Integrating 16S rRNA Gene Sequencing and Metabolomics Analysis to Reveal the Mechanism of L-Proline in Preventing Autism-like Behavior in Mice

All experimental procedures were approved by the Laboratory Animal Welfare and Animal Experimental Ethical Committee of China Agricultural University (Approval Number: AW30803202-4-5, Beijing, China). Fourteen-week-old C57BL/6J female and male mice were purchased from Vital River Laboratories (Beijing, China) and housed in specific pathogen-free conditions at the Animal Centre (SYXK (Jing) 2020-0052), maintained at a relative humidity of 40–70%, a temperature of 22 ± 2 °C, and a 12-h light-dark cycle.

Adult male and female C57BL/6J mice were bred using a monogamous breeding scheme at night, with pregnancy confirmed by the presence of vaginal plugs the following morning, designated as embryonic day 0.5 (E0.5). Pregnant mice were individually housed and randomized into three groups: SAL, VPA, and PRO-VPA groups. Mice in the SAL and VPA groups, along with their offspring, were provided with sterile water, whereas those in the PRO-VPA group received sterile water supplemented with 1% (w/w) L-proline (99% purity, Sigma). On embryonic day 12.5 (E12.5), pregnant mice in the VPA and PRO-VPA groups were intraperitoneally injected (i.p.) with 600 mg/kg of VPA (98% purity, Sigma, St. Louis, MO, USA), while mice in the SAL group received an equivalent volume of 0.9% saline. Male offspring were weaned at postnatal day 21, resulting in 15 mice in the postnatal SAL group, 19 mice in the postnatal VPA group, and 11 mice in the postnatal PRO-VPA group. All animals had ad libitum access to food and water during the experimental period.

At 18 weeks of age, the offspring mice underwent behavioral testing in the following sequence: open-field test, self-grooming test, and three-chambered social test. Following completion of the behavioral assessments, the mice were euthanized at 23 weeks of age after a 6-h fasting period (from 8:00 a.m. to 2:00 p.m.). Euthanasia was conducted under isoflurane anesthesia via cervical dislocation in accordance with humane protocols. Subsequently, cecal content samples were collected and stored at −80 °C for subsequent analysis. Figure 1A illustrates a detailed timeline of the animal experimental procedures.

A NanoDrop 2000 spectrophotometer manufactured by Thermo Fisher Scientific (Wilmington, DE, USA) was utilized to quantify the purity and concentration of the total DNA extracted from cecal content samples. Subsequently, the 16S rRNA gene segments corresponding to the V3–V4 regions were amplified from the extracted DNA utilizing the primers 806R (5′-GGACTACHVGGGTWTCTAAT-3′) and 338F (5′-ACTCCTACGGGAGGCAGCAG-3′). Amplification was achieved through the execution of 27 cycles of polymerase chain reaction, with each cycle consisting of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 45 s. For paired-end sequencing of the amplified products, the Illumina MiSeq sequencing platform (Illumina, San Diego, CA, USA), along with PE300 reagents provided by Majorbio Biomedical Technology Co. (Shanghai, China), was employed.

The raw data were processed on the Majorbio Cloud platform (https://cloud.majorbio.com/↗, Shanghai, China). Following demultiplexing, the sequences were merged using FLASH (version 1.2.11) and subsequently underwent quality filtering with fastp (version 0.19.6). High-quality sequences were then denoised utilizing the DADA2 plugin within the QIIME2 pipeline (version 2020.2). The resultant sequences, termed amplicon sequence variants (ASVs), represent the conventional nomenclature for DADA2-denoised sequences. For ASV classification, the built-in Naive Bayes consensus classifier within QIIME2 was utilized, with the SILVA 16S rRNA database (version 138) serving as the reference for ASV categorization.

The α diversity indices at the ASV level were calculated using the QIIME2 software (version 2020.2). Statistical significance was set at p < 0.05, with false-discovery rate (FDR)-adjusted p-values calculated using Welch’s correction to account for potential biases. Principal coordinates analysis (PCoA) visualizations were generated on the Majorbio Cloud platform (https://cloud.majorbio.com/↗ accessed on 11 December 2024, Shanghai, China), followed by a weighted UniFrac distance-based analysis of similarity (ANOSIM) to assess clustering patterns. Differences in bacterial taxa, ranging from the phylum to the genus level, were evaluated using a linear discriminant analysis (LDA) coupled with an LDA effect size (LEfSe) analysis, applying two filters: p < 0.05 and LDA score > 4. Given the potential for high false-positive rates associated with LEfSe, as previously reported [19], an additional analysis of the composition of microbiomes (ANCOM) was conducted using the ANCOM 4.0.2 package in R to refine the identification of bacterial taxa enriched across sample groups. The significance of ANCOM was defined as W > 0.6. Finally, PICRUSt2 was employed to predict the functional capabilities of the gut microbial communities.

Shanghai Majorbio Bio-pharm Technology Co., Ltd. conducted an untargeted metabolomic analysis of cecal content samples. For metabolite extraction, a solution consisting of methanol and water in a volume ratio of 2:1 (v/v), containing internal standards, was employed. Each sample was supplemented with 400 μL of the extraction solvent and subsequently milled in a grinder for a duration of 6 min at −10 °C with a frequency of 50 Hz. This was followed by a low-temperature ultrasonic extraction step for 30 min at 5 °C with an ultrasonic frequency of 40 KHz. Subsequently, the samples were allowed to equilibrate at −20 °C for 30 min, followed by centrifugation at 13,000× g for 15 min at 4 °C. The resultant supernatant was then pipetted for further analytical procedures.

The instrumental platform utilized for the untargeted metabolomics analysis in this study was the Ultra High Performance Liquid Chromatography Tandem Fourier Transform Mass Spectrometry UHPLC-Exploris 240 system (Thermo Fisher, USA). For liquid chromatography (LC) separation, an ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 µm; manufactured by Waters, Milford, CT, USA) was employed. Mobile phase A comprised 95% water and 5% acetonitrile (supplemented with 0.1% formic acid), while mobile phase B consisted of 47.5% acetonitrile, 47.5% isopropanol, and 5% water (also containing 0.1% formic acid). The injection volume was set to 3 μL, and the column temperature was maintained at 40 °C. The samples were ionized through electrospray, and the mass spectrometry signals were acquired in both positive (ES+) and negative (ES−) ion scanning modes.

The raw data were processed utilizing Progenesis QI v3.0 software (Waters Corporation, Milford, CT, USA). Subsequent to normalization based on total peak intensity, the processed data were uploaded and imported into the ropls R package (version 1.6.2), where they underwent multivariate data analysis. This included partial least squares discriminant analysis (PLS-DA) to investigate alterations in metabolic profiles across different groups. The Student’s t-test was employed to determine the p between pairs of metabolite groups. Metabolites were screened based on a variable importance in projection (VIP) > 1.0 and a p < 0.05. The Human Metabolome Database (HMDB) was utilized for annotating the identified differential metabolites. Differential metabolites between the VPA and PRO-VPA groups were imported into the Kyoto Encyclopedia of Genes and Genomes (KEGG) database for further analysis of their functions and metabolic pathways. Only pathways with a p < 0.05 were considered as significantly altered.

The sample size of the litters was decided from previous works on the VPA model in this mouse strain and behavioral analyses [20] and all the male offspring from different litters were selected for the behavioral analyses. The determination of the number of animals for the molecular analyses was guided by the previous, similar literature, in which sample sizes of n = 3–5 are typically employed for the analysis of gut microbiota and metabolomes [16,21,22]. No statistical methods were used to predetermine the sample size. No outliers were considered, and all collected data were used in statistical analyses.

The data were presented as the mean ± standard error of the mean (SEM). Comparisons involving a single variable were performed using an unpaired, two-tailed Student’s t-test. Differences among multiple groups with a single variable were assessed using a one-way analysis of variance (ANOVA). All statistical analyses were conducted utilizing GraphPad Prism 9.5 software (GraphPad, San Diego, CA, USA). The level of significance was denoted as follows: * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.

The preventive impact of L-proline on behavioral disorders reported in the prenatal VPA-exposed male mice was investigated by assessing locomotor activity, repetitive stereotypical behavior, and social dysfunction. Initially, an open-field test was employed to evaluate the effect of L-proline supplementation on locomotor activity. In this test, the VPA-exposed mice exhibited significantly decreased total distances traveled, travel speed, and activity time (Figure 1B–E) compared to mice in the SAL control group, aligning with previous findings [23,24]. Notably, L-proline supplementation significantly prevented the abnormal reduction in total distance traveled and speed observed in the open-field test among mice with ASD characteristics (Figure 1B–E). ASD patients often display repetitive stereotypical behaviors and social impairments. Intriguingly, our results indicated that the VPA-treated mice spent significantly more time on self-grooming compared to the SAL group, an effect that was substantially prevented by L-proline supplementation (Figure 1G,H).

Given that social dysfunction is a prominent characteristic of individuals with ASD and that mice exposed to VPA exhibit impaired social behavior, a three-chamber social test was conducted to investigate the sociability and preference for social novelty of these mice. To assess sociability, the time spent interacting with the stranger mice 1 (S1) versus an empty wired cage (E) was evaluated. As anticipated, mice in the SAL control group displayed normal social behavior, evident by their preference for interacting with the stranger mice. Conversely, the VPA-treated mice did not exhibit a preference for the stranger mice over empty cages, indicating impaired sociability, which was ameliorated by L-proline supplementation (Figure 1J,K). Additionally, the VPA-treated mice failed to demonstrate a normal preference for social novelty. Dietary L-proline restored social novelty preferences in the VPA-treated mice, with the performance of mice in the PRO-VPA group being comparable to that of the SAL group (Figure 1L,M).

Consequently, these findings demonstrate that dietary L-proline significantly prevented ASD-like behavioral disorders in the VPA-induced ASD mouse model, as evidenced by improvements in social communication deficits and reductions in repetitive behavior.

Recent experimental studies conducted on mice have provided evidence that gut microbiota can play a modulatory role in brain-centered disorders, including ASD [25]. Additionally, dietary L-proline has been identified as a potential regulator of both the gut microbiome and metabolome [16]. To investigate whether L-proline supplementation could alter the gut microbial composition in mice treated with VPA, we conducted an analysis of microbial communities using 16S rRNA sequencing. Initially, we assessed the impact of L-proline supplementation on the α diversity and β diversity of the gut microbiota in a VPA-induced ASD mouse model.

The α diversity indices were evaluated at the ASV level. Our findings revealed that community diversity was significantly elevated in the VPA group compared to the SAL control group, as indicated by a higher Shannon index and a lower Simpson index (Figure 2A,B). However, the VPA treatment did not significantly affect the community coverage (assessed by the Coverage index, Figure 2C) or the community richness (assessed by the Sobs index, Figure 2D) of the gut microbiota. Furthermore, L-proline supplementation did not exert a statistically significant effect on α diversity in the ASD mouse model (Figure 2A–D). PCoA was utilized to examine the community structures of the gut microbiota. The results demonstrated that the gut microbiota of mice in the VPA group was distinct from that of the SAL group (R = 0.8120, p = 0.004, Figure 2E). Moreover, the gut microbiota of mice in the PRO-VPA group was separated from that of the VPA group (R = 0.8000, p = 0.004, Figure 2F), suggesting that L-proline supplementation significantly modulated the gut microbial community structure in the VPA-treated mice.

The gut microbiome health index (GMHI) was employed as a metric to assess the health status of the microbiota [26]. Extending this observation, we further discovered that the GMHI was decreased in mice from the VPA group compared to those in the SAL group (Figure 2G). Notably, the L-proline treatment was found to significantly elevate the GMHI in the ASD mouse model (Figure 2H). Additionally, the microbial dysbiosis index (MDI) was calculated to quantify the degree of microbial imbalance. The results indicated a significant increase in MDI in the ASD mice compared to those in the SAL group (Figure 2I). Concurrently, L-proline supplementation significantly decreased the MDI in the VPA-induced ASD mouse model (Figure 2J). Collectively, these findings suggest that the L-proline treatment can markedly enhance the health status of the microbiota and ameliorate the degree of gut microbiota dysbiosis.

LDA was integrated with LEfSe to identify the ASVs that were primarily responsible for the compositional differences observed in the gut microbial communities. A comprehensive analysis yielded a total of 38 bacterial clades, characterized by LDA > 4 and p < 0.05, which were utilized to elucidate the gut microbial variations across different treatment groups, spanning from the phylum to the genus level (Figure 3A,B). The dominant species in the SAL group were Firmicutes, Patescibacteria, and Deferribacterota at the phylum level; Bacilli, Coriobacteriia, Deferribacteres, and Saccharimonadia at the class level; Erysipelotrichales, Coriobacteriales, Clostridia_vadinBB60_group, Saccharimonadales, and Deferribacterales at the order level; Erysipelotrichaceae, norank_o__Clostridia_vadinBB60_group, Eggerthellaceae, Saccharimonadaceae, and Deferribacteraceae at the family level; and norank_o__Clostridia_vadinBB60_group, Mucispirillum, Candidatus_Saccharimonas, and Ileibacterium at the genus level. The dominant species in the VPA group were Bacillales at the order level; Bacillaceae at the family level; and Bacillus and Lachnospiraceae_UCG-006 at the genus level. The dominant species in the PRO-VPA group were Verrucomicrobiota at the phylum level; Clostridia and Verrucomicrobiae at the class level; Clostridiales, Lachnospirales, and Verrucomicrobiales at the order level; Akkermansiaceae, Clostridiaceae, and Lachnospiraceae at the family level; and Akkermansia, Dubosiella, Lachnospiraceae_NK4A136_group, and UBA1819 at the genus level.

Given the potential for high false-positive rates associated with LEfSe, the ANCOM method was employed as an additional tool to refine the identification of bacterial taxa enriched across group comparisons [19]. At a significance threshold of W > 0.6, the analysis revealed differential abundances of two phyla and ten genera of gut bacteria among the SAL, VPA, and PRO-VPA groups (Figure 3C). Notably, two phyla (Patescibacteria and Verrucomicrobiota) and four genera (Ileibacterium, Candidatus_Saccharimonas, Lachnospiraceae_UCG-006, and Akkermansia) were also identified as differentially abundant bacterial taxa in the LEfSe analysis (Figure 3A,B). These findings indicated that ANCOM refined the identification of enriched bacterial taxa, confirming that Patescibacteria at the phylum level and Ileibacterium and Candidatus_Saccharimonas at the genus level were differentially enriched in the SAL group; Lachnospiraceae_UCG-006 at the genus level was enriched in the VPA group; and Verrucomicrobiota at the phylum level and Akkermansia at the genus level were enriched in the PRO-VPA group. Collectively, these results suggested that the L-proline treatment significantly altered the gut microbiome composition in the VPA-induced ASD mouse model (Figure 3C).

The regulatory impacts of L-proline on microbiota-derived metabolites were examined through a metabolomics approach. PLS-DA revealed significant disparities in metabolite composition among the SAL, VPA, and PRO-VPA groups, suggesting unique gut metabolomic alterations in the ASD mice subsequent to L-proline administration (Figure 4A). The predictive accuracy of the PLS-DA model was confirmed by permutation test results (Figure 4B). Volcano plots were employed to elucidate the differential metabolites between the SAL and VPA groups, as well as between the PRO-VPA and VPA groups, thereby highlighting the modulated metabolites (Figure 4C,D). Specifically, compared to the SAL group, the VPA group exhibited 547 up-regulated and 95 down-regulated metabolites (Figure 4C). Following L-proline intervention, the PRO-VPA group demonstrated 564 up-regulated and 306 down-regulated metabolites (Figure 4D). Figure 4E illustrates the Venn diagram of up-regulated metabolites in the VPA vs. SAL comparison and down-regulated metabolites in the PRO-VPA vs. VPA comparison, accompanied by a heat map depicting their 124 overlapped metabolites. Similarly, Figure 4F displays the Venn diagram of down-regulated metabolites in the VPA vs. SAL comparison and up-regulated metabolites in the PRO-VPA vs. VPA comparison, along with a heat map of their 25 overlapped metabolites. These observations indicate that L-proline exerts a reversal effect on the regulatory influence of the VPA treatment on gut metabolites in mice, leading to a gut metabolic profile in the PRO-VPA group that more closely resembles that of the SAL group (Figure 4E,F).

To gain a deeper understanding of the molecular mechanisms through which L-proline alleviates ASD-like behavioral abnormalities, this study primarily concentrated on the 870 differential metabolites identified between the PRO-VPA and VPA groups. Metabolite annotation was conducted using the HMDB, revealing that 237 metabolites could not be annotated at the Superclass level. The remaining 633 metabolites were categorized as follows: lipids and lipid-like molecules (28.28%), organic acids and derivatives (21.33%), organoheterocyclic compounds (14.85%), phenylpropanoids and polyketides (10.74%), organic oxygen compounds (8.37%), benzenoids (5.37%), nucleosides, nucleotides, and analogues (3.79%), alkaloids and derivatives (2.05%), lignans, neolignans, and related compounds (0.95%), organic nitrogen compounds (0.47%), organophosphorus compounds (0.16%), and an additional 3.63% that were not classifiable within these categories (Figure 5A).

A rigorous exploration into the significance of these metabolites was conducted through KEGG pathway enrichment analysis. An analysis of the 870 differential metabolites between the PRO-VPA and VPA groups using KEGG enrichment revealed that L-proline exerted a significant influence on 27 pathways in the VPA-treated mice (Figure 5B, p < 0.05). Subsequently, to gain further insights into the biological pathways involved in the metabolism of the differentially expressed metabolites and their functional roles, pathway enrichment and topology analyses were performed using MetaboAnalyst (Version 5.0). Based on the identified metabolites and their concentration changes, eight metabolic pathways emerged with a pathway impact value > 0.1, which served as the threshold for relevance. These eight metabolic pathways, which were significantly regulated by L-proline in mice exhibiting ASD-like characteristics, encompassed nucleotide metabolism, citrate cycle (TCA cycle), tryptophan metabolism, alpha-linolenic acid metabolism, pyrimidine metabolism, taurine and hypotaurine metabolism, pyruvate metabolism, and phenylalanine, tyrosine, and tryptophan biosynthesis (Figure 5C, p < 0.05 and impact value > 0.1). The differential metabolites between the VPA and PRO-VPA groups involved in the eight metabolic pathways are presented in Table 1.

Furthermore, the findings from the metabolomic analysis were substantiated and corroborated by the predicted functional analysis of the gut microbiota. To gain insight into how microbial alterations induced by L-proline supplementation influenced mouse metabolism, PICRUSt was utilized to predict the potential functional pathways that could be affected by the gut microbiota. As anticipated from the microbiome data, the L-proline treatment in mice exhibiting ASD-like characteristics resulted in a significant decrease in the pathways of nucleotide metabolism, taurine and hypotaurine metabolism, and pyruvate metabolism, while it increased the pathways of alpha-linolenic acid metabolism and phenylalanine, tyrosine, and tryptophan biosynthesis (Figure 5D). Additionally, the pyruvate metabolic pathway exhibited a tendency to be reduced in the PRO-VPA group compared to the VPA group, albeit with a marginal significance (p = 0.0884). Collectively, these results demonstrated that the metabolomic analyses aligned with several key observations from the microbiome studies, offering valuable insights into the preventive effects of L-proline on ASD-like behavioral disorders.

To gain a deeper understanding of the relationship between the drastically altered gut microbiota and gut metabolites, we conducted an analysis to identify correlations between the 31 differential metabolites involved in eight metabolic pathways significantly regulated by L-proline (as shown in Table 1), and the differentially enriched gut bacteria (comprising two phyla and ten genera, as determined by ANCOM). This analysis was performed using the Spearman correlation method. In constructing the correlation network, only strong (|coefficient| > 0.6) and statistically significant (p < 0.05) correlations were considered for inclusion (Figure 6). The findings indicate that the network comprises 394 statistically significant correlations among the 31 differential metabolites and 12 differential gut bacteria (Figure 6), thereby providing a visual depiction of the intricate interdependencies between the gut microbiota and metabolites.

Several central nodes emerged in the network, forming numerous connections and, thus, likely serving as hubs of physiologically significant interactions. Based on the node degree, the top the gut microbiota identified were Clostridium_sensu_stricto_1, Ileibacterium, and norank_f__Eubacterium_coprostanoligenes_group in descending order (Figure 6). Notably, the genus Clostridium_sensu_stricto_1 exhibited extensive connectivity within the network, demonstrating correlations with 25 nodes included in the correlation analysis. Furthermore, Ileibacterium and norank_f__Eubacterium_coprostanoligenes_group displayed significant and strong correlations with 24 nodes, respectively (Figure 6). Moreover, according to the node degree from high to low, protocatechuic acid, 2-isopropylmalic acid, and malonic acid represented the top three differential metabolites, showing significantly strong correlations with 25, 24, and 24 nodes, respectively (Figure 6). These findings suggest robust associations between the gut microbiota and gut metabolites influenced by L-proline. Collectively, these results imply that L-proline’s capacity to prevent autism-like behaviors may be linked to its regulatory effects on the interactions and dynamic balances between gut microbial composition and gut metabolites.

Data supporting reported results can be requested after publication from the corresponding author.