In the primate dorsolateral prefrontal cortex (DLPFC), the density of excitatory synapses decreases by 40–50% during adolescence. Although such substantial circuit refinement might underlie the adolescence-related maturation of working memory performance, its functional significance remains poorly understood. The consequences of synaptic pruning may depend on the properties of the eliminated synapses. Are the synapses eliminated during adolescence functionally immature, as is the case during early brain development? Or do maturation-independent features tag synapses for pruning? We examined excitatory synaptic function in monkey DLPFC during postnatal development by studying properties that reflect synapse maturation in rat cortex. In 3-month-old (early postnatal) monkeys, excitatory inputs to layer 3 pyramidal neurons had immature properties, including higher release probability, lower α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA)/N-methyl-D-aspartate (NMDA) ratio, and longer duration of NMDA-mediated synaptic currents, associated with greater sensitivity to the NMDA receptor subunit B (NR2B) subunit–selective antagonist ifenprodil. In contrast, excitatory synaptic inputs in neurons from preadolescent (15 months old) and adult (42 or 84 months old) monkeys had similar functional properties. We therefore conclude that the contribution of functionally immature synapses decreases significantly before adolescence begins. Thus, remodeling of excitatory connectivity in the DLPFC during adolescence may occur in the absence of widespread maturational changes in synaptic strength.
The extended period of adolescence in primates is characterized by both marked changes in emotion, cognition, and behavior (Steinberg 2005) and a substantial refinement of cortical circuits. In the neocortex of both monkeys and humans, exuberant excitatory synapses produced by rapid synaptogenesis during the perinatal period are eliminated during adolescence when massive synaptic pruning occurs (Rakic et al. 1986; Huttenlocher and Dabholkar 1997). In the monkey dorsolateral prefrontal cortex (DLPFC), this developmental trajectory in excitatory synaptic density is paralleled by similar changes in the density of dendritic spines, the principal site of excitatory inputs to cortical pyramidal cells (Anderson et al. 1995).
The substantial remodeling of excitatory connections in the DLPFC during adolescence appears to contribute to the normal functional maturation of this region. For example, structural neuroimaging studies in humans demonstrated that adolescence is associated with a substantial decrease in cortical gray matter thickness (Giedd et al. 1999), a decrease usually interpreted to result from the massive synaptic pruning occurring during this period (Gogtay et al. 2004). Neuropsychological studies combined with functional neuroimaging showed that the marked improvements in working memory function during adolescence are accompanied by an increased recruitment of DLPFC activity (Luna et al. 2004; Crone et al. 2006). In macaque monkeys, the contribution of the DLPFC to behavioral performance in tasks that require working memory becomes fully mature near the end of adolescence (Goldman and Alexander 1977).
Synaptic pruning during adolescence has also been linked to neurodevelopmental models of schizophrenia because the clinical symptoms, including working memory impairments, typically arise near the end of adolescence, and early adulthood (Lewis and Levitt 2002). The exuberant synapses present in the DLPFC before adolescence have been hypothesized to compensate for a dysfunction in excitatory transmission due to alterations in the expression of certain synaptic proteins (Mirnics et al. 2001; Owen et al. 2005).
Understanding how reductions in excitatory synaptic density during adolescence could contribute to such age- and disease-related changes in the function of the DLPFC depends, in part, on the knowledge of the functional properties of the synapses present before adolescence-related pruning. One possibility is that the eliminated synapses are functionally immature. For example, during early brain development some synapses are selectively stabilized by activity-dependent processes, which increase their strength, whereas weak immature synapses that are not strengthened are eliminated (Katz and Shatz 1996; Waites et al. 2005; Le Be and Markram 2006). Consistent with this view, the dendritic spines preferentially eliminated during adolescence in mouse neocortex have immature morphological features (Zuo, Lin, et al. 2005). Alternatively, it might be that the functional maturation of most synapses in the DLPFC occurs before adolescence and that another factor, such as the neuronal source of the inputs, somehow marks functionally mature synapses for elimination (Woo et al. 1997).
In order to begin to discriminate between these 2 alternatives, we studied during postnatal development the physiological properties of excitatory synapses in a living slice preparation of the monkey DLPFC, focusing on properties that reflect glutamate synapse maturation in rat cortex. We found that early in postnatal development, a significant proportion of excitatory inputs have immature properties. In addition, the contribution of immature synapses declines significantly before the onset of adolescence. Consequently, before synaptic pruning begins, the functional properties of excitatory synaptic inputs are indistinguishable from those observed in neurons from adult monkeys. Our results therefore indicate that the changes in DLPFC circuits associated with the maturation of working memory performance during adolescence involve a decrease in excitatory synapse number in the absence of substantial maturational changes in synaptic strength.
Brain Slice Preparation
These experiments were carried out with tissue obtained from 14 female rhesus monkeys (Macaca mulatta) supplied by the University of Pittsburgh Primate Research Center. Housing and experimental procedures were conducted in accordance with United States Department of Agriculture and National Institutes of Health guidelines and with approval of the University of Pittsburgh's Institutional Animal Care and Use Committee. All animals 42 months of age or less were bred at this facility. As described previously (Cruz et al. 2003), animals were housed with their mothers until 6 months of age when they were placed by groups in the same social setting. Older animals were housed either in pairs or in single cages in the same setting. All animals were experimentally naive at the time of entry into this study.
To determine the functional properties of excitatory synapses, we prepared brain slices at 4 postnatal time points. The first group of animals (n = 5 animals and 5 hemispheres) was studied at 3 months of age, when the period of rapid synaptogenesis is close to complete and the densities of excitatory synapses and layer 3 pyramidal cell spines are at peak values (Rakic et al. 1986; Anderson et al. 1995). The second group of animals (n = 3 animals and 6 hemispheres) was studied at 15 months of age, just prior to the onset of the period of rapid synaptic elimination, when synapse and spine densities are not appreciably different from those present at 3 months of age. The third group of animals (n = 4 animals and 8 hemispheres) was studied at 42 months of age, when the period of adolescence-related pruning of excitatory synapses and spines is essentially complete. A fourth group of animals (n = 2 animals and 3 hemispheres) was studied at 84 months of age to determine whether the functional properties of excitatory synapses are stable across early adulthood.
Tissue blocks containing portions of DLPFC areas 9 and 46 were obtained from one or both hemispheres of each animal. Each of the 3-month-old and one of the 84-month-old animals were deeply anesthetized and perfused transcardially with cold artificial cerebrospinal fluid (ACSF) solution of the following composition (in mM): sucrose 210.0, NaCl 10.0, KCl 1.9, Na2HPO4 1.2, NaHCO3 33.0, MgCl2 6.0, CaCl2 1.0, glucose 10.0, and kynurenic acid 2.0; pH 7.3–7.4 when bubbled with 95% O2–5% CO2, as previously described (Gonzalez-Burgos et al. 2004). The brain was then rapidly removed and bilateral DLPFC blocks were prepared. For all other animals, an initial tissue block was removed from one hemisphere using a previously described surgical procedure (Gonzalez-Burgos et al. 2004) and then a second DLPFC tissue block was removed 1–2 weeks later following the transcardial cold ACSF perfusion procedure described above. When 2 tissue blocks were removed per animal in separate surgical procedures, the locations of the blocks were offset in the rostral–caudal axis, so that nonhomotopic portions of the DLPFC were studied from each hemisphere. Previous studies have shown that the first procedure does not alter the physiological or anatomical properties of the neurons and local circuits present in the tissue obtained in the second hemisphere (Gonzalez-Burgos et al. 2000).
Cortical slices (350 μm thick) were cut in the coronal plane using a vibrating microtome (VT1000S, Leica Microsystems, Nussloch, Germany) in ice-cold ACSF. Immediately after cutting, slices were transferred to an incubation chamber maintained at room temperature and filled with a solution of the following composition (in mM): NaCl 126.0, KCl 2.0, Na2HPO4 1.2, glucose 10.0, NaHCO3 25.0, MgCl2 6.0, and CaCl2 1.0; pH 7.3–7.4 when bubbled with 95% O2–5% CO2.
For recording, slices were submerged in a chamber superfused at a rate of 2–3 ml/min with a solution that was bubbled with 95% O2/5% CO2 and maintained at 30–32 °C. The superfusion solution had the following composition (in mM): NaCl 126.0, KCl 2.5, Na2HPO4 1.2, Na2HCO3 25.0, glucose 10.0, CaCl2 2.0, MgCl2 1.0, and bicuculline methiodide 0.02. In some experiments, gabazine (0.02 mM) was used instead of bicuculline, to block γ-aminobutyric acid A (GABAA) receptor channels. Using infrared differential interference contrast video microscopy (Stuart et al. 1993), whole-cell recordings were obtained from visually identified pyramidal neurons in layer 3 of DLPFC areas 9 and 46. Recording micropipettes pulled from borosilicate glass had a resistance of 3–5 MOhm when filled with a solution of the following composition (in mM): CsGluconate 120.0, NaCl 10.0, ethyleneglycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid 0.2, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid 10.0, MgATP 4.0, NaGTP 0.3, NaPhosphocreatine 14.0, and biocytin 0.5%. The pH of the pipette solution was adjusted to 7.2–7.3 using CsOH. Recordings were performed using Axoclamp 200A, Axopatch 1C, or Axopatch 1D amplifiers (Axon Instruments, Union City, CA) operating in voltage-clamp mode without series resistance compensation. Signals were low-pass filtered at 3 kHz, digitized at 10 or 20 kHz, and stored on disk for off-line analysis. Data acquisition and analysis were performed using customer-made programs written in LabView (National Instruments, Austin, TX).
The ability to voltage clamp the postsynaptic membrane at steady-state potentials was assessed in each cell by recording the excitatory postsynaptic currents (EPSCs) at different somatic holding potentials. The reversal potential of the EPSCs typically had positive values close to 0 mV. If the EPSC reversal potential had values positive to +15 mV, the recordings were not considered for data analysis. During the experiment, the series resistance was continuously monitored measuring the current evoked by a short voltage step. If the series resistance increased more than 20%, the recordings were excluded from data analysis.
Focal Extracellular Stimulation
To elicit α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptor–mediated EPSCs (AMPA EPSCs) or N-methyl-D-aspartate (NMDA) receptor–mediated EPSCs (NMDA EPSCs), focal extracellular stimulation was applied using theta glass pipettes (1.5 mm outside diameter, Warner Instruments Corporation, Hamden, CT) pulled to a tip diameter of 2–3 μm and filled with freshly oxygenated extracellular solution. Chlorided silver wires placed inside each compartment of the theta glass were connected to a stimulus isolation unit (Model A350D-A, World Precision Instruments, Sarasota, FL) to apply bipolar focal stimulation. The stimulation electrodes were placed in layer 3 at ∼100 μm above (or below) and ∼50–100 μm lateral to the soma of the recorded neurons, by means of motorized micromanipulators (Model MP-285, Sutter Instruments Co., Novato, CA). Stimulation current parameters and electrode position were adjusted to elicit small amplitude responses resembling monosynaptic EPSCs, as described previously (Gonzalez-Burgos et al. 2000). Stimuli were applied at a baseline frequency of 0.1 Hz.
GABAA receptor–mediated currents were blocked by application of bicuculline methiodide or gabazine (20 μM) to the extracellular solution. To block the NMDA component of the EPSCs, 100 μM of D,L (-)-2-amino-5-phosphonopentanoic acid (D,L-AP5) was used. AMPA receptor–mediated EPSCs were blocked by application of 20 μM 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). To compare the probability of glutamate release across age groups, we examined the rate of blockade of NMDA EPSCs by 25 μM of (+)-5-methyl-10,11-dihydro-5H-dibenzo(a,d)cyclohepten-5,10-imine maleate (MK801). All reagents were obtained from Sigma-Aldrich (St. Louis, MO).
Analysis of the AMPA/NMDA Ratio of EPSCs
To determine the AMPA/NMDA ratio of EPSCs evoked by extracellular stimulation, cells were initially held at −80 mV and then the holding potential was changed in +20 mV increments until reaching +40 mV. At least 20 sweeps were collected at each holding potential and averaged for data analysis. After recording EPSCs at +40 mV, the holding potential was returned to −80 mV and the EPSCs recorded at −80 mV at the beginning and end of the voltage-clamp protocol were compared. At the baseline stimulation frequency of 0.1 Hz employed in our experiments, no evidence of induction of synaptic plasticity was observed after voltage-clamping neurons at positive membrane potentials. These results are in agreement with previous findings showing that induction of plasticity by pairing presynaptic stimulation with postsynaptic depolarization requires a stimulation frequency higher than 0.1 Hz (Isaac et al. 1997; Rumpel et al. 1998; Montgomery et al. 2001; Barria and Malinow 2005). Changes of more than 10% in the EPSC amplitude recorded at −80 mV were exclusively due to changes in the series resistance. Such recordings were excluded from data analysis in the present study. The EPSCs recorded at −80 mV were readily blocked by application of the AMPA receptor antagonist CNQX and were not affected by application of the NMDA antagonist D,L-AP5. Thus, the peak amplitude of EPSCs recorded at −80 mV was used to estimate the AMPA receptor contribution to the EPSCs. To determine the NMDA receptor contribution, in individual neurons the amplitude of the current recorded at a holding potential of +40 mV was measured over a time period when the AMPA EPSC recorded at −80 mV decayed to 5% of its peak amplitude. When EPSCs were recorded at +40 mV the AMPA component decayed completely by that time because the driving force is significantly smaller at +40 compared with −80 mV. Therefore, small voltage-dependent changes in the decay kinetics of the AMPA EPSC as those previously reported (Cathala et al. 2005) do not affect our measurement of the NMDA EPSC. In some neurons, the AMPA/NMDA ratio was estimated by measuring the area under the AMPA and NMDA EPSC waveforms, which represents the synaptic charge transfer. The EPSC recorded at −80 mV was used to estimate the area of the AMPA EPSC waveform. The NMDA EPSC area was estimated by measuring the area of the EPSC recorded subsequently from the same neuron at a holding potential of +40 mV and after the AMPA EPSC was blocked by application of CNQX. Stimulus intensity, stimulus duration, and the position of the stimulation electrode were kept constant before and after CNQX application to stimulate the same presynaptic fibers. The area under the curve was calculated between EPSC onset and the time at which the EPSC decayed to 5% of its peak amplitude.
Analysis of the Kinetics of AMPA and NMDA EPSCs
To compare the decay kinetics of AMPA EPSCs between age groups, EPSCs were recorded at a holding potential of −80 mV in the presence of a GABAA-receptor antagonist (either bicuculline or gabazine, 20 μM). To compare the decay kinetics of NMDA EPSCs, EPSCs were recorded at a holding potential of +40 mV. For both AMPA and NMDA EPSCs, at least 20 consecutive sweeps were recorded and averaged, and curve fitting was performed on the average trace. Single-exponential decay functions were fit to the data using curve-fitting routines in Igor (Wavemetrics, Lake Oswego, OR).
Rate of NMDA EPSC Block by MK801
MK801 is an open-channel blocker that inhibits NMDA receptor–activated currents in an essentially irreversible manner (Huettner and Bean 1988). Therefore, the rate of NMDA EPSC block by MK801 at a constant interstimulus interval is directly proportional to the probability that each stimulus evokes glutamate release (Pr), leading to a faster rate of NMDA EPSC block by MK801 (Rosenmund et al. 1993). Assessing the actual probability of release (Pr) requires knowledge of parameters related to the time course of the glutamate concentration transient in the synaptic cleft and the kinetics of NMDA channel opening (Rosenmund et al. 1993). Because such parameters were not available in the present study, only the rate of NMDA EPSC block was used for comparison of the Pr between age groups. NMDA EPSCs were isolated by application of bicuculline or gabazine and CNQX, 20 μM each. NMDA EPSCs typically had a stable peak amplitude of 50–150 pA when evoked at a frequency of 0.1 Hz. In preliminary experiments (n = 5), the stimulation was interrupted and resumed a few minutes later to verify that pausing the stimulation did not affect the NMDA EPSC amplitude (Fig. 6A). Before MK801 application, stimulation was paused, and the cells were held at −80 mV to avoid NMDA channel opening. MK801 (25 μM) was applied for at least 15 min to allow equilibration of its concentration in the extracellular space. Stimulation was then resumed to evoke NMDA EPSCs at a holding potential of +40 mV in the continuous presence of MK801 (Fig. 6A). In contrast to previously published findings (Wasling et al. 2004), we found that the amplitude of the first NMDA EPSC recorded after resuming the stimulation in the presence of MK801 was essentially identical to that of the last EPSC recorded right before interrupting the stimulation (Fig. 6A). For data analysis, NMDA EPSC amplitude in each experiment was normalized relative to that of the first NMDA EPSC recorded after resuming stimulation in the presence of MK801 (Fig. 6B). Single- or double-exponential decay functions were fit to the data using curve-fitting routines in Igor (Wavemetrics, Lake Oswego, OR).
All layer 3 pyramidal cells in which dendritic spine density was measured were located in dorsolateral PFC areas, specifically in the dorsal portion of area 9 or in area 46, in the medial bank of the principal sulcus. In slices from 15-month-old animals, 6 out of 11 neurons were located in area 9, and 4 out of 11 cells were in area 46. In slices from 42-month-old animals, 3 out of 12 neurons were located in area 9, and 6 out of 12 neurons in area 46. For the remaining cells (1/11 in 15-month-old and 3/12 in 42-month-old animals), the location in area 9 versus area 46 could not be determined reliably. Layer 3 pyramidal neurons were filled with 0.5% biocytin during recording and after recording, the slices were fixed in 4% paraformaldehyde and stored in an antifreeze solution (1:1, glycerol:ethylene glycol in 0.1 M phosphate buffer) at −80 °C until processed for visualization of the biocytin label. Slices were resectioned at 60 μm, incubated with 1% H2O2, and immersed in blocking serum containing 0.5% Triton X-100 for 2 h at room temperature. Slices were then incubated with Avidin Biotinylated enzyme complex-peroxidase and developed using the Nickel-enhanced diaminobenzidine chromogen to visualize the biocytin-labeled pyramidal neurons.
Previous studies have shown that the changes in excitatory synapse density during postnatal development occur throughout the entire dendritic tree of layer 2/3 pyramidal neurons in the monkey DLPFC (Anderson et al. 1995). In order to avoid any potential confounds associated with proximity to the recording site at the soma, the analysis of spine density was focused on the most distal portions of the apical dendrite (apical tuft) in layers 1 and 2. Using the Neurolucida tracing system (MicroBrightField, Williston, VT), a sampling frame (200 μm deep by 400 μm wide), containing 50 sampling squares (40 μm side), was superimposed on the image of the apical dendritic tuft. The top border of the sampling frame was oriented parallel to and 30 μm below the pial surface and was centered relative to the main trunk of the apical dendrite. Each sampling frame had 42% of the squares randomly selected for sampling. A different frame was used for sampling dendritic spines in each neuron. The selected portions of the apical dendrite were reconstructed using a 63× oil immersion objective with differential interference contrast optics using Neurolucida (MicroBrightField, Williston, VT) with the diameters of the dendritic shafts recorded. The total number of spines counted per neuron was divided by the total length of dendritic shaft sampled to obtain an estimation of dendritic spine density per cell. The total number of spines counted per neuron varied from 129 to 949 spines. Spine density was not correlated with the total dendritic length sampled in each neuron (Pearson correlation coefficient r = −0.04, P = 0.83). Mean spine density per neuron was then compared across age groups.
The method used for counting spines may underestimate the true spine density due to “hidden” spines, oriented perpendicular to the plane of section. Therefore our spine counts are relative rather than absolute. Most likely, however, this issue applies equally to all cells examined and is independent of age group. For example, because the number of hidden spines is related to dendritic diameter, age-related differences in diameter could bias the obtained relative spine densities. However, previous studies (e.g., Anderson et al. 1995) of animals in the age range used herein did not detect age-related differences in diameter of layer 3 pyramidal cell dendrites and no differences in mean dendrite diameter were found in the present study. Furthermore, previous studies using 3-dimensional reconstructions (see for example http://synapse-web.org/) of pyramidal cell dendrites did not find a particular bias in spine distribution around the dendritic shaft in developing or mature pyramidal neurons (Fiala and Harris 1999; Bourne et al. 2007). Thus, it seems unlikely that our relative spine density measures were substantially biased by age-dependent differences in the fraction of spines oriented perpendicular to the plane of section.
Previous studies demonstrated that changes in estrogen levels are associated with dendritic spine density in rat hippocampal pyramidal neurons and that estradiol administration increases spine density in pyramidal neurons from the DLPFC of ovariectomized monkeys (Tang et al. 2004; Hao et al. 2006). To determine if gonadal steroids levels could confound dendritic spine measures in this study, we measured the serum levels of estradiol and progesterone on each day when brain tissue was obtained from the 42- and 84-month-old animals. Spine density did not differ (P = 0.47) between or within animals when slices were prepared during the luteal or follicular phases of the menstrual cycle.
The data are expressed as mean ± standard deviation. Statistical significance of the difference between group means was determined employing Student's t-test or analysis of variance (ANOVA) followed by comparisons, as indicated in each particular case. Differences were considered significant when P < 0.05.
The Density of Dendritic Spines in Layer 3 Pyramidal Neurons Significantly Declines between 15 and 42 Months of Age
During adolescence, spine density significantly decreases throughout the dendritic tree of layer 3 pyramidal neurons in the monkey DLPFC (Anderson et al. 1995) in a manner consistent with the elimination of excitatory synapses in layers 1 to 4 (Rakic et al. 1986; Bourgeois et al. 1994). In contrast to spine density, the size of layer 3 pyramidal cell bodies or the length of their dendrites are close to adult values at 3 months of age and essentially identical to adult values at 15 months of age (Anderson et al. 1995). The absence of changes in layer 3 pyramidal cell size suggests that the significant decrease in spine density during adolescence represents a reduction in the total number of excitatory synaptic inputs. To confirm that an adolescence-related decrease in spine density was present in our neuron sample, we determined the spine density in the distal (apical tuft) dendrites of layer 3 pyramidal neurons that were filled with biocytin during electrophysiological recordings (Fig. 1A).
The spines in the distal dendrites of layer 3 neurons displayed various morphologies (Fig. 1B). However, using light microscopy, the morphology of many individual spines could not be reliably characterized. Therefore, spine density was measured independently of morphological spine type (Fig. 1C). Mean spine density was significantly lower (17%) in layer 3 pyramidal neurons from 42-month-old compared with the 15-month-old animals (Fig. 1D). It was previously shown that in this pyramidal cell class, spine density decreases by 40–50% by the end of adolescence when averaged across dendritic membrane compartments (Anderson et al. 1995). However, the actual magnitude of this decrease depends on the dendritic compartment studied. For example, in the apical tuft dendrites, spine density decreases by ∼15% between 18 and 32 months of age (Anderson et al. 1995). Therefore, the 17% decrease in spine density in the distal apical dendrites between 15 and 42 months of age reported here is consistent with the age-related changes described in previous studies.
Postsynaptic Function Changes between 3 and 15 Months of Age
In the developing rat cortex, immature synapses contain NMDA receptors but possess few or no AMPA receptors and are therefore weak or silent at the resting membrane potential (Malinow and Malenka 2002). Glutamate synapse maturation involves activity-dependent potentiation by AMPA receptor insertion into the postsynaptic membrane (Liao et al. 2001; Zhu and Malinow 2002; Lu and Constantine-Paton 2004; Mierau et al. 2004), which gradually decreases the proportion of synapses with low AMPA receptor content (Rumpel et al. 1998, 2004). To determine if the relative contribution of AMPA receptors to the EPSCs increases in layer 3 pyramidal neurons during maturation of the primate DLPFC, we first determined the AMPA/NMDA ratio in EPSCs evoked by focal extracellular stimulation (see Methods). Figure 2A shows examples of EPSCs from neurons of the 3- and 15-month-old age groups, recorded at somatic holding potentials of −80 and +40 mV. As displayed in Figure 2B, the AMPA/NMDA ratio increased significantly between 3 and 15 months of age, but did not show further changes at 42 and 84 months of age. An increase in the AMPA/NMDA EPSC ratio is consistent with a significant decrease in the proportion of functionally immature synapses between 3 and 15 postnatal months.
In rat neocortex and hippocampus, the NMDA EPSC decay accelerates during early postnatal development in coincidence with synaptic circuit maturation (Flint et al. 1997). Therefore, we tested whether the duration of pharmacologically isolated NMDA EPSCs in layer 3 pyramidal neurons differs between age groups. Figure 3A shows examples of NMDA EPSCs recorded from pyramidal cells in slices of the 3- and 15-month-old age groups, with exponential function fits superimposed on the current traces. We found that the time constant of exponential decay of the NMDA EPSCs decreased significantly between 3 and 15 months of age (Fig. 3B). After 15 months, the time constant did not change further (Fig. 3B). These results are consistent with a maturation of NMDA receptor function at glutamatergic inputs onto layer 3 pyramidal cells, between 3 and 15 months of age.
The decrease in the duration of the NMDA EPSCs observed in our studies (Fig. 3B) may contribute to the increase in the AMPA/NMDA ratio when the NMDA contribution to this ratio is measured at late phases in the decay of EPSCs recorded at positive holding potentials (Fig. 2B). If so, then AMPA/NMDA ratio and NMDA EPSC duration should be inversely correlated when measured from the same neuron. To test this possibility, we determined whether the AMPA/NMDA ratio and the NMDA EPSC duration were correlated in a subsample of neurons (n = 29) in which both parameters were measured sequentially from the same cell. As shown in Figure 3C, there was no significant correlation between AMPA/NMDA ratio and NMDA EPSC duration (Pearson correlation coefficient r = −0.11, P = 0.57), showing that the AMPA/NMDA ratio did not show an association with the NMDA EPSC duration. To further test the idea that the AMPA/NMDA ratio increases independently of the decrease in NMDA EPSC duration, in a subsample of neurons (3 months, n = 11; 15 months n = 12; 42 months, n = 5; 84 months, n = 9), we determined the AMPA/NMDA ratio by recording from each neuron isolated AMPA and NMDA EPSCs and by estimating the area under the AMPA and NMDA EPSC waveforms separately. Measurement of the NMDA EPSC area provides an estimation of the total NMDA current contribution to the synaptic charge transfer, including the contribution of both peak current and decay, independent of differences in decay time between age groups. The mean AMPA/NMDA ratio calculated based on EPSC areas increased by about 81% between 3 and 15 months and remained higher than 3 months in the older age groups, although the differences between group means did not reach statistical significance (3 months: 0.22 ± 0.16, n = 11; 15 months: 0.40 ± 0.24, n = 12; 42 months: 0.30 ± 0.13, n = 5; 84 months: 0.37 ± 0.23, n = 9; F3,28 = 2.50, P = 0.08, Single-factor ANOVA). The increase in the AMPA/NMDA area ratio observed between 3 and 15 months was similar to the statistically significant increase observed for the ratio of currents (about 60% increase between 3 and 15 months, Fig. 2B). Moreover, the AMPA/NMDA area ratio in the 3-month-old group was significantly different from that of the other age groups after they were pooled (P = 0.03). Together, the results shown in Figure 3 suggest that in addition to an acceleration of the NMDA EPSC decay there is an actual increase in the relative contribution of AMPA receptors between 3 and 15 months of age.
Multiple factors may contribute to the acceleration of the NMDA EPSC decay during development. For instance, developmentally regulated modifications in synaptic structure (Cathala et al. 2005) may shorten the duration of the glutamate concentration transient in the synaptic cleft by changing the kinetics of glutamate release or the speed of transmitter clearance (Diamond 2005; Koike-Tani et al. 2005; Takahashi 2005). Changes in the cleft glutamate concentration transient may not alter synaptic NMDA receptor activation (Lester et al. 1990). However, NMDA EPSCs are mediated in part by extrasynaptic NMDA receptors (Lozovaya et al. 2004; Thomas et al. 2006), the activation of which may depend on the duration of the glutamate concentration transient (Lozovaya et al. 1999; Arnth-Jensen et al. 2002). If the acceleration of NMDA EPSC decay is in part due to changes in the time course of the synaptic glutamate concentration transient, then the decay of AMPA EPSCs should also be affected (Cathala et al. 2005; Koike-Tani et al. 2005). To test this possibility, we examined the decay kinetics of AMPA EPSCs recorded from monkey DLPFC layer 3 pyramidal neurons (Fig. 4A). We found that the decay time constant of AMPA EPSCs did not differ significantly between age groups (Fig. 4B) when assessed in neurons that displayed significant acceleration of NMDA EPSC decay between 3 and 15 months. These results indicate that the acceleration of EPSC decay is specific for NMDA EPSCs.
An important factor regulating NMDA EPSC duration during development in rat hippocampus and neocortex is a switch in the subunit composition of the NMDA receptor complex. Early in development, NMDA receptor subunit B (NR2B) subunits are prevalent and determine a longer NMDA EPSC duration. Conversely, NR2A subunits are more abundant in the mature cortex (Carmignoto and Vicini 1992; Erisir and Harris 2003; Liu et al. 2004) and produce shorter NMDA EPSCs (Monyer et al. 1994; Sheng et al. 1994; Flint et al. 1997; Fu et al. 2005). To test whether the decrease in NMDA EPSC duration with age observed here is associated with a decrease in the contribution of NR2B subunits, we determined the effects of ifenprodil, a selective antagonist of NR2B subunit–containing NMDA receptors (Dingledine et al. 1999). Application of ifenprodil significantly reduced the amplitude of NMDA EPSCs recorded from neurons of 3-, 15-, or 42-month-old animals (Fig. 5A). However, the depression of NMDA EPSC amplitude by ifenprodil was significantly strong in neurons from 3-month-old animals (Fig. 5B). In addition, ifenprodil decreased the duration of NMDA EPSCs, but this effect was statistically significant only in neurons from 3-month-old animals (Fig. 5C). Our findings thus suggest that the decrease in NMDA EPSC duration during postnatal development of monkey DLPFC is at least in part due to a decreased contribution of NR2B receptor subunits.
EPSCs Display Immature Presynaptic Properties at 3 Months of Age
Our results suggest that excitatory inputs onto neurons from 3-month-old animals have a significantly larger proportion of synapses with relatively low AMPA receptor content (Figs. 2 and 3). Immature synapses with low AMPA receptor content may have low failure rate (Montgomery et al. 2001) and higher Pr than AMPA receptor-rich synapses (Yanagisawa et al. 2004). Maturation of presynaptic properties is usually associated with a decrease in the Pr (Bolshakov and Siegelbaum 1995; Kumar and Huguenard 2001; Wasling et al. 2004; Zhang 2004). Thus, we determined if the maturation of postsynaptic properties observed in this study between 3 and 15 months of age was paralleled by changes in Pr. To compare the Pr between age groups, we determined the rate of block of NMDA EPSCs by MK801, an irreversible open-channel blocker (Huettner and Bean 1988) that blocks NMDA EPSCs with faster kinetics the higher Pr is (Rosenmund et al. 1993). As in previous studies, the amplitude of NMDA EPSCs recorded from layer 3 pyramidal cells in monkey DLPFC was reduced progressively by stimulation in the continuous presence of 25 μM MK801 (Fig. 6A). Consistent with the presence of synaptic subpopulations with heterogeneous Pr, in each age group double-exponential decay functions (Fig. 6B) fit the data better than single-exponential decay functions. At later stages of the experimental protocol, the time course of NMDA EPSC block largely overlapped between age groups (Fig. 6B). Conversely, the initial rate of NMDA EPSC block was faster in neurons from 3-month-old animals (Fig. 6B), as revealed by an analysis of the parameters of exponential curve fitting. Tau F, the time constant of faster exponential decay, was significantly shorter in neurons from 3-month-old animals than in neurons from the other age groups (Fig. 6C). Comparisons of the remaining fitting parameters showed no statistically significant differences (Table 1). These data suggest that neurons from 3-month-old animals have a subpopulation of synapses in which Pr is higher than in synapses of any of the other age groups. Therefore, the analysis of the rate of NMDA EPSC blockade by MK801 suggest that the contribution of synaptic inputs with immature presynaptic properties decreases significantly prior to adolescence, sometime between 3 and 15 months of age.
|3 months (n = 19)||15 months (n = 17)||42 months (n = 15)||84 months (n = 10)||F||P value|
|FB||0.45 ± 0.05||0.49 ± 0.07||0.37 ± 0.04||0.52 ± 0.14||0.450||0.717|
|tau F (s)||16.4 ± 4.1||38.9 ± 11.1||37.6 ± 8.8||56.9 ± 19.8||3.838||<0.01|
|SB||0.56 ± 0.04||0.48 ± 0.07||0.60 ± 0.05||0.46 ± 0.15||0.706||0.550|
|tau S (s)||463 ± 81||1386 ± 1340||642 ± 158||609 ± 478||1.693||0.173|
|3 months (n = 19)||15 months (n = 17)||42 months (n = 15)||84 months (n = 10)||F||P value|
|FB||0.45 ± 0.05||0.49 ± 0.07||0.37 ± 0.04||0.52 ± 0.14||0.450||0.717|
|tau F (s)||16.4 ± 4.1||38.9 ± 11.1||37.6 ± 8.8||56.9 ± 19.8||3.838||<0.01|
|SB||0.56 ± 0.04||0.48 ± 0.07||0.60 ± 0.05||0.46 ± 0.15||0.706||0.550|
|tau S (s)||463 ± 81||1386 ± 1340||642 ± 158||609 ± 478||1.693||0.173|
Note: For each age group, the time course of NMDA EPSC blockade by MK801 was fit by a double-exponential decay function of the form: E(t) = FB × exp(−t/tau F) + SB × exp(−t/tau S) (Fig. 3C). In this equation, E is the amplitude of the NMDA EPSC; t is time after resuming stimulation in the presence of MK801 25 μM; FB and SB represent the fractions of the initial NMDA EPSC that are blocked with faster and slower time course, respectively, by MK801; tau F and tau S are the time constants of faster and slower exponential decay, respectively. Data are shown as mean ± standard error. Statistical significance of the differences in each parameter between age groups was determined using an F test and differences were considered significant if P < 0.05. Pairwise comparisons showed that tau F at 3 months was significantly different from tau F at 15 months (P = 0.034), from tau F at 42 months (P = 0.049) and from tau F at 84 months (P = 0.008). In contrast, no significant differences were found between the tau F values of age groups older than 3 months.
If the fraction of NMDA channels blocked by MK801 differs between EPSCs recorded from neurons of the different age groups, then the rate of NMDA EPSC blockade may be dependent not only on the Pr but also on the efficacy of MK801 to block the NMDA channels once glutamate is released (Hessler et al. 1993; Rosenmund et al. 1993). Because we employed a saturating concentration of MK801 (25 μM), age-related differences in binding affinity are unlikely to influence the fraction of NMDA channels blocked by MK801. To rule out the possibility that a different MK801-binding affinity or other factors may determine a different fractional block of NMDA channels, we determined the effects of MK801 on the NMDA EPSC decay. Acceleration of the NMDA EPSC decay by MK801 is an indicator of the fraction of channels blocked once glutamate is released (Hessler et al. 1993; Rosenmund et al. 1993) because MK801 blocks the NMDA channels not only at the EPSC peak but also during deactivation (Chen et al. 1999). To determine if the efficacy of NMDA channel block differed between age groups, we tested the effect of MK801 on the NMDA EPSC decay time constant in neurons from 3-, 15-, and 42-month-old animals. We found that MK801 (25 μM) significantly accelerated the decay time of NMDA EPSCs. For each cell, the decay time was computed for the average of the last 5 NMDA EPSCs recorded prior to MK801 application and also for the first 5 NMDA EPSCs recorded after resuming stimulation in the presence of MK801 (3 months, control decay time: 84 ± 30 ms, MK801 decay time: 47 ± 24 ms, n = 10 cells; 15 months, control decay time: 49 ± 9 ms, MK801 decay time: 30 ± 15 ms, n = 10 cells; 42 months, control decay time: 63 ± 12 ms, MK801 decay time: 48 ± 30 ms, n = 9 cells). Two-factor ANOVA indicated a significant effect of MK801 (F = 18.6, P = 0.00007), a significant effect of age (F = 7.27, P = 0.00165), but an absence of significant interaction between MK801 and age effects (F = 1.03, P = 0.36171). These results indicate that the efficacy of MK801 for blocking NMDA channels did not differ substantially between age groups, thus suggesting that differences in Pr are the most likely determinant of the faster kinetics of NMDA EPSC block in neurons from 3-month-old animals.
We determined the functional properties of excitatory synapses onto layer 3 pyramidal neurons during postnatal development of monkey DLPFC. We found that between 3 and 15 months of age, when synaptic density is constant, the contribution of synapses with immature functional properties appears to decrease significantly, as suggested by the following findings: 1) the relative contribution of AMPA receptors to the EPSCs increased; 2) the speed of NMDA EPSC decay increased; 3) the NMDA EPSCs became less sensitive to the NR2B subunit–selective antagonist ifenprodil; and 4) the Pr, as assessed by the rate of NMDA EPSC block by MK801, decreased. In contrast, the properties of EPSCs at 15 months were indistinguishable from those of EPSCs recorded from adult neurons at 42 or 84 months of age. Thus, our data suggest that functional maturation of excitatory synaptic inputs onto layer 3 pyramidal cells of monkey DLPFC occurs prior to the adolescence-related reduction in excitatory synaptic input described in previous studies.
Changes in Synaptic Function during Postnatal Development through Adolescence
An increase in the AMPA/NMDA ratio such as that found here is commonly associated with postsynaptic maturation (Durand et al. 1996; Wu et al. 1996; Isaac et al. 1997; Zhu and Malinow 2002; Franks and Isaacson 2005). Although consistent with an increase in the AMPA receptor contribution, the AMPA/NMDA ratio may increase via a decrease in the NMDA contribution. In the developing frog optic tectum (Wu et al. 1996) and during activity-dependent maturation of thalamocortical synapses (Isaac et al. 1997), the AMPA/NMDA ratio increases via an increased AMPA but constant NMDA component. In contrast, in rat olfactory cortex the AMPA/NMDA ratio increases, in part by an increase in AMPA contribution but mostly from NMDA downregulation (Franks and Isaacson 2005). The relative contribution of AMPA increase versus NMDA downregulation to the age-dependent increase in ratio observed in this study remains to be determined.
In the neonatal rat hippocampus, nascent synapses are rapidly silenced by removal of AMPA receptors from the postsynaptic membrane resulting from presynaptic action potential firing (Groc et al. 2006). We did not directly address the contribution of activity-dependent silencing of AMPA-containing nascent synapses to the lower AMPA/NMDA ratio observed in neurons from 3-month-old animals. However, the AMPA EPSC amplitude was typically stable during stimulation at 0.1 Hz and showed no evidence of silencing mechanisms during our recordings. Changes in the AMPA/NMDA ratio may be also due to an increase in the Pr at presynaptically silent synapses, rather than to postsynaptic mechanisms (Voronin and Cherubini 2004). Our data, however, suggest that the Pr decreases between 3 and 15 months of age. We therefore conclude that the differences in AMPA receptor contribution reported here are more likely due to postsynaptic mechanisms.
We found that EPSCs recorded from neurons of 3-month-old animals had both lower AMPA/NMDA EPSC ratios and longer NMDA EPSC decay time due to a more prominent contribution of NR2B subunits. These results are consistent with recent data indicating that immature synapses with low AMPA receptor content contain only NR2B-subtype NMDA receptors and that activity-dependent maturation produces synaptic insertion of AMPA receptors and of NMDA receptors containing ifenprodil-insensitive NR2A subunits (Nakayama et al. 2005). Expression of NR2B-containing NMDA receptors favors the induction of synaptic plasticity bidirectionally (i.e., either potentiation or depression), via specific properties of the NR2B-mediated intracellular Ca2+ transients or by the efficient association of NR2B-containing receptors with Ca2+/Calmodulin protein kinase II (Barria and Malinow 2005). Thus, the apparent decrease in the contribution of NR2B subunits with maturation would imply a relative decrease in the plasticity of synapses in the cortex of adult primates. Nevertheless, our findings are consistent with in situ mRNA hybridization studies indicating that NR2B subunits expression is still significant in the adult primate neocortex (Scherzer et al. 1998).
The peak open-channel probability for NR2B subunit–containing NMDA receptors is 5 times lower than that of NR2A-containing receptors (Chen et al. 1999). Thus, MK801 could be less effective in inhibiting NMDA EPSCs in neurons from 3-month-old animals. A less efficient block of NR2B-mediated responses by MK801 would lead to slower EPSC block independent of the Pr, thus leading to underestimation of the differences in Pr between the 3-month-old and other age groups. However, previous studies showed that MK801 similarly inhibits the amplitude of NR2B- and NR2A-mediated currents (Chen et al. 1999), suggesting that the on-rate of MK801 block does not differ between NMDA receptor subtypes. Consistent with this idea are previous data showing that blockade of NR2B-containing NMDA receptors with ifenprodil does not affect the rate of NMDA EPSC amplitude block by MK801 (Yanagisawa et al. 2004). Furthermore, we found that MK801 similarly accelerated the decay of NMDA EPSCs recorded from layer 3 pyramidal neurons of 3-, 15-, and 42-month-old animals (see results). Our data suggest that the efficacy of NMDA channel block by MK801 does not differ between age groups, supporting the conclusion that the faster rate of NMDA EPSC block in neurons from 3-month-old animals is due to synapses with higher probability of glutamate release.
Implications for Models of Synapse Elimination during Adolescence
During early brain development, synapses are stabilized via activity-dependent strengthening, leading to elimination of immature synapses that remain weak (Le Be and Markram 2006). In mouse neocortex, a significant proportion of morphologically immature synapses are eliminated during adolescence (Zuo, Lin, et al. 2005; Zuo, Yang, et al. 2005), suggesting that, in rodents, similar mechanisms could underlie synapse elimination during early development and during adolescence. Adolescence represents approximately a similar proportion of the total lifespan in mammals (Clancy et al. 2001); however, lifespan is substantially longer in primates than in rodents. Therefore, if similar mechanisms underlie synapse elimination during early development and adolescence in primates, then such mechanisms must remain active for an exceptionally prolonged time. For example, in macaque monkeys adolescence begins around 2–3 years of age (Lewis 1997). Our data seem to indicate that most synapses are mature prior to adolescence and, therefore, that the mechanisms of synapse elimination differ between early development and adolescence. It is possible, however, that synapses eliminated during adolescence are immature when considering molecular or ultrastructural characteristics that are not reflected in the functional properties measured in this study.
Our experiments cannot directly determine if the changes in functional properties observed between 3 and 15 months of age are due to maturation of persistent synapses or synapse replacement. Although no changes in synapse density occur in layer 3 of the monkey DLPFC between 3 and 15 months of age (Bourgeois et al. 1994; Anderson et al. 1995), functionally mature synapses could replace immature ones maintaining total synaptic density constant. During periods of constant spine density in adulthood, most spines are long lasting and only a small fraction is eliminated or added from pyramidal cell dendrites in mouse neocortex (Grutzendler et al. 2002; Trachtenberg et al. 2002; Holtmaat et al. 2005; Zuo, Lin, et al. 2005; Zuo, Yang, et al. 2005). However, newly added spines eventually form synapses with axonal boutons, demonstrating that synaptogenesis occurs in the neocortex of adult mice (Holtmaat et al. 2006; Knott et al. 2006). A recent study suggests that synaptogenesis may be an actively ongoing process throughout life also in the primate neocortex (Stettler et al. 2006). Therefore, a higher rate of synapse elimination than synaptogenesis may lead to the decrease in synapse number during primate adolescence. Conversely, those rates would be balanced in adult primate neocortex, when there is no net change in synapse density, although the rate of synapse turnover is surprisingly high, about 7% per week (Stettler et al. 2006).
Our experiments revealed that EPSCs had relatively immature properties in neurons from 3-month-old animals, an age when synaptogenesis still proceeds at a fast rate, and many newly formed synapses are being added to layer 3 pyramidal cell dendrites (Rakic et al. 1986; Bourgeois et al. 1994; Anderson et al. 1995). Collectively, our findings and those of previous anatomical studies suggest that around this age in primates, functional maturation of single synapses and synaptic circuits is coincident and perhaps causally related. For instance, formation and plasticity of ocular dominance columns, thought to proceed via synapse stabilization and elimination (Katz and Crowley 2002), in monkey visual cortex occurs during the first 12 weeks after birth (LeVay et al. 1980; Horton and Hocking 1997). This time window overlaps with the period of active synaptogenesis in layer 4 of monkey visual cortex (Bourgeois and Rakic 1993). Our data also suggest that global maturational changes in excitatory synaptic function happen before adolescence and that remodeling of excitatory circuits during adolescence happens via mechanisms that seem to be independent of significant maturational changes in synaptic strength.
An intriguing possibility is that refinement of excitatory circuits during adolescence results from dynamic excitation–inhibition interactions. For instance, administration of benzodiazepine modulators that potentiate GABAA receptor–mediated transmission significantly alters the remodeling of excitatory connections during critical periods of cortical development (Hensch 2005). In contrast to excitatory synapses, no significant changes in the density of inhibitory synapses are observed in the neocortex during adolescence (Rakic et al. 1986; De Felipe et al. 1997). However, the efficacy of GABAergic transmission may nevertheless change during adolescence because during this developmental period substantial changes occur in the levels of expression of GABAA receptors, GABA transporters and parvalbumin, an interneuron-specific calcium-binding protein (Lewis et al. 2004). Whether or not GABA-mediated transmission can regulate excitatory synaptic pruning during adolescence remains to be investigated.
Our results are relevant to current neurodevelopmental models of schizophrenia based in the observations that the expression of mRNAs for proteins that control the efficacy of synaptic transmission is altered in the illness (Mirnics et al. 2001). According to this model, the exuberant number of cortical synapses present prior to adolescence compensates for this synaptic dysfunction, and the pruning of these synapses during adolescence leads to the expression of the clinical consequences of the synaptic dysfunction. Our findings indicate that synaptic properties do not differ pre- and postpruning, at least in the normal monkey DLPFC, suggesting that prior to adolescence, functionally mature synapses might provide compensation via an excess in number.
Young Investigator Award from NARSAD to GG-B; National Institute of Mental Health (MH45156).
We thank Ms Olga Krimer and Ms Mary Brady for their help with reconstruction of dendritic trees and spines. Conflict of Interest: None declared.