PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Neurotoxicol Teratol. Author manuscript; available in PMC 2013 July 1.
Published in final edited form as:
PMCID: PMC3404268
NIHMSID: NIHMS389512

Effect of chronic glutathione deficiency on the behavioral phenotype of Gclm(−/−) knockout mice

Abstract

Enhanced oxidative stress or deficient oxidative stress response in the brain is associated with neurodegenerative disorders and behavioral abnormalities. Previously we generated a knockout mouse line lacking the gene encoding γ-glutamylcysteine ligase modifier subunit (GCLM). Gclm(−/−) knockout (KO) mice are viable and fertile, yet exhibit only 9–35% of wild-type levels of reduced glutathione (GSH) in tissues, making them a useful model for chronic GSH depletion. Having the global absence of this gene, KO mice—from the time of conception and throughout postnatal life—experience chronic oxidative stress in all tissues, including brain. Between postnatal day (P) 60 and P100, we carried out behavioral phenotyping tests in adults, comparing male and female Gclm(−/−) with Gclm(+/+) wild-type (WT) littermates. Compared with WT, KO mice exhibited: subnormal anxiety in the elevated zero maze; normal overall exploratory open-field activity, but slightly more activity in the peripheral zones; normal acoustic startle and prepulse inhibition reactions; normal novel object recognition with increased time attending to the stimulus objects; slightly reduced latencies to reach a random marked platform in the Morris water maze; normal spatial learning and memory in multiple phases of the Morris water maze; and significantly greater hyperactivity in response to methamphetamine in the open field. These findings are in general agreement with two prior studies on these mice and suggest that the brain is remarkably resilient to lowered GSH levels, implying significant reserve capacity to regulate reactivity oxygen species—but with regional differences such that anxiety and stimulated locomotor control brain regions might be more vulnerable.

Keywords: acoustic startle response, elevated zero maze, GSH, reduced glutathione, locomotor activity, Morris water maze, novel object recognition

Introduction

Oxidative stress occurs when oxygen free radicals production exceeds cellular defense mechanisms (Dalton et al., 1999). The brain generates large amounts of reactive oxygen species (ROS) during cellular metabolism but has antioxidant systems to maintain redox homeostasis. The primary antioxidants are glutathione (GSH) (Dalton et al., 2004), ascorbate (vitamin C) (Harrison and May, 2009), lipoic acid (De Araujo et al., 2011), uric acid (Amaro et al., 2008), carotenes (Obulesu et al., 2011), tocopherols/tocotrienols (vitamin E) (Clarke et al., 2008), ubiquinone/ubiquinol (coenzyme Q) (Dhanasekaran and Ren, 2005), and melatonin (Reiter et al., 2010).

GSH is the most abundant cellular non-protein thiol (2–4 mM in neurons and slightly higher in glia) (Rice and Russo-Menna, 1998). GSH scavenges free radicals and serves as a cofactor for antioxidant enzymes such as glutathione peroxidases and glutathione S-transferases (Meister, 1991). GSH is also important in defense against toxicity from endogenous and exogenous electrophiles (Monks et al., 1999; Fonnum and Lock, 2004) and serves as a signaling molecule through its redox regulation of protein functions (Dalle-Donne et al., 2007). GSH may modulate synaptic transmission via redox-sensitive proteins—including NMDA receptors (Sucher and Lipton, 1991), GABA-A receptors (Amato et al., 1999), calcium-activated K+ channels (Tang et al., 2003), voltage-gated calcium channels (Brown et al., 2005), and glutamate transporters (Volterra et al., 1994; Trotti et al., 1997).

GSH is a synthesized by γ-glutamate combining with cysteine, catalyzed by γ-glutamylcysteine ligase (GCL), followed by addition of glycine via glutathione synthase. The GCL holoenzyme is a heterodimer of the 72.8-kDa catalytic subunit (GCLC) and the 30.8-kDa modifier subunit (GCLM). GCLM maintains GSH homeostasis by optimizing the catalytic efficiency of the holoenzyme properties GCLC (Chen et al., 2005). Both genes are expressed in mouse embryo, fetus, yolk sac, and placenta (Diaz et al., 2002; Diaz et al., 2004). In liver, lung, pancreas, erythrocytes, and plasma, GSH levels in Gclm(−/−) knockout (KO) mice are 9–16% of WT (Yang et al., 2002). Except when challenged with oxidant stress, Gclm(−/−) animals are viable and fertile and exhibit no overt phenotype, making them a useful model of GSH deficiency (McConnachie et al., 2007). Fetal fibroblasts derived from Gclm(−/−) mice are sensitive to oxidative stress and oxidant-induced cell death (Yang et al., 2002). Primary astrocytes from Gclm(−/−) mice show diminished cell viability and higher rates of apoptosis when exposed to an oxidizing agent, the polybrominated flame retardant DE-71 (Giordano et al., 2008).

Disorders associated with GSH deficiency include Parkinsonism, amyotrophic lateral sclerosis, Alzheimer disease, epilepsy and other neurodegenerative diseases (Aoyama et al., 2008). Individuals carrying rare polymorphic variants of the GCLC gene, with low enzymatic activity and decreased intracellular GSH, manifest symptoms ranging from hemolytic anemia to neurological disorders (Ristoff and Larsson, 1998). Other polymorphisms of the GCLC and GCLM genes, that result in modestly lowered GSH levels, have been associated with schizophrenia (Tosic et al., 2006).

The major objective of this study was to assess the causal relationship between chronic GSH deficiency and behavioral abnormalities using male and female adult Gclm(−/−) mice. We previously phenotyped other genetically-modified mice (Schaefer et al., 2009; Curran et al., 2011; Skelton et al., 2011) using a battery of behavioral tests that included a multi-phase version of the Morris water maze known to detect subtle differences in hippocampal dependent spatial learning and memory (Vorhees and Williams, 2006). Herein we determined the effect of chronic GSH deficiency using a similar approach.

Materials and Methods

Animals

Gclm(−/−) knockout mice were previously generated in our lab (Yang et al., 2002). These mice were backcrossed eight times to C57BL/6J (B6) mice to establish a >99.8% B6 background; Gclm(+/−) heterozygotes were bred in order to generate WT controls and KO offspring. Plug-positive females were removed from breeding and housed separately throughout gestation and lactation until dams were removed from their litters on postnatal (P) 28 (birth was counted as P0). Offspring were genotyped and WT and KO mice within each litter were retained for testing. Behavioral testing was conducted between P60–P100. Mice were maintained on regular lab chow and filtered water ad lib on a 14:10 light:dark cycle (lights on at 700 h). The protocol under which this project was conducted was approved by the Institutional Animal Care and Use Committee.

GSH, GSSG, and ascorbate assays

Whole brain was collected from WT and KO mice (N=4 per group) at age P60. Mice were euthanized by CO2, whole brain was excised, snap-frozen in liquid nitrogen, and stored in −80° C. At the time of assay, frozen whole brain was ground in nine volumes of ice-cold redox-quenching buffer (20 mM HCl and 5 mM diethylenetriaminepentaacetic acid) containing 5% trichloroacetic acid. This mixture was then centrifuged at 12,000 × g for 10 min, and the resulting supernatant was used for biochemical measurements. GSH and GSSG levels were determined spectrophotofluorometrically, using the fluorescent probe o-phthalaldehyde, as described (Senft et al. 2000). Another aliquot of the supernatant was treated with dithionite to convert GSSG to GSH, and the total GSH + GSSG was determined. Ascorbate levels were determined by spectrophotometric measurement of ferritin iron released by ascorbate as described (Zannoni et al. 1974). Briefly, an aliquot of the supernatant was mixed sequentially with 85% H3PO4, 1% 2,20-bipyridine, and 3% ferric ammonium sulfate; after incubation at room temperature for 30 min, absorbance was determined at 525 nm. Concentrations of GSH, GSSG, and ascorbate in each sample were interpolated from known standards. Results were reported as μmole/g tissue.

Behavior

Gclm(−/−) mice were compared with WT mice using a battery of behavioral tests (Vorhees, 1996; McIlwain et al., 2001; Bailey et al., 2006; Paylor et al., 2006; Crawley, 2007; Crawley, 2008). Not more than one male and one female/litter/genotype were tested (16–20/group) beginning on P60 as follows: Week-1 = elevated zero maze (EZM), open-field locomotor activity, acoustic startle response (ASR) with prepulse inhibition (PPI). Week-2 = novel-object-recognition (NOR). Week-3 = Morris water maze (MWM) cued; Week-4 = MWM hidden acquisition. Week-5 = MWM hidden reversal. Week-6 = MWM hidden shift. Week-7 = locomotor activity with (+)-methamphetamine (1 mg/kg; subcutaneously) challenge. All tests were performed during the light portion of the light:dark cycle. Not all litters were perfectly balanced by sex and genotype making a balance-by-litter experimental design impossible; however, by testing no more than one male and one female of any one genotype from any given litter, control of any litter effects was maintained.

Monoamine neurotransmitter

Dopamine, serotonin and their respective metabolites DOPAC and 5-HIAA were measured in prefrontal cortex, hippocampus, and neostriatum (caudate and putamen) using the methods described (Grace et al., 2010). Brains were collected from Gclm WT and KO mice (N = 6 to 14 per group). In brief, tissue concentrations of DA, DOPAC, 5-HT, and 5-HIAA were quantified using high-pressure liquid chromatography with electrochemical detection. Tissue weights were determined prior to homogenization in 50 volumes of 0.2 N perchloric acid and centrifuged for 6 min at 10,000 × RCF. Aliquots of 20 μl were injected onto a C18-column (MD-150, 3 × 150 mm; ESA, Chelmsford, MA) connected to a Coulochem detector (25A, Chemsford, MA) and an integrator recorded the peak heights for each injection. The potentials of the E1 and E2 on the analytical cell (model 5014B) of the Coulochem were −150 and +160 mV, respectively. The mobile phase consisted of 35 mM citric acid, 54 mM sodium acetate, 50 mg/l of disodium ethylenediamine tetraacetate, 70 mg/l of octanesulfonic acid sodium salt, 6% (v/v) methanol, 6% (v/v) acetonitrile, pH 4.0, and pumped at a flow rate of 0.4 ml/min. Quantification of analytes was calculated on the basis of standards.

Elevated zero-maze

The EZM is a circular runway (105-cm diameter), 72 cm above the floor with a 10-cm circular path divided equally in four quadrants, designed to assess anxiety based on avoidance of open spaces. Two opposite quadrants have 28-cm walls and two have 1.3-cm clear acrylic curbs. Mice were video-recorded for 5 min and later scored by an observer blind to treatment group for the following behaviors: Latency to leave the closed quadrant, time-in-open, head-dips, and zone-crossings (Shepherd et al., 1994).

Locomotor activity

Open-field arenas were 40 × 40 cm with 16 infrared LED-photocells in x- and y-planes spaced 2.5 cm apart. Mice were tested for 1 h (Accuscan Instruments; Columbus, OH) and photobeam interruptions were analyzed in 5-min intervals. The test assesses exploration and habituation, as well as fear of open spaces—as reflected by central zone activity.

Acoustic startle response with prepulse inhibition

SR-LAB test chambers (San Diego Instruments; San Diego, CA) were used with a 5-min acclimation period with background mixed frequency noise of 68 dB. Following acclimation, mice received a 4 × 4 Latin square protocol of four trial types repeated three times: no-stimulus, startle signal (SS), 74-dB prepulse (PP) + SS, or 76-dB PP + SS. The inter-trial interval was 8 s and the inter-stimulus interval was 70 ms (from PP onset to SS onset). The startle signal was a mixed-frequency white noise burst (120 dB SPL, 20 ms). Prepulse signals were also 20 ms duration. Peak response amplitude (Vmax) was the dependent measure analyzed. This test assesses the basic acoustic startle reflex pathway as modified by higher brain regions using PPI, which reflects sensorimotor gating and, as such, one aspect of attention.

Novel object recognition

Mice were habituated to arenas (91-cm diameter) for 4 days. On the test day, two different identical objects were presented until 30 s of observation had accrued. Observation by the animal was defined as being within 1 cm of the object and oriented toward it, but excluded climbing on it. This method was used in order to try to equalize attention time to objects, so that the denominator in all group comparisons is the same (Clark et al., 2000). One hour later, an exact copy of the familiar object and a novel object were presented until 30 s of observation had accrued (up to 10 min). Objects were counter-balanced to avoid confounding by object preference. The dependent variable was the percent time attending to the novel versus the familiar object. This is a test of memory, based on retention and recognition of a previously seen object.

Morris water maze

The tank was 122-cm diameter and the method similar to that described (Vorhees and Williams, 2006). Day-1: six cued trials with the start and platform positions fixed. The purpose of this procedure was to teach the mice to find, climb on, and remain on the platform until removed. Days 2–6: two trials/day each with a random start and random platform position. During all cued trials, curtains were closed around the pool to obscure distal cues. The escape platform was 10-cm in diameter and contained an orange ball mounted 10 cm above the surface on a metal rod to provide a direct proximal cue to the platform’s location. Following completion of cued trials, mice received three phases of hidden-platform testing. Each of these phases consisted of four trials/day for 6 days with a 30-s probe trial on Day-7 (see (Curran et al., 2011). Each phase used a progressively smaller platform (10, 7, 5 cm in diameter for acquisition, reversal, and shift, respectively). During acquisition the platform was located in the southwest quadrant, during reversal in the northeast quadrant, and during shift in the northwest quadrant—relative to the position where the experimenter stood (arbitrarily designated as south). The platform was positioned halfway between the center and wall of the tank. On hidden platform trials, start positions were pseudo-randomized with the restriction that on each day there was one trial from each of four possible distal start locations. To avoid start positions close to the target quadrant, only distal start positions were used, i.e., two cardinal and two ordinal positions farthest from the platform, e.g., when the platform was in the southwest position, start locations were either cardinal (north or east) or ordinal (northwest or southeast). Data were collected automatically by an overhead video camera attached to a closed-circuit monitor and a computer with tracking software (Smart-track, San Diego Instruments; San Diego, CA). Extracted data files were analyzed on platform trials for latency to reach the platform, cumulative distance to the platform, path length, and swim speed and on probe trials with the platform removed for platform-site crossovers, average distance to the platform site, quadrant preference, and swim speed. This is a test of spatial or allocentric learning with probe trials, which reflects spatial reference memory.

Locomotor activity with drug challenge

Prior to drug challenge, mice were re-habituated to the open-field test chambers with no drug given for 30 min. After 30 min, the test was briefly interrupted and each mouse removed, injected subcutaneously with (+)-methamphetamine HCl (1.0 mg/kg, prepared as the free base in a volume of 1 ml/kg of saline), and returned to the test chamber and activity monitored for an additional 120 min.

Statistical analyse

Behavioral data were analyzed using mixed-linear analysis of variance (ANOVA) with repeated-measures, or by analysis of co-variance (ANCOVA) in the case of locomotor activity after drug challenge using prechallenge activity as the covariate. Data were analyzed using SAS procedures (version 9.2, SAS Institute; Cary, NC). Follow-up analyses of interactions were conducted using slice-effect ANOVAs. Because there were only two groups, a posteriori pairwise group comparisons were analyzed by t-test for independent samples. Data on antioxidant levels were analyzed by t-test for independent samples (two-tailed). Neurotransmitter data were analyzed by general linear model ANOVA. Data are presented as least square (LS) means ± SEM. Results were considered significant if P ≤0.05.

Results

GSH and GSSG levels in brain

At the age (P60) when mice were evaluated by behavioral tests, GSH and GSSG levels (Fig. 1A & B) in whole brain from Gclm(−/−) KO mice were 35% and 58%, respectively, of that from WT mice. The resulting GSH/GSSG ratio was diminished ~40% in KO mice compared with WT mice (Fig. 1C); these data confirm a state of chronic oxidative stress in Gclm(−/−) brain. We did not observe any compensatory increase of ascorbate, another major antioxidant that interplays with GSH (Mãrtensson and Meister, 1991), in the KO brains (Fig. 1D).

Fig. 1
GSH and ascorbate levels in brain. Whole-brain homogenates prepared from males (N=4) at P60 were assayed for GSH, GSSG, and ascorbate levels. The bars represent means and brackets represent S.E.M. *P <0.05 vs WT.

Body weight

Gclm(+/+) and Gclm(−/−) offspring as adults showed slight but significant bodyweight differences (F(1,72) = 6.40, P <0.02). At P100, the LS mean ± SEM of male WT and KO mice were 26.3 ± 0.4 g (N=16) and 25.6 ± 0.4 g (N=18), respectively, and female WT and KO mice were 20.7 ± 0.4 g (N=18) and 19.3 ± 0.4 g (N=21), respectively. KO mice showed no overtly different phenotype and appeared as healthy as WT mice throughout the experiment.

Monoamine neurotransmitters

There were no significant differences in levels of dopamine, serotonin or their metabolites in the three brain regions analyzed (neostriatum, hippocampus, and prefrontal cortex) (data not shown).

EZM

There was a significant genotype main effect for time-in-open (F(1,63) = 4.88, P <0.05) but no interaction with sex (Fig. 2A). Similarly, there was a main effect of genotype for head-dips (F(1,63) = 6.16, P <0.02) (Fig. 2B). There was also a significant main effect of genotype for latency to first open-zone entry (F(1,63) = 3.88, P <0.05). For latency to first open-zone entry, there was also a significant genotype × sex interaction (F(1,63) = 6.37, P<0.02). Slice-effect ANOVA on the interaction showed no differences between males, but a significant difference between females (P <0.01; Fig. 2C) in which female KO mice had shorter latencies to first open-zone entry than female WT mice. For the number of zone crossings, there was no significant genotype or genotype × sex interaction (Fig. 2D).

Fig. 2
Elevated zero maze (5 min). (A) Percent time spent in open quadrants; (B) number of head-dips over the side of open quadrants; (C) latency (s) to first open quadrant entry; (D) number of zone crossings. Number of mice per group (males (M)/females (F)): ...

Open-field locomotor activity

For the 1-h open-field locomotor activity test, there were no significant effects of genotype, but a trend was seen (P <0.09) in which KO mice were slightly more active than WT mice. Sex, interval, and the sex × interval effects were significant but did not interact with genotype. We also analyzed central and peripheral distance (successive beam interruptions converted to cm) using a genotype × region × sex interval ANOVA. There were main effects of gene (F(1,77) = 5.24, P < 0.03), sex (P < 0.01), region (P <0.0001), and interval (P <0.0001). There were also significant interactions of genotype × region (F(1,77) = 6.11, P <0.02), region × sex (F(1,77) = 12.07, P <0.001), sex × interval (F(11,847) = 3.87, P <0.001), region × interval (F(11,847) = 112.95, P <0.0001), region × interval × sex (F(11,847) = 4.36, P <0.0001); the remaining 2-, 3-, and 4-way interactions were not significant. In order to follow-up the genotype × region interaction, separate ANOVAs were conducted for each region.

No significant effects of genotype or genotype interactions with sex or interval on central distance were found (Fig. 3A). With regard to peripheral distance, there was a significant genotype main effect (F(1,77.1) = 5.87, P <0.02) and genotype × interval interaction (F(11,697) = 1.83, P <0.05). Slice-effect ANOVAs for each interval showed differences at multiple intervals and overall (Fig. 3B). KO animals were more active in the periphery than WT controls.

Fig. 3
Locomotor activity (60 min). Number of successive photocell beam interruptions converted to distance (cm) traveled during the 60 min test session in 5-min intervals. (A) Distance traveled in the central zone. (B) Distance traveled in the peripheral zone. ...

ASR with PPI

There were no significant effects of genotype or interaction of genotype with other factors (sex or prepulse) on acoustic startle response amplitude, with or without a prepulse present (not shown), but there was a significant inhibitory effect of prepulse that was intensity-dependent (F(2,152) = 259.34, P <0.0001).

NOR

During the familiarization trial, KO and WT mice showed no significant difference in their preference for one object over the other, i.e., both groups on average spent approximately 50% of their time attending to each of the two identical objects. On the novelty test trial, no significant effect of genotype on novel object preference and no genotype × sex interaction (Fig. 4A) were found, i.e., of the 30 s spent attending to the objects, both groups equally preferred the novel (65%) compared with the familiar (35%) object. However, it was noted that an unusual number of animals failed to complete the task, i.e., failed to accumulate 30 s of exploration of the objects during the test period. We therefore analyzed the proportion of animals failing to complete the test (Fisher’s Exact Test). There was a significant difference (Fig. 4B) in failed trials as a function of genotype (P <0.03): 54.0% of WT mice failed to accumulate 30 s of object time, whereas only 27.9% of KO mice failed to do so.

Fig. 4
Novel object recognition (A) Percent time attending to novel object during the retention interval. (B) Percentage of mice that failed to attend to either object for at least 30 s during the retention interval (time limit = 10 min). Number of mice per ...

MWM, cued platform

Day-1 involved six trials, with the same start and goal positions. The purpose was to train mice to remain on the platform because mice on early trials often jump off the platform and resume swimming the first few times they find it. No significant genotype effect or genotype × trial or sex interaction was seen (Fig. 5A).

Fig. 5
Morris water maze, cued, test. (A) Day-1 latency (s) to the goal: mice were given six trials to reach the platform marked with an orange ball affixed to a 10 cm pole that protruded above the water with the start and goal unchanged on all six trials. ...

Days 2–6 involved two trials per day, with the start and goal randomly positioned for each trial. The purpose was to train the mice to search for the platform using only the cue because no other strategy provided reliable information as to the platform’s location. There was a significant genotype effect (F(1,75.1) = 3.81, P = 0.05; Fig. 5B) in which KO mice, regardless of sex, took less time to reach the platform than WT mice. There were no significant genotype interactions with sex, trial, or the combination of both. There was a sex main effect (P <0.02) because males had shorter latencies than females of both genotypes.

MWM, hidden platform

There were no genotype or interactions with genotype on indices of associative learning (latency, path length, and cumulative distance) to find the hidden platform during acquisition (phase-1) or shift (phase-3) testing (Figs. 6A & C for cumulative distance), i.e., all groups showed typical mouse learning curves. There was no genotype main effect on reversal (phase-2) for path length or latency, but there was a significant genotype × day effect on cumulative distance from the platform [F(5,268) = 2.22, P <0.05]. Genotype × sex slice-effect ANOVAs on each day failed to show a significant genotype or genotype × sex interaction on any single day (Fig. 6B). For swim speed, there were no significant main effects or interactions on any phase of the test suggesting that KO mice had no sensorimotor or motivation deficits that might interfere with their ability to locate the platform.

Fig. 6
Morris water maze, hidden platform, test. Cumulative distance from the starting point to the platform (cm) during the three phases of the test. (A) Acquisition, (B) reversal, and (C) shift. During reversal there was a significant genotype × day ...

MWM, probe trials

We measured average distance to the platform site, number of site-crossovers, swim speed, and percent time and percent distance in the target quadrant. For average distance, there were no significant effects of genotype or interactions on the acquisition, reversal, or shift probe trials. Similarly, there were no genotype effects on probe swim speed, or percent time or percent distance in the target quadrant.

For crossovers (number of times a mouse swam over the spot where the platform had previously been), there were no genotype effects found on the acquisition or shift probe trials, but an effect was observed on the reversal probe trial. Whereas the genotype main effect was not significant, the genotype × sex interaction was (F(1,72) = 4.92, P <0.03; Fig. 7). Slice-effect ANOVA on males showed no significant genotype effect, whereas for females the genotype effect was significant. KO females had more crossovers than WT females (P <0.05), suggesting that they were better than WT females at remembering the location where the platform had been previously.

Fig. 7
Morris water maze, reversal, probe trial. Probe trials were given 24 h after the last platform trial of each phase of MWM testing. For platform-site crossovers, there were no significant effects on the acquisition or shift probe trials, but there was ...

Locomotor activity with drug challenge

During the 30-min re-habituation phase, we found no significant main effect of genotype or genotype × interval, genotype × sex, or genotype × sex × interval.

For the post-challenge activity, there was a significant main effect of genotype (F(1,74.5) = 7.47, P <0.01; Fig. 8) but no genotype interactions with sex, interval, or the combination of all three factors. KO mice showed an exaggerated hyperactivity in response to methamphetamine compared with WT mice. The effect was maximal from 20 to 70 min post-challenge but was statistically significant only when averaged across all intervals (Fig. 8, inset).

Fig. 8
Locomotor activity before and after drug challenge with methamphetamine. Number of horizontal photobeam interruptions per 5-min interval. Mice were re-exposed to the test chambers for 30 min prior to drug administration in order to re-habituate them to ...

Discussion

In the present study we used a test battery to obtain a broad spectrum of possible behavioral effects in a mouse model with chronic oxidative stress from GSH reduction. Because GCLM modifies the regulation of GCLC and therefore regulates GSH levels that serve to buffer ROS, we postulated that Gclm(−/−) KO mice would be at greater risk than WT mice for enhanced CNS damage—given the dramatic decreases in brain GSH (Fig. 1). In contrast to this prediction, KO mice performed within the range of WT mice on most of the tests given and, on a few tests, performed better than WT mice. For example, KO mice displayed better completion rates in the NOR test, had shorter escape latencies on random cued trials in the MWM, and (female) KO mice had more site crossovers on the reversal-probe trial. Compared with WT mice, there were some tests that suggested abnormalities in KO mice: mild hyperactivity in the peripheral zones of the open-field test, diminished anxiety in the EZM, and exaggerated hyperactivity in response to a challenge dose of methamphetamine. These findings represent a moderate behavioral phenotype, which was unexpected—given the degree of GSH depletion in the brain of KO mice.

Our results agree, in some respects, with findings from another study using the Gclm(−/−) mouse (Steullet et al., 2010). This group reported increased oxidative stress, reduced GABAergic signaling and a significant decrease in parvalbumin-immunoreactive interneurons in the CA3 and dentate gyrus of the ventral hippocampus with no damage seen in the dorsal hippocampus. This pathology is consistent with our behavioral phenotype, because it would not affect spatial learning and memory, but could produce behaviors associated with the onset of schizophrenia which has been linked to reduced GSH in humans (Tosic et al. 2006).

Steullet et al. (2010) used a battery of behavioral tests as well. They tested object recognition as we did, except they evaluated both novel-place and novel-object recognition and they used a 10-min retention test interval compared with our 1-h retention test. They found that KO mice explored objects more than WT mice even during the familiarization phases; this finding is somewhat consistent with our finding that more KO than WT mice reached the object-exploration criterion. Similar to our experiment, they found no evidence of impaired novel object preference during retention, suggesting no impairment of object memory.

During appetitive alternation T-maze testing, used as a test of spatial working memory, Steullet et al. (2010) reported no significant differences between KO and WT mice. We did not test for working memory; therefore no comparison of this outcome is possible. They also used the MWM, as we did, to test for spatial learning and they included a reversal-learning phase as we did. Similar to our data, they found no evidence of deficits in MWM cued- or hidden-platform spatial learning on acquisition or reversal phases; Steullet et al. (2010) also showed no evidence of impaired reference memory on the post-acquisition probe trial. Hence, these data are consistent with ours, in that KO mice exhibit no impairment of cued- or hidden-platform performance, including no effects on probe trial tests of memory.

Steullet and coworkers (2010) used several measures of anxiety. One was a novel food suppression test. They found a trend in KO mice to approach food in a novel environment more quickly than WT mice. In another test, they used the elevated plus maze (EPM) and found, as we did in the EZM, that KO mice entered open areas more than WT mice. To further test anxiety, they used the light-dark test and found that KO mice made more zone transitions and spent more time in the light than did WT mice; this is consistent with the EPM and EZM data—indicating that KO mice are less anxious or have increased risk taking compared with WT mice.

Exhibiting less evidence of anxiety may be a positive or negative characteristic. It may be a positive attribute in the context of a preexisting anxious state; however, when an organism is euthymic, exhibiting abnormally low anxiety may be maladaptive and lead to risk-taking or absence of adaptive fear of high risk environments, which for rodents can be open spaces. For rodents, such behavior might place them at greater risk to predation; in humans, lack of appropriate fear is often associated with excessive risk-taking that can lead to dangerous actions that may result in injury to themselves or others.

Steullet et al. (2010) also tested Gclm(−/−) mice for conditioned, as well as approach-avoidance, fear. Day-1 of the procedure was to expose mice to a neutral stimulus (tone) paired with foot shock as conditioning. On day-2 the mice were placed back in the same chamber and assessed for contextual fear. On day-3, mice were placed in a different chamber for an adaptation period with no tone, then tone was presented and freezing behavior recorded. They found that KO mice showed decreased freezing to both the conditioned context and cue—suggesting less conditioning, or reduced fear, compared with WT mice. This is consistent with the lowered anxiety found using the tests of anxiety described above.

More recently, another group reported on the behavioral phenotype of Gclm(−/−) KO mice, also using a behavioral test battery (Cole et al., 2011). They compared KO with WT mice on tests of ASR/PPI, rotorod, automated open-field, and MWM acquisition and reversal. Three of these tests are nearly identical to what we did, viz., ASR/PPI, automated open-field, and MWM. Similar to our study, they found no changes in KO mice on ASR/PPI, open-field, or MWM acquisition, reversal or probe, even though several procedural details differed from our methods. They did not separate open-field activity into central and peripheral sub-regions or use a psychostimulant drug-challenge in the open-field, thus it is unknown if the mice they tested would have similar effects to those observed in this study. They tested their mice on a rotorod for motor coordination (which was not in our battery of tests), and they found no differences.

Cole et al. (2011) did not carry out tests of anxiety or conditioned fear and, hence, did not detect the diminished anxiety or fear-conditioning that we and Steullet et al. (2010) found. On balance, across the two previous behavioral phenotyping studies and the present data, there is a large measure of agreement. All three studies find that Gclm(−/−) KO mice show the following: no evidence of cognitive/learning/memory impairment, no motoric deficits measured as spontaneous locomotion, forced rotorod walking, or changes in swimming ability (indexed as swim speed); however, KO mice were found to display decreased anxiety and fear-associated learning, as well as an altered response to a dopaminergic stimulant.

These experiments—as in the cases of Steullet et al. (2010) and Cole et al. (2011)—were behavioral characterizations and were done without knowledge that these other similar studies were in progress. Ours, by being reported last and without benefit of seeing the findings of the other two, may appear confirmatory, to which we would concur. However, interlaboratory confirmation of a genetic phenotype is important and strengthens the previous observations. In addition, the present data add outcomes not found in the previous data on this mouse model. First, we tested both males and females and showed that for Gclm-deletion, there are no striking sexually dimorphic behavioral differences caused by disruption of this gene, and we found that males and females are similarly affected. More significantly, we found that Gclm KO mice over-respond to a moderate dose of methamphetamine, a drug that predominantly induces dopamine overflow. Methamphetamine in adult animals works through four mechanisms: (1) inhibiting the dopamine transporter (DAT) which blocks DA reuptake, (2) inhibiting VMAT2 which blocks synaptic vesicle reuptake, (3) reversing DAT resulting in DA efflux, and (4) inhibiting MAO metabolism of DA. The fact that Gclm(−/−) mice have a differential response to a drug acting through these mechanisms provides an avenue for follow up experiments to determine if DA is central to how disruption of the Gclm gene affects CNS function.

The findings also offer possible clues to brain antioxidant mechanisms. First, ablation of the Gclm gene lowers brain GSH levels (Fig. 1). If this were the brain’s sole or principal antioxidant control mechanism—one would expect severe neurotoxicity and neurobehavioral deficits. Because this was not observed in three independent studies, loss of GCLM suggests that the brain has redundant antioxidant systems to protect itself from ROS damage, even when the GSH system is compromised. Although brain ascorbate is not compensatorily increased, neither is it decreased in KO mice (Fig. 1); thus, existing ascorbate levels might provide partial protection against GSH loss from ROS damage in a manner similar to that shown in newborn rats (Mãrtensson and Meister, 1991). An alternative possibility for protection against oxidants might be melatonin, which has been shown to protect against oxidative stress particularly in dopaminergic pathways (Ma et al., 2009; Hashimoto et al., 2011). If this is the case, the altered response to methamphetamine suggests that such protection is incomplete.

The data further show that, despite the brain’s resilience to lowered GSH levels, not all neural pathways are protected. Instead, the findings suggest that there are regions that are vulnerable to chronic GSH reduction. Given the changes in anxiety and conditioned fear, that both depend on the amygdala, the data across studies suggest that this region is less capable of buffering the effect of ROS exposure than other regions. Interestingly, methamphetamine depletes GSH in the amygdala of mice (Achat-Mendes et al., 2007) and increases levels of GSH-related proteins in rats (Iwazaki et al., 2008), suggesting an important role for GSH in regulating function in this brain region. The neostriatum, on the other hand, showed a different effect: under basal conditions when the mouse was in a novel environment, no effects in total movement were observed; however, a small and variable increase in movement in the perimeter of the apparatus was seen, but when locomotion was augmented, driven by exposure to the dopaminergic agonist methamphetamine, a much larger difference emerged. This differential response is unlikely to be attributable to nonspecific effects because—even though we did not include a saline-challenged control condition—we previously showed that such a control (saline injection) induces no significant locomotor response (Ehrman et al., 2006).

Human disorders associated with GSH depletion include some that are neurodegenerative, such as amyotrophic lateral sclerosis, Parkinson disease, and Alzheimer disease (Aoyama et al., 2008), pathologies with overt symptoms not found in the Gclm-null mice. Therefore, it may be important to test aged Gclm(−/−) animals to determine whether or not greater impairments are found following life-long GSH depletion. It is also possible that humans with mutations that deplete GSH might be at higher risk to exposure from neurotoxic agents that increase ROS. For example, Gclm(−/−) mice are more susceptible to the pollutant 2,3,7,8-tetrachlorodibenzo-p-dioxin (Chen et al., 2011), and rats having decreased GSH are more susceptible to manganese neurotoxicity (dos Santos et al., 2010).

Further experiments may profit by using the Gclm(−/−) mouse in combination with other models of decreased antioxidant capability, such as the Gulo(−/−) knockout mouse that cannot synthesize the antioxidant ascorbic acid. We have recently found that Gulo(−/−) mice also exhibit a behavioral phenotype that is relatively subtle, yet different in many respects from that seen herein with the Gclm(−/−) mouse (Chen et al., 2012). However, there is one striking commonality between the Gclm(−/−) and Gulo(−/−) KO mice: both show an exaggerated hyperactivity in response to methamphetamine. This similarity suggests that the neostriatum is particularly susceptible to ROS-induced neuronal injury. Given that the striatum is a region of high dopamine abundance, it may be conjectured that dopaminergic terminals are more susceptible to neurotoxic effects than terminals containing other neurotransmitters. Or, it may be that, because dopamine and glutamate interact to facilitate the release of one other, and high glutamate concentrations can cause excitotoxicity, perhaps lowered GSH places other neurotransmitter cell-types at risk—culminating in dopamine receptor up-regulation and increased response to methamphetamine. Crossing the Gclm(−/−) and Gulo(−/−) mouse lines to create a double-knockout mouse could be one way to test this hypothesis, because ablation of both genes may create a mouse with an extreme locomotor response to methamphetamine and perhaps to other pharmacological probes as well. Given the well-known interactions between GSH and ascorbate (Mãrtensson and Meister, 1991), we would also gain additional information concerning these important antioxidant defense systems in the CNS.

Highlights

  • Oxidative stress is associated with brain and behavioral abnormalities
  • We generated mice lacking glutamylcysteinyl ligase modifier GCLM which reduces glutathione (GSH) antioxidant activity
  • KO mice exhibited anxiety changes in the elevated plus maze and increased peripheral zone locomotor activity
  • KO mice had shortened latencies to reach the cued platform in the Morris water maze but were normal on hidden platform trials of spatial learning and memory
  • KO mice showed exaggerated hyperactivity in response to methamphetamine challenge
  • We conclude that some brain regions are resilient low GSH, while others that influence anxiety and locomotion are more susceptible

Acknowledgments

We thank Mary Moran for statistical help. Supported by NIH grants P30-ES006096 (D.W.N.), T32-DK059803 (C.P.C.), T32-ES007051 (C.P.C.), R01-ES008147 (D.W.N.), R01-ES014403 (D.W.N.).

Footnotes

Author Conflict of Interest Statement:

None.

COI statement

The authors of this paper declare no conflicts of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • Achat-Mendes C, Anderson KL, Itzhak Y. Impairment in consolidation of learned place preference following dopaminergic neurotoxicity in mice is ameliorated by N-acetylcysteine but not D1 and D2 dopamine receptor agonists. Neuropsychopharmacology. 2007;32:531–541. [PubMed]
  • Amaro S, Planas AM, Chamorro A. Uric acid administration in patients with acute stroke: a novel approach to neuroprotection. Expert Rev Neurother. 2008;8:259–270. [PubMed]
  • Amato A, Connolly CN, Moss SJ, Smart TG. Modulation of neuronal and recombinant GABA(A) receptors by redox reagents. J Physiol. 1999;517 ( Pt 1):35–50. [PubMed]
  • Aoyama K, Watabe M, Nakaki T. Regulation of neuronal glutathione synthesis. J Pharmacol Sci. 2008;108:227–238. [PubMed]
  • Bailey KR, Rustay NR, Crawley JN. Behavioral phenotyping of transgenic and knockout mice: practical concerns and potential pitfalls. ILAR J. 2006;47:124–131. [PubMed]
  • Brown ST, Scragg JL, Boyle JP, Hudasek K, Peers C, Fearon IM. Hypoxic augmentation of Ca2+ channel currents requires a functional electron transport chain. J Biol Chem. 2005;280:21706–21712. [PubMed]
  • Chen Y, Krishan M, Nebert DW, Shertzer HG. Glutathione-deficient mice are susceptible to TCDD-induced hepatocellular toxicity but resistant to steatosis. Chem Res Toxicol. 2011 E-pub Dec 6 2011. [PubMed]
  • Chen Y, Shertzer HG, Schneider SN, Nebert DW, Dalton TP. Glutamate cysteine ligase catalysis: dependence on ATP and modifier subunit for regulation of tissue glutathione levels. J Biol Chem. 2005;280:33766–33774. [PubMed]
  • Chen Y, Curran CP, Nebert DW, Patel KV, Williams MT, Vorhees CV. Effects of vitamin C deficiency during postnatal development on adult behavior: functional behavioral phenotype of Gulo(−/−) mice. Genes Brain Behav. 2012 (In press) [PubMed]
  • Clark RE, Zola SM, Squire LR. Impaired recognition memory in rats after damage to the hippocampus. J Neurosci. 2000;20:8853–8860. [PubMed]
  • Clarke MW, Burnett JR, Croft KD. Vitamin E in human health and disease. Crit Rev Clin Lab Sci. 2008;45:417–450. [PubMed]
  • Cole TB, Giordano G, Co AL, Mohar I, Kavanagh TJ, Costa LG. Behavioral characterization of Gclm(−/−) knockout mice, a model for enhanced susceptibility to oxidative stress. J Toxicol. 2011;2011:157687. [PMC free article] [PubMed]
  • Crawley JN. Behavioral phenotyping of transgenic and knockout mice. Hoboken, NJ: John Wiley & Sons; 2007. What’s wrong with my mouse?
  • Crawley JN. Behavioral phenotyping strategies for mutant mice. Neuron. 2008;57:809–818. [PubMed]
  • Curran CP, Nebert DW, Genter MB, Patel KV, Schaefer TL, Skelton MR, Williams MT, Vorhees CV. In utero and lactational exposure to PCBs in mice: adult offspring show altered learning and memory depending on Cyp1a2 and Ahr genotypes. Environ Health Perspect. 2011;119:1286–1293. [PMC free article] [PubMed]
  • Dalle-Donne I, Rossi R, Giustarini D, Colombo R, Milzani A. S-glutathionylation in protein redox regulation. Free Radic Biol Med. 2007;43:883–898. [PubMed]
  • Dalton TP, Chen Y, Schneider SN, Nebert DW, Shertzer HG. Genetically altered mice to evaluate glutathione homeostasis in health and disease. Free Radic Biol Med. 2004;37:1511–1526. [PubMed]
  • Dalton TP, Shertzer HG, Puga A. Regulation of gene expression by reactive oxygen. Annu Rev Pharmacol Toxicol. 1999;39:67–101. [PubMed]
  • De Araujo DP, Lobato RF, Cavalcanti JR, Sampaio LR, Araujo PV, Silva MC, Neves KR, Fonteles MM, Sousa FC, Vasconcelos SM. Contributions of antioxidant activity of lipoic acid in reducing neurogenerative progression of Parkinson’s disease: a review. Int J Neurosci. 2011;121:51–57. [PubMed]
  • Dhanasekaran M, Ren J. The emerging role of coenzyme Q-10 in aging, neurodegeneration, cardiovascular disease, cancer and diabetes mellitus. Curr Neurovasc Res. 2005;2:447–459. [PubMed]
  • Diaz D, Krejsa CM, Kavanagh TJ. Expression of glutamate-cysteine ligase during mouse development. Mol Reprod Dev. 2002;62:83–91. [PubMed]
  • Diaz D, Krejsa CM, White CC, Charleston JS, Kavanagh TJ. Effect of methylmercury on glutamate-cysteine ligase expression in the placenta and yolk sac during mouse development. Reprod Toxicol. 2004;19:117–129. [PubMed]
  • dos Santos AP, Milatovic D, Au C, Yin Z, Batoreu MC, Aschner M. Rat brain endothelial cells are a target of manganese toxicity. Brain Res. 2010;1326:152–161. [PMC free article] [PubMed]
  • Ehrman LA, Williams MT, Schaefer TL, Gudelsky GA, Reed TM, Fienberg AA, Greengard P, Vorhees CV. Phosphodiesterase 1B differentially modulates the effects of methamphetamine on locomotor activity and spatial learning through DARPP32-dependent pathways: evidence from PDE1B-DARPP32 double-knockout mice. Genes Brain Behav. 2006;5:540–551. [PubMed]
  • Fonnum F, Lock EA. Contributions of excitotoxicity, glutathione depletion, and DNA repair in chemically-induced injury to neurons: exemplified with toxic effects on cerebellar granule cells. J Neurochem. 2004;88:513–531. [PubMed]
  • Giordano G, Kavanagh TJ, Costa LG. Neurotoxicity of a polybrominated diphenyl ether mixture (DE-71) in mouse neurons and astrocytes is modulated by intracellular glutathione levels. Toxicol Appl Pharmacol. 2008;232:161–168. [PMC free article] [PubMed]
  • Grace CE, Schaefer TL, Herring NR, Graham DL, Skelton MR, Gudelsky GA, Williams MT, Vorhees CV. Effect of a neurotoxic dose regimen of (+)-methamphetamine on behavior, plasma corticosterone, and brain monoamines in adult C57BL/6 mice. Neurotoxicol Teratol. 2010;32:346–355. [PMC free article] [PubMed]
  • Harrison FE, May JM. Vitamin C function in the brain: vital role of the ascorbate transporter SVCT2. Free Radic Biol Med. 2009;46:719–730. [PMC free article] [PubMed]
  • Hashimoto KI, Ueda S, Ehara A, Sakakibara SI, Yoshimoto K, Hirata K. Neuroprotective effects of melatonin on the nigrostriatal dopamine system in the zitter rat. Neurosci Lett. 2011 E-pub Oct 28. [PubMed]
  • Iwazaki T, McGregor IS, Matsumoto I. Protein expression profile in the amygdala of rats with methamphetamine-induced behavioral sensitization. Neurosci Lett. 2008;435:113–119. [PubMed]
  • Ma J, Shaw VE, Mitrofanis J. Does melatonin help save dopaminergic cells in MPTP-treated mice? Parkinsonism Relat Disord. 2009;15:307–314. [PubMed]
  • Mãrtensson J, Meister A. Glutathione deficiency decreases tissue ascorbate levels in newborn rats: ascorbate spares glutathione and protects. Proc Natl Acad Sci U S A. 1991;88:4656–4660. [PubMed]
  • McConnachie LA, Mohar I, Hudson FN, Ware CB, Ladiges WC, Fernandez C, Chatterton-Kirchmeier S, White CC, Pierce RH, Kavanagh TJ. Glutamate-cysteine ligase modifier subunit deficiency and gender as determinants of acetaminophen-induced hepatotoxicity in mice. Toxicol Sci. 2007;99:628–636. [PubMed]
  • McIlwain KL, Merriweather MY, Yuva-Paylor LA, Paylor R. The use of behavioral test batteries: effects of training history. Physiol Behav. 2001;73:705–717. [PubMed]
  • Meister A. Glutathione deficiency produced by inhibition of its synthesis, and its reversal: applications in research and therapy. Pharmacol Ther. 1991;51:155–194. [PubMed]
  • Monks TJ, Ghersi-Egea JF, Philbert M, Cooper AJ, Lock EA. Symposium overview: the role of glutathione in neuroprotection and neurotoxicity. Toxicol Sci. 1999;51:161–177. [PubMed]
  • Obulesu M, Dowlathabad MR, Bramhachari PV. Carotenoids and Alzheimer Disease: an insight into therapeutic role of retinoids in animal models. Neurochem Int. 2011 Jun 7; [Epub ahead of print] [PubMed]
  • Paylor R, Spencer CM, Yuva-Paylor LA, Pieke-Dahl S. The use of behavioral test batteries, II: effect of test interval. Physiol Behav. 2006;87:95–102. [PubMed]
  • Reiter RJ, Manchester LC, Tan DX. Neurotoxicants: free-radical mechanisms and melatonin protection. Curr Neuropharmacol. 2010;8:194–210. [PMC free article] [PubMed]
  • Rice ME, Russo-Menna I. Differential compartmentalization of brain ascorbate and glutathione between neurons and glia. Neuroscience. 1998;82:1213–1223. [PubMed]
  • Ristoff E, Larsson A. Patients with genetic defects in the gamma-glutamyl cycle. Chem Biol Interact. 1998;111–112:113–121. [PubMed]
  • Schaefer TL, Vorhees CV, Williams MT. Mouse plasmacytoma-expressed transcript-1 knockout-induced 5-HT disruption results in a lack of cognitive deficits and an anxiety phenotype complicated by hypoactivity and defensiveness. Neuroscience. 2009;164:1431–1443. [PMC free article] [PubMed]
  • Shepherd JK, Grewal SS, Fletcher A, Bill DJ, Dourish CT. Behavioural and pharmacological characterisation of the elevated “zero-maze” as an animal model of anxiety. Psychopharmacology (Berl) 1994;116:56–64. [PubMed]
  • Skelton MR, Schaefer TL, Graham DL, Degrauw TJ, Clark JF, Williams MT, Vorhees CV. Creatine transporter (CrT; Slc6a8) knockout mouse as a model of human CrT deficiency. PLoS One. 2011;6:e16187. [PMC free article] [PubMed]
  • Steullet P, Cabungcal JH, Kulak A, Kraftsik R, Chen Y, Dalton TP, Cuenod M, Do KQ. Redox dysregulation affects the ventral but not dorsal hippocampus: impairment of parvalbumin neurons, gamma oscillations, and related behaviors. J Neurosci. 2010;30:2547–2558. [PubMed]
  • Sucher NJ, Lipton SA. Redox modulatory site of the NMDA receptor-channel complex: regulation by oxidized glutathione. J Neurosci Res. 1991;30:582–591. [PubMed]
  • Tang XD, Xu R, Reynolds MF, Garcia ML, Heinemann SH, Hoshi T. Haem can bind to and inhibit mammalian calcium-dependent Slo1 BK channels. Nature. 2003;425:531–535. [PubMed]
  • Tosic M, Ott J, Barral S, Bovet P, Deppen P, Gheorghita F, Matthey ML, Parnas J, Preisig M, Saraga M, Solida A, Timm S, Wang AG, Werge T, Cuenod M, Do KQ. Schizophrenia and oxidative stress: glutamate cysteine ligase modifier as a susceptibility gene. Am J Hum Genet. 2006;79:586–592. [PubMed]
  • Trotti D, Nussberger S, Volterra A, Hediger MA. Differential modulation of the uptake currents by redox interconversion of cysteine residues in the human neuronal glutamate transporter EAAC1. Eur J Neurosci. 1997;9:2207–2212. [PubMed]
  • Volterra A, Trotti D, Floridi S, Racagni G. Reactive oxygen species inhibit high-affinity glutamate uptake: molecular mechanism and neuropathological implications. Ann N Y Acad Sci. 1994;738:153–162. [PubMed]
  • Vorhees CV. Design considerations in the use of behavioral test batteries for the detection of CNS dysfunction in laboratory animals. Ment Retard Dev Dis Res Rev. 1996;2:227–233.
  • Vorhees CV, Williams MT. Morris water maze: procedures for assessing spatial and related forms of learning and memory. Nat Protoc. 2006;1:848–858. [PMC free article] [PubMed]
  • Yang Y, Dieter MZ, Chen Y, Shertzer HG, Nebert DW, Dalton TP. Initial characterization of the glutamate-cysteine ligase modifier subunit Gclm(−/−) knockout mouse. Novel model system for a severely compromised oxidative stress response. J Biol Chem. 2002;277:49446–49452. [PubMed]