BefA, a microbiota-secreted membrane disrupter, disseminates to the pancreas and increases β cell mass

Abstract

Microbiome dysbiosis is a feature of diabetes, but how microbial products influence insulin production is poorly understood. We report the mechanism of BefA, a microbiome-derived protein that increases proliferation of insulin-producing β cells during development in gnotobiotic zebrafish and mice. BefA disseminates systemically by multiple anatomic routes to act directly on pancreatic islets. We detail BefA's atomic structure, containing a lipid-binding SYLF domain, and demonstrate that it permeabilizes synthetic liposomes and bacterial membranes. A BefA mutant impaired in membrane disruption fails to expand β cells, whereas the pore-forming host defense protein, Reg3, stimulates β cell proliferation. Our work demonstrates that membrane permeabilization by microbiome-derived and host defense proteins is necessary and sufficient for β cell expansion during pancreas development, potentially connecting microbiome composition with diabetes risk.

Keywords: BefA; SYLF domain; diabetes; membrane permeabilization; microbiota; β cell proliferation.

Figure 1.. The structure of BefA reveals the SYLF domain that confers its function.

(A) Ribbon diagram of BefA in two orientations, rotated by 180°. Amino terminal helices labelled H1-H5. (B) Scale 2D domain architecture map of BefA. Black rectangles denote truncation sites. (C) Overlay of BefA structure (green) with I-TASSER model of the K. aerogenes homolog (black). Residues 99 and 133 in pink sticks. (D) Scale schematics of BefA⁹⁹ and BefA¹³³ truncations. (E) Representative 2D slices from the center of primary islets of ins:gfp larvae treated with purified BefA or BefA⁹⁹. Insulin expressing cells = green, nuclei = white, Scale bar = 40 μm (F) Total β-cells, n≥15 larvae per treatment from pooled replicate experiments. Boxplot whiskers = 95%. Lowercase letters indicate post hoc means testing (Tukey). In all figures, groups with different letter designations are statistically different with a p value of ≤ 0.05; groups with the same letter are not significantly different.

Figure 2.. BefA induces membrane permeabilization

(A) Representative vesicles showing BefA induced vesiculation. Scale bars = 10 μm (B) Percentage of multi-vesicular vesicles (C) Concentration of vesicles per nL. n = 8 image stacks or ≥ 457 total vesicles per treatment in B,C. (D) Model of predicted vesiculation and fragmentation over time (E-H) Leakiness of neutral (E,F,H) or negatively charged (G) vesicles measured as percentage of maximal dye release by detergent, water control = blue, 1μM BefA = light green, 5μM BefA = dark green, 1μM BefA⁹⁹ = gray, 5μM BefA⁹⁹ = black, and 5μM BefA^R195A = brown. Vesicle diameters were less than or equal to 100 nm (E) or 400 nm (F-H). Lines trace mean of duplicate experiments, error bars = std. dev. Two-way ANOVA followed by Tukey multiple comparisons ****p<0.0001 (I&J) Measurements of Bacillus sp. (I) and Staphylococcus sp. (J) cell permeabilization, lines trace the mean of 3 replicate experiments, error bars = the min and max. In contrast to E-H, permeabilization is detected by dye quenching upon release from cells; water control =blue, 7 μM BefA = green, 3 μM Reg3α = purple, 0.08% SDS detergent control =red. (K) Representative images of Bacillus sp. incubated with mCh (top) or mCh-BefA (bottom). Magenta = mCh protein, green = DNA, DIC = differential interference contrast. Scale bar = 10 μm (L) Mean mCH fluorescence intensity on Bacillus in K, n≥33 individual bacterial cells per treatment (M) Total β-cells, n≥25 larvae per treatment. (N) Representative 2D slices from center of islets of ins:gfp larvae: CV, or Reg3γ treated. β-cells = green, nuclei = white, Scale bar = 40 μm (O) Total β-cells, n≥19 larvae per treatment. (B,C,L) Error bars = std. dev., p values results shown are Student’s unpaired T-test. (M,O) Boxplot whiskers = 95% confidence interval of the data from pooled replicate experiments. Lowercase letters indicate post hoc means testing (Tukey).

Figure 3.. BefA interacts directly with host β-cells.

(A) Total β-cells, n≥10 larvae per treatment. (B) Larval gut cells, mNG = green, insulin = magenta, nuclei = blue, DIC overlay = gray. White arrowheads = mNG puncta associated with non-insulin expressing cells. Yellow arrowheads = mNG puncta associated with β-cells. Scale bar = 50 μm (C) Representative 2D slices from larval islet explants. β-cells = green. Scale bar = 40 μm (D) Total β-cells, n≥20 islet explants per treatment. (E) Total β-cells, n≥18 larvae per treatment. BefA concentration in picograms (pg). (F&G) 4 dpf cartoon larvae, gray dotted outline denotes trunk region in images. Swim bladder (sb) = large white dashed outline, intestinal lumen = small white dashed outline, pancreas (pan) = solid white outline, ib = intestinal bulb, pi = primary islet. Scale bar = 200 μM (F) Top panel = merge of bottom panels with DIC overlay. Nkx2.2+ endocrine tissue = green, mCh-BefA = magenta. * = reflection of mCh off swim bladder (G) Larvae fed BODIPY FL-C₂ (green), DIC overlay in gray. Scale bar = 200 μm. White box (left) indicates region of right zoom inset, Scale bar = 20 μm (H&I) Total β-cells, (H) n≥17, (I) n≥6 larvae per treatment (J) Representative 2D slices through primary islets of WT and cloche mutant larvae. β-cells = magenta, vasculature = green, nuclei = blue, Scale bar = 100 μm (K) Total β-cells, n≥15 larvae per treatment. (L) The thermal stability of BefA (green) and lysozyme (gray),*T_m value of BefA = 68° C. (A, D, E, H, I, K) Boxplot whiskers = 95% confidence interval of data from pooled replicate experiments. Lowercase letters indicate post hoc means testing (Tukey).

Figure 4.. BefA’s direct activity on host beta-cells is conserved in mice.

(A) Average ratio of insulin to whole pancreas area, n≥5 pups per treatment, except ABX + BefA where n=3. (B) Representative cross sections of neonatal mouse pancreata, ABX = Antibiotic treated, nuclei = blue, insulin + tissue = brown. Because insulin staining in GF mice is faint, dashed brown outlines were added to help distinguish insulin⁺ areas. Scale bar = 500 μm (C) Average ratio of insulin to whole pancreas area of MA pups, n=12 pups per treatment, *p value for Student’s unpaired T-test result <0.05. (D) β-cell mass in milligrams (mg) of adult MA mice, n≥8 mice per treatment. **p value for Student’s unpaired T-test result <0.01. (E) Pancreas mass in grams (g) of adult MA mice, n≥8 mice per treatment. (F) Western blots against mCherry on βTC-6 cell culture treated with either mCh (29 kDa) or mCh-BefA⁹⁹ (45 kDa) proteins. Lanes 1 & 10: ladder. Lanes 2–5: mCh treated. Lanes 6–9: mCh-BefA⁹⁹ treated. Lane 11: purified mCh. Lane 12: purified mCh-BefA. Lanes 13–16: untreated. Arrow: mCh-BefA⁹⁹ band within cell membrane fraction. (G) βTC-6 cultures treated with either mNG (top) or mNG-BefA⁹⁹ (bottom), β-cells = magenta, mNG = green, nuclei = blue, DIC overlay = gray. Scale bar = 25 μm (H) Neonatal mouse islets after 48-hour treatment with mNG-BefA⁹⁹ (top and middle rows), or mNG only (bottom row). nuclei = blue, mNG = green (either punctate as pointed out by white arrowheads or outer membrane associated as denoted by yellow arrowhead), and EdU (nuclear) = green, right = merge, Scale bar = 100 μm (I-K) Total EdU labelled cells per islet, n≥22 islets per treatment. Whiskers = std. dev. (L) Average ratios of insulin to whole pancreas area. (M) Nonfasted blood glucose (mg/dL). (N) Basal serum insulin (ng/mL). (L-N) **p values for Student’s unpaired T-test results <0.01, n≥6 mice per treatment. (A,C-E, L-N) Box plot whiskers = min and max of data from pooled replicate experiments. (A, I-K) Lowercase letters indicate post hoc means testing (Tukey).