Expression of a nuclear inhibitor of PP1 (NIPP1*) in mature granule cells.
To test for evidence of activity-dependent competition for survival between new neurons and pre-existing granule cells, we used a mouse model where the plasticity of mature granule cells is enhanced through transgenic expression of a constitutively active form of the nuclear inhibitor of protein phosphatase 1 (NIPP1*)Morishita 2001 Genoux 2002 Koshibu 2009 Graff 2010 . NIPP1* expression in mature granule cells has been previously shown to enhance synaptic plasticity of granule cells in the adult DG, leading, for example, to an increased amplitude of long-term potentiation (LTP)Koshibu 2009 Graff 2010 . To selectively direct NIPP1* expression to mature granule cells, we used a cell type-specific approach where NIPP1* is expressed under the control of the CaMKIIα promoter and the reverse tetracycline (Tet)-controlled transactivator 2 (CaMKIIα-driven rtTA2 x TetO-NIPP1*/EGFP; hereafter called tg-NIPP1*; Figure A)Gossen 1992 Koshibu 2009 . We found that the CaMKIIα promoter is not active in newborn granule cells expressing the microtubule-associated protein doublecortin (DCX) that is transiently expressed in newborn neurons but not in mature granule cellsPlumpe, 2006 , and that transgene expression upon Doxycycline (DOX) treatment was highly selective in mature granule cells (NeuN+, DCX-) in the DG as measured by expression of EGFP fused to NIPP1* (Figure A and data not shown). Thus, transgenic NIPP1* expression within the DG is restricted to mature granule cells past the expression stage of DCXPlumpe, 2006  making this model suitable for testing the effects of enhanced mature granule cell plasticity on NSPC proliferation, fate determination, and stable integration of newborn neurons.
Tg-NIPP1* expression in mature granule cells does not affect NSPC proliferation.
As previous reports suggested that NSPC proliferation is regulated by neuronal activityDeisseroth 2005 Song 2012 Song 2013 , we first analyzed if NSPC proliferation is affected upon tg-NIPP1* expression. We compared DOX-fed tg-NIPP1* mice with their single transgenic (lacking the TetO-NIPP1*/EGFP transgene) and non-DOX-fed tg-NIPP1* littermates as controls. Non-DOX-fed single as well as double transgenic mice and DOX-fed single and double transgenic mice were analyzed separately to exclude leakiness of the transgene expression and to test for a potential influence of DOX alone. We did not observe significant differences in the rate of proliferation and number of newborn neurons as measured using DCX expression and thymidine analogue labeling (BrdU, CldU, and IdU) between any of the control groups in any experiment (Supplemental Figure 1A, B). Upon 2 weeks of DOX treatment, NSPC proliferation was analyzed using a single BrdU pulse (Figure B) 14 h prior to killing the animals. Using this approach, we observed no significant difference between DOX-fed tg-NIPP1* and their respective controls (Con: 1332 ± 145; DOX-tg-NIPP1*: 1340 ± 125 BrdU+ cells per DG, n.s). We next analyzed if tg-NIPP1* expression affects early neuronal cell death or fate specification of newborn cells. Using the thymidine analogue IdU, we labeled cells 1 week prior to analyses and found no difference between tg-NIPP1* mice and respective controls (Figure C) (Con: 1184 ± 163; DOX-tg-NIPP1*: 1138 ± 122 IdU+ cells per DG, n.s.). In addition, we found that virtually all IdU-labeled cells expressed the neuronal markers DCX and Prox1 (Figure C and data not shown), suggesting that fate determination and differentiation of newborn neurons is not affected in DOX-tg-NIPP1* mice.
Enhanced plasticity of the mature granule cell circuit impairs survival of newborn neurons.
We then tested if the total number of immature neurons in tg-NIPP1* mice is affected upon DOX treatment. DCX starts to be expressed in late dividing NSPCs and expression lasts for approximately 3 weeks after neuronal birthKempermann 2004 Plumpe, 2006 . Strikingly, the number of DCX-expressing cells was substantially reduced in DOX-tg-NIPP1* mice (Con: 13182 ± 662.2; DOX-tg-NIPP1* 9810 ± 828.8 DCX+ cells per DG, p <0.05), indicating a loss of immature neurons at later stages when activity-dependent survival occurs (Figure D)Tashiro 2006 . We used a complementary approach to confirm the loss of new neurons based on DCX analyses and injected the thymidine analogue CldU in DOX-tg-NIPP1* mice and respective controls. Animals were killed 3 weeks later and the number of CldU+ cells analyzed. Corroborating the DCX results, we found a significant drop in CldU-labeled cells in DOX-tg-NIPP1* mice compared to controls (Con: 808 ± 100 control; DOX-tg-NIPP1*: 465 ± 101 CldU+ cells per DG, p <0.05) (Figure E). Interestingly, these results suggest that neuronal loss occurs around the time when newborn neurons start to form dendritic spines and excitatory synapsesZhao 2006 Toni 2007  suggesting that synaptic competition, impaired by enhanced plasticity of the mature granule circuit through NIPP1* expression, may be critical for stable integration into the DG network.
Reduced dendritic complexity in newborn neurons in tg-NIPP1* mice.
We next analyzed if the length and branching of newborn neurons are affected in DOX-tg-NIPP1* mice, using these measures as a proxy for neuronal integrationTronel 2010 . To analyze the morphology of newborn neurons, we injected retroviruses expressing GFP under chicken beta-actin promoter stereotactically into the DG of control and tg-NIPP1* mice and killed the animals 3 weeks laterZhao 2006 . Using this approach, we found that dendritic length was significantly reduced in DOX-tg-NIPP1* mice compared to controls (Figure F) (Con: 476.8 µm ± 18.5; DOX-tg-NIPP1*: 368.1 µm ± 22.8 average dendritic length, p <0.001). Further, dendrites extending from neurons born in DOX-tg-NIPP1* mice had substantially fewer branches compared to controls (Figure F) (Con: 8.43 ± 0.52; DOX-tg-NIPP1*: 6.14 ± 0.35, branch points per neuron, p <0.01). In contrast, we found that axonal growth into area CA3, which is reached by axons extending from newborn neurons before first spines are formedZhao 2006  was not altered in DOX-tg-NIPP1* mice (Supplemental Figure 2) (Con: 1122.74 ± 91.73 µm; DOX-tg-NIPP1* 1225.73 ± 42.36 µm, n.s.). After finding that dendrites extending from newborn neurons in DOX-tg-NIPP1* were shorter and less complex, we next analyzed the number of dendritic spines, the main place for excitatory synapses of excitatory neurons, in DOX-tg-NIPP1* and control mice (Figure G). On the dendritic segments analyzed, the number of spines per µm dendrite was similar between groups (Con: 0.469 ± 0.04; DOX-tg-NIPP1*: 0.423 ± 0.02 spines/µm, n.s.) (Figure GToni 2007 ). Furthermore, the subtype of spines did not differ between DOX-tg-NIPP1* and control mice (Con: 0.761 ± 0.07; DOX-tg-NIPP1*: 0.898 ± 0.07 mushroom spines/µm; Con: 0.136 ± 0.01; DOX-tg-NIPP1*: 0.105 ± 0.09 stubby spines/µm; Con: 0.256 ± 0.03; DOX-tg-NIPP1*: 0.227 ± 0.02 thin spines/µm, n.s.). However, given that dendrites are shorter in DOX-tg-NIPP1* mice, we reasoned that despite similar spine density the total number of excitatory inputs is reduced per newborn neuron in DOX-tg-NIPP1* mice compared to controls. To estimate the number of spines per cell, we multiplied spines per µm with the calculated dendritic length. It has to be considered, however, that spines are not uniformly distributed on the dendrites; so any calculations can only be a rough estimation (e.g., within the granule cell layer (GCL), only few spines are formed). Given that dendrites are substantially longer in controls than in DOX-tg-NIPP1* mice, we estimated a reduction of synaptic input of approximately 50% of newborn neurons in DOX-tg-NIPP1* compared to controls (195.7 in control, 130.6 in DOX-tg-NIPP1* calculated spines per cell). Next, we analyzed synapse formation of new neurons in DOX-tg-NIPP1* on the ultrastructural levelToni 2007 . Again, we used retroviruses expressing GFP to label newborn neurons in DOX-tg-NIPP1* mice and controls. Synapse formation was analyzed 3 weeks after stereotactic injection of retroviruses and after immunhistochemical conversion of the GFP signal into osmiophilic DAB precipitate. We then used focused ion beam scanning (FIBS)- electron microscopy (EM) to reconstruct spines and their environment in three dimensions (n = 1 mouse per genotype). In total, we analyzed 73.2 µm of two dendrites containing 36 synapses in DOX-tg-NIPP1* (n = 1). Using this approach, we found that spines of newborn neurons in DOX-tg-NIPP1* mice formed synapses with visible postsynaptic densities, indicating functional connectivity, and engaged in multiple synapse boutons (MSBs), suggesting that synapse formation is not fundamentally altered in DOX-tg-NIPP1* mice (Figure H)Toni 2007 .
We used a genetic approach to test if the survival of new neurons in the adult DG is influenced by the plasticity of the mature granule cell circuit. Strikingly, we found that dendritic integration and subsequent survival of newborn granule is impaired with enhanced plasticity of the mature granule cell circuit, suggesting that competition for synaptic integration is critical in regulating the survival of new neurons. New neurons need to receive synaptic input for their survival, and structural data suggested that filopodia extending from dendrites of newborn neurons initially grow towards existing synapses between mature granule cells and presynaptic axons in the perforant path originating from neurons in the entorhinal cortex, leading to the formation of MSBsTashiro 2006 Toni 2007 . With time, MSBs are then presumably exchanged by single synapse boutons (SSBs) suggesting that new neurons that successfully compete for synaptic partners stably integrateToni 2007 . The survival of newborn neurons can be enhanced, for example, by environmental enrichment that appears to enhance excitability in the DG circuitKempermann 1997 Eckert 2010 . Similarly, the reduced survival of new neurons lacking the GluN1 subunit of the NMDA receptor can be partially rescued by global blockade of NMDA-dependent activityTashiro 2006 . However, characterizing a potential competition by manipulating the balance of excitability selectively between new and mature granule cells had not been tested experimentally. We achieved this by cell type-specific expression of NIPP1* in mature granule cells and found that enhancing plasticity of the mature granule cell circuitGenoux 2002 Koshibu 2009  impairs the survival of newborn granule cells. The growth of dendrites extending from newborn granule cells was substantially impaired in NIPP1*-expressing mice, whereas synapse formation appeared to be unaltered as analyzed using conventional light microscopy and electron microscopy. This led to a strongly reduced number of dendritic spines of newborn neurons, suggesting that their overall excitatory synaptic input is decreased. Since the transgene is expressed in all CaMKII-expressing cells, including those of the entorhinal cortex, we cannot exclude an effect of more globally changed activity of neuronal circuits. Without longitudinal imaging of synapse formation, which is currently technically not feasible at the required resolution, it cannot be proven that new neurons fail to survive due to impaired synaptic competition. However, our data strongly supports the hypothesis that new neurons need to compete for synaptic partners to ensure proper integration and survivalToni 2007 . Interestingly, it has been shown that during the first 3-6 weeks after their birth, new granule cells are highly excitable and show a higher degree of plasticity compared to mature granule cellsSchmidt Hieber_2004 Ge 2007 Marin Burgin_2012 Brunner 2014 . This unique feature has been attributed to their special functional properties with emerging evidence supporting the idea that adult neurogenesis in the DG is not a process for mere cell replacement but that new neurons exert their function at least partially due to these special propertiesAimone 2011 Sahay 2011 Deng 2013 . However, it is also reasonable to speculate that the phase of heightened excitability may also be important to ensure the integration of new neurons, giving them a competitive advantage to form synapses with axons arising from the entorhinal cortex. Our data supports this idea by showing that the survival of new neurons is impaired with increased plasticity of the mature granule cell circuit, whereas NSPC proliferation and initial steps of fate determination and specification are unaltered. Thus, we here experimentally support a long-standing hypothesis, indicating that synaptic competition represents a key mechanism regulating the integration and survival of newborn granule cells in the adult mammalian hippocampus.