|Home | About | Journals | Submit | Contact Us | Français|
Experience rearranges anatomical connectivity in the brain, but such plasticity is suppressed in adulthood. We examined the turnover of dendritic spines and axonal varicosities in the somatosensory cortex of mice lacking Nogo Receptor 1 (NgR1). Through adolescence, the anatomy and plasticity of ngr1 null mice are indistinguishable from control, but suppression of turnover after age 26 days fails to occur in ngr1−/− mice. Adolescent anatomical plasticity can be restored to one-year old mice by conditional deletion of ngr1. Suppression of anatomical dynamics by NgR1 is cell autonomous, and is phenocopied by deletion of Nogo-A ligand. Whisker removal deprives the somatosensory cortex of experience-dependent input and reduces dendritic spine turnover in adult ngr1−/− mice to control levels, while an acutely enriched environment increases dendritic spine dynamics in control mice to the level of ngr1−/− mice in a standard environment. Thus, NgR1 determines the low set point for synaptic turnover in adult cerebral cortex.
Environmental experience plays a major role in sculpting brain architecture and synaptic connectivity. The malleability of synaptic connections is extensive during adolescent “critical periods”. In adults, anatomical plasticity is restricted (Grutzendler et al., 2002; Holtmaat et al., 2005; Trachtenberg et al., 2002; Yang et al., 2009; Zuo et al., 2005). Discrete rearrangement of adult synaptic connections has been implicated in learning and memory for motor, sensory and contextual tasks (Hofer et al., 2009; Holtmaat et al., 2006; Lai et al., 2012; Trachtenberg et al., 2002; Wilbrecht et al., 2010; Xu et al., 2009; Yamahachi et al., 2009; Yang et al., 2009). Environmentally determined patterns of electrical activity drive patterns of synaptic change, but the molecular basis for the set point, or gain, of anatomical dynamics is not known.
NgR1 was originally identified as a mediator of myelin-dependent restriction of recovery from injury (Fournier et al., 2001). In healthy brain, NgR1 was shown to be essential in closing the critical period for ocular dominance plasticity by single unit electrophysiology in anesthetized mice after monocular deprivation (McGee et al., 2005). In dissociated neuron and brain slice studies, NgR1 has a role in acute electrophysiological plasticity and synaptic morphology (Delekate et al., 2011; Lee et al., 2008; Raiker et al., 2010; Wills et al., 2012; Zagrebelsky et al., 2010). Here, we examined whether the in vivo anatomical dynamics of adult mouse synaptic structure is regulated by NgR1 expression. We find that NgR1 gates the age-dependent restriction of anatomical plasticity in the adult cortex, and that NgR1 determines the set point for experience-driven synaptic turnover.
Using ngr1 mutants transgenic for Thy1-YFP-H, we assessed NgR1 regulation of dendritic spine dynamics in vivo. NgR1 deletion does not alter the pattern of cortical dendrite branching, the density of dendritic spines, or the morphology of spines (Fig. S1A–D). Repeated transcranial two-photon confocal imaging of anesthetized mice over 2 weeks revealed dendritic spine gains and losses from layer 5 pyramidal dendrites in the superficial 100 μm of S1 barrel field cortex (Fig. 1A, B). Compared to controls, the gains and losses of ngr1 −/− dendritic spines over a 14-day period are more than doubled (P<0.001). Similarly increased spine turnover was observed in M1 and V1 of ngr1 −/− mice (Fig. S1E). The greater spine dynamics occur without change in total spine density, emphasizing the necessity for time-lapse imaging. Separate from spine plasticity, branch extensions or retractions are rare for pyramidal neurons, and not different in ngr1 −/− mice (not shown).
When spines first protrude they are typically transient and quickly lost, with only a small subset becoming persistent and gaining the ultrastructure of synapses (Holtmaat et al., 2006; Holtmaat et al., 2005; Knott et al., 2006; Trachtenberg et al., 2002). Learning paradigms or sensory enriched environments increase short-term spine turnover, and also the stabilization of new spines into persistent spines (Holtmaat et al., 2006; Xu et al., 2009; Yang et al., 2009). In the adult, persistent spines are the overwhelming majority; a smaller pool of transient spines turns over frequently. Transient spines account for ~80% of all spine changes during 2 days, and serve as the basis for novel connectivity (see Detailed Methods, Holtmaat et al., 2005). Here, spines were classified as persistent if they were observed on two imaging sessions at days 0 and 2. The 14-day survival of persistent spines from day 2 to 16 is decreased in mice lacking NgR1, with greater persistent spine loss over 2 weeks, 10.6±2.6% in ngr1−/− vs. 3.7±0.4% in control, P<0.05 (Fig. 1C). In addition, the 2-day gain and loss of spines is greater in mice lacking NgR1, with gains of 6.0±0.6% in ngr1−/− vs. 1.9±0.4% in control, P<0.005 (Fig. 1D). Thus, endogenous NgR1 restricts the initial protrusions and retractions preceding synapse formation and promotes the persistence of established spines likely to have formed synapses.
As WT mice transition from late adolescence to adulthood, there is a dramatic increase in synaptic stability (Grutzendler et al., 2002; Holtmaat et al., 2005; Zuo et al., 2005) that coincides with myelination of V1 (McGee et al., 2005), and S1 (Fig. S1G, H). Oligodendrocytic Nogo-A protein also increases during this time (Wang et al., 2002b). The accelerated spine turnover in adult ngr1−/− mice (Fig. 1A–D) is similar to that in juvenile WT mice (Yang et al., 2009; Zuo et al., 2005). We therefore tested if NgR1-dependent restriction of spine turnover is age-dependent (Fig. 1E). Starting at P26, control mice show rapid spine turnover during 2 weeks. Turnover slows by P35 and reaches a stable adult pattern by P45, that is not altered at P180 or P360, matching the progression of intracortical myelination. Dendritic spine turnover in ngr1−/− mice is identical to controls at P26–40, but shows no change from P26 to P180. When gains and losses are considered separately (Fig. S1I, J), it is clear that age-dependent restriction of gains is NgR1-dependent. A non-significant trend suggests that a minor component of age-dependent stabilization against losses is NgR1-independent. Overall, the development of spines is normal through P26 without NgR1, but an adult gate on plasticity is lacking in ngr1−/− mice.
Next, we considered whether NgR1 is required during the transition from juvenile plasticity to adult stability, or if the protein is continuously required in the adult cortex to suppress plasticity. We used a conditionally targeted ngr1 allele, ngr1-flx (Wang et al., 2011). Temporal control was provided by an actin promotor transgene that drives ubiquitous expression of a Cre fusion protein with a mutant version of the estrogen-receptor (ERT2) (Hayashi and McMahon, 2002). Tamoxifen treatment leads to efficient ngr1 gene rearrangement and near total loss of mRNA and protein within 2 weeks (Fig. S1F and (Wang et al., 2011)). Mice with flox’ed ngr1 alleles with or without Actin-Cre-ERT2 transgene were allowed to develop with endogenous levels of NgR1. At P330, the mice received tamoxifen, to delete NgR1 from the Cre subgroup. One month later, dendritic spine stability was assessed over 2 weeks. Even at this advanced age, deletion of NgR1 increases dendritic spine turnover to the level observed in adolescent mice (Fig. 1E, P<0.001 vs. control, and n.s. vs. P26–40). Thus, constitutive NgR1 signaling reversibly limits synaptic turnover in the adult cerebral cortex.
We considered whether NgR1 regulation of post-synaptic stability in adult cortex was coupled with similar changes in pre-synaptic stability, or if there was selective action in dendrites. We first determined the types of presynaptic fibers labeled in cortical layer I of Thy1-YFP-H mice. Using described morphological criteria (De Paola et al., 2006), we found that the vast majority of labeled axons are consistent with recurrent cortical fibers from layer V and layer II/III (A3 subtype, 98.7±0.7% of total). Pre-synaptic specializations along these fibers were imaged over a 14-day interval in the S1 barrel field cortex in 6–7 month old mice (Fig. 1G). Consistent with previous reports (De Paola et al., 2006), axonal varicosities are more stable than dendritic spines. Critically, axonal specializations are at least twice as dynamic in ngr1−/− mice than in ngr1+/− mice (Fig. 1H; P<0.005). The gains and losses of axonal varicosities are increased equally. While varicosity turnover is increased, we observed no genotype difference in the frequency of rare axonal branch extensions and retractions (not shown). Overall, the requirement of NgR1 for axonal stability closely matches its role in stabilizing dendritic spines in adult cortex.
Several ligands for NgR1 have been described, including Nogo-A, MAG and OMgp from oligodendrocytes (reviewed in Akbik et al., 2012), chondroitin sulfate proteoglycans (Dickendesher et al., 2012) and LGI1 (Thomas et al., 2010). To assess whether the prototypical Nogo ligand might regulate dendritic spines in vivo, we exposed 19–22DIV dissociated cortical neurons to a Nogo-A polypeptide (22 kDa, (Huebner et al., 2011)) during time-lapse spinning disc confocal imaging of myristoyl-GFP transfected cells (Fig. 2A). Without added Nogo-22, spine dynamics are similar between WT and ngr1−/− cultures. This parallels what we observe in juvenile mice in vivo because 19–22 DIV dissociated cultures are unmyelinated (not shown). Acute treatment with 100 nM Nogo-22 protein reduces the appearance of new dendritic spines by 80% (Fig. 2B, P<0.002). Acute Nogo-22 treatment also leads to an increase in spine loss. In contrast to the differential effect of acute Nogo-22 treatment on gains and losses, both spine gains and losses are restricted in adult mice (Fig. 1, Table S1). We hypothesized that chronic treatment with Nogo-22 in vitro might mimic the chronic effect of myelin-inhibition in vivo. Pre-treating 19–22 DIV dissociated cortical cultures with Nogo-22 for 6 days leads to a decrease in both spine gains and losses (Fig. 2C), mimicking adult wild-type spine dynamics in vivo. This suggests that NgR1 inhibition of spine gain, paralleling inhibition of neurite extension, is the primary event in vivo, with steady state changes in spine loss being secondary and compensatory in vivo. Nogo-22 induced changes are not detectable in ngr1 −/− cultures (Fig. 2B) and are dose-dependent (Fig. S2).
Given the acute in vitro action of Nogo-22 through NgR1 to prevent dendritic spine gain, we utilized Nogo-A/B null mice to determine whether this ligand is required for NgR1 stabilization of dendritic spines in adult mice. Using the Thy1-YFP-H marker, dendritic spine gains over 2 weeks are increased more than 2-fold in nogo-A/B null mice relative to control at P180 (Fig. 2C, D; P<0.001). A balanced increase in spine losses is observed in nogo-A/B −/− mice (Fig. 2D), and the greater turnover of Nogo-A/B null axonal varicosities parallels that of dendritic spines (Fig. 2E, F). Thus, loss of the Nogo-A/B ligand phenocopies the rapid juvenile-type of synaptic turnover observed in NgR1-deficient adult mice. To examine a genetic interaction between Nogo-A/B and NgR1, we assessed the turnover of dendritic spines in compound heterozygotes (Fig. 2E). The single heterozygotes are indistinguishable from WT mice, but the nogo-A/B, ngr1 double heterozygous mice exhibit 14-day spine turnover that is similar to the increased rate in homozygous mutants.
NgR1 might alter anatomical plasticity indirectly, for example by changing activity patterns globally. Alternatively, the protein might function cell autonomously to limit synapse dynamics in cells expressing NgR1. If NgR1 acts cell autonomously and locally, an enrichment in post-synaptic densities might be expected. Fractionation of adult cerebral cortex reveals enrichment of NgR1, but not Nogo-A/B, in post-synaptic densities (Fig. 3A). This is consistent with previous immuno-electron microscopy demonstrating NgR1 in pre- and post-synaptic structures (Wang et al., 2002b). To assess cell autonomous action in vitro, we mixed ngr1−/− and wild type neurons in different proportions prior to Nogo-22 exposure. The GFP marker was included in a small percentage of cells of one genotype plated with a 24-fold excess of neurons from one or the other genotype. The response of the imaged cell depends on the expression of NgR1 in that cell and not on the expression in the majority of cells throughout the culture (Fig. 3B, C). Even when 96% of the culture is wild type, rare ngr1−/− neurons do not respond to Nogo-22 (Fig. 3C). In vitro, NgR1 acts cell autonomously to mediate the acute action of diffusely applied Nogo-22.
To determine if NgR1 acts cell autonomously in vivo, we utilized SLICK transgenic mouse containing a Thy1 promotor that drives a fluorescent marker and Cre-ERT2 in a subset of cortical neurons (Young et al., 2008). After crossing onto the ngr1 flx/flx background, tamoxifen treatment provides labeled NgR1 null neurons in a milieu of unlabeled WT neurons (Fig. 3D). Compared to ngr1 flx/+ neurons, dendritic spine gains and losses for the single cell NgR1-deleted neurons are increased (Fig. 3E, F; P<0.01). In this paradigm, axonal varicosity turnover for NgR1 null neurons is also increased in the wild type host (Fig. 3G). Thus, NgR1 acts cell autonomously in vitro and in vivo to limit the turnover of both presynaptic and postsynaptic anatomical specializations.
While NgR1 is unique as a gene selectively titrating adult dendritic spine dynamics, inflammation and experience produce a similar phenotype in adult mice. We examined whether NgR1 gene deletion is redundant or synergistic with these interventions. Surgically implanting glass windows in the skull produce mild gliosis and greater dendritic spine turnover than does a thinned skull, which does not expose dura matter (Fig. 4A) (Xu et al., 2007). Our studies of M1 cortex confirm that dendritic spine turnover is substantially increased by the open craniotomy window relative to the thinned skull preparation (Fig. 4B; P<0.05). Importantly, NgR1 deletion increases dendritic spine turnover by a similar proportion in both the open and thinned skull preparations (Fig. 4B; P<0.01). This is not explained by differential degrees of gliosis, since craniotomy-induced inflammation is NgR1-independent (Fig S3A, B), consistent with previous observations using murine models of spinal cord injury or stroke (Kim et al., 2004; Lee et al., 2004; Wang et al., 2006). These data demonstrate the robustness of the ngr1−/− phenotype and additivity with inflammation to increase anatomical plasticity.
Next, we considered how altered sensory experience interacts with NgR1 deletion. Dendritic spine turnover is reported to increase during the first several days after mice are exposed to an enriched environment, after which turnover returns to baseline (Yang et al., 2009). Similarly, we observed an increase in 2-day spine turnover from 5.0±0.6% in standard caging (SE) to 13.5±1.5% in a new environment strongly enriched (EE) for whisker stimulation (Fig. 4C; P<0.001). The environment-driven increase requires vibrissal somatosensory input because it is eliminated by whisker trimming at the time of transition to the enriched environment (Fig. 4C). Strikingly, the elevated level of dendritic spine turnover in ngr1−/− mice is not further increased by EE (Fig. 4C).
On the one hand, this might imply that ngr1−/− neurons are no longer sensitive to experience-dependent inputs. Alternatively, the threshold for experience-dependent plasticity may be lower; the standard cage may provide a level of stimulation capable of eliciting full spine turnover in ngr1−/− but not control mice. To separate these possibilities, we trimmed the whiskers of ngr1−/− and control mice for 12 days and then assessed 2-day S1 dendritic spine turnover in the sensory-deprived state. For control mice, there was a slight decrease in dendritic spine turnover (Fig. 4C), indicating that our SE provides little effective experiential drive to S1 dendritic spine turnover. In contrast, two-day spine turnover for ngr1−/− mice is suppressed to control levels by whisker trimming (Fig. 4C, P<0.01). We conclude that SE provides a level of vibrissal input adequate to drive experience-dependent dendritic spine changes in ngr1−/− mice, but not controls.
Previous studies have reported activity-dependent down regulation of NgR1 (Josephson et al., 2003; Wills et al., 2012), raising the possibility that EE might reduce NgR1 level to instruct enhanced anatomical plasticity. To assess this hypothesis, we analyzed total NgR1 levels and NgR1 enrichment in the PSD as a function of age or experience, and observed no differences (Fig. S3C–F). While myelin and perineuronal net formation are maturation and activity-dependent (McRae et al., 2007; Wake et al., 2011), NgR1 is a continuously present restrictive factor for anatomical plasticity in the adult cortex, with instructive clues likely to be derived from electrical activity patterns. We also investigated whether NgR1 acts upstream or downstream of experience-driven immediate early gene induction in the barrel field (Bisler et al., 2002). The rapid whisker-dependent induction of cFOS is not altered by the absence of NgR1 (Fig. S3G, H). Thus, initial acute responses to sensory input, as marked by cFOS, are distinct from NgR1 action. In contrast, the downstream events required to reorganize and imprint that input anatomically are achieved more easily without NgR1. This is consistent with NgR1 serving as a brake on anatomical dynamics.
If SE engages dendritic spine dynamics for NgR1 null mice that are similar to those of control mice in acute EE, then the characteristics of spine turnover should be similar. The most robust effect of acute EE in control mice is an increase in the rate of stabilizing new spines into persistent spines, and NgR1 null mice in the SE fully mimic this pattern (Fig. 4D, E; P<0.001). Thus, endogenous NgR1 restricts the initial protrusions and retractions preceding synapse formation, as well as the subsequent maturation of spines leading to synapse formation and persistence. Mice lacking NgR1 respond to chronic SE with a pattern of dendritic spine dynamics that matches precisely the physiological pattern of acute EE in control mice.
We considered whether the enhanced anatomical sensitivity to experience alters behavioral responses of ngr1−/− mice. Both motor learning (Xu et al., 2009; Yang et al., 2009) and extinction of fear conditioning (Lai et al., 2012) are reported to correlate with cortical spine dynamics. To assess motor learning rate, we examined conditional ngr1 null mice on the rotarod (Fig. 4F). Training improves the latency to fall more rapidly in mice lacking NgR1. In tone-associated fear conditioning, both genotypes are rapidly conditioned during Day 1 (Fig. 4G, H). During the first 30 minutes of extinction on Day 2, both groups show decreased freezing (Fig. 4I). By Day 3, when new spine gains have started to form and to maintain extinction (Lai et al., 2012), the mice lacking NgR1 show significantly (P<0.001) more pronounced extinction of freezing to the previously conditioned tone (Fig. 4I). This is due to enhanced extinction in the ngr1−/− mice, rather than impaired memory of the conditioned tone, because both genotypes show freezing at 10 days if not subjected to the extinction protocol (Fig. 4G, J). Thus, ngr1 mutant mice show enhanced learning in two paradigms linked to cortical spine turnover.
We report that NgR1 gates the transition from rapid anatomical turnover during adolescence to slower adult dendritic spine dynamics. NgR1 is required continuously to suppress synaptic turnover, functions cell autonomously, and responds to Nogo ligand. While Nogo-A/B is required, MAG, OMgp, CSPGs and LGI1 are alternate NgR1 ligands (Dickendesher et al., 2012; Liu et al., 2002; Thomas et al., 2010; Wang et al., 2002a) that may have co-essential roles to limit synaptic rearrangement in the cerebral cortex. CSPGs are known to participate in electrophysiological plasticity of the cerebral cortex (Pizzorusso et al., 2002). The alternate Nogo receptor, PirB, has also been implicated in experience dependent plasticity (Atwal et al., 2008; Syken et al., 2006). Limits on recovery after injury may share a common pathway with plasticity (Akbik et al., 2012). The transition from rapid spine turnover in juvenile mice to the anatomical stability of adult synapses coincides with the timing of NgR1 ligand presentation. Intracortical myelination and perineuronal net formation both occur at the final stages of cortical maturation, near P30, when adult spine stability is achieved.
As a molecular determinant for the low set point for anatomical plasticity in the adult, NgR1 restricts the effect of experience on cortical anatomy. Highly enriched artificial environments enhance the spine turnover of control mice during the first two days, with a subsequent return to baseline dynamics (Yang et al., 2009). In contrast, chronic housing under standard conditions is adequate to drive rapid turnover in NgR1 null mice. From these observations, the cerebral cortex lacking NgR1, with its augmented spine dynamics, is predicted to have greater sensory map refinement and greater map plasticity. Indeed, the observation of adult ocular dominance plasticity of single units under urethane anesthesia in adult ngr1−/− mice, but not adult wild type mice (McGee et al., 2005), is consistent with these observations. Adult NgR1 null mice exhibit juvenile levels of plasticity for acoustic preference as well as ocular dominance (Yang et al., 2012).
Inflammation, but not an enriched environment, is additive with NgR1 deletion to increase spine dynamics, suggesting that various experience-dependent modalities will saturate at different levels. For fearful responses, conditioning is associated with cortical spine loss, and extinction with spine gain (Lai et al., 2012). Both spine protrusion and fear extinction are age-dependent and limited in the adult (Gogolla et al., 2009). We show that NgR1 antagonizes spine protrusion (Figs. 1E, S1I) and stabilizes fear memories (Fig. 4I) in the adult. It will be of interest to determine whether accelerated cortical spine loss correlates with extinction of fear memories in NgR1 mutants. Together, these data identify NgR1 as a novel therapeutic target for the extinction of pathological fear and anxiety.
With regard to recovery from adult CNS injury, such as traumatic spinal cord injury and ischemic stroke, NgR1 limits recovery by inhibiting both axonal sprouting and long-distance regeneration (reviewed in Akbik et al., 2012). Here we show that NgR1 mutants have a decreased threshold for imprinting experience-dependent plasticity (Fig. 4C), and accelerated learning in a motor training paradigm linked to cortical spine turnover (Fig. 4F). In the setting of rehabilitation, this suggests that antagonizing NgR1 decreases the threshold to reacquire motor skills via plasticity of neuronal connectivity. Together the findings identify cortical plasticity as a substrate for NgR1-dependent restriction of functional recovery after a CNS lesion. Reactivating juvenile rates of cortical plasticity by NgR1 loss-of-function therefore represents a novel therapeutic mechanism to maximize recovery and rehabilitation after neurological injury. Moreover, synaptic anatomical dynamics provide a basis for synergy between NgR1 blockade and training effects.
ngr1 null, ngr1 flx/flx conditional mutants, and nogo-A/B null mice(Kim et al., 2003; Kim et al., 2004; McGee et al., 2005; Wang et al., 2011) were bred with Thy1-EGFP-M or Thy1-YFP-H mice (Feng et al., 2000) or (SLICK) mice (Young et al., 2008).
Neuronal structures were monitored in the cerebral cortex of anesthetized mice with a transcranial or open-skull approach. Images were acquired with two photon laser scanning microscope with a 40X objective. Neurites were images were obtained from Layer 1 of cerebral cortex from neurites traced to a cell body in L5 in vivo. Images of dendritic spines and axonal boutons were analyzed in three dimensions blind to the experimental conditions.
Environmental enrichment, rotarod and fear conditioning were by standard procedures. Detailed Methods are provided in the Supplemental Materials.
We thank J. Grutzendler for instruction on thin skull techniques, C. Portera-Cailliau for instruction on open skull surgeries, S. Tomita for instruction on subcellular fractionation, A. Holtmaat and A. McGee for helpful discussion, and S. Sodi and Y. Fu for technical assistance. F.A. is supported by an Institutional N.I.H. Medical Scientist Training Program grant. S.M.S. is a member of the Kavli Institute for Neuroscience at Yale University. We acknowledge research support from the N.I.H. and the Falk Medical Research Trust to S.M.S.
DISCLOSURE. S.M.S. is a co-founder of Axerion Therapeutics, seeking to develop PrP- and NgR-based therapeutics.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.