Postnatal Developmental Trajectories of Neural Circuits in the Primate Prefrontal Cortex: Identifying Sensitive Periods for Vulnerability to Schizophrenia

Introduction

Substantial evidence supports the view that schizophrenia is a disorder of neurodevelopment,¹ with deficits in cognition often presenting and progressing years before the onset of psychosis.² Working memory (WM), the ability to transiently maintain and manipulate a limited amount of information in order to guide thought or behavior, is a prototypical cognitive deficit in schizophrenia.³ Performance on WM tasks depends, at least in part, upon dorsolateral prefrontal cortex (DLPFC) activity,⁴ and both WM performance and associated patterns of DLPFC activity continue to mature through late adolescence,⁵ the period during which psychosis onset is most frequent.

In schizophrenia, the DLPFC exhibits altered activation during WM tasks⁶ and smaller gray matter volume.⁷ Interestingly, the normal decline in DLPFC gray matter that occurs during adolescence appears to be accentuated in schizophrenia, suggesting alterations in the pruning of excitatory synapses that occur during adolescence in the DLPFC of healthy individuals.⁸ Together, these findings emphasize the importance of understanding the developmental trajectories of DLPFC circuits that subserve WM because these may reveal which elements of the circuitry are preferentially vulnerable to environmental events at specific stages of development.

Risk for schizophrenia appears to result, in part, from a range of environmental exposures that may occur at different stages of postnatal development, defined for the purposes of this review as the period extending from birth to the end of adolescence. These exposures may include obstetrical complications,⁹ minority group position and urban residence during childhood, and frequent cannabis use during early adolescence.¹⁰ Knowledge of how and when these environmental exposures could alter the development of DLPFC circuitry, and thereby contribute to impaired WM, is critical for understanding the pathogenesis of schizophrenia.

Macaque monkeys provide an excellent model system for examining the postnatal refinements in DLPFC circuitry that contribute to WM maturation. As in humans, monkeys exhibit a progressive age-related improvement in WM performance that extends through adolescence. The ability to perform WM tasks first appears between 2 and 4 months of age in rhesus monkeys and then progressively improves to reach adult levels of performance after 3 years of age.¹¹ This improved performance with age appears to be dependent upon an increased engagement of the DLPFC in task performance. For example, reversible inactivation of the DLPFC does not impair WM performance in monkeys 9–16 months of age, produces modestly disrupted performance in animals 19- to 31-month old, and substantially impairs performance in animals over 3 years of age.^12,13 In electrophysiological studies, some DLPFC neurons exhibit elevated firing rates during the delay period of WM tasks, and the loss of this delay-related activity is associated with errors in task performance.¹⁴ Between 12 and 36 months of age, the proportion of DLPFC neurons that exhibit delay period activity doubles, suggesting that developmental changes in DLPFC circuitry facilitate the recruitment of more neurons to this functional role.¹³

Consequently, in this review, we consider how developmental refinements in glutamatergic and γ-aminobutyric acid (GABA)-ergic circuits in the monkey DLPFC may provide insight into the contribution of different environmental risk factors to the structural and molecular alterations found in the DLPFC in individuals with schizophrenia.

Glutamatergic Circuit Alterations in Schizophrenia

In the DLPFC, pyramidal neurons are the principal excitatory output cells and the targets of the majority of glutamatergic-containing axon terminals.¹⁵ While the number of pyramidal neurons appears to be unchanged in schizophrenia,^16,17 a reduction in the amount of cortical neuropil has been reported.¹⁸ The reduced neuropil may be the consequence of fewer axon terminals, as suggested by findings in postmortem studies of schizophrenia of lower levels of proteins present in axon terminals,^19,20 and of fewer dendritic spines, the principal targets of excitatory synapses to pyramidal neurons.^21,22

Cortical pyramidal neurons can be categorized based on their laminar position and the principal target of their axonal projection. For example, many layers 2 and 3 pyramidal neurons send collaterals within the same cortical area (intrinsic) or between cortical areas (associational) (figure 1A). In contrast, layer 5 pyramidal neurons tend to project to the striatum and other subcortical regions and those from layer 6 tend to project to the thalamus. In the DLPFC of subjects with schizophrenia, dendritic spine density was significantly lower in deep layer 3 pyramidal cells, relative to both normal and psychiatrically ill comparison subjects, whereas spine density was only modestly lower in superficial layer 3 pyramidal neurons and unchanged in pyramidal neurons in layers 5 and 6.^22,23 Consistent with these findings, the somal volume of deep layer 3 pyramidal neurons was lower in schizophrenia in the DLPFC and other cortical regions,^24–27 whereas the volume of layer 5 pyramidal neurons was unchanged²⁸ and the layer 3 findings did not appear to be attributable to medication use or length of illness.²⁹ In concert, these findings suggest that in the DLPFC and other cortical regions in schizophrenia, basilar dendritic spine density is lower and somal volume is smaller in deep layer 3 pyramidal neurons (table 1), these alterations are specific to or at least most prominent in deep layer 3, and these differences reflect the underlying disease process and not confounding factors.

Postnatal Development of Glutamatergic Circuitry in the DLPFC

Dendritic spine density on layer 3 pyramidal neurons in monkey DLPFC undergoes marked changes across postnatal development.³⁰ Spine density increases substantially during late gestation and the first 3 months postnatal, reaches a plateau that is maintained until 15–18 months of age and then declines during adolescence until stable adult levels are achieved at 3–4 years of age (figure 2A). Consistent with the findings that dendritic spines are the main site of excitatory synaptic input onto pyramidal cells and that all mature dendritic spines contain an excitatory synapse,³¹ the density of excitatory synapses (defined morphologically by the presence of a thick postsynaptic density) changes in a similar age-related fashion in both monkey and human DLPFC (figure 2A).^32,33 In humans, this synaptic pruning could contribute to the decrease in cortical gray matter thickness that occurs during adolescence.⁸ Importantly, the late developmental refinements in excitatory synapse and spine number are more pronounced in layer 3 than in layers 5–6,³² consistent with the laminar location of pyramidal cell alterations in schizophrenia.

During early brain development, synaptic elimination appears to be directed at synapses that are functionally immature.³⁴ Immature glutamatergic synapses are relatively weak and their maturation involves an activity-dependent increase in strength of excitation. Such activity-dependent strengthening might underlie synapse stabilization and thus mark for elimination the immature synapses that are not strengthened. Understanding if the synapses that are eliminated during adolescence in the primate DLPFC are functionally immature is essential for determining how synaptic pruning informs the developmental improvements observed in WM ability. For example, if the eliminated synapses are functionally immature, then they may not contribute to, detract from, the functional properties of DLPFC circuitry required for WM. That is, neurotransmission at immature glutamatergic synapses has (1) very low α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptor contribution, rendering these synapses silent at the resting membrane potential and (2) relatively high probability of glutamate release, such that these synapses are not able to be repetitively activated without quickly exhausting glutamate vesicle pools.^35,36 On the other hand, pruning of functionally mature synapses could help constrain the range of inputs that a given pyramidal cell receives and thereby perhaps contribute to the improved accuracy and greater resistance to distracters on WM tasks observed with age.^37,38

This question was addressed in a study of living slices of the DLPFC prepared from monkeys of different ages.³⁹ Excitatory inputs to layer 3 pyramidal neurons in 3-month-old monkeys had immature functional properties, including a higher probability of glutamate release and lower AMPA/NMDA receptor ratio. In contrast, the excitatory synaptic inputs to layer 3 pyramidal neurons from 15-month-old monkeys (preadolescence and prepruning) had mature functional properties that were similar to those observed in neurons from 42-month-old (postadolescence and postpruning) and 84-month-old (adult) animals (figure 2A). Therefore, the contribution of functionally immature synapses appears to decrease significantly before synapse elimination begins and to remain essentially constant thereafter. These data suggest that the substantial remodeling of excitatory connectivity of the primate DLPFC during adolescence primarily involves the elimination of mature synapses and that some other factor, such as the neuronal source of input or the postsynaptic target, somehow tags mature synapses for pruning. For example, of the axon collaterals furnished by layer 3 pyramidal neurons in monkey DLPFC (figure 1A), the intrinsic branches appear to be pruned to a greater degree than the associational branches.⁴⁰ Although speculative, the elimination of excessive functional glutamatergic synapses might reduce the number of inputs from distracting stimuli during WM performance, improve the circuit’s ability to process relevant stimuli, and thus contribute to the improved WM performance observed during adolescence.

GABAergic Circuit Alterations in Schizophrenia

Multiple studies have consistently found that transcript levels for the 67-kDa isoform of the GABA-synthesizing enzyme glutamic acid decarboxylase (GAD67) are decreased in the DLPFC of subjects with schizophrenia.⁴¹ In contrast, messenger RNA (mRNA) levels for the 65-kDa isoform of GAD (GAD65) have been reported to be unchanged, or only slightly lower, in schizophrenia.^42–44 DLPFC GABA neurons can be differentiated on the basis of a number of molecular, physiological, and anatomical properties (figure 1B). For example, the neuropeptide cholecystokinin (CCK) is heavily expressed in GABA neurons that do not contain either the calcium-binding protein parvalbumin (PV) or the neuropeptide somatostatin (SST).⁴⁵ The axon terminals of CCK-containing large basket neurons, which target selectively pyramidal neuron cell bodies, contain type I cannabinoid receptors (CB1R) (figure 1C)^46,47; interestingly, mRNA and protein levels of CB1R are lower in schizophrenia.^48,49

The affected GABA neurons in schizophrenia also include the PV subclass, which can be subdivided into 2 major types based on the principal target of their axon terminals (figure 1C). Axon terminals from PV basket cells target the perisomatic region (soma and proximal dendrites) and those from chandelier cells exclusively target the axon initial segments (AIS) of pyramidal neurons.⁵⁰ In the DLPFC of individuals with schizophrenia, although the number of PV neurons appears to be unchanged,⁵¹ expression of PV mRNA is decreased.⁵² Furthermore, approximately half of the PV mRNA containing neurons lack detectable levels of GAD67 mRNA.⁵² In contrast, the subclass of interneurons that expresses the calcium-binding protein calretinin and comprises ∼50% of GABA neurons appears to be unaffected in schizophrenia.⁵²

In the DLPFC of individuals with schizophrenia, immunoreactivity for the GABA membrane transporter (GAT1) appears to be reduced in the characteristic chandelier cell axon terminal arrays referred to as cartridges.⁵³ In the axon initial segment (AIS) of pyramidal cells that receive synaptic inputs from chandelier neuron axon cartridges, immunoreactivity for the GABA_A receptor α2 subunit, which is normally the dominant GABA_A receptor α subunit present in the AIS,⁵⁴ is markedly increased in schizophrenia.⁵⁵ In addition, the detectability of ankyrin-G, a key structural protein in the AIS, is also lower in the DLPFC in schizophrenia.⁵⁶ Interestingly, each of these alterations in chandelier neuron connectivity is most pronounced in layer 3. In contrast, the density of PV-labeled axon terminals, presumably from basket neurons, and the expression of the GABA_A receptor α1 subunit, which is postsynaptic to PV basket cell inputs, are both reduced in DLPFC layer 3 in schizophrenia^43,57,58 (table 1). These changes appear to be specific to the pathophysiology of schizophrenia because they are not observed in individuals with other psychiatric disorders or in monkeys exposed chronically to antipsychotic medications.^29,58

Postnatal Development of GABAergic Circuitry

Environmental Exposures During Sensitive Periods and the Risk of Schizophrenia

A range of early life environmental exposures may increase the risk for the later development of schizophrenia (figure 2C). While some of these risk factors cover an extended period of development, their impact on the risk of developing schizophrenia may be amplified during discrete periods of postnatal development that have enhanced sensitivity due to the high rate of change in specific elements of DLPFC circuitry. For example, the apparent dose-related effect of cannabis use on schizophrenia risk may be most prominent with repeated exposure prior to the age of 16 years,⁹⁵ which corresponds with the developmental period when CB1R protein levels are increasing in the middle layers of the DLPFC,⁶² suggesting that vulnerability to cannabis exposure is most pronounced when these cortical circuits are being refined. Activation of CB1R by exogenous cannabinoids, such as the principal psychoactive component in cannabis, Δ-9-tetrahydrocannabinol, suppresses GABA release from the axon terminals of CCK neurons⁹⁶ (figure 1C). The repetitive use of cannabis during adolescence may blunt the developmental increase in CB1R protein that normally occurs during this period of development,⁶² perhaps contributing to the lower levels of CB1R protein observed in schizophrenia.^47,49 Furthermore, unlike the synapse-specific depolarization-induced suppression of inhibition that normally occurs when a pyramidal neuron synthesizes endogenous cannabinoids on demand in response to synaptic activity,⁵⁹ exogenous cannabinoids may persistently and widely suppress CB1R-containing inputs onto pyramidal neurons in a non-specific manner. This widespread suppression of GABA release from CCK neurons may subsequently contribute to excessive pyramidal cell activity, resulting in altered excitatory-inhibitory balance. To prevent runaway excitation and cell injury, affected pyramidal neurons may compensate by evoking an exaggerated pruning of excitatory synapses and eliminating dendritic spines in order to restore excitatory-inhibitory balance in the DLFPC; such a scenario might account for the lower pyramidal neuron spine density present in schizophrenia. Moreover, the output of CCK neurons also influences the activity of PV neurons,⁹⁷ suggesting that cannabis use during adolescence could indirectly disrupt the normal development of PV basket and chandelier cells in schizophrenia.

This complex and protracted maturation of local neural circuits, especially in layer 3 of the DLPFC, provides a number of opportunities for any disturbances, even subtle ones, to have their effects amplified because they alter the trajectories of the developmental events that follow. Indeed, the similarities between the components of DLPFC circuitry that exhibit marked changes during postnatal development and those reported to be altered in postmortem studies of schizophrenia suggest that the alterations of these markers in schizophrenia reflect a disturbance in their patterns of development and might explain how a range of environmental factors (figure 2C) could all be associated with an increased risk of schizophrenia later in life. That is, different exposures might each alter, albeit in different ways, the neural circuitry that is essential for WM function, resulting in WM performance levels in schizophrenia that resemble the immature levels seen in children.^37,98 The different environmental exposures during discrete sensitive periods of development could thus each alter the maturation of DLPFC circuitry, leading to a common outcome of impaired cognition in schizophrenia but with differences in the nature and severity of the impairment across individuals depending upon the type and timing of the exposure.

Recently, it has been suggested that schizophrenia be reconceptualized as a syndrome with psychosis as a late, potentially preventable outcome of the illness.⁹⁹ The findings summarized in this review highlight that DLPFC circuits are dynamic in their structural and molecular properties from birth through adolescence, with substantial changes in occurring in markers of both glutamatergic and GABAergic neurotransmission. Importantly, in the primate DLPFC, these developmental trajectories appear to be temporally associated with a range of environmental risk factors for schizophrenia. This association raises the possibility that early intervention to interrupt or reverse pathological circuit development may help reduce the decline in cognitive function and alter the course of the illness.

Funding

National Institute of Health grants (MH051234, DA023109, MH084053).