Salmonella enterica Serovar Typhimurium Exploits Cycling through Epithelial Cells To Colonize Human and Murine Enteroids

Early Salmonella Typhimurium infection involves luminal growth and IEC invasion. The steps of IEC invasion comprise (i) Salmonella Typhimurium flagellar motility, (ii) binding and TTSS-1-dependent invasion, (iii) intraepithelial replication within an SCV and/or in the cytosol, (iv) inflammasome-driven expulsion of infected IECs, and in some cases (v) bacterial breaching of the epithelial barrier. To validate enteroid microinjection as a model for these infection cycle events, human jejunum enteroids were established, injected with wild type (WT) Salmonella Typhimurium (SL1344) constitutively expressing mCherry (rpsM-mCherry) (60), and imaged by time-lapse microscopy (Fig. 1A; time-lapse movies related to the figures are available at https://doi.org/10.17044/scilifelab.12998570).

Microinjected Salmonella Typhimurium maintained flagellar motility, reached the epithelial surface within seconds after injection (Fig. 1A), and engaged in near-surface swimming (Fig. 1B). Tracking of individual bacteria revealed approximately similar levels of motile bacteria and average swimming speeds in the enteroid lumen and in the inoculum (Fig. 1B and C). Following injection, we observed a gradual increase in the numbers of fluorescent bacteria in the enteroid lumen, highlighting luminal replication over several hours post-injection (p.i.) (Fig. 1D). Luminal expansion with broadly similar kinetics was observed in spheroid human enteroids (Fig. 1D) and in morphologically more elaborate murine (C57BL/6 jejunum origin; two independently established enteroid lines) and human enteroids with clearly distinguishable crypt and villus domains (see Fig. S1 in the supplemental material). However, since microinjection of multilobulated enteroids was technically more challenging and quantification was more accurate in sphere-shaped enteroids with a clearly visible lumen, we used the latter for further analyses.

Salmonella Typhimurium invasion foci within IECs were detected shortly after injection and continued to accumulate during the following hours (Fig. 1E, left). To investigate the timing of SCV establishment inside IECs, we injected enteroids with Salmonella Typhimurium expressing SPI-2-inducible green fluorescent protein (GFP) upon host cell invasion (pssaG-GFP) (15). The appearance of GFP-positive intraepithelial foci at ∼3 to 4 h p.i. confirmed successful SCV establishment in some IECs (Fig. 1E, right; also, see Fig. S2A and B), and the expansion of these foci demonstrated intraepithelial replication (Fig. 1F). Applying a reporter induced upon access to the cytosolic metabolite glucose-6-phosphate (puhpT-GFP) (18), we assessed the existence of Salmonella Typhimurium in the IEC cytosol. In agreement with recent observations in primary IECs by others (53), puhpT-GFP-positive foci were few and barely detectable in the enteroid model (Fig. S2C). Furthermore, restriction of intraepithelial Salmonella Typhimurium included the frequent expulsion of bacterium-containing IECs into the enteroid lumen, detectable as early as ∼60 to 90 min p.i. (Fig. 1G; Fig. S2B and D). While the Salmonella Typhimurium population was contained within the epithelial lining during the first hours (Fig. 1D), breaching of the epithelial barrier was observed at later time points (Fig. 1H). Bacterial escape as a rule originated from an IEC invasion focus (Fig. 1H) and could be seen as a simple proxy for systemic Salmonella Typhimurium spread in the enteroid model.

Microbe-host interactions in organotypic models can exhibit considerable donor-to-donor variation (61–63). To exclude donor-specific effects, the Salmonella Typhimurium infectious cycle was replicated in a second independently established human enteroid line (Fig. S3). In addition, we repeated the experimental series in wild type murine enteroids. Also in this model, flagellar motility, near-surface swimming, luminal expansion, IEC binding and invasion, infected-IEC expulsion, and breaching of the epithelial barrier could be robustly detected and occurred with kinetics similar to in human enteroids (Fig. S2E and F; Fig. S4). Intraepithelial replication was observed to a lesser extent than in human enteroids, indicating that bacterial restriction within murine IECs might be even more efficient (Fig. S4E). Altogether, our findings confirm that human and murine enteroid microinjection recapitulate the predominant steps of the Salmonella Typhimurium gut infection cycle. The high temporal resolution enables dissection of how distinct events in the infection cycle are causally linked to each other.

We next addressed the contribution of Salmonella Typhimurium virulence factors to enteroid colonization. TTSS-1 encoded by Salmonella Typhimurium SPI-1 mediates translocation of effector proteins into host cells to promote bacterial invasion (16). To assess the impact of TTSS-1, human enteroids were injected with either fluorescently tagged Salmonella Typhimurium WT or a mutant deficient for a critical TTSS-1 structural component (ΔinvG) (64) at different bacterial doses (low, <20 Salmonella Typhimurium organisms per enteroid; intermediate, 20 to 50 per enteroid; high, >50 per enteroid). Microinjected enteroids were followed by time-lapse microscopy, and colonization was quantified based on the increase in bacterial fluorescence within the enteroids over time.

Neither adverse effects of microinjection on enteroid health nor an increase in enteroid-associated fluorescence could be observed in mock-injected controls (Fig. 2A and B; Fig. S3A). This confirmed that an increase in fluorescence can be attributed to bacterial population expansion. Over time, injection of either Salmonella Typhimurium strain resulted in diffuse cytopathic effects, such as a variable degree of enteroid shrinkage, which was most consistently noted for Salmonella Typhimurium WT high-dose injections (Fig. 2C, E, and G). The kinetics of enteroid colonization for both Salmonella Typhimurium strains increased with the dose of bacteria injected (Fig. 2C to H; see Fig. S5A for individual curves). Notably, however, Salmonella Typhimurium WT expanded quickly and by 9 h p.i. had filled the enteroid lumen even upon low-dose injection. In contrast, the Salmonella Typhimurium ΔinvG mutant population expanded significantly more slowly at all doses tested, and final fluorescence intensities reached only ∼1/3 to 1/2 of the level observed for Salmonella Typhimurium WT at 9 h p.i. (Fig. 2C to H). A similar attenuation of Salmonella Typhimurium ΔinvG colonization capacity was observed also in murine enteroids (Fig. S5B and S6). These results were unexpected, since this TTSS-1-deficient strain has no detectable growth defect in rich medium (65) and since TTSS-1-deficient Salmonella Typhimurium can colonize the gut lumen of permissive mice (11, 15), although prior in vivo studies did not allow the high temporal resolution of the present experiments. As anticipated, we found that Salmonella Typhimurium ΔinvG did not invade IECs upon enteroid microinjection (Fig. S7A). Moreover, a Salmonella Typhimurium strain maintaining an intact TTSS-1 but lacking four TTSS-1 effectors (SipA, SopB, SopE, and SopE2; referred to here as Salmonella Typhimurium Δ4) (66) both was incapable of invading IECs (Fig. S7B and C) and exhibited attenuated enteroid colonization kinetics similar to that of Salmonella Typhimurium ΔinvG (Fig. S8). These observations led us to postulate that TTSS-1 activity boosts Salmonella Typhimurium colonization kinetics in enteroids by enabling IEC invasion.

Flagellar motility is another key virulence determinant during early Salmonella Typhimurium infection of the intestine, allowing the bacterium to penetrate the mucus layer, reach the epithelial surface, and engage in TTSS-1-dependent IEC invasion in vivo (3, 4, 6). To assess the involvement of flagellar motility in enteroid colonization, we injected human and murine enteroids with fluorescent Salmonella Typhimurium WT or a nonmotile mutant lacking the flagellar motor protein MotA (ΔmotA) but maintaining structurally intact flagella and quantified bacterial colonization kinetics.

The lack of motility was verified by tracking Salmonella Typhimurium WT and Salmonella Typhimurium ΔmotA bacteria within the enteroid lumen. The quantification of Salmonella Typhimurium WT motility recapitulated our previous findings (Fig. 3A to C, WT; compare to Fig. 1C and Fig. S4B, enteroid), whereas no bacteria moving at a speed greater than 3 μm/s were detected for Salmonella Typhimurium ΔmotA (Fig. 3A to C). Strikingly, however, no difference in enteroid colonization kinetics (based on the quantification of bacterial fluorescence intensities) was observed between the two strains at any bacterial dose or time point in human or murine enteroids (Fig. 3D and E; also, see Fig. S9 for individual curves). These findings surprisingly suggest that flagellar motility is not required for efficient bacterial colonization of enteroids.

As we had found TTSS-1 to both drive IEC invasion and boost enteroid colonization, hence linking these two phenomena, we speculated that Salmonella Typhimurium ΔmotA was able to successfully invade IECs after reaching the epithelial surface independent of flagellar motility. Therefore, we analyzed the vertical distribution of the bacterial population within the enteroid lumen by confocal microscopy. Flagellar motility of the Salmonella Typhimurium WT enabled rapid access to the epithelium (Fig. 1A to C), and consequently, no enrichment of bacteria at any specific location could be observed (Fig. 3F). In sharp contrast, Salmonella Typhimurium ΔmotA displayed a clear accumulation atop the epithelium at the bottom plane of the enteroid, within minutes after inoculation (Fig. 3F). This local enrichment was further enhanced during the following ∼2 h p.i. (Fig. 3F). Finally, analysis of the vertical distribution of SPI-2-positive invasion foci following microinjection with Salmonella Typhimurium ΔmotA pssaG-GFP confirmed IEC invasion at the bottom plane at 4 h p.i. (Fig. 3F). This suggests that nonmotile Salmonella Typhimurium can reach the epithelial surface swiftly by gravitational sedimentation within enteroids, thereby allowing IEC invasion in the absence of functional flagella.

We next investigated how TTSS-1 and flagellar motility impact breaching of the epithelial barrier. In our setup, this could be approximated by the basolateral escape of bacteria previously confined within the boundaries of the enteroid (Fig. 1H; also see Fig. S3F and S4G). To this end, human and murine enteroids were microinjected with fluorescently labeled Salmonella Typhimurium ΔinvG, Salmonella Typhimurium Δ4 (i.e., ΔsipA ΔsopBEE2), or Salmonella Typhimurium ΔmotA, in each case with parallel Salmonella Typhimurium WT injections into enteroids within the same dome as controls. Breaching of the epithelial barrier was defined as the point in the time-lapse series where bacterial escape from the enteroid occurred in at least one location (using the Kaplan-Meier model for analysis) (Fig. 4). Bacterial breaching as a rule occurred earlier and was more distinct in murine than in human enteroids (Fig. 4).

TTSS-1-deficient Salmonella Typhimurium ΔinvG breached the epithelial barrier less frequently and at later time points p.i. than the Salmonella Typhimurium WT at all doses, with median times of confinement on average ∼1.5 to 4 h longer (Fig. 4A and B). While this difference constituted a trend in human enteroids (Fig. 4A), it was strikingly evident in the more sensitive murine enteroid model, and particularly upon high-dose injection (Fig. 4B). Salmonella Typhimurium Δ4 again behaved broadly similarly to Salmonella Typhimurium ΔinvG (significantly longer time of confinement) (Fig. 4C), suggesting that both the TTSS-1 apparatus itself and the TTSS-1 effectors that drive IEC entry promote Salmonella Typhimurium breaching of the epithelial barrier in enteroids. Intriguingly, Salmonella Typhimurium ΔmotA showed an even more pronounced delay in the time of confinement (∼2 to 6 h longer than for Salmonella Typhimurium WT in parallel infections) and a markedly reduced frequency of bacterial escape events (Fig. 4D and E). These findings held true in both human and murine enteroids and at all three doses tested (Fig. 4D and E). Hence, our combined results suggest that while only TTSS-1 activity boosts luminal colonization, both TTSS-1 activity and flagellar motility promote breaching of the epithelial barrier in Salmonella Typhimurium-microinjected enteroids.

To address the causal relationship between TTSS-1-dependent IEC invasion and luminal colonization, we took advantage of the temporal resolution of our microinjection model to determine whether and how the invasive bacterial population contributes to luminal growth. To that end, enteroids were injected with Salmonella Typhimurium WT SPI-2–GFP, and the fate of GFP-positive Salmonella Typhimurium was followed over time. In accordance with our earlier results (Fig. 1E), GFP-positive foci could be detected within IECs at ∼3 to 4 h p.i. (Fig. 5A). Following intraepithelial expansion, SPI-2–GFP-positive Salmonella Typhimurium began reemerging as distinct packages in the enteroid lumen through IEC expulsion (Fig. 5A). Eventually, intraepithelial bacteria released from expelled IECs populated the entire enteroid lumen, filling it completely by ∼6 to 8 h p.i. (Fig. 5A). Quantification of epithelial and luminal fluorescence intensities individually revealed significant bacterial expansion in both compartments (Fig. 5B). Strikingly, the luminal fluorescence intensity was found to exceed the epithelial one at 8 h p.i. (P = 0.015) (Fig. 5B). This indicates a contribution of the intraepithelial Salmonella Typhimurium population to luminal colonization in enteroids.

To further pinpoint the extent to which reemergence of IEC-residing Salmonella Typhimurium can contribute to luminal colonization, enteroid coinfections were performed with a 1:1 mixture of constitutively fluorescent Salmonella Typhimurium ΔinvG (carrying rpsM-mCherry) and Salmonella Typhimurium WT carrying the SPI-2–GFP construct. As in the single-strain infections, SPI-2–GFP-positive Salmonella Typhimurium WT organisms were primarily found within the epithelium at early time points (2 to 4 h p.i.), whereas the Salmonella Typhimurium ΔinvG strain was confined to the enteroid lumen at all time points of the imaging series (Fig. 5C to E). Salmonella Typhimurium ΔinvG fluorescence gradually increased in the lumen (Fig. 5E). Importantly, Salmonella Typhimurium WT SPI-2–GFP fluorescence in the lumen accumulated with similar kinetics and potency (Fig. 5E).

At least two mechanisms might account for the contribution of Salmonella Typhimurium IEC invasion to enteroid lumen colonization. The cycles of IEC invasion, intraepithelial expansion, and reemergence through IEC expulsion (Fig. 5A to E) could by themselves expand the total luminal pool of Salmonella Typhimurium. Alternatively, some consequence of TTSS-1-dependent IEC invasion (e.g., luminal accumulation of IEC debris or hampered antimicrobial peptide [AMP] secretion) could change the conditions for bacterial growth in the lumen itself, which would benefit any Salmonella Typhimurium organisms residing in that compartment. To distinguish between these two scenarios, we injected enteroids with a genetically tagged mixed consortium (65), consisting of three invasive Salmonella Typhimurium WT strains (tags A to C) and three noninvasive Salmonella Typhimurium ΔinvG strains (tags D to F), along with a fluorescently (rpsM-mCherry) labeled Salmonella Typhimurium ΔinvG tracer strain for visualization (Fig. 5F). Luminal Salmonella Typhimurium organisms were extracted at ∼16 h p.i., and the population structure was analyzed by real-time quantitative PCR (qPCR), using primers specific for tags A to F.

The enteroid colonization dynamics of the fluorescent Salmonella Typhimurium ΔinvG tracer strain were broadly similar to what was previously observed for the corresponding single-strain infections (Fig. 5G; Fig. S10A), and plating to determine the CFU further confirmed successful expansion of the genetically tagged luminal Salmonella Typhimurium population (Fig. S10B). Bacterial escape was observed in ∼45% of the enteroids and, as anticipated, mainly involved Salmonella Typhimurium WT tag A to C strains, but expansion of the escaper population was efficiently suppressed by gentamicin added to the surrounding medium (Fig. S10A to C). Importantly, the population structure of the luminal Salmonella Typhimurium population revealed a consistent increase in the abundance of all Salmonella Typhimurium WT strains by 16 h p.i., whereas the relative strain abundances in mock infection samples (overnight growth of the mixed consortium in IntestiCult medium in the absence of enteroids) remained unchanged (Fig. 5H). These results replicate the ∼2- to 3-fold colonization advantage observed for the Salmonella Typhimurium WT strain in single-strain injections (compare Fig. 5H with Fig. 2D, F, and H). The fact that the growth advantage of Salmonella Typhimurium WT was maintained in coinfections with lumen-confined Salmonella Typhimurium ΔinvG refutes the idea that an altered luminal environment can explain the link between TTSS-1 and luminal colonization. Altogether, our findings demonstrate instead that a cycle(s) of TTSS-1-dependent IEC invasion, intraepithelial replication, and reemergence through IEC expulsion potently complements planktonic Salmonella Typhimurium growth for colonization of the enteroid lumen.

Human jejunal enteroids were generated from tissue resected in the course of bariatric surgery, subsequent to each subject’s giving informed consent. Personal data were pseudonymized before further processing of tissue specimens in the laboratory. The procedures were approved by the local governing body (Etikprövningsmyndigheten, Uppsala, Sweden) under license number 2010-157 with addendum 2010-157-1 (2018-06-13). The maintenance of laboratory mice and experimentation involving murine intestinal tissue were approved by the local governing body (Uppsala Djurförsöksetiska Nämnd, Uppsala, Sweden) under license number C6/16.

All strains used in this study had a Salmonella enterica serovar Typhimurium SL1344 background (SB300; streptomycin resistant) (74). Besides the wild type (Salmonella Typhimurium WT), the previously described ΔinvG (64) and ΔsipA ΔsopBEE2 (referred to here as Salmonella Typhimurium Δ4) mutants (66) were used. The ΔmotA mutant was generated via transfer of a previously described deletion (75) from a Salmonella Typhimurium 14028 strain (C1172) to the SL1344 background by P22 transduction. Chloramphenicol-resistant, isogenically tagged Salmonella Typhimurium WT (tags A to C) and Salmonella Typhimurium ΔinvG (tags D to F) strains were used in an earlier study (65). The pFPV-mCherry (rpsM-mCherry; Addgene plasmid number 20956) (60), pM975 (pssaG-GFPmut2) (15, 22), and pZ1400 (puhpT-GFP) (18) reporter plasmids were previously used and validated. For infections, Salmonella Typhimurium cultures were grown overnight for 12 h in LB–0.3 M NaCl (Sigma-Aldrich) with appropriate antibiotics, followed by subculturing in the same medium without antibiotics at a 1:20 dilution for 4 h. Prior to microinjection, the inoculum was reconstituted in antibiotic-free complete human or mouse IntestiCult medium (StemCell) at a concentration of 5 × 10⁸ to 1 × 10⁹ CFU/ml.

Human jejunal enteroid cultures were established from tissue resected during bariatric surgery performed on otherwise healthy subjects. After resection, the tissue was transported in ice-cold phosphate-buffered saline (PBS; Gibco) until it was opened and fastened to a Styrofoam cushion. Particulate material was removed by washing with cold PBS, and surgical scissors were used to separate the mucosa from the muscle layer. An ∼6- by 6-mm tissue piece excised from the mucosa was washed several times with PBS, minced with surgical scissors, and passed through a 1-ml pipette tip. The minced mucosa was centrifuged and washed once more with cold PBS before incubation in gentle cell dissociation reagent (StemCell) with gentle shaking on ice for 30 min. Following another centrifugation step and resuspension in cold Dulbecco’s modified Eagle medium (DMEM)–F-12 (Gibco) supplemented with 0.25% bovine serum albumin (BSA; Gibco), epithelial crypts were detached by vigorous pipetting. When the resulting suspension had been passed through a 70-μm cell strainer, the crypt concentration was enumerated. The number of crypts required to yield a density of 250 to 750 crypts/dome were centrifuged, resuspended in Matrigel (Corning; product number 356230)–25% DMEM–F-12 and seeded as 50-μl domes in multiwell plates. After solidification at 37°C for 10 min, complete human IntestiCult supplemented with 10 μM Y-27632 (Sigma-Aldrich) and 100 U/ml penicillin-streptomycin (PenStrep; Gibco) was added. Cultures were maintained in a 5% CO₂ atmosphere at 37°C, and after the first 2 days in culture, Y-27632 was omitted. From then onward, the medium was exchanged every 3 to 4 days. At day 8 to 10 after establishment, the best-looking enteroids were expanded further using the procedure for continuous enteroid subculturing (see below). Murine enteroids of C57BL/6 jejunal origin were established according to a previously published protocol (76), embedded in 50-μl Matrigel domes containing ∼40% complete mouse IntestiCult, and overlaid with IntestiCult supplemented with PenStrep after solidification. Newly established enteroids were frozen at passage 2 in DMEM–F-12–10% fetal bovine serum (FBS; Thermo Fisher Scientific)–10% dimethyl sulfoxide (DMSO; Sigma-Aldrich) and cryopreserved in liquid nitrogen gas phase.

For maintenance culturing, both newly generated human and murine enteroids, as well as previously described murine jejunal enteroids (C57BL/6 background) (76), were thawed from cryopreserved stocks and embedded in 50 μl Matrigel domes as described above. After the domes had been allowed to solidify for 10 min at 37°C, they were overlaid with complete human or mouse IntestiCult supplemented with PenStrep. During the first 2 to 3 days after thawing, the culture medium additionally contained 10 μM Y-27632. Cultures were maintained at 37°C in 5% CO₂, and fresh medium was added every 2 to 3 days. Enteroids were passaged at a 1:3 to 1:12 splitting ratio every 5 to 10 days by breaking up the Matrigel domes through extensive pipetting and incubation in gentle cell dissociation reagent with rocking at 20 rpm. The extracted enteroid fragments were washed once in DMEM–F-12–0.25% BSA and re-embedded in Matrigel domes as described above. Enteroids from passages 4 to 30 were used for experimentation.

Salmonella Typhimurium microinjection into human and murine enteroids was performed 4 to 5 days (human enteroids) or 2 to 3 days (murine enteroids) after enteroids had been passaged and embedded in 50-μl elongated, loaf-shaped, ∼90 to 100% Matrigel domes seeded in a 35-mm glass-bottom dish (no. 1.5 coverslip; 20-mm glass diameter, uncoated; MatTek P35G-1.5-20-C). The culture medium was replaced with antibiotic-free complete human or mouse IntestiCult prior to infection. For microinjections of barcoded Salmonella Typhimurium consortia, the medium was replaced with complete human IntestiCult containing 6 μg/ml gentamicin. Microinjection needles were generated from 1.0-mm filamented glass capillaries (World Precision Instruments; no. BF100-78-10; Borosilicate, 1 mm wide, 100 mm long, with filament) using a micropipette puller (Sutter Instruments; P-1000; settings: heat = ramp + 5; pull = 60; velocity = 80; delay = 110; pressure = 200) beveled at a 30° angle on a fine-grit diamond lapping wheel. Needles were loaded with the prepared inoculum by fluidic force and mounted on a microinjector (MINJ-FLY; Tritech Research) in a micromanipulator (uMP-4; Senapex). A 0.02- to 0.2-s air pressure pulse was applied to inject enteroids with the respective dose of Salmonella Typhimurium. The infectious dose was in each case estimated by eye, based on the number of fluorescent particles emerging from the needle.

For barcoded-consortium infections, bacterial subcultures of three tagged Salmonella Typhimurium WT (tags A to C) and three tagged Salmonella Typhimurium ΔinvG (tags D to F) strains, as well as one fluorescently labeled Salmonella Typhimurium ΔinvG strain (rpsM-mCherry), were prepared as described above, mixed at a 1:1:1:1:1:1:1 ratio, and reconstituted in antibiotic-free complete human IntestiCult. Microinjection of ∼40 enteroids per replicate with a total number of ∼200 to 1,000 Salmonella Typhimurium organisms per enteroid was performed as described above. Microinjected enteroids were incubated at 37°C and 5% CO₂ for ∼16 h. For the mock-injected sample, 1 μl of the mixed consortium inoculum was added to a 35-mm glass-bottom dish containing 2 ml antibiotic-free complete human IntestiCult, and the dish was incubated in parallel with the microinjected enteroids. Following overnight incubation, the medium surrounding the Matrigel dome was removed from the dish and saved for enrichment of the escaped population. The dome containing the injected enteroids was washed three times in prewarmed DMEM–F-12 before the enteroids were extracted from the Matrigel by gentle pipetting in ice-cold DMEM–F-12–0.25% BSA using cut pipette tips. After two washes in ice-cold DMEM–F-12–0.25% BSA, the enteroids were broken up mechanically by vigorous pipetting, and luminal bacteria were harvested in 3 ml LB containing 12.5 μg/ml chloramphenicol (Cm; Sigma-Aldrich). For CFU plating, bacterial suspensions extracted from the enteroid lumen and the microinjection supernatant were serially diluted and plated on LB agar containing 12.5 μg/ml Cm.

For tag quantification, the bacterial populations recovered from the organoid lumen and microinjection supernatant, as well as a 1:3,000 dilution of the inoculum and mock-injected samples, were enriched for 15 h in 3 ml LB containing 12.5 μg/ml Cm. Half of the enrichment culture was used for genomic DNA extraction using the GenElute bacterial genomic DNA kit (Sigma-Aldrich). Quantitative PCR analysis with the Maxima SYBR green/ROX qPCR master mix (2×) (Thermo Fisher Scientific) was performed on a Bio-Rad CFX 384 instrument using 9 ng of genomic DNA (gDNA) and tag-specific primers as previously described (65, 77). The relative abundances of all strains were calculated as 2^−ΔCT, where ΔC_T was defined as the difference in cycle threshold (C_T) value compared to that of Salmonella Typhimurium WT tag A in the same sample (i.e., the relative abundance of Salmonella Typhimurium WT tag A in each sample was set to 1). To express the population structure as a percentage, the summed relative abundances of all strains in each sample were set to 100%.

Microinjected enteroids were imaged on a custom-built microscope based on an Eclipse Ti2 body (Nikon), using a 60×, 0.7 numerical aperture Plan Apo Lambda air objective (Nikon) and a back-lit sCMOS (scientific complementary metal oxide semiconductor) camera with a pixel size of 11 μm (Prime 95B; Photometrics). The microscope chamber was maintained at 37°C in a moisturized 5% CO₂ atmosphere. Bright-field images were acquired using differential interference contrast (DIC), and fluorescence was imaged using the excitation light engine Spectra-X (Lumencor) and emission collection through a quadruple band pass filter (89402; Chroma). For bacterial tracking, the microinjected enteroids as well as a 1:700 to 1:1,000 dilution of the inoculum were imaged at 300- to 500-ms intervals for 20 frames in total. To quantify the initial fluorescence intensity, each enteroid was imaged at the middle plane immediately after microinjection. Live imaging of microinjected enteroids at the middle and/or bottom plane started 10 to 120 min p.i., and time-lapse images were acquired every 5 min for up to 12 h (human enteroids), every 3 min for up to approximately 8 h (murine enteroids), or every 30 min for 16 h (barcoded infections). Confocal (rpsM-mCherry) and wide-field (pssaG-GFP) z-stacks of microinjected enteroids were acquired immediately after microinjection as well as at 1.5 to 2 h p.i. (rpsM-mCherry) using an X-Light V2 L-FOV spinning-disk module with a pinhole size of 60 μm (CrEST Optics) or at 4 h p.i. (pssaG-GFP), respectively. Time-lapse movies related to the figures can be found at https://doi.org/10.17044/scilifelab.12998570.

For motility analysis, single bacteria in the inoculum and within the lumen of microinjected enteroids were tracked using the TrackMate plugin (78) in Fiji (a version of ImageJ) (79). Relative fluorescence intensities were determined in Fiji by manually outlining the enteroid cross-section at the middle plane for each time point and quantifying the fluorescence within this area at 30-min intervals, whereby the fluorescence was normalized to the initial intensity immediately after microinjection for the respective enteroid. Fluorescence intensity profiles were determined in Fiji. Background subtraction in confocal and wide-field z-stack images was also performed in Fiji. The time point of bacterial escape from the enteroid lumen was defined as the time p.i. when the first visible fluorescent Salmonella Typhimurium were observed outside the epithelial boundary. For quantification of separate epithelial and luminal fluorescence intensities, the epithelial and luminal regions were defined based on manual measurements of the epithelial thickness and definition of the enteroid outline at the middle plane for each enteroid at 0, 2, 4, 6, and 8 h p.i. Next, a Gaussian blur filter was applied with a standard deviation of 5 μm, and the Gaussian blur was subtracted from the image to reach a uniform background signal close to 0 in the epithelial, luminal, and outside regions. The fluorescence intensity in each region (epithelium, lumen, and outside) was then quantified, and epithelial and luminal fluorescence intensities were normalized to the outside fluorescence intensity at 0 h p.i. for the respective enteroid.

Where applicable, statistical significance was determined by two-way analysis of variance (ANOVA) with Tukey’s honestly significant difference (HSD) post hoc test applying the functions aov() and TukeyHSD() in RStudio (80). For analysis of bacterial escape from the enteroid lumen, survival analysis according to the Kaplan-Meier model was performed using the functions Surv(), survfit(), and survdiff() in the survival package for RStudio (81), and statistical significance was assessed by the log rank test.