PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
J Neurosci. Author manuscript; available in PMC Oct 24, 2013.
Published in final edited form as:
PMCID: PMC3684185
NIHMSID: NIHMS471950
Homeostatic Responses Fail to Correct Defective Amygdala Inhibitory Circuit Maturation in Fragile X Syndrome
Rebecca L. Vislay,1 Brandon S. Martin,1 Jose Luis Olmos-Serrano,1,2 Sebila Kratovac,1 David L. Nelson,3 Joshua G. Corbin,1* and Molly M. Huntsman1,4*
1Center for Neuroscience Research, Children's National Medical Center, 111 Michigan Ave NW, 6th Floor Main Building, Washington, DC 20010 USA
2Department of Anatomy and Neurobiology, Boston University School of Medicine, 72 East Concord St, L1004, Boston, MA, 02118 USA
3Department of Molecular and Human Genetics, Baylor College of Medicine, 1 Baylor Plaza, BCM225, Houston, TX, 77030 USA
4Department of Pharmaceutical Sciences, Skaggs School of Pharmacy, Department of Pediatrics, School of Medicine, University of Colorado | Anschutz Medical Campus 12850 E. Montview Blvd., Aurora, CO 80045 USA
* Co-Corresponding Authors: Molly M. Huntsman, PhD Departments of Pharmaceutical Sciences and Pediatrics University of Colorado | Anschutz Medical Campus 12850 E. Montview Blvd, V20-3121 Aurora, CO 80045 Phone: 303-724-7456 ; Molly.Huntsman/at/UCDenver.edu Joshua G. Corbin, PhD Center for Neuroscience Research Children's National Medical Center 111 Michigan Ave, NW Room M7639 Washington, DC 20010 USA Phone: 202-476-6281 ; JCorbin/at/cnmcresearch.org
Fragile X Syndrome (FXS) is a debilitating neurodevelopmental disorder thought to arise from disrupted synaptic communication in several key brain regions including the amygdala - a central processing center for information with emotional and social relevance. Recent studies reveal defects in both excitatory and inhibitory neurotransmission in mature amygdala circuits in Fmr1-/y mutants, the animal model of FXS. However, whether these defects are the result of altered synaptic development or simply faulty mature circuits remains unknown. Using a combination of electrophysiological and genetic approaches, we show the development of both pre- and postsynaptic components of inhibitory neurotransmission in the FXS amygdala is dynamically altered during critical stages of neural circuit formation. Surprisingly, we observe that there is a homeostatic correction of defective inhibition, which, despite transiently restoring inhibitory synaptic efficacy to levels at or beyond those of control, ultimately fails to be maintained. Using inhibitory interneuron-specific conditional knockout and rescue mice, we further reveal that Fragile X Mental Retardation Protein (FMRP) function in amygdala inhibitory microcircuits can be segregated into distinct pre- and postsynaptic components. Collectively, these studies reveal a previously unrecognized complexity of disrupted neuronal development in FXS and therefore have direct implications for establishing novel temporal and region-specific targeted therapies to ameliorate core amygdala-based behavioral symptoms.
Defects underlying neurodevelopmental disorders, including FXS, are widely believed to lie at the level of the synapse (Zoghbi, 2003;Ebert and Greenberg, 2013). In FXS, these profound changes include alterations in both excitatory and inhibitory neurotransmission across multiple brain regions, including the amygdala (Huber et al., 2002;Bear et al., 2004;Olmos-Serrano et al., 2010). Excitatory synaptic transmission in FXS is strongly altered by misregulated metabotropic glutamate receptor (mGluR) signaling, a phenomenon observed widely throughout the brain (Wilson and Cox, 2007;Desai et al., 2006;Pilpel et al., 2009;Zhang et al., 2009;Suvrathan et al., 2010). In addition to these defects, a growing body of evidence has revealed complementary and profound defects in inhibitory neurotransmission (Gibson et al., 2008;Centonze et al., 2008). In the amygdala, our previous work revealed significant decreases in GABA production and the numbers of inhibitory synapses (Olmos-Serrano et al., 2010).
Defects in neuronal communication in the FXS brain most likely stem from altered processes of synaptogenesis and circuit formation. This is supported by studies of Fmr1-/y knockout mice. For example, in the cerebral cortex the morphological immaturity that characterizes adult excitatory neurons suggests alterations in synaptic maturation and pruning (Comery et al., 1997). Indeed, dynamic analysis of spine formation in mutant layer V barrel cortex pyramidal neurons revealed similar abnormalities at critical periods of synaptogenesis (Nimchinsky et al., 2001). Consistent with these developmental defects, critical period barrel cortex plasticity is altered in Fmr1-/y mutants and coincides with a persistence in the number of N-methyl-D-aspartate (NMDA) receptor-dominated silent synapses, an indicator that this circuit may fail to properly mature (Harlow et al., 2010). Taken together, these studies suggest that FMRP plays a crucial role in processes required for the maturation of excitatory neurons.
However, the processes that lead to the establishment of synaptic defects with regard to inhibitory neurotransmission and in relevant regions such as the amygdala, remains unexplored. In this study, we directly address this question by examining the developmental progression of GABAergic inhibitory neurotransmission in the basolateral nucleus of the amygdala (BLA) during the critical window of synaptogenesis. We find that the development of inhibitory neurotransmission in the Fmr1-/y mutant BLA is dramatically altered in a complex manner. Underlying these changes are deficiencies in both key pre- and postsynaptic developmentally-regulated processes, including production of synaptic GABA and GABAA receptor (GABAAR) maturation. Most strikingly, we find that although GABA neurotransmission is initially decreased during early periods of synaptogenesis, at later times there is a transient increase in specific components of inhibitory synaptic function. Ultimately, however, this temporally-restricted upregulation of inhibitory synaptic mechanisms fails to be maintained and inhibitory neurotransmission returns to deficient levels. Moreover, these defects can be genetically segregated into pre- and postsynaptic components based on whether or not FMRP expression is either specifically conditionally knocked out or rescued in inhibitory neurons. Together, these data reveal novel mechanisms of altered trajectories of abnormal development of amygdala inhibitory networks in FXS and also provide insight into the most efficacious therapeutic strategies for treating amygdala-based symptoms at different pediatric stages.
Animal Use
All experiments were performed under protocols approved by the IACUC animal usage committee at Children's National Medical Center. Control Fmr1 and Fmr1-/y knockout mice on the congenic FVB background were obtained from Jackson Laboratories (Stock #4828 and #4624). Conditional knockout (Fmr1cKO) and conditional rescue (Fmr1cON) animals were maintained on the C57BL/6 background and generated by crossing male Dlx5/6Cre mice to female mice containing either a floxed portion of the Fmr1 gene to generate conditional knockouts or a floxed Neo cassette interrupting the Fmr1 gene for conditional rescues (Mientjes et al., 2006). For conditional analyses, mice were were genotyped (Transnetyx Inc.) and only males positive for Dlx5/6Cre and the conditional mutant alleles were selected for experiments. Male littermates positive only for Dlx5/6Cre and that did not contain the conditional mutant allele were used to rule out any effect of Cre expression alone.
Slice preparation
Male mice were deeply anesthetized with carbon dioxide (CO2). Brains were removed and placed in oxygenated ice-cold sucrose slicing solution, (in mM: 234 sucrose, 11 glucose, 26 Na2HCO4*H2O, 2.5 KCl, 1.25 NaH2PO4*H2O, 10 MgSO4*7H2O, and 0.5 CaCl2*H2O) for 1-2 minutes. Coronal slices (300μm) containing the BLA were then incubated for 45-60 minutes in pre-warmed (35°C) artificial cerebrospinal fluid (aCSF) containing, in mM: 126 NaCl, 26 NaHCO3, 10 glucose, 2.5 KCl, 1.25 NaH2PO4*H2O, 2 MgCl2*7H2O, and 2 CaCl2*H2O. Slices were allowed to equilibrate to room temperature for approximately 25 minutes following incubation.
Electrophysiology
Slices were visualized using a fixed-stage upright microscope (Olympus BX61WI) outfitted with 4X air and 60X water immersion objectives. Cells were identified under bright field illumination with the aid of a CCD camera (Hamamatsu) and Slidebook imaging software package (Intelligent Imaging Innovations). All recordings were performed at room temperature (21-22°C) and slices were continuously perfused with oxygenated aCSF. Similar to our previous study, the reason for room temperature experiments is for identification and proper decay analysis of isolated IPSCs during periods of high frequency and in the presence of agonists that elongate IPSCs (Olmos-Serrano et al., 2010). Whole cell patch clamp recordings were made using pipettes with resistances in the range of 2.0-3.5 MΩ. Recordings were made using a Multiclamp 700A amplifier (Molecular Devices). Data was collected at a sampling frequency of 10kHz, digitized, and acquired using custom-written programs in pClamp10 (Molecular Devices). Principal excitatory neurons in the BLA were initially identified visually by their large cell bodies. When not in the presence of TTX, cell identification was confirmed by injecting short, 600ms hyperpolarizing and depolarizing pulses while in current clamp mode as previously described (Olmos-Serrano et al., 2010). sIPSCs and mIPSCs were recorded using an intracellular solution containing, in mM: 70 K-gluconate, 70 KCl, 10 HEPES, 10 EGTA, 2 MgCl2, 4 Mg-ATP, and 0.3 Na-GTP, equilibrated to pH = 7.3 and 290 mOsm. Inhibitory currents from principal cells were voltage-clamped at -70mV and isolated by blocking ionotropic glutametergic synaptic transmission by inclusion of 20μM 6,7-dinitroquinoxaline-2,3-dione (DNQX) and 50μM DL-2-amino-5-phosphonopentanoic acid (APV; both obtained from Tocris Biosciences) to the continuously-circulating aCSF. The perfusion rate was approximately 2 mL/min. For recordings of mIPSCs, 1μM TTX (Alomone Labs) was added to the bath to block sodium action potentials. In some cases the frequency of mIPSCs were not lower than sIPSC frequency because the slices are typically quiescent and recordings were obtained at room temperature. Additional drugs used where indicated included (final concentrations): 200nM zolpidem, 100nM clonazepam, 1μM NO-711, and 100μM (1,2,5,6-tetrahydropiridin-4-yl) methylphosphonic acid (TPMPA), all acquired from Tocris Biosciences. All recordings were continuous with 5 minutes pre-drug, 5 minute drug wash-in and 5 minute wash-out. A wash-in time of 3 minutes was observed before “drug-IPSCs” were collected for analysis. For analyses of biophysical recordings, each distribution of values was first tested for normality using the Shapiro-Wilk test. Based on these results, a parametric two-tailed heteroscedastic t-test or a non-parametric Mann-Whitney test were used to determine the statistical significance between normally and non-normally distributed groups, respectively. The Bonferroni method was used to adjust the alpha values in these tests to correct for multiple comparisons of cells across different ages/conditions.
Data Analysis
Individual synaptic events were identified visually using prewritten routines in pClamp10 (Molecular Devices). Averaged sIPSCs or mIPSCs were fitted using a double exponential: f(t)=Afaste-t/τfast-Aslowe-t/τslow. Fitted traces were used to determine the weighted decay time constant: τd,w=[(Afastfast)+Aslowslow)]/(Afast+Aslow). Drug “enhancement” measures were calculated by taking the difference between the decay constant measured prior to drug application and that measured during drug application and are expressed as the percent difference.
For mIPSC quantal analysis (Fig. 5), amplitudes were binned into 5pA bins and plotted as histograms. The fitting function was determined using a least squares algorithm (Edwards et al., 1990). 50-60 iterations of the algorithm were run to identify the fitting with the minimum error function between the histogram peaks and fitting function (> 1.5%). The resulting function was deconvolved into separate underlying Gaussian functions. The number of Gaussian functions was determined for each cell and differences in the average number of Gaussians between mutant and control at each different age were assessed using two-sample Mann-Whitney test. Peak-scaled nonstationary noise analysis was conducted as previously described (De Koninck and Mody, 1994). Briefly, the average of all mIPSC events was calculated and normalized to the peak of each individual event. These normalized averages were subtracted from the individual event waveforms and the remaining current amplitudes binned into 5pA bins. The mean current (Im) and variance (σ2) calculated for each bin. The variance was plotted versus the mean current and the resulting distribution fitted using the parabolic equation: σ2=iIm-Im2/N, where i is the unitary current and N is the number of postsynaptic receptors. All fitting and statistical analyses were performed using custom-written routines in MATLAB (Mathworks).
Figure 5
Figure 5
mIPSC quantal analysis reveals smaller postsynaptic GABAAR patch sizes in Fmr1-/y inhibitory synapses. A, B. Representative mIPSC amplitude histograms for control (A) and Fmr1-/y (B) at P10, P14, P16, and P21. Dashed lines demarcate the least squares (more ...)
Immunohistochemistry
Mice were transcardially perfused with 4% paraformaldehyde (PFA). Brains were fixed overnight, and coronal sections of brains embedded in 4% Agar (Fisher Scientific, Pittsburg, PA) were cut using a vibratome (Leica). For immunohistochemistry, following antibodies were used: GAT1 (1:200; Abcam), and secondary antibodies for immunofluorescence (cy3 at 1:1000; Jackson Immunoresearch, PA)
Alterations in the development of inhibitory neurotransmisssion in the Fmr1-/y mutant amygdala
The biophysical properties of GABAAR mediated inhibitory neurotransmission follow a stereotypical progression of changes over the course of the first three postnatal weeks that have been observed across several brain regions, including the cerebellum (Tia et al., 1996;Vicini et al., 2001), the hippocampus (Hollrigel and Soltesz, 1997), and the thalamus (Huntsman and Huguenard, 2000). During this time, sIPSCs recorded in rodent brain slices decrease in amplitude, increase in frequency, and exhibit faster decay kinetics. To determine whether this developmental pattern is similar in the amygdala and to test whether it is altered in Fmr1-/y mutant animals, sIPSCs were recorded from excitatory principal neurons in the BLA using the whole-cell patch clamp technique at postnatal days (P) 10, 14, 16, and 21; four key time points during this critical postnatal period of development. The developmental trajectory of sIPSCs in the control BLA follows that which has been observed in other brain regions (Fig.1A; Table 1). sIPSCs from Fmr1-/y mutant neurons, however, displayed dramatic differences compared to control at a number of ages explored. At P10, which represents earlier stages of synaptogenesis, sIPSC amplitude was decreased (Fig. 1B), as is sIPSC frequency (Fig. 1C). However, these alterations did not remain constant as development progresses and sIPSC amplitude and frequency were surprisingly increased at P14 in Fmr1-/y mutants. In contrast, at P16, sIPSC amplitude returned to that observed in control but frequency remained high. By P21, both sIPSC amplitude and frequency returned to below-control levels. The decay constant did not differ significantly from control at P14 and P16, yet it was increased at P10 and again at P21 (Fig. 1D).
Figure 1
Figure 1
The development of inhibitory neurotransmission is dynamically altered in the Fmr1-/y mutant BLA. A. Recordings of spontaneous inhibitory postsynaptic currents (sIPSCs) recorded from BLA excitatory neurons. B-D. Histograms of pooled data for sIPSC biophysical (more ...)
Table 1
Table 1
Summary sIPSC/mIPSC Amplitude, Frequency, and Decay Values
These dramatic differences in sIPSCs at different time points could be the result of alterations in processes dependent on and/or independent of action potential firing. To examine this, miniature IPSCs (mIPSCs) were examined in the presence of tetrodotoxin (TTX, 1 μM), which blocks action potentials. mIPSCs in control neurons exhibit the stereotypical developmental progression of decreased amplitude, increased frequency, and faster decay kinetics (Fig. 1E-G, Table 1). Similar to sIPSCs, at P10 mIPSC amplitude and frequency were significantly lower in Fmr1-/y mutant neurons compared to control. At P14, however, these differences were no longer observed. Similarly, no change was observed in either of these parameters at P16. By P21, mIPSC amplitude and frequency returned to levels below those of control. Similar to our previous studies (Olmos-Serrano et al., 2010), at P21 the mIPSC decay constant did not differ between the genotypes. Additionally, we see no differences at the other three time points indicating that both receptor and non-receptor based alterations likely affect decay kinetics.
Collectively these analyses also revealed that defects in inhibitory neurotransmission are present at P10 but this is followed by a transient restoration of inhibitory synaptic activity at P14-P16. At these ages, action potential independent Fmr1-/y mutant mIPSC frequencies are indistinguishable from control, while action potential dependent sIPSCs frequencies are increased, suggesting that the transient increase in inhibitory efficacy is at least in part due to increased firing of action potentials. By P21, however, both the activity-dependent and activity-independent increases are no longer present and inhibitory synaptic transmission in Fmr1-/y mutants is again reduced.
Alterations in presynaptic mechanisms affecting the amount of synaptic GABA production and release could also contribute to the dramatic changes in inhibitory neurotransmission that we observe in Fmr1-/y mutants. Relative levels of synaptic GABA can be examined using the competitive GABAA receptor antagonist (1,2,5,6-tetrahydropyridin-4-yl)methylphosphinic acid (TPMPA). The efficacy of this antagonist is inversely proportional to the concentration of GABA at the synapse (Barberis et al., 2005). Thus, a stronger decrease in mIPSC amplitude in the presence of TPMPA signifies a decreased level of synaptic GABA. In the presence of 100μM TPMPA, a greater attenuation of mIPSC amplitude was observed in principal neurons from Fmr1-/y mutants at age P10 compared to control percent change in mIPSC amplitude (Fig. 2; Table 2). Interestingly, this difference was not observed at P14 or P16 where TPMPA reduced mIPSC amplitudes to the same degree in mutant and control. By P21, however, mIPSC amplitudes in Fmr1-/y cells were again more sensitive to TPMPA than control. Thus, at P10 and P21, there was a decrease in the concentration of GABA in the synapse in Fmr1-/y mutants, however, between P14 and P16 GABA levels transiently recovered to that of control.
Figure 2
Figure 2
Dynamic changes in the levels of synaptic GABA concentration at inhibitory synapses in the Fmr1-/y BLA. The ability of TPMPA to attenuate the amplitude of mIPSCs is inversely proportional to the concentration of GABA localized to the synapse. A. Averages (more ...)
Table 2
Table 2
Pharmacological Assessment of IPSCs utilizing TPMPA, NO-711, Zolpidem and Clonazepam
The expression of the major presynaptic GABA reuptake transporter, GAT1, is developmentally-regulated and increases as animals age (Swanwick et al., 2006). An increase in GAT1 function could lead to the lower levels of synaptic GABA we observed in Fmr1-/y mutants at P10 and P21. To assess any possible changes in GAT1-mediated GABA reuptake, Fmr1-/y mutant sIPSCs were recorded from control and mutant BLA principal neurons in the presence and absence of 1μM NO-711, a GABA reuptake blocker that acts specifically on GAT1 (Fig. 3A-C; Table 2). Blockage of GAT1 with NO-711 leads to an increase in the decay constant indicating a greater efficacy of NO-711 and thus a likely increase in GAT1 function. Consistent with previous findings that GAT1 expression is very low in younger animals (Avila et al., 2011), no significant increase in decay constant was observed with NO-711 application in either control or Fmr1-/y mutant sIPSCs recorded at P10 (Fig. 3A-C). At P21, however, significant increases in decay constant were observed in both control and mutant sIPSCs (Fig. 3B,C). This change, however, was significantly greater in Fmr1-/y mutants than in control, suggesting increased GAT1-mediated GABA reuptake in these neurons (Fig. 3C). In order to determine if increased GAT1 function at P21 is related to transporter protein levels, we compared GAT1 expression in control and Fmr1-/y mutant animals (Fig. 3D,E). Qualitative immunohistochemical detection of GAT1 protein indicated an increase in BLA GAT1 expression in P21 Fmr1-/y mutants. Immunohistochemical detection of GAT1 protein indicated an increase in BLA GAT1 expression in P21 Fmr1-/y mutants.
Figure 3
Figure 3
Developmental increase of GABA transporter function and expression. A,B., Averages of >150 sIPSC events normalized for amplitude before (black and gray traces for control and Fmr1-/y, respectively) and after (red traces) the application of 1μM (more ...)
In addition to these GABAAR-independent changes, alterations in the number and/or subtype specification of GABAARs could also contribute to changes in inhibitory neurotransmission. Toward this goal, we first examined the development of the postsynaptic GABAARs using subunit selective pharmacology. Previous studies have revealed that the subunit composition of postsynaptic GABAA receptors changes in several brain regions during the first two postnatal weeks (Tia et al., 1996;Huntsman and Huguenard, 2000;Paysan et al., 1994; Vicini et al., et al., 2001). Typically, the population of GABAARs transitions from one dominated by those containing α2/α3 subunits early in postnatal development to mostly α1 subunit containing receptors at more mature ages. A major determinant of decay kinetics of IPSCs depends primarily on GABAAR subunit composition and, since early α2/α3GABAA receptors have slower kinetics than mature α1GABAARs, IPSC decay constants normally decrease over development ((Rudolph et al., 1999; Verdoorn 1995; Hollrigel and Soltesz. 1997) and Figs. 1D,,G). The activity of these two different subtypes of GABAARs can be distinguished pharmacologically using the benzodiazepine site agonists zolpidem and clonazepam, which act to increase IPSC decay constants by selectively targeting α1 or α2/α3 GABAARs, respectively (Huntsman and Huguenard, 2006;Galanopoulou, 2008). We therefore examined the status of developmental GABAAR subtype specification in the Fmr1-/y mutant BLA by recording sIPSCs from principal excitatory neurons in the presence of either 100nM zolpidem or 200nM clonazepam across developmental time.
Consistent with the normal developmental increase in the prevalence of α1GABAARs observed in other brain regions, sIPSC sensitivity to zolpidem steadily increased across the four developmental time points in control cells (Fig. 4A, black symbols; Table 2). In Fmr1-/y principal neurons, however, this profile was severely disrupted, with zolpidem sensitivity remaining low throughout development (Fig. 4A, gray symbols). With clonazepam, as expected for control cells sensitivity decreased as development progresses (Fig. 4B, black symbols). This normal pattern of clonazepam sensitivity was also altered in Fmr1-/y mutant neurons (Fig. 4B, gray symbols) where sensitivity was initially lower than control at P10, increased to control levels at P14 and P16, and then remained higher than control at P21. These data indicate probable alterations in subunit density early in development (P10) and altered receptor subunit composition later (P21) in the FXS amygdala.
Figure 4
Figure 4
GABAA receptor subunit-selective pharmacology of sIPSCs is disrupted in the Fmr1-/y mutant BLA. A. Sensitivity to the α1GABAAR agonist zolpidem is reduced in Fmr1-/y mutants at postnatal ages. sIPSCs were recorded in the presence of 200nM zolpidem, (more ...)
Additional information regarding the density and efficacy of postsynaptic GABAAR populations can be obtained from examination of mIPSC amplitude distributions. Variability in mIPSC amplitudes is directly correlated with the number of postsynaptic GABAARs (Nusser et al., 1997). Furthermore, the shape of these distributions suggest that the number of GABAARs at different synapses (GABAAR patch size) may be quantally distributed (Edwards et al., 1990). As such, the overall structure of these distributions can best be described by the sum of underlying Gaussian functions where each underlying function corresponds to a different GABAAR patch size. To examine putative differences in the population of GABAAR patch sizes between control and Fmr1-/y, mIPSC amplitudes were first binned into 5pA subgroups (Fig. 5). These distributions were then best fit using a least squares algorithm to determine the function best describing them (Edwards et al., 1990). The number of underlying Gaussian functions indicates the number of different GABAAR patch sizes. Overall, mIPSC distributions from Fmr1-/y mutant cells were best fit by fewer Gaussian functions at all ages tested (Fig. 5). However, when examined across cells, the difference in the number of best-fit Gaussians between control and mutant mIPSC distributions was only significant at P10 and P21 (Fig. 5C). This suggests that, at these ages, there was significantly less variability in the size of postsynaptic GABAAR patch sizes in Fmr1-/y mutant cells with smaller patch sizes dominating. To more directly assess the average number of GABAARs per synapse in mutant versus control at different ages, peak-scaled nonstationary noise analysis was performed on the mIPSC amplitudes (De Koninck and Mody, 1994). This type of analysis provides accurate estimations of the average number of GABAARs per synapse. When performed using mIPSC amplitudes pooled from neurons within each age and genotype, this analysis suggested that there were significantly fewer GABAARs per synapse for Fmr1-/y as compared to control at P10 and P21, but not at P14 or P16 (Fig. 6). In combination with the finding that mutant neurons have a smaller variety of GABAARs numbers per synapse (Fig. 5), this result suggested that Fmr1-/y principal neurons have a preponderance of small postsynaptic GABAAR populations per synapse. Given that the number of postsynaptic GABAARs was directly correlated with overall synapse size (Nusser et al., 1997), these results suggested that at P10 and P21, inhibitory synapses onto principal excitatory neurons in the Fmr1-/y BLA are smaller and therefore contain less GABAARs than control neurons. Interestingly, the number of postsynaptic GABAARs per synapse did not significantly differ at P14 and P16 (Fig. 6D), possibly explaining, in part, the finding that inhibitory synaptic efficacy at these ages was comparable to control, at least in the absence of action potential firing.
Figure 6
Figure 6
Peak-scaled nonstationary noise analysis suggests significantly higher unit currents and lower numbers of GABAARs in Fmr1-/y at P10 and P21 but not at P14 or P16. A, B. Current variance (σ2)-mean current (Im) plots for control (A) and Fmr1-/y (more ...)
Taken together, these results show that both pre- and postsynaptic mechanisms involved in inhibitory neurotransmission including synaptic GABA production, GABA reuptake mechanisms, GABAAR receptor subtype specification and numbers are significantly and dynamically disrupted in the Fmr1-/y BLA. In addition, between P14 and P16 numerous components of inhibitory neurotransmission are upregulated including action potential firing, synaptic GABA concentration and receptor number as well as changes in the predominance of specific GABAAR subtypes.
Inhibitory cell-autonomous and non-cell-autonomous mechanisms contribute to inhibitory neurotransmission deficits
The above results reveal a complex pattern of altered inhibitory synaptic development in which an initial defect in inhibition is transiently corrected by the upregulation of specific pre- and post-synaptic mechanisms critical to GABAergic synaptic function. However, these changes in inhibition may or may not critically depend on the cell-autonomous function of FMRP in inhibitory neurons. To investigate the contribution of the lack of FMRP in inhibitory neurons compared to the lack of FMRP in all other cells except inhibitory neurons on inhibitory synaptic defects, we utilized previously generated mice in which Fmr1 expression can be conditionally knocked out (Fmr1cKO) or restored (Fmr1cON) (Mientjes et al., 2006). To conditionally knockout or restore FMRP specifically in inhibitory interneurons, Fmr1cKO and Fmr1cON mice were separately crossed to Dlx5/6Cre mice which drives Cre recombinase expression solely in forebrain inhibitory neurons starting at mid-neurogenesis (Stenman et al., 2003). In this analyses, we compared sIPSCs recorded from BLA principal neurons across five groups of mice: wild type control, Fmr1-/y (full KO), Dlx5/6cre (cre control), Dlx5/6cre;Fmr1cKO and Dlx5/6cre; Fmr1cON (Fig. 7 and Table 3) at P10, P14 and P21. For each age and genotype we measured sIPSC amplitude, frequency, decay constant and zolpidem enhancement of conditional genotypes as compared to Dlx5/6cre controls. Significant changes were observed in a number of criteria and varied across time and genotype. At P10 and P21, Dlx5/6cre;Fmr1cKO mice did not show significant changes in sIPSC amplitude, frequency, decay constant or zolpidem enhancement. At P14, however, Dlx5/6cre;Fmr1cKO mice exhibited an increase in frequency but no change in amplitude, decay constant or zolpidem enhancement. In contrast, Dlx5/6cre; Fmr1cON rescue animals displayed a greater number of significant changes in sIPSC characteristics at all ages. These included changes in frequency and decay constant at all ages (P10, P14 and P21) as well as changes in amplitude and zolpidem enhancement at P21.
Figure 7
Figure 7
Comparison of biophysical parameters for sIPSCs in conditional knockout and rescue animals to control and full KO animals. Histograms of pooled data of sIPSCs recorded from P10 (A), P14 (B) and P21 (C) time points. Amplitude, frequency and decay constant (more ...)
Table 3
Table 3
Development of sIPSC Characteristics in BLA Principal Neurons from Inhibitory Cell-Specific Conditional Fmr1 Knockout and Rescue Animals
Combined, the results of sIPSC recordings in conditional Fmr1 mutants show that both inhibitory cell- autonomous and non-autonomous mechanisms are disrupted in the full Fmr1-/y KO. Conditional knockout of FMRP solely in interneurons (Dlx5/6Cre;Fmr1cKO) results in a normalization across ages of post synaptic responses as evidenced by normal amplitude, decay and zolpidem enhancement. The latter finding reveals that restoration of FMRP in excitatory neurons can rescue the postsynaptic α1GABAAR defect observed in the full Fmr1-/y KO. In contrast, conditional rescue of FMRP solely in interneurons (Dlx5/6cre; Fmr1cON) results in enhanced frequency, without postsynaptic rescue of decay constant or zolpidem enhancement. This suggests that restoration of FMRP in interneurons causes an enhancement of inhibitory neuronal efficacy above and beyond that of control. Interestingly, in both conditional knockout and conditional rescue genotypes, frequency is increased at P14. Taken together with P14 analyses of the full Fmr1-/y KO (Fig. 1), this suggests a transient increase either in the number of synapses, neurotransmitter release, and/or firing rate of inhibitory neurons, represents restorative mechanisms that can still occur whether FMRP is expressed solely in inhibitory neurons or excitatory neurons.
The most striking finding presented here is that the state of inhibitory neurotransmission varies dramatically at different points in the first few weeks of postnatal life in the FXS amygdala. Quite interestingly, initial inhibitory defects are transiently corrected during the postnatal period of increased synaptogenesis in the second postnatal week (Jacobson, 1991). The drastic differences in inhibitory synaptic function over time are correlated with different levels of specific pre- and post-synaptic defects, the most profound being alterations in the functional maturation of α1GABAARs, GABAAR numbers, and amount of synaptic GABA. These alterations act in concert to define the status of inhibitory synaptic efficacy overall, and transient increases in synaptic GABA and the upregulation of α2/3GABAARs appear to be part of homeostatic responses that temporarily restores inhibition to normal levels at P14-P16. This transient restoration importantly demonstrates that, at least during periods of development characterized by significant synaptogenesis, in the FXS amygdala the machinery necessary for normal inhibitory neurotransmission is at least partially intact, or can be compensated for. While the exact mechanism of this transient increase in inhibition remains unknown, our data suggest that this compensation occurs via a transient increase in interneuron firing. However, increased GABA release and/or overabundance of inhibitory connections may also play a role. Furthermore, Fmr1 conditional analyses demonstrate that amygdala inhibitory dysfunction in Fmr1-/y mutants is comprised of distinct pre- and post-synaptic components.
Homeostasis and the dynamic nature of synaptic defects
Normally, levels of synaptic input are globally regulated throughout a network by means of “synaptic scaling” (Turrigiano, 1999). Such homeostatic modifications act to maintain a specific level of synaptic input onto a given neuron in a manner dependent on the amount of activity in the entire network (Turrigiano et al., 1998;Desai et al., 2002;Kilman et al., 2002). Here, we observe a dramatic, developmentally-dependent transient increase in inhibitory synaptic input onto excitatory principal neurons in the FXS amygdala that is reminiscent of homeostatic changes observed when global network activity is increased in normal networks (Hartman et al., 2006;Gonzalez-Islas and Wenner, 2006;Peng et al., 2010). This compensatory upregulation of inhibition (reported as frequency and amplitude) appears to occur in response to the decreased overall inhibitory activity present earlier at P10.
Our data suggest that at least five critical mechanisms appear to contribute to this homeostasis (Fig. 8). First, the increase in the frequency of GABAergic synaptic inputs to principal neurons in the developing FXS BLA between P14-P16 could be the result of an increased number of functional inhibitory synapses. This is consistent with the finding that enhanced excitatory activity, as might be present at early postnatal stages in the FXS BLA, signals the upregulation of factors, such as Brain-Derived Neurotrophic Factor (BDNF), that have been shown to promote inhibitory synapse formation and stabilization (Peng et al., 2010). Second, the frequency of GABAergic inputs increases to levels beyond normal when action potentials are not blocked. We interpret this result as a homeostatic increase in inhibitory interneuron firing. Third, immature α2/3GABAARs, while deficient early in postnatal development in Fmr1-/y, rise to normal levels and remain higher than normal in mature circuits. It is possible that these immature α2/3GABAARs compensate for the absence of α1GABAARs. Fourth, the overall relative number of postsynaptic GABAARs also temporarily increases to normal levels as determined by quantal analysis. Fifth, we demonstrate that the concentration of GABA at the synapse is similarly temporarily restored during the same window of time. This suggests that presynaptic mechanisms governing GABA synthesis, packaging into vesicles, and/or vesicle release may also be enhanced homeostatically. Therefore, we suggest that both an increase in inhibitory interneuron activity (and/or synapse number) and the enhancement of basic mechanisms involved in inhibitory synaptic formation and function contribute to the transient developmental correction of inhibitory defects in the FXS amygdala. However, despite retaining the ability to achieve normal levels of inhibition in the absence of FMRP, this apparent homeostatic response is unable to maintain proper inhibitory function as the BLA network matures and ultimately returns to diminished levels at P21.
Figure 8
Figure 8
Summary of altered developmental inhibitory synaptic processes in FXS. The top set of panels depicts development across the four postnatal time points in the wildtype BLA and the bottom set shows the case of FXS. Our results suggest that at least five (more ...)
Inhibitory cell autonomous and non-cell autonomous mechanisms of reduced GABAergic efficacy
It has recently been shown that several FMRP target mRNAs are involved in the formation and function of key pre- and postsynaptic elements (Darnell et al., 2011). Consistent with this, we observe that several functional components of GABAergic synaptic transmission – both pre- and postsynaptic – are disrupted in the FXS BLA. These defects may be affected differentially based on whether or not FMRP is expressed solely in inhibitory neurons. To address this, we generated conditional Fmr1 animals in which FMRP expression is either specifically knocked out or restored in inhibitory neurons.
Our results reveal that GABAAR maturation appears to be entirely independent of FMRP expression in interneurons. When FMRP expression is knocked out in inhibitory cells only, sIPSC decay constants are normal and exhibit a normal sensitivity to zolpidem. Conversely, restoring FMRP in interneurons only leads to decay constants with values similar to those observed in the full Fmr1 knockout. These decay constants are also insensitive to zolpidem. The most likely explanation is that FMRP is critically required in postsynaptic BLA principal cells, but is not required in inhibitory neurons, for GABAAR maturation. This lack of FMRP in BLA excitatory neurons appears to primarily affect the process leading to the rise in α1-sensitive receptors that characterizes mature BLA inhibitory synapses.
The lack of FMRP expression in the full Fmr1 knockout affects processes underlying IPSC amplitude and frequency in more complex ways, however. While several factors influence IPSC amplitude, the amount of GABA released into the synapse and the number of postsynaptic GABAARs are believed to contribute the most (De Koninck and Mody, 1994). The results of experiments in the full Fmr1 knockout demonstrate that both of these are reduced in the FXS amygdala. Measurements from conditional Fmr1 knockout animals show that FMRP expression may be affecting the level of synaptic GABA and the number of GABAARs independent of whether or not it is absent either in inhibitory neurons or in excitatory principal neurons. This is perhaps not surprising given the dependence of GABAAR expression on neuronal activity (Ives et al., 2002;Kilman et al., 2002). However, the observation that the restoration of FMRP in inhibitory neurons in conditional rescue mice does not rescue GABAAR maturation defects interestingly suggests that FMRP function in inhibitory neurons is not critical for postsynaptic receptor maturation. Thus, importantly, defective inhibitory efficacy in the FXS amygdala is comprised of genetically separable, distinct pre- and post-synaptic components
Consequences for the treatment of FXS and autism-related developmental disorders
With regard to considering treatment for FXS, the results presented here suggest that developmental age should be given significant consideration. As our studies focused on the amygdala, they are particularly relevant with regard to the treatment of core behavioral symptoms such as heightened anxiety, disrupted responses to fearful stimuli, and abnormal social functioning (Olmos-Serrano and Corbin, 2011; Martin and Huntsman, 2012). Differences in the gross anatomy of certain regions of the brain in boys with FXS have been detected at as early as one year of age and change as development progresses (Hoeft et al., 2011). Our results suggest that the maturation of neural circuitry on a fine scale in the amygdala has similar developmental differences.
Second, the data we present have implications regarding the types of pharmacological agents that may or may not be helpful, either with regard to their efficacy or their propensity for causing prohibitive side effects, in patients with FXS. Clonazepam, for instance, is often prescribed to treat epilepsy, which has a high rate of co-morbidity with FXS (Leung and Ring, 2013). However, it has been observed that clonazepam treatment can increase the severity of some autistic symptoms or have otherwise enhanced side effects in individuals with FXS (Gillberg, 1995). Our results suggest a possible explanation for these outcomes: GABAA receptor maturation may be defective in FXS patients and, as such, there may be a preponderance of immature, clonazepam-sensitive α2/α3-containing GABAA receptors. If this is the case, clonazepam would have increased effects in FXS patients compared to the neurotypical population. Finally, we show that GABA reuptake is enhanced in the mature FXS amygdala. Therefore, GABA transporters may be viable drug targets for the treatment of FXS symptoms caused by amygdala dysfunction in patients beyond early childhood, when the system has matured. GABA reuptake inhibitors, such as the anti-epileptic drug tiagabine, are well-tested and readily available.
In conclusion, we demonstrate that several mechanisms involved in GABAergic synaptic transmission in the amygdala are affected by the absence of FMRP. Overall, it appears that these mechanisms act in concert to reduce the amount of synaptic GABA as well as decrease the strength of the postsynaptic response mediated by GABAARs. By increasing the amount of synaptic GABA or increasing GABA sensitivity in the amygdala, it may be possible to alleviate some of the more debilitating symptoms of FXS and other developmental disorders. However, the presence or absence of these defects will likely critically depend on the age of the affected individual, highlighting an important additional consideration when developing therapeutic strategies, especially in pediatric cases.
Acknowledgements
This work was supported by grants from NIH/NINDS R01 NS053719 (MMH), Autism Speaks (JGC and MMH), Neurodevelopmental disorders fund (JGC), FRAXA (JGC and MMH), NICHD T32 training grant (RLV), and NIH HD024064 (DLN). The authors wish to acknowledge the members of the Huntsman and Corbin labs for useful discussions and Stefano Vicini and Vittorio Gallo for advice and critical reading of the manuscript.
Footnotes
Conflict of Interest: The authors declare no conflict of interest.
  • Avila MA, Real MA, Guirado S. Patterns of GABA and GABA Transporter-1 immunoreactivities in the developing and adult mouse brain amygdala. Brain Res. 2011;1388:1–11. [PubMed]
  • Barberis A, Lu C, Vicini S, Mozrzymas JW. Developmental changes of GABA synaptic transient in cerebellar granule cells. Mol Pharmacol. 2005;67:1221–1228. [PubMed]
  • Bear MF, Huber KM, Warren ST. The mGluR theory of fragile X mental retardation. Trends Neurosci. 2004;27:370–377. [PubMed]
  • Centonze D, Rossi S, Mercaldo V, Napoli I, Ciotti MT, De Chiara V, Musella A, Prosperetti C, Calabresi P, Bernardi G, Bagni C. Abnormal striatal GABA transmission in the mouse model for the fragile X syndrome. Biological Psychiatry. 2008;63:963–973. [PubMed]
  • Comery TA, Harris JB, Willems PJ, Oostra BA, Irwin SA, Weiler IJ, Greenough WT. Abnormal dendritic spines in fragile X knockout mice: maturation and pruning deficits. Proc Natl Acad Sci U S A. 1997;94:5401–5404. [PubMed]
  • Darnell JC, Van Driesche SJ, Zhang C, Hung KY, Mele A, Fraser CE, Stone EF, Chen C, Fak JJ, Chi SW, Licatalosi DD, Richter JD, Darnell RB. FMRP stalls ribosomal translocation on mRNAs linked to synaptic function and autism. Cell. 2011;146:247–261. [PMC free article] [PubMed]
  • De Koninck Y, Mody I. Noise analysis of miniature IPSCs in adult rat brain slices: properties and modulation of synaptic GABAA receptor channels. J Neurophysiol. 1994;71:1318–1335. [PubMed]
  • Desai NS, Casimiro TM, Gruber SM, Vanderklish PW. Early postnatal plasticity in neocortex of Fmr1 knockout mice. J Neurophysiol. 2006;96:1734–1745. [PubMed]
  • Desai NS, Cudmore RH, Nelson SB, Turrigiano GG. Critical periods for experience-dependent synaptic scaling in visual cortex. Nat Neurosci. 2002;5:783–789. [PubMed]
  • Ebert DH, Greenberg ME. Activity-dependent neuronal signalling and autism spectrum disorder. Nature. 2013;493:327–337. [PMC free article] [PubMed]
  • Edwards FA, Konnerth A, Sakmann B. Quantal analysis of inhibitory synaptic transmission in the dentate gyrus of rat hippocampal slices: a patch-clamp study. J Physiol. 1990;430:213–249. [PubMed]
  • Galanopoulou AS. GABA(A) receptors in normal development and seizures: friends or foes? Curr Neuropharmacol. 2008;6:1–20. [PMC free article] [PubMed]
  • Gibson JR, Bartley AF, Hays SA, Huber KM. Imbalance of neocortical excitation and inhibition and altered UP states reflect network hyperexcitability in the mouse model of fragile X syndrome. J Neurophysiol. 2008;100:2615–2626. [PubMed]
  • Gillberg C. Clinical Child Neuropsychiatry. Cambridge University Press; Cambridge, UK: 1995. Interventions and Treatments. pp. 326–339.
  • Gonzalez-Islas C, Wenner P. Spontaneous network activity in the embryonic spinal cord regulates AMPAergic and GABAergic synaptic strength. Neuron. 2006;49:563–575. [PubMed]
  • Harlow EG, Till SM, Russell TA, Wijetunge LS, Kind P, Contractor A. Critical period plasticity is disrupted in the barrel cortex of FMR1 knockout mice. Neuron. 2010;65:385–398. [PMC free article] [PubMed]
  • Hartman KN, Pal SK, Burrone J, Murthy VN. Activity-dependent regulation of inhibitory synaptic transmission in hippocampal neurons. Nat Neurosci. 2006;9:642–649. [PubMed]
  • Hoeft F, Walter E, Lightbody AA, Hazlett HC, Chang C, Piven J, Reiss AL. Neuroanatomical differences in toddler boys with fragile x syndrome and idiopathic autism. Arch Gen Psychiatry. 2011;68:295–305. [PubMed]
  • Hollrigel GS, Soltesz I. Slow kinetics of miniature IPSCs during early postnatal development in granule cells of the dentate gyrus. J Neurosci. 1997;17:5119–5128. [PubMed]
  • Huber KM, Gallagher SM, Warren ST, Bear MF. Altered synaptic plasticity in a mouse model of fragile X mental retardation. Proc Natl Acad Sci U S A. 2002;99:7746–7750. [PubMed]
  • Huntsman MM, Huguenard JR. Nucleus-specific differences in GABA(A)-receptor-mediated inhibition are enhanced during thalamic development. J Neurophysiol. 2000;83:350–358. [PubMed]
  • Huntsman MM, Huguenard JR. Fast IPSCs in rat thalamic reticular nucleus require the GABAA receptor beta1 subunit. J Physiol. 2006;572:459–475. [PubMed]
  • Ives JH, Drewery DL, Thompson CL. Neuronal activity and its influence on developmentally regulated GABA(A) receptor expression in cultured mouse cerebellar granule cells. Neuropharmacology. 2002;43:715–725. [PubMed]
  • Jacobson M. Formation of dendrites and development of synaptic connections. In: Rao MS, Jacobson M, editors. Developmental Neurobiology. Plenum Press; New York, NY: 1991. pp. 223–284.
  • Kilman V, van Rossum MC, Turrigiano GG. Activity deprivation reduces miniature IPSC amplitude by decreasing the number of postsynaptic GABA(A) receptors clustered at neocortical synapses. J Neurosci. 2002;22:1328–1337. [PubMed]
  • Leung HT, Ring H. Epilepsy in four genetically determined syndromes of intellectual disability. J Intellect Disabil Res. 2013;57:3–20. [PubMed]
  • Martin BS, Huntsman MM. Pathological plasticity in fragile X syndrome. Neural Plast. 2012;2012:1–12.
  • Mientjes EJ, Nieuwenhuizen I, Kirkpatrick L, Zu T, Hoogeveen-Westerveld M, Severijnen L, Rife M, Willemsen R, Nelson DL, Oostra BA. The generation of a conditional Fmr1 knock out mouse model to study Fmrp function in vivo. Neurobiol Dis. 2006;21:549–555. [PubMed]
  • Nimchinsky EA, Oberlander AM, Svoboda K. Abnormal development of dendritic spines in FMR1 knock-out mice. J Neurosci. 2001;21:5139–5146. [PubMed]
  • Nusser Z, Cull-Candy S, Farrant M. Differences in synaptic GABA(A) receptor number underlie variation in GABA mini amplitude. Neuron. 1997;19:697–709. [PubMed]
  • Olmos-Serrano JL, Corbin JG. Amygdala regulation of fear and emotionality in fragile X syndrome. Dev Neurosci. 2011;33:365–378. [PMC free article] [PubMed]
  • Olmos-Serrano JL, Paluszkiewicz SM, Martin BS, Kaufmann WE, Corbin JG, Huntsman MM. Defective GABAergic neurotransmission and pharmacological rescue of neuronal hyperexcitability in the amygdala in a mouse model of fragile X syndrome. J Neurosci. 2010;30:9929–9938. [PMC free article] [PubMed]
  • Paysan J, Bolz J, Mohler H, Fritschy JM. GABAA receptor alpha 1 subunit, an early marker for area specification in developing rat cerebral cortex. J Comp Neurol. 1994;350:133–149. [PubMed]
  • Peng YR, Zeng SY, Song HL, Li MY, Yamada MK, Yu X. Postsynaptic spiking homeostatically induces cell-autonomous regulation of inhibitory inputs via retrograde signaling. J Neurosci. 2010;30:16220–16231. [PubMed]
  • Pilpel Y, Kolleker A, Berberich S, Ginger M, Frick A, Mientjes E, Oostra BA, Seeburg PH. Synaptic ionotropic glutamate receptors and plasticity are developmentally altered in the CA1 field of Fmr1 knockout mice. J Physiol. 2009;587:787–804. [PubMed]
  • Rudolph U, Crestani F, Benke D, Brunig I, Benson JA, Fritschy JM, Martin JR, Bluethmann H, Mohler H. Benzodiazepine actions mediated by specific gamma-aminobutyric acid(A) receptor subtypes. Nature. 1999;401:796–800. [PubMed]
  • Stenman J, Toresson H, Campbell K. Identification of two distinct progenitor populations in the lateral ganglionic eminence: implications for striatal and olfactory bulb neurogenesis. J Neurosci. 2003;23:167–174. [PubMed]
  • Suvrathan A, Hoeffer CA, Wong H, Klann E, Chattarji S. Characterization and reversal of synaptic defects in the amygdala in a mouse model of fragile X syndrome. Proc Natl Acad Sci U S A. 2010;107:11591–11596. [PubMed]
  • Swanwick CC, Murthy NR, Mtchedlishvili Z, Sieghart W, Kapur J. Development of gamma-aminobutyric acidergic synapses in cultured hippocampal neurons. J Comp Neurol. 2006;495:497–510. [PMC free article] [PubMed]
  • Tia S, Wang JF, Kotchabhakdi N, Vicini S. Developmental changes of inhibitory synaptic currents in cerebellar granule neurons: role of GABA(A) receptor alpha 6 subunit. J Neurosci. 1996;16:3630–3640. [PubMed]
  • Turrigiano GG. Homeostatic plasticity in neuronal networks: the more things change, the more they stay the same. Trends Neurosci. 1999;22:221–227. [PubMed]
  • Turrigiano GG, Leslie KR, Desai NS, Rutherford LC, Nelson SB. Activity-dependent scaling of quantal amplitude in neocortical neurons. Nature. 1998;391:892–896. [PubMed]
  • Verdoorn TA. Formation of heteromeric gamma-aminobutyric acid type A receptors containing two different alpha subunits. Mol Pharmacol. 1994;45:475–480. [PubMed]
  • Vicini S, Ferguson C, Prybylowski K, Kralic J, Morrow AL, Homanics GE. GABA(A) receptor alpha1 subunit deletion prevents developmental changes of inhibitory synaptic currents in cerebellar neurons. J Neurosci. 2001;21:3009–3016. [PubMed]
  • Wilson BM, Cox CL. Absence of metabotropic glutamate receptor-mediated plasticity in the neocortex of fragile X mice. Proc Natl Acad Sci U S A. 2007;104:2454–2459. [PubMed]
  • Zhang J, Hou L, Klann E, Nelson DL. Altered hippocampal synaptic plasticity in the FMR1 gene family knockout mouse models. J Neurophysiol. 2009;101:2572–2580. [PubMed]
  • Zoghbi HY. Postnatal neurodevelopmental disorders: meeting at the synapse? Science. 2003;302:826–830. [PubMed]