S. Typhimurium localizes at FAE in an ex vivo caecum model

It is well established from animal studies that enteric pathogens prefer the FAE as a gateway to invade the host and cause infections6,7,8. This is difficult to replicate in vitro, even with organotypic cultures that mimic some in vivo electrophysiological features29. We use our recently developed ex vivo mouse caecum model25 (Fig. 1a) to test whether differently tagged E. coli (derived from K12) and S. Typhimurium (derived from virulent 14028S) (Fig. 1b and Supplementary Table 1) show preferential targeting in the caecal epithelia. E. coli tagged with dTomato preferred the villi and avoided the FAEs (Fig. 1c–f and Extended Data Fig. 1a), while S. Typhimurium tagged with EGFP showed a preference for the FAEs, where they amassed (Fig. 1c–f and Extended Data Fig. 1a). These different tropisms were confirmed by quantifying the spatial fluorescence intensity profiles (Fig. 1g,h and Extended Data Fig. 1b). Merging channels and calculating spatial Salmonella vs E. coli ratios showed exclusive colonization of Salmonella in the FAEs (P < 0.001) (Fig. 1i and Extended Data Fig. 1c). Since targeting FAE is common among enteric pathogens, these data suggest a specific ‘sorting’ mechanism aiding S. Typhimurium targeting (Fig. 1j,k).

Fig. 1: S. Typhimurium amasses in FAE and E. coli avoids the FAE.

a, Schematic illustrating the S. Typhimurium (expressing EGFP) vs E. coli (expressing dTomato) competitive targeting experiment setup in an ex vivo mouse caecum model. A freshly isolated mouse caecum was mounted in a silicone gel plate with its luminal side facing up. Tweezers point to a Peyer’s patch (details in Methods). b, A confocal image shows the inoculum of E. coli (red) vs S. Typhimurium (green) mixture (20:1, 108 c.f.u.s ml–1 in mouse Ringer’s solution). c–f, Bright-field images of the mucosal epithelium of a mouse caecum shows the organization of FAE (white dotted enclosure) and villi (white triangle) (c), RFP fluorescence image of E. coli expressing dTomato (d), GFP fluorescence image of S. Typhimurium expressing EGFP (e) and the overlay (f). g, Enlargement of the yellow dashed area in f, showing that S. Typhimurium (green) preferably colonized FAE (white dotted enclosure), while E. coli (red) are dominantly associated with villus epithelium (white triangle). h, Normalized fluorescence profiles and green/red fluorescence ratio (thick grey line indicated by an arrow) of the line scan in g, showing difference in S. Typhimurium (green) and E. coli (red) spatial distributions between FAE (circle) and inter- and extrafollicular villus epithelium (triangle). i, Mean green/red fluorescence intensity (G/R) ratios associated with FAE or villus epithelium plotted in common logarithm (n = 6 mice, P < 0.001 by unpaired, two-tailed Student’s t-test). Box tops indicate the 75th percentile, box bottoms indicate the 25th percentile, centre lines indicate median, and whiskers indicate maximum and minimum. Dashed line indicates the ratio of 1. j, Cartoon showing microscopic view with two highlighted FAEs (dashed enclosures); S. Typhimurium in green and E. coli in red. k, Summary of the finding that S. Typhimurium (green) navigates to and accumulates in FAE, which E. coli (red) avoids and stays away from, through an unknown sorting mechanism (question marks).

Source data

Active ionic currents loop between FAE and villus epithelium

Recently, we observed a difference in TEPs between the FAE and surrounding villus epithelium25. This led us to hypothesize that a regional electrical field might influence the preferential targeting of pathogens, as S. Typhimurium and non-pathogenic E. coli migrate differently in response to an electrical field27,28. To test this, we mapped the bioelectric activities in murine caecal epithelia (Fig. 2a,b). With a vibrating probe to profile the extracellular current densities (JI)30, we recorded net outward currents in the FAE and net inward currents in the surrounding villi (Extended Data Fig. 2a–c). The extracellular currents were 0.527 ± 0.091 µA cm–2 (mean ± s.e.) and −0.606 ± 0.040 µA cm–2 in the FAE and villi, respectively (P < 0.01) (Fig. 2c). These recordings reproduced the current circuit between these functionally different epithelia that we observed in our previous study25 (Extended Data Fig. 2d). Next, to dissect the main ionic sources of the current, we perturbed the fluxes of sodium (Na+) and chloride (Cl−), two essential ions for membrane and epithelial bioelectricity20,31. We started by using broad-spectrum Na+ and Cl− channel blockers, amiloride32 and 4,4’-diisothiocyano-2,2’-stilbenedisulfonic Acid (DIDS)33, respectively. While the differences between the FAE and villi were still significant (P < 0.01, for both drugs), the current density at the villi was not significantly altered by either drug (P > 0.05, compared with the no drug control in both cases). In the FAE, the typical outward current remained in the presence of 10 µM amiloride in mouse Ringer’s solution, but it was significantly decreased, reversing to an inward current of −0.173 ± 0.060 µA cm–2 (mean ± s.e.) when bathed with 200 µM DIDS in mouse Ringer’s solution (P < 0.01, compared with the no drug control) (Fig. 2c). On the basis of these measurements, we conclude that: (1) regional ionic currents loop by entering the absorptive villi and exiting the FAE (Extended Data Fig. 2e); (2) the sustained ionic currents depend on active channel function prevailing in the mucosal epithelium because ionic currents were absent in fixed tissues (Fig. 2c); (3) the large inward current reflects the collective absorption of major electrolytes (Na+, K+, Cl− and so on) in the villus epithelium and the outward current in FAE results from chloride conductance (Fig. 2d). Thus, we further explored the chloride dependency in gut bioelectricity.

Fig. 2: Robust ionic currents emerge from ion channel activities at murine caecum epithelia.

a, Schematic of the experimental setup. Forks indicate vibrating probes and the sites where current densities were measured. b, A mouse caecum under a dissecting microscope as viewed from the luminal side, showing an intact Peyer’s Patch containing a cluster of follicles (dashed enclosures) surrounded by villi. Forks indicate vibrating probes and the sites where current densities were measured. c, Peak ionic current densities (JI) in the absence (CTRL) or presence of a general ENaC inhibitor (AMIL) or chloride channel inhibitor (DIDS). Formalin-fixed mouse caeca (‘Fixed’) served as control. Each data point represents the average of 3 to 5 FAE or villus epithelium from each mouse (n = 4, 13, 4, 4, 7, 8, 4, respectively, from left to right). **P < 0.01 by one-way ANOVA with post hoc Tukey HSD test. Box tops indicate the 75th percentile, box bottoms indicate the 25th percentile, centre lines indicate median, and whiskers indicate maximum and minimum. d, A cartoon depicts ionic flows in caecal FAE and around villus epithelium as detected by vibrating probes. Arrows indicate the flow directions and sizes are approximate. e, Peak ionic current density (JI) in the absence (CTRL) or presence of a CFTR inhibitor (CFTR(i)). Each dot represents the average of 3 to 5 FAE or villus epithelium from each mouse (n = 7, 7, 7, 9, respectively, from left to right). Box tops indicate the 75th percentile, box bottoms indicate the 25th percentile, centre lines indicate median, and whiskers indicate maximum and minimum. *P < 0.05, **P < 0.01 by one-way ANOVA with post hoc Tukey HSD test. f, Ionic flows in the presence of a CFTR inhibitor (CFTR(i)). Note the reversed ionic flow around the FAE due to reduced secretion of chloride or bicarbonate. g, Schematic illustrating critical roles of major ion channels and CFTR in generating the ionic flows around the FAE.

Source data

Regional ionic current flow is CFTR regulated

We sought to determine which Cl− channel has a major contribution to the circuit. CFTR acts as an anion channel22,34 that is involved in the osmotic balance of the mucus via the efflux of Cl− anions from the epithelia in many systems, including the intestines35. In the intestine, CFTR mediates Cl−, HCO3− (bicarbonate) and fluid secretion, with bicarbonate neutralizing luminal acidity. We hypothesized that differential activity of the CFTR could underlie the current loop in the caecum epithelia. We blocked this channel with a selective CFTR inhibitor (10 µM CFTR(inh)-172 (ref. 36) in mouse Ringer’s solution) and measured JI in both epithelia. While the villi maintained a more widely ranged (−1.066 to −3.740 µA cm–2) and robust inward current (−1.875 ± 0.371 µA cm–2, mean ± s.e.), the FAE reversed its current from outward to inward (−0.602 ± 0.432 µA cm–2) (P < 0.05, compared with the no drug control) (Fig. 2e). This pattern is similar to that observed with the use of the generic Cl− channel blocker DIDS (which also inhibits CFTR37) (Fig. 2c), showing that CFTR is a key contributor to the Cl− flux and, consequently, to the overall electric current circuit. Taken together, this means that a de facto current circuit is dependent on, or at least regulated by the CFTR-driven Cl− efflux (Fig. 2f). Although the hierarchical approach (first broad-spectrum and then specific Cl− channel inhibitors) points to a role of Cl− flux, we cannot exclude the transport of HCO3− by CFTR as an additional contributor, to some extent, to the currents detected (Fig. 2g). The reversal of the currents in FAE is also supported by the CFTR expression profile since CFTR expression is increased in mucosal epithelial cells that are near lymph nodules35.

Spatial V m patterns mirror opposing ionic flows

The ionic currents at the tissue level suggest that the enteric cells constituting different epithelia per se could have a spatially different membrane potential. Current can traverse epithelia via paracellular (between cells) and/or cellular (traversing cells) paths20. If electrogenic ion flux flows through a cell, variations in Vm will occur. To test whether FAE and villi have differential Vm, we used the voltage-sensitive dye DiBAC4(3)38. After incubation with the dye, we imaged a homogeneous polarization in the mouse caecum (Extended Data Fig. 3a,b), except for the Peyer’s patch (Extended Data Fig. 3c,d). Specifically, within the Peyer’s patch, we observed relatively positive (that is, depolarized) potentials in the villi, and relatively negative (that is, polarized) potentials in the FAEs (Fig. 3a). These differences are reliable within the same and across different mouse Peyer’s patches (P < 0.001) (Fig. 3b,c). Interestingly, the live dye evidenced well-defined intercellular zones of similar relative potential, negative at the FAEs relative to the positive villi (Fig. 3d). This indicates that while traversing the tissue during their circuit, ions move through the cells (rather than in an exclusively paracellular pathway), which alters their membrane potential. Importantly, the Vm profile matches the anionic efflux from the CFTR at the villi (Fig. 2e–g). A steady efflux of negative charges from the villi renders a more depolarized Vm and a steady influx of negative charges into the FAEs will maintain a more polarized Vm (Fig. 3e). Therefore, a regional Vm pattern (Fig. 3e) mirrors the ionic currents (Fig. 2g) as the electrogenic anionic charges flow through the cells.

Fig. 3: Regional pattern of cell membrane potentials in the FAE and villus epithelium.

a, Bright-field, live fluorescence and merged images of a mouse caecum, showing a Peyer’s patch stained with membrane potential-sensitive probe DiBAC4(3) (also see Extended Data Fig. 3). Enlargement of the yellow dashed area (bottom right panel) highlights a follicle (white dotted enclosure) surrounded by densely stained villus epithelium (white triangle), showing that the villus epithelium is electrically more positive than the FAE. b, Fluorescence intensity profile of the line scan in a (top right panel), showing a spatial difference in cellular membrane potential between FAE (circle) and inter- and extrafollicular villus epithelium (triangle). c, Relative quantitation of resting Vm of FAE and villus epithelium by DiBAC4(3) fluorescence fraction (%). A higher fraction means a more depolarized area. Quantification was based on observations from multiple FAEs and corresponding villus regions across two independent experiments (n = 4 mice, P < 0.001 by unpaired, two-tailed Student’s t-test). Box tops indicate the 75th percentile, box bottoms indicate the 25th percentile, centre lines indicate median, and whiskers indicate maximum and minimum. d, Pseudocoloured map of the Peyer’s patch region in a with a fire look-up table scale, showing the electrically negative FAEs surrounded by the relatively more positive villus epithelium. e, A cartoon to suggest how CFTR and flow of Cl− or HCO3− influence cellular membrane potential at FAE and surrounding villus epithelium (details in the main text).

Source data

Lateral bioelectric fields between FAE and villus epithelium

The observed pattern of extracellular electric currents suggested a regional lateral electrical field with the cathode in the FAEs and the anode in the neighbouring villi (Extended Data Fig. 2e). To complete the overall circuit, these extracellular currents must be balanced in subepithelial current corridors that, from Ohm’s law20, can only emerge in the presence of voltage drops underneath the FAEs and villi. To test this, we measured TEP by positioning glass microelectrodes39 in the FAE and villi of an ex vivo mouse caecum model (Fig. 4a,b). We recorded a significant gradient of inside-negative TEP in FAEs and the surrounding villi, with a larger potential in the latter (P = 0.033) (Fig. 4c,d). Similar recordings in rat ileal epithelium and Peyer’s patch reproduced this differential TEPs in the FAE and villi (P = 0.011) (Extended Data Fig. 4a–d). The polarity of the TEP is relative to the reference microelectrode, located in the bathing media; with this, we measured an inside-negative TEP. Crucially, there is a consistently larger potential in the villi than in the FAEs, demonstrating a lateral voltage drop that fuels the luminal and subepithelial currents (Fig. 4d and Extended Data Fig. 4c). As for JI, TEP is an active bioelectrical property of epithelia because they are abolished in fixed tissues (Fig. 4d). We also profiled the TEP across the Peyer’s patch and found that the interfollicular villi and villi away from the follicles have a similar TEP in both mouse (P = 0.350) (Fig. 4e) and rat (P = 0.970) (Extended Data Fig. 4d). Taken together, our extracellular and transepithelial data suggest that the subepithelial current flows from the villi towards the FAE, then exits FAE and enters the villi (Fig. 2d), completing the circuit and generating a local lateral electrical field in the gut mucosa (Fig. 4f).

Fig. 4: Spatial difference in transepithelial potential generates a lateral potential gradient between FAE and villus epithelium.

a, A cartoon depicting the TEP experiment setup. b, A mouse caecum under a dissecting microscope, showing a glass electrode (yellow arrowhead) approaching an interfollicular villus (white triangle) surrounding a follicle (white dotted enclosure). c, Typical TEP traces recorded in the FAE or villus epithelium. d, The basal TEP of both villi and FAE were negative in the mouse caeca and significantly larger in the villi than in FAE (P = 0.033, by unpaired, two-tailed Student’s t-test). Each data point represents the average of 3 to 5 FAE or villus epithelium from each mouse (n = 7). Formalin-fixed mouse caeca (‘Fixed’) served as control (n = 3), which is not subjected to statistical analysis. e, The major villus away from FAE indicated as ‘av’ in b (n = 6) and the interfollicular villus surrounding FAE indicated as ‘iv’ in b (n = 4) have similar TEPs (P = 0.350, by unpaired, two-tailed Student’s t-test). For d and e, box tops indicate the 75th percentile, box bottoms indicate the 25th percentile, centre lines indicate median, and whiskers indicate maximum and minimum. f, Schematic illustration of the spatially distinctive TEPs and the generation of a lateral bioelectric field between FAE and surrounding villus epithelium.

Source data

E. coli and S. Typhimurium galvanotax in opposing directions

The presence of a regional electrical field raises the interesting possibility of galvanotaxis-driven targeting of local enterobacteria. To test this, we first selected well-established representatives of commensal and pathogenic bacteria, Escherichia coli and S. Typhimurium, respectively25,40. Next, we subjected these bacteria to an endogenous-like electrical field in vitro (Fig. 5a,b), either sequentially (Supplementary Videos 1 and 2) or simultaneously (Supplementary Video 3). Without an electrical field, both types of bacteria migrate randomly (Fig. 5c), with their averaged directedness (<cosθ>, defined in Methods) values close to 0 (Fig. 5d). In the presence of an electrical field, E. coli cells presented a directedness of −0.995 ± 0.001 (mean ± s.e.) and S. Typhimurium a directedness of 0.994 ± 0.001 (Fig. 5d), showing a robustly biased migration of all cells towards the anode and cathode, respectively (Fig. 5c). The migratory speed (spanned distance over elapsed time) of both bacteria was around threefold faster in the presence of an electrical field (P < 0.01, E. coli or S. Typhimurium with electrical field compared with no electrical field). Intriguingly, E. coli migrated significantly faster than S. Typhimurium (5.848 ± 0.158 µm s–1 versus 4.083 ± 0.083 µm s–1 (mean ± s.e.), respectively) in the presence of an electrical field (P < 0.01) (Fig. 5e). Therefore, the O-antigen-deficient E. coli K12 and the smooth, virulent S. Typhimurium 14028S have opposing responses to the same electric cue (Supplementary Video 3).

Fig. 5: A physiological electrical field drives opposing directional migration of S. Typhimurium to the cathode and E. coli to the anode in vitro.

a, Experimental setup. b, Enlargement of the dashed area in a. c, Migration trajectories over 6 s of E. coli (red) and S. Typhimurium (green) in the absence (No EF) or presence of an electrical field (2 V cm−1) with the field polarity as shown. d,e, Quantification of directedness (cosθ: negative to the anode or left, positive to the cathode or right) (d), and migration speed (µm s−1) (e) of E. coli and S. Typhimurium in the absence (No EF) or presence (With EF) of electrical field. Each circle represents an individual cell (n = 57, 53, 64, 65, respectively, from left to right). Box tops indicate the 75th percentile, box bottoms indicate the 25th percentile, centre lines indicate median, and whiskers indicate maximum and minimum. **P < 0.01, by multiple unpaired, two-tailed Student’s t-test. f, The bacterial surface’s electrical property and flagellar propelling action determine migration direction in the galvanotaxis of E. coli and S. Typhimurium. Model based on ref. 28 and this work. Dotted arrows indicate the direction and relative size of passive electrophoretic motilities of either bacterial bodies or flagellar filaments. Solid arrows indicate the direction and relative speed of bacterial migration under a 2 V cm−1 electrical field in the shown polarity. Circular arrows indicate flagellar rotations in the counterclockwise direction propelling the bacteria along a straight trajectory.

Source data

Motile bacteria swim directionally by arranging their filament flagella on one end of the cell (the prospective back) in a bundle. These flagella rotate counterclockwise to propel the bacteria along a straight trajectory41 by generating a thrust force in the piconewton range42. To further explore the role of flagella in the galvanotaxis of S. Typhimurium, we tested a flagellar mutant strain (∆fliC, fljB::MudJ)43. Most of these mutants were non-motile and unresponsive to the applied electrical field, confirming that flagella are indeed essential for electrical field-guided galvanotaxis in S. Typhimurium27 (Supplementary Video 4). To investigate whether an applied electrical field can induce passive electrophoretic movement of the S. Typhimurium flagellar bundle ahead of the cell body, we initiated our examination with latex beads. These negatively charged beads exhibited slow migration towards the anode under our galvanotaxis experiment conditions, as observed at specific focal planes (Supplementary Video 5). Subsequent staining of Salmonella flagella, pre and post electrical field application, was carried out using antibody targeting Salmonella O- and H-antigens (Supplementary Table 1). In the absence of an electrical field, the flagella exhibited random orientation. However, upon electrical field exposure, the flagella predominantly repositioned to the anode side, trailing the bacterial body (Extended Data Fig. 6a–e). These findings not only reinforce Adler’s model28 but also elucidate a potential mechanism for directional galvanotaxis in motile Salmonella via its flagella (Fig. 5f).

To test whether other commensal bacteria can respond to an applied electrical field in vitro, we conducted galvanotaxis assays with Bacillus subtilis (B. subtilis). Surprisingly, this commensal did not perform robust galvanotaxis (Extended Data Fig. 7a and Supplementary Video 6), despite being biased towards the cathode (directedness with electrical field vs no electrical field: 0.385 ± 0.081 vs 0.117 ± 0.095 (mean ± s.e.), P = 0.034) (Extended Data Fig. 7b). Unlike S. Typhimurium that migrated straight towards the cathode (Fig. 5d) with increased speed (Fig. 5e), the migration speed of B. subtilis did not increase (electrical field vs no electrical field: 0.567 ± 0.030 vs 0.558 ± 0.028 µm s–1 (mean ± s.e.), P = 0.839) (Extended Data Fig. 7c) and remained one order of magnitude slower than that of S. Typhimurium (Fig. 5e). Hence, B. subtilis, one of the most abundant commensals in the human gut, do not undergo robust directional migration when exposed to a small electrical field.

The bacterial galvanotaxis is not, or at least not fully due to electrophoresis since both E. coli and S. Typhimurium are negatively charged and they migrated to the anode in a higher electrical field when fixed by formaldehyde28,44; neither is it due to fluid flow because our experiments were conducted in sealed microfluidic chambers (Fig. 5b). On the basis of these data, we hypothesize that the E. coli K12 and the enteric pathogen S. Typhimurium 14028S may act and move differentially in the vicinity of the intestinal epithelia in response to an existing, naturally occurring bioelectrical signal.

S. Typhimurium galvanotaxis is independent of chemotaxis

Previous research has shown that S. Typhimurium invades the murine ileum Peyer’s patches by detecting gradients of host-derived chemoattractants. This process was contingent upon the flagellar apparatus and specific chemotaxis protein receptors11,45. To probe the role of chemotaxis in galvanotaxis-facilitated migration, we executed an in vitro galvanotaxis assay with a chemotaxis-deficient mutant in the background of the S. Typhimurium 14028S, specifically lacking the methyl-accepting chemotaxis protein CheB46. This mutant displayed marked directional migration towards the cathode (Supplementary Video 7), aligning with the movement pattern of the wild-type strain (Supplementary Video 2). This suggests that CheB is not crucial for bacterial galvanotaxis27 and that the mutant might still navigate effectively to the FAE. Subsequently, mouse caecum explants were exposed to a 1:1 mixture of E. coli (K12, dTomato-expressing) and an S. Typhimurium wild-type strain (14028S, EGFP-expressing). A comparable experiment was set up with another S. Typhimurium wild-type strain (mCherry-expressing) against either the cheB mutant (EGFP-expressing) or a non-motile flagellar mutant (EGFP-expressing) as controls. At 30 min post incubation, epithelium-associated bacteria were recovered from FAEs isolated using fine biopsy punches. Quantitative analysis revealed that the S. Typhimurium recovery rate was about five times that of the E. coli from the FAE (P = 0.063) (Extended Data Fig. 8a). The flagellar mutant exhibited a 20-fold lower recovery than its wild-type counterpart (P = 0.021) (Extended Data Fig. 8a), underscoring the pivotal role of flagella in congregating at the FAE. Importantly, the cheB mutant recovery rate mirrored that of the wild-type S. Typhimurium (P = 0.937) (Extended Data Fig. 8a), suggesting that, unlike flagella, CheB is inconsequential in this bioelectricity-driven event in our ex vivo setup. The competitive index analysis and its subsequent data comparison (P = 0.019) (Extended Data Fig. 8b) further substantiate this notion.

CFTR modulates S. Typhimurium localization

Having revealed the CFTR-regulated regional electrical fields, demonstrated the opposing directional migration of E. coli and S. Typhimurium under physiological electrical fields, and established that S. Typhimurium galvanotaxis operates independently of chemotaxis, we next investigated whether disrupting the endogenous electrical fields would affect S. Typhimurium localization in the FAE. To test this, we performed a competitive tropism assay in our ex vivo mouse caecum model using a mixture of differentially tagged E. coli K12 and S. Typhimurium 14028S (Fig. 6a). In unperturbed endogenous electrical fields, confocal microscopy revealed that dTomato-tagged E. coli is predominantly localized in the villi, avoiding the FAEs, while EGFP-tagged S. Typhimurium showed a preference for the cathodic FAEs (Fig. 6b). Quantification of the spatial distribution of S. Typhimurium and E. coli via a green/red fluorescence intensity ratio confirmed their respective preferences (P < 0.01, Fig. 6d). Notably, the bacterial tropism towards anodic villi and cathodic FAEs aligns with the lateral potential gradient (Fig. 4f), the Vm pattern (Fig. 3d) and the robust directional galvanotaxis in vitro (Fig. 5c). We validated that the bioelectricity-modulated bacterial targeting is an active biological process, by including fluorescently labelled latex beads in the inoculum in a 1:2 (bead/bacteria) ratio (Extended Data Fig. 9a). The beads showed a relatively homogeneous distribution in the FAEs and villi (Extended Data Fig.