Separate groups of C57Bl/6 mice were exposed to running wheels for 0, 2, 4, 7, 10 or 14 days (n = 8 per exposure group). (A) Performance on a Y-maze task of hippocampus-dependent cognition shows a duration-response curve with exercise-induced improvements beginning after 7 days of running exposure. (B) qRT-PCR analyses show decreases in levels of BMP4 mRNA and increases in noggin mRNA in the hippocampus preceding the onset of improvements in Y-maze performance. (C) Examples of Western blots for noggin, BMP4 and PCNA protein in the hippocampus after increasing durations of exposure to elective exercise. (D) Quantification of the Western analyses of hippocampal BMP4, noggin and PCNA shows a duration-dependent relationship, with the onset of changes in BMP4 and noggin levels preceding those for the cell proliferative marker, PCNA. * Differs from control (day 0) group at p<0.01

(A–D) Unbiased stereological sampling of coronal sections through the hippocampus and quantification of cells co-labeled with (A) CldU and Dcx, (B) IdU and Dcx, (C) CldU and Sox2, and (D) IdU and Sox2. (E) Table detailing the early/late progenitor identity of cells dividing at different points in time for different genotypes and exposure to running. The number in pink on the top is the number of CldU labeled cells per field, the blue number on the top is the number of IdU labeled cells, the number in red on the left is the number of Sox2 labeled cells, and the number in green on the left is the number of DCX labeled cells. Black numbers indicate cells marked for division which are also labeled as a given progenitor subtype. Percentages indicate the fraction of dividing cells within a given lineage species for each group of mice. * Differs from wild type naive group at p<0.01. The marked reductions in the number of proliferative Sox2-labeled and Dcx-labeled cells in the NSE-BMP4 mice indicates that BMP signaling acts on both early and later populations of precursors. The increase in both CldU and IdU labeling of the Dcx labeled cells in the NSE-noggin and WT running mice suggests that inhibition of BMP signaling increases proliferation of later precursors as well as earlier ones.

(A) Schematic of the experimental design. All 4 groups of animals (N = 8 per group) received CldU for 2 days. One group of wild type animals was exposed to elective exercise for 8 days while the other groups were maintained in standard housing. On the last 2 days all 4 groups of animals received IdU. Coronal sections through the dentate gyrus were then immunostained for DCX (green), CldU (pink), IdU (blue), and Sox2 (red: panels B–E) or GFAP (red: panels F–I). Panels B–E and F–I show representative merged images for all sets of four labels. Individual channels comprising merged images are shown in Figure S2. (B,F) Wild type, standard housing; (C,G) Wild type, running; (D,H) NSE-noggin. (E,I) NSE-BMP4. (J–M) Unbiased stereological sampling and quantification per 40X field of cells labeled for (J) CldU; (K) (IdU); (L) Sox2; (M) DCX. Analyses of double labeling are shown in Figure 3. Analysis of coronal sections labeled for nestin is presented in Figure S3. * Differs from wild type naïve group at p<0.001 or from other groups as indicated at p<0.01 by Bonferroni/Dunn pair-wise comparison.

Mice were trained on a water maze task requiring daily adjustment to a new target location. (A) All groups performed equally on four trials during a single day when the platform was visible. (B) With a hidden platform, all groups performed equally on the first trial. However thereafter NSE-noggin mice were greatly improved and NSE-BMP4 mice were impaired in the time to find the hidden platform. (C) Daily averages of 4 trials per day for the time required to find a platform hidden each day in a new location. Noggin transgenic mice show superior adaptation to the novel platform location each day between the primer and training trials, and show improved reference learning across the 4 daily trials. BMP4 transgenic mice perform significantly worse than wild type on both measures. (D) The daily average distance swam to reach the platform mirrored the time to reach the platform within each group, indicating equal swimming velocities for all mice. * Differs from wild type at p<0.01 by Bonferroni/Dunn pair-wise comparison. **Significant main effect by multiple ANOVA for group x day comparisons with all groups at p<0.05. The daily average distance swam to reach the platform mirrored the time to reach the platform within each group,

(A–I) Two month old wild type mice were implanted with Alzet minipumps delivering noggin, AraC, noggin + AraC, or vehicle intraventricularly (n = 10 per group) following the experimental paradigm in Figure 7A. Coronal sections through the dentate gyrus were then immunostained for DCX (green), CldU (pink), IdU (blue), and GFAP (red). Panels A–D show representative merged images for all four labels. (A) Saline; (B) Noggin; (C) Noggin + AraC; (D) AraC (E–H) (E–I) Unbiased sampling and quantification of cells labeled for (E) CldU; (F) (IdU); (G) DCX; (H) CldU + DCX; (I) IdU + DCX. (J–L) Noggin infusion into wild type mice improved scores on hippocampus-dependent behavioral tests in a manner focused on the DG and dependent upon ongoing cell proliferation. (J) Y-maze scores. (K) NOR scores. (L) Latency by day on Morris water maze testing. Analyses of hippocampus-independent behavioral changes and of coronal sections labeled for nestin are provided in Figures S12 and S13. * Differs from saline infusion control at p<0.01 by Bonferroni/Dunn pair-wise comparison. **Significant main effect by multiple ANOVA for group x day comparisons with all groups differing from wild type at p<0.05.

Wild type, NSE-noggin, and NSE-BMP4 mice (n = 10 per group per experiment) were tested on a series of learning and memory tasks to measure baseline levels of hippocampal cognitive performance, processing and recall. (A) Performance on a Y-maze test measuring spontaneous alternation as a hippocampus-dependent exploration strategy shows significant gains in NSE-noggin mice and significant deficits in NSE-BMP4 mice relative to wild type controls. (B) NSE-noggin mice have significantly improved object memory on the novel object recognition test after multiple delay periods, and (C) significantly better spatial memory on the novel location recognition test relative to wild type animals. NSE-BMP4 mice are impaired on all tests, never significantly exceeding chance levels of performance. (D) Fear conditioning tests for hippocampus-independent cued memory and hippocampus-dependent contextual memory show that the cognitive phenotypes and the effects of BMP signaling are specific to the hippocampus. All mice perform equally on the cued memory test, while the percentage of time freezing on the contextual memory test is increased for NSE-noggin mice and decreased for NSE-BMP4 mice relative to wild type controls. * Differs from wild type group at p<0.01 by Bonferroni/Dunn pair-wise comparison. **Significant main effect by multiple ANOVA for group x delay comparisons with all groups differing from each other at p<0.05.

(A–I) Groups of 2 month old NSE-BMP4 mice and age-matched WT controls (N = 16 per group) received CldU for 2 days. Half of each group was exposed to elective exercise for the last 8 days of the 10-day trial, while the other half was maintained in standard housing. On the final 2 days, all animals received IdU. Coronal sections through the dentate gyrus were then immunostained for DCX (green), CldU (pink), IdU (blue), and GFAP (red). Panels A–D show representative merged images for all four labels; (A) Wild type, standard housing; (B) Wild type, running; (C) NSE-BMP4, standard housing; (D) NSE-BMP4, running. (E–I) Unbiased sampling and quantification of cells labeled for (E) CldU; (F) (IdU); (G) DCX; (H)CldU + DCX; (I) IdU + DCX per 40X field. (J–K) Other groups of NSE-BMP4 mice were housed with running wheels for 0, 4, 7, 10 or 14 days and were analyzed for (J) hippocampus-dependent Y-maze performance and (K) hippocampal levels of BMP4, noggin and PCNA proteins. Exercise failed to alter either behavior or levels of any of the proteins in the NSE-BMP4 animals. * Differs from wild type naïve group at p<0.01 by ANOVA and Bonferroni/Dunn pair-wise comparison.

(A) Two month old NSE-Noggin mice were implanted with an Alzet minipump delivering cytosine arabinoside (AraC) or vehicle intraventricularly (n = 10 per group) for 15 days. A Y-maze test performed on day 3 during a pilot experiment and tests in non-operated mice were used to establish baseline behavior. NOR testing was performed on day 6 and CldU was administered from days 6–8. Thereafter the animals were habituated to the Morris water maze (MWM), and MWM testing was performed on days 11–15. IdU was administered on days 14 and 15, and a Y-maze test was performed on day 15 prior to sacrifice/perfusion of the animals. (B–G) Coronal sections through the dentate gyrus were then immunostained for DCX (green), CldU (pink), IdU (blue), and Sox2 (red). Panels B–C show representative merged images for all four labels. (B) Saline infusion; (C) AraC infusion. (D–G) Unbiased stereological sampling and quantification of cells labeled for (D) CldU; (E) (IdU); (F) DCX; (G) Sox2, per 40X field. (H–J) AraC infusion impaired hippocampus-dependent cognitive abilities of NSE-noggin mice. (H) Latency by day on Morris water maze testing. (I) Y-maze scores. (J) NOR scores. * Differs from vehicle infusion control at p<0.01 by Student's t-test or ANOVA with Bonferroni/Dunn pair-wise comparison. **Significant main effect by multiple ANOVA for group x day comparisons differing from vehicle control at p<0.05.