Glycine Reverses Behavioral Deficits in a Mouse Model for Psychosis With 4 Copies of the Gldc Gene

ABSTRACT

A duplication/triplication copy number variant in the 9p24.1 chromosomal region with the additional gene copies being located on a small supernumerary marker chromosome has been identified in patients with psychosis. Mice genetically engineered to harbor 9p24.1 duplications or triplications have been shown to display schizophrenia‐like phenotypes, including deficits in startle habituation, latent inhibition, working memory, and social interaction and a reduction in the dendritic spine density. Genetic fine‐mapping traced these phenotypes to a duplication or triplication of the Gldc gene, that is, to the presence of three or four functional copies of the Gldc gene. The enzyme glycine decarboxylase (GLDC) degrades glycine, which is a co‐agonist at the NMDA receptor. In mice with 4 copies of Gldc, extracellular glycine concentrations have been reported to be reduced, while total glycine concentrations were unaltered. Here, we tested the hypothesis that chronically administered glycine could revert phenotypic changes observed in mice with 4 copies of Gldc. We found that 1.3 g/kg glycine administered in the drinking water reversed the startle habituation deficit, the spatial working memory deficit in Y‐maze, the sociability deficit and the latent inhibition deficit, while it had a minimal effect on the density of dendritic spines. We conclude that oral administration of glycine is sufficient to reverse some of the behavioral deficits in mice with 4 copies of Gldc but has a very limited effect on dendritic spine density.

Mice with 4 copies of the glycine decarboxylase (Gldc) gene display startle habituation deficits, which are reversed by treatment with glycine. In control‐treated wild type mice (A), the response to the last startle stimuli is lower than the response to the first startle stimuli. In control‐treated mice with 4 copies of the 9p24.1 genes including Gldc (B) or mice with 4 copies of Gldc gene (C), the response to the first and last startle stimuli is indistinguishable. Whereas glycine treatment abolishes startle habituation in wild type mice (A), it restores startle habituation in mice with 4 copies of the 9p24.1 genes (B) and mice with 4 copies of the Gldc gene (C).

Abbreviations

CNV

copy number variant

LCMS

liquid chromatography–mass spectrometry

PFC

prefrontal cortex

1. Introduction

Genetic factors have been shown to increase the risk of developing psychosis in humans. While common genomic variants are responsible for a large number of cases of psychosis with each variant having a small effect size [1], some rare genomic copy number variants (CNV) can confer a much higher relative risk for schizophrenia and bipolar disorder [2]. Two related patients, one with bipolar disorder with psychotic features and the other with schizoaffective disorder, have been found to harbor a small supernumerary marker chromosome, which duplicates 13 genes from the 9p24.1 chromosomal region (for a total of 3 copies) and triplicates (for a total of 4 copies) the GLDC gene encoding the enzyme glycine decarboxylase [3, 4]. GLDC is the rate‐limiting enzyme of the glycine cleavage system, and extra copies of GLDC could potentially result in lower glycine levels in the brain, where glycine is a co‐agonist at the excitatory NMDA receptor [5]. Although this genomic variant is too rare to achieve genome‐wide significance without additional cases, it has the potential to elucidate disease biology by providing a direct link between a structural genomic mutation and pathophysiology, specifically the hypothesis that NMDA receptor hypofunction is an important deficiency in the pathophysiology of schizophrenia [6]. In order to evaluate whether an increase in the copy number of the affected 9p24.1 genes elicits any changes in brain function, mouse lines were generated that in the homozygous state harbor 4 copies of the 9p24.1 genes (9p24.1 CNV mice). These mice displayed behavioral deficits, including startle habituation deficits, latent inhibition deficits, working memory deficits in the Y‐maze and water T‐maze, as well as sociability and social preference deficits in a three‐chamber social interaction test [7]. Subsequently, mouse lines with increased copy numbers of only the Gldc gene and with increased copy numbers of the other 9p24.1 genes (except Gldc) were generated and analyzed. Mice with additional copies of the Gldc gene have been shown to have a phenotype identical to that of the 9p24.1 CNV mice while mice harboring 2 extra copies of the other 9p24.1 genes (excluding Gldc) were behaviorally indistinguishable from wild type mice [7]. The findings suggest that an increase in the copy number of the Gldc gene to 3 copies or 4 copies alone is sufficient to elicit behavioral phenotypes that are consistent with a mouse model for psychosis.

Several lines of genetic evidence, including genome‐wide association studies and exome sequencing studies have suggested a role of the glutamatergic synapse in the pathophysiology of schizophrenia [6, 8, 9]. Dysfunction of glutamatergic neurotransmission via the NMDA receptor has been postulated to be a central contributor to the etiology of schizophrenia [10, 11]. In addition to the binding of the agonist glutamate, activation of the NMDA receptor depends on the binding of glycine or D‐serine at the glycine binding site. Both co‐agonists regulate NMDA receptor function with time and space constraints [12]. For example, in adult animals, in the CA1 subregion of the hippocampus D‐serine is the preferred co‐agonist, and in the dentate gyrus glycine is the preferred co‐agonist [12]. Indeed, we have shown that in mice with 4 copies of Gldc, the extracellular concentration of glycine is reduced in the dentate gyrus but not in CA1, and likewise long‐term potentiation is absent in the dentate gyrus but not in CA1 [7]. A knockout of the gene encoding serine racemase which converts L‐serine to D‐serine resulted in strongly reduced levels of D‐serine, which led to impaired glutamatergic neurotransmission and behavioral deficits [13]. With the administration of D‐serine, some of the phenotypes were rescued [14, 15]. This raises the possibility that the observed deficit caused by additional copies of Gldc could be reversed by glycine administration.

In the patients with psychosis harboring additional copies of 14 9p24.1 genes, including a total of 4 copies of Gldc, adding glycine on top of standard antipsychotic therapy with clozapine led to significant improvements in psychotic and mood symptoms [16]. However, naturally the studies in the patients were limited to the use of questionnaires, and it is unknown whether glycine alone would have similar effects. Moreover, it is not known whether chronic administration of glycine can reverse deficits induced by a structural mutation at an adult age. Testing of glycine alone in the patients, that is, withholding standard antipsychotic treatment, is essentially impossible for ethical reasons. Thus, animal studies are needed to fill this gap in knowledge. In the current study in mice, we aimed to evaluate whether it is possible to reverse at least some of the phenotypes caused by additional copies of the 9p24.1 genes with administration of glycine alone. Moreover, we wanted to evaluate whether glycine administration can reverse the reduced density of dendritic spines that has been observed previously in mice with 4 copies of the Gldc gene. Our hypothesis was that the administration of glycine alone would be sufficient to reverse schizophrenia‐like phenotypes in mice genetically engineered to harbor additional copies of Gldc (either mice with 4 copies of the entire 9p24.1 CNV or mice with 4 copies of Gldc alone). This would be in line with the idea that the phenotypes observed with additional copies of Gldc may arise to a large degree through a deficit in glycine.

2. Methods

2.1. Animals

All experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals of the National Research Council and approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Illinois Urbana‐Champaign. Our studies used mice with 3 genotypes on the C57BL/6J background: mice with 4 copies of the mouse homologs of the 9p24.1 genes including Gldc (4c 9p24.1) (Rln1, Plgrkt, Cd274, Pdcd1lg2, Ric1, Emrp1, Mlana, 9930021J03Rik, Ranbp6, Il33, Trpd52l3, Uhrf2, Gldc); mice with 4 copies of the Gldc gene (4c Gldc); and wild‐type controls. All 4c 9p24.1 mice were bred from homozygous pairings. All 4c Gldc mice and all wild‐type mice except those in behavioral group batch 6 were born and raised by females with 3 copies of Gldc to minimize any potential genotype‐dependent parental effects. Only wild‐type mice from behavioral group batch 6 (Figure 1) were offspring of C57BL/6J mice purchased from the Jackson Laboratory. The genotypes of 4c Gldc mice and wild‐type mice were identified by copy number PCR when needed [7]. All animals were approximately 3 months old at the start of the experiments. Animals from each genotype were randomly assigned to the glycine treatment group or the control group. For behavioral experiments, each treatment group from each genotype included at least eight males and eight females (Figure 1A). 24 additional mice were used for the biochemical analysis of blood serum and brain samples (Figure 1B). Mice were typically group‐housed. Behavioral experiments were done in 6 batches (Figure S1). Similar numbers of male and female mice were used in each experimental group, except for the measurement of dendritic spines, which was only performed in male mice.

FIGURE 1.

Composition of study groups and experiment schedule. (A) Composition of behavioral group and Liquid Chromatography‐Mass Spectrometry (LC–MS) group. In the behavioral group batch 6, 10 wild type (wt) mice in the control treatment group are only used in latent inhibition experiments. (B) Typical experimental schedule for all mice in the behavioral group and LC–MS group. No data collection was done before day 21 of glycine treatment and behavioral experiments were at least 2 days apart. PPI, pre‐pulse inhibition; SIT, social interaction test.

2.2. Glycine Treatment

In a previous clinical study on human patients with the 9p24.1 duplication/triplication, the target dose of glycine of 0.8 g/kg/day could not be maintained due to gastrointestinal symptoms [16]. Since in mice chronic administration of glycine at a dose of 1.3 g/kg/day was reported previously [17], we used this dose and did not observe any sign of side effects. Mice received either their regular drinking water or 1.3 g/kg body weight of glycine (PharmaGrade, Ajinomoto) in their drinking water. Mice were weighed at least once every 5 days, and the drinking volume of the cage was measured at least once every 2 days. The glycine concentration in the drinking water of each cage was adjusted once every 2 days based on the drinking volume and mouse weight to maintain a dose of 1.3 g glycine/kg/day. Previous clinical studies have shown that glycine augmentation of standard treatment with clozapine in human patients with the 9p24.1 CNV may need weeks before maximal effects could be observed [16]. We thus decided to study the effects of glycine after chronic administration for at least 21 days before any data collection for behavioral experiments or sample collection. In addition, mice also received 1.3 g/kg glycine dissolved in water by gavage 3 h before data collection in most behavioral experiments or sample collection, to ensure consistent glycine intake immediately prior to experiments (Figure 1B). The only exception to this rule was that the last three rounds of the Y‐maze experiment in behavioral group batch 6 did not receive water or glycine gavage before the experiment.

2.3. Blood Serum and Brain Sample Collection and Analysis

Blood samples were collected via retroorbital bleeding and let stand for 30 min at room temperature to allow clotting. The clotted blood samples were then centrifuged at 2,000 × g for 10 min at 4°C. Serum was collected from the supernatant. Blood serum samples were collected from all 24 animals in the LCMS group before administration of glycine and stored in a −80°C freezer. Mice then received 1.3 g/kg glycine in the drinking water as described above or standard drinking water for at least 21 days (Figure 1B). After this treatment, mice received 1.3 g/kg glycine or water by gavage. Three hours after the gavage, mice were sacrificed for the collection of serum samples and brain samples. Both hippocampus and PFC were collected and the homogenates were made to have the same wet brain weight per volume. All samples were then sent to the Metabolomics Center at the Roy J. Carver Biotechnology Core of our university on the same date for analysis by LCMS. The following amino acids were measured by LCMS: alanine, arginine, asparagine, aspartic acid, citrulline, cystine, GABA, glutamine, glutamic acid, glycine, histidine, isoleucine, kynurenine, kynurenic acid, leucine, lysine, methionine, ornithine, phenylalanine, proline, serine, taurine, threonine, tryptophan, tyrosine, valine, and in addition caprolactam. Due to the loss of the records on the weight of brain samples collected, we were unable to calculate the absolute concentration of amino acid per wet weight of the brain sample. Thus, relative values are shown.

2.4. Behavioral Experiments

Behavioral experiments followed the general timeline outlined below (Figure 1B).

2.4.1. Pre‐Pulse Inhibition and Habituation of Acoustic Startle

Deficits in pre‐pulse inhibition [18] and startle habituation [19] have been observed in patients with schizophrenia. A Startle Reflex system (Kinder Scientific Inc.) (chamber dimension: 40 cm × 35 cm × 49 cm, not illuminated) was used to evaluate pre‐pulse inhibition of acoustic startle closely following an established protocol [7, 20]. In short, the animals were habituated to the restrainer, the restraint and the chamber in the first 3 days of the experiment (3 × 10 min for restrainer and the restraint, 5 min for the chamber). On day 4, the startle input–output function was measured from 73 dB to 120 dB (approximately 5 min total). On day 5, pre‐pulse inhibition and startle habituation data were collected (approximately 35 min total). Average startle responses were measured from 12 to 120 dB startle pulses scattered throughout the experiment. Percentage pre‐pulse inhibition was calculated as the percentage of reduction of response to the startle pulse with 73–85 dB pre‐pulses 60 ms or 120 ms prior to the startle pulse. Startle habituation data were calculated based on 6 startle pulses that were presented at the beginning and 6 startle pulses presented at the end of the experiment.

2.4.2. Y‐Maze

A Y‐shaped maze (gray, 53 cm × 61 cm, arms: 30 cm × 8 cm × 15 cm) was used to test the spatial memory of the mice. The arena was illuminated from above (100 lx measured at the center of the arena). Each animal was started on the same arm of the Y‐maze and allowed free exploration for 8 min. Entries into each arm were recorded and analyzed. Animals with normal spatial memory will remember the arms previously visited and show a tendency to alternate [21]. Trials with less than three entries were discarded. The percentage of “correct”, that is, alternating arm entries (triads) was calculated for each animal. For behavioral group batch 6, the last three rounds of the Y‐maze were done without oral gavage prior to the experiment and executed by different experimenters, with the last two rounds by the same experimenter.

2.4.3. Latent Inhibition to Conditioned Freezing

Latent inhibition is a phenomenon demonstrating associative learning. It measures the ability to ignore irrelevant stimuli, for example, reduced learning in response to a stimulus in a group receiving preexposure to the stimulus without reinforcement compared to a group without preexposure. Latent inhibition, a process related to selective attention, is attenuated in actively psychotic schizophrenia patients [22]. Based on established protocol [7, 23], we measured latent inhibition to conditioned freezing using chambers from Med‐Associates, Inc. (Chamber dimension: 26 cm × 32 cm × 25 cm, 200 lx when light switched on). Animals in each group defined by genotype and glycine treatment were divided into a preexposure subgroup and a non‐preexposure subgroup. On day 1, animals in the preexposure subgroup were presented with 30 pure tones (20 s, 90 dB, 5000 Hz, with 30 s intervals) while animals in the non‐preexposure subgroup were not. On day 2, for 5 times 80 s apart, all animals received a pure tone (20 s, 90 dB, 5000 Hz) followed by a two‐second 0.7 mA foot shock. On day 3, all animals were presented with 6 min of pure tone (90 dB, 5000 Hz) after 3 min in the chamber. The contextual setups (floor, roof, and odor) of the cages were different for all 3 days to avoid freezing behavior being triggered by anything other than the tone. Freezing behavior to the pure tone on day 3 was analyzed for latent inhibition, calculated as [(freezing duration during the tone)/(tone duration) × 100%]−[(freezing duration before the tone start)/(3 min) × 100%].

2.4.4. Social Interaction Test

A transparent three‐chamber arena (41 cm × 63 cm × 21.5 cm arena divided into three equal chambers, illuminated from above to 100 lx) was used to test the sociability and social preference of mice. Our experiment follows an established protocol [24] used in a previous publication from our lab [7]. Holes are available between each chamber to allow mice free access to all chambers at any time. The arena was placed on a platform surrounded by black curtains to block unintended environmental factors. Each animal was given 5 min to explore the arena with an empty wire cup in both the left and the right chamber. Then, a stranger mouse was placed under one wire cup and the experimental mouse was given 10 min to explore. Finally, another stranger mouse was placed under the other wire cup, and the experimental mouse was given another 10 min to explore. All stranger mice are sex‐matched and of similar age as the experimental mice. Animal movement was recorded from above which was tracked and analyzed using the EthoVision XT 15 software (Noldus Information Technology, BV).

2.4.5. Dendritic Spine Counting

Some animals from the behavioral groups were used for Golgi staining and analysis of dendritic spine after the behavioral experiments. The superGolgi Kit (Bioenno Tech, LLC) was used following the manufacturer's instructions for Golgi staining. In short, mice were perfused with saline and the brain was freshly harvested and processed using solutions in the staining kit according to instructions from the manufacturer. After staining, brains were sectioned with a vibratome into 200 μm slices. Stained brain sections were imaged using a Keyence BZ‐X810 fluorescence microscope with bright field imaging. Z‐stacks were collected every 0.3 μm with a 100× objective lens. Due to our metallic staining, 3D imaging with a confocal microscope was ineffective; thus dendritic spine density was quantified as spine number per length of dendrite instead of spine number per volume. Dendritic spine density was analyzed for non‐primary dendrites in the dentate gyrus using the Reconstruct software closely following a published protocol [25]. In short, a section no shorter than 15 μm of the dendrite was selected, and the length was measured using z‐stack images. The number of dendritic spines was manually counted with z‐stack images, and the spine density was calculated as the number of dendritic spines over the length of dendrite.

2.5. Statistical Analysis

ANOVA (two‐way or three‐way) was used to compare multiple data groups that involve multiple factors [26] followed by Tukey's multiple comparisons test for post hoc analysis. For two‐way ANOVA, the Shapiro–Wilk test was used to test data for normality [27]. For three‐way ANOVA, statistical tests for data normality were not available; thus a QQ plot was visually inspected to check for data normality. Bartlett's test was used to test for equal variance. If the data violate the normality assumption or equal variance assumption of ANOVA, the Scheirer–Ray–Hare (SRH) test followed by Dunn's test was used instead of two‐way ANOVA [28]. The Greenhouse–Geisser correction was applied for the three‐way ANOVA repeated test. For the habituation to acoustic startle and latent inhibition experiments, analysis by ANOVA resulted in non‐significant startle habituation and latent inhibition effects in wild‐type mice under control treatment in Tukey's multiple comparisons test. Additionally, due to the very nature of these experimental paradigms, to assess the behavior in question in these experiments we have to compare two data sets (first versus last startle pulses in the startle habituation experiment, pre‐exposure versus non‐pre‐exposure in the latent inhibition) which makes t‐tests appropriate to assess whether the phenomena of startle habituation and latent inhibition were present in an experimental group, for example, in wild‐type mice without glycine treatment, as in principle, this determination is independent of the results obtained for other groups. Thus, paired t‐test, Welch's t‐test (for unpaired comparisons), Wilcoxon test (for paired data sets that failed the normality test) [29] and Mann–Whitney test (for unpaired data sets that failed the normality test) [29] were used to prevent interference from data in other groups. GraphPad Prism 10 software was used for statistical analysis.

3. Results

3.1. Glycine Levels After Chronic Glycine Treatment

In order to assess the delivery of glycine, we analyzed 48 serum samples, 24 brain samples from the hippocampus, and 24 brain samples from the PFC after oral administration of glycine (Figure 1B). As expected, the glycine level in the serum is significantly increased by the treatment (Figure 2) (SRH test, for factor treatment p = 0.0398). However, the interaction between the factor genotype and the factor treatment was not significant (SRH test, p = 0.902), and a post hoc test did not show a significant difference (Dunn's test, p > 0.999). The relatively limited effect of chronic glycine administration on glycine levels may be due to a high peripheral degradation of glycine in the liver, which is known to have a high level of Gldc expression [30], and potentially to the development of tolerance to glycine. For both the hippocampus and the PFC, glycine levels under control treatment were similar in wild‐type, 4c 9p24.1, and 4c Gldc mice (Figure S2A,B). Somewhat surprisingly, glycine administration did not increase glycine levels in the homogenates of the hippocampus or the PFC in any genotype studied (Figure S2A,B). In any case, glycine levels in homogenates from the brain do not seem to be significantly influenced by an increased copy number of Gldc or by the glycine treatment, suggesting that the behavioral deficits previously observed in mice with 4 copies of the 9p24.1 genes or 4 copies of Gldc and any potential effect of glycine treatment are not facilitated by significant changes in glycine concentrations in homogenates, which likely largely reflect intracellular glycine. However, we have previously shown that extracellular glycine concentrations in the dentate gyrus (but not in CA1) are reduced in 4c Gldc mice [7]. Other than glycine, the LCMS also measured 26 other amino acids (see Materials and Methods); however, only the concentration of kynurenic acid (KYNA) was found to be increased in the hippocampus after glycine treatment (for factor treatment F = 13.66, p = 0.0017, two‐way ANOVA) (Figure S2C). KYNA is a nonselective competitive antagonist at all three ionotropic glutamate receptors with a particularly strong affinity for the glycine modulatory site of the NMDAR, which was found to be lowered in patients with the 9p24.1 CNV previously [16]. Potentially, elevated glycine levels caused by our glycine treatment may have induced elevation of KYNA to balance NMDAR function.

FIGURE 2.

Glycine levels in serum before and after administration of glycine. Glycine levels in serum for each group. (n = 4 for all control or glycine treatment groups, n = 8 for all before treatment groups, SRH test).

3.2. Startle Habituation Deficit Induced by 4 Copies of Gldc Is Rescued by Glycine

Previous studies in our lab have found a small pre‐pulse inhibition deficit in 4c 9p24.1 mice and 4c Gldc mice only at a pre‐pulse intensity of 73 dB (and not at 77 dB, 81 dB or 85 dB) and a startle habituation deficit [7]. Normally, an animal will show a reduced response to a non‐threatening stimulus after repeated presentations of the stimulus, which is referred to as startle habituation. We tested our mice with the intention to evaluate whether glycine treatment could reverse prepulse inhibition and/or habituation deficits. Using pre‐pulse intensities from 73 dB to 85 dB and a pre‐pulse timing of 60 ms or 120 ms prior to the startle pulse, no significant change of pre‐pulse inhibition between genotypes or within a given genotype in the presence or in the absence of glycine treatment was observed (Figure S3). With respect to habituation to startle pulses, although three‐way ANOVA analysis did show a significant interaction between the factor genotype, the factor glycine treatment and the factor habituation effect (F = 5.821, p = 0.004), Tukey's multiple comparisons test did not show a significant difference between startle responses to the first pulses and to the last pulses for wild type control treatment mice (p = 0.0743). To determine whether pairs of data for first pulses versus last pulses in the same animal of an experimental group were significantly different, we performed t‐tests. In the absence of glycine treatment, the wild type mice displayed a significantly lower response to the last 6 startle pulses compared to the first 6 startle pulses (Wilcoxon test, p = 0.0082) (Figure 3A). Interestingly, with glycine treatment, wild type mice showed significantly reduced startle responses to the first 6 startle pulses compared to control treatment (Mann–Whitney t‐test, p = 0.0353), which is comparable to the responses to the last 6 pulses, that is, there is no significant habituation (paired t‐test, t = 1.562, p = 0.1380) (Figure 3A). Both the 4c 9p24.1 and the 4c Gldc genotypes showed no habituation in the absence of glycine treatment (Figure 3B,C), (paired t‐test, t = 1.382, p = 0.1847, for 4c 9p24.1; Wilcoxon test, p = 0.1674 for 4c Gldc), which is consistent with our previous studies [7]. With glycine treatment, the responses to the last startle pulses were significantly lower than the responses to the first startle pulses, that is, the startle habituation deficit induced by the mutant 4c Gldc and 4c 9p24.1 genotypes was reversed by glycine administration (paired t‐test, t = 4.038, p = 0.0007, for 4c 9p24.1; Wilcoxon test, p = 0.012 for 4c Gldc). These results suggest that glycine administration is capable of reversing a behavioral deficit caused by two additional copies of the Gldc gene.

FIGURE 3.

Effect of chronic glycine treatment on habituation of acoustic startle in mice with 4 copies of Gldc. A‐C. Habituation to acoustic startle in (A) wild type mice (n = 19 for control treatment, n = 17 for glycine treatment, paired t‐test, Wilcoxon test, Mann–Whitney test), (B) 4c 9p24.1 mice (n = 18 for control treatment, n = 20 for glycine treatment, paired t‐test), and (C) 4c Gldc mice (n = 18, Wilcoxon test). The bars indicate the startle responses to the first 6 startle pulses (First) and the last 6 startle pulses (Last). *p < 0.05, **p < 0.01, ***p < 0.001.

3.3. Effect of Glycine Treatment on Working Memory in the Y‐Maze

Previously, we reported that in 4c 9p24.1 and 4c Gldc mice the percentage of correct alternations in a Y‐maze is reduced compared to wild type [7]. Initially, we performed the Y‐maze experiment with oral gavage prior to testing, and we did not observe a working memory deficit in 4c 9p24.1 and 4c Gldc mice in a Y‐maze alternation experiment when oral gavage is given before the experiment or 12 days after the last oral gavage (Figure S4). Administration of glycine did not have any significant effect in these experiments (Figure S4) (two‐way ANOVA, for Figure S4A for factor treatment F = 2.71, p = 0.1028, for factor genotype F = 1.854, p = 0.1619, for interaction between factor genotype and factor treatment F = 1.048, p = 0.3544, for Figure S4B for factor genotype F = 0.012, p = 0.9132, for treatment F = 1.72, p = 0.1889, for interaction between factor genotype and factor treatment F = 0.4968, p = 0.6113). We hypothesized that potentially, the stress of repeated gavage procedures might interfere with the Y‐maze performance of our mice and a long resting period after the most recent gavage may be needed to restore normal performance in wild type mice and a genotypic difference between wild type mice and 4c Gldc mice under control treatment. When the Y‐maze experiment was done 24 days after the last oral gavage, we were able to replicate the genotypic difference previously observed (Figure 4) (two‐way ANOVA, Tukey's multiple comparisons test, p < 0.0001 for wt control treatment versus 4c Gldc control treatment). Chronic glycine treatment was able to significantly improve the alternation percentage of 4c Gldc mice (Tukey's multiple comparisons test, p = 0.0017) to a level that is still significantly lower than that for the wild type mice with control treatment (Tukey's multiple comparisons test, p < 0.0001). Notably, the alternation percentage for the wild type mice is significantly lower after glycine treatment (Tukey's multiple comparisons test, p = 0.0002). With wild type mice having a reduced alternation percentage after glycine treatment, the alternation percentage of the 4c Gldc mice after glycine treatment is not significantly different from that of wild type mice after glycine treatment (Tukey's multiple comparisons test, p = 0.9601). Overall, the glycine treatment at least partially reversed the working memory deficit in the Y‐maze in 4c Gldc mice and induced a working memory deficit in wild type mice (two‐way ANOVA, for factor gene F = 46.94, p < 0.0001, for interaction between factor gene and factor treatment F = 37.99, p < 0.0001).

FIGURE 4.

Effect of chronic glycine treatment on spatial working memory in the Y‐maze in mice with 4 copies of Gldc. The percentage of correct alternations is shown for the Y‐maze experiments in behavioral group batch 6 on day 60 and 87 that did not receive oral gavage before the experiment (n = 10 per group, two‐way ANOVA, Tukey's multiple comparisons test). **p < 0.01, ***p < 0.001, ****p < 0.0001.

3.4. Deficit in Latent Inhibition to Conditioned Freezing Is Reversed by Glycine Treatment

Wild type mice display latent inhibition to conditioned freezing, that is, preexposure to a tone reduces the freezing in response to the tone after fear conditioning to the tone. We previously detected a latent inhibition deficit in 4c 9p24.1 and 4c Gldc mice [7]. We wanted to assess whether glycine treatment can restore this deficit. With three‐way ANOVA, the wild type control treatment group did not show a significant difference between the non‐pre‐exposed (nPE) group and the pre‐exposed (PE) group (three‐way ANOVA, Tukey's multiple comparisons test, p = 0.306). To compare relevant PE and nPE data pairs, Welch's t‐test was used. With control treatment, wild type mice show a significant latent inhibition effect (Figure 5) (p = 0.0169, Welch's t‐test). For 4c 9p24.1 mice and 4c Gldc mice, consistent with our previous studies [7], nPE and PE groups not treated with glycine showed a similar level of freezing, that is, they displayed a latent inhibition deficit (Figure 5) (p = 0.184 for 4c 9p24.1 mice, p = 0.2358 for 4c Gldc mice, Welch's t‐test). With glycine treatment, while 4c 9p24.1 mice still displayed no significant difference between nPE and PE groups (p = 0.0926, Welch's t‐test), 4c Gldc mice showed significantly less freezing in the PE group compared to the nPE group (p = 0.0004, Welch's t‐test), that is, a strong latent inhibition effect (Figure 5). Unexpectedly, in the wild type mice the latent inhibition effect was no longer significant after glycine treatment (p = 0.0944, Welch's t‐test). When we combined the data for the 4c 9p24.1 mice and the 4c Gldc mice, analysis with two‐way ANOVA showed that glycine treatment reversed the latent inhibition deficit in the mutant mice with an increased copy number of the Gldc gene (Figure S5) (two‐way ANOVA, for interaction F = 2.352, p = 0.1292, Tukey's multiple comparisons test, p = 0.2525 for control treatment, p = 0.0001 for glycine treatment). In summary, our data indicate that glycine is able to reverse the latent inhibition deficit caused by two extra copies of the Gldc gene (Figure 5).

FIGURE 5.

Effect of chronic glycine treatment on latent inhibition to conditioned freezing in mice with 4 copies of Gldc. Percentage freezing difference between the 6 min in the presence of the tone and the preceding 3 min in the absence of the tone for all experimental groups. (n = 15 for wt control treatment pre‐exposed (PE), n = 14 for wt control treatment non‐pre‐exposed (nPE), n = 13 for wt glycine treatment pre‐exposed, n = 14 for wt glycine treatment non‐pre‐exposed, n = 8 for 4c 9p24.1 control treatment pre‐exposed, n = 9 for 4c 9p24.1 control treatment non‐pre‐exposed, n = 10 for 4c 9p24.1 glycine treatment pre‐exposed, n = 9 for 4c 9p24.1 glycine treatment non‐pre‐exposed, n = 10 for 4c Gldc control treatment pre‐exposed, n = 8 for 4c Gldc control treatment non‐pre‐exposed, n = 15 for 4c Gldc glycine treatment pre‐exposed, n = 13 for 4c Gldc glycine treatment non‐pre‐exposed, Welch's t‐test). *p < 0.05, ***p < 0.001.

3.5. Glycine Treatment Reverses Sociability Deficit in 4c Gldc Mice

Previous studies in our lab have identified a sociability deficit and a social preference deficit in mice with 4 copies of Gldc [7]. Our current experiment was designed to test the effect of chronic glycine treatment on both deficits. In the habituation phase, all groups had no preference for one of the lateral compartments of the three‐chamber apparatus (Figure S6A) (three‐way repeated measures ANOVA, for factor side of chamber F = 0.0462, p = 0.8322). 4c Gldc control treatment mice displayed no difference in interaction time between a stranger mouse and the empty restrainer, that is, a deficit in sociability (three‐way repeated measures ANOVA, Tukey's multiple comparisons test, p = 0.2015); in contrast, the wildtype control treatment mice preferred the stranger mouse over the empty restrainer (Tukey's multiple comparisons test, p = 0.0404) (Figure 6). After glycine treatment, the wildtype mice no longer showed preference for the stranger animal (Tukey's multiple comparisons test, p = 0.9074) but the sociability deficit in the 4c Gldc mice was reversed (Tukey's multiple comparisons test, p = 0.0136) (Figure 6). Overall, the data show an effect of chronic glycine treatment but the interaction between genotype and treatment was not enough to be statistically significant (three‐way ANOVA, for factor treatment, F = 8.081, p = 0.0108; for interaction between factor treatment, factor genotype and factor side of the chamber F = 2.667, p = 0.1198). In the social preference test, the wild type control treatment group did not show a preference for the novel stranger mice (repeated measure three‐way ANOVA, Tukey's multiple comparisons test, p = 0.9986), obviating further analysis (Figure S6B).

FIGURE 6.

Effect of chronic glycine treatment on sociability. Bar graph showing comparison of time spent in chamber with stranger mice under restrainer and empty restrainer. (n = 10 per group, three‐way ANOVA repeated measures, Tukey's multiple comparisons test). *p < 0.05.

3.6. Glycine Treatment Partially Relieves Dendritic Spine Density Deficit in Dentate Gyrus of 4c Gldc Mice

Mice with 4 copies of Gldc have been shown to have a reduced density of dendritic spines in the dentate gyrus [7]. We investigated whether glycine treatment could reverse this decrease in 4c Gldc mice. Consistent with our previous observations, 4c Gldc mice showed a reduced dendritic spine density in the dentate gyrus for both neuron average (two‐way ANOVA, for factor genotype F = 13.19, p = 0.0007, for factor treatment F = 0.4226, p = 0.5191, for interaction F = 0.07046, p = 0.7919, Tukey's multiple comparisons test, p = 0.0441 for wt control treatment versus 4c Gldc control treatment) and branch average (two‐way ANOVA, for factor genotype F = 19.46, p < 0.0001, for factor treatment F = 0.6238, p = 0.4317, for interaction F = 0.104, p = 0.7478, Tukey's multiple comparisons test, p = 0.0072 for wt control treatment versus 4c Gldc control treatment) (Figure 7). Glycine treatment did not significantly increase the dendritic spine density of either wild type mice or 4c Gldc mice (Figure 7). However, the dendritic spine density in the 4c Gldc mice treated with glycine was no longer significantly different from that of the wild type mice in the absence of glycine treatment (Tukey's multiple comparisons test, neuron average: p = 0.1743, branch average: p = 0.0616) and for neuron average, 4c Gldc mice after glycine treatment had similar dendritic spine density as wild type mice after glycine treatment (Tukey's multiple comparisons test, p = 0.0909) (Figure 7). Thus, while glycine treatment does not result in a clear increase of dendritic spine density in 4c Gldc mice, it may have some limited effects on dendritic spine density and partially relieve the dendritic spine density deficit in the 4c Gldc mice.

FIGURE 7.

Effect of glycine treatment on dendritic spine density in the dentate gyrus in mice with 4 copies of Gldc. Dendritic spine density of non‐primary dendrites in the dentate gyrus in 4c Gldc mice for each branch (A) (n = 22 for wt control treatment, n = 24 for all other groups, two‐way ANOVA, Tukey's multiple comparisons test) or each neuron (B) (n = 11 for wt control treatment, n = 12 for all other groups, two‐way ANOVA, Tukey's multiple comparisons test). *p < 0.05, **p < 0.01.

4. Discussion

This study has shown that monotherapy with glycine in mice that have additional copies of the Gldc gene can reverse startle habituation, latent inhibition, working memory, and social interaction deficits, while the effects on dendritic spine density in the dentate gyrus are very discrete and their significance is therefore less clear. A previous study in patients with a 9p24.1 CNV that results in 4 copies of GLDC has demonstrated that orally administered glycine, given in addition to standard antipsychotic treatment with clozapine, can reduce psychotic and mood symptoms but not cognitive symptoms [16]. The current study revealed that deficits in mice with 4 copies of Gldc (either 4 copies of the entire 9p24.1 chromosomal region including Gldc or 4 copies of Gldc alone) can be reversed at least partially by treatment with glycine in the absence of clozapine. Notably, despite the 1.3 g/kg dose of glycine administered to mice being higher than the target dose of 0.8 g/kg or the tolerable dose for the human patients [16], we did not detect an increase in glycine concentrations in homogenates from the hippocampus or PFC after glycine administration (Figure S2A,B). Another study has shown increased cortical glycine levels in rats that received a single dose of 2 g/kg glycine [31]. It is noteworthy that in our studies glycine also had unexpected effects in the wild type mice: reduction of the startle response, of the percentage of alternations in the Y‐maze, and of sociability in the social interaction test. These effects could be considered indicative of mild toxicity of chronic administration of 1.3 g/kg glycine in mice.

NMDA receptor activation is regulated by glial‐neuronal cross‐talk. Astrocytes take up glucose from the blood and convert it to L‐serine, which is transported out of the astrocytes and into the neurons, where it is converted to the co‐agonist D‐serine. In astrocytes, L‐serine is also converted to the co‐agonist glycine. According to a current model, D‐serine is released primarily by neurons and glycine by astrocytes [32]. Interestingly, in the brain, Gldc is expressed mostly in astrocytes as part of the glycine cleavage system and plays an important role in the degradation of glycine [33, 34, 35]. Thus, it is conceivable that patients or mice with additional copies of Gldc might degrade glycine faster or might degrade more glycine, and that chronic glycine treatment would elevate the availability of glycine as a co‐agonist at the NMDA receptor, reversing deficits induced by the additional copies of the gene encoding this enzyme [4, 7].

Somewhat surprisingly, our LCMS data did not show a significant difference in glycine levels between genotypes or after glycine administration in brain tissue homogenates. It is reasonable to assume that most of the glycine detected in such homogenates represents intracellular glycine. This indicates that there is not a dramatic decline in intracellular glycine levels as one might have predicted with potentially increased glycine degradation by 4 copies of Gldc. A mouse line with a heterozygous knockout of the alanine‐serine‐cysteine‐1 (Asc‐1) transporter which controls glycine levels in the brain has been reported to display a 50% reduction in glycine levels in the brain [36]. The Asc‐1 heterozygous KO mouse line displays neurological complications (hyperekplexia) that have not been observed in the patients or in the mice with extra copies of the Gldc gene. Thus, it is possible that Gldc‐containing CNVs may have much more restricted effects on local and/or extra‐ or subcellular glycine levels or availability that are more relevant for modulation of NMDA receptors and related behavioral phenotypes. Indeed, using a fluorescent glycine sensor, it was found that in the dentate gyrus, but not in the CA1 hippocampal subregion, the level of extracellular glycine is reduced in 4c Gldc mice [7]. In line with these observations, LTP was reduced in the dentate gyrus, but not in CA1, of mice with 4 copies of Gldc [7]. While we have not determined extracellular glycine levels in this study, our findings are consistent with the notion that extracellular glycine deficits can be reversed with exogenous glycine.

It is noteworthy that kynurenic acid (KYNA) levels were increased in hippocampal homogenates of mice treated with glycine compared to mice not treated with glycine (Figure S2). The kynurenine pathway of tryptophan degradation is known to impact NMDA receptor function [37]. KYNA is a competitive, broad‐spectrum antagonist of glutamate receptors at high micromolar concentrations, with the greatest affinity for the glycine binding site of the NMDA receptor [38]. The significance of the increase in KYNA levels after glycine treatment is currently unclear. Potentially, increased KYNA levels after glycine treatment could, at least in part, reduce the binding of glycine to the glycine site of the NMDA receptor, and limit the effects of glycine. KYNA has also been found to be a non‐competitive inhibitor at the α7 nicotinic acetylcholine receptor (α7nAChR) [38]. The CHRNA7 gene is one of six genes in the 15q13.3 microdeletion that has been associated with an increased risk of developing schizophrenia [39, 40], and a positive allosteric modulator of the α7nAChR has been shown to rescue a schizophrenia‐associated brain endophenotype in mice with a 15q13.3 microdeletion [41]. Thus, one might speculate whether increased KYNA levels after glycine treatment may potentially have negative effects on cognition. More research would be needed to address this question.

Psychiatric disorders like schizophrenia and bipolar disorder are complex diseases at several levels. Not only have dozens and potentially hundreds of genes been found to contribute to the risk of developing these disorders [8], but it is also known that disease‐related deficits may develop in multiple biochemical pathways. Although having extra copies of Gldc appears to be a simple pathway‐specific change, we have previously shown that the activity of multiple biochemical pathways is altered in the 4c Gldc mice [7], and such changes would be expected to play a role in the manifestation of observed behavioral and other phenotypes. One such change is decreased mitochondrial respiration in the dentate gyrus in 4c Gldc mice [7]. Given that Gldc is found in mitochondrial membranes [33], metabolic pathways in mitochondria could be an important target for future investigations of the pathophysiology of psychosis and for the development of novel treatments.

The behavioral deficits and the reversal of those deficits by glycine could also provide valuable clues on the contribution of specific brain regions to disease pathology. For example, the latent inhibition effect has been shown to be dependent on the activity of the dentate gyrus [23], and our results on latent inhibition may thus reflect a relative glycine deficit and its potential reversal in the dentate gyrus. Moreover, as in the brain, Gldc is mostly expressed in astrocytes [33], manipulation of astrocytic function could be a valuable approach to investigate the etiology of pathophysiological changes in our 4c Gldc animal model. It will be interesting to study whether astrocytes in specific brain regions, e.g., in the dentate gyrus of the hippocampus contribute to behavioral functions altered in the mice with 4 copies of Gldc. If this is the case, new therapeutic strategies for the treatment of psychosis may be developed that would primarily target astrocytes.

In summary, our study provides evidence that monotherapy with glycine is sufficient to reverse schizophrenia‐like phenotypes in a mouse model in which an increased rate of glycine catabolism is the primary genetically induced deficit. To our knowledge, this is one of the rare examples where schizophrenia‐like deficits can be reversed in a mouse model by specifically targeting the underlying genetic deficit responsible for these phenotypes pharmacologically. While our results may be viewed as supporting the use of glycine monotherapy in patients with 4 copies of GLDC or in patients with other glycine‐related deficits, our studies caution that glycine, at the dose that we used, has unwanted toxic effects, as in some behavioral paradigms (startle habituation, latent inhibition, sociability, Y‐maze) its actions were opposite to the ones that it has in mice with 4 copies of Gldc.

Author Contributions

Muxiao Wang: conceptualization, formal analysis, investigation, methodology, writing – original draft, writing – review and editing. Maltesh Kambali: formal analysis, investigation, methodology, writing – review and editing. Jinrui Lyu: investigation, methodology, writing – review and editing. Rajasekar Nagarajan: formal analysis, investigation, methodology, writing – review and editing. Uwe Rudolph: conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, writing – review and editing.

Funding

This work was supported by the National Institute of Mental Health of the National Institutes of Health by award number R56MH11264 to U.R. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Ethics Statement

Animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Illinois Urbana‐Champaign.

Conflicts of Interest

U.R. served on the Scientific Advisory Board of Damona Pharmaceuticals. The other authors report no competing interests.

Supporting information

Figure S1: Animal ID and batch assignments for behavioral experiments.

Figure S2: Normalized glycine levels in hippocampal and prefrontal cortex (PFC) homogenates and normalized kynurenic acid (KYNA) levels in hippocampal homogenates without and with chronic administration of oral glycine.

Figure S3: Effect of glycine treatment on prepulse inhibition of acoustic startle in mice with 4 copies of Gldc.

Figure S4: Effect of chronic glycine treatment with oral gavage on spatial working memory in the Y‐maze in mice with 4 copies of Gldc.

Figure S5: Combined analysis of latent inhibition to conditioned freezing in 4c 9p24.1 mice and 4c Gldc mice.

Figure S6: Side preference in the habituation phase of the social interaction test and effect of chronic glycine treatment on social preference.

Wang M., Kambali M., Lyu J., Nagarajan R., and Rudolph U., “Glycine Reverses Behavioral Deficits in a Mouse Model for Psychosis With 4 Copies of the Gldc Gene,” Pharmacology Research & Perspectives 13, no. 6 (2025): e70202, 10.1002/prp2.70202.

Data Availability Statement

Primary data are available in the Harvard Dataverse repository (https://dataverse.harvard.edu/api/access/datafile/13198294).