Adult cortical plasticity depends on an early postnatal critical period

See allHide authors and affiliations

Science  24 Jul 2015:
Vol. 349, Issue 6246, pp. 424-427
DOI: 10.1126/science.aaa8481

Keeping synaptic plasticity plastic

Neuronal synapses in the brain adjust according to shifting demands as we experience the world. This synaptic plasticity forms the basis for critical periods in the visual and somatosensory systems. Greenhill et al. have now found, in mice, a critical period for the development of plasticity itself. At the core is a protein that in its mutant form is associated with schizophrenia. Disrupting this protein's function temporarily during early development caused a failure in brain plasticity in adult mice.

Science, this issue p. 424


Development of the cerebral cortex is influenced by sensory experience during distinct phases of postnatal development known as critical periods. Disruption of experience during a critical period produces neurons that lack specificity for particular stimulus features, such as location in the somatosensory system. Synaptic plasticity is the agent by which sensory experience affects cortical development. Here, we describe, in mice, a developmental critical period that affects plasticity itself. Transient neonatal disruption of signaling via the C-terminal domain of “disrupted in schizophrenia 1” (DISC1)—a molecule implicated in psychiatric disorders—resulted in a lack of long-term potentiation (LTP) (persistent strengthening of synapses) and experience-dependent potentiation in adulthood. Long-term depression (LTD) (selective weakening of specific sets of synapses) and reversal of LTD were present, although impaired, in adolescence and absent in adulthood. These changes may form the basis for the cognitive deficits associated with mutations in DISC1 and the delayed onset of a range of psychiatric symptoms in late adolescence.

Disrupted in schizophrenia 1 (DISC1) is a protein that, when mutated, predisposes the human carrier for a number of mental disorders including schizophrenia, bipolar disorder, recurrent major depression, and autism (1, 2). DISC1 interacts with a surprisingly large number of signaling molecules, including phosphodiesterase 4, glycogen synthase kinase 3, kalirin-7, fasciculation and elongation protein ζ 1, kendrin, lissencephaly 1 (Lis1), and nudE neurodevelopment protein 1–like 1 (Nudel) (38). DISC1 affects diverse aspects of neuronal development, such as proliferation, migration, and neurite extension. In addition, DISC1 is known to be expressed in cortical neurons during both development and adulthood (9) and to reside at the postsynaptic density (6, 1012), although very little is understood of the role it plays there. In this study, working with mice, we asked whether DISC1 protein-protein interactions early in development are critical for synaptic plasticity in adulthood. We disrupted transiently DISC1’s interaction with Lis1 and Nudel during early development, at a time after cortical neurogenesis and cell migration [which are complete by about postnatal day 7 (P7) in the mouse] but before synaptogenesis and dendrite formation dominate.

We studied adult plasticity in the mouse barrel cortex, a primary sensory cortical area that receives tactile information from a normal array of 40 large whiskers. We removed all but one whisker on one side of the face of adult mice (13) to invoke cortical plasticity. The single-whisker experience normally leads to expansion of the cortical territory responding to the spared whisker (Fig. 1A). To manipulate DISC1 interactions with Lis1 and Nudel, we used a conditional transgenic mouse expressing the DISC1 C-terminal domain (DISC1cc; residues 671 to 852), which interacts with Lis1 and Nudel (1416) in a tamoxifen-sensitive construct. Within this system, a single tamoxifen injection affects DISC1 signaling for 6 to 48 hours (P7 to P9) (15). Spatial expression of DISC1cc is restricted to excitatory neurons in the forebrain by the calcium/calmodulin-dependent protein kinase II subunit α (αCaMKII) promoter, and its activity is controlled by tamoxifen. We studied the effect of a single subcutaneous injection of tamoxifen at P7 on single-whisker plasticity in adulthood (age range P70 to P130).

Fig. 1 Plasticity is impaired in adults by transient impairment of DISC1 C-terminal interactions at P7.

(A) Whisker deprivation and expansion of spared whisker domain (orange area) (13). (B) Weighted vibrissae dominance index (WVDI) for spared versus principal whiskers across experimental groups (total n = 52 mice, 496 cells; naïve mice, black bars; deprived mice, gray bars). WVDI increases with deprivation except for in DISC1cc mice injected with tamoxifen at P7 [F7,51 = 10.6, P < 0.001, three-way analysis of variance (ANOVA)]. Tamoxifen only affected plasticity in DISC1cc mice and not wild types [interaction between genotype and tamoxifen (P < 0.0005)]. (C) Spared (D1) whisker response increased with deprivation (gray bars), directly correlated with WVDI [correlation coefficient (R) = 0.93; P < 0.0001]. (D) The WVDI increases in DISC1cc mice injected with tamoxifen on P11 to P13 [t12 = 4.97, P < 0.05; black, naïve; gray, deprived) but only attains levels seen in wild types (red square) when injected at P28]; WT mice injected on P7 were not different from DISC1cc mice injected on P28 (t12 = 0.61, P = 0.45) (interaction between age and deprivation F2,2 = 10.46, P < 0.0003, ANOVA). The WT control data are plotted at P45 for clarity (red, deprived; blue, naïve), but all mice were injected with tamoxifen on P7. All plasticity values were measured in adulthood.

We found that adult DISC1cc mice injected with tamoxifen at P7 and with whiskers intact developed a normal barrel pattern, as well as normal cortical layers, cell density, and receptive fields (fig. S1). However, experience-dependent potentiation invoked by whisker loss was absent in DISC1cc mice injected with tamoxifen [F7,51 = 6.9, P < 0.001, three-way analysis of variance (ANOVA)] (see Fig. 1). Plasticity in cortical layers 2 and 3 (L2/3) was normal in wild-type mice receiving tamoxifen, which indicated that tamoxifen only acted in conjunction with the mutant protein and not by perturbing estrogen signaling (t11 = 2.9, P < 0.02). Plasticity was also normal in DISC1cc mice given just vehicle at P7, which indicated that background levels of DISC1cc availability are effectively zero (Fig. 1) (t11 = 2.4, P < 0.05). (Note that the mutated ligand-binding domain fused to DISC1cc does not bind natural estrogen, only tamoxifen.) The weighted vibrissae dominance index was unchanged in whisker-deprived DISC1cc animals receiving tamoxifen, because the spared whisker responses did not potentiate (Fig. 1 and fig. S2), and consequently, the spared whisker domain did not expand into the deprived barrels surrounding the D1 barrel (fig. S3). The lack of plasticity in the DISC1cc mice was robust across two background strains (Fig. 1 and fig. S4). These results show that normal DISC1 interaction with Lis1 and Nudel is vital during a brief period in neonatal development for the adult animal to exhibit experience-dependent plasticity.

Transient disruption of DISC1/Lis1/Nudel interactions later in development had a smaller effect on L2/3 plasticity. Disrupting DISC1 C-terminal interactions at P11 to P13 reduced plasticity less than it did at P7 and had no effect at P28 (Fig. 1D). This indicates that a critical period exists in early development with long-lasting consequences for plasticity expressed much later in adulthood.

We studied the early development of the DISC1cc mice to see where the defect originated. We found that disrupting DISC1 C-terminal signaling at P7 retarded dendritic elongation and elaboration of dendritic branching (figs. S5 and S6), but both had recovered by P21. The paired-pulse ratio, which is a measure of presynaptic maturation in the L4 to 2/3 pathway (17) was also delayed (fig. S7). Retardation of neuronal development demonstrates the immediate effect of disrupting DISC1 C-terminal interactions at P7 but does not explain the long-lasting loss of adult plasticity.

The long-lasting effects of transient disruption of DISC1/Lis1/Nudel interactions were to be found at the level of the spines rather than the dendrites. At the start of the critical period for adult plasticity, the neurons highest-order basal dendrites are mainly second- and third-order branches and are destined to become 50% of adult basal dendrites (Fig. 2 and figs. S5 and S6). We found lower spine density on second- and third-order dendritic spines in DISC1cc mice at P28 (t31 = 2.36, P < 0.03; and t41 = 3.82, P < 0.0005, respectively) and P50 (t30 = 4.78, P < 0.0001; and t43 = 4.66, P < 0.0001, respectively). The fourth- and higher-order dendrites, which mainly develop after the period during which we impaired DISC1 C-terminal interactions, showed normal spine density at P28 (t20 = 0.96, P = 0.35; interaction between dendrite order and genotype F4,104 = 4.48, P < 0.005) and at P50 (t30 = –1.318, P = 0.20; interaction between dendrite order and genotype F4,117 = 7.29, P < 0.0001). The spine density deficit was only found on basal dendrites, not on apical dendrites (F4,131 = 0.86, P = 0.49).

Fig. 2 Enduring effects of transient impairment of DISC1 C-terminal interactions at P7 on dendritic spines.

(A) Example of L2/3 dendrites showing spines and dendritic order. Scale bar, 10 μm. (B) DISC1cc mice had lower spine density on second- and third-order dendrites at P28 and P50. Spine density was lower at P8 on third-order dendrites in DISC1cc mice. (*P < 0.05, **P < 0.01, ***P < 0.001.) (C and D) DISC1cc mice had a lower density of mushroom spines and a higher density of thin spines on second- and third-order dendrites. (ANOVA: interaction between dendrite order and genotype for mushroom spines: F1,124 = 58.64, P < 0.0001 and for thin spines F1,124 = 7.40, P < 0.01).

The period when DISC1 C-terminal signaling is critical for adult plasticity (P7 to P13) corresponds to a period of rapid synaptogenesis across the brain, as well as in L2/3 of barrel cortex (18), when experience is necessary for AMPA insertion within synapses (19). Altered neonatal experience during this period leads to defocused receptive fields in adulthood (20). As the size of spine heads are correlated with their AMPA receptor content (21, 22), we investigated spine head size and classification. At P50, there were fewer mushroom spines (both as a percentage of the whole and in absolute terms) on the second- and third-order dendrites of DISC1cc mice than on their first-, fourth-, and fifth-order dendrites (t35.8 = 8.76, P < 0.0001), and fewer than on the second- and third-order dendrites of wild-type mice (t49.9 = 8.72, P < 0.0001). Furthermore, there were more thin spines on the second- and third-order dendrites in the DISC1cc mice (t75 = 3.07, P < 0.005 compared with wild types, and t68 = 4.10, P < 0.0005 compared with other dendrite orders within the DISC1cc mice). Finally, the spine heads were smaller on the thin spines in the DISC1cc mouse second- and third-order dendrites than in the wild types (t76 = 3.31, P < 0.01) (Fig. 2). These findings imply a lower level of AMPA receptor insertion in DISC1 mutants.

We investigated synaptic function further in DISC1 mutants and found that, whereas the AMPA/N-methyl-d-aspartate (NMDA) ratio followed a normal developmental trajectory up to P14, it diverged at P28 (t18 = 2.33, P < 0.05) and did not recover by P50 (t18 = 3.29, P < 0.01) (Fig. 3A). Consistent with this finding, silent synapses were present in DISC1 L2/3 cells at P50 (Fig. 3B), whereas in wild types they had converted to functional synapses by this age (23). The NMDA component of the synaptic response was also immature and contained a higher proportion of GluN2B versus GluN2A subunits than normal (t10 = 3.9, P < 0.005) (Fig. 3C) (24, 25). In contrast, inhibition appeared to be unaffected in DISC1cc mice (fig. S8). Low levels of GluN2A and AMPA receptors are consistent with the spine size defects, which implies that glutamate receptor insertion is affected by transient disruption of DISC1 C-terminal interactions in early development. These factors predict that synaptic plasticity should also be deficient in DISC1cc mice (21, 22, 25).

Fig. 3 Persistent functional consequences of transient impairment of DISC1 C-terminal interactions at P7.

(A) The AMPA/NMDA ratio of WT and DISC1cc mice diverge after P14, and this ratio in DISC1cc mice remains at a low level into adulthood (P50 WT ratio is 8.29 ± 0.97; DISC1 ratio is 4.52 ± 0.61; and t15 = 3.29, P < 0.01). (B) Of P50 DISC1cc recordings, 50% had minimal-stimulus excitatory postsynaptic current success rates at +40mV higher than rates at –70mV, indicative of the presence of silent synapses. (C) NMDA currents in adult DISC1cc mice (red bar) showed enhanced sensitivity to ifenprodil application when compared with WT mice (black bar; DISC1cc 0.67 ± 0.08, WTs 1.01 ± 0.04, t10 = 3.9, P < 0.005). (D) LTD in WT mice was consistent at P14, P28, and P50 but was absent at P100 [average depression of 47 ± 13% at P14 (P < 0.01), 47 ± 6% at P28 (P < 0.001), 36 ± 6% at P50 (P < 0.001), and –1 ± 14% at P100 (P > 0.05)]. (E) Normalized peak excitatory postsynaptic potential amplitude is plotted versus time. Transient expression of mutant DISC1cc at P7 abolishes the capability for intercolumnar LTP in L2/3 at P28 and (F) at P50 (effect of genotype F1,74 = 14.27, P < 0.0003 and not age F1,74 = 0.13, P < 0.71, ANOVA). The percentage of cells showing statistically significant LTP drops from 33% in WT mice to 5% in DISC1cc mice (P28) and 43% in WT mice to 9% in DISC mice at P50. (G) Average LTD values are similar in WT and DISC1cc mice (F1,18 = 3.44, P < 0.08, ANOVA), although the percentage of cells showing LTD drops from 90% in wild types to 40% in DISC1cc mice. (H) Complete whisker deprivation unmasks the reversal of LTD that depends on adenosine 3′,5′-monophosphate–dependent protein kinase (26), and this is unaffected in the adult mouse by P7 DISC1cc activation (F1,19 = 0.16, P < 0.87, ANOVA).

On investigation, we found that long-term potentiation (LTP) was absent in the P7 tamoxifen-treated DISC1cc mice at P28 and P50 (Fig. 3, E and F), which indicated that development of LTP was abolished rather than delayed. Long-term depression (LTD) was affected although not abolished: The time-course of LTD was slower and the probability of LTD induction was lower in DISC1cc mice (fig. S9), although it was possible to induce LTD in the mutants as in the wild types (Fig. 3G). This suggested that it might be possible to reverse LTD in these synapses despite their lack of LTP. Previous studies had shown that 7 days of complete bilateral whisker deprivation can occlude LTD in the barrel cortex and reset the synapses to a state that favors LTD reversal (26). We found that it was possible to reverse LTD in the completely whisker-deprived DISC1cc mouse (Fig. 3H). These results show that developmental disruption of DISC1 signaling blocks or impairs selective aspects of synaptic plasticity.

Adult plasticity is different from many forms of developmental plasticity (27, 28). In the somatosensory and visual cortex, adult plasticity is dependent on CaMKII and closely related to LTP (29, 30). Developmental forms of synaptic plasticity, such as tumor necrosis factor-α–dependent synaptic scaling, experience-dependent depression, and LTD are gradually reduced with age (30, 31). We found that the normal period of LTD expression in the barrel cortex ends between P50 and P100 (Fig. 3D), after which LTD and reversal of LTD are not available modes of plasticity. Therefore, the loss of adult LTP only becomes critical at an age when developmental forms of plasticity have decreased to low levels. The latent effect of ablating prospective adult plasticity during an early critical period may therefore help explain the late onset of some schizophrenia symptoms.

How might a loss of LTP affect psychiatric conditions? Working memory is defective in schizophrenia and relies on persistent modes of network firing (32). Persistent neuronal activity requires formation of attractor states in neuronal networks, as has recently been shown in monkey prefrontal cortex during context-dependent integration of visual information (33). Therefore, a loss of plasticity such as we describe here is likely to disrupt working memory function by preventing formation of stable attractor states.

The C-terminal domain of DISC1 expressed in the DISC1cc mouse is known to reduce wild-type DISC1-Nudel and DISC1-Lis interactions (15). DISC1-Nudel interactions are thought to depend on the C-terminal domain’s ability to form dimeric and tetrameric states (16). DISC1 and Nudel interact strongly at P7, less so by P16, and negligibly by 6 months (14). The DISC1-Nudel complex is therefore available to be disrupted only when spines form rapidly on cortical neurons during the critical period we describe here for adult plasticity. Nudel and DISC1 also both bind to Lis1 (14), and Lis1 haploinsufficiency has been shown to decrease spine density, specifically on second- and third-order dendrites (34), in striking similarity to the present results. Human induced pluripotent stem cells from schizophrenia and depression sufferers carrying a DISC1 C-terminal mutation also exhibit deficits in synapse formation (35). By restricting DISC1 C-terminal dysfunction to a short period of development, we have been able to show that adult plasticity (i) depends on synapse formation during this early critical period, (ii) cannot be recovered despite continued expression of normal DISC1, and (iii) is independent of DISC1-Nudel interactions in adulthood.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

References (3646)


  1. Acknowledgments: We acknowledge the support of the UK Medical Research Council and U.S. National Institute of Mental Health, NIH, to K.F. for this work. We also thank T. Gould for histology, A. Siva for discussions on DISC1cc, and S. Butt for suggestions on methods of tamoxifen delivery. Supplement contains additional data.
View Abstract

Navigate This Article