1. Neuroscience
Download icon

Opposite regulation of inhibition by adult-born granule cells during implicit versus explicit olfactory learning

  1. Nathalie Mandairon  Is a corresponding author
  2. Nicola Kuczewski
  3. Florence Kermen
  4. Jérémy Forest
  5. Maellie Midroit
  6. Marion Richard
  7. Marc Thevenet
  8. Joelle Sacquet
  9. Christiane Linster
  10. Anne Didier
  1. CNRS UMR 5292, INSERM U1028, Université de Lyon, France
  2. Cornell University, United States
Research Article
  • Cited 5
  • Views 749
  • Annotations
Cite this article as: eLife 2018;7:e34976 doi: 10.7554/eLife.34976

Abstract

Both passive exposure and active learning through reinforcement enhance fine sensory discrimination abilities. In the olfactory system, this enhancement is thought to occur partially through the integration of adult-born inhibitory interneurons resulting in a refinement of the representation of overlapping odorants. Here, we identify in mice a novel and unexpected dissociation between passive and active learning at the level of adult-born granule cells. Specifically, while both passive and active learning processes augment neurogenesis, adult-born cells differ in their morphology, functional coupling and thus their impact on olfactory bulb output. Morphological analysis, optogenetic stimulation of adult-born neurons and mitral cell recordings revealed that passive learning induces increased inhibitory action by adult-born neurons, probably resulting in more sparse and thus less overlapping odor representations. Conversely, after active learning inhibitory action is found to be diminished due to reduced connectivity. In this case, strengthened odor response might underlie enhanced discriminability.

https://doi.org/10.7554/eLife.34976.001

Introduction

Brain representations of the environment constantly evolve through learning mediated by different plasticity mechanisms. Several lines of evidence suggested that adult neurogenesis is a plasticity mechanism mediating changes in representations of sensory information (Lledo and Valley, 2016). Experience-dependent survival of adult-born neurons has received recent attention as a mechanism to modulate pattern separation and perceptual discrimination in the hippocampus and olfactory bulb (OB) (Sahay et al., 2011). In the OB, current hypotheses about the mechanism underlying this enhanced discrimination capability focus on increased inhibitory processes mediated by GABAergic adult-born neurons (Moreno et al., 2009; Alonso et al., 2012), with more integrating cells delivering more inhibition, leading to sparser odor representations. The data presented here challenge the current hypothesis by showing that the same rate of adult-born cell survival in the OB can enhance fine olfactory discrimination by either increasing or decreasing OB output sparseness. In the olfactory system, both passive (implicit perceptual learning in response to repeated exposure) and active (explicit associative learning in response to reinforcement) learning can improve discrimination between similar odorants (Mandairon et al., 2006a; Moreno et al., 2009). Both forms of learning have been shown to modulate neural activity in the OB (Buonviso et al., 1998; Kay and Laurent, 1999; Doucette et al., 2011), and to increase survival of inhibitory adult-born interneurons (Moreno et al., 2009; Sultan et al., 2010; Mandairon et al., 2011; Sultan et al., 2011). However, exactly how adult-born neurons shape the output of the OB to support enhanced discrimination in these two paradigms is currently unknown. To address this question, we focused on how the number of integrated adult-born cells and/or the synaptic contacts established by adult-born neurons in the network could account for experience-induced changes in perception (Moreno et al., 2009; Daroles et al., 2016).

The data reported here, combining morphological analysis and optogenetic stimulation of adult-born neurons with mitral cells activity recording, indicate that both forms of learning similarly enhanced the number of adult-born neurons formed in the OB. However, these adult-born neurons showed higher density of dendritic spines compared to controls in implicit learning only, leading to stronger inhibition of mitral cells. In contrast, in explicit learning, adult-born neurons formed weaker connections to mitral cells, resulting in disinhibition of mitral cells in learning conditions compared to control. These experiments indicate that a same number of adult-born inhibitory neurons are able to positively or negatively modulate inhibition in the OB, depending on the synaptic integration mode and dictated by the behavioral significance of the odorant.

Results

Implicit learning increased adult-born granule cell survival and spine density

To induce implicit learning, animals were exposed to two odorants, which are not spontaneously discriminated: +limonene (+lim) and - limonene (-lim). These were placed in two tea balls in the home cage for one hour per day over 10 days as previously described (Moreno et al., 2009, 2012) (Enriched animals; Figure 1A). Controls were exposed to empty tea balls (Non-enriched animals). We used a habituation/cross habituation test to assess the effect of the enrichment on discrimination between +lim and –lim (see methods) (Moreno et al., 2009, 2012). Significant habituation was observed in both the control (n = 20; Friedman test p<0.0001) and enriched groups (n = 25; Friedman test p<0.0001). Enriched animals discriminated +lim and -lim at the end of the enrichment procedure as evidenced by increased investigation times between the last habituation trial (OHab4) and the test trial (OTest) (Wilcoxon test for discrimination p=0.004) whereas the controls did not (Wilcoxon test for discrimination p=0.74) (Figure 1B and Figure 1—source data 1).

We used Bromodeoxy-uridine (BrdU) incorporation to quantify adult-born neuron survival and lentiviral transduction to follow their morphological integration into the network (GFP encoding lentivirus was injected 8 days before enrichment to allow for adult-born neurons to migrate from the subventricular zone to the OB; Figure 1A). As expected, implicit learning increased adult-born granule cell survival (p=0.042, parametric Anova followed by Bonferroni post hoc test; Non-Enriched n = 6 and Enriched n = 4, see methods for global statistics strategy), (Table 1; Figure 1C and Figure 1—source data 1; Figure 1—figure supplement 1A) as well as their responsiveness to the learned odors as assessed by BrdU/Zif268 co-expression (p=0.0078, parametric Anova followed by Bonferroni post hoc tests; Non-Enriched n = 7 and Enriched n = 6), (Table 1; Figure 1D and Figure 1—source data 1; Figure 1—figure supplement 1B) (Moreno et al., 2009). Spine density of labeled adult-born neurons was analyzed on the apical (site of interactions with mitral/tufted (M/T) cells) and the basal (site of input of centrifugal fibers) dendrites of adult-born granule cells (Figure 1E). Spine density increased after implicit learning on both the apical (p=0.0003, Kruskal-Wallis Anova followed by FDR-corrected permutation tests; Non-Enriched: 48 segments, four mice and Enriched: 30 segments, four mice) (Table 1; Figure 1F and Figure 1—source data 1) and basal dendritic domains (p=0.0015, Kruskal-Wallis Anova followed by FDR-corrected permutation tests; Non-Enriched: 59 segments, seven mice and Enriched: 48 segments, seven mice), (Table 1; Figure 1G and Figure 1—source data 1). In addition, 5 to 6 weeks after enrichment, once the enriched animals had forgotten the task, the density of BrdU-positive cells and the spine density on adult-born cells had returned to control values (Figure 1—figure supplement 2).

Table 1
Summary of statistical comparisons described in the text.

For normal data, Anova followed by parametric Bonferroni post hoc test were used. For data that did not reach normality, Kruskall-Wallis Anova followed by FDR-corrected permutation tests were used. *p<0.05; **p<0.001; ***p<0.0001 and =: not different

https://doi.org/10.7554/eLife.34976.002
Cond vs EnrPC vs Non-EnrPC vs EnrPC vs CondCond vs Non-EnrEnr vs Non-Enr
BrdU
Anova F(3,18)=6.63,
p=0.003
Cond = Enr

(p=0.99)
PC = Non-Enr

(p=0.99)
PC < Enr
*
(p=0.024)
PC < Cond
*
(p=0.025)
Cond > Non-Enr
*
(p=0.048)
Enr > Non-Enr
*
(p=0.042)
BrdU/Zif268
Anova
F(3,23)=9.25, p=0.0003
Cond = Enr

(p=0.99)
PC = Non-Enr

(p=0.99)
PC < Enr
*
(p=0.006)
PC < Cond
*
(p=0.005)
Cond > Non-Enr
*
(p=0.0073)
Enr > Non-Enr
*
(p=0.0078)
Tbx21/Zif268
Anova
F(3,15)=7.33, p=0.002
Cond > Enr
*
(p=0.015)
PC = Non-Enr

(p=0.053)
PC = Enr

(p=0.99)
PC < Cond
*
(p=0.044)
Cond = Non-Enr

(p=0.99)
Enr < Non-Enr
*
(p=0.02)
Spine density Apical
Kruskal-Wallis
H(3, N = 240)=44.55
p<0.0001
Cond < Enr
***
(p=0.0003)
PC = Non-Enr

(p=0.59)
PC < Enr
*
(p=0.001)
PC > Cond
*
(p=0.00255)
Cond < Non-Enr
*
(p=0.0132)
Enr > Non-Enr
**
(p=0.0003)
Spine density Basal
Kruskal-Wallis
H(3, N = 187)=20.15
p<0.0001
Cond = Enr
(p=0.36)
PC > Non-Enr **
(p=0.0006)
PC = Enr

(p=0.36)
PC = Cond

(p=0.09)
Cond > Non-Enr
*
(p=0.042)
Enr > Non-Enr
*
(p=0.0015)
sIPSCs
(Frequency)
Kruskal-Wallis
H(3, N = 177)=10.68
p=0.013
Cond = Enr

(p=0.55)
PC > Non-Enr
**
(p=0.0006)
PC = Enr

(p=0.13)
PC > Cond

(p=0.06)
Cond = Non-Enr

(p=0.13)
Enr > Non-Enr

(p=0.06)
sIPSCs
(Amplitude)
Kruskal-Wallis
H(3, N = 175)=21.30
p<0.0001
Cond < Enr
*
(p=0.0048)
PC = Non-Enr

(p=0.48)
PC = Enr

(p=0.7)
PC > Cond
*
(p=0.015)
Cond < Non-Enr
*
(p=0.018)
Enr = Non-Enr

(p=0.48)

Inhibition on mitral cells is increased after implicit learning

We then analyzed how the observed neurogenic changes affect OB output by investigating the overall level of inhibition of the M/T cells. Spontaneous inhibitory postsynaptic current (sIPSC) frequency recorded on M/T cells showed a trend toward increase after implicit learning (p=0.06, Kruskal-Wallis Anova followed by FDR-corrected permutation test, Enriched, n = 50 and Non-Enriched, n = 48) (for detailed statistics see Table 1), (Figure 1H and Figure 1—source data 1) with no change in amplitude (Table 1, Enriched n = 48 and Non-Enriched, n = 46). To isolate the specific role of adult-born granule cells in the inhibition of M/T cells among the global granule cell population, we then optogenically activated channelrhodopsin expressing adult-born granule cells in OB slices (adult-born granule cells were transduced with Lenti-hSyn-ChR2EYFP 25- 30 days before testing) and recorded the evoked inhibitory post-synaptic current (eIPSC) in M/T cells (Enriched n = 27 cells from three mice; Non-Enriched n = 25 cells from three mice) (Figure 1I, left). Data analysis indicated an overall effect of light stimulation of granule cells on M/T cells eIPSC frequency (light effect, parametric Anova, F(1,198)=4.37, p=0.037, paired t-tests for light versus pre light: p=0.0025 in the enriched group, p=0.2 in the non-enriched group) (Figure 1—figure supplement 3A). Besides, eIPSC frequency was higher in enriched compared to Non-enriched animals (learning effect, parametric ANOVA, F(1,198)=4.47, p=0.03, t-test for difference between Enriched and non-enriched under light stimulation, p=0.04) (Figure 1—figure supplement 3A). No change in eIPSC amplitude was observed (parametric ANOVA, light effect F(1,160)=2.66, p=0.1, learning effect F(1,160)=1.72, p=0.19) (Figure 1—figure supplement 3B). To identify the individual M/T cells actually responding to adult-born granule cell stimulation, we performed a statistical comparison of the pre- and post-light IPSC occurrence across repeated stimulations (see Materials and methods). Results indicated that the percentage of light-responding M/T cells is marginally increased in enriched compared to non-enriched animals (unilateral Chi squared, p=0.07; Figure 1I right and Figure 1—source data 1). Additionally, we showed that the biophysical properties of adult-born granule cells (resting potential, membrane resistance, membrane capacitance and input/output curves) were unaffected by learning (Figure 1—figure supplement 4). Thus, the results from these experiments suggest that increased structural and/or functional connectivity is responsible for increased granule-to-M/T cell inhibition in enriched versus non-enriched groups, rather than any modification of the intrinsic properties of adult-born granule cells. To assess the functional outcome of this increased inhibition, we investigated in vivo M/T cell responsiveness to the learned odorant (+lim). Enriched and non-enriched animals were exposed to +lim 7 days after learning, and the percentage of odor-activated M/T cells (co-expressing the specific M/T cell marker Tbx1 (Imamura et al., 2011; Mitsui et al., 2011) and immediate early gene Zif268) was quantified (Figure 1—figure supplement 5). We found that implicit learning decreased the number of odor-activated M/T cells compared to non-enriched controls (587 ± 29 M/T cells counted per animal; Enr, n = 5 and Non-Enr, n = 4; Bonferroni-corrected test p=0.02; Table 1, Figure 1J and Figure 1—source data 1), resulting in a sparser representation of learned odorants at the OB output. This finding is highly consistent with the increased inhibition of M/T cells observed in slice recordings. Altogether, in the context of implicit learning, adult-born granule cells provide more inhibitory contacts on M/T cells, leading to increased sparseness of M/T odor responses. As a consequence, these findings strongly support the hypothesis that a decreased overlap between learned odor representations mediates the observed increase in odor discrimination (Chu et al., 2016).

Explicit learning increased adult-born granule cell survival but reduced spine density

Theoretically, discrimination could also be achieved through increasing the resolution of the odor representations (Aimone et al., 2011), in other words through increasing the response to the learned odorant (Doucette and Restrepo, 2008). Indeed, adding information to the representation of the stimulus could ultimately help discrimination, and we hypothesized that such a mechanism could be involved in associative learning where the odor is reinforced by a positive reward. We thus trained mice on a two odorized hole-board apparatus to associate +lim with a food reward while -lim was not reinforced (Conditioned group, Cond) (Mandairon et al., 2006a). Control animals were exposed to the same odorants randomly associated with the reward (pseudo-conditioned group) (Mandairon et al., 2006a; Sultan et al., 2010) (Figure 2A). Training took place over 5 days (four trials/day) and was evidenced by an increase of correct choices in the conditioned (n = 20; Friedman test day effect p=0.03) but not in the pseudo-conditioned group (PC) (n = 20; Friedman test day effect p=0.57; Figure 2B and Figure 2—source data 1). Using the habituation/cross habituation test, we confirmed that conditioned but not pseudo-conditioned animals discriminated +lim from –lim as they did with implicit learning (Figure 2—figure supplement 1). As with implicit learning, explicit learning induced an increase in the survival of adult-born granule cells (n = 6/group; Bonferroni-corrected test p=0.025; Table 1; Figure 2C and Figure 2—source data 1) as well as an increase in their responsiveness to the learned odor (+lim), as assessed by BrdU/Zif268 co-expression (n = 7/group; Bonferroni-corrected test p=0.005; Figure 2D and Figure 2—source data 1). However, in contrast to implicit learning, analysis of the fine morphology of adult-born neuron dendrites showed a decrease in spine density on the apical dendrites after conditioning compared to pseudo-conditioning (Cond 106 segments, four mice and PC 56 segments, five mice; non-parametric corrected test p=0.0025; Figure 2E and Figure 2—source data 1), while no such change was observed in the basal domain (Cond: 37 segments, four mice and PC: 43 segments, four mice; non-parametric corrected test p=0.09; Figure 2F and Figure 2—source data 1). 42 days after conditioning, when discrimination had returned to control levels (Figure 2—figure supplement 2A), the spine density had also returned to control values (Figure 2—figure supplement 2B). This indicated that the morphological changes of adult-born neurons paralleled discrimination performance.

Inhibition on mitral cells is decreased after explicit learning

We further analyzed the impact of adult-born neurons on M/T cell activity by recording sIPSCs from M/T cells in OB slices. sIPSCs amplitude was decreased (p=0.015, Kruskal-Wallis Anova followed by FDR-corrected permutation test; Table 1) while their frequency was marginally affected (p=0.06, Kruskal-Wallis Anova followed by FDR-corrected permutation test; Table 1) (Figure 2G and Figure 2—source data 1) suggesting a global reduction of M/T cell inhibition in the OB in conditioned versus pseudo-conditioned animals. To analyze specifically the impact of adult-born granule cells on M/T cells, we recorded IPSCs on M/T cells in response to light stimulation of channelrhodopsin-expressing adult-born granule cells. The light stimulation of adult-born granule cells induced an increase in M/T cell eIPSC frequency in both conditioned and pseudo-conditioned animals compared to pre-light (light effect, parametric Anova, F(1,198)=4.37, p=0.037, paired t-tests for light versus pre light, p=0.015 in the pseudo-conditioned group, n = 46 cells from five mice; p=0.001 in the conditioned group, n = 34 cells from three mice) (Figure 2—figure supplement 3A) with no change in amplitude (Figure 2—figure supplement 3B). Besides, eIPSC frequency was lower in conditioned compared to pseudo-conditioned animals (learning effect, parametric Anova, F(1,198)=4.47, p=0.03, t-test for difference between conditioned and pseudo-conditioned groups under light stimulation, p=0.0015; Pseudo-conditioned n = 22 cells from two mice; Conditioned n = 28 cells from two mice) (Figure 2—figure supplement 3A). Analysis of individual cell responses showed the percentage of M/T cells responding to light stimulation of adult-born granule cells by an increased eIPSC frequency was lower in conditioning compared to pseudo-conditioning (Unilateral Chi squared test, p=0.007), (Figure 2H and Figure 2—source data 1). The biophysical properties of adult-born granule cells (resting potential, membrane resistance, membrane capacitance and input/output curves) were not affected by explicit learning (Figure 2—figure supplement 4). As a consequence, we concluded that inhibitory inputs to M/T cells were decreased after explicit learning. The functional outcome of this decreased inhibition was assessed using Tbx21/Zif268 co-expression in mitral cells (540 ± 20 mitral cells counted per animal; n = 5 animals/group), and showed a significant increase in mitral cell responsiveness to the learned odor, (Bonferroni-corrected test p=0.04; Figure 2I and Figure 2—source data 1) suggesting that in this case, odor representations at the OB output were not rendered sparser by learning.

Comparison between implicit and explicit learning

Using a two-factor Anova, we found that explicit and implicit learnings similarly increased adult-born survival (learning versus controls F(1,18)=19.72, p=0.0003), regardless of the type of learning (implicit versus explicit F(1, 18)=0.21, p=0.64 and no interaction F(1,18)=0.006, p 0.9) (Figure 1C and Figure 2C). The same was true for adult-born cells responsiveness to the learned odor (learning effect F(1,23)=27.70, p<0.0001, no effect of the type of learning F(1, 23)=0.23, p=0.87, no interaction F(1,23)=0.0001, p=0.98) (Figure 1D and Figure 2D). However, implicit learning resulted in stronger inhibition on M/T cells than did explicit learning as shown by lower sIPSC amplitude in conditioned versus enriched groups (p=0.0048, FDR-corrected permutation test, Table 1). In addition, adult-born granule cells bear more spines in enriched compared to conditioned groups in their apical domain which interacts with mitral cells (p=0.0003, FDR-corrected permutation test, Table 1, Figure 1F,G; Figure 2E,F). In line with this, M/T cells displayed lower responsiveness to the learned odorant in enriched compared to conditioned animals (p=0.015, Bonferroni post hoc test, Tbx21/Zif268, Table 1Figure 1J and Figure 2I).

Figure 1 with 5 supplements see all
Behavioral and neural effects of implicit learning.

(A) Experimental design for implicit learning (S: Sacrifice; Test: Habituation/Cross Habituation task). (B) Behavior. Habituation/cross habituation task indicated that +lim and -lim are not discriminated in control non-enriched group. In contrast, enrichment allows discrimination as observed by the significant increase of investigation time between the last habituation trial (OHab4) and the presentation of the second odorant of the pair (Otest). (C) Adult-born cell (BrdU-positive cell) density is increased after implicit learning. (D) The percentage of odor-responsive adult-born cells (expressing Zif268) is increased after implicit learning. (E) Spine density of adult-born neuron transduced by Lenti hSyn ChR2EYFP is analyzed in the apical and basal domains. (F) Spine density in the apical domain is increased after implicit learning. (G) Spine density in the basal domain is increased after implicit learning. (H) Representative traces of sIPSC recorded on mitral cells for Enr and Non-Enr animals (up). sIPSC frequency is increased after implicit learning while no modification is observed for sIPSC amplitude (down). (I) Left, experimental design for studying the connectivity of adult-born neurons on M/T cells. Middle, example of the effect of optogenic stimulation of adult-born neurons on M/T cells IPSC (top connected M/T cell, bottom unconnected M/T cell, superposition of 10 traces). Right, percentage of M/T cells exhibiting a significant response to light stimulation of adult-born neurons. (J) The percentage of M/T cells (Tbx21positive) expressing Zif268 is decreased after implicit learning. *:p<0.05; #p=0.07.

https://doi.org/10.7554/eLife.34976.003

Interestingly, when comparing the controls for each learning group (pseudo-conditioned versus non-enriched) (Table 1), they appeared to differ. More precisely, sIPSC frequencies were higher in the pseudo-conditioned compared to the non-enriched animals (p=0.0006, FDR-corrected permutation test). Consistent with this, the number of odor-activated M/T cells tended to be smaller in the pseudo-conditioned than the non-enriched animals (p=0.053 Bonferroni post hoc test, Table 1). These differences could be explained by the fact that the pseudo-conditioned animals, in contrast to the non-enriched animals were exposed to the odorants throughout the pseudo-conditioning procedure.

Finally, we observed that the pseudo-conditioned animals shared cellular similarities with enriched animals (similar sIPSC frequency, percentage of odor-activated M/T cells and basal spine density) (Table 1) despite the fact that they do not show behavioral discrimination.

Discussion

The findings reported here reveal that enhanced odor discrimination following implicit and explicit learning is achieved through different mechanisms. While the number of integrated adult-born granule cells was similar in both forms of learning, they differed in the synaptic integration mode of adult-born neurons and their effect on M/T cell responses to odor. Implicit learning increased spine density on adult-born granule cells (apical and basal dendritic domains), in agreement with previous studies (Daroles et al., 2016; Zhang et al., 2016) and increased inhibition of mitral cells, consistent with reduced number of mitral cells responding to the learned odorant. Increased number of spine in the basal domain is suggestive of an enhanced connectivity between inputs from centrifugal projections and adult-born granule cells, possibly leading to more global excitation of adult-born granule cells (Moreno et al., 2012; Lepousez et al., 2014). More apical spines increase feedback inhibition between M/T and granule cells increasing local inhibition. These data suggest that in response to implicit learning, structural plasticity of adult-born cells mediates an increased feedback and central inhibition on mitral cells to support perceptual discrimination of odorants. This view is strongly supported by our previous report of enhanced paired-pulse inhibition in the OB after implicit learning (Moreno et al., 2009), and of the loss of learning upon blockade of neurogenesis (Moreno et al., 2009). In addition to increased spine density, the increase in the number of adult-born cells after implicit learning is also likely contributing to the enhancement of inhibition on mitral cells.

In contrast to the effects of implicit learning, a decrease in spine density in the apical domain of adult-born neurons is accompanied by a decrease in sIPCS amplitude in mitral cells after explicit learning. In addition, an overall increase rather than a decrease of mitral cells activation was observed in response to the learned odorant compared to pseudo-conditioned animals. Reduced synaptic contacts on the apical dendrites of adult born neurons reduce local feedback inhibition leading to an enhanced response of M/T cells to the learned odorants.

To summarize, the effects of implicit and explicit learning on M/T odor responses are opposite: an overall sparser response to the learned odor after implicit learning and an overall increased response to the conditioned odor after explicit learning, while similar numbers of adult-born neurons are present. Because new adult-born granule cells replace older ones (Imayoshi et al., 2008), replacing pre-existing granule cells by new ones with fewer synaptic contacts with mitral cells (in conditioned animals) would result in a global pool of granule cells delivering less local inhibition in response to the conditioned odor. In contrast, replacing granule cells by new cells making more local and global synaptic contacts with mitral cells (enriched animals) would result in a shift toward more inhibition in the network. This is consistent with the experimental observations. Computational modeling suggested that reinforced inhibition reduces the overlap between similar odor representations at the level of mitral cells (Mandairon et al., 2006b). As an alternative mechanism leading to behavioral discrimination, decreased inhibition could lead to enhanced representations of a conditioned odor only, improving the resolution of the representation of behaviorally significant stimuli (Aimone et al., 2011). This proposed mechanism that could be at play in explicit learning is consistent with the associative coding features of the OB (Doucette et al., 2011; Fletcher, 2012). While these evidences favor a prominent role of adult born neurons in shaping mitral cell activity in learning, the role of pre-existing interneurons in mitral cell response plasticity remains to be investigated.

In conclusion, the data presented here challenge current hypotheses by showing that the same rate of adult-born cell survival in the OB can enhance fine discrimination by either increasing or decreasing OB output, thus revealing a new facet of the adaptability of the network provided by adult neurogenesis.

Materials and methods

Animals

Adult male C56Bl6/J mice (Charles River, L’arbresles, France) aged 8 weeks at the beginning of the experiments were used in this experiment. They were housed in groups of five in standard laboratory cages with water and food ad libitum (except during the explicit learning) and were kept on a 12-hr light/dark cycle (at constant temperature of 22°C). All behavioral training was conducted in the afternoon (14:00-17:00). Experiments were done following procedures in accordance with the European Community Council Directive of 22nd September 2010 (2010/63/UE) and the National Ethics Committee (Agreement DR2013-48 (vM)). Every effort was made to minimize suffering.

Implicit (perceptual) learning

Experimental design

Request a detailed protocol

After a 10-day enrichment period, the mice were tested for their ability to discriminate between +limonene (+lim) and – limonene (- lim) using an olfactory cross-habituation test (see below). This group was compared to a non-enriched control group. To label the adult-born cells, the mice were injected with BrdU or GFP expressing lentivirus 8 days before the beginning of the enrichment period. They were sacrificed 25 (Figures 1 and 2) or 60 (Figure 1—figure supplement 2 and Figure 2—figure supplement 2) days after these injections.

Figure 2 with 4 supplements see all
Behavioral and neural effects of explicit learning.

(A) Experimental design for explicit learning (S: Sacrifice, Test:Habituation/Cross Habituation Task). (B) From D1 to D5 of training, the percentage of correct choices increased in conditioned (Cond) but not pseudo-conditioned animals (PC) indicating that learning occurred only in the conditioned animals (C) Adult-born cell (BrdU-positive cell) density is increased after explicit learning (D) The percentage of odor-responsive adult-born cells (expressing Zif268) is increased after explicit learning. (E). Spine density of the apical domain decreased after explicit learning. (F) Spine density of the basal domain is unchanged after explicit learning. (G) Representative traces of sIPSC for Cond and PC (up). sIPSC frequency and amplitude are decreased after explicit learning (down). (H) Percentage of mitral cells exhibiting a significant response to light stimulation of adult-born granule cells. (I) The percentage of mitral cells (Tbx21) expressing Zif268 is higher in Cond versus PC animals. *p<0.05; the data are expressed as mean values ± SEM.

https://doi.org/10.7554/eLife.34976.010

Odor enrichment and control

Request a detailed protocol

Odor enrichment consisted of exposure to + and - limonene (purity >97%; Sigma-Aldrich Corp., Sr. Louis, MO, USA) for 1 hr per day over 10 consecutive days. The odors were presented simultaneously on two separate swabs containing 100 µl of pure odor placed in two separate tea balls hanging from the cover of the animal’s cage. The non-enriched control mice were housed under the same conditions except that the two tea balls were left empty.

Olfactory habituation/cross habituation test

Request a detailed protocol

We used a cross-habituation test to assess discrimination. Briefly, the task assesses the degree to which mice are able to discriminate between odorants by habituating them to an odorant (OHab) and measuring their cross-habituation to a second odorant (OTest). If the test odorant is not discriminated from the habituation one, it will not elicit an increased investigation time by the mouse. We presented each odor in a tea ball hanging on the cover of the cage containing 60 µl of the diluted odor (1 Pa) on filter paper (Whatman #1) for 50 s. There was a 5 min pause between presentations. Odors are renewed between each test. For each animal, each odorant of the pair was used alternatively as the habituation or test odorant. The amount of time that the mice investigated the odorant was recorded during all trials as previously described (Moreno et al., 2009, 2012). We determined (1) if the investigation time decreased from OHab1 to OHab4 (habituation); and (2) if the investigation time during OTest was significantly higher than for OHab4 (discrimination). These behavioural experiments were conducted blind with regard to experimental group.

Explicit (associative) learning

Experimental design

Request a detailed protocol

The mice learned to discriminate +lim and - lim by training them to associate +lim with a food reward (see below). This group of mice were compared to a pseudo-conditioned control group. After a 5-day period of conditioning, the mice were tested for their ability to discriminate between +limonene (+lim) and – limonene (- lim) using an olfactory cross-habituation test (see above). They were injected with BrdU or with GFP lentivirus 13 days before training and sacrificed 25 (Figure 1 and Figure 2) or 60 days after injection. During the olfactory learning experiments, water was continuously available, but the mice were deprived of food (~20% reduction of daily consummation, leading to a 10% reduction in body weight) for 5 days before the shaping session.

Shaping

Request a detailed protocol

The mice were first trained to retrieve a reward (a small bit of sweetened cereal; Kellogg’s, Battle Creek, MI, USA) by digging through the bedding. The mouse was put in the start area of the two hole-board apparatus and allowed to dig for 2 min. During the first few trials, the reward was placed on top of the bedding in one of the holes. After the mice successfully retrieved the reward several times, it was successively buried deeper and deeper in the bedding. Shaping was considered to be complete when a mouse could successfully retrieve a reward buried deep in the bedding.

Conditioning

Request a detailed protocol

Conditioning consisted of 5 sessions (1/day) of 4 trials (2 min/trial, inter trial interval 15 min). For each trial, the mouse was placed in the start area and had to retrieve a reward now systematically associated with the +lim (20 µl of pure odorant). To avoid spatial learning, the rewarded dish was randomly placed in one of the two holes; the other hole contained – lim (20 µl of pure odorant) and no reward. In the pseudo-conditioned group, the reinforcement was randomly associated with either the +lim or the - lim. For each trial, correct choices (first nose poke in the odorized hole) were recorded as indicative of learning.

Lentivirus injections

Request a detailed protocol

The mice were anesthetized by injecting a mixture of 50 mg/kg ketamine and 7.5 mg/kg xylazine (intraperitoneal) and then secured in a stereotactic instrument (Narishige Scientific Instruments, Tokyo, Japan). For the optogenetic experiments, they were injected bilaterally in the subventricular zone (AP +1 mm, ML ±1 mm, DV −2.3 mm) with a AAV5-hSyn-ChR2EYFP viral vector (150 nl per side; Addgene 26973P). This construct was a kindly donated by the Deisseroth laboratory and produced by the Penn Vector Core facility (titer 1,3 1013 UI/ml). For the neuronal morphology experiments, a Lenti-PGK-GFP (Addgene #12252; titer 2 109 UI/ml) viral vector was injected at the same coordinates (200 nl per site). All injections were performed at a rate of 100 nl/min using a programmable syringe controller (KD Scientific Inc. Holliston, USA).

Histology

Tissue preparation

Request a detailed protocol

Mice randomly taken from the behavioral groups were placed in a clean cage for 1 hr. To investigate immediate early gene expression in response to the learned odorant (+lim), the mice were presented with a tea ball containing 100 µl of pure odorant for 1 hr. One hour after the end of the odor stimulation, they were deeply anesthetized (pentobarbital, 0.2 ml/30 g) and killed by intracardiac perfusion of 50 ml of fixative (4% paraformaldehyde in phosphate buffer, pH 7.4). Their brains were removed, postfixed, cryoprotected in sucrose (20%), frozen rapidly, and then stored at −20°C before sectioning (40 µm for neuronal morphology and 14 µm for BrdU analysis) with a cryostat (Reichert-Jung, NuBlock, Germany).

Image analysis and quantification

Request a detailed protocol

Morphological analysis of adult-born cells was performed after GFP immunohistochemisty (chicken GFP antibody, 1:1000, Anaspec TEBU, ref: 55423). Images were acquired on a Zeiss pseudo-confocal system. For analysis of spine density, images were taken blind to the identity of the experimental group with a 100x objective (lateral and z-axis resolutions were 60 and 200 nm, respectively). Dendritic processes and spines were then analyzed using NeuronStudio software (Mandairon et al., 2006a; Rodriguez et al., 2008). This software allows for automated detection of three-dimensional (3D) neuronal morphology (dendrites and spines) from confocal z-series stacks on a spatial scale. Because spine density assessment was challenging for the automated detection in our model, we counted them manually with the help of the 3D reconstruction. All morphology analyses were done blind with regard to the experimental group.

Assessment of neurogenesis

5-Bromo-2-deoxyuridine (BrdU) administration

Request a detailed protocol

BrdU (Sigma-Aldrich) (50 mg/kg in saline, 3 times at 2 hr intervals) was injected 13 days before the behavioral training began.

BrdU immunohistochemistry

Request a detailed protocol

The protocol has been previously described (Mandairon et al., 2006a). Brain sections were first incubated in Target Retrieval Solution (Dako, Trappes, France) for 20 min at 98°C. After cooling for 20 min, they were treated with 0.5% Triton X-100 (Sigma-Aldrich) in PBS for 30 min, then for 3 min with pepsin (0.43 U/ml in 0.1 N HCl, Sigma-Aldrich). Endogenous peroxidases were blocked with a solution of 3% H2O2 in 0.1 M PBS. Sections were then incubated for 90 min in 5% normal horse serum (Vector Laboratories, Burlingame, CA, USA), in 5% BSA (Sigma-Aldrich) and 0.125% Triton X-100 to block nonspecific binding and were incubated overnight at 4°C in a mouse anti-BrdU primary antibody (1:100; Millipore; ref: MAB4072). Sections were then incubated in a horse biotinylated anti-mouse secondary antibody (1:200; Vector Laboratories) for 2 hr and processed with avidin-biotin-peroxidase complex (ABC Elite Kit, Vector Laboratories) for 30 min. Finally, sections were reacted in 0.05% 3,3-diaminobenzidine-tetra-hydrochloride (Sigma-Aldrich), 0.03% NiCl2, and 0.03% H2O2 in Tris-HCl buffer (0.05 M, pH 7.6), dehydrated in graded ethanols, and cover-slipped in DPX.

BrdU-positive cell quantification

Request a detailed protocol

All cell counts were conducted blind with regard to experimental group. Data were collected with the help of mapping software (Mercator Pro; Explora Nova, La Rochelle, France), coupled to a Zeiss microscope (Carl Zeiss, Oberkochen, Germany). BrdU-positive cells were counted in the granule cell layer of the OB on six sections distributed along the olfactory bulb (14 µm thick, 70 µm intervals). The number of positive cells was divided by the surface of the region of interest to yield the total densities of labeled cells (labeled profiles/µm2). All BrdU-positive cell counts were done blind with regard to the experimental group.

Double-labeling analysis

Request a detailed protocol

To determine the phenotype of BrdU-positive cells of the OB, BrdU/neuronal nuclei (NeuN) double-labeling was performed using a rat anti-BrdU (1:100; Abcys ref: Abc117-7513) and a mouse anti-NeuN (1:500, Millipore MAB337). For functional involvement of newborn neurons, Zif268/BrdU double-labeling was performed using a rabbit anti-Zif268 antibody (1:1000, Santa Cruz Biotechnology, Egr-1 SC:189). The appropriate secondary antibodies, coupled to Alexa 633 (Molecular Probes, Eugene, OR, USA) to reveal BrdU and Alexa 488 (Molecular Probes) to reveal other markers, were used. BrdU-positive cells were examined for co-labeling with NeuN and Zif268 (80–100 cells/animal, n = 5 animals/group). The double-labeled cells were observed and analyzed by pseudo-confocal scanning microscopy using a Zeiss microscope equipped with an Apotome. All double-labeled cell counts were done blind with regard to the experimental group.

Zif268-Tbx21 double labeling

Request a detailed protocol

The activation of mitral cell in response to the learned odorant (+lim) was assessed using co-labeling with Zif268 and Tbx21 (1:20,000; provided by Y. Yoshihara, RIKEN). The appropriate secondary antibodies (Molecular Probes, Eugene, OR, USA) were used. All Zif268-Tbx21 cell counts were done blind with regard to the experimental group.

Brain slice electrophysiological experiments

Request a detailed protocol

For brain slice experiments, 20 additional mice were injected with lentivirus expressing GFP as described previously and submitted to implicit or explicit learning. Electrophysiological recording and data analysis were done blind with respect to experimental group.

The animals were anesthetized with intraperitoneal injection of 50 µl of ketamine (50 mg/ml) between 25 and 30 days after the lentivirus injections and killed by decapitation. The head was quickly immersed in ice-cold (2–4°C) carbogenized artificial cerebrospinal fluid (cACSF; composition: 125 mM NaCl, 4 mM KCl, 25 mM NaHCO3, 0.5 mM CaCl2, 1.25 mM NaH2PO4, 7 mM MgCl2 and 5.5 mM glucose; pH = 7.4) oxygenated with 95% O2/5% CO2. The osmolarity was adjusted to 320 mOsm with sucrose. OBs were removed and cut into horizontal slices (400 µm thick) using a Leica VT1000s vibratome (Leica Biosystems, France). These were then incubated in a Gibb’s chamber at 30 ± 1°C using an ACSF solution with a composition similar to the cACSF, except that the CaCl2 and MgCl2 concentrations were 2 mM and 1 mM, respectively. Slices were transferred to a recording chamber mounted on an upright microscope (Axioskop FS, Zeiss) and were continuously perfused with oxygenated ACSF (4 ml/min) at 30 ± 1°C. Neurons were visualized using a 40X objective (Zeiss Plan-APOCHROMAT) and a Hamamatsu ‘Orca Flash 4.0’ camera. Measurements were performed with an RK 400 amplifier (BioLogic, France). The data were acquired with a sampling frequency of 25 kHz on a PC-Pentium D computer using a 12-bit A/D-D/A converter (Digidata 1440A, Axon Instruments) and PClamp10 software (Axon Instruments). The junction potential was corrected offline. Patch-clamp configurations were achieved with borosilicate pipettes (o.d.: 1.5 mm; i.d.: 1.17 mm; Clark Electromedical Instruments). Membrane properties and firing input/output curves of adult born neurons were recorded with the following intracellular solution: in mM (131 K-gluconate; 10 HEPES; 1 EGTA; 4 ATP-Na2+; 0.3 GTP-Na3; 1 MgCl2; 10 phosphocreatine). To record the spontaneous IPSCs (sIPSC) on mitral cells and those evoked by optogenetic stimulation of adult-born granule cells (see below) the recording pipette was filled with the following intracellular solution (in mM; 135 caesium gluconate, 10 KCl, 10 Hepes, 1 EGTA, 0.1 CaCl2, 2 MgATP and 0.4 GTP-Na3) adjusted with 1 M CsOH to pH 7.3. The ECl was −85 mV and both sIPSC and light evoked IPSCs were recorded as outward currents at holding potential of 0 mV.

Optogenetic experiment

Request a detailed protocol

EYFP expression in adult-born granule cells was detected with excitation (470/40 nm) and emission (dichroic mirror: 495; 525/50 nm) band pass filters (Zeiss filter set 38 HE) 25–30 days after their birth. Optogenetic stimulation of adult-born cells was produced by 10 ms illumination by a blue LED (Dual Port OptoLED CAIRN, UK) with the excitation spectrum centered at 470 nm at an inter-stimulus interval of 10 s. For each recorded mitral cell the connectivity with adult-born granule cells was evaluated by statistically comparing the occurrence of IPSCs in the 50 ms preceding the light stimulation with the IPSC occurrence in the first 50 ms following the beginning of light stimulation, across repeated light stimulations (mean = 16 stimulations per cell, range 8–52). Pre-light and post-light IPSC frequency was calculated as follows: the total number of IPSC in the 50 ms preceding and following the beginning of light stimulation was divided by the number of repetitions and multiplied by 0.05 s.

Statistics

All analyses were performed using Statistics. A Kolmogorov-Smirnov test was used to assess normality of measured parameters. For all parametric tests, homogeneity of variance was tested using Levene’s test.

For each habituation/cross habituation task, a non-parametric Friedman Anova was performed to determine whether mice exhibited habituation, and a paired Wilcoxon test (comparing Ohab4 and Otest) to test the discrimination abilities. Discrimination was indicated by a significant increase in investigation time during the test trial. For conditioning task, a non-parametric Friedman Anova was used to assess the effect of training days on the percentage of correct choice.

Non-parametric Kruskal-Wallis Anova including the four experimental groups (Enr, Non-Enr, PC and Cond) followed by FDR-corrected non-parametric permutation tests based on 1000000 artificial groups, were used for sets of data that did not reach normality (Table 1).

For normally distributed data sets, parametric Anovas followed by Bonferroni-corrected post hoc tests were used (Table 1).

In addition, for eIPSC data, in order to take into account the factors ‘light’ (pre and post light), ‘group of learning’ (implicit versus explicit) and ‘learning’ (Cond or Enr versus PC or Non-Enr), we normalized the data (ln(x + 1)) and performed a 3-factor Anova followed by an unilateral paired t-test for comparison between light and no light condition or an unpaired t-test for the comparison between learning and control condition. In addition, individual cell analysis of the effect of light was performed by comparing the occurrence of pre and post light eIPSC across repetitions of light stimulation by unilateral Chi squared tests.

No statistical methods were used to predetermine sample sizes, but our sample sizes were similar to those reported in previous publications. Data collection and animal assignation to the various experimental groups were randomized.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24

Decision letter

  1. Naoshige Uchida
    Reviewing Editor; Harvard University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your work entitled "Opposite regulation of inhibition by adult-born granule cells in response to implicit versus explicit olfactory learning" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor.

Summary:

Mandairon and colleagues examined the effects of two types of olfactory learning on granule cell neurogenesis, spine density of adult-born granule cells (GCs) and mitral cell (MC) activity and physiology. The learning paradigms compared are (1) presentation of two different odors [(+)- and (-)-limonene] in the home cage (1 hour/day, 10 days) ("implicit learning"), and (2) odor-reward associative learning using a food-digging task (explicit [associative] learning).

The data suggest that implicit learning increased inhibitory drive onto mitral cells (assessed based on the frequency of IPSC after a short pulse of ChR2 stimulation of adult-born GCs) and decreased odor-evoked activity of the mitral cells (as measured by immediate early gene expressions). These changes were accompanied by increased spine densities both at the apical and basal dendrites of GCs. In contrast, in explicit learning, the frequency of light-evoked IPSCs was decreased in MCs, and odor-evoked activity in MCs is increased. These changes were accompanied by a decrease in spine density at apical but not basal dendrites of GCs.

Although both implicit and explicit learning induced increased discriminability in the test phase as tested by the same assay (habituation paradigm), the authors found different morphological and physiological changes in GCs and MCs. All the reviewers agreed these are potentially very interesting findings. However, all pointed substantive concerns regarding the use of statistical methods and the robustness of the data. Specifically, some conclusions were made without correcting for multiple comparisons. Furthermore, statistical significance of some of the analyses appeared to be due to outlier points, raising the issue of robustness. After discussing these concerns, the reviewers thought that it is very likely that the authors need to perform additional experiments, or significantly change their conclusions (because some results might not be statistically significant after correcting for multiple comparisons). It was the consensus of the reviewers that addressing these issues will take more than 2-3 months. We therefore decided that we cannot consider the current manuscript further at least in the present form. If the authors improve their statistical arguments either by revising the manuscript to use more statistically sound analysis or by performing more experiments to increase the sample size, we are willing to consider a revised manuscript as a new submission.

Essential revisions:

1) All the reviewers raised their concerns regarding statistical analyses. Most importantly, the authors perform multiple pair-wise comparisons. In such cases, a significance criterion should be corrected for multiple comparisons (e.g. by using Bonferroni correction). When appropriate, a one-way ANOVA (analysis of variance) followed by post-hoc pair-wise comparisons is preferred. Many conclusions do not appear to hold after these corrections. This would mean that some conclusions need to be revised or the authors need to perform additional experiments to increase the sample size. More detailed criticisms and suggestions can be found in the reviewers' comments appended below.

2) Some of the statistical significance appear to be due to some outlier points in the data (e.g. reviewer 1, point #1; reviewer 2, point #2). The authors need to consider whether the results hold even removing these points. To make these results more reliable, we expect that a larger sample size is required to draw statistically-sound conclusions. The reviewers also thought that the conclusion that adult-born neurons are mediating the enhanced inhibition onto MCs requires more work:

3) How do authors control for number of cells expressing ChR2 virus or normalize counts to interpret the ChR2 stimulation data (number of spines vs. number of cells)? Are equivalent numbers of cells infected in both groups? The increase in survival induced by learning should bias the number of Chr2 expressing GCs in OB.

4) Following reversal of changes in apical spine density (but with extra adult-born GCs still present), is the effect of ChR2 stimulation on MCs lost?

5) The link between behavior (higher discrimination) and the changes at the level of neurogenesis, morphology and physiology remain largely unclear. We understand that this study is not designed to examine the mechanisms that connect these changes with behavior, but the reviewers found some of the discussions very loose. We would like the authors to state these links more carefully, so that the manuscript becomes more scholarly. For example, the authors mention "pattern separation" multiple times and discuss that increased or decreased inhibition may be causally related pattern separations, and in turn, increased discriminability. However, whether pattern separation causally underlies increased discrimination or whether changes in inhibitory tones can lead to pattern separations remain totally unknown (also see point #4 by reviewer 3). The present study as well as existing literature does not specify specific mechanisms underlying increased discriminability after implicit or explicit olfactory learning.

Reviewer #1:

In this manuscript, the authors describe bidirectional changes in the spine density of olfactory bulb adult-born granule cells (aGCs), that depend on learning paradigms which the authors have established previously.

The used paradigms appear subtle and physiological, since they rely on few odor exposures. Nevertheless, the changes can be observed in electrophysiological recordings in brain slices even though the relation of the investigated mitral cells to odor-activated glomeruli is not known. Thus, the phenomena might reflect a general response of an otherwise deprived olfactory system.

My general criticism is that there are counterintuitive results that would require stronger statistics to be firmly established. Therefore, the authors should increase sample size, try to sample cells more specifically (see below) and thoroughly discuss their results, possibly requiring a longer format.

1) Statistical significance appears to be an issue with some results – would unilateral naris occlusion be feasible, which might allow for pairwise comparisons across hemibulbs? Or a higher number of animals?

While the increased spine density (Figure 1F) looks fine, all the other changes (Figure 1G, Figure 2E) show highly overlapping distributions, perhaps because some MCs in the sample might not have been involved in limonene-evoked bulbar activation. Would it be possible to preferentially record from bulbar parts with strong limonene activation? Or even better fluorescently label MC/GCs expressing IEGs and record from them?

There is a substantial amount of eIPSC recordings with high frequency in the 30-60 Hz range in PC condition (Figure 2H). This range is not observed in Figure 1J, even not in the enriched case. Were the high frequency responses recorded in slices from the same animal? Strikingly, the very same issue arises also for the sIPSC frequency and amplitude (Figure 2G) – for both there are several data points in the high range for PC that are not observed in non-enr or enr conditions. If we now doubt the PC data for this reason, what about the statistical comparison between non-enr and cond? Unfortunately, this comparison is missing from Table 1 – by eye between Figure 1F and Figure 2E almost certainly there is no difference for the apical spine density. Which casts doubt on the central finding of reduced connectivity – at least anatomically.

2) What is the effect of the increase in number of adult born GCs also in the associative learning – how exactly can an addition of interneurons reduce inhibitory drive? As mentioned above, the decrease in spine numbers on the apical dendrite is barely significant. Nevertheless, there is a striking overall decrease in inhibitory drive of mitral cells, especially substantially reduced sIPSC frequency and amplitude (please label the examples from Figure 2G and Figure 1H top in the distributions shown below). If we believe these data (but see above), what is going on here? Synaptic plasticity – reduction in release probability and/or quantal size at the GC-MC contacts – also for preexisting GCs? Were the lost contacts located close to the MC somata and thus the efficiency is strongly reduced?

3) Missing of important reference: Zhang, Huang and Hu, (2016): increased spine density on GC dendrites following odor enrichment in Xenopus tadpoles.

Reviewer #2:

This report by Mandairon and colleagues is a significant new look on the function of adult born GABAergic neurons in the olfactory bulb. The community as a whole thinks in two ways about the types of learning that occur with passive exposure and conditioning. On the one hand, we all know that these are different types of learning and on the other we often refer to them as if they are the same (under the umbrella term of Learning). Similarly, for learning-based effects on the survival and integration of new granule cells in the OB. This paper shows that the ability to discriminate hard-to-discriminate odors after the two types of learning may depend in opposite fashion on neurogenesis. For passive learning, it appears that adult-born GC connections with MCs are strengthened in the external plexiform layer (at the reciprocal synapse). The authors conclude that this may be related to the effect known as sparsening. For active learning, the effects are opposite. The adult born GC effects on MCs are decreased. The argument is rigorous and rational. My only major concern regards the many pairwise statistical tests (see below). The results reported here, if improved statistical methods confirm them, may constitute one of the most significant findings in the field of olfactory learning as it relates to neurogenesis.

1) There are many more implied comparisons than reported in Table 1. With 4 groups of mice, this makes 6 total comparisons for each factor (cond vs. PC, cond vs. enr, cond vs. non-enr, PC vs enr, PC vs non-enr, enr vs non-enr). With the multiple comparisons within each factor, the threshold for significance should be 0.00833. There is a pretty good argument that all of the factors should be combined in a single analysis, which then multiplies the comparisons and lowers the threshold value, or at least that like-kind factors be grouped (spine density apical/basal as one group, e/sIPSP frequency as another). This would make the p threshold smaller by a factor of 2 at least. Please provide justification in the methods for not performing the multiple comparisons adjustment or do the analysis with the corrected p thresholds. In addition to lowering the threshold, a more conservative analysis might provide additional insight.

2) This is not a disagreement but rather a different interpretation. It is the adult born cells that show the effects reported (and assuming that the earlier born cells do not). In the passive condition, the result is that new GCs inhibit MCs more at the apical dendrites where primary processing happens. Also, we know that these new cells are relatively specific to the enriched odors (previous work by the first and second to last authors). Is it possible that these new connections serve to help the mice ignore the conditioned stimuli and that the enhanced discrimination ability is a side-effect of this now-active ignorance? On the other hand, in the case of active learning there are more cells born and integrated (perhaps associated with the learned odors). Because these cells survive, one assumes that the inputs from higher order areas are strong. The signal is amplified, so one might expect that the now meaning-based odor perception is accomplished in the AON or PC or both, rather than in the OB.

Reviewer #3:

Mandairon and colleagues examined the effects of passive (odor exposure, habituation) and active (odor reinforced with reward) olfactory learning on granule cell neurogenesis, apical and basal dendritic spines of adult-born granule cells and mitral cell activity and physiology. The authors find/claim that passive learning increased neurogenesis and consequently, increased inhibitory drive onto mitral cells (assessed using ChR2 stimulation of adultborn GCs) and decreased odor-evoked activity of the mitral cells. In contrast, the authors show that active learning produces a decrease in apical dendritic spines without affecting basal spines. Interestingly, the decrease in apical dendritic spine density was reversed with restoration of discrimination levels to baseline over time. Furthermore, odor-evoked activity in MCs was increased and ChR2 light evoked IPSC frequency of the mitral cells was decreased.

This is an interesting study that begins to delineate how different kinds of learning affect GC-MC connectivity. Because passive and active learning may affect inputs and outputs of adult born and developmentally generated GCs and mitral cells (along with physiological properties of mitral cells), it is not clear how adultborn GCs are solely driving changes in inhibition onto mitral cells under these different learning conditions.

For example, does blockade of adult GC neurogenesis under active or passive learning conditions (over the learning period) eliminate the reported effects on mitral cells?

1) It is not clear why Mann-Whitney was used over a one-way ANOVA for comparisons of apical or basal dendritic spine density.

2) Figure 1G: Effect on basal spine density appears to be driven by one datapoint. Does basal spine density return to baseline after 42 days?

3) How do authors control for number of cells expressing ChR2 virus or normalize counts to interpret the ChR2 stimulation data (number of spines vs. number of cells)? Are equivalent numbers of cells infected in both groups? The increase in survival induced by learning should bias the number of Chr2 expressing GCs in OB.

4) Following reversal of changes in apical spine density (but with extra adultborn GCs still present), is the effect of ChR2 stimulation on MCs lost?

5) Although the authors examine activity of mitral cells (sparseness), evidence examining population based coding or input-output transformations is critically needed to justify the use of term pattern separation or interpret the data within this framework.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Opposite regulation of inhibition by adult-born granule cells in response to implicit versus explicit olfactory learning" for further consideration at eLife. Your revised article has been favorably evaluated by Gary Westbrook (Senior Editor), a Reviewing editor (Naoshige Uchida), and three reviewers.

This is a resubmission in which the authors examined the effects of two types of olfactory learning on granule cell neurogenesis, spine density of adult-born granule cells (GCs) and mitral cell (MC) activity and physiology. They make various interesting findings. All the reviewers thought that the manuscript has greatly improved, but still raised some relatively minor concerns. We believe these points can be addressed without additional experiments, by revising the manuscript including adding discussions on some caveats or future directions. Please see below the individual points raised by the reviewers.

Reviewer #1:

This is a much improved paper that shows a striking difference in inhibition provided by adult-born GCs depending on the type of learning.

Is it possible that some part of the character of the effect is because of the odor similarity? Might it look different if the odors were easier to discriminate/less overlapping? I suggest the authors qualify the results under the class of fine odor discrimination. A nice follow-up study (not for this paper) would be to compare in each condition (implicit vs explicit), or at least the explicit condition, the effect of discrimination difficulty.

Reviewer #2:

I recommend the manuscript for publication.

Please discount the significance of the Daroles et al. study as evidence supporting a role for an increase in spine density in implicit learning. This is because FMRPcKO mice show an elevation in spine density at baseline (Figure 3) and therefore, it is not clear whether the failure to increase spine density further following learning or the elevation prior to learning is the culpable factor. Additionally, FMRP has numerous functions within neurons, that when disrupted, may be responsible for behavioral phenotype in implicit learning.

The authors acknowledge "However, while these evidences favor a prominent role of adult born neurons in shaping mitral cell activity in learning, the role of pre-existing interneurons in mitral cell response plasticity remains to be investigated". However, this possibility is absent from discussion. Please address this concern in addition to acknowledging "potential changes in mitral cell properties' also as a potential mechanism.

Reviewer #3:

The major concerns have been addressed appropriately, including a careful point-by-point response to the reviewers' concerns. In particular the statistics have been improved both in terms of n and of the tests, and the figures also are way more convincing. In my view the manuscript is now acceptable for publication, upon minor changes.

https://doi.org/10.7554/eLife.34976.018

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

Essential revisions:

1) All the reviewers raised their concerns regarding statistical analyses. Most importantly, the authors perform multiple pair-wise comparisons. In such cases, a significance criterion should be corrected for multiple comparisons (e.g. by using Bonferroni correction). When appropriate, a one-way ANOVA (analysis of variance) followed by post-hoc pair-wise comparisons is preferred. Many conclusions do not appear to hold after these corrections. This would mean that some conclusions need to be revised or the authors need to perform additional experiments to increase the sample size. More detailed criticisms and suggestions can be found in the reviewers' comments appended below.

We would like to thank the referees and the editor for their comments and suggestions on statistics that greatly enhance the strength of our conclusions. We have entirely revised the statistical analysis and added new data in response to reviewer’s comments. Please see detailed replies below.

2) Some of the statistical significance appear to be due to some outlier points in the data (e.g. reviewer 1, point #1; reviewer 2, point #2). The authors need to consider whether the results hold even removing these points.

The new analysis performed with larger samples confirmed the main results. See responses to reviewer 1 and 2.

To make these results more reliable, we expect that a larger sample size is required to draw statistically-sound conclusions. The reviewers also thought that the conclusion that adult-born neurons are mediating the enhanced inhibition onto MCs requires more work:

We added new data (new animals, new neurons). See table below:

Added n=total n=
BrdU PC26
n=animalsCond26
Non-Enr26
Enr04
BrdU/Zif268 PC27
n=animalsCond27
Non-Enr27
Enr16
Morpho Apical PC856/5
n=fragments/animalsCond0106/4
Non-Enr448/4
Enr030/4
Morpho Basal PC043/4
n=fragments/animalsCond037/4
Non-Enr1859/7
Enr1248/7
sIPSC Frequency PC23/346/5
n=cells/animalsCond28/134/3
Non-Enr24/348/6
Enr23/350/6
sIPSC Amplitude PC23/346/5
n=cells/animalsCond28/134/3
Non-Enr24/346/6
Enr23/348/6

Table A: List of additional data

3) How do authors control for number of cells expressing ChR2 virus or normalize counts to interpret the ChR2 stimulation data (number of spines vs. number of cells)? Are equivalent numbers of cells infected in both groups? The increase in survival induced by learning should bias the number of Chr2 expressing GCs in OB.

In line with the reviewer’s comment and according to our BrdU counts (Figure 1 and Figure 2D), we agree that enrichment or conditioning should result in an increased number of ChR2+ cells (we did not count these cells on the slices used for electrophysiology because the thickness of the sections (400 µm) makes it difficult in our view to provide reliable cell count). This leads to the conclusion that the increased inhibition on M/T cells following implicit learning could be accounted for by an increased number of adult-born cells. This issue is now discussed in the manuscript (Discussion section).

However, this does not explain how the very same number ChR2 GCs turns out to produce decreased inhibition on M/T cells in animal subjected to explicit learning. This is why we think that learning induces additional plasticity (such as changes in apical spine density onto M/T cells).

4) Following reversal of changes in apical spine density (but with extra adult-born GCs still present), is the effect of ChR2 stimulation on MCs lost?

Extra adult-born GCs expressing BrdU are no longer present 42 days after explicit learning as shown in a previous publication from our lab (Sultan et al., 2010). We now added new data to the present paper showing that extra adult-born GCs are no longer present 42 days after implicit learning, at a time where changes in apical spine density and behavioral discrimination are no longer observed (new Figure 1—figure supplement 2).

5) The link between behavior (higher discrimination) and the changes at the level of neurogenesis, morphology and physiology remain largely unclear. We understand that this study is not designed to examine the mechanisms that connect these changes with behavior, but the reviewers found some of the discussions very loose. We would like the authors to state these links more carefully, so that the manuscript becomes more scholarly. For example, the authors mention "pattern separation" multiple times and discuss that increased or decreased inhibition may be causally related pattern separations, and in turn, increased discriminability. However, whether pattern separation causally underlies increased discrimination or whether changes in inhibitory tones can lead to pattern separations remain totally unknown (also see point #4 by reviewer 3). The present study as well as existing literature does not specify specific mechanisms underlying increased discriminability after implicit or explicit olfactory learning.

In response to this comment, we have focused and rewritten the Discussion section.

We suggest that differences specifically in apical dendrite synapses onto granule cells create more or less feedback inhibition, which leads to specific changes in M/T responses to a learned odorant. After implicit learning, overall inhibition would be increased, learning to sharper odor representations in the OB. In contrast, after explicit learning, inhibition in response to the conditioned odor would be decreased, leading to stronger odor responses to the conditioned odor.

Reviewer #1:

In this manuscript, the authors describe bidirectional changes in the spine density of olfactory bulb adult-born granule cells (aGCs), that depend on learning paradigms which the authors have established previously.

The used paradigms appear subtle and physiological, since they rely on few odor exposures. Nevertheless, the changes can be observed in electrophysiological recordings in brain slices even though the relation of the investigated mitral cells to odor-activated glomeruli is not known. Thus, the phenomena might reflect a general response of an otherwise deprived olfactory system.

The reviewer rightly points out that relation of the mitral cells recorded in the electrophysiological experiments to the odor-activated glomeruli is unknown. However, most odorants are documented to activate a rather large number of glomeruli (from 15 to 35 glomeruli (Vincis et al., 2012). Previous studies in our lab have shown that Limonene (and other odorants) evokes broad activation of the granule cell layer as reported using Zif268 expression (Mandairon et al., 2008; Moreno et al., 2014). Similarly, the pattern of addition of new neurons to the olfactory bulb show some degree of odor/spatial specificity but nevertheless also covers distributed regions of the olfactory bulb (Mandairon et al., 2006, Alonso et al., 2006, Sultan et al., 2010, Moreno et al., 2012). We thus agree that there is indeed a broad effect of the learning paradigms we used on the bulbar network that could be underlined by broadly distributed odor-responding cells. We also acknowledge that olfactory deprivation of laboratory mice may enhance the contrast between the learning condition and controls.

My general criticism is that there are counterintuitive results that would require stronger statistics to be firmly established. Therefore, the authors should increase sample size, try to sample cells more specifically (see below) and thoroughly discuss their results, possibly requiring a longer format.

As reported above we have substantially increased the number of animals/neurons for most experiments and performed the required better statistical methods (see Table A above).

1) Statistical significance appears to be an issue with some results – would unilateral naris occlusion be feasible, which might allow for pairwise comparisons across hemibulbs? Or a higher number of animals?

Because of contralateral projections between olfactory bulbs via the anterior olfactory nucleus, and the fact that learning after unilateral occlusion transfers to the other side (Guthrie et al., 1990), we would not be confident that the occluded side of the brain would be a good control. Instead, we have performed additional experiments to increase animal/neuron numbers (see Table A above).

While the increased spine density (Figure 1F) looks fine, all the other changes (Figure 1G, Figure 2E) show highly overlapping distributions, perhaps because some MCs in the sample might not have been involved in limonene-evoked bulbar activation. Would it be possible to preferentially record from bulbar parts with strong limonene activation? Or even better fluorescently label MC/GCs expressing IEGs and record from them?

Statistical analysis is revised and performed on the new data sets. All morphological data (including those of Figure 1F, Figure 1G and Figure 2E) have been re analyzed using non parametric Kruskall-Wallis Anova (Statistica) including the four groups (Enr, Non Enr, PC and Cond) and pair-wise comparisons are done by FDR-corrected permutation tests (R). Please also note that data points have been separated on the figures to make their distribution more visible (see also Source data files including all experimental values submitted with the manuscript). The new analysis confirmed the main findings.

As stated above, the odor responding adult-born cells are broadly distributed in the OB and represent an average of ~60% of total number of BrdU labelled cells (BrdU/Zif268 co expressing cells, Figure 1D and Figure 2D) and we agree that this heterogeneity may increase data variability and thereby reduce the chance to reveal an effect of learning rather than the opposite.

Finally, the ultimate experimental way to address this issue could be fluorescently labeled MC or GCs expressing IEG as suggested by the reviewer. However, to our knowledge and appreciation, in the available inducible systems, tagging is irreversible and the time window allowing expression of the IEG (several hours) is still too long to enable specific labelling of the cells responding to the odorant stimulation of interest and not to other olfactory cues. Thus, even though these designs have proven highly valuable tools for investigating memory processes, they seem less appropriate for tracking cells responding directly to sensory stimuli in physiological environment (Reijmers et al., 2007, Guenthner, 2013, Denny et al., 2014).

There is a substantial amount of eIPSC recordings with high frequency in the 30-60 Hz range in PC condition (Figure 2H). This range is not observed in Figure 1J, even not in the enriched case. Were the high frequency responses recorded in slices from the same animal?

For PC (Figure 2H), eIPSCs were recorded from 20 cells in 2 different animals. Among the 7 values above 30 Hz, 2 were recorded from animal 1 and 5 from animal 2. For information, 19 cells from 2 animals were recorded for Cond.

Pre-light data have been added to the figure (new Figure 1—figure supplement 3 and figure 2—figure supplement 3). In order to take into account the factors “light” (pre and post light), “group of learning” (implicit versus explicit) and “learning” (Cond or Enr versus PC or Non-Enr), we normalized the data (ln(x+1)) and performed a 3-factor Anova followed by paired when appropriate or unpaired t-tests.

To strengthen the analysis further, the percentage of “connected” mitral cells, ie cells significantly responding to the light stimulation, has been recalculated based on a statistical analysis of the change in IPSC frequency between pre and post light stimulation (rather than according to an arbitrary threshold in our previous analysis). The latter analysis is now presented in new Figure 1I and Figure 2H.

Strikingly, the very same issue arises also for the sIPSC frequency and amplitude (Figure 2G) – for both there are several data points in the high range for PC that are not observed in non-enr or enr conditions.

Regarding sIPSC frequency (Figure 2G), we have added new animals and recording (see Table A). Total numbers of cells/animals per group now amount to: Non Enr 48/6), Enr 50/6, PC 46/5, Cond 34/3. New data are shown on new Figure 2G, (PC and Cond) and Figure 1H (enr and non–enr).

Statistical analysis now also includes FDR-corrected permutation tests in which 1000 000 artificial groups were tested. The analysis yield a level of significance at p=0.06 (after correction) for the differences in sIPSCs frequency between Cond and PC and between enr and non-enr. Regarding amplitudes, the new analysis confirmed a lower sIPSC amplitude in Cond versus PC. See new Figure 1H and Figure 2G.

If we now doubt the PC data for this reason, what about the statistical comparison between non-enr and cond? Unfortunately, this comparison is missing from Table 1 – by eye between Figure 1F and Figure 2E almost certainly there is no difference for the apical spine density. Which casts doubt on the central finding of reduced connectivity – at least anatomically.

Regarding apical spine density, we performed a Kruskall Wallis Anova including the 4 groups and FDR-corrected permutation tests on the new data set. Importantly, Cond showed a significantly lower apical spine density than non-enr (p=0.013). See new Figure 1F and Figure 2E and new Table 1 in the manuscript that now includes all comparisons.

2) What is the effect of the increase in number of adult born GCs also in the associative learning – how exactly can an addition of interneurons reduce inhibitory drive? As mentioned above, the decrease in spine numbers on the apical dendrite is barely significant. Nevertheless, there is a striking overall decrease in inhibitory drive of mitral cells, especially substantially reduced sIPSC frequency and amplitude (please label the examples from Figure 2G and Figure 1H top in the distributions shown below). If we believe these data (but see above), what is going on here? Synaptic plasticity – reduction in release probability and/or quantal size at the GC-MC contacts – also for preexisting GCs? Were the lost contacts located close to the MC somata and thus the efficiency is strongly reduced?

The apical spine density in the conditioned group is significantly lower than in the pseudoconditionned group, p=0.0017 (Anova including the 4 groups and FDR-corrected permutation tests).

We added representative images (new Figure 1 and Figure 2).

The new adult-born granule cells replace older ones (Imayoshi et al., 2008). Our hypothesis, now developed in the discussion is that “replacing pre-existing granule cells by new ones with fewer synaptic contacts with mitral cells (in conditioned animals) would result in a global pool of granule cells delivering less local inhibition in response to the conditioned odor. In contrast, replacing granule cells by new cells making more local and global synaptic contacts with mitral cells (enriched animals) would result in a shift toward more inhibition in the network. This is consistent with experimental observations.”

3) Missing of important reference: Zhang, Huang and Hu (2016): increased spine density on GC dendrites following odor enrichment in Xenopus tadpoles.

This reference has been added

Reviewer #2:

This report by Mandairon and colleagues is a significant new look on the function of adult born GABAergic neurons in the olfactory bulb. The community as a whole thinks in two ways about the types of learning that occur with passive exposure and conditioning. On the one hand, we all know that these are different types of learning and on the other we often refer to them as if they are the same (under the umbrella term of Learning). Similarly, for learning-based effects on the survival and integration of new granule cells in the OB. This paper shows that the ability to discriminate hard-to-discriminate odors after the two types of learning may depend in opposite fashion on neurogenesis. For passive learning, it appears that adult-born GC connections with MCs are strengthened in the external plexiform layer (at the reciprocal synapse). The authors conclude that this may be related to the effect known as sparsening. For active learning, the effects are opposite. The adult born GC effects on MCs are decreased. The argument is rigorous and rational. My only major concern regards the many pairwise statistical tests (see below). The results reported here, if improved statistical methods confirm them, may constitute one of the most significant findings in the field of olfactory learning as it relates to neurogenesis.

We thank the referee for his/her very positive comment on the manuscript. We reply below to his/her specific comments.

1) There are many more implied comparisons than reported in Table 1. With 4 groups of mice, this makes 6 total comparisons for each factor (cond vs. PC, cond vs. enr, cond vs. non-enr, PC vs enr, PC vs non-enr, enr vs non-enr). With the multiple comparisons within each factor, the threshold for significance should be 0.00833. There is a pretty good argument that all of the factors should be combined in a single analysis, which then multiplies the comparisons and lowers the threshold value, or at least that like-kind factors be grouped (spine density apical/basal as one group, e/sIPSP frequency as another). This would make the p threshold smaller by a factor of 2 at least. Please provide justification in the methods for not performing the multiple comparisons adjustment or do the analysis with the corrected p thresholds. In addition to lowering the threshold, a more conservative analysis might provide additional insight.

Anova including the 4 groups (Cond, PC, Enr, non-Enr) followed by Bonferroni post hoc tests was performed for analyzing the BrdU, BrdU/Zif268, Tbx21/Zif268. For apical spine density, basal spine density and sEPSC frequency and amplitude, non-parametric Anovas followed by FDR-corrected permutation tests were performed. This was done on the new data sets including new animals and/or cell counting and recording.

For eIPSCs, in order to take into account the factors “light” (pre and post light), “group of learning” (implicit versus explicit) and “learning” (Cond or Enr versus PC or Non-Enr), we normalized the data (ln(x+1)) and performed a 3-factor Anova followed by paired or selected unpaired t-tests when appropriate. In addition, individual cell analysis of the effect of light was performed by comparing the number pre and post light IPSC across repetitions of light stimulation by unilateral Chi squared tests. These changes in statistical analyses are now included in the Materials and methods and Results sections.

2) This is not a disagreement but rather a different interpretation. It is the adult born cells that show the effects reported (and assuming that the earlier born cells do not). In the passive condition, the result is that new GCs inhibit MCs more at the apical dendrites where primary processing happens. Also, we know that these new cells are relatively specific to the enriched odors (previous work by the first and second to last authors). Is it possible that these new connections serve to help the mice ignore the conditioned stimuli and that the enhanced discrimination ability is a side-effect of this now-active ignorance? On the other hand, in the case of active learning there are more cells born and integrated (perhaps associated with the learned odors). Because these cells survive, one assumes that the inputs from higher order areas are strong. The signal is amplified, so one might expect that the now meaning-based odor perception is accomplished in the AON or PC or both, rather than in the OB.

We thank to the referee for his/her stimulating comment. Although the proposed hypothesis cannot be ruled out based on our data, and if we understood well, we could argue that during enrichment, we previously reported that the locus Coeruleus is solicited (Rey et al., 2011, Moreno et al., 2012) suggesting that daily olfactory stimulation triggers arousal, a notion is rather opposite to ignoring stimuli. Now, whether “active” ignorance is a component of arousal remains an open question. Regarding associative learning, we agree that the AON or PC may play an important part in perceiving the meaning odorants. However, the OB is still crucial to implement this amplification as suggested by the present data, and to maintain it since forgetting correlates to the loss of adult-born cells (Sultan et al., 2010, Sultan et al., 2011 and present data).

Reviewer #3:

Mandairon and colleagues examined the effects of passive (odor exposure, habituation) and active (odor reinforced with reward) olfactory learning on granule cell neurogenesis, apical and basal dendritic spines of adult-born granule cells and mitral cell activity and physiology. The authors find/claim that passive learning increased neurogenesis and consequently, increased inhibitory drive onto mitral cells (assessed using ChR2 stimulation of adultborn GCs) and decreased odor-evoked activity of the mitral cells. In contrast, the authors show that active learning produces a decrease in apical dendritic spines without affecting basal spines. Interestingly, the decrease in apical dendritic spine density was reversed with restoration of discrimination levels to baseline over time. Furthermore, odor-evoked activity in MCs was increased and ChR2 light evoked IPSC frequency of the mitral cells was decreased.

This is an interesting study that begins to delineate how different kinds of learning affect GC-MC connectivity. Because passive and active learning may affect inputs and outputs of adult born and developmentally generated GCs and mitral cells (along with physiological properties of mitral cells), it is not clear how adultborn GCs are solely driving changes in inhibition onto mitral cells under these different learning conditions.

For example, does blockade of adult GC neurogenesis under active or passive learning conditions (over the learning period) eliminate the reported effects on mitral cells?

From our previous work on perceptual learning, we know that blocking neurogenesis abolishes the increase in paired-pulse inhibition of mitral cells and GAD expression in the OB (Moreno et al., 2009). In addition, blocking neurogenesis (Moreno et al., 2009) or structural plasticity in adultborn cell (Daroles et al., 2015) prevents implicit learning. Regarding associative learning, we have shown that blocking neurogenesis during learning altered recall of the discrimination task (5 days post learning, a delay similar to that in the present study). Finally, in the present work, we show that the morphological changes in spine density on granule cells return to pre-learning level after one month, consistent with the loss of the behavioral response. However, while these evidences favor a prominent role of adult born neurons in shaping mitral cell activity in learning, the role of pre-existing interneurons in mitral cell response plasticity remains to be investigated.

1) It is not clear why Mann-Whitney was used over a one-way ANOVA for comparisons of apical or basal dendritic spine density.

The statistical analysis was entirely revised. Anova followed by FDR-corrected permutation tests were performed. (See above, reply to editor comments and to referees #1 & #2.)

2) Figure 1G: Effect on basal spine density appears to be driven by one datapoint.

We have added new data in this data set (more neurons analyzed) and statistical analysis was performed by non-parametric Anova followed by FDR-corrected permutation tests:

-basal spine density differed between Enr and non-enr (Figure 1G) (Kruskall Wallis H(3, 187)=20.15, p<0.0001, enr versus non enr p=0.0015)

-PC did not differed from Cond (PC versus Cond p=0.09).

Does basal spine density return to baseline after 42 days?

Data regarding basal spine density after 42 days has been added to the manuscript. Basal spine density does return to pre learning value (new graph in Figure 1—figure supplement 2).

3) How do authors control for number of cells expressing ChR2 virus or normalize counts to interpret the ChR2 stimulation data (number of spines vs. number of cells)? Are equivalent numbers of cells infected in both groups? The increase in survival induced by learning should bias the number of Chr2 expressing GCs in OB.

In line with the reviewer’s comment and according to our BrdU counts (Figure 1 and Figure 2D), we agree that enrichment or conditioning should result in an increased number of ChR2+ cells (we did not count these cells on the slices used for electrophysiology because the thickness of the sections (400 µm) makes it difficult in our view to provide reliable cell count). This leads to the conclusion that the increased inhibition on M/T cells following implicit learning could be accounted for by an increased number of adult-born cells. This issue is now discussed in the manuscript (Discussion section).

However, this does not explain how the very same number ChR2 GCs turns out to produce decreased inhibition on M/T cells in animal subjected to explicit learning. This is why we think that learning induces additional plasticity (such as changes in apical spine density onto M/T cells).

This view is supported by our previous work (Daroles et al., 2016) showing that blockade of the increase in spine density is sufficient to prevent perceptive learning.

4) Following reversal of changes in apical spine density (but with extra adultborn GCs still present), is the effect of ChR2 stimulation on MCs lost?

Extra adult-born GCs expressing BrdU are no longer present 42 days after explicit learning as shown in a previous publication from our lab (Sultan et al., 2010). We now added new data to the present paper showing that extra adult-born GCs are no longer present 42 days after implicit learning, at a time where changes in apical spine density and behavioral discrimination are no longer observed (new Figure 1—figure supplement 2).

5) Although the authors examine activity of mitral cells (sparseness), evidence examining population based coding or input-output transformations is critically needed to justify the use of term pattern separation or interpret the data within this framework.

We agree that pattern separation strictly refers to the process of input-output transformation and we have modified the discussion accordingly.

[Editors' note: the author responses to the re-review follow.]

Reviewer #1:

This is a much improved paper that shows a striking difference in inhibition provided by adult-born GCs depending on the type of learning.

Is it possible that some part of the character of the effect is because of the odor similarity? Might it look different if the odors were easier to discriminate/less overlapping? I suggest the authors qualify the results under the class of fine odor discrimination. A nice follow-up study (not for this paper) would be to compare in each condition (implicit vs explicit), or at least the explicit condition, the effect of discrimination difficulty.

Done.

Reviewer #2:

I recommend the manuscript for publication.

Please discount the significance of the Daroles et al. study as evidence supporting a role for an increase in spine density in implicit learning. This is because FMRPcKO mice show an elevation in spine density at baseline (Figure 3) and therefore, it is not clear whether the failure to increase spine density further following learning or the elevation prior to learning is the culpable factor. Additionally, FMRP has numerous functions within neurons, that when disrupted, may be responsible for behavioral phenotype in implicit learning.

This part of the sentence has been removed from the Discussion section.

The authors acknowledge "However, while these evidences favor a prominent role of adult born neurons in shaping mitral cell activity in learning, the role of pre-existing interneurons in mitral cell response plasticity remains to be investigated". However, this possibility is absent from discussion. Please address this concern in addition to acknowledging "potential changes in mitral cell properties' also as a potential mechanism.

This sentence has been added to the Discussion section.

https://doi.org/10.7554/eLife.34976.019

Article and author information

Author details

  1. Nathalie Mandairon

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Investigation, Methodology, Writing—original draft, Writing—review and editing
    For correspondence
    nathalie.mandairon@cnrs.fr
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7935-5215
  2. Nicola Kuczewski

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Data curation, Formal analysis, Writing—original draft
    Competing interests
    No competing interests declared
  3. Florence Kermen

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Data curation, Formal analysis, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Jérémy Forest

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  5. Maellie Midroit

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  6. Marion Richard

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Methodology, Writing—original draft
    Competing interests
    No competing interests declared
  7. Marc Thevenet

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Formal analysis, Methodology
    Competing interests
    No competing interests declared
  8. Joelle Sacquet

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Data curation, Formal analysis, Methodology
    Competing interests
    No competing interests declared
  9. Christiane Linster

    1. Computational Physiology Lab, Cornell University, Ithaca, United States
    2. Department of Neurobiology and Behavior, Cornell University, Ithaca, United States
    Contribution
    Conceptualization, Writing—original draft
    Competing interests
    No competing interests declared
  10. Anne Didier

    Lyon Neuroscience Research Center, Neuroplasticity and Neuropathology of Olfactory Perception Team, CNRS UMR 5292, INSERM U1028, Université de Lyon, Lyon, France
    Contribution
    Conceptualization, Formal analysis, Supervision, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3118-9961

Funding

Centre National de la Recherche Scientifique

  • Nathalie Mandairon

Université Claude Bernard Lyon 1

  • Nathalie Mandairon

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by the CNRS, Inserm, and Lyon 1 University. We would like to thank K Deisseroth for the gift of the channelrhodopsin construct, C Benetollo from the Neurogenetic and Optogenetic Platform of the CRNL for lentiviral production and G Froment, D Nègre and C Costa from the lentivector production facility/SFR BioSciences de Lyon (UMS3444/US8). We thank Y Yoshihara (Riken Brain Science Institute, Saitama Japan) for the gift of the Tbx21 antibody. We thank I Caillé for her comments and S Scotto-Lomasesse for her help with the morphological analysis. We would like to thank Y Chelminski, A Auguste and A Ferréol for helping during behavioral procedure and cell morphology analysis.

Ethics

Animal experimentation: Experiments were done following procedures in accordance with the European Community Council Directive of 22nd September 2010 (2010/63/UE) and the National Ethics Committee (Agreement DR2013-48 (vM)). Every effort was made to minimize suffering.

Reviewing Editor

  1. Naoshige Uchida, Harvard University, United States

Publication history

  1. Received: January 10, 2018
  2. Accepted: February 12, 2018
  3. Version of Record published: February 28, 2018 (version 1)

Copyright

© 2018, Mandairon et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 749
    Page views
  • 162
    Downloads
  • 5
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)