Bacteria require phase separation for fitness in the mammalian gut

Abstract

Therapeutic manipulation of the gut microbiota holds great potential for human health. The mechanisms bacteria use to colonize the gut therefore present valuable targets for clinical intervention. We now report that bacteria use phase separation to enhance fitness in the mammalian gut. We establish that the intrinsically disordered region (IDR) of the broadly and highly conserved transcription termination factor Rho is necessary and sufficient for phase separation in vivo and in vitro in the human commensal Bacteroides thetaiotaomicron. Phase separation increases transcription termination by Rho in an IDR-dependent manner. Moreover, the IDR is critical for gene regulation in the gut. Our findings expose phase separation as vital for host-commensal bacteria interactions and relevant for novel clinical applications.

Fig. 1.. B. thetaiotaomicron Rho harbors an intrinsically disordered region required for fitness in the murine gut.

(A) The B. thetaiotaomicron Rho protein harbors a 303 amino acid-long domain absent from well-characterized homologues, such as that from E. coli. Predicted to be an intrinsically disordered region (IDR), the identified domain is immediately adjacent to Rho’s RNA binding domain (RBD). (B) AlphaFold (38, 39) prediction of the structures of the E. coli and B. thetaiotaomicron Rho proteins (source UniProt) highlighting the IDR in the latter. The N- and C-termini of the structures are indicated with N and C, respectively. (C) Relative abundance of isogenic B. thetaiotaomicron strains expressing wild-type (GT1504) or ΔIDR (GT1506) Rho at the indicated times in the gut of germ-free mice (n= 5 Swiss Webster mice); the strains were present in a 1:1 ratio in the inoculum. (D) Relative abundance of isogenic B. thetaiotaomicron strains harboring wild-type (AK310) or ΔIDR (AK312) Rho and 13 species representing the major phyla in the human gut during gut colonization in germ-free mice (n=5 C57BL/6 mice); the B. thetaiotaomicron strains were supplied at ~1:1 ratio (2×10⁸ CFU wild-type Rho-vs 1.5×10⁸ CFU ΔIDR Rho-expressing strains) in the inoculum. For (C) and (D), bacterial abundance was measured by qPCR of genomic DNA recovered from mouse fecal samples over time. SD error bars are shown.

Fig. 2.. Rho exhibits IDR-dependent phase separation in vitro.

(A) Differential Interference Contrast (DIC) microscopy of wild-type Rho and ΔIDR Rho proteins at the indicated protein concentrations reveals droplet formation by the former but not by the latter. (B) DIC microscopy of wild-type Rho, ΔIDR Rho, and IDR proteins (2.5 μM) in the presence of total B. thetaiotaomicron RNA extract (25 ng/μl). (C) DIC and fluorescence microscopy of wild-type (5 μM) and IDR (2.5 μM) proteins with fluorescently labeled (TM-rhodamine) total B. thetaiotaomicron RNA extract (12.5 ng/μl and 100 ng/μl, respectively). (D) Time-lapse DIC microscopy of wild-type Rho (10 μM) droplet fusion in the presence of total B. thetaiotaomicron RNA extract (50 ng/μl). (E) FRAP of wild-type Rho_mNeonGreen (2.5μM) with 12.5 ng/μl total B. thetaiotaomicron RNA extract at 50 mM KCl. (F) Normalized fluorescence recovery of FRAP experiment (E). Mean and SD (dashed lines) of n=28 droplets are shown. (G) Left panel: DIC microscopy of wild-type Rho protein at the indicated concentrations in the presence of total B. thetaiotaomicron RNA extract at the indicated concentrations. Right panel: Schematic of droplet formation data shown in left panel. Experiments were carried out in the presence of 150 mM KCl unless indicated otherwise. n=3 independent experiments. Scale bars: (A, B, C, G): 10 μm; (D, E): 1 μm.

Fig. 3.. Rho exhibits IDR-dependent phase separation in vivo.

(A) Fluorescence microscopy of immunostained isogenic B. thetaiotaomicron strains expressing HA-tagged wild-type Rho (AK82), ΔIDR Rho (AK86), IDR (AK393), or BT4338 (GT1481). Bacteria were grown until mid-exponential phase in minimal media with glucose (Glu), then subjected to 30 min carbon starvation (No C) and/or 5 min 5% 1,6-hexanediol (Hex). (B) Fluorescence microscopy and 3D reconstructions using cryo-ET of bacteria harvested from the gut of germ-free mice monocolonized with B. thetaiotaomicron expressing HA-tagged wild-type Rho (AK82) or ΔIDR Rho (AK86). n=3 mice per strain. (C) Quantification of protein clustering data shown in (B). (Data points represent clustering values of individual cells from three independent experiments; n=405). (D) Quantification of protein clustering data shown in (A). (Data points represent clustering values of individual cells from three independent experiments; n=75). Scale bars: Immunofluorescence in (A-B): 1 μm, cryo-ET in (B): 200 nm. p values: unpaired t-test in (C), Fisher’s LSD test was performed only for depicted pairwise comparisons (D).

Fig. 4.. IDR-dependent LLPS control of Rho transcription termination in vitro.

(A) DIC microscopy of wild-type Rho protein (2.5μM) in the presence of 50 ng/μl of mgtA RNA, roc RNA, or rocmgtA RNA, and 150 mM KCl. (B) DIC microscopy of wild-type Rho protein used in the in vitro transcription assays under LLPS-promoting (100 mM KCl) or -non-promoting (200 mM KCl) conditions. In vitro transcription of DNA templates corresponding to mgtA (C), roc (E), and rocmgtA (G) in the presence of wild-type Rho or ΔIDR Rho +/− Rho inhibitor bicyclomycin (BIC) under LLPS-promoting (100 mM KCl) or -non-promoting (200 mM KCl) conditions. Arrows indicate runoff (RO) and termination (TR) products. (D, F, H) Relative transcript abundance (termination/run off) for data presented in (C, E, G), respectively. Scale bar: 10 μm. 1 μM Rho protein was used in (B-H). Transcription from the three templates was driven by the λp_R promoter. NusG (300 nM) was included in all transcription assays. p values: Fisher’s LSD test was performed only for depicted pairwise comparisons, n≥3 independent experiments, SD error bars are shown. Bands shown in (C) are from the same gel, but during preparation of this figure two lanes between lanes 1 and 2 and one lane between lanes 6 and 7 were cropped out from the original image (see Fig. S16 for uncropped image).

Fig. 5.. B. thetaiotaomicron exhibits IDR-dependent gene expression in the murine gut.

RNA-seq analysis was performed in cecal contents of ex-germfree mice (C57BL/6 mice, n=5) monocolonized with strains harboring wild-type Rho (AK310) or ΔIDR (AK312) Rho. (A) Volcano plot of gene RNA abundance in wild-type vs ΔIDR Rho strains as log2-fold change vs - log(p). Genes >2-fold upregulated (blue) or downregulated (orange) in ΔIDR background are highlighted. (FDR-adjusted p-value <0.05). (B) Heat map of genes differentially expressed in strains harboring wild-type Rho vs ΔIDR Rho based on molecular function. (C) KEGG pathway enrichment analysis of differentially expressed in strains harboring wild-type Rho vs ΔIDR Rho. Identified pathways with enrichment p-value <0.05 and number of genes per pathway shown. (D) Log2 fold change of genes identified in (C).

Fig. 6.. LLPS of B. thetaiotaomicron Rho governs B. thetaiotaomicron gene expression in the gut.

(A) Conditions that elicit IDR-dependent LLPS of Rho (highlighted with a blue circle) increase transcription termination by wild-type Rho. Some RNAs can be terminated with low efficiency even without LLPS (i.e., mgtA, purple schematic) but other RNAs require LLPS for termination (i.e., roc, orange schematic). (B) When B. thetaiotaomicron experiences specific stress conditions, Rho exhibits LLPS in an IDR-dependent manner, which increases Rho-dependent transcription termination thereby altering global gene expression in ways that further B. thetaiotaomicron fitness in the gut.