Abstract
Severe thyroid hormone (TH) deficiency during critical phases of brain development results in irreversible neurological and cognitive impairments. The mechanisms accounting for this are likely multifactorial, and are not fully understood. Here we pursue the possibility that one important element is that TH affects basal and activity-dependent neurotrophin expression in brain regions important for neural processing. Graded exposure to propylthiouracil (PTU) during development produced dose-dependent reductions in mRNA expression of nerve growth factor (Ngf) in whole hippocampus of neonates. These changes in basal expression persisted to adulthood despite the return to euthyroid conditions in blood. In contrast to small PTU-induced reductions in basal expression of several genes, developmental PTU treatment dramatically reduced the activity-dependent expression of neurotrophins and related genes (Bdnft, Bdnfiv, Arc, and Klf9) in adulthood and was accompanied by deficits in hippocampal-based learning. These data demonstrate that mild TH insufficiency during development not only reduces expression of important neurotrophins that persists into adulthood but also severely restricts the activity-dependent induction of these genes. Considering the importance of these neurotrophins for sculpting the structural and functional synaptic architecture in the developing and the mature brain, it is likely that TH-mediated deficits in these plasticity mechanisms contribute to the cognitive deficiencies that accompany developmental TH compromise.
Thyroid hormone (TH) is essential for normal development of the central nervous system; severe TH deficiency in the fetus and neonate produces severe neurological deficits including cognitive impairment and mental retardation (1–3). However, even mild TH insufficiency in pregnant women can produce cognitive deficits in the offspring (1, 4). Recently, animal models have confirmed that modest degrees of maternal TH insufficiency produce dose-dependent changes in molecular, anatomical, electrophysiological, and behavioral effects in the offspring (5–8). Despite advances in our recognition that the developing brain is sensitive to mild TH insufficiency, the specific mechanisms by which TH affects brain development, and the developmental events that are disturbed by TH insufficiency, remain poorly understood. This is important both because an understanding of the mechanisms and developmental events controlled by TH may provide insight into strategies for treating children of women with TH insufficiency but will also identify the kinds of molecular and developmental endpoints that could be captured in tests of chemicals that may interfere with TH action.
TH receptors are ligand-activated transcription factors (9) that regulate the expression of specific genes. This regulation is cell-type and developmental stage specific, making the study of TH action in brain development complex. We and others have previously identified TH-responsive elements in the cortex and hippocampus of the neonatal rat (6, 10–12) and postulated that neurotrophins may be important mediators of TH action in the developing hippocampus (11, 13, 14). This possibility is provocative, in part because neurotrophins in general are strongly implicated in mechanisms of neuronal plasticity during development and in adulthood. The mechanisms of neuronal plasticity during brain development are similar to those proposed to occur during learning and neurotrophin-based molecular signals subserving both are conserved from development to adulthood. TH insufficiency may interfere with these mechanisms. The present studies were designed to test this hypothesis in 3 ways. First, we examined the effect of TH insufficiency in pregnant rats on the basal expression of a number of neurotrophin genes in young (neonatal) and adult offspring. Secondly, learning was assessed because learning initiates alterations in glutamate-dependent excitatory synaptic transmission that are subsequently stabilized through brain-derived neurotrophic factor (BDNF)-induced structural changes at the synapse (15–17). We examined learning and memory in adult offspring of hypthyroxenemic dams using specific hippocampally mediated spatial learning and contextual fear conditioning paradigms. Finally, we examined the effect of induction of long-term potentiation (LTP), the glutamate-dependent cellular correlate of associative learning, on neurotrophin expression. Our findings reveal what may be a critical deficit in the brains of adult offspring exposed to developmental hypothyroxinemia. Specifically, although neurotrophin gene expression was only modestly affected by developmental hypothyroxinemia, hippocampal-based learning was impaired, and LTP-induced neurotrophin expression was severely suppressed by this treatment. Thus, the ability of the brain to respond to neural activity may be an important mechanism by which developmental hypothyroidism alters neuronal network formation in the developing brain and contributes to cognitive deficits in the adult.
Materials and Methods
Developmental exposure
Pregnant Long-Evans rats were obtained from Charles River on gestational day 2 and housed individually in standard plastic hanging cages in an Association for Assessment and Accreditation of Laboratory Animal Care-certified animal facility. All animal treatments were in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Animal rooms were maintained on a 12-hour light, 12-hour dark schedule, and rats were permitted free access to food (Purina 5008 rat chow) and deionized water. Beginning on gestational day 6 dams (15/dose group) were administered via the drinking water, 0, 1, 2, 3, or 10 parts-per-million (ppm, 0.0001%, 0.0002%, 0.0003%, and 0.001%, wt/vol) of the antithyroid agent propylthiouracil (PTU) (Sigma) in deionized water until postnatal day (PN)21. The day of birth was designated PN0, and all litters were culled to 10 pups (5 males and 5 females to the extent possible) on PN4. One male and 1 female from each litter were killed on PN14 and PN90 and blood and hippocampus collected. With the exception of the high-dose group, pups were weaned on PN21, transferred to plastic hanging cages (2/cage), and permitted free access to food (Purina 5001) and filtered tap water. Dams exposed to 10 ppm were removed from PTU water on PN21 but remained with pups for an additional week to increase survivability of the pups before weaning. Dam weights were monitored frequently throughout pregnancy and offspring weights were recorded during the first postnatal month and once again on PN90. Dams were killed when pups were weaned and blood collected for TH analysis.
Hippocampus collected from male offspring (n = 8 litters/dose group) killed on PN14 and PN90 were flash frozen for gene expression analysis. Two behavioral tests sensitive to hippocampal dysfunction were conducted on independent groups of adult male offspring. Finally, LTP was induced and activity-induced expression of a set of neurotrophin-related genes were assessed.
Behavioral assessments
Two hippocampal-dependent memory tests were evaluated in different groups of adult male offspring. Spatial learning using the Morris water maze (MWM) began at approximately 3 months of age in 1 male from each of 13–14 litters/dose group according to previously described procedures (18). Briefly, 4 trials per day for 5 consecutive days were conducted with animals introduced into the pool from a different start point on every trial each day. Animals were allowed to swim freely in search of an escape platform hidden below the surface of the water, and once found, to rest upon it for 15 seconds. If not found within 60 seconds, animals were guided to it by the experimenter. On trials 6, 11, 14, and 20, the platform was removed and reinserted after 60 seconds of free swim, and animals rested on it for 15 seconds before removal from the pool. The time spent searching the correct quadrant and the number of times the animal's path crossed the platform position were recorded. Cue testing with a visible platform raised above the water surface was conducted on the sixth day of testing.
Beginning at approximately 5 months of age, trace fear conditioning was conducted in a different set of male offspring than used for MWM. Testing was performed in 9–12 animals/dose from 6–11 litters with no more than 2 animals represented from any given litter. Procedures followed slight modifications to those employed previously in our laboratory in hypothyroid animals (7) and were identical to those described by Oshiro et al (19). Briefly, a total of 2 training trials were given. After a 2-minute baseline period, a compound 15-second light/tone stimulus was presented, terminating 30 seconds before (the trace interval) administration of a brief, mild footshock (0.5 seconds, 1 mA). An intermittent noncontingent “distractor” flashing light stimulus was randomly presented throughout the training period. Activity was monitored via a motion detector. The next day, animals were returned to the same training box and activity monitored for 5 minutes as an index of contextual learning. About 1–2 hours later, animals were placed in a different testing box with visual, tactile, and olfactory cues distinct from the training/context testing box and housed in a different test room. Activity was monitored for 2 minutes before and after presentation of the compound light/tone stimulus previously paired with footshock to evaluate conditioning to cue.
Surgical procedures
Between 9 and 12 months of age, LTP was induced in a subset of animals previously tested in MWM or fear conditioning several months earlier. Animals were anesthetized with urethane (1.5–2 g/kg, ip), mounted in a stereotaxic frame, stimulating electrodes were lowered into the angular bundle of the perforant path of the left hemisphere according to standard procedures described previously (20). A recording electrode was lowered into the ipsilateral dentate gyrus (DG). Optimal depth placement was achieved through electrophysiological monitoring of the response evoked in the DG after single pulse perforant path stimulation. Only animals for which an optimal waveform morphology, minimum field potential amplitude of 3 mV, an excitatory postsynaptic potential (EPSP) threshold of less than 200 μA, and stable recordings over the course of electrophysiological monitoring were included in analysis for activity-dependent gene expression. These criteria were achieved in offspring from 8, 6, 10, and 8 animals from the 0-, 2-, 3-, and 10-ppm dose groups, respectively. The positive component of the field potential provides an index of synaptic activation resulting from the summed EPSPs. The slope of the EPSP was calculated as the rate of amplitude change for the initial positive component of the field potential before population spike (PS) onset.
Electrophysiological recordings and induction of LTP
Once optimal electrode placement was achieved, responses were monitored for approximately 2 hours to ensure stability. Responses were amplified, digitized (33-kHz sampling rate), averaged using LabWindows (National Instruments) and custom designed software, and stored on a PC for later analysis. Once stable, intensity was adjusted to produce a PS 30% of the maximal recorded at 1500 μA, and a series of sweeps were collected at 10-second intervals. Five 5-sweep averages sampled at 5-minute intervals were collected at this “test stimulus” intensity just before application of LTP-inducing trains. Train parameters effective in inducing BDNF expression (21) were applied to the perforant path (4 8-pulse train bursts, 400 Hz, 1500 μA delivered at 10-s intervals, repeated 3 times at 5-min intervals). Sampling of 5-sweep averages at 15-minute intervals at the pretrain test stimulus intensity resumed 1 minute after the final train delivery and continued for the next 5 hours when animals were killed by decapitation. The brain was removed from the skull and the hippocampus immediately dissected. The DG was separated from the CA (Cornu Ammonis) regions (primarily CA1 was harvested) from both the stimulated and the unstimulated hemispheres according to procedures outlined in Hagihara et al (22). Briefly, on removal from the brain, the hippocampus was laid medial surface up on a cold plate moistened with ice cold artificial cerebrospinal fluid. A 30 gauge hypodermic needle was used to section along the fissure that separates DG from CA regions. DG and CA1 were collected, flash frozen and stored at −80°C until processing.
Quantitative RT-PCR
RNA was extracted from whole hippocampus in 1 set of animals killed on PN14 and PN90, and from the DG and CA1 regions of the stimulated and nonstimulated hemispheres of animals after LTP induction using Tri reagent (Sigma) according to manufacturer's instructions. RNA pellets were resuspended in nuclease-free H20 and the concentration determined on a Nanodrop ND-100. Two micrograms of total RNA were initially digested using 2-U DNase I (Promega). Some of this RNA was quantified using a Quant-iT RiboGreen RNA assay kit (Life Technologies). Another portion of the DNased RNA was reverse transcribed using a High Capacity cDNA Archive kit (Life Technologies) according to the provided protocol. Amplification was performed on an Applied Biosystems 7900HT Fast Real-Time PCR System in duplicate using 25-ng cDNA and TaqMan Universal PCR Master Mix in a total volume of 12 μL. PCR cycling conditions were 2 minutes at 50°C, 10 minutes at 95°C, then 40 cycles of 15 seconds at 95°C and 1 minute at 60°C. The TaqMan primers used are listed in Supplemental Table 1. Selection of β-2-microglobulin (B2m) and gyceraldehyde-3-phosphate dehydrogenase (Gadph) as endogenous reference genes was based on expression at uniform levels across all treatments as determined by ANOVA. Relative quantification was measured by the delta-delta cycle threshold (ddCt) method and mean fold change (FC) calculated by normalizing to the mean dCt of the control group.
In the first experiment, probes for BDNF-total (Bdnft), nerve growth factor (Ngf), and neurotrophin factor 3 (Ntf3) were evaluated in whole hippocampus. The BDNF exon iv (Bdnfiv) was also examined as it has been specifically implicated in activity-dependent plasticity (13). Another target gene for the calcium-binding protein, parvalbumin (Pvalb), was included as a positive control. In the second experiment, gene expression was performed independently in the CA1 and DG regions of the hippocampus, including probes for the neurotrophins above in addition to other activity-dependent immediate early genes Arc (activity-regulated gene, also known as Arg3.1), early growth factor 1 (Egr-1) (also known as Zif268, Krox24), mammalian target of rapamycin (mTor), and the TH-dependent transcription factor Klf9. Gene expression in the nonstimulated hemisphere was taken as estimate of “basal” expression and the reference for induction by LTP in the stimulated hemisphere.
Functional magnetic resonance imaging has revealed bilateral hippocampal activation after unilateral stimulation to induce LTP (23). Bilateral increases in the expression of neurotrophins have also been reported after unilateral LTP (21), raising a potential confound in the use of the unstimulated hemisphere as the reference for expressing LTP-induced augmentation in neurotrophin expression. For this reason expression of a subset of gene targets (Arc, Bdnfiv, Klf9, and Ngf) was examined in the DG harvested from anesthetized but not implanted animals (surgical controls) and dCts compared with those obtained from the nonstimulated hemisphere in the 0-ppm group.
Statistical analyses
Standard general linear model ANOVA was conducted using SAS (version 9.2; SAS Institute). Baseline EPSP and PS amplitudes, dCts in DG contrasting surgical control unstimulated hemisphere expression levels, activity counts during training, and testing in fear conditioning were evaluated using one-way ANOVAs with α-level set to P < .05. LTP and MWM acquisition were assessed with repeated measures ANOVA. Contrast tests within ANOVA were performed for each dose group relative to control if significant main effects or interactions were observed. In MWM, quadrant dwelling time was also evaluated by determining search time that was greater than chance (25%) for each dose group using t-tests. For gene expression of whole hippocampus, statistical significance of FC from control was determined by one-way ANOVA followed by the Dunnett's post hoc test. Repeated measures ANOVAs were conducted for DG and CA1 regions independently in the samples taken after LTP. Factors included in the analyses were dose (which collapses across hemisphere), stimulation (which collapses across dose), and dose × hemisphere interaction terms. Significant main effects of dose indicate PTU-induced alteration in gene expression. Significant main effects of hemisphere indicate LTP-induced alteration in gene expression. Most importantly, significant dose × hemisphere interactions indicate differential effects of PTU on activity-induced gene expression. Because a number of ANOVAs was performed to evaluate the expression of many genes in the same samples, an α-level of 0.025 was applied to reduce experiment-wise error rate for both whole hippocampus and LTP-induced gene expression experiments.
Results
Characterization of the low-dose PTU model
Serum TH and body weights in dams and pups from this cohort of animals were previously reported in Johnstone et al (24) but for completeness are summarized in Supplemental Table 2. Body weight of the dams was not altered, body weight of pups was reduced in the 10-ppm dose group from PN4, remaining below controls weights throughout lactation. The 2- and 3-ppm dose groups did not differ from controls until PN21 when a small but significant reduction was seen. Body weight deficits disappeared in adulthood for these dose levels, but persisted in the 10-ppm dose group. Maternal T4 values measured at weaning of the pups were dose dependently reduced, TSH was increased in the 2- and 3-ppm dose groups. As described in Materials and Methods, pups were weaned 1 week later in the high-dose group to optimize pup survival during which time dam and pup TH and TSH had completely recovered. No changes in maternal T3 measures were observed at any dose. Pup serum T4 on PN14 was dose dependently reduced and to a greater degree than that observed in the dam. Reductions in serum T3 in pups were restricted to the high-dose group.
PTU effects on neurotrophin gene expression in hippocampus of the neonate and adult
We selected the neurotrophins Bdnft, Bdnfiv, Ngf, and Ntf3 because of their role in neurodevelopment and activity-dependent plasticity. No changes in expression of the neurotrophins Bdnft, Bdnfiv, and Ntf3 were detected in the hippocampus of offspring of PTU-treated dams (all P > .50) (Figure 1A). Ngf was reduced in a dose-dependent manner in the neonate at the 2 highest dose levels (P < .0001), a suppression that persisted in the adult hippocampus at PN90 (P < .0003) (Figure 1B). Although no change in the expression of Bdnfiv was observed in the neonate (P > .065), surprisingly, small but statistically significant increases were detected in the adult hippocampus at the 2 highest dose levels (P < .0024). Pvalb was included as a positive control as it has previously been shown to be markedly reduced in the hippocampus by perinatal exposure to graded levels of PTU (6, 10). Pvalb expression was reduced at all dose levels in the neonatal (P < .0001) (Figure 1A), but was not different from control levels in the adult hippocampus (P > .78) (Figure 1B).
Neurotrophins in neonate and adult hippocampus. Mean (±SEM, n = 8/dose group) FC in expression level of 4 neurotrophin genes in the whole hippocampus from male offspring on (A) PN14 and (B) PN90. The expression level of Ngf was dose dependently reduced on PN14 (P < .0001), and this reduction persisted into adulthood (P < .0003). No change was detected in the other neurotrophins assessed on PN14 (Bdnft, Bdnfiv, and Ntf3, all P > .05) or on PN90 with the exception of a slight but statistically significant increase in Bdnfiv expression (P < .0024). The gene probe for the calcium-binding protein Pvalb was included as a positive control gene (see text) and was dose dependently reduced at all dose levels in the neonate (P < .0001) but exhibited no change in adulthood. *, P < .05, in Dunnett's t test conducted after significant effect of dose was detected in ANOVA with stringent α-level of 0.025 applied to reduce experiment-wise error rate.
Behavioral assessments
Morris water maze
PTU-exposed animals exhibited spatial learning deficits as assessed in the MWM. Latency to find the escape platform was longer across several trials in the PTU-exposed animals relative to controls (dose, P < .0001; dose × trial interaction, P < .009) (Figure 2A). Acquisition deficits were most pronounced in animals exposed to 10-ppm PTU (mean contrast test in ANOVA, 0 vs 10 ppm, P < .0001), but impairments were also evident in the 3-ppm dose group (0 vs 3 ppm, P < .039), and limited to the early trials. Performance improved with additional training, but deficits persisted in the high dose group. Temporary removal of the escape platform during “probe” trials intermittently inserted throughout acquisition, support these observations. Deficits were seen at all dose levels on the first probe trial in that a greater than chance dwelling time in the correct quadrant (ie, evidence of learning) was seen only in controls (Figure 2B). Fewer crossings over the position where the escape platform had been reliably placed on previous trails were also observed in dosed animals (Figure 2C) on the first 2 probe trials (trials 6 and 11, both P < .01). No differences between groups were seen on performance in cue learning when the platform was made visible or in swim speed (data not shown).
Morris Water Maze. A, Mean (±SEM, n = 13–14/dose group) latency to find the hidden platform was increased in 3- and 10-ppm animals indicative of impaired acquisition. Significant main effects of dose and trial were detected in the overall analysis, and 10- and 3-ppm dose groups were different from controls in contrast tests collapsed across trial (+, P < .03; ++, P < .0007). Deficits were restricted to the early learning trials in the 3-ppm dose group but persisted in the high dose group. Duncan t test contrast to control; *, P < .05 for 10 ppm; #, P < .05 for 3 ppm on individual trials. B, Probe trials were interspersed throughout training on trials 6, 11, 15, and 20. Chance performance was seen in all dosed groups but not in controls on the in initial probe. *, P < .05, indicates significantly different from chance performance (25% time in correct quadrant, dotted line). C, Mean number of platform crossings were significantly lower in PTU-treated animals relative to controls (dose, P < .001), and these differences were more prevalent in the first (P < .001) and second (P < .01) probe trials.
Distract trace fear conditioning
Learning impairments were evident in trace fear conditioning, confirming and extending a previous report by Gilbert (7). No differences in baseline activity counts were seen during training (data not shown), but animals placed back into the context previously associated with shock did not suppress their activity to the same extent as controls (dose, P < .003) (Figure 3A). Cue learning, as evidenced by suppression of activity during the trace period, also appeared to be impaired in exposed offspring, although the overall ANOVA failed to reveal an overall effect of dose (P = .10). Mean contrast tests at each dose however suggest impairment at the 3-ppm (P < .04) and 10-ppm (P < .03) dose groups (Figure 3B).
Trace Fear Conditioning. A, Mean (±SEM, n = 9–12/dose group) activity counts after reintroduction to the training box in which shock was received 24 hours earlier were higher in all PTU-treated groups relative to controls, indicative of a deficit in context learning (dose, P < .0028). No differences were detected in baseline activity levels during training (data not shown). B, Mean (±SEM) percent of baseline activity counts corresponding to the trace interval after tone/light stimulus 24 hours after training but presented in a unique context. Higher activity counts were suggestive of impaired learning to cue but the overall effect of dose failed to reach statistically significant levels (P = .10).
DG field potentials in response to PTU
Baseline synaptic transmission for all animals with single electrode penetrations, minimal response amplitudes, and thresholds criteria are summarized in Table 1. As expected from previous work (7, 20), the maximal amplitude of the EPSP slope (P < .0034) and PS (P < .0189) of the field potential recorded from the DG were reduced in PTU-exposed animals. The EPSP slope of the test stimulus was also smaller as a function of developmental PTU exposure (P < .0372), but the stimulus intensity required to evoke it did not differ by dose (P > .36). Distinct from previous work, these reductions in excitatory synaptic transmission were statistically limited to the highest dose group, likely reflecting the smaller sample size of the present study based on strict criteria for inclusion for gene analysis (ie, single hemisphere penetration only). Analysis of baseline synaptic transmission with all animals included (ie, penetration of both hemispheres) echoed deficits at lower doses reported previously (data not shown).
Mean (±SEM) EPSP Slope of Probe Stimulus, Set at Intensity Required to Produce a PS 30% of Maximal, the Intensity Required to Evoke Probe, EPSP Slope, and PS at Maximal Stimulus Intensity of 1500 μA
| PTU Dose (ppm) | n | Probe Intensity (μA) | Probe EPSP Slope (mv/ms) | Maximal EPSP Slope (mv/ms) | Maximal Population Spike (mv) |
|---|---|---|---|---|---|
| 0 | 8 | 533.8 ± 48.7 | 3.4 ± 0.5 | 5.2 ± 0.5 | 10.4 ± 1.9 |
| 2 | 6 | 435.0 ± 48.9 | 3.3 ± 0.7 | 5.4 ± 0.8 | 10.6 ± 2.2 |
| 3 | 10 | 428.0 ± 54.4 | 3.0 ± 0.4 | 4.8 ± 0.7 | 9.4 ± 1.6 |
| 10 | 8 | 537.1 ± 73.5 | 1.5 ± 0.2a | 2.1 ± 0.3a | 3.5 ± 0.6a |
| PTU Dose (ppm) | n | Probe Intensity (μA) | Probe EPSP Slope (mv/ms) | Maximal EPSP Slope (mv/ms) | Maximal Population Spike (mv) |
|---|---|---|---|---|---|
| 0 | 8 | 533.8 ± 48.7 | 3.4 ± 0.5 | 5.2 ± 0.5 | 10.4 ± 1.9 |
| 2 | 6 | 435.0 ± 48.9 | 3.3 ± 0.7 | 5.4 ± 0.8 | 10.6 ± 2.2 |
| 3 | 10 | 428.0 ± 54.4 | 3.0 ± 0.4 | 4.8 ± 0.7 | 9.4 ± 1.6 |
| 10 | 8 | 537.1 ± 73.5 | 1.5 ± 0.2a | 2.1 ± 0.3a | 3.5 ± 0.6a |
P < .05 Dunnett's t test after significant effect of dose in ANOVA.
Mean (±SEM) EPSP Slope of Probe Stimulus, Set at Intensity Required to Produce a PS 30% of Maximal, the Intensity Required to Evoke Probe, EPSP Slope, and PS at Maximal Stimulus Intensity of 1500 μA
| PTU Dose (ppm) | n | Probe Intensity (μA) | Probe EPSP Slope (mv/ms) | Maximal EPSP Slope (mv/ms) | Maximal Population Spike (mv) |
|---|---|---|---|---|---|
| 0 | 8 | 533.8 ± 48.7 | 3.4 ± 0.5 | 5.2 ± 0.5 | 10.4 ± 1.9 |
| 2 | 6 | 435.0 ± 48.9 | 3.3 ± 0.7 | 5.4 ± 0.8 | 10.6 ± 2.2 |
| 3 | 10 | 428.0 ± 54.4 | 3.0 ± 0.4 | 4.8 ± 0.7 | 9.4 ± 1.6 |
| 10 | 8 | 537.1 ± 73.5 | 1.5 ± 0.2a | 2.1 ± 0.3a | 3.5 ± 0.6a |
| PTU Dose (ppm) | n | Probe Intensity (μA) | Probe EPSP Slope (mv/ms) | Maximal EPSP Slope (mv/ms) | Maximal Population Spike (mv) |
|---|---|---|---|---|---|
| 0 | 8 | 533.8 ± 48.7 | 3.4 ± 0.5 | 5.2 ± 0.5 | 10.4 ± 1.9 |
| 2 | 6 | 435.0 ± 48.9 | 3.3 ± 0.7 | 5.4 ± 0.8 | 10.6 ± 2.2 |
| 3 | 10 | 428.0 ± 54.4 | 3.0 ± 0.4 | 4.8 ± 0.7 | 9.4 ± 1.6 |
| 10 | 8 | 537.1 ± 73.5 | 1.5 ± 0.2a | 2.1 ± 0.3a | 3.5 ± 0.6a |
P < .05 Dunnett's t test after significant effect of dose in ANOVA.
DG LTP
Distinct from previous reports with this low-dose PTU model (7, 20), a more robust stimulation regimen was followed to maximally induce LTP and neurotrophin expression (see Figure 4A). The EPSP slope amplitude was increased immediately after train delivery in all groups, but the 3- and 10-ppm dose groups demonstrated a significantly smaller magnitude of LTP at these early time points (Figure 4B). All groups demonstrated a decay from maximally induced LTP over the next 5 hours, with declines to pretrain baseline in the 2 highest dose groups. The 2-ppm group tracked a similar path to that observed in controls and could not be discriminated statistically from the 0-ppm dose group. These observations are supported by the results of the repeated measures ANOVA which revealed a significant main effect of dose (P < .0061) and time (P < .0001). The dose × time interaction trended towards but failed to reach statistical significance (P < .0894). Mean contrast tests to examine the significant main effect of dose revealed significant differences from controls for the 3-ppm (P < .0491) and the 10-ppm (P < .0064) dose levels, with no difference between 0- and 2-ppm dose group (P > .47).
Long Term Potentiation. LTP was reduced in adult offspring of PTU-exposed dams. A, Train stimulation protocol optimized to induce LTP-dependent neurotrophin expression. Field potential from the adult offspring of a control dam taken before (dotted line), immediately after, and at 15-minute intervals after delivery of a series of high frequency trains to induce a robust and long lasting (>5 h) LTP. A leftward shift in the slope of the rising phase of the EPSP and a dramatic increase in the size of the negative component of the potential (PS) are indicative of LTP. B, Mean (±SEM) percent of baseline (BL) EPSP slope amplitude before (BL), immediately after (h = 0), and at 15-minute intervals for the next 5 hours. Only animals with single penetration in one hemisphere, minimal response amplitudes of 3mV, EPSP thresholds of less than 200 μA, and stable recordings throughout the pre- and posttrain periods were included (see Table 1). Significant main effects of dose (P < .0061) and time (P < .0001) were detected. There was no difference in the magnitude of LTP seen between the 0- and 2-ppm dose groups (P > .47), but LTP was significantly impaired in both the 3-ppm (+, P < .05) and 10-ppm (++, P < .0064) dose groups.
Basal and activity-dependent induction of neurotrophin expression
Minor differences in basal expression were detected in DG and CA1
PTU-induced effects on basal levels of expression were examined in the nonstimulated hemisphere as these could influence interpretation of any observed effects of stimulation. In general, findings were consistent with results in whole hippocampus of the adult offspring. Ngf was reduced in a dose-dependent manner in both DG and CA1 subregions (both P < .006) (Figure 5, A and B), reflecting the same pattern as seen in whole hippocampus (Figure 1B). The very modest increase in expression of Bdnfiv seen in whole hippocampus (Figure 1B) was not detected under conditions where DG and CA1 regions were independently assessed (both P > .36) (Figure 6, A and B). However, basal levels of Bdnft (Figure 6C) and Ntf3 (Figure 5C) that were unchanged in whole hippocampus (Figure 1B), were slightly reduced in the DG (both P < .02) but not significantly altered in CA1 (both P > .05) (Figures 5D and 6D). Similarly, basal expression levels of the transcription factor Klf9 were reduced in DG (P < .0001) (Figure 6E), but unchanged in CA1 (P > .31) (Figure 6F). Basal expression levels of these 3 transcripts, Bdnft, Ntf3, and Klf9, exhibited a nonmonotonic pattern of change in the DG, with reductions at lower doses and return to controls levels at the highest PTU dose (Figure 6C and E and 5C, respectively). No effects on basal levels of the immediate early genes Arc (P > .72) (Figure 6G), mTor, or Erg-1 (P > .80) (Supplemental Figure 1, A and B) were seen in DG. Basal levels of Arc in area CA1 were variable, displaying a large increase at the lower doses and approaching control levels at the highest dose (P < .007) (Figure 6H). Despite these changes in basal expression levels that were often restricted to DG, the magnitude of change paled in comparison with the large increases in neurotrophins induced by LTP as described below.
Basal but not activity-dependent changes in Ngf and Ntf3. Only animals passing strict criteria were included in genomic analysis (text and Figure 4), resulting in sample sizes of 8 0-ppm, 6 2-ppm, 10 3-ppm, and 8 10-ppm animals. Basal levels of expression of Ngf were dose dependently reduced in PTU-exposed rats in both DG (A) and CA1 (B) regions (dose, both P < .0035). Stimulation to induce LTP produced a large fold increase in expression of Ngf in the DG (hemisphere, P < .0001), but this was not differentially altered in PTU-exposed animals, because no dose × hemisphere interaction was detected (P > .93). Only modest changes in Ngf expression in CA1 followed LTP stimulation, reductions rather than increases, were observed (hemisphere, P < .0011), but no differential effect of PTU was detected (dose × hemisphere interactions, P > .05). C, Ntf3 expression was reduced in response to LTP stimulation in both DG (C) and CA1 (D) (significant effect of hemisphere, both P < .0001), but as with Ngf, not differentially impacted by PTU exposure. Significant main effects of hemisphere are depicted by a line with asterisk above LTP grouping. Mean contrast tests for basal expression revealed by Dunnett's t test, *, P < .05 after significant main effect of dose.
Basal and LTP-induced expression of BDNF-related transcripts are suppressed by PTU. Basal levels of expression of Bdnfiv (A), and Arc (G) in the DG were largely unchanged in PTU-exposed animals. Modest reductions were observed in Bdnft (C, P < .008) and Klf9 transcripts (E, P < .0001). Stimulation to induce LTP produced large increases in expression of all of these transcripts in the DG of control animals that were dramatically blunted in those exposed to PTU. Activity-dependent changes induced by LTP are ascertained by significant main effects of hemisphere depicted as line with asterisk in LTP panels (all P < .0001). Differential effects in control vs PTU-exposed groups are supported by significant dose × hemisphere interactions (all P < .005) followed by mean contrast tests (#, P < .05, mean contrast Dunnett t test relative to LTP-FC in 0-ppm group). Although significant main effects of hemisphere were also evident in area CA1 (B, D, F, and H), these were smaller in magnitude and not differentially affected by PTU (no significant dose × hemisphere interactions).
Activity-dependent expression in DG induced by LTP
Activity-dependent changes in gene expression were examined by contrasting expression levels in the stimulated and nonstimulated hemispheres. Statistically, LTP-induced changes in gene expression are confirmed by a main effect of hemisphere (line above LTP histograms in Figures 5 and 6 indicates significant change from basal levels). LTP induced robust increases in expression of Ngf, Bdnfiv, Bdnft,Klf9, and Arc, and in the DG (LTP, all P < .0001) (Figures 5A and 6, A, C, E, and G). A significant effect of hemisphere was also obtained for Ntf3, but a reduction rather an increase over basal levels was seen with LTP (P < .0001) (Figure 5C). These data indicate that the regulation of these genes is modulated by neuronal activity and are consistent with previous reports investigating DG LTP using in situ hybridization to examine neurotrophin expression (21, 25, 26).
However, no differential effects of hypothyroidism on activity-dependent induction of these neurotrophins was detected. Differential effects of PTU exposure on activity-dependent induction of gene transcripts are indicated by significant dose × hemisphere interactions and summarized in Figure 6, left panel. Only under conditions of a significant interaction were mean contrast tests conducted to identify PTU-induced effects on activity-dependent induction. A strong suppression in activity-induced expression of both Bdnf transcripts was evident at all dose levels of PTU tested (LTP, both P < .005) (Figure 6, A and C) and in Arc expression at the 2 highest dose levels (P < .025) (Figure 6G). Expression of Klf9 was reduced at the 2 lower doses, with some recovery evident in the highest dose group (P < .0007) (Figure 6E), mirroring the pattern seen under basal conditions for this gene. Although Ngf was increased (Figure 5A), and Ntf3 reduced (Figure 5C) as a function of LTP, neither were differentially affected by PTU under stimulation conditions (dose × hemisphere interaction, P > .16). Neither were the immediate early genes Erg1 and mTor, downstream targets of BDNF, induced by LTP or affected by PTU in the basal or stimulated state (Supplemental Figure 2).
Site specificity
Hippocampal CA subregions of both hemispheres were assessed as a control for site specificity of stimulation-induced changes. As in DG, LTP-inducing stimulation up up-regulated expression of Bdnfiv, Bdnft, Klf9, and Arc in the CA1 region (significant main effect of hemisphere, P < .004) (Figure 6, B, D, F, and H), and modestly reduced expression of Ngf and Ntf3 (P < .0001) (Figure 5, B and D). Significant dose × hemisphere interactions in the analyses of CA1 were largely absent, indicating no differential effect of PTU on activity-dependent induction of Ngf, Ntf, Bdnfiv, Bdnft, and Klf9. Although a significant interaction was seen in area CA1 for Arc (P < .002) (Figure 6H), it was not sufficiently robust to statistically distinguish any PTU dose group from controls. Importantly, all changes induced by stimulation and affected by developmental PTU exposure were of a much smaller magnitude in CA1 relative to DG (LTP) (Figures 5 and 6, left vs right panels).
Surgical controls
A third “surgical control experiment” was conducted to evaluate expression of a subset of these gene targets in DG harvested from anesthetized but not implanted animals (n = 9). A comparison of dCts revealed very similar levels of expression of Bdnfiv and Klf9 in the DG of surgical control and the nonstimulated hemisphere of 0-ppm control animals undergoing LTP (both P > .35). However, expression of Arc was lower (dCt means = 7.59 ± 0.18 and 6.07 ± 0.17, P < .0001) and Ngf slightly higher (6.83 ± 0.08 and 7.28 ± 0.05, P < .0002) in the nonstimulated DG relative to the surgical control DG, respectively. The latter is consistent with previous reports of bilateral hippocampal activation and induction of Ngf after unilateral administration of LTP-inducing stimulation (21, 23). Although these differences in basal levels of expression in animals undergoing surgery may result in an underestimation of the magnitude of activity-dependent increase in expression for some transcripts, they are not likely to alter the pattern of change observed or counter the conclusions drawn.
Discussion
In this study, we have demonstrated that small changes in neurotrophin expression were detected under basal conditions, but activity-dependent induction of neurotrophin-related gene expression was severely impaired by developmental hypothyroxinemia. This deficit was accompanied by impairments in synaptic plasticity and in tests of hippocampal-dependent learning and memory. These findings suggest that disruption of neurotrophin signaling may represent a fundamental deficiency underlying neurological impairments induced by developmental TH insufficiency.
Basal levels of neurotrophins
Basal levels of expression of neurotrophins at the doses examined were largely unaffected, either in whole hippocampus of the neonate and adult, or in DG and CA1 subdivisions of adult hippocampus. Ngf was a noteworthy exception - expression levels of this gene were reduced in the hippocampus of the neonate, changes persisted to adulthood, and a pattern was recapitulated with microdissections of hippocampal subregions in an independent group of animals. The findings in the neonatal hippocampus are consistent with reports by others after more severe TH insufficiency (27–29). As such, Ngf may represent a robust biomarker of altered TH-dependent action during hippocampal development that imparts a lingering signature that can be detected after recovery of TH status in serum. By contrast, transcripts of Bdnft and Bdnfiv were not changed in the hippocampus of the neonate, and although very modest increases in Bdnfiv expression were detected in the adult hippocampus, these were not reproduced in independent assessments of DG and CA1. So although neurotrophins have been previously implicated in TH-induced neurotoxicity, changes in basal levels of expression are not readily apparent in response to moderate TH insults.
Hippocampal learning and memory
Both tests of hippocampal-dependent memory were impaired in adult offspring of hypothyroxenemic dams. Spatial learning acquisition deficits were most profound in the high-dose animals, but longer latencies to find the escape platform were also present in animals exposed to 3 ppm, replicating previous findings with this task (20), and consistent with impairments reported using other spatial learning tests (30, 31). Insertion of probe trials throughout the acquisition phase revealed that all PTU-exposed animals, but not controls, did not differ from chance performance in quadrant dwelling time on the first probe (trial 6 of acquisition), and the 2 higher doses remained at chance levels on subsequent probe trials. Less time in the correct quadrant was also accompanied by fewer platform position crossings at all dose levels relative to controls on the first probe trial. Similarly, in a trace fear conditioning paradigm, conditioning to context was impaired at all dose levels. Conditioning to cue failed to reach statistical significance but showed a trend with a similar profile of impairment to that observed for context conditioning. Training in similar hippocampal-based learning paradigms increases Bdnf mRNA expression (15) and impairment of these tasks is observed in mice with hippocampus-specific deletion of BDNF (32) Learning impairments in PTU-exposed animals may reflect deficiencies in BDNF-related activity-dependent signaling pathways as described below.
Activity-dependent induction of neurotrophins in controls
A very robust stimulation paradigm was used to elicit maximal LTP and optimize induction of activity-dependent gene expression. Impairments in EPSP slope LTP were observed in PTU-exposed offspring, consistent with previous reports from our laboratory using a similar dosing regimen but a much milder stimulation (7, 20). Also consistent with previous LTP reports, control animals exhibited large increases in the expression of Ngf, both Bdnf transcripts, the immediate early gene Arc, and reductions in Ntf3 expression in response to LTP (16, 17, 21, 25, 33–36). We report for the first time LTP-induced increases in the TH-dependent transcription factor Klf9, a TH-responsive gene first identified in amphibians (37), and subsequently observed in neonatal rat brain (38). Klf9 is predominantly expressed in the outer fields of granules cells of the developing rat DG, but can be induced by seizures in the adult and has been proposed as a regulator of activity-dependent integration of neurons into the DG during adult neurogenesis (39). Our data support this role for Klf9 as a regulator of activity-dependent neuronal change in the DG. They extend the observations of Scobie et al (39) demonstrating Klf9 activation in response to the milder form of neuronal activation that accompanies LTP. Collectively, these findings further implicate TH as a modulator of activity-dependent plasticity that subserves learning and memory.
TH insufficiency and activity-dependent induction of neurotrophins
The activity-dependent induction in Bdnft, Bdnfiv, Arc, and Klf9 expression in control animals was severely blunted in animals experiencing developmental TH insufficiency. During LTP, Bdnf and its downstream target Arc translocate to the dendrites and accumulate at active synapses in response to neuronal activation (17, 40). These actions serve to support both the persistent enhancements in synaptic transmission of LTP and the structural remodeling of the active synaptic zone (16, 33, 36, 41), and were clearly compromised by TH insufficiency. Robust LTP-induced increases in expression of the transcription factor Klf9 were also strongly suppressed in DG of PTU-exposed animals. DG LTP is impaired in the Klf9 knockout mouse (39). To our knowledge the present data are not only the first report of Klf9 induction by LTP stimulation in control animals, but also the first to reveal persistent impairments in its basal and activity-induced expression by developmental hypothyroidism.
Interestingly, a nonmonotonic pattern in basal and activity-induced expression was observed in the DG for Klf9 and the Bdnft transcript, the significance of which is unclear. It is possible that the delayed weaning in the high-dose group (see Materials and Methods) may have served as an early enrichment, countering some of the effects of hypothyroidism (42). Alternatively, the severity of the hypothyroidism associated with 10-ppm PTU may have been sufficient to activate injury-induced responses of neurotrophins and related genes (43, 44).
Differential effects of TH insufficiency on neurotrophins
Not all LTP-responsive neurotrophins were altered in the same way in response to TH insufficiency. Although basal expression of Ngf and Ntf3 was reduced in the DG of PTU-exposed animals, and both transcripts were responsive to LTP stimulation, distinct from BDNF transcripts, no differential effect of PTU on activity-dependent induction of expression was detected. Neither were the immediate early genes Egr-1 and mTor impacted by hypothyroidism or LTP. Egr-1 and mTor are BDNF-associated transcripts that are rapidly and transiently induced during synaptic plasticity, memory consolidation, and recall (40, 41, 45–47). Failure to demonstrate changes in these transcripts is not surprising given the rapid onset and transient responses of these genes, a profile that was not compatible with the single 5-hour sampling time of the present study chosen to optimize BDNF expression.
Which came first, LTP impairments or neurotrophin deficiencies?
The suppression of activity-dependent up-regulation of Bdnft, Bdnfiv, Arc, and Klf9 are largely consistent with the observed impairment in LTP induction and decay. It is difficult to ascertain whether the LTP impairments are the consequence of deficiencies in mechanisms controlling activation of critical activity-dependent genes, or whether the deficiencies in gene activation are the consequence of failed LTP. However, comparable levels of LTP between controls and the lowest dose group in the presence of suppressed gene expression suggest that developmental TH insufficiency permanently disrupts the underlying machinery controlling activity-dependent plasticity. Although the nonmonotonic pattern of “recovery” of expression of Bdnft and Klf9 at the highest dose was not reflected in functional measures of LTP, spatial and context learning deficits were evident at all doses of PTU. These data indicate that the milder degrees of TH disruption sufficient to suppress activity-dependent genes are associated with persistent neurological impairments in the adult.
Implications of TH-dependent disruption of neurotrophin-based plasticity
The mechanisms underlying plasticity involve strengthening (LTP) and weakening (long-term depression) of synaptic connections, synaptogenesis, structural remodeling, and neurogenesis (27, 36, 48). All these processes are active in the developing brain and are reliant on neurotrophin pathways. These same biological substrates are also conserved in the adult brain where they support learning and memory. Direct infusion of BDNF into the hippocampus enhances synaptic strength, modulates the induction of LTP, and induces structural changes that support learning. Blockade or hippocampus-specific deletion of BDNF impairs spatial memory and LTP, and learning induces a time-dependent induction of BDNF expression (15, 17, 32, 34, 36, 41, 49–51). Compromise of this key neurotrophin pathway by developmental TH insufficiency may contribute to altered connectivity in the developing hippocampus (11, 52–54) in addition to reducing structural (27, 55–57), physiological, and behavioral plasticity in the adult (7, 20). As such, disruption of neurotrophin-based activity-dependent processes as reported here represents a unifying mechanism whereby TH can influence structural and functional synaptic architecture of the developing brain and impair learning in the adult.
Willoughby et al (3) recently described deficiencies in memory function and hippocampal size in children of adolescent age born to hypothyroxinemic women. Román et al (58) reported an association between maternal hypothyroxinemia and increased risk of autistic symptoms, including cognitive impairment in offspring. Others have argued, largely based on models of severe TH deficiencies, that diminished BDNF and related molecules are responsible for abnormal dendritic and synaptic differentiation and altered network connectivity (14, 59) and underlie the clinical manifestations described by Willoughby et al (3) and Román et al (58). The present data represent the first direct empirical support of TH-mediated disruption of activity-dependent neurotrophin signaling essential for synaptic plasticity and support this contention. They demonstrate further that these deficits are induced after moderate degrees of TH insufficiency and that TH-mediated deficiencies in neurotrophin signaling may underlie neurological and cognitive deficits in both the neonate and adult offspring after maternal hypothyroxenemia.
Conclusions
In conclusion, TH insufficiency decreased basal expression levels of the neurotrophin Ngf in the hippocampus of the hypothyroxenemic neonate, and these decrements persisted to adulthood despite complete recovery of TH status. In contrast to relatively minor TH-dependent changes in basal expression of other neurotrophins, dramatic deficiencies in the induction of BDNF and related activity-dependent plasticity markers in adult offspring were unveiled by LTP stimulation. In addition, LTP-induced increases in the expression of the TH-responsive transcription factor Klf9 and its vulnerability to developmental TH insufficiency were demonstrated for the first time. These alterations in expression of activity-dependent genes after moderate degrees of developmental TH insufficiency were accompanied by deficits in spatial learning and context fear conditioning. Considering the importance of these neurotrophin-based plasticity mechanisms in synaptogenesis and network connectivity in the developing brain, it is likely that TH-mediated impairments contribute to altered structural and functional pathways in both the developing and the adult brain.
Acknowledgments
We thank Andrew Johnstone, Masashi Hasegawa, Joan Hedge, and Michelle Taylor for technical assistance. We also thank Dr Stephen Lasley, Dr Christopher Lau, and Dr William Mundy for their comments on an earlier version of this manuscript. We especially thank the comments of Dr R. Thomas Zoeller in revision of this article. This document has been subjected to review by the National Health and Environmental Effects Research Laboratory and approved for publication. Approval does not signify that the contents reflect the views of the Agency, nor does mention of trade names or commercial products constitute endorsement or recommendation for use.
Disclosure Summary: The authors have nothing to disclose.
Abbreviations
- BDNF
brain-derived neurotrophic factor
- CA
Cornu Ammonis
- DG
dentate gyrus
- dCt
delta cycle threshold
- EPSP
excitatory postsynaptic potential
- FC
fold change
- LTP
long-term potentiation
- MWM
Morris water maze
- PN
postnatal day
- ppm
parts-per-million
- PS
population spike
- PTU
propylthiouracil
- TH
thyroid hormone.
