The Tis21 knockout mice were of the C57BL/6 (B6) strain and had a replacement of the entire exon II of the Tis21 gene. Mutant mice had been generated previously, as described (Park et al., 2004). Genotyping of mice was routinely performed by polymerase chain reaction (PCR), using genomic DNA from tail tips, as described (Farioli-Vecchioli et al., 2009). Mice were maintained under standard specific-pathogen-free conditions; all animal procedures were carried out on male mice and completed in accordance with the Istituto Superiore di Sanita’ (Italian Ministry of Health; authorizations DM 442-2016-PR and DM 04/2013) and current European (directive 2010/63/EU) Ethical Committee guidelines.

Differentiated (28-day-old) neurons in the dentate gyrus were detected in Tis21 wild-type and knockout mice after treatment with five daily i.p. injections of bromodeoxyuridine (BrdU; 95 mg/kg b.wt., 10 μl/g b.wt.; Sigma–Aldrich, S.Louis, MO, USA) from P60 to P64, followed by perfusion at P88. In Morris water maze (MWM) experiments BrdU was administered at P60 (two daily injections, 95 mg/kg i.p) and analysis was performed 21 days later. Fluoxetine (10 mg/Kg b.wt., 10 μl/g b.wt.; Tocris Bioscience, Bristol, United Kingdom) or vehicle in control groups (sterilized water) was administered daily by i.p. injection to 2-month-old mice for 18 days (in proliferation and differentiation experiments) or, in experiments aimed to measure the Ts length, to 39-day-old mice for 21 days. Treatments are summarized in Supplementary Table S1. Brains were collected after transcardiac perfusion with 4% paraformaldehyde (PFA) in PBS—DEPC and kept overnight in PFA. Brains of 14-day-old mice were directly fixed by immersion in PFA. Afterwards, brains were equilibrated in sucrose 30% and cryopreserved at −80°C.

Immunohistochemistry was performed on serial free-floating coronal sections cut at 40 μm thickness in a cryostat at −25°C from brains embedded in Tissue-Tek OCT (Sakura, Torrence, CA, USA). The sections were then processed immunohistochemically for multiple labeling with BrdU and other cellular markers using fluorescent methods.

Proliferating subpopulations of progenitor cells were identified using a goat polyclonal antibody against DCX (Santa Cruz Biotechnology, Santa Cruz, CA, USA; SC-8066; 1:300) and a mouse monoclonal against nestin (Millipore, Temecula, CA, USA; MAB353; 1:100).

BrdU incorporation was visualized by denaturing DNA through pretreatment of sections with 2N HCl 45 min at 37°C, followed by 0.1 M sodium borate buffer pH 8.5 for 10 min. The sections were then incubated overnight at 4°C with a rat monoclonal antibody against BrdU (AbD Serotech, Raleigh, NC, USA; MCA2060; 1:400), together with other primary antibodies, as indicated: rabbit polyclonal antibody against Calretinin (Swant, Bellinzona, Switzerland; 7699/4, 1:200), mouse monoclonal antibody raised against NeuN (Millipore, Temecula, CA, USA; MAB377; 1:300).

Proliferating, immature and terminally differentiated neurons infected with retroviruses were visualized by means of chicken polyclonal antibody against GFP (green fluorescent protein; Aves Labs, Tigard, OR, USA; GFP-1010; 1:100) together with either a rabbit monoclonal antibody against Ki67 (Thermo Scientific, Fremont, CA, USA; SP6; 1:150), or a rabbit polyclonal antibody against Calretinin (Swant, Bellinzona, Switzerland; 7699/4, 1:200) and a mouse monoclonal antibody raised against NeuN (Millipore, Temecula, CA, USA; MAB377; 1:300). The NeuroD1 antibody used in cell cycle studies was a goat polyclonal (R&D Systems, Minneapolis, MN, USA; AF2746; 1:200).

Secondary antibodies used to visualize the antigen were all, with the exception of the secondary anti-GFP, from Jackson ImmunoResearch (West Grove, PA, USA) as follows: a donkey anti-rat antiserum conjugated to tetramethylrhodamine isothiocyanate (TRITC) (BrdU), a donkey anti-rabbit antiserum conjugated to Cy2 (calretinin), a donkey anti-mouse antiserum conjugated to Alexa-647 (NeuN), a donkey anti-rabbit antiserum Cy3-conjugated (Ki67 and calretinin), a donkey anti-mouse Cy2-conjugated (nestin) and a donkey anti-goat Alexa 647-conjugated (DCX or NeuroD1). The secondary antibody to visualize GFP was an anti-chicken antiserum fluorescein-conjugated (Aves Labs). Nuclei were counterstained by Hoechst 33258 (Sigma–Aldrich; 1 μg/ml in PBS). Supplementary Table S1 summarizes the cell staining used in the different experiments.

Images of the immunostained sections were obtained by laser scanning confocal microscopy using a TCS SP5 microscope (Leica Microsystem) equipped with Laser Diode 405, Ar 488, HeNe 543, HeNe 633. Analyses were performed in sequential scanning mode to rule out cross-bleeding between channels and using combinations of three different lasers simultaneously.

For BrdU detection, see “Immunohistochemistry” Section above. Concerning IdU and CldU immunohistochemistry, sections were washed five times in TBS (Tris-buffered saline), pretreated with 2N HCl for 10 min, then washed five times in TBS. Antibodies were incubated in TBS containing 5% normal donkey serum and 0.25% Triton X-100 (Vega and Peterson, 2005).

Halogenated thymidine analogs were detected as follows: for IdU using the mouse monoclonal anti-BrdU (BD Biosciences, San Jose, CA, USA; BD 44; 1:500); for CldU using the rat monoclonal anti-BrdU (AbD Serotec, Raleigh, NC, USA; MCA2060; 1:250). Antibody specificity was analyzed in mice that were injected with either IdU or CldU alone. Secondary antibodies used to visualize the antigens were a donkey anti-rat antiserum conjugated to TRITC (CldU) and a donkey anti-mouse antiserum conjugated to Cy2 (IdU), all from Jackson ImmunoResearch (West Grove, PA, USA).

S-phase length, measured in hours (Ts), was obtained employing the equation developed by others (Vega and Peterson, 2005; Brandt et al., 2012). In order to measure Ts in 2-month-old and 2-week-old mice, we administrated IdU intraperitoneally (57.5 mg/kg in 0.02 N NaOH, 0.9% NaCl saline solution; MP Biomedicals, Solon, OH, USA), and 3 h later CldU (42.5 mg/kg b.wt. in 0.9% NaCl saline solution; Sigma–Aldrich). The injected volumes were adjusted to the body weight. Forty-five minutes after CldU, injected animals were killed for immunohistochemistry. The Ts of the NeuroD1⁺ progenitors was obtained through triple immunofluorescence (NeuroD1⁺IdU⁺CldU⁺).

Stereological analysis of the number of cells was performed on one-in-eight series of 40-μm free-floating coronal sections (320 μm apart), which were analyzed by confocal microscopy to count cells expressing the indicated marker throughout the whole rostrocaudal extent of the dentate gyrus. To obtain the total estimated number of cells within the dentate gyrus, positive for each of the indicated markers, the average number of positive cells per section was multiplied by the total number of 40-μm sections comprising the entire dentate gyrus (about 50–60 sections), as described (Gould et al., 1999; Jessberger et al., 2005; Kee et al., 2007; Farioli-Vecchioli et al., 2008, 2012). Thus, about seven sections per mouse and at least three animals per group were analyzed. Confocal single plane images and Z-stacks with orthogonal projections of the immunostained sections were obtained using a TCS SP5 confocal laser scanning microscope (Leica Microsystems).

In Figures 3D, 4D (fluoxetine and MWM experiments, respectively) cell numbers were calculated as ratios of stage 5 or stage 6 neurons to the total number of BrdU⁺ cells. The virus-infected cells (Figures 5–7) were calculated as percentage ratios between GFP⁺Ki67⁺ or GFP⁺Calretinin^±NeuN⁺ cells and the total number of infected cells (GFP⁺) in each section, and then averaged. In each virus-injected mouse were analyzed at least 15 sections throughout the whole extent of the dentate gyrus; the minimum number of GFP⁺ cells identified per mouse was 12. Cell number analyses were performed manually by trained experimenters using the I.A.S. software to record positive cells (Delta Sistemi, Rome, Italy).

The retroviral vector pCAG-IRES-GFP, kindly provided by Dr. Chichung Lie (Institute of Developmental Genetics, Germany; Jessberger et al., 2008), was used to express only in dividing neural cells the cDNA of Tis21 (i.e., the murine sequence) and NeuroD2, as well as the shId3-190 RNA interfering sequence. The construct pCAG-IRES-GFP-Tis21 had been generated as previously described (Farioli-Vecchioli et al., 2014a). The construct pCAG-IRES-GFP-shId3 was generated by cloning in the SfiI-5′/PmeI-3′ sites of pCAG-IRES-GFP the whole 64-mer containing the 19-nucleotide siRNA from the pSUPER.retro-neo-GFP-shId3-190 vector, preceded by the H1 promoter and the TATA box sequence driving the transcription. pSUPER.retro-neo-GFP-shId3-190 was previously generated by us and the 19-nucleotide siRNA sequence specific to mouse Id3 was shown to efficiently reduce the Id3 mRNA levels (Farioli-Vecchioli et al., 2014a).

The same procedure was used to obtain the pCAG-IRES-GFP-shLUC retroviral vector, containing an sh-LUC control sequence specifically targeting the luciferase gene (Farioli-Vecchioli et al., 2014a). The pCAG-IRES-GFP-NeuroD2 sequence was obtained by cloning the full open reading frame of NeuroD2 mouse cDNA in the sites SfiI-5′/PmeI-3′ of pCAG-IRES-GFP. The shId3, shLUC and NeuroD2 sequences cloned into pCAG-IRES-GFP vector had been synthesized by MWG (Ebersberg, Germany) and were checked by DNA sequencing.

The different retroviruses, including pCAG-IRES-GFP-shLUC, were propagated as previously described (Farioli-Vecchioli et al., 2008). The concentrated virus solution (10⁸ pfu/ml) was infused (1.5 μl at 0.32 μl/min) by stereotaxic surgery into the right and left dentate gyrus of anesthetized P60 Tis21-null mice (anteroposterior = −2 mm from bregma; lateral = ±1.5 mm; ventral = 2.0 mm). Infected, GFP-positive cells were counted throughout the whole extent of the dentate gyrus.

All the retroviruses generated were amphotropic and replication-deficient. Their manipulation and stereotactic injection in mice were approved by the Italian Ministry of Health (authorizations RM/IC/Op2/06/008 and 04/2007) and performed using BSL-2 and ABSL-2 containments.

For the training in the MWM, the same protocol as in Farioli-Vecchioli et al. (2009) was used, with minor modifications. The training was carried out in a circular swimming pool of 1.3 m in diameter, filled with opaque (white) water kept at constant temperature (25 ± 1°C). The apparatus was located in a room containing prominent extra-maze cues. A hidden 15-cm-diameter platform was used. We measured the time employed by the mouse to reach the hidden platform (s). The training consisted of 28 trials (four trials per day, lasting a maximum of 60 s, with an inter-trial interval of 30 min), with the platform maintained in the same position. Mice were left for 15 s upon the platform at the end of each trial. In the event a mouse did not reach the platform within 60 s, it was gently directed to the platform by the experimenter and left for 15 s upon it (this occurred only sporadically on the first day). No probe test was performed. Only male mice (2 months of age; at least five per group) were used for the experiment. Mice behavior was recorded and analyzed offline by the EthoVision software (Noldus Information Technology, Wageningen, Netherlands).

Two-way ANOVA was used to compare the effects in all experiments on wild-type and knockout mice of fluoxetine treatment (including cell cycle length) and of MWM training; individual between-group comparisons, where appropriate, were carried out by Fisher’s PLSD ANOVA post hoc test. The variance of data in the fluoxetine and MWM experiment that were calculated as ratio of differentiated neurons (stage 5 or stage 6) to the total number of bromodeoxyuridine⁺ (BrdU⁺) cells (Figures 3D, 4D), were instead analyzed with the Kruskall-Wallis test, which accounts for the assumption of non-normal distribution; individual between-group comparisons where then performed with the non parametric Mann-Whitney U test which does not require the assumption of normal distribution. Each experimental group analyzed was composed of at least three animals. Mann-Whitney U test was also used to analyze the percent values of retrovirally infected dentate gyrus cells, as the distribution of percent data may not comply with the assumption of a normal distribution. These analyses were performed using the StatView 5.0 software (SAS Institute, Cary, NC, USA). Supplementary Table S1 summarizes mice number, experimental design and statistical tests used for each experiment. Supplementary Table S2 summarizes two-way ANOVA and post hoc analyses.

Differences were considered statistically significant at p < 0.05. All data were expressed as mean values ± SEM.

As a first step, we checked in our system the known ability of fluoxetine to induce the proliferation of progenitor cells of the dentate gyrus (Encinas et al., 2006; Crowther and Song, 2014; Bolijn and Lucassen, 2015). We treated 2-month-old mice with fluoxetine daily for 18 days, and then we analyzed the ongoing proliferation of dentate gyrus cells. We observed that the treatment with fluoxetine significantly increased the total number of proliferating stem and progenitor cells in the dentate gyrus (detected as Ki67⁺) in wild-type as well as in mutant mice, relative to the respective control-treated (with water) mice (Tis21^WT vs. Tis21^WT + FLX, 82% increase, p < 0.0001 PLSD test; Tis21^KO vs. Tis21^KO + FLX, 40% increase, p = 0.0003 PLSD test; two-way ANOVA, fluoxetine treatment effect F_(1,143) = 31.8, p < 0.0001; Figures 1A–C). Moreover, as expected, the knockout of Tis21 in itself led to increased proliferation of the total proliferating cells (Tis21^WT vs. Tis21^KO, 55% increase, p = 0.005; Figures 1A,C). We further checked the increase of proliferation in the subpopulations of stem/progenitor cells. Type-1-2a cells were identified as nestin⁺DCX⁻, type-2b progenitor cells as nestin⁺ and DCX⁺, while type-3 cells as DCX⁺and nestin⁻ (Filippov et al., 2003; Kronenberg et al., 2003; Kempermann et al., 2004). We observed that in wild-type as well as in mutant mice the number of proliferating type-2b and type-3 progenitor cells was increased by fluoxetine treatment (type-2b, Ki67⁺/nestin⁺/DCX⁺ cells: Tis21^WT vs. Tis21^WT + FLX, 55% increase, p = 0.007; Tis21^KO vs. Tis21^KO + FLX, 37% increase, p = 0.016 PLSD test, two-way ANOVA, treatment effect F_(1,143) = 13.2, p = 0.0004; type-3, Ki67⁺/nestin⁻/DCX⁺ cells: Tis21^WT vs. Tis21^WT + FLX, 73% increase, p = 0.013 PLSD test; Tis21^KO vs. Tis21^KO + FLX, 43% increase, p = 0.002 PLSD test, two-way ANOVA, treatment effect F_(1,143) = 15.6, p = 0.0001; Figures 1A,D).

Then, we analyzed whether fluoxetine was in itself able to enhance the proliferation of progenitor cells of the dentate gyrus by modifying the length of the cell cycle. For this, we calculated the S phase length (Ts)—which is the main determinant of cell cycle duration—using a new accurate procedure (Brandt et al., 2012; Farioli-Vecchioli et al., 2014b; Figures 2A–D). We treated 39-day-old mice with fluoxetine daily for 21 days; this protocol was used in order to maximize the proliferative effects of fluoxetine. It turned out that fluoxetine treatment was unable to alter the Ts of the wild-type proliferating progenitor cells in 2-month-old mice, whereas Ts was significantly shortened in Tis21 knockout mice treated with fluoxetine, relative to either Tis21 knockout mice control-treated or wild-type control-treated (Tis21^KO vs. Tis21^KO + FLX, p = 0.02 PLSD test, two-way ANOVA, treatment effect F_(1,164) = 2.04, p = 0.15; Tis21^WT vs. Tis21^KO + FLX, p = 0.033 PLSD test; Figures 2A–C). We also checked whether fluoxetine affected the Ts of the subpopulation of progenitor cells at the threshold of differentiation, i.e., the type-3 neuroblasts, which express NeuroD1 (Roybon et al., 2009). These cells in fact, since they are at a pre-differentiation stage, may represent an interesting target for our analysis. However, no significant change of Ts induced by fluoxetine was observed in NeuroD1⁺ cells (two-way ANOVA, treatment effect F_(1,102) = 1.10, p = 0.29; Figure 2D). Moreover, the ablation of Tis21 in itself did not alter Ts in 2-month-old dentate gyrus progenitor cells, relative to wild-type mice (Figures 2C,D). However, considering that Tis21, in addition to the intrinsic pro-differentiative function, has antiproliferative properties (Farioli-Vecchioli et al., 2009), it seemed appropriate to test whether the deletion of Tis21 could in itself modify the cell cycle length at an earlier age —2 weeks— when the proliferation in the dentate gyrus is stronger (Gilley et al., 2011). We found that indeed the Ts of the total population of stem/progenitor cells in 14-day-old Tis21 knockout mice resulted significantly shorter than in wild-type mice (about 19% decrease, p = 0.0005; Figures 2E,F). Furthermore, a significant decrease was observed also for the Ts of the subpopulation of NeuroD1⁺ progenitor cells (21% decrease, p = 0.023; Figures 2E,G).

Taken together, these data indicated that: (i) fluoxetine enhances the proliferation of the adult wild-type dentate gyrus progenitor cells, but is able to accelerate the S phase only in mice ablated of Tis21 (probably because of a permissive condition consequent to the absence of the antiproliferative Tis21); and that (ii) the knockout of Tis21 in itself causes an acceleration of the S phase although detectable only in young mice.

Having ascertained that fluoxetine exerts a powerful proliferative effect in mice of about 2-months of age on Tis21 knockout and wild-type dentate gyrus progenitor cells, we studied the effect of fluoxetine treatment on their terminal differentiation.

P60 mice were treated with fluoxetine for 18 days and injected with BrdU daily for the last 5 days of fluoxetine treatment. Dentate gyrus cells were analzyed at the end of the treatment, in this way monitoring essentially the proliferative effect of fluoxetine on progenitor cells. We observed that, although fluoxetine induced in both wild-type and knockout mice an increase of the number of immature and terminally differentiated neurons, fluoxetine was not able to restore their rate of differentiation (measured over the total number of BrdU⁺ cells) in mutant mice (data not shown and Supplementary Table S1).

Thus, we modified the protocol of treatment by adding at the end an off treatment period of 10 days, in order to allow the differentiation of the newly generated progenitor cells, and to better assess whether fluoxetine is able to accelerate the differentiation of progenitor cells in mutant mice (Supplementary Figures S1A–C).

We observed that fluoxetine significantly increased in both wild-type and knockout mice the number of stage 5 (BrdU⁺/Calretinin⁺/NeuN⁺, hereafter named as BrdU⁺/stage 5, Tis21^WT vs. Tis21^WT + FLX, p = 0.02 PLSD test; Tis21^KO vs. Tis21^KO + FLX, p = 0.009 PLSD test, two-way ANOVA, treatment effect F_(1,121) = 11.74, p = 0.0008; Supplementary Figure S1C) and stage 6 neurons (BrdU⁺/Calretinin⁻/NeuN⁺, hereafter named as BrdU⁺/stage 6, Tis21^WT vs. Tis21^WT + FLX, p = 0.008 PLSD test; Tis21^KO vs. Tis21^KO + FLX, p = 0.005 PLSD test, two-way ANOVA, treatment effect F_(1,121) = 15.23, p = 0.0002; Supplementary Figure S1C).

Notably, while the rate of differentiation after fluoxetine, measured as the ratio of terminally differentiated neurons to the total number of BrdU⁺ 10- to 15-day-old neurons, was restored to normal values in stage 5 neurons, the rate of stage 6 neurons remained lower in mutant mice relative to wild-type, indicating that no rescue of the defect of terminal differentiation occurred (BrdU⁺/stage 6 in total BrdU⁺: Tis21^WT + FLX vs. Tis21^KO + FLX, p < 0.0001, Mann-Whitney U test, Kruskall-Wallis (d.f. 3) H = 45.42, p < 0.0001; Supplementary Figure S1D).

Next, we decided to modify the protocol by anticipating BrdU injections at the beginning of the treatment, in order to detect the initial amplification of the different cohorts of neurons throughout the process of maturation during the treatment with fluoxetine, which is also endowed with a prodifferentiative action (Wang et al., 2008). We left the off treatment period of 10 days at the end, to allow the complete differentiation of the newly generated progenitor cells. Such protocol would highlight the effect of fluoxetine treatment not only on the proliferation and differentiation but also on the survival of progenitor cells (Encinas et al., 2006).

As a preliminary test, we checked whether the new conditions, with five daily injections of BrdU from the first day of treatment, allow to detect the proliferative effect of fluoxetine (Supplementary Figures S2A–C). In fact, a previous study indicated that an increase of BrdU⁺ cells by fluoxetine was detected between 5 days and 14 days of treatment, after, however, a single injection of BrdU before treatment (Malberg et al., 2000). We observed that BrdU⁺ cells are significantly increased by fluoxetine treatment for 7 days (BrdU⁺, Tis21^WT vs. Tis21^WT + FLX, one-way ANOVA F_(1,70) = 4.6, p = 0.035; Supplementary Figure S2C), plausibly before a fluoxetine-dependent survival action may take place (Caiaffo et al., 2016). Therefore, our protocol appears to be suited to detect the effect of fluoxetine on both proliferation and differentiation of progenitor cells, with also a prosurvival component that certainly plays a part during the treatment.

Thus, we induced the process of neurogenesis by treating P60 mice with fluoxetine for 18 days, followed by a period of 10 days in the absence of treatment, to allow the differentiation of the newly generated progenitor cells. New neurons were labeled by five daily injections of BrdU performed during the first 5 days of fluoxetine treatment (Figures 3A,B; see scheme 3B). This protocol, from Wang et al. (2008) with modifications, allows detection of new 23- to 28-day-old differentiated neurons, generated with or without the neurogenic stimulus of fluoxetine.

We observed an increase of immature stage 5 neurons in Tis21 knockout mice, relative to wild-type (BrdU⁺/stage 5 neurons, p = 0.028 PLSD test; Figures 3A,C) and a decrease of terminally differentiated stage 6 neurons (BrdU⁺/stage 6 neurons, p = 0.03; Figures 3A,C), as expected. Fluoxetine significantly increased in wild-type and in mutant mice the number of differentiated neurons, either immature (BrdU⁺/stage 5, Tis21^WT vs. Tis21^WT + FLX, p = 0.007 PLSD test; Tis21^KO vs. Tis21^KO + FLX, p = 0.0013 PLSD test, two-way ANOVA, treatment effect F_(1,80) = 18, 466, p < 0.0001; Figure 3C), or terminally differentiated (BrdU⁺/stage 6, Tis21^WT vs. Tis21^WT + FLX, p < 0.0001 PLSD test; Tis21^KO vs. Tis21^KO + FLX, p = 0.0004 PLSD test, two-way ANOVA, treatment effect F_(1,79) = 72.355, p < 0.0001; Figure 3C). The increase of 28-day-old neurons confirms that the cell cycle acceleration of progenitor cells induced by fluoxetine enhances the generation of viable neurons. Fluoxetine treatment, however, was unable to functionally counteract the impairment of differentiation occurring in Tis21 knockout mice, as the number of terminally differentiated neurons (BrdU⁺/stage 6) observed in mutant mice after fluoxetine treatment remained significantly lower than in wild-type mice treated with fluoxetine (Tis21^KO + FLX vs. Tis21^WT + FLX, p < 0.0001 PLSD test; Figure 3C).

Moreover, if we visualize in Figure 3D the rate of differentiation by analyzing the ratio of terminally differentiated neurons (BrdU⁺/stage 6) to the total number of BrdU⁺ 23- to 28-day-old neurons, this was slightly but significantly changed by fluoxetine in wild-type, indicating a mild pro-differentiative effect by fluoxetine. However, again, no change by fluoxetine occurred in the ratio of immature or terminally differentiated neurons of mutant mice (BrdU⁺/stage 5 in total BrdU⁺: Tis21^WT vs. Tis21^WT + FLX, p = 0.14 or Tis21^KO vs. Tis21^KO + FLX, p = 0.09 Mann-Whitney U test, Kruskall-Wallis (d.f. 3) H = 38.92, p < 0.0001; BrdU⁺/stage 6 in total BrdU⁺: Tis21^WT vs. Tis21^WT + FLX, p = 0.017 or Tis21^KO vs. Tis21^KO + FLX, p = 0.30 Mann-Whitney U test, Kruskall-Wallis (d.f. 3) H = 30.96, p < 0.0001; Figure 3D). In fact, the pattern of increase of stage 5 neurons and the decrease of stage 6 neurons elicited by Tis21 knockout, relative to wild-type mice, remained the same also after fluoxetine treatment (BrdU⁺/stage 5 in total BrdU⁺: Tis21^WT + FLX vs. Tis21^KO + FLX, p < 0.0001, Mann-Whitney U test; BrdU⁺/stage 6 in total BrdU⁺: Tis21^WT + FLX vs. Tis21^KO + FLX, p < 0.0001, Mann-Whitney U test; Figure 3D).

As a whole, this indicates that the defect of terminal differentiation of stage 5 into stage 6 neurons in Tis21 knockout mice is refractory to rescue by the proliferative and pro-differentiative neurogenic stimulus of fluoxetine, and suggests that this process is separate from the expansion of progenitor cells.

As a second step, we tested whether a different neurogenic stimulus —namely, spatial learning in the MWM— could rescue the differentiation defect of Tis21 knockout dentate gyrus. It has in fact been demonstrated that spatial learning activates neurogenesis and promotes the survival and incorporation of new neurons into the dentate gyrus (Gould et al., 1999). The neurogenic effect of spatial learning appears more pronounced in 1-week-old neurons, as determined by BrdU birth-dating (Epp et al., 2013). We decided to highlight the process of neuron differentiation, by analyzing the new neurons generated 21 days after two daily injections of BrdU, whereas the MWM training was performed from day 8 to day 14, when the proliferative phase was almost concluded (Figures 4A,B, see scheme 4B). This timeline was chosen following, with modifications, published protocols (Dupret et al., 2007; Epp et al., 2007). The experiments were performed in 2-month-old mice.

The MWM was performed as described in Farioli-Vecchioli et al. (2009) with minor modifications. In this task, mice learn across daily sessions to find a hidden escape platform using extra-maze visual cues. Tis21^KO and Tis21^WT mice performed equally in the task (Supplementary Figure S3). MWM training elicited in the dentate gyrus of both wild-type and Tis21 knockout mice a significant increase of immature neurons (BrdU⁺/stage 5, Tis21^WT vs. Tis21^WT + MWM, p = 0.042; Tis21^KO vs. Tis21^KO + MWM, p < 0.0001 PLSD test, two-way ANOVA, MWM effect F_(1,270) = 27.17, p < 0.0001; Figure 4C), as well as of terminally differentiated neurons (BrdU⁺/stage 6, Tis21^WT vs. Tis21^WT + MWM, p < 0.002 PLSD test; Tis21^KO vs. Tis21^KO + MWM, p = 0.011 PLSD test, two-way ANOVA, MWM effect F_(1,270) = 15.35, p = 0.0001; Figure 4C). Notably, after MWM training the difference between the number of wild-type and Tis21 knockout terminally differentiated neurons was reduced to a non-significant level, suggesting that some rescue of terminal differentiation occurred (BrdU⁺/stage 6, Tis21^WT + MWM vs. Tis21^KO + MWM, p = 0.070 PLSD test; Figure 4C). Nevertheless, when we analyzed the ratio of stage 5 or 6 neurons to the total number of BrdU⁺ neurons, no change was exerted by MWM training (BrdU⁺/stage 5 in total BrdU⁺: Tis21^WT vs. Tis21^WT + MWM, p = 0.87 or Tis21^KO vs. Tis21^KO + MWM, p = 0.37 Mann-Whitney U test, Kruskall-Wallis (d.f. 3) H = 30.41, p < 0.0001; BrdU⁺/stage 6 in total BrdU⁺: Tis21^WT vs. Tis21^WT + MWM, p = 0.68 or Tis21^KO vs. Tis21^KO + MWM, p = 0.57 Mann-Whitney U test, Kruskall-Wallis (d.f. 3) H = 36.17, p < 0.0001; Figure 4D). Hence, the Tis21 knockout-induced pattern of increase of stage 5 neurons and decrease of stage 6 neurons, relative to wild-type mice, remained the same also after MWM training, indicating that the defect in the differentiation rate of Tis21 knockout neurons was not rescued (BrdU⁺/stage 5 in total BrdU⁺: Tis21^WT + MWM vs. Tis21^KO + MWM, p < 0.0001, Mann-Whitney U test; BrdU⁺/stage 6 in total BrdU⁺: Tis21^WT + MWM vs. Tis21^KO + MWM, p < 0.0001, Mann-Whitney U test; Figure 4D).

As a way to assess the specificity of impairment of the differentiation observed in Tis21 knockout mice, we sought to rescue it by injecting in the dentate gyrus retroviruses expressing (or silencing) either Tis21 (Figure 5) or genes regulating neural differentiation at specific steps, namely, Id3 (Figure 6) and NeuroD2 (Figure 7). The gene Id3, which negatively regulates neural differentiation (Lyden et al., 1999; Andres-Barquin et al., 2000; Yokota, 2001), has been previously shown to be strongly repressed by Tis21 in dentate gyrus neurons (Farioli-Vecchioli et al., 2009); whereas NeuroD2 is involved in the process of dentate gyrus development and survival (Olson et al., 2001) and is amongst the bHLH genes target of the inhibitory activity of Id3.

The infection of the dentate gyrus of Tis21-null adult mice (at P60) with a retrovirus expressing Tis21 and GFP (pCAG-IRES-GFP-Tis21; Figures 5A–E) occurred only in proliferating progenitors; however, the analysis of infected cells (GFP⁺) was performed 5 days after infection, hence cells were detected that either continued to proliferate or had differentiated. The pCAG-IRES-GFP-Tis21 virus effectively inhibited the proliferation of dividing progenitors (GFP-Tis21⁺Ki67⁺/total GFP-Tis21⁺ vs. GFP-empty⁺Ki67⁺/total GFP-empty⁺ p = 0.015, Mann Whitney U test, used hereafter for virus infection analyses; Figures 5A,C); in parallel, the high percentage of immature stage 5 neurons and the low percentage of terminally differentiated stage 6 neurons present in Tis21-null dentate gyrus was reversed (stage 5: GFP-Tis21⁺stage 5/total GFP-Tis21⁺ vs. GFP-empty⁺stage 5/total GFP-empty⁺ p = 0.037; stage 6: GFP-Tis21⁺stage 6/total GFP-Tis21⁺ vs. GFP-empty⁺stage 6/total GFP-empty⁺ p = 0.001; Figures 5D,E). These data clearly indicate that the defect of differentiation of dentate gyrus progenitor cells is specifically dependent on the loss of Tis21.

Given that non-specific stimuli triggering the whole neurogenic machinery, either by activation of the serotoninergic pathway, or promoting the process of differentiation through cognitive training, did not rescue the defect of terminal differentiation, we now tested the specific silencing of Id3. In fact Id3 is an inhibitor of the proneural bHLH genes (Lyden et al., 1999; Yokota, 2001), and is negatively regulated at transcriptional level by Tis21 in neural cells (Farioli-Vecchioli et al., 2009).

Thus, we infected the dentate gyrus cells of Tis21-null mice with a retrovirus expressing GFP and an shRNA targeting specifically Id3 (GFP-shId3), which had been previously generated and checked for its efficiency in silencing Id3 expression (Farioli-Vecchioli et al., 2009; Figures 6A–E). We observed that the percentage of dividing progenitors observed in Tis21-null mice was not affected by the infection with the GFP-shId3 retrovirus, as compared to a control retrovirus expressing an shRNA targeting luciferase (GFP-shLUC, Farioli-Vecchioli et al., 2014a; GFP-shId3⁺Ki67⁺/total GFP-shId3⁺ vs. GFP-shLUC⁺Ki67⁺/total GFP-shLUC⁺ p = 0.35; Figures 6A,C). In contrast, the percent accumulation of immature neurons and the lower percent ratio of terminally differentiated neurons occurring in Tis21-null dentate gyrus were reversed by GFP-shId3 retrovirus (stage 5: GFP-shId3⁺stage 5/total GFP-shId3⁺ vs. GFP-shLUC⁺stage 5/total GFP-shLUC⁺ p = 0.006; stage 6: GFP-shId3⁺stage 6/total GFP-shId3⁺ vs. GFP-shLUC⁺stage 6/total GFP-shLUC⁺ p = 0.043; Figures 6D,E). Of note, the measurement of stage 5/6-infected cells (i.e., GFP⁺) was performed in all infection experiments as ratio to the total number of GFP⁺ cells, and is therefore a percentage analysis analogous to that used to measure FLX-treated progenitor cells above background in Figure 3D. Thus, the silencing of Id3 expression specifically rescues the Tis21 knockout-dependent defect of differentiation of dentate gyrus progenitor cells, without effect on their proliferation. This reveals that Id3 is a specific regulator of the terminal differentiation of dentate gyrus progenitor cells, where it acts under control of Tis21 (Farioli-Vecchioli et al., 2009), and that proliferation and differentiation in this system are genetically dissociable.

We further tested whether the proneural gene NeuroD2 could specifically rescue the defect of terminal differentiation of Tis21-null neurons. This choice was based on the fact that NeuroD2 has been shown to induce neural differentiation in embryonic stem cells (Sugimoto et al., 2009) and to be selectively expressed in stage 6 dentate gyrus neurons, i.e., with Tis21 (Roybon et al., 2009; Attardo et al., 2010).

Thus, we generated a retrovirus expressing GFP and the NeuroD2 coding region (GFP-NeuroD2) and used it to infect the dentate gyrus cells of Tis21-null mice. We noticed a manifest decrease, albeit below significance, of the proliferating stem/progenitor cells expressing the NeuroD2 virus, relative to control-infected cells (GFP-NeuroD2⁺Ki67⁺/total GFP-NeuroD2⁺ vs. GFP-empty⁺Ki67⁺/total GFP-empty⁺ p = 0.15; Figures 7A,C). In contrast, the higher percentage of stage 5 immature neurons and the lower percentage of terminally differentiated neurons occurring in Tis21-null dentate gyrus were overturned in cells infected by the NeuroD2 retrovirus (stage 5: GFP-NeuroD2⁺stage 5/total GFP-NeuroD2⁺ vs. GFP-empty⁺stage 5/total GFP-empty⁺ p = 0.0004; stage 6: GFP-NeuroD2⁺stage 6/total GFP-NeuroD2⁺ vs. GFP-empty⁺stage 6/total GFP-empty⁺ p = 0.044; Figures 7D,E). This indicates that NeuroD2, similarly to Id3, is a specific regulator of the terminal differentiation of dentate gyrus progenitor cells, able to rescue the defect of differentiation in Tis21-null dentate gyrus neurons.