Using engineered Lactobacillus bacteria to deliver GLP-1 peptide by mouth for treating type 2 diabetes

In this study, the plasmid pUC57-usp45-GPA was commercially synthesized. The target gene fragment encoding GPA was subsequently cloned into the pMFH vector, as illustrated in Fig. 1a and Fig. S1a and b. The resulting pMFH-GPA plasmid was then transformed into E. coli BL21 for expression. Following fermentation, the expression of pMFH-GPA fusion protein was induced by Isopropyl β-D-thiogalactoside (IPTG) and confirmed through tricine-SDS–PAGE and Western blot analysis (Fig. S1c and d). The pMFH-GPA fusion protein, with a molecular weight of 22 kDa (Table S3), has a theoretical isoelectric point (pI) of 9.83, as determined using the ExPASy ProtParam tool (https://web.expasy.org/protparam/↗). The fusion protein was subsequently cleaved using cyanogen bromide (CNBr) and purified. The crude GPA peptide was further purified by high-performance liquid chromatography (HPLC) and validated by Western blot analysis (Fig. 1b). The purified GPA peptide has a molecular weight of 8.1 kDa and a theoretical pI of 10.93, as detailed in Table S3.

To evaluate the bioactivity of GPA peptide in rats, the presence of GPA in the blood was analyzed using HPLC and Western blot analysis 24 h after intraperitoneal administration of 0.02 mg/kg of body weight GPA (Fig. 1c). To further assess the bioactivity of GPA peptide in mice, a glucose tolerance test (GTT) was conducted in C57BL/6J mice. Compared to the control group, GPA administration significantly improved blood glucose tolerance and reduced AUC value (P < 0.05, Fig. 1d and e).

To further evaluate the bioactivity of GPA peptide in db/db mice, random blood glucose levels were measured. As illustrated in Fig. 1f, the random blood glucose levels in db/db mice were decreased gradually within 3 h post-injection, followed by a gradual increase, returning to the baseline levels by 12 h. In addition, fasting blood glucose levels were monitored in db/db mice before and after treatment with 0.2 mg/kg body weight GPA peptide. As shown in Fig. 1g, the peptide significantly reduced fasting blood glucose levels 24 h after administration. These findings demonstrate the glucose-lowering effects of the GPA peptide in both normoglycemic and diabetic murine models.

The target gene fragment encoding usp45-GPA was cloned into the pMG36e vector (Fig. 1h; Fig. S2 and S3), enabling the successful construction of the recombinant probiotic strain, Lgs^GPA. As a control, a recombinant probiotic strain harboring the empty pMG36e vector, Lgs^36e, was also generated (Fig. S4).

The expression of the GPA peptide in the supernatant of the recombinant L. gs strains was confirmed by Western blot analysis (Fig. 1i) and quantified using an ELISA assay (Fig. 1j). To assess the bioactivity of the secreted GPA peptide, the expression of β-cell markers PDX-1 and insulin release were evaluated in Min6 cells treated with the medium of the recombinant L. gs strains. Liraglutide was used as a positive control. Compared with the Lgs^36e group, the Lgs^GPA group exhibited a significant increase in the expression of PDX-1 (Fig. 1k). As shown in Fig. 1l, although the insulin secretion induced by the medium of Lgs^GPA was slightly less pronounced than that induced by 100 nmol/L liraglutide, it still significantly stimulated insulin secretion compared to the medium of Lgs^36e (P < 0.01). These results demonstrate the functional activity of the GPA peptide secreted by Lgs^GPA in promoting β-cell marker expression and insulin release in vitro.

To evaluate the therapeutic potential of Lgs^GPA, its bioactivity was investigated in db/db mice. Compared to the Lgs^36e group, the fasting blood glucose levels in the Lgs^GPA group exhibited a decreasing trend starting from the 2nd week. Although this reduction was not statistically significant during the 4th and 5th weeks, by the 6th week, the fasting blood glucose levels in the Lgs^GPA group were significantly lower than those in the Lgs^36e group (Fig. 2b). Furthermore, Lgs^GPA significantly improved glucose tolerance, as evidenced by a notable reduction in the AUC value (P < 0.001), compared to the Lgs^36e group (Fig. 2c). A similar trend was observed in the insulin tolerance test (ITT), with the AUC values mirroring those of the GGT (Fig. 2d). In addition, Lgs^GPA significantly reduced serum levels of glycosylated hemoglobin A1c (GHbA1c) compared to the Lgs^36e group (Fig. 2e).

To assess the broader metabolic effects of Lgs^GPA, various blood biochemical parameters were measured in db/db mice. As shown in Fig. 2, Lgs^GPA had no significant impact on liver function, as indicated by serum levels of aspartate aminotransferase (AST) (Fig. 2f), alanine aminotransferase (ALT) (Fig. 2g), and the ratio of AST/ALT (Fig. 2h). Similarly, kidney function, measured by serum creatinine (CRE) (Fig. 2i) and blood urea nitrogen (BUN) levels (Fig. 2j), remained unaffected. Lgs^GPA also did not alter serum levels of total cholesterol (TC) (Fig. 2l) or low-density lipoprotein cholesterol (LDL-C) (Fig. 2n). However, compared to the Lgs^36e group, Lgs^GPA significantly increased serum levels of high-density lipoprotein cholesterol (HDL-C) (Fig. 2m) (P < 0.05) and significantly decreased serum levels of triglyceride (TG) (Fig. 2k) (P < 0.05). These findings highlight the potential of Lgs^GPA to improve glycemic control and lipid metabolism in a diabetic mouse model without adversely affecting liver or kidney function.

To explore the therapeutic potential of Lgs^GPA, its impact on the gut microbiome of db/db mice was analyzed. A total of 376,173 high-quality reads and 1,717 operational taxonomic units (OTUs) were obtained. The α diversity analysis, including the ace index, sobs index, Chao1 index, Shannon index, Pielou evenness, and Simpson index at the OTU level, revealed no significant differences between the Lgs^GPA and Lgs^36e groups (Fig. S6a). The Venn diagram revealed 94 common OTUs between the two groups, with 14 unique OTUs in the Lgs^36e group and 7 unique OTUs in the Lgs^GPA group, respectively (Fig. 3b). The β diversity analysis of principal coordinates (PCoA) demonstrated partial separation between the two groups (Fig. 3c).

Taxonomic analysis revealed differences in microbial composition between the groups. At the phylum level, Firmicutes and Bacteroidota were dominant taxa. In the Lgs^GPA group, the average relative abundance of Firmicutes decreased, while that of Bacteroidota increased, resulting in a reduction of the Firmicutes/Bacteroidota (F/B) ratio, although these changes were not statistically significant (Fig. S5a and S6b). At the order level, Bacteroidales, Lactobacillales, and Lachnospirales were predominant. The relative abundance of Clostridiales (P = 0.09) increased, while Desulfovibrionales (P = 0.08) and Erysipelotrichales (P = 0.08) decreased in the Lgs^GPA group compared to the Lgs^36e group (Fig. S5b and S6c). At the family level, Muribaculaceae, Lactobacillaceae, and Rikenellaceae were the most abundant. Notably, the relative abundance of Erysipelotrichaceae, Enterococcaceae, and the combined abundance of these two potentially pernicious bacteria was significantly reduced (P < 0.05) in the Lgs^GPA group (Fig. S5c and S6d) compared to the Lgs^36e group. At the genus level, unclassified_f__Muribaculaceae, Ligilactobacillus, and Lactobacillus were predominant (Fig. S5d). The relative abundances of unclassified_f__Erysipelotrichaceae and Intestinimonas were significantly lower (P < 0.05) in the Lgs^GPA group (Fig. 3e and 4a) compared to the Lgs^36e group.

Correlation analysis identified three bacterial taxa associated with two clinical parameters (highlighted in red boxes). At the genus level, unclassified_o__Coriobacteriales, Enterorhabdus, Corynebacterium, Lachnospiraceae_UCG-001, and Psychrobacter (marked in blue) were identified as potential targets of Lgs^GPA in ameliorating hyperlipidemia or hyperglycemia in db/db mice (Fig. 4b).

Linear discriminant analysis effect size (LEfSe) analysis was utilized to estimate microbiome differences at the genus level between the two groups (Fig. 3d). Only taxa with linear discriminant analysis (LDA) scores above 2 were presented, revealing 15 differentially abundant bacteria between the two groups. In addition, Student’s t-test identified six bacteria that differed between the groups, with two (marked in blue) ranking among the top 50 (Fig. 3e), suggesting that the Lgs^GPA may improve intestinal conditions in db/db mice.

Functional predictions based on the 16S amplicon sequencing data, combined with the Clusters of Orthologous Genes (COGs) and KEGG databases, indicated that bacterial functions were primarily related to metabolic pathways (Fig. S7). The COGs database predicted that bacterial functions were predominantly involved in metabolism, with amino acid metabolism being the third most abundant (Fig. S7a). Similarly, 16S rRNA sequencing data combined with KEGG functional predictions indicated that bacterial colony function is primarily related to metabolic pathways, especially carbohydrate metabolism (Fig. S7b and c). Further analysis using KEGG data at a deeper level (3rd Level) confirmed the association of bacterial functions with metabolic pathways (Fig. S7d). These findings suggest that Lgs^GPA modulates the gut microbiota in db/db mice, potentially contributing to its therapeutic effects on metabolic dysregulation.

We confirmed the genetic stability of the recombinant L. gs strain across three generations cultured from stool samples using PCR (Fig. S8). As illustrated in Fig. 5a, rats were intragastrically administered the recombinant L. gs strains for 1 week. To verify the bioactivity of Lgs^GPA in the gastrointestinal tract, the presence of the GPA peptide in the blood of rats was analyzed utilizing HPLC and Western blot analysis (Fig. 5b) following intragastric administration of recombinant Lgs^GPA. To further evaluate the potential preventive value of Lgs^GPA, its impact on the stool microbiome of SD rats was investigated. A total of 455,489 high-quality reads and 173 OTUs were obtained. The Venn diagram revealed that 643 common OTUs were shared between the two groups, with 87 unique OTUs in the Lgs^36e group and 168 unique OTUs in the Lgs^GPA group (Fig. S9a). The α diversity analysis, including the ace index, sobs index, Chao1 index, Shannon index, Pielou evenness, and Simpson index at the OTU level, showed no significant difference between the two groups (Fig. S9c). However, the β diversity analysis using PCoA indicated partial separation between the two groups (Fig. S9b). These results suggest that Lgs^GPA may enhance microbial community diversity.

Taxonomic analysis revealed differences in microbial composition between the groups. At the phylum level, Firmicutes and Bacteroidota were the dominant taxa. In the Lgs^GPA group, the average relative abundance of Firmicutes decreased, while that of Bacteroidota increased, resulting in a reduction in the F/B ratio, although these changes were not statistically significant (Fig. S9d). At the family level, Muribaculaceae, Lactobacillaceae, and Akkermansiaceae were the most abundant (Fig. 5c). At the genus level, norank_f_Muribaculaceae, Lactobacillus, and Akkermansia were predominant. Notably, the average relative abundance of Akkermansia was higher in the Lgs^GPA group compared to the Lgs^36e group, although this increase was not statistically significant (Fig. 5d).

Functional predictions based on 16S amplicon sequencing data, combined with the COGs and KEGG databases, indicated that bacterial functions were primarily associated with metabolic pathways (Fig. S10). The COGs database predicted that bacterial functions were mainly involved in metabolism, with amino acid metabolism representing the third most abundant category (Fig. S10a). Similarly, 16S rRNA sequencing data combined with KEGG functional predictions indicated that bacterial colony function is primarily related to metabolism pathways, particularly carbohydrate metabolism (Fig. S10b and c). Further analysis using KEGG at a deeper level (3rd Level) confirmed the association of bacteria with metabolic pathways (Fig. S10d). These findings suggest that Lgs^GPA modulates the gut microbiota in SD rats, potentially contributing to its preventive and therapeutic effects through metabolic regulation.

For this study, the bacterial strains and plasmids used are given in Table S1, primers from Genscript Co. (Nanjing, China) are given in Table S2. DNA restriction enzymes (Xba I, Pst I, EcoR, and BamH I), Dpn I, alkaline phosphatase (Calf intestine, CIAP), and the T4 DNA ligase for ligation were supplied by Takara Co. (Dalian, China). Mini-plasmid kit was from Tiangen Co. (Beijing, China). GLP-1 and pancreatic duodenal homeobox-1 (PDX-1) antibody were from Abclonal Co. (Wuhan, China). Min6 cell line was from Servicebio Co. (Wuhan, China). RPMI Medium 1640 basic and Fetal Bovine Serum (FBS) were from Gibco (New York, USA). Liraglutide was from Aikang Biopharmaceutical (Jiangsu, China). Other chemicals and reagents used were analytical grade.

The plasmids pMFH-GPA and pMG36e-usp45-GPA (pMG36e-GPA) were constructed by our lab. All the constructed plasmids were verified with DNA sequencing by Tianyihuiyuan Co. (Wuhan, China). Detailed experimental protocols are described in the supplementary information. The pMFH-GPA fusion gene was transformed into E. coli DH5α, E. coli BL21 using the heat shock method. The pMG36e-usp45-GPA fusion gene was transformed into E. coli MC1061 using the heat shock method. After preparation of the L. gs-competent cells, the pMG36e and pMG36e-GPA vectors were transformed into L. gs competent cells by electroporation (47). The recombinant L. gs is named Lgs^36e and Lgs^GPA, respectively.

The recombinant strain E. coli BL21 transformed with pMFH-GPA was induced by 0.3 mmol/L β-D-isopropylthiogalactopyranoside (IPTG). The cells were collected by centrifugation at 3,500 × g for 20 min. Pelleted cells were resuspended in moderate Tris-HCl buffer (20 mmol/L Tris, 100 mmol/L NaCl, pH 8.0). Then, the cells were lysed by homogenization. The pellets were collected by centrifugation at 12,000 × g for 20 min. The pellet was dissolved in 6 mol/L urea (pH 8.0). After sonication, the supernatants were collected by centrifugation at 7,500 × g for 20 min. An equal volume of pre-cooled ethanol (−20°C) was added to precipitate protein. The pellets were collected by centrifugation at 12,000 × g for 20 min. The supernatants were detected by tricine sodium dodecyl sulfate-polyacrylamide gel electrophoresis (tricine-SDS-PAGE) (48, 49) or Western blot analysis. After CNBr treatment and cleavage of the fusion protein (50), the GPA peptide was purified by HPLC and detected by Western blot analysis.

Min6, a mouse islet cell line, was cultured in RPMI Medium 1640 with 10% FBS and cultured at 37°C in 5% CO₂ in a humidified incubator. After 80%–90% confluency, the cells were seeded in 24-well plates. After 80%–90% confluency, cells were rapidly rinsed twice with Hank’s balanced salt solution (HBSS) (51), incubated by HBSS with 3 mmol/L glucose for 2 h. Then, the cells were treated with the sterile medium of the recombinant L. gs, which was diluted with an equal volume of RPMI Medium 1640 for 2 h. Liraglutide (100 nmol/L) mixed with RPMI Medium 1640 was used as a positive control. Then, the supernatants were collected by centrifugation (10 min, 3,000 × g, 4°C), and the levels of insulin in the supernatants were detected with insulin enzyme-linked immunosorbent assay (ELISA) kits (Ruixin Co., Quanzhou, China). For the samples of Western blot analysis, the cells were treated with the above mixes for 24 h, then the cells were collected to analyze the expression of PDX-1.

All animals were kept and cared for in accordance with the guidelines of the Hubei University of Technology Animal Care and Use Committee. The animals were fed a standard food diet. When treated with recombinant L. gs, mice or rats were gavaged with microbiota solution (1–5 × 10⁸ CFU bacteria) (52).

For GTT, mice were fasted for 12 h with free access to autoclaved water, and then they were intragastrically administered glucose at 2 g/kg (C57BL/6J mice) or 1 g/kg (db/db mice) of body weight. For ITT, mice were fasted for 6 h with free access to autoclaved water, and then they were intraperitoneally injected with 1.5 U/kg body weight insulin. The blood glucose levels were measured at 0, 15, 30, 60, 90 and 120 min, and the area under the curve (AUC) was calculated after the mice were treated with glucose/insulin (53). A glucometer (Sinocare, China) was used to measure the blood glucose of mice.

HPLC analysis was carried out with the protocol described previously (8) with slight modifications. Briefly, the GPA was separated or detected at a flow rate of 1 mL/min. The gradient program was set as follows: 0–5 min, 95% buffer A (0.1% trifluoracetic acid in acetonitrile), and 5% buffer B (0.1% trifluoracetic acid in H₂O); 5–30 min, 95%-5% buffer A, 5%-95% buffer B; 30–40 min, 95%–100% buffer A. The peaks were collected, and they were detected by Western blot analysis.

All animals were anesthetized after an overnight fast. Then, blood samples were rapidly collected. After centrifugation (10 min, 3,000 × g, 4°C), serum samples were stored at −80°C. The concentrations of alanine aminotransferase (ALT), aspartate transaminase (AST), creatinine (CRE) and urea (BUN), total cholesterol (TC), total triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and glycosylated hemoglobin A1c (GHbA1c) in serum were measured by corresponding commercial kits (Jiancheng Co., Nanjing, China) according to the instructions, respectively.

The recombinant Lgs^36e and Lgs^GPA were amplified in culture, respectively. Clarified supernatants (300 mL) by centrifugation (15 min, 10,000 × g, 4°C) were precipitated with 10% trichloroacetic acid for 30 min on ice, and the pellets were washed twice in ice-cold ethanol/ether (1:1), and the pellets were dissolved in 1 × loading buffer (54). The samples were separated by tricine-SDS-PAGE (48, 49). The cell pellets from 3 mL of bacterial culture were collected by centrifugation (15 min, 10,000 × g, 4°C). The pellets were added to 300 µL PBS and then lysed by sonication, and they were centrifuged (15 min, 10,000 × g, 4°C). The supernatants were added to 1/4 vol 5 × loading buffer and heated at 95°C for 10 min. The samples were separated by tricine-SDS-PAGE. The concentration of GPA in the supernatants secreted by Lgs^GPA was measured using the GLP-1 ELISA kit (Ruixin Co., Quanzhou, China).

After the above samples were separated by tricine-SDS-PAGE or tricine-SDS-PAGE. For Western blot analysis, electrophoresis gels were blotted onto PVDF membranes (pore size 0.22 µm), and they were blocked with 5% milk. The anti-GLP-1 (1:1,000) or anti-PDX-1 (1:1,000) specific antibodies were used to detect the expression of target proteins or peptides. HRP-conjugated secondary antibodies were utilized for ECL visualization by the fully automatic image analysis system (Biotanon, China).

Stool samples were collected as described above. They were collected at the same time points (10:00 a.m.) to minimize circadian influences on the microbiome. The compositions of the stool microbiome were sent to Majorbio Biotechnology (Shanghai, China). After DNA extraction, the primer pair 515F and 806R targeting V4 regions (31) of the 16S rRNA gene was used for amplification. The resulting amplicons were purified. The purified products were pooled to construct the Miseq library, and then Miseq sequencing. Subsequently, bioinformatic analysis was processed with the Majorbio Cloud platform. To compare the relative abundance of gut microbiota among different groups, analyses were conducted at the phylum, order, family, and genus levels (55). The raw sequencing data were deposited into the NCBI database (submission ID SUB14383914, BioProject ID PRJNA1102494; submission ID SUB14384293, BioProject ID PRJNA1139654).

All data were presented as mean ± SEM. The unpaired two-tailed Student’s t test was used for two-group comparisons. Pairwise comparisons of microbiota (dis)similarity by LDA and LEfSe analysis were compared using the Mann-Whitney U test. One-way ANOVA followed by Tukey’s multiple comparison test for multiple group comparisons using GraphPad Prism 8 Software (San Diego, CA, USA), and *P < 0.05, **P < 0.01, ***P < 0.001 represented statistically significant.

Raw reads reported in this paper are deposited in the National Center for Biotechnology Information’s Sequence Read Archive (SRA) database under the BioProject ID numbers PRJNA1102494↗ and PRJNA1139654↗.