Lactobacillus fermentum (recently re-classified as Limosilactobacillus fermentum; Zheng et al., 2020) strains PS150 and GR1009 were both isolated from the same fermented sausage (Liu et al., 2019). Bacterial strains used in the present study were anaerobically cultured in Mann Rogosa Sharp (MRS) broth (Criterion, Hardy Diagnostics, Santa Maria, CA, USA) at 37°C for 18 h. The cells were harvested by centrifugation at 4°C, 10,000 g for 10 min. For the animal model, the cells were resuspended in MRS broth containing 12.5% glycerol and adjusted to a final concentration of 10¹⁰ colony-forming units per milliliter (CFU/ml), and then aliquoted and stored at −80°C. Before oral administration, bacterial stocks were removed from −80°C storage and thawed in water bath at 37°C for 1 h and then centrifuged at 10,000 g for 10 min at 4°C. The supernatant was discarded, and the pellet was resuspended in phosphate-buffered saline (PBS). For heat-killed L. fermentum PS150 (HK-PS150), the harvested cells were resuspended in PBS and adjusted to a final concentration of 10¹⁰ CFU/ml, and then heated at 80°C for 30 min in the water bath. The heat-treated samples were also stored at −80°C before use.

Bacterial genomic DNA was prepared by phenol extraction. briefly, the harvested cell pellets were resuspended in genome extraction buffer (200 mM Tris–HCl, 80 mM EDTA (pH 9.0), 2% w/v sodium dodecyl sulfate). The suspensions were then supplied with an equal volume of phenol and 0.1 mm glass beads. The mixtures were lysed with a FastPrep FP120 homogenizer (Q-Biogene, Carlsbad, CA, USA), and genomic DNA was extracted using phenol-chloroform extraction. The quality of DNA extracts was validated using NanoDrop spectrophotometer. Random amplification of polymorphic DNA (RAPD) and enterobacterial repetitive intergenic consensus (ERIC) PCR were performed using Takara Taq polymerase (Takara Bio Inc., Shiga, Japan) with designated primers In accordance with the manufacturer’s instructions (Versalovic et al., 1991). The arbitrary sequence RAPD-B (5′-AACGCGCAAC-3′) was used in RAPD, and the primer pair ERIC1 (5′-ATGTAAGCTCCTGGGGATTCAC-3′) and ERIC2 (5′-AAGTAAGTGACTGGGGTGAGCG-3′) were used in ERIC PCR. RAPD and ERIC PCR products were analyzed by electrophoresis In a 1% agarose gel, followed by SYBR safe staining (Thermo fisher, Waltham, MA, USA).

The bacterial culture was spread on MRS agar plates and incubated anaerobically at 37°C for 48 h. After incubation, colonies were scraped from the agar plates for DNA extraction. A DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) was used to extract genomic DNA for MinION long-read sequencing. For Illumina sequencing, DNeasy UltraClean microbial kits (Qiagen) were used to prepare genomic DNA. Genomic DNA was extracted according to the manufacturer’s protocol provided by the vendor. The quality of extracted DNA was examined using Qubit4 (Thermo Fisher, Waltham, MA, USA) and a UV spectrophotometer. The shotgun sequencing library for Illumina sequencing was prepared using the Nextera DNA Flex library with DNA CD Indexes (Illumina, CA, USA) following standard protocols. Quality control of the sequencing libraries was performed using an Agilent 2,100 Bioanalyzer (Agilent, CA, USA). Sequencing was performed using an iSeq100 (Illumina). De-multiplexing and trimming were performed using a Basespace (Illumina). For long-read sequencing, a shotgun library for genomic DNA was constructed using the Rapid Barcoding kit SQK-RBK004 (Oxford Nanopore Technologies, Oxford, UK). The sequencing library was sequenced, base- called, and debarcoded on a MinION Mk1C (Oxford Nanopore Technologies). De novo genome assembly was performed using Unicucler v0.4.8 (Wick et al., 2017). The completeness of the resulting assembly was validated using Bandage (Wick et al., 2015) and further examined by mapping Illumina short reads using the CLC Genomic Workbench (Qiagen). The genomes were annotated using the NCBI Prokaryotic Genome Annotation Pipeline. Carbohydrate gene clusters were analyzed using the dbCAN2 meta server and MAUVE (Darling et al., 2004; Yin et al., 2012; Zhang et al., 2018). The CLC Genomics Workbench (Qiagen) was used for single nucleotide polymorphism (SNP) analysis. To identify inverted repeat sequences, the flanking sequence of the IS256 transposase was analyzed using the palindrome function of the EMBOSS server (Rice et al., 2000).

The PCR primer set Eps1F (5′-ATCCCACCCACATGACGTTC-3′) and Eps1R (5′-AGTTTATCCGCACGAGGAGT-3′) were designed according to specific DNA sequences (located in 113,863–113,882 and 128,276 – 128,295) respectively in the chromosome of PS150. The estimated amplicon of Eps1F/Eps1R was 14,433 bp in PS150 and 2,303 bp in GR1009. Amplification of eps1 was carried out using a long PCR enzyme mix (Thermo Fisher, Waltham, MA, USA) following the manufacturer’s three-step cycling protocol. First, the molecular size of the PCR products was analyzed by electrophoresis in 1% agarose gel, followed by SYBR safe staining (Thermo Fisher, Waltham, MA, USA). The DNA sequence of the PCR products was then bidirectionally confirmed by Sanger sequencing using the Eps1F and Eps1R primer set.

Male C57BL/6J mice (6 weeks old) were purchased from the National Laboratory Animal Center (Taipei, Taiwan). The mice were housed in the Laboratory Animal Center of the National Yang Ming Chiao Tung University. The room was maintained at a constant temperature (22 ± 1°C) and humidity (55–65%) with a 12 h light/dark cycle. The mice were fed ad libitum with a standard chow diet and sterilized water. All experiments were conducted in accordance with relevant guidelines and regulations, and were approved by the Institutional Animal Care and Use Committee of National Yang Ming Chiao Tung University (IACUC No. 1080702).

For the comparative assessment of PS150, GR1009 and HK-PS150, the bacterial stocks were thawed and recovered at room temperature before use. The time- and dose-dependent properties of PS150 have been confirmed, the mice used in this experiment were intragastrically administered 0.2 ml PBS or bacterial suspensions (10⁹ CFU/day) for 13 days (Lin et al., 2019). On day 14, the mice were administered PBS, 10 mg/kg diphenhydramine (DIPH, Sigma, Saint Louis, MO, USA), or bacterial suspensions 30 min before the intraperitoneal injection of pentobarbital sodium (50 mg/kg, Sigma). Following the injection, the mice were tested for righting reflex by flipping upside down. The time between the pentobarbital injection and the loss of righting reflex was defined as the sleep latency. The time required to restore the righting reflex was defined as the sleep duration, whereas the time spent between righting action and voluntary movement was defined as the recovery time. For the functional validation of (C-EPS), the mice were orally administered PBS, MRS broth, or C-EPS (0.4 mg/day) for 13 days. On day 14, the mice were treated as described above for evaluating sleep conditions.

All mice were placed individually in clean cages until defecation. Immediately after the mice defecated, the stool samples were collected into 1.5 ml microtubes containing 0.3 ml of RNAlater solution (Invitrogen) and then frozen at −80°C. For DNA extractions, samples were homogenized and washed twice with PBS and then lysed with 0.2 mm glass beads using a FastPrep FP120 homogenizer (Q-Biogene, Carlsbad, CA, USA). Following centrifugation at 12,000 g for 5 min at 4°C, then the supernatant (0.4 ml) was collected. The bacterial genomic DNA was extracted using phenol–chloroform extraction, and subsequently, the concentration and quality of the DNA were determined using the NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).

Next generation sequencing and data processing were performed by BioTools Co. Ltd. (Taipei, Taiwan). Briefly, the 16S rRNA V3–V4 region was amplified via PCR using a region-specific primer set (341F: 5′-CCTACGGGNGGCWGCAG-3′, 806R: 5′- GACTACHVGGGTAT CTAATCC-3′) according to the 16S Metagenomic Sequencing Library Preparation procedure (Illumina). The quality of the indexed PCR product was evaluated using the Qubit 4.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA) and Qsep100™ system (Bioptic). An equal amount of the indexed PCR product was combined to create the sequencing library, which was ultimately sequenced on an Illumina MiSeq platform to generate paired 300-bp reads. After demultiplexing each sample, the primer and adapter sequences were removed from the paired-end reads using the QIIME2 cutadapt plugin. To construct the Amplicon Sequence Variants (ASVs), a denoising pipeline was executed using the QIIME2 DADA2 plugin (v2020.11). The community composition was differentially analyzed using the linear discriminant analysis effect size (LEfSE) method (Segata et al., 2011), which employs a non-parametric Kruskal-Wallis test and Wilcoxon rank-sum test algorithm to identify bacterial taxa whose relative abundance significantly differs between control and experimental groups. Following identification of these taxa, LEfSe uses Linear discriminant analysis (LDA) to assess the effect size of each differentially abundant taxon. In the present study, taxa with LDA score (log₁₀) > 2 was considered significant.

The bacterial suspensions of PS150 and GR1009 were dispensed and grown on glass slides placed in a 24-well plate. After centrifugation at 2,000 g for 5 min, the samples were primarily fixed by 100 mM PBS (pH 7.2) containing 2.5% glutaraldehyde at room temperature for 1 h. The samples were rinsed three times with 5 min incubations in PBS and subsequently fixed by PBS containing 4% paraformaldehyde and 2.5% glutaraldehyde at room temperature for 30 min. After rinsed three times with 5 min incubations in the PBS, the samples were incrementally dehydrated in a series of washes in 30 and 50% ethanol for 10 min each at room temperature, 75% ethanol overnight at 4°C, 85 and 95% ethanol for 10 min each at room temperature, and 100% ethanol three times for 20 min each at room temperature. Finally, the samples were dried by critical-point drying in liquid CO₂ and sputter-coated with gold and observed by the scanning electron microscopy (JEOL JSM-7600F).

Bacterial culture of PS150 was prepared as previously described (Liu et al., 2019). The overnight culture was heated at 80°C for 1 h, and the bacterial cells were removed by centrifugation at 7,000 g for 30 min. The polysaccharide component was then precipitated using 3× volume of anhydrous ethanol at 4°C for 24 h. The precipitants were then collected at 7,000 g for 30 min and washed with 70% ethanol. The residual ethanol was removed by evaporation, and the EPS-containing extract was dissolved in ddH₂O and stored at −80°C until use. The recovery rate of saccharide content and protein contamination during the purification process were monitored using phenol sulfuric acid (Masuko et al., 2005) and the Bradford method, respectively.

The assembled genomes of PS150 (GR1008) and GR1009 were uploaded to the NCBI Bioproject (PRJNA702613).

Results were described as mean ± standard error of the mean (SEM). Graphs were generated and statistical analysis performed using the software GraphPad Prism 9.0. All data were analyzed by one-way ANOVA using the Tukey post hoc test or unpaired t-test.

PS150 exhibited a white, opaque, and circular colony when cultured anaerobically on a MRS agar plate (Figure 1A). From the same source as PS150, we isolated another L. fermentum strain that exhibited a flat, transparent colony with a rough surface and undulated margin, named GR1009. In contrast to the apparent “ropy” phenotype of PS150, the colony of GR1009 was not slimy. Furthermore, after overnight anaerobic incubation in MRS broth, the viscosity of the liquid culture was assessed by centrifugation. PS150 cells formed a puffy layer after centrifugation whereas GR1009 cells could be sedimented into a solid pellet (Figure 1B). Despite apparent differences in colony morphology between PS150 and GR1009, it is intriguing that the DNA fingerprinting profiles thereof were similar (Figure 1C), suggesting that both strains share a common origin.

According to our laboratory’s previous study, we have demonstrated the potential sleep-improving effects of PS150 in a pentobarbital-induced sleep mouse model (Lin et al., 2019). In this study, we aimed to investigate whether GR1009 possess the hypnotic efficacy; in addition, we were also interested in the possible effect of PS150 as a postbiotic. We used an 80°C water bath to kill PS150 cells. The PS150, HK-PS150, and GR1009 were then examined in a pentobarbital-induced sleep mouse model to characterize their hypnotic functions. Followed the experiment design, the sleep latency, sleep duration, and recovery time were measured to evaluate the sleep initiation, persistence, and awakening, respectively (Figure 2A). Compared with the PBS group, intervention with PS150 (175.50 ± 26.68 s) and HK-PS150 (183.20 ± 20.79 s) exhibited a significant effect on reducing sleep latency, the reducing effect were more than GR1009 (236.10 ± 46.26 s) and the control drug, DIPH [223.90 ± 40.87 s; Figure 2B, F(4, 45) = 10.18, df = 4, p < 0.0001]. On sleep duration, both PS150 (5262.00 ± 1465.04 s) and HK-PS150 (5577.10 ± 1228.98 s) groups exhibited a trend in elongating sleep duration, only GR1009 group (4224.33 ± 541.57 s) has no obvious difference compared to the PBS group [3975.70 ± 588.15 s; Figure 2C, F(4, 44) = 5.072, df = 4, p = 0.0019]. Notably, both PS150 and HK-PS150 exhibited significantly distinct effects on recovery time compared to DIPH and GR1009 [Figure 2D, F(4, 43) = 8.587, df = 4, p < 0.0001]. Administration of PS150 (17.63 ± 6.95 s) and HK-PS150 (29.90 ± 24.79 s) demonstrated a significant reduction on recovery time, while neither DIPH (90.90 ± 42.99 s) nor GR1009 (113.60 ± 54.56 s) demonstrated this effect. These results suggest that HK-PS150 cells are postbiotics that possess heat-stable bioactive effectors. Furthermore, despite sharing a common origin between PS150 and GR1009, the hypnotic effect of GR1009 was found to be impaired compared to that of PS150. As GR1009 exhibited ineffectiveness of hypnotic activity, we hypothesized that the expression of the bioactive effectors is reduced or absent in GR1009 cells.

Next, we investigated the composition of the gut microbiota among the PBS, PS150 and HK-PS150 groups. At the phylum level, Firmicutes and Bacteroidetes were the most predominant bacterial phyla, the ratio of Firmicutes/Bacteroidetes (F/B) have no significant difference between the three experimental groups. Compared to the PBS group, Actinobacteria level was increased in PS150 (Fisher’s exact test, p = 0.0414) but not in HK-PS150 group. Furthermore, Bacteroidetes and Tenericutes levels were also slightly increased in both PS150 and HK-PS150 group. In contrast, the relative abundance of Patescibacteria significantly decreased in PS150 (Fisher’s exact test, p = 0.0127) or HK-PS150 (Fisher’s exact test, p = 0.0127) administration (Figure 3A).

At the genus level, we obtained 12 differentially abundant taxa after PS150 or HK-PS150 oral administration. As shown in Figure 3B, the abundances of Blautia, Dubosiella, and Lachnospira were increased in PS150 group, while Candidatus_Soleaferrea, Candidatus_Saccharimonas, Lachnospiraceae_UCG_006, Ruminococcaceae_UCG_005, and Tyzzerella were decreased compared with the PBS group. Compared with the HK-PS150 group and the PBS group, the relative abundances of Blautia, [Eubacterium]_nodatum_group, and Parabacteroides were increased in the HK-PS150 group while that of Gordonibacter was decreased. We then analyzed the gut microbiota composition at the species level in the three experimental groups, Lactobacillus_fermentum and Clostridium sp. Culture_54 levels were significantly increased after PS150 or HK-PS150 administration (Figures 3D,E). LEfSe analysis was further applied and identified different microbiome signatures between groups, revealing the gut microbiota composition was altered in both PS150 and HK-PS150 treated mice (Supplementary Figure S1).

We utilized a comparative approach to delineate genomic differences between PS150 and GR1009. Whole-genome sequences of both PS150 and GR1009 were determined using a combination of Nanopore and Illumina sequencing. The results revealed that the genome size of GR1009 was slightly smaller than that of PS150, with approximately the same GC content (Supplementary Table S1). No plasmids were detected in either strain. Single nucleotide polymorphism (SNP) analysis was performed by mapping Illumina sequencing reads of both strains to their assembled genomes. We discovered a total of 11 SNPs, wherein five SNPs were non-synonymous polymorphisms (Table 1). In addition, we identified five insertion–deletion mutations (indels) using whole-genome BLAST, wherein three indels were insertion sequence (IS) elements (Table 2).

As mentioned above, whole-genome sequencing revealed that GR1009 genome size is slightly smaller than PS150. Compared with PS150, a 13-kb deletion was detected in the GR1009 genome (Table 2). Functional annotation revealed that this deletion region in GR1009 is an EPS gene cluster (referred to as eps1) which contains 13 genes that fulfill the requirement for EPS biosynthesis, including a putative priming glycosyltransferase (GT), four GTs for sugar subunit synthesis, chain length determinant proteins, a polymerase for repeat unit synthesis, and a flippase for product exportation (Figure 4A; Table 3). We used polymerase chain reaction (PCR) and Sanger sequencing to reconfirm the deletion region and nucleotide sequence in PS150 and GR1009, respectively (Figure 4B). As shown in Figure 4A, a putative upstream inverted repeat (IRL), belonging to an IS256 transposable element (locus tag: JYQ66_00485), was detected adjacent to the deletion site. The results suggest that GR1009 might loss its esp1 cluster region during transposon jumping.

Given the discovery of eps1 deletion in GR1009, we further assessed other potential EPS biosynthesis systems of PS150 by dbCAN2 prediction program.¹ As the result, 11 carbohydrate-related gene clusters with various functions were identified in PS150, three of which featured multiple GTs, and possible exportation mechanisms were predicted as EPS biosynthetic gene clusters, including the eps1. The second putative EPS biosynthetic gene cluster (referred to as eps2) comprises four GTs and two transporters (Supplementary Table S2).

The third putative EPS biosynthetic gene cluster (referred to as eps3) comprises two glycosyltransferases for subunit synthesis, one polymerase for repeat unit polymerization, one chain length determinant protein, and a flippase for exportation (Supplementary Table S3). Sequence alignment revealed that both eps2 and eps3 of GR1009 were identical to that of PS150. Accordingly, loss of eps1 was the only difference we found in the EPS biosynthetic related genes between both strains.

To confirm the comparative approach results, PS150 and GR1009 were further tested for EPS production level. We used scanning electron microscopy (SEM) to investigate whether EPS was present in bacterial colony which obtained from the MRS agar plate. As shown in Figure 5A, we observed the presence of EPS-like matrix that heavily coated PS150 cells (yellow arrow), while GR1009 cells exhibited a relatively clean background, revealing potential differences in the EPS production level between PS150 and GR1009. To further qualify the EPS yield of PS150 and GR1009, we purified EPS-like matrix of PS150 and GR1009 from bacteria colonies on MRS agar plates and determined the sugar content by the phenol-sulfate method. As expected, PS150 contains a higher sugar content than GR1009. The EPS yield of GR1009 (1.65 ± 0.11 mg/g/biomass) was significantly lower than that of PS150 (6.84 ± 0.55 mg/g/biomass, p < 0.001; Figure 5B).

As describe above, we demonstrated that GR1009 shown significantly lower EPS level than PS150. Based on the result, we further hypothesized that the EPS of PS150 is responsible for the reduced sleep recovery (Figure 6). To confirm the hypothesis, the hypnotic function of the EPS-containing crude extract (C-EPS) derived from PS150 was compared with PBS in a pentobarbital-induced sleep mouse model. Furthermore, we took unfermented MRS into comparison to ensure that the hypnotic function was not triggered by components in our culture medium. It showed significant reduction in sleep latency after treatment with the PS150-derived C-EPS [Figure 6B, F(3, 16) = 14.11, df = 3, p < 0.0001]. As live PS150 and HK-PS150, the C-EPS exhibited an increasing trend in terms of sleep duration [Figure 6C, F(3, 16) = 10.98, df = 3, p = 0.0004]. In addition, C-EPS treatment significantly reduced sleep recovery [Figure 6D, F(3, 16) = 35.29 df = 3, p < 0.0001]. The reduced sleep latency and recovery time revealed that the C-EPS was sufficient to reproduce the hypnotic effects of PS150. In contrast, the unfermented MRS did not affect any parameters in the present animal study. To sum up, our data demonstrated that GR1009 has lower EPS production level than PS150 due to eps1 loss; most importantly, implied that EPS plays an important role in the hypnotic effect of PS150.