary but not tertiary dendrites drive cognitive flexibility through decreased CA3 activation requires replication by larger studies using more advanced labeling and manipulation techniques. Such studies could shed additional light on the mechanisms by which such immature new neurons may contribute to cognitive flexibility.
6. Conclusions The present findings are distinct as they provide novel evidence that immature DG neurons without tertiary dendrites are specifically implicated in cognitive flexibility. In combination with decreased CA3 activation in the flipped PAT, the functional contribution of these immature DG neurons may involve the inhibition of postsynaptic CA3 neurons containing traces of previously salient conditioned contextual memories. Studies using more precise techniques to modulate specific cell types and label new neurons and their downstream CA3 targets are warranted before more definitive conclusions regarding the specific role of immature new neurons in cognitive flexibility can be made.
Declarations of interest None.
7
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
[14] N.S. Burghardt, et al., Adult-born hippocampal neurons promote cognitive flexibility in mice, Hippocampus 22 (9) (2012) 1795–1808. [15] N. Guo, et al., Dentate granule cell recruitment of feedforward inhibition governs engram maintenance and remote memory generalization, Nat. Med. (2018). [16] S. Ruediger, et al., Learning-related feedforward inhibitory connectivity growth required for memory precision, Nature 473 (7348) (2011) 514–518. [17] L. Restivo, et al., Development of adult-generated cell connectivity with excitatory and inhibitory cell populations in the Hippocampus, J. Neurosci. 35 (29) (2015) 10600–10612. [18] T. Herdegen, et al., Basal expression of the inducible transcription factors c‐Jun, JunB, JunD, c‐Fos, FosB, and Krox‐24 in the adult rat brain, J. Comp. Neurol. 354 (1) (1995) 39–56. [19] T.D. Perera, et al., Antidepressant-induced neurogenesis in the hippocampus of adult nonhuman primates, J. Neurosci. 27 (18) (2007) 4894–4901. [20] J.P. Brown, et al., Transient expression of doublecortin during adult neurogenesis, J. Comp. Neurol. 467 (1) (2003) 1–10. [21] K.K. Kim, R.S. Adelstein, S. Kawamoto, Identification of neuronal nuclei (NeuN) as Fox-3, a new member of the Fox-1 gene family of splicing factors, J. Biol. Chem. 284 (45) (2009) 31052–31061. [22] M.W. Miller, R. Nowakowski, Use of bromodeoxyuridine-immunohistochemistry to examine the proliferation, migration and time of origin of cells in the central nervous system, Brain Res. 457 (1) (1988) 44–52. [23] A.A. Swan, et al., Characterization of the role of adult neurogenesis in touch-screen discrimination learning, Hippocampus 24 (12) (2014) 1581–1591. [24] J.S. Snyder, et al., Adult-born hippocampal neurons are more numerous, faster maturing, and more involved in behavior in rats than in mice, J. Neurosci. 29 (46) (2009) 14484–14495. [25] R.L. Redondo, et al., Bidirectional switch of the valence associated with a hippocampal contextual memory engram, Nature 513 (7518) (2014) 426–430. [26] M.S. Fanselow, From contextual fear to a dynamic view of memory systems, Trends Cogn. Sci. 14 (1) (2010) 7–15. [27] A. Beyeler, et al., Divergent routing of positive and negative information from the Amygdala during memory retrieval, Neuron 90 (2) (2016) 348–361. [28] P. Namburi, et al., Architectural representation of Valence in the limbic system, Neuropsychopharmacology 41 (7) (2016) 1697–1715. [29] S. Ge, et al., A critical period for enhanced synaptic plasticity in newly generated neurons of the adult brain, Neuron 54 (4) (2007) 559–566. [30] L.A. Mongiat, et al., Reliable activation of immature neurons in the adult hippocampus, PLoS One 4 (4) (2009) e5320. [31] A. Marín-Burgin, et al., Unique processing during a period of high excitation/inhibition balance in adult-born neurons, Science 335 (6073) (2012) 1238–1242. [32] Y. Gu, et al., Optical controlling reveals time-dependent roles for adult-born dentate granule cells, Nat. Neurosci. 15 (12) (2012) 1700–1706. [33] C.E. Ribak, L. Seress, D.G. Amaral, The development, ultrastructure and synaptic connections of the mossy cells of the dentate gyrus, J. Neurocytol. 14 (5) (1985) 835–857. [34] C. Vivar, et al., Monosynaptic inputs to new neurons in the dentate gyrus, Nat. Commun. 3 (2012) 1107. [35] J.H. Chancey, et al., Hilar mossy cells provide the first glutamatergic synapses to adult-born dentate granule cells, J. Neurosci. 34 (6) (2014) 2349–2354. [36] N. Sah, et al., Running reorganizes the circuitry of one-week-old adult-born hippocampal neurons, Sci. Rep. 7 (1) (2017) 10903. [37] M.S. Espósito, et al., Neuronal differentiation in the adult hippocampus recapitulates embryonic development, J. Neurosci. 25 (44) (2005) 10074–10086. [38] M. Boldrini, et al., Human hippocampal neurogenesis persists throughout aging, Cell Stem Cell 22 (4) (2018) 589–599 e5.
Disclosures Dr. Coplan is a speaker for Pfizer, Forest, BMS, GSK, Eli Lilly and Sunovion. He has received grants from Pfizer Pharmaceuticals, GSK, Corcept and Neurocrine. He has served on the advisory board of Pfizer, Otsuka, Lundbeck and Corcept. Mr. Webler, Dr. Fulton, and Dr. Perera have no interests to disclose. Funding source This research was supported by SUNY Downstate Medical Center Dean’s Grant (J.D.C.) and KO8 MH70954 (T.D.P). Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.neulet.2019.134414. References [1] D.R. Dajani, L.Q. Uddin, Demystifying cognitive flexibility: implications for clinical and developmental neuroscience, Trends Neurosci. 38 (9) (2015) 571–578. [2] T.W. Robbins, Shifting and stopping: fronto-striatal substrates, neurochemical modulation and clinical implications, Philos. Trans. R. Soc. Lond., B, Biol. Sci. 362 (1481) (2007) 917–932. [3] T. Hedden, J.D. Gabrieli, Shared and selective neural correlates of inhibition, facilitation, and shifting processes during executive control, Neuroimage 51 (1) (2010) 421–431. [4] C. Anacker, R. Hen, Adult hippocampal neurogenesis and cognitive flexibility—linking memory and mood, Nat. Rev. Neurosci. 18 (6) (2017) 335. [5] J.K. Leutgeb, et al., Pattern separation in the dentate gyrus and CA3 of the hippocampus, science 315 (5814) (2007) 961–966. [6] T.J. McHugh, et al., Dentate gyrus NMDA receptors mediate rapid pattern separation in the hippocampal network, Science 317 (5834) (2007) 94–99. [7] L. Wiskott, M.J. Rasch, G. Kempermann, A functional hypothesis for adult hippocampal neurogenesis: avoidance of catastrophic interference in the dentate gyrus, Hippocampus 16 (3) (2006) 329–343. [8] A. Sahay, et al., Increasing adult hippocampal neurogenesis is sufficient to improve pattern separation, Nature 472 (7344) (2011) 466. [9] C. Clelland, et al., A functional role for adult hippocampal neurogenesis in spatial pattern separation, Science 325 (5937) (2009) 210–213. [10] J.B. Aimone, W. Deng, F.H. Gage, Resolving new memories: a critical look at the dentate gyrus, adult neurogenesis, and pattern separation, Neuron 70 (4) (2011) 589–596. [11] N.B. Danielson, et al., Distinct contribution of adult-born hippocampal granule cells to context encoding, Neuron 90 (1) (2016) 101–112. [12] L.J. Drew, et al., Activation of local inhibitory circuits in the dentate gyrus by adult‐born neurons, Hippocampus 26 (6) (2016) 763–778. [13] T.D. Perera, et al., Role of hippocampal neurogenesis in mnemonic segregation: implications for human mood disorders, World J. Biol. Psychiatry 14 (8) (2013) 602–610.
8
Contents lists available at ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
Research article
Maturational phase of hippocampal neurogenesis and cognitive flexibility a
b
c
Ryan D. Webler , Sasha Fulton , Tarique D. Perera , Jeremy D. Coplan
d,⁎
T
a
Department of Psychology, University of Minnesota, MN, United States Mount Sinai Medical Center, New York, NY, United States Contemporary Care LLC, United States d Department of Psychiatry & Behavioral Sciences, State University of New York, Downstate Medical Center, United States b c
A R T I C LE I N FO
A B S T R A C T
Keywords: Hippocampus Neurogenesis Cognitive flexibility BrdU-NeuN Doublecortin
Introduction: Pattern separation aids cognitive flexibility by reducing interference between closely related memories. Dentate gyrus (DG) neurogenesis may facilitate pattern separation by blocking memory retrieval via inhibition of non-neurogenic downstream CA3 neurons. We hypothesized that immature adult-born DG neurons would be associated with decreased CA3 activation and increased cognitive flexibility. Method: Two groups of adult male rats were tested either on the place avoidance task (PAT) (unflipped condition) or a subtly altered-PAT (flipped condition). Four weeks prior, the rats were injected with the mitotic marker BrdU. Immature new neurons were detected by the microtubule protein doublecortin (DCX). Cells that took up BrdU and expressed NeuN were identified as relatively more mature neurons. Synaptic activation was determined by c-Fos expression. Adaptation to the flipped versus unflipped condition reflected a measure of cognitive flexibility. Results: CA3 but not DG c-Fos was lower in the flipped versus unflipped condition [p = 0.002]. CA3 c-Fos correlated inversely with flipped task performance and immature (DCX) neurons with primary and secondary but not tertiary dendrites or more mature (BrdU + NeuN) new neurons. CA3 c-Fos was a significant predictor for the flipped versus unflipped condition specifically for DCX versus BrdU-NeuN neurons. Conclusion: Immature new neurons (DCX+) without tertiary dendrites may be preferentially implicated in cognitive flexibility relative to more mature new neurons (BrdU-NeuN). In combination with decreased CA3 activation in the flipped PAT, the functional contribution of these immature DG neurons may involve the inhibition of postsynaptic CA3 neurons containing traces of previously salient conditioned memories.
1. Introduction Adaptive behavior requires cognitive flexibility, the ability to efficiently shift between mental processes [1]. To date, areas of the prefrontal cortex have mostly been implicated in cognitive flexibility [2,3]. Recently however, the putative contribution of the hippocampus has been detailed [4]. The hippocampus plays a key role in pattern separation, the process of distinguishing closely related information [5,6]. Pattern separation serves cognitive flexibility by reducing interference between closely related information sets [7]. A developing literature suggests that newly derived, adult-born dentate gyrus (DG) neurons play a critical role in hippocampal mediated pattern separation [8–10]. These highly excitable neurons are more likely to activate to novel cues [11] and may
reduce interference between closely related information sets by dampening activation patterns corresponding to closely related past experiences, allowing for the encoding of novel information [12]. The developmental stage and mechanism by which new DG neurons interact with mature hippocampal neurons to form a substrate for cognitive flexibility is unclear. We have previously used the active place avoidance task (flipped PAT) to address these questions [13]. The flipped PAT, which ‘flips’ the shock zone 180 degrees from the initially learned position, is particularly sensitive to both hippocampal activation and neurogenesis [14]. In Perera et al. (2013), immature new neurons were identified using the microtubule doublecortin (DCX) and relatively more mature new neurons were identified using the mitotic marker bromodeoxyuridine (BrdU) and the neuronal marker NeuN. Our results suggested that the flipped PAT directly invokes immature new
⁎ Corresponding author at: Department of Psychiatry & Behavioral Sciences, SUNY Downstate Medical Center, 450 Clarkson Ave, Brooklyn, NY, 11203, United States. E-mail address: [email protected] (J.D. Coplan).
https://doi.org/10.1016/j.neulet.2019.134414 Received 20 August 2018; Received in revised form 11 July 2019; Accepted 5 August 2019 Available online 17 August 2019 0304-3940/ © 2019 Elsevier B.V. All rights reserved.
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
original task wherein a previously learned shock zone location was flipped 180 degrees (n = 11), and (2) an unflipped control condition (n = 7). The unflipped control condition implicated retrieval of the original shock location memory; the flipped condition putatively required suppression of this original memory and the encoding of a new shock location memory, invoking cognitive flexibility. The degree to which the animal can “adapt” to its new environment is a task putatively requiring cognitive flexibility, of which a plausible biological substrate should be identifiable.
neurons (DCX labeled) and that combined activation in the DG/CA3 regions may have components that are functionally implicated in cognitive flexibility. However, one important weakness of this study was that hippocampal activation (measured via c-Fos) was collapsed into a single measure combining the DG and CA3 regions. The combining of these distinct sub-regions obscured the significance of a negative correlation between immature DCX neurons and activation in a collapsed DG/CA3 region. Moreover, the correlation between the combined DG/ CA3 activation and a negatively predicted task performance on the flipped PAT was also ambiguous. The present analysis endeavored to identify the role of specific hippocampal sub-regions (DG and CA3) in cognitive flexibility and the developmental stage (immature DCX or relatively mature BrdU-NeuN) at which immature new neurons contribute to cognitive flexibility. In addition, we sought to identify the manner in which immature new DG neurons and CA3 activation interact to facilitate cognitive flexibility. Data from several studies shed light on interactions between these regions and their interactive functional significance. Specifically, these studies suggest that engram specificity, a pre-requisite for cognitive flexibility, requires DG feed-forward inhibition onto CA3 [15–17]. These findings necessitated an approach that distinguishes the DG and CA3 sub-regions. We now hypothesize that CA3 activation would correlate negatively with performance on the task invoking cognitive flexibility (flipped PAT) but not the task invoking recent spatial memory (unflipped PAT). Additionally, consistent with the feedforward inhibitory hypothesis and our previous work implicating DCX neurons in cognitive flexibility, we hypothesized that subjects with high counts of immature, DCX but not BrdU-NeuN marked new neurons would have reduced CA3 c-Fos in the flipped task requiring new memory formation and the putative blockade of a closely related memory (cognitive flexibility).
2.3. Neurogenesis and CA3 activation Cellular activity in the DG and CA3 regions was measured via expression of the early gene precursor, c-Fos [18]. c-Fos has a half-life of 2 h; rats were euthanized within 1 h of task completion on Day 4 to allow for task induced c-Fos activation. Neurogenesis was measured using previously published methods [19]. The microtubule protein double-cortin (DCX) stains DG neurons that have not yet completed the maturational point of tertiary dendrite formation (i.e. immature new neurons) [20]. The rats were injected with BrdU (100 mg/kg, i.p., for 3 days) 4 weeks prior to randomization. BrdU-NeuN co-labelled neurons were identified as relatively more mature new neurons [21,22]. These cells were assumed to be at least 5 weeks old given the timing of BrdU injection. Two independent raters (ICC > 0.90) who were blinded to animal identity, counted DCX expressing immature new neurons, c-Fos expressing synaptically activated mature neurons, and BrdU-NeuN co-labeled mature new neurons in every twentieth section (average of 15 per animal) of the left subgranular zone (SGZ) defined as a 50-mm band at the border of the granule cell layer and hilus. We calculated the density of DCX, BrdU-NeuN, and c-Fos labeled cells by dividing the total number of counted cells by the volume of SGZ in the counted sections (length of SGZ of section = 50 mm width = 40 mm depth/thickness before post-mounting). Of note, c-Fos was only counted in DG neurons that were BrdU negative as that would provide a counter-balanced relationship to CA3, where no new neurons expressing the mitotic marker were anticipated. Each neurohistological count was performed on separate hippocampal sections.
2. Materials and methods All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of SUNY Downstate Medical Center in compliance with NIH and national standards for the ethical treatment and use of vertebrate animals in research. These subjects were previously reported on in Perera et al. (2013). However, the relationship between immature new neurons as reflected by DCX-stained neurons in comparison to more mature new neurons reflected by BrdU-NeuN-stained neurons as predictors of CA3 c-Fos was not directly examined. Additionally, as previously described, hippocampal sub-regions (DG and CA3) were not properly disambiguated. In the current study, we examined, using general linear models which had not previously been employed, whether DCX-stained neurons vis-à-vis BrdU-NeuN-stained neurons were differentially predictive of CA3 c-Fos in the modified (flipped) versus unmodified (unflipped) condition of the PAT.
3. Statistical methods Data was analyzed using general linear modeling (Statistica 12.0) with post-hoc GLMs, generalized linear/non-linear models, and Pearson’s correlations. Due to technical difficulties, we have available 94% of study values and therefore we have had to omit samples, for instance in general linear models with repeated measures analyses. See Appendix Table 1 for total number of samples available for all study variables broken down by condition. For the first analysis, we used DG c-Fos and CA3 c-Fos as repeated measures, the flipped versus unflipped condition as the categorical variable, and individual DCX counts as the predictor variable. The GLM was then set up as a factorial ANOVA where an effect was provided for ROI c-Fos, for the flipped versus unflipped condition, and the interactive effect of ROI c-Fos* condition effect. When introducing DCX counts as a predictor variable, the factorial ANOVA now contained the triple interactive term – condition* ROI c-Fos * DG DCX interaction. This term would allow us to determine whether DG DCX would differentially predict CA3 c-Fos vis-à-vis DG c-Fos as a function of condition and ROI. Post-hoc generalized linear/non-linear models were separately performed for the flipped versus unflipped condition to examine for differential patterns. Finally, post-hoc Pearson’s correlations were performed contrasting DG DCX versus DG c-Fos to CA3 c-Fos in the flipped versus unflipped condition. For the next GLM DG DCX and DG BrdU-NeuN counts were used as repeated measures, condition (flipped versus unflipped) was entered as the categorical variable, and CA3 c-Fos was used as the predictor
2.1. Active place avoidance task In brief, we exposed Long-Evans male rats (N = 18) to the PAT, a paradigm that employs a continuously revolving circular arena to measure spatial segregation. In this task, rats avoid a stationary foot shock segment of the arena by distinguishing a salient, stationary spatial cue linked with a safe location, from a non-salient, rotating cue identified through deposition of rodent olfactory cues. [13]. Conditioned avoidance performance was assessed by: (1) time to first enter the shock-zone; (2) number of times entering into the shock zone; and 3) time to second entry into the shock zone. 2.2. Unflipped vs. Flipped modification of the PAT After undergoing training on the place avoidance task (PAT), rats were randomized to one of two conditions: (1) a ‘flipped’ version of the 2
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Table 1 General Linear Model Analysis of Dependent Repeated Measures Variables CA3 c-Fos and DG c-Fos in a Factorial ANOVA Design (see Fig. 1 and Results).
Group DCX Group*DCX Error R1 R1*Group R1*DCX R1*Group*DCX Error
DCX Error R1 R1*DCX total Error
Degr. Of Freedom
F
p
1 1 1 10 1 1 1 1 10
0.85 0.87 0.87
0.38 0.37 0.37
0.23 6.61 0.64 5.93
0.64 0.03 0.44 0.04
Degr. Of Freedom
F
p
1 8 1 1 8
3.21
0.111
18.94 15.95
0.002 0.004
Fig. 1. Immediate Gene c-Fos Expression in CA3 section of Hippocampus versus the Subgranular Zone (SGZ) of the Dentate Gyrus (DG) in relation to a Flipped versus Unflipped Version of the Place Avoidance Task. Legend: Whereas CA3 c-Fos was reduced in comparison to DG SGZ c-Fos in the flipped condition, there were no differences for c-Fos in the unflipped condition between CA3 and SGZ. See text and Table 1 for numerical results.
DG=Dentate Gyrus; Group = “flipped” versus “unflipped” condition. DCX=doublecortin-stained new neurons in the subgranular zone of the DG. Significant Results are Bolded. Post-hoc GLM: Flipped condition. No significant effects observed for unflipped condition.
revealed that, although the relationship between CA3 c-Fos and BrdUNeuN in the flipped condition was a trend using Pearson’s correlations, when examined non-parametrically, analyses revealed a significant inverse relationship (Spearman r = -0.89, N = 7, p = 0.007) whereas in the flipped condition there was no relationship between the variables (r = 0.22, N = 11, p = 0.50) leading to a significant condition*CA3 cFos interactive effect [Wald Statistic (1) = 4.60; p = 0.03; lower 95% CI = -0.01, upper 95% CI = 0.0005]. Moreover, in the flipped condition, the inverse relation between DG DCX and CA3 c-Fos differed significantly from the DG BrdU-NeuN to CA3 c-Fos relationship (the latter correlation: r = 0.22, N = 11, p = 0.50); F(1,9) = 16.46, p = 0.004]. An Omnibus GLM Analysis using DG DCX and DG BrdUNeun as the Repeated Measure, condition as the categorical variable and CA3 c-Fos as the independent predictor variable in a factorial ANOVA Design, revealed a triple interactive effect [SGZ Marker*condition*CA3 c-Fos interactive effect: F (114) = 4.83, p = 0.045]. In Appendix Fig. 1 we examine the relationship between DG c-Fos and DG DCX compared to the DG c-Fos and DG BrdU-NeuN relationship in the flipped versus non-flipped condition. No significant relationships were noted for the four correlations examined. In summary, CA3 c-Fos was specifically reduced in the flipped condition, and the magnitude of CA3 c-Fos reductions were inversely related to DCX counts in the DG. The latter inverse relation described was specific only to the flipped condition. By contrast, there was an inverse relationship between CA3 c-Fos and DG BrdU-NeuN only in the unflipped condition and that relationship was shown to be specific to the unflipped condition (Table 3).
variable. Variables were formatted into a factorial design so that the triple interactive effect could be evaluated. Finally, the relationship between CA3 c-Fos, DG c-Fos, DG DCX, and DG BrdU-NeuN were examined in relationship to the cognitive performance outcomes in the flipped versus unflipped condition. Because of the unusual distribution seen with the behavioral outcomes (e.g. time to enter cannot be greater than 600 s), a generalized linear/non-linear model was employed, in addition to Pearson’s correlations, to assess the relationships between neurohistological measures and cognitive performance. 4. Results 4.1. CA3 c-Fos as a function of flipped versus unflipped version of the PAT task In a model using the dependent regions-of-interest measures CA3 cFos and DG c-Fos in a factorial design with condition as a categorical variable (flipped versus unflipped) and continuous predictor variable, doublecortin stained new neurons of the subgranular zone of the dental gyrus, there was a main GLM effect for a condition*ROI interactive effect [F(110) = 6.60, p = 0.027] and a condition*ROI*Doublecortin triple interactive effect. The former result is illustrated by the observation that in a post-hoc GLM, in comparison to the “unflipped” condition where no ROI differences were noted, in a post-hoc GLM for the “flipped” condition, CA3 c-Fos was reduced in comparison to DG cFos [F(1,8) = 18.94, p = 0.002] (see Fig. 1, Table 1). As CA3 c-Fos and DG c-Fos were not standardized, using a post-hoc generalized linear/ non-linear model in a factorial design (Appendix Table 3), there was a condition effect [Wald Statistic (1) = 5.55; p = 0.02] indicating reduced CA3 c-Fos in the flipped versus unflipped condition [Lower CL 95.0% = -3.11; Upper CL 95.0% = -0.28]. By contrast, using the identical model, (see Appendix Table 3B) c-Fos of the dentate gyrus showed no significant differences as a function of condition. In the Fig. 2 scatterplot, inspection of the top two panels demonstrated that CA3 c-Fos was inversely related to DG DCX in the flipped condition (r= -0.79; N = 11, p = 0.003) whereas the same relationship in the unflipped condition was a direct non-significant correlation [(r= -0.22, N = 7, p = 0.62); [Condition * CA3 c-Fos interactive effect [Wald Statistic (1) = 5.49; p = 0.02; lower 95% CI = 0.0006, upper 95% CI = 0.007] (Table 2a, Fig. 2). Analysis of the lower two panels
4.2. Doublecortin-Stained Neurons with or without tertiary dendrites The next analysis addressed whether it was the more mature form of the DCX expressing neuron exhibiting tertiary dendrites in comparison to the more immature form of DCX-expressing neurons that only stained for primary and secondary dendrites. Using a GLM an overall condition* immature vs relatively more mature new neurons* CA3 c-Fos interaction was noted [F(111) = 6.32, p = 0.03] (See Fig. 3, Table 4). Post hoc-testing of the flipped condition revealed an effect for maturation of new neurons*c-Fos interactive effect indicating that it was the DCX neurons with primary or secondary dendrites that correlated inversely with CA3, rather than the DCX stained new neurons with tertiary dendrites [F(1,6) = 9.11, p = 0.023] (see Fig. 3). Moreover, when examining DCX neurons without tertiary dendrites, it was only in 3
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Fig. 2. Relationship of Maturational Phase of Adult Born Dentate Gyrus New Neurons to CA3 c-Fos. Legend: CA3 c-Fos was inversely related to DG DCX c-Fos in the flipped condition whereas the same relationship in the unflipped condition was a direct non-significant correlation. Although the relationship between CA3 c-Fos and BrdU-NeuN in the flipped condition was a trend, when examined non-parametrically it revealed a significant inverse relationship whereas in the flipped condition there was no relationship between the variables leading to a significant condition*CA3 interactive effect. See text and Table 2 for numerical results.
the flipped condition that there was a relation with CA3 c-Fos [F(111) = 4.88, p = 0.049]. These post-hoc results were replicated using a generalized linear/non-linear model.
4.3. Functional performance Pearson’s correlations revealed a direct relationship between CA3 cFos and number of entries into the shock zone during flipped condition (r = 0.61, N = 11, p = 0.046). However, this effect was not different from the unflipped condition (see Appendix Table 2 for all correlations 4
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
would differentially contribute to cognitive flexibility compared to those with tertiary dendrites. As hypothesized, both the individual variation of observed differences in CA3 activity and the proliferation of immature new DG neurons had functional implications on cognitive flexibility vis-à-vis task performance. The first measure of task performance, number of shock zone entries, was directly related to CA3 c-Fos in the flipped condition. A second measure, time to first entry, was inversely predicted by CA3 cFos in the flipped but not unflipped condition. The third measure, time to second entry, was positively predicted by total DG DCX (not simply DCX neurons with primary and secondary dendrites as noted in above findings) only in the flipped condition. Of the three significant analyses of neurohistological measures and task performance, the latter two were statistically distinguishable from the unflipped condition. No relationships in the unflipped condition were significant. Thus, low CA3 c-Fos and high DG DCX were predictive of outcomes in the cognitive flexibility condition (flipped task) but not the control condition (unflipped task). Because of the timing of our staining procedure, even the most mature BrdU-NeuN co-labelled new neurons were presumed to be at most 5 weeks of age. The functional dissociation that emerged between DCX and BrdU-NeuN neurons is thus especially striking given the narrow developmental window for transitioning neurohistological marker characterization in populations of new neurons in this sample. Dendritic findings lend additional support to these findings, suggesting that the most developmentally immature new DG neurons may contribute to cognitive flexibility. Previous work by Swan et al. (2014) implicates new neurons over one month old in cognitive flexibility [23]. While Swan and colleagues used mice, the current study used Long-Evans rats. There is data to suggest that new rat neurons mature more quickly than new mice DG neurons, outpacing their development by one or more weeks [24]. Thus, cross-species differences in the rate of new DG cell development may partially explain why new DG neurons older than one month were implicated in Swan et al. (2014) but not in the current study. Additionally, differences in task valence – Swan et al. (2014) used reward learning while the current report used aversive learning – need to be considered. Though DG engrams may be valence independent [25,26], divergent hippocampal afferents instantiate positive and negative value [27,28]. It is unclear whether these afferents differentially implicate immature DG neurons across various stages of development. The vast majority of new rat DG neurons express NeuN by 4 weeks [20,24]. New DG neurons between 4–6 weeks old have properties that confer increased plasticity, including decreased inhibition and higher plasticity in input and output synapses [17,29–32]. Thus, plasticity differences between immature DCX neurons and relatively more mature ≥ 5-week old BrdU-NeuN neurons are counterintuitive. However, we note that for this form of cognitive flexibility (measured via flipped PAT performance), it is the more immature stage of new DCX neurons that accounts for the effect. It is argued – discussed below—that a plausible interaction with mossy cells in the hilar area may account for the early forms of neuroplasticity observed in the current study. Mossy cells are large hilar neurons with a prominent commissural projection from ipsilateral hilar area to the contralateral DG and form the major source of afferents to the contralateral inner molecular layer [33]. Adult born DG cells receive monosynaptic input from a number of cells, including mossy cells [34]. Mossy cells provide new DG neurons with their first glutamatergic and disynaptic GABAergic inputs [35]. The mossy cells also have the capacity, via GABAergic interneurons, to synapse on primary and secondary dendrites of adult born DG cells [36]. Therefore, one mechanism by which immature new neurons with only primary and secondary dendrites may have modulated CA3 activation is through interactions with mossy cells. The ability of these latter cells to provide functionally relevant information in the flipped version of the PAT to the contralateral DG has to be considered. However, because the current study lacks mossy cell data, this
Table 2 Relationship of CA3 c-Fos to Subgranular Zone Neuronal Maturational Phase Markers: Doublecortin versus BrdU-NeuN Generalized Linear/Nonlinear Model. a. Dependent Variable : Total Doublecortin – Top Panel Effect
Degr. Of Freedom
Wald Stat
p
Condition CA3 c-Fos Condition*CA3 c-Fos
1 1 1
0.83 11.75 5.49
0.3630 0.0006 0.0191
b. Dependent Variable: BrdU-NeuN – Bottom Panel Effect
Degr. Of Freedom
Wald Stat
p
Condition CA3 c-Fos Condition*CA3 c-Fos
1 1 1
6.48 1.04 4.60
0.011 0.307 0.032
Table 3 General Linear Model Analysis Using DG DCX and DG BrdU-Neun as the Repeated Measures, Condition as the Categorical Variable and CA3 C-Fos as the Independent Predictor Varaible in a Factorial ANOVA Design.
Group CA3 c-Fos Group*CA3 c-Fos Error R1*Group R1*CA3 c-Fos R1*Group*CA3 c-Fos Error
Degr. Of Freedom
F
p
1 1 1 14 1 1 1 14
0.00 8.01 1.69
0.986 0.013 0.215
1.09 6.07 4.83
0.315 0.027 0.045
between functional performance on PAT and neurohistological markers). CA3 c-Fos inversely predicted time to entry into the shock zone in the flipped condition (r = -0.69, N = 11, p = 0.017) whereas in the unflipped condition this effect was not significant [(r = 0.07, N = 7, p = 0.86; condition* CA3 c-Fos interactive effect Wald statistic (df = 1) = 5.03, p = 0.02, 95%ile lower CI = 0.0023, higher CI = 0.035] (see Table 5a). For DG DCX, a direct relationship to second time to enter was observed in the flipped condition (r = 0.61, N = 11, p = 0.046) whereas in the unflipped condition there was no relationship [(r = -0.26, N = 7, p = 0.56); condition* DG DCX interactive effect – Wald Statistic (df = 1) = 4.95, p = 0.03, 95%ile lower CI = -0.0014, higher CI = -0.00009] (See Table 5b). Of note, DCX neurons with or without tertiary dendrites did not predict PAT performance (see Appendix Table 2). DG BrdU without NeuN or BrdU c-Fos was not predictive of PAT outcomes. Thus, only CA3 c-Fos and DG DCX total were predictive of cognitive outcomes specifically in the flipped condition. The directionality of these effects is consistent with our hypotheses – both high DG DCX and low CA3 c-Fos were predictive of better cognitive outcomes specifically in the flipped condition (Figs. 4 and 5). 5. Discussion This study examined the effect of individual variation in the number of developmentally distinct adult-born populations of new neurons DCX compared to BrdU-NeuN staining neurons - and influence on tasks invoking cognitive flexibility and recent spatial memory. Unexpectedly, the proliferation of especially immature DCX neurons with only primary and secondary but not tertiary dendrites were inversely correlated with CA3 activity in the condition implicating cognitive flexibility (flipped condition) but not the condition implicating recent spatial memory (unflipped condition). Although we expected DCX neurons to be linked to decreased CA3 activation and cognitive flexibility, we did not anticipate that DCX neurons with primary and secondary dendrites 5
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Fig. 3. The Relationship between Doublecortin-Stained Neurons with or without Tertiary Dendrites and CA3 c-Fos as a function of Experimental Condition. Legend: Post hoc-testing of the flipped condition revealed an effect for maturation of new neurons*c-Fos interactive effect indicating that it was the DCX neurons with primary or secondary dendrites that correlated inversely with CA3, rather than the DCX stained new neurons with tertiary dendrites in the flipped condition. Moreover when examining DCX neurons without tertiary dendrites, it was only in the flipped condition that there was a relation with CA3 c-Fos.
conflictive learning as studies using more precise, optogenetics techniques have done [17]. The time-course of our study also bears mention. As previously stated, because of the length of our study, the oldest new neurons were ≥ 5-weeks old. Thus, we were unable to compare the functioning of DCX neurons and still relatively immature BrdU-NeuN neurons against truly mature new DG neurons. The lack of supplementary morphological data must also be acknowledged. The use of only BrdU-NeuN and
explanation remains speculative. Our findings should be interpreted with caution in light of several limitations. First, we did not use a naïve group of animals that followed the same experimental paradigm but stayed in their home cage on Day 4. Such a design would have allowed us to compare correlations without assuming a putative hippocampal mechanism serving cognitive flexibility. Additionally, we did not examine the particular subpopulations of CA3 neurons (e.g. GABAergic or pyramidal) engaged during 6
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
Table 4 Relationship of Doublecortin-Stained Neurons with or without Tertiary Dendrites in Relationship to CA3 c-Fos. a. Repeated Measures (R1) is Immature DCX Neurons versus Mature DCX (tertiary dendrites)
R1*condition R1*CA3 c-Fos R1* condition * CA3 c-Fos Error
Degr. Of Freedom
F
p
1 1 1 11
5.84 2.04 6.32
0.03 0.18 0.03
b. Post-hoc: Only flipped condition Effect
Degr. Of Freedom
F
p
c-Fos (CA) Error R1*c-Fos (CA) Error
1 6 1 6
9.61
0.021
9.11
0.023
Fig. 4. Low CA3 c-Fos predicts less errors on “Flipped” Active Avoidance Paradigm. Legend: There was a direct relationship between CA3 c-Fos and number of entries into the shock zone during flipped condition. However, this effect was not different from the unflipped condition.
c. Post-hoc Immature Neurons only
Condition CA3 c-Fos Condition * CA3 c-Fos Error
Degr. Of Freedom
F
1 1 1 11
2.51 4.22 4.88
0.142 0.064 0.049
Table 5 Generalized Linear/Non-Linear Models. a. Relationship between CA3 c-Fos and time to enter shock zone in the flipped versus unflipped condition
Condition CA3 c-Fos Condition*CA3 c-Fos
Degr. Of Freedom
Wald Stat
1 1 1
2.54 5.26 5.03
0.11 0.02 0.02
b. Relationship between DG DCX and Time 2 to Enter Shock Zone.
Condition DG DCX Condition*DG DCX
Degr. Of Freedom
Wald Stat
1 1 1
5.33 5.41 4.95
Fig. 5. Higher Rates of Dentate Gyrus Doublecortin Predict Delayed Time (Time 2) to Entrance to Shock Zone on the “Flipped” Active Avoidance Paradigm. Legend: For DG DCX, a direct relationship to second time to enter was observed in the flipped condition (r = 0.61, N = 11, p = 0.046) whereas in the unflipped condition there was no relationship [(r = -0.26, N = 7, p = 0.56); condition* DG DCX interactive effect – Wald Statistic (df = 1) = 4.95, p = 0.03, 95%ile lower CI = -0.0014, higher CI = -0.00009] (See Table 5b).
0.02 0.02 0.03
Significant results bolded.
DCX to distinguish developmental differences between DG new cell populations is potentially problematic. Though BrdU and NeuN are frequently used to mark mature adult born DG neurons, and DCX is used to mark immature new DG neurons, some studies find that these putative differentiators co-express in neurons ranging from 1 to 4 weeks old, suggesting that molecular expression and morphological features should be used to mark the developmental stage of DG neurons [37]. However, recent work highlights a clearer distinction between DCX and NeuN expression and suggests that these measures may accurately differentiate immature from mature new DG neurons [38]. Though no recorded new neurons co-expressed both NeuN and DCX in this study, the use of additional morphological features to more clearly delineate maturational phase would have strengthened this analysis. Our unexpected finding that immature new neurons with primary and secondary but not tertiary dendrites drive cognitive flexibility through decreased CA3 activation requires replication by larger studies using more advanced labeling and manipulation techniques. Such studies could shed additional light on the mechanisms by which such immature new neurons may contribute to cognitive flexibility.
6. Conclusions The present findings are distinct as they provide novel evidence that immature DG neurons without tertiary dendrites are specifically implicated in cognitive flexibility. In combination with decreased CA3 activation in the flipped PAT, the functional contribution of these immature DG neurons may involve the inhibition of postsynaptic CA3 neurons containing traces of previously salient conditioned contextual memories. Studies using more precise techniques to modulate specific cell types and label new neurons and their downstream CA3 targets are warranted before more definitive conclusions regarding the specific role of immature new neurons in cognitive flexibility can be made.
Declarations of interest None.
7
Neuroscience Letters 711 (2019) 134414
R.D. Webler, et al.
[14] N.S. Burghardt, et al., Adult-born hippocampal neurons promote cognitive flexibility in mice, Hippocampus 22 (9) (2012) 1795–1808. [15] N. Guo, et al., Dentate granule cell recruitment of feedforward inhibition governs engram maintenance and remote memory generalization, Nat. Med. (2018). [16] S. Ruediger, et al., Learning-related feedforward inhibitory connectivity growth required for memory precision, Nature 473 (7348) (2011) 514–518. [17] L. Restivo, et al., Development of adult-generated cell connectivity with excitatory and inhibitory cell populations in the Hippocampus, J. Neurosci. 35 (29) (2015) 10600–10612. [18] T. Herdegen, et al., Basal expression of the inducible transcription factors c‐Jun, JunB, JunD, c‐Fos, FosB, and Krox‐24 in the adult rat brain, J. Comp. Neurol. 354 (1) (1995) 39–56. [19] T.D. Perera, et al., Antidepressant-induced neurogenesis in the hippocampus of adult nonhuman primates, J. Neurosci. 27 (18) (2007) 4894–4901. [20] J.P. Brown, et al., Transient expression of doublecortin during adult neurogenesis, J. Comp. Neurol. 467 (1) (2003) 1–10. [21] K.K. Kim, R.S. Adelstein, S. Kawamoto, Identification of neuronal nuclei (NeuN) as Fox-3, a new member of the Fox-1 gene family of splicing factors, J. Biol. Chem. 284 (45) (2009) 31052–31061. [22] M.W. Miller, R. Nowakowski, Use of bromodeoxyuridine-immunohistochemistry to examine the proliferation, migration and time of origin of cells in the central nervous system, Brain Res. 457 (1) (1988) 44–52. [23] A.A. Swan, et al., Characterization of the role of adult neurogenesis in touch-screen discrimination learning, Hippocampus 24 (12) (2014) 1581–1591. [24] J.S. Snyder, et al., Adult-born hippocampal neurons are more numerous, faster maturing, and more involved in behavior in rats than in mice, J. Neurosci. 29 (46) (2009) 14484–14495. [25] R.L. Redondo, et al., Bidirectional switch of the valence associated with a hippocampal contextual memory engram, Nature 513 (7518) (2014) 426–430. [26] M.S. Fanselow, From contextual fear to a dynamic view of memory systems, Trends Cogn. Sci. 14 (1) (2010) 7–15. [27] A. Beyeler, et al., Divergent routing of positive and negative information from the Amygdala during memory retrieval, Neuron 90 (2) (2016) 348–361. [28] P. Namburi, et al., Architectural representation of Valence in the limbic system, Neuropsychopharmacology 41 (7) (2016) 1697–1715. [29] S. Ge, et al., A critical period for enhanced synaptic plasticity in newly generated neurons of the adult brain, Neuron 54 (4) (2007) 559–566. [30] L.A. Mongiat, et al., Reliable activation of immature neurons in the adult hippocampus, PLoS One 4 (4) (2009) e5320. [31] A. Marín-Burgin, et al., Unique processing during a period of high excitation/inhibition balance in adult-born neurons, Science 335 (6073) (2012) 1238–1242. [32] Y. Gu, et al., Optical controlling reveals time-dependent roles for adult-born dentate granule cells, Nat. Neurosci. 15 (12) (2012) 1700–1706. [33] C.E. Ribak, L. Seress, D.G. Amaral, The development, ultrastructure and synaptic connections of the mossy cells of the dentate gyrus, J. Neurocytol. 14 (5) (1985) 835–857. [34] C. Vivar, et al., Monosynaptic inputs to new neurons in the dentate gyrus, Nat. Commun. 3 (2012) 1107. [35] J.H. Chancey, et al., Hilar mossy cells provide the first glutamatergic synapses to adult-born dentate granule cells, J. Neurosci. 34 (6) (2014) 2349–2354. [36] N. Sah, et al., Running reorganizes the circuitry of one-week-old adult-born hippocampal neurons, Sci. Rep. 7 (1) (2017) 10903. [37] M.S. Espósito, et al., Neuronal differentiation in the adult hippocampus recapitulates embryonic development, J. Neurosci. 25 (44) (2005) 10074–10086. [38] M. Boldrini, et al., Human hippocampal neurogenesis persists throughout aging, Cell Stem Cell 22 (4) (2018) 589–599 e5.
Disclosures Dr. Coplan is a speaker for Pfizer, Forest, BMS, GSK, Eli Lilly and Sunovion. He has received grants from Pfizer Pharmaceuticals, GSK, Corcept and Neurocrine. He has served on the advisory board of Pfizer, Otsuka, Lundbeck and Corcept. Mr. Webler, Dr. Fulton, and Dr. Perera have no interests to disclose. Funding source This research was supported by SUNY Downstate Medical Center Dean’s Grant (J.D.C.) and KO8 MH70954 (T.D.P). Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.neulet.2019.134414. References [1] D.R. Dajani, L.Q. Uddin, Demystifying cognitive flexibility: implications for clinical and developmental neuroscience, Trends Neurosci. 38 (9) (2015) 571–578. [2] T.W. Robbins, Shifting and stopping: fronto-striatal substrates, neurochemical modulation and clinical implications, Philos. Trans. R. Soc. Lond., B, Biol. Sci. 362 (1481) (2007) 917–932. [3] T. Hedden, J.D. Gabrieli, Shared and selective neural correlates of inhibition, facilitation, and shifting processes during executive control, Neuroimage 51 (1) (2010) 421–431. [4] C. Anacker, R. Hen, Adult hippocampal neurogenesis and cognitive flexibility—linking memory and mood, Nat. Rev. Neurosci. 18 (6) (2017) 335. [5] J.K. Leutgeb, et al., Pattern separation in the dentate gyrus and CA3 of the hippocampus, science 315 (5814) (2007) 961–966. [6] T.J. McHugh, et al., Dentate gyrus NMDA receptors mediate rapid pattern separation in the hippocampal network, Science 317 (5834) (2007) 94–99. [7] L. Wiskott, M.J. Rasch, G. Kempermann, A functional hypothesis for adult hippocampal neurogenesis: avoidance of catastrophic interference in the dentate gyrus, Hippocampus 16 (3) (2006) 329–343. [8] A. Sahay, et al., Increasing adult hippocampal neurogenesis is sufficient to improve pattern separation, Nature 472 (7344) (2011) 466. [9] C. Clelland, et al., A functional role for adult hippocampal neurogenesis in spatial pattern separation, Science 325 (5937) (2009) 210–213. [10] J.B. Aimone, W. Deng, F.H. Gage, Resolving new memories: a critical look at the dentate gyrus, adult neurogenesis, and pattern separation, Neuron 70 (4) (2011) 589–596. [11] N.B. Danielson, et al., Distinct contribution of adult-born hippocampal granule cells to context encoding, Neuron 90 (1) (2016) 101–112. [12] L.J. Drew, et al., Activation of local inhibitory circuits in the dentate gyrus by adult‐born neurons, Hippocampus 26 (6) (2016) 763–778. [13] T.D. Perera, et al., Role of hippocampal neurogenesis in mnemonic segregation: implications for human mood disorders, World J. Biol. Psychiatry 14 (8) (2013) 602–610.
8