Dietary protein increases T-cell-independent sIgA production through changes in gut microbiota-derived extracellular vesicles – Nature Communications

High protein feeding promotes high lamina propria IgA production and higher secretion of luminal sIgA

To determine how dietary macronutrients might affect sIgA and thus host-microbiota mutualism, we fed mice on one of 10 isocaloric diets with defined ratios of macronutrients in the ranges 5–60% protein, 20–75% fat, and 20–75% carbohydrate, for at least 6 weeks (Fig. 1a and Supplementary Table 1).

Fig. 1: High protein feeding promotes high lamina propria IgA production and higher luminal sIgA.
figure 1

Animals were fed one of ten isocaloric diets encompassing a macronutrient range of protein (5–60%), carbohydrate (20–75%), and fat (20–75%) for 6 weeks. a Visual representation of the composition of the diets used in this study. Each diet is represented by one circle each and their localisation on the x axis and on the y axis define their proportion of protein and of carbohydrate, respectively. The proportion of fat is indicated by the colour range as illustrated in the legend. b Contribution of macronutrient composition to small intestinal luminal sIgA (n = 7–8 per diet, quantified by ELISA) was modelled by mixture model and represented on a right-angled mixture triangle comprising of carbohydrate (y axis), protein (x axis) and fat (hypotenuse) with small intestinal content IgA concentration (ng/ml, numbers on isolines) as the response variable. Red represents high levels of sIgA while blue represents low levels of sIgA in the nutrient mixture space. Each dot represents one of the 10 diets used for modelling response surface. c Scatter bar plot of sIgA from mice fed on a high-protein (HP), high-carbohydrate (HC) or high-fat (HF) diet as determined by ELISA (n = 7 or n = 8 mice per diet for HP/HC and HF diet respectively; HP vs. HC p = 0.0009, HP vs. HF p = 0.0002). d Mixture model of plasma IgA represented on a right-angled mixture triangle and (e) corresponding scatter bar plot (n = 7 or n = 8 mice per diet for HP, and HC/HF diet respectively). f Representative immunofluorescence staining of IgA (green) in the small intestine counterstained with DAPI (blue) from mice fed on an HP, HC, or HF diet for 5 weeks. The scale bar represents 40 µm. g Total number of B220IgA+ IgA plasma cells in the small intestine lamina propria as determined by flow cytometry (n = 8 mice per group; HP vs. HC p = 0.0489, HP vs. HF p = 0.0002). h, i Gene expression of (h) Ccl28 (HP vs. HC p = 0.0187, HP vs. HF p = 0.0017) and (i) Pigr (HP vs. HC p = 0.0002, HP vs. HF p = <0.0001) in whole small intestine tissue was determined by qPCR from mice fed on either a HP, HC, or HF diet (n = 8 mice per group). Data are represented as mean ± SEM. Results represent n = 2 (ei) and n = 3 independent experiments (c). *p < 0.05, **<0.01, ***<0.001, ****<0.0001 by ordinary one-way ANOVA followed by Tukey’s multiple comparisons test.

The impact of macronutrient composition on gut luminal concentrations of sIgA was visualised using a proportion-based nutritional geometry approach, as described previously15,16. Mixture models were used to determine the effects of diet composition on luminal sIgA concentration, as quantified by ELISA (n = 8 per diet). Predicted effects of diet on sIgA were mapped onto a right-angled mixture triangle plot, where protein concentration in the diet is represented on the x axis, fat on the y axis and carbohydrate on the hypotenuse (Fig. 1b). Regions of the nutrient mixture space appearing in red demonstrate high levels of sIgA while areas in deep blue represent low levels of sIgA, and values on the isolines indicate the modelled concentration of sIgA (ng/ml). Of the 4 models fitted to sIgA concentration (relating to the first to fourth-order of the Scheffé polynomials, described by Lawson and Wilden17), model 1 was the most appropriate as indicated by the lowest Akaike Information Criterion (AIC) value (Supplementary Table 2), which reveal that dietary protein concentration was the principal predictor of sIgA, there being a clearly graded increase in sIgA with increasing protein concentration, irrespective of fat or carbohydrate content (R2 = 0.8076). Linear regression revealed that this effect was driven significantly by protein, but not by carbohydrate or fat intake, and was independent of total caloric intake (Supplementary Fig. 1a–d). Accordingly, when the three diets located at the apices of the mixture space in Fig. 1a were compared, the highest concentrations of gut luminal sIgA were observed in mice fed on the highest protein diet (HP; P60 C20 F20), compared to those fed on a diet high in fat (HF; P5 C20 F75) or high in carbohydrate (HC; P5 C75 F20) (Fig. 1c).

The shape of the response surface for plasma IgA was markedly different from that of sIgA, with plasma IgA being lowest on diets containing high-protein coupled with both low fat and low carbohydrate (Fig. 1d, e and Supplementary Table 3, R2 = 0.7243). Macronutrient composition was not a predictor of plasma IgM concentrations, as the null model was determined to be most favourable by AIC (Supplementary Table 4). This suggests that high luminal sIgA under high-protein feeding conditions was not linked to a systemic increase in basal B cell activity and that this effect was, therefore, mucosa-specific.

To determine whether elevated sIgA levels in the gut lumen of HP-fed mice were due to higher local production of IgA in the lamina propria, we performed immunofluorescence staining with anti-IgA on frozen sections of small intestine isolated from mice fed on HP, HC, or HP diets. Immunofluorescence analysis showed that HP feeding led to the highest expression of IgA in the lamina propria, compared to HC or HF feeding (Fig. 1f). Consistent with this, mice fed on an HP diet had a significantly greater number of B220IgA+ plasma cells in the small intestinal lamina propria (Fig. 1g), as determined by flow cytometry (gating strategy presented in Supplementary Fig. 1e). Prior to differentiation into IgA-producing plasma B cells, B cells are recruited to the gut by the gut epithelial chemokine CCL28. By qPCR analysis, we found that mice fed on HP diet had significantly higher intestinal gene expression of the B cell gut homing chemokine, CCL28 (Fig. 1h). Additionally, the expression of the gene encoding for pIgR, involved in the transport of sIgA to the lumen via the epithelium, was significantly increased under HP feeding conditions (Fig. 1i).

Finally, we observed that a minimum of 5 weeks on HP feeding was necessary to stably increase sIgA levels (Supplementary Fig. 1f) and that this effect was a reversible process, as sIgA levels in HP-fed animals decreased when switched to an HF diet (Supplementary Fig. 1g).

Together, these data show that protein is the major macronutrient driving sIgA production in the gut lumen and that this is reversible. An HP diet promotes the expression of the B cell gut homing chemokine, CCL28, and accordingly, a higher presence of IgA in the small intestine lamina propria. Finally, an HP diet also promotes increased expression of pIgR, the transporter of IgA in the gut lumen, consistent with the highest concentration of luminal sIgA.

High protein feeding promotes IgA production through T-cell-independent mechanisms

IgA-producing plasma cells in the lamina propria either originate from IgM-producing B cells that differentiate locally in the lamina propria, or from IgA-expressing plasmablasts induced in gut-associated lymphoid structures such as the Peyer’s patches or the mesenteric lymph nodes, which migrates back into the lamina propria to differentiate into mature plasma cells. To identify the origin of the IgA+ plasma cells that are increased under HP feeding conditions, we assessed by flow cytometry the proportion of total B cells, as well as IgA plasmablasts in both Peyer’s patches, the main site of T-cell-dependent IgA plasma cell induction, and mesenteric lymph nodes of mice fed on HP, HC or HF diets. We found that proportions of total B cells in both the Peyer’s patches (Fig. 2a) and mesenteric lymph nodes (Fig. 2b) were similar between groups. Likewise, no changes in the proportion of IgA+B220+ plasmablast were observed between groups in both the Peyer’s patches (Fig. 2c, d) and the mesenteric lymph nodes (Supplementary Fig. 2a). This suggests that HP feeding does not increase T-cell-dependent IgA production, as confirmed by the similar proportions of GL7+CD95+ germinal centre B cells in both the Peyer’s patches (Fig. 2e, f) and the mesenteric lymph nodes (Supplementary Fig. 2b). To confirm this, we depleted CD4+ T cells in mice fed on an HP diet from the start of the dietary intervention and found that mice treated with isotype or anti-CD4 depleting antibodies had comparable levels of sIgA (Supplementary Fig. 2c, d). Mice fed on a control diet had lower sIgA than HP-fed mice regardless of the depletion of CD4+ T cells. Together, our results show that HP feeding promotes high levels of small intestine sIgA via a T cell-independent pathway.

Fig. 2: High protein feeding promotes IgA production through T-cell-independent mechanisms.
figure 2

Mice were fed on either a high-protein (HP), high-carbohydrate (HC) or high-fat (HF) diet for 6 weeks. a Proportion of B220+ B cells in the Peyer’s patches (n = 8 mice per diet) and in the (b) mesenteric lymph nodes (MLN; n = 4, n = 8, and n = 7 mice per diet for HP, HC and diet, respectively) were determined by flow cytometry. c Representative flow cytometry plots representing the proportion of B220+IgA+ plasmablasts in the Peyer’s patches and (d) corresponding scatter bar graph (n = 8 mice per diet). e Representative flow cytometry plots and graph representing the proportions of CD95+GL7+ germinal centre B cells in the Peyer’s patches and (f) corresponding scatter bar graph (n = 8 mice per diet). Data are represented as mean ± SEM. Results represent n = 2 independent experiments.

High-protein feeding modulates the lamina propria cytokine environment, specifically APRIL, to favour IgA production

T-cell-independent induction of IgA CSR results from the complex interaction between gut microbes, host gut epithelial cells, immune cells and the stromal cells of the lamina propria. TLR activation in epithelial cells leads to the production of A Proliferation-inducing Ligand (APRIL) and B cell-activating factor (BAFF), the major cytokines in B cell IgA CSR, as well as thymic stromal lymphopoietin (TSLP). TSLP further amplifies this signal by inducing APRIL and BAFF production by dendritic cells18.

To determine whether diet composition affects these cytokines, we quantified their expression levels in the small intestines of mice fed on HP, HC and HF diets by qPCR. The expression of April was significantly higher under HP feeding conditions, approximately twofold greater than HC- and HF-fed mice (Fig. 3a). Similarly, Baff was highly expressed under HP feeding conditions, compared to HF feeding conditions but HC-fed mice had levels of Baff expression similar to HP-fed mice (Fig. 3b). This suggests that increased expression of April might account for elevated sIgA levels under HP feeding conditions. Like Baff, Tslp expression was significantly lower in HF-fed mice, whilst HP- and HC-fed mice had similar higher levels of expression (Fig. 3c). TSLP biases T-cell differentiation towards Th2 T cells which are characterised by their production of the cytokine IL-4. IL-4 has been shown to promote IgA CSR as well as pIgR expression. Indeed, mice fed on an HP diet had significantly elevated expression of Il4, while HF-fed mice had the lowest expression (Fig. 3d). Among other key cytokines involved in IgA CSR induction, IL-10 and TGF-beta produced by epithelial cells or dendritic cells are co-signals necessary for BAFF and APRIL to mediate their effects. We found that the expression of Tgfb was significantly higher in HP-fed mice and Il10 was significantly lower in HF-fed mice compared to HC groups (Fig. 3e, f).

Fig. 3: High protein feeding promotes a pro-IgA cytokine environment, specifically APRIL in the lamina propria.
figure 3

Mice were fed on either a high-protein (HP), high-carbohydrate (HC), or high-fat (HF) diet for 6 weeks and intestinal ileal gene expression of a April (HP vs. HC p = <0.0001, HP vs. HF p = <0.0001), b Baff (HP vs. HF p = 0.0005, HC vs. HF p = <0.0135), c Tslp (HP vs. HF p = 0.0005, HC vs. HF p = 0.0082), d Il4 (HP vs. HF p = 0.0033), e Il10 (HC vs. HF p = 0.0481) and f Tgfb (HP vs. HC p = 0.0481, HP vs. HF p = 0.0214) was determined by qPCR (ac, e n = 7 and n = 8 mice per diet for HP and HC/HF diet respectively, d n = 6 and n = 4 mice per diet for HP/HC and HF diet respectively, f; n = 7 and n = 8 mice per diet for HP/HC and HF diet respectively). Data are represented as mean ± SEM. Results represent n = 2 independent experiments. *p < 0.05, **<0.01, ***<0.001, ****<0.0001 by ordinary one-way ANOVA followed by Tukey’s multiple comparisons test.

These data demonstrate that HP feeding induces the highest expression of cytokines involved in IgA CSR, and that these key cytokines are lowest under HF feeding.

Dietary intervention significantly affects the small intestine microbiota composition

Diet is one of the most influential factors driving gut microbiota composition19, which in turn modulates host IgA responses. To characterise the impact of HP, HC and HF diet feeding on the small intestinal gut microbiota composition, we performed 16 S rRNA DNA sequencing of small intestine luminal samples. The analysis demonstrated equal sequencing reads between samples (raw data presented in Supplementary Table 5 and corresponding graphs in Supplementary Fig. 3a, b). Furthermore, rarefaction analysis revealed adequate sequencing depth (total number of sequencing reads), with a horizontal asymptote evident for all samples (Supplementary Fig. 3c), demonstrating that our sequencing depth far exceeded what is required to uncover all observations (number of unique taxa, or ASV) present in each sample.

The diversity of the gut microbiota is often used as a marker of a “healthy microbiome”. We found that bacterial richness was similar across all diets (Fig. 4a). However, HF-fed animals had lower diversity measures such as evenness and Inverse Simpson index, where Inverse Simpson index was significantly lower when compared to the HP group (Fig. 4a). By principal component analysis (PCA) of Aitchison distance, a compositionally appropriate between-sample distance metric, mice fed on the different diets had significantly distinct gut microbiota composition (Fig. 4b) as determined by PERMANOVA analysis (Supplementary Table 6). Likewise, principal coordinates analysis (PCoA) of UniFrac distances, a between-sample distance metric that considers phylogenetic relatedness, showed similar results, with HF feeding having the most distinct bacterial communities in both unweighted and weighted UniFrac PCoA analyses (Supplemental Fig. 3d, e and Supplementary Table 6).

Fig. 4: Dietary intervention significantly affects the small intestine microbiota composition.
figure 4

Mice were fed on either a high-protein (HP), high-carbohydrate (HC), or high-fat (HF) diet for 6 weeks and DNA was extracted from small intestinal content for 16 S rRNA gene sequencing (n = 8 mice per diet). Diversity of the small intestinal microbiome was determined by a Richness, Evenness and Inverse Simpson Index (HP vs. HF p = 0.0269) with *p < 0.05 by ordinary one-way ANOVA followed by Tukey’s multiple comparisons test. b Differences in the structure of the small intestinal microbiota communities were determined by principal component analysis (PCA) of Aitchison’s distance (Euclidean distance of centre-log transformed counts). Relative abundance of bacteria in the small intestine is represented at the c phylum and d genus level (showing only the top 10 genera). Relative abundance of (e) Akkermansia (HP vs. HC p = 0.032/p = 0.013, HP vs. HF p = 0.002/p = 0.003, HC vs. HF p = 0.002/p = 0.003), f Bifidobacterium (HP vs. HF p = <0.0001/p = 0.001, HC vs. HF p = <0.0001/p = 0.001), and (g) Allobaculum (HP vs. HF p = <0.0001/p = 0.001, HC vs. HF p = <0.0001/p = 0.001) with, **<0.01, ***<0.001 by Aldex2 test (Welch’s t test/Wilcoxon test with Benjamini–Hochberg corrected false discovery rate). Data are represented as mean ± SEM.

At the phylum level, HF feeding was associated with a higher ratio of Firmicutes:Bacteroidetes, as well as increased representation of Verrucomicrobia, compared to the microbiota of HP- and HC-fed mice (Fig. 4c). On the other hand, Proteobacteria was underrepresented in the microbiome of HP-fed mice (Fig. 4c). To investigate this further, we applied the ALDEx2 statistical test at the genus level (Fig. 4d) to identify differentially abundant genera (Supplementary Tables 7-9). Consistent with the over-representation of bacteria from the phylum Verrucomicrobia in HF-fed mice (Fig. 4c), Akkermansia, which can utilise host mucus as a carbon source20, was significantly higher compared to both HP- and HC- fed mice (Fig. 4e). The microbiota of the HP-fed mice was characterised by the increased abundance of bacteria from the phylum Actinobacteria (Fig. 4c) with the over-representation of bacteria from the genus Bifidobacterium (Fig. 4f). Finally, HC feeding was characterised by the over-representation of bacteria from the genus Allobacullum (Fig. 4g). Of note, we did not identify segmented filamentous bacteria or Mucispirillum in our animals (Supplementary Tables 7-9), which are two genera known to induce strong T-cell-dependent sIgA responses6.

Overall, macronutrient composition significantly impacts both the alpha and beta diversity of the small intestine microbiome, with fat appearing to be the dominant driver.

EV derived from high-protein-fed microbiota activate epithelial TLR4 and promote the expression of PIGR and APRIL

While we identified that each diet had an impact on the gut microbiota, we then investigated the mechanisms through which the HP microbiota promote the host sIgA response. TLR4 is one of the main TLRs expressed by the small intestinal enterocytes and TLR4 signalling is the major pathway promoting T-cell-independent IgA responses9. To determine whether TLR4 signalling was linked to the effect of HP in vivo, we fed wild type (WT) versus Tlr4-/- mice on HP diets for 6 weeks and measured sIgA levels in the small intestine. The absence of TLR4 abrogated the effects of HP on sIgA levels (Supplementary Fig. 4a), confirming that an HP diet promotes sIgA production via TLR4-dependent mechanisms.

Increased TLR4 signalling may be attributable to either higher expression of TLR4, increased concentration of its ligand, or increased exposure to its ligand through defects in the mucus layer or tight junction proteins. Mice fed on an HP, HC or HF diet had similar expression of TLR4 in the small intestine (Supplementary Fig. 4b), as well as comparable bacterial loads as determined by DNA concentration and total 16 S copy number (Supplementary Fig. 4c, d). We also quantified the levels of endotoxin in each diet and found low levels of endotoxin, which were comparable between the different diets (EU/ml of 0.0137, 0.053 and below the detection limit for HP, HC and HF diet, respectively). Furthermore, the expression of gut barrier-related markers Tjp1 (tight junction protein 1) and Ocln (occludin) were similar between groups (Supplementary Fig. 4e), as was the integrity of the mucus layer (Supplementary Fig. 4f) and Muc2 (mucin 2) expression (Supplementary Fig. 4g). These results indicate that increased activation of TLR4 and sIgA production under HP-feeding conditions are not linked to impaired gut integrity, higher bacterial load or higher Tlr4 expression.

Typically, the mucus layer mediates an efficient physical separation of microbes from the gut epithelium. We, therefore, hypothesised that in vivo activation of TLR4 might be mediated by smaller structures containing PAMPs, rather than through physical interaction with whole bacteria. Bacterial extracellular vesicles (EV) are small spherical structures less than 300 nm in diameter, produced by the budding of the membrane and consequently contain PAMPs. EV can contain nucleic acid, protein, metabolites and other molecules as cargo and are a key component of bacterial communication21. EV derived from Bacteroides fragilis has been shown to promote regulatory T-cell differentiation through the stimulation of TLR22, highlighting the ability of EV to traverse the mucus layer to reach the host. As such, we investigated whether EV derived from the microbiota of HP-fed mice might activate TLR and promote cytokines involved in IgA CSR.

To determine whether the small intestinal environment of HP-fed mice could differentially activate TLR4 signalling, we used the reporter cell line, HEK-Blue mTLR4, in which the intensity of TLR4 activation is assessed through a colorimetric assay. These cells were incubated with the non-bacterial fraction of small intestine luminal contents isolated from mice fed on HP, HC or HF diets. HEK cells stimulated with HP small intestine content had significantly higher levels of TLR4 activation than those stimulated with either HC or HF small intestine content (Fig. 5a), suggesting the highest potential for HP-fed mice to activate host cells.

Fig. 5: EV derived from high protein-fed microbiota activate epithelial TLR4 and promote the expression of PIGR and APRIL.
figure 5

Mice were fed on either a high-protein (HP), high-carbohydrate (HC) or high-fat (HF) diet for 6 weeks. a Small intestinal content (diluted at 1:200) (n = 5 and n = 8 mice per diet for HP/HC and HF diet, respectively) was incubated overnight with HEK-TLR4 cell line and TLR4 activation was quantified at 630 nm. (HP vs. HC p = 0.0017, HP vs. HF p = 0.0038). b Small intestine microbiota-derived EV were characterised by Nanoparticle Tracking Analysis (n = 2 per diet pooled from n = 4 mice each) and represented as an XY plot (left): 0–500 nm vs. particle number/mL), or polar plot (right): angular axis represents particle size between 0 and 300 nm and radial axis represents particle number/mL). c Small intestine microbiota-derived extracellular vesicles (EV) were incubated at physiological ratio (~2:1:1 of HP:HC:HF), or at 1:1:1 ratio (HP diluted) overnight with the HEK-Blue TLR4 cell line and TLR activation measured at 630 nm (n = 8 mice per condition; HP vs. HC p = 0.0058, HP vs. HF p = 0.0019, HP vs. HP diluted p = 0.007). df HT-29 were stimulated with small intestine microbiota-derived EV isolated from HP (2 × 109 EV per well) or HC/HF (1 × 109 EV per well) fed animals for 16 h and expression of d CCL28 (Un vs. HP p = 0.0006, HP vs. HC p = 0.0372, HP vs. HF p = 0.0105), e PIGR (Un vs. HP p = 0.0003, HP vs. HC p = <0.0001, HP vs. HF p = 0.0001) and f APRIL (Un vs. HP p = 0.0165, HP vs. HF p = 0.0036) were quantified by qPCR (n = 6 wells per condition). g HT-29 cells were incubated with 2 × 109 small intestine microbiota-derived EV from HP-fed animals for 16 hours in the presence or absence of 10 µM of NF-κB inhibitor BAY11-7082, or BAY11-7082 alone and expression of CCL28 (−EV + BAY11-7082 vs. +EV-BAY11-7082 p = 0.0103, +EV-BAY11-7082 vs. +EV + BAY11-7082 p = 0.0106), PIGR (−EV + BAY11-7082 vs. +EV-BAY11-7082 p = 0.0131, +EV-BAY11-7082 vs. +EV + BAY11-7082 p = 0.0032) and APRIL (−EV + BAY11-7082 vs. +EV-BAY11-7082 p = 0.0021, +EV-BAY11-7082 vs. +EV + BAY11-7082 p = 0.0443) were quantified by qPCR (n = 6 wells per condition). h Representative ex vivo epi-fluorescence imaging of gastrointestinal tract 6 hours post-oral administration of DiD-stained EV or PBS. i Presence of DiD-EV (yellow) and DAPI (blue) were assessed by fluorescent microscopy from sections of small intestine isolated from mice 6 hours after oral administration, with PBS as control. Scale bar represents 100 µm. j Mice were intragastrically administered with 1–3 × 1010 EV daily for 5 weeks and small intestine sIgA quantified by ELISA (n = 5 mice per group; p = 0.0319 by two-tailed unpaired t test). Data are represented as mean ± SEM. Results represent n = 2 independent (ah) and n = 1 independent experiment (i, j). *p < 0.05, **<0.01, ***<0.001, ****<0.0001 by ordinary one-way ANOVA followed by Tukey’s multiple comparisons test unless otherwise stated.

The luminal environment contains bacterial metabolites, food derivatives and bacterial EV, but bacterial EV are the only candidates containing bacterial motifs able to activate TLR4. Thus, we characterised the number and size distribution of microbiota-derived EV from mice fed on different diets by nanoparticle tracking analysis (NTA). While microbiota-derived EV isolated from the small intestine luminal content had a similar size distribution across the diets (Fig. 5b), we found a 2-fold increase in concentration under HP feeding conditions (Fig. 5b). To determine whether microbiota-EV isolated from mice fed on the different diets differentially activated TLR4, we incubated HEK-Blue mTLR4 cells with similar doses of small intestine microbiota-derived EV relative to those observed in vivo (2:1:1 ratio of HP:HC:HF). Consistent with previous results showing the highest potential of small intestine content to activate TLR4, we found that purified EV derived from HP microbiota stimulated TLR4 and to the highest extent (Fig. 5c). However, when the HEK-Blue mTLR4 cells were incubated with the same concentration of EV from each group, the activation of TLR4 was similar (Fig. 5c). Thus, the higher activation of TLR4 by microbiota-derived EV under HP feeding conditions was due to a quantitative rather than a qualitative effect.

To determine the impact of microbiota-derived EV on host sIgA production and translocation, we quantified the gene expression of CCL28, APRIL and PIGR in HT-29 cells incubated with vehicle control (PBS) or with small intestine microbiota-derived EV from mice fed on HP, HF and HC at physiological levels (HP > HC = HF). EV derived from HP small intestine microbiota significantly upregulated CCL28, PIGR and APRIL, compared to cells treated with PBS or EV derived from HC and HF microbiota, and close to significance for APRIL when compared to EV derived from HC microbiota (Fig. 5d–f). To confirm whether HP-derived EV mediated these effects via TLR signalling, we cultured HT-29 cells with EV in the presence of BAY11-7082, an inhibitor of NF-κB, the transcription factor involved in TLR signalling12. The addition of BAY11-7082 abrogated the effect of HP-EV on the expression of PIGR, APRIL and CCL28 (Fig. 5g), suggesting that EV mediated their effects via TLR signalling. We confirmed these in vitro findings using caecum microbiota-derived EV (Supplementary Fig. 4h–k), a site containing a higher density and purity of bacteria.

To determine whether microbiota-derived EV could reach the host small intestine epithelium in vivo, we isolated microbiota-EV from mice fed on normal chow, fluorescently labelled them with DiD, and administered them by gavage to another set of mice. Ex vivo imaging of gut tissue by IVIS revealed higher epi-fluorescence intensity in the gastrointestinal tract of mice receiving DiD-EV, compared to the PBS control, showing that DiD-EV readily reached the small intestine (Fig. 5h). There was also a presence of DiD-EV in both the lumen and mucosa of these mice, as determined by fluorescence microscopy (Fig. 5i). Finally, we showed that daily administration of purified microbiota-derived EV for 5 weeks in vivo could increase sIgA levels (Fig. 5j).

Together, these data highlight a close interaction between bacterial EV and host cells in vivo as well as a role for gut microbiota-derived EV on the regulation of host genes involved in sIgA regulation via TLR activation.

Succinate promotes bacterial ROS and bacterial EV production

In addition to TLR activation, the gut microbiota can also modulate host sIgA production through the generation of metabolites, such as short-chain fatty acids23,24,25. To determine whether such mechanisms are involved, we quantified microbiota-derived metabolites by NMR and focused on metabolites that are significantly increased under HP feeding conditions (Supplementary Tables 10-11). Of the major bacterial metabolites detected, we found that succinate was significantly elevated in both the small intestine luminal content (Fig. 6a), and caecal content (Supplementary Fig. 5a) in HP-fed mice, compared to mice fed on HC and HF diets. Of note, succinate was not detectable in the diets used in this study (Supplementary Fig. 5b). We incubated HT-29 cells with succinate to determine whether it could directly affect CCL28, APRIL and PIGR expression. We found that, unlike EV-stimulated HT-29, succinate could not directly induce the expression of CCL28, APRIL or PIGR expression (Supplementary Fig. 5c). As such, we hypothesised that changes to the gut metabolite environment would affect the gut microbiota, rather than the host directly to elicit host sIgA response.

Fig. 6: Succinate promote bacterial ROS and bacterial EV production and is associated with worse DSS-induced colitis.
figure 6

a The concentration of succinate in the small intestine luminal content of mice fed on a high-protein (HP), high-carbohydrate (HC) or high-fat (HF) diet for 6 weeks was quantified by NMR spectroscopy (n = 7 and n = 8 mice per diet for HP/HF and HC group respectively; HP vs. HC p = <0.0001, HP vs. HF p = <0.0001). b E. coli were grown in the presence of increasing concentration of succinate (0–10 mM) for 2 h and ROS production quantified by the conversion of 2’,7’-dichlorofluorescein diacetate to 2’,7’-dichlorofluorescein (n = 5 independent culture per condition; 0 vs. 0.1 mM p = 0.0026, 0 vs. 1 mM p = 0.0131, 0 vs 10 mM p = <0.0001, 0.01 vs. 0.1 mM p = 0.0021, 0.01 vs. 1 mM p = 0.0108, 0.01 vs. 10 mM p = <0.0001). c E. coli was grown in the presence of succinate (0, 1, or 10 mM) for 16 h and extracellular vesicles (EV) isolated and quantified by NTA (n = 5–6 per condition) and represented as XY plot (left): 0–500 nm vs. particle number/mL or polar plot (right): angular axis represents particle size between 0 and 300 nm and radial axis represents particle concentration in number/mL. d Mice received 3% DSS in drinking water for 6 days (upper) and colitis development was scored daily (lower) (n = 6 mice per group; p = <0.0001 by two-way ANOVA) and (e) at endpoint, faecal EV were isolated and quantified by Nanoparticle Tracking Analysis (n = 6 mice per condition) and represented as an XY plot (left): 0–500 nm vs. particle number/mL, or polar plot (right): angular axis represents particle size between 0 and 300 nm and radial axis represents particle concentration in number/mL. fh Mice were mice fed on a HP, HC, or HF diet for 6 weeks before induction of DSS colitis (3% DSS in drinking water for 6 days) and (f) clinical colitis development was scored daily (HP vs. HC p = <0.001, HP vs. HF p = <0.0001) and g colonic histological score assessed (HP vs. HC p = <0.0001, HP vs. HF p = <0.0001) and h representative haematoxylin and eosin-stained colonic sections (n = 6 mice per group). ik Mice were administered 100 mM pH-adjusted succinate in drinking water for 3 weeks before induction of DSS colitis (3% DSS in drinking water for 6 days) and i clinical colitis development was scored daily (p = 0.0005 by two-way ANOVA) and j histological scores quantified at endpoint on colonic section (p = <0.0001 by two-tailed unpaired t test) with k representative haematoxylin and eosin-stained colonic sections shown. (n = 7 mice per group). Scale bar represents 100 µm. Data are represented as mean ± SEM. Results represent n = 3 independent (b), n = 2 (be) and n = 1 independent experiments (fk). *p < 0.05, **<0.01, ***<0.001, ****<0.0001 and were analysed by ordinary one-way ANOVA followed by Tukey’s multiple comparisons test unless otherwise stated.

Like any cell types, bacteria can produce EV in response to stress signals such as reactive oxygen species (ROS)21. Succinate, which we found to be highly upregulated in HP-fed mice, has been shown to increase ROS in host cells such as macrophages26. High levels of microbiota-produced succinate under HP-feeding conditions might promote bacterial ROS, which would, in turn, stimulate the release of EV. To test this hypothesis, we incubated E. coli with increasing concentrations of succinate and quantified ROS. Strikingly, succinate increased ROS production in a dose-dependent manner (Fig. 6b) without affecting bacterial growth (Supplementary Fig. 5d). To determine the impact of succinate on EV production, we cultured E. coli with succinate for 16 h and found that succinate significantly increased EV production by NTA at the high concentrations (Fig. 6c). These results uncover a mechanism through which gut bacterial vesiculation is regulated by the metabolites in the environment, particularly succinate.

High bacterial EV and high succinate production under HP feeding correlates with worse DSS-induced colitis

Increased succinate has also been reported in human inflammatory bowel disease (IBD), as well as in DSS-induced colitis in mice27,28. Accordingly, we found that mice treated with DSS had higher gut bacterial EV production with similar size distribution (Fig. 6d, e). As observed under HP feeding conditions, IBD is not only characterised by high gut succinate levels27 but also increased TLR activation by gut microbial PAMPs29. We found that DSS colitis was exacerbated under conditions of high succinate, either through HP feeding (Fig. 6f–h) or direct administration of succinate in drinking water (Fig. 6i–k). Our results suggest succinate is a key mediator that can contribute to an inflammatory intestinal environment by promoting the production of microbiota EV that can activate TLR4. This model is summarised in Fig. 7.

Fig. 7: Model of high protein diet in the induction of sIgA response.
figure 7

High protein diet feeding promotes succinate production by the gut microbiota. High levels of luminal succinate induce gut bacterial cellular stress and reactive oxygen species (ROS) production, which promote vesiculation and increased production of microbiota-derived extracellular vesicles. Microbiota-derived extracellular vesicles can directly activate TLR4 expressed on the gut epithelium, activating downstream NFκB signalling that results in increased expression of APRIL, CCL28 and PIGR, which drives the T-cell-independent sIgA response. Increased TLR4 signalling also potentiates pro-inflammatory responses, resulting in more severe DSS-induced colitis. Increased intestinal succinate observed in IBD patients may contribute to disease pathology by promoting microbiota-derived extracellular vesicle production and the TLR4/NFκB signalling pathway.

Together, these results highlight a potential pathway involved in IBD severity through the increased production of the bacterial metabolite, succinate, in the induction of bacterial EV that can in turn activate TLR signalling and modulate host gene expression and inflammatory responses.

#Dietary #protein #increases #Tcellindependent #sIgA #production #gut #microbiotaderived #extracellular #vesicles #Nature #Communications

Leave a Comment

Your email address will not be published.