Bacterial Swarmers Enriched during Intestinal Stress Ameliorate Damage

Abstract

Background and Aims:

Bacterial swarming, a collective movement on a surface, has rarely been associated with human pathophysiology. This study aims to define a role for bacterial swarmers in amelioration of intestinal stress.

Methods:

We developed a polymicrobial plate agar assay to detect swarming and screened mice and humans with intestinal stress and inflammation. From chemically induced colitis in mice, as well as humans with inflammatory bowel disease, we developed techniques to isolate the dominant swamers. We developed swarm-deficient but growth and swim-competent mutant bacteria as isogenic controls. We performed bacterial re-inoculation studies in mice with colitis, fecal 16S, and meta-transcriptomic analyses, as well as in vitro microbial interaction studies.

Results:

We show that bacterial swarmers are highly predictive of intestinal stress in mice and humans. We isolated a novel Enterobacter swarming strain, SM3, from mouse feces. SM3 and other known commensal swarmers, in contrast to their mutant strains, abrogated intestinal inflammation in mice. Treatment of colitic mice with SM3, but not its mutants, enriched beneficial fecal anaerobes belonging to the family of Bacteroidales S24–7. We observed SM3 swarming associated pathways in the in vivo fecal meta-transcriptomes. In vitro growth of S24–7 was enriched in presence of SM3 or its mutants; however, since SM3, but not mutants, induce S24–7 in vivo, we conclude that swarming plays an essential role in disseminating SM3 in vivo.

Conclusions:

Overall, our work identifies a new but counterintuitive paradigm in which intestinal stress allows for the emergence of swarming bacteria; however, these bacteria act to heal intestinal inflammation.

Lay Summary

Bacterial swarmers are a feature of a stressed intestine, and when present in high abundance protect from intestinal inflammation in a microbiome dependent manner, facilitated by its distinct form of motility.

Graphical Abstract

Introduction

Bacterial motility is essential in mucosal colonization and has long been associated with virulence and pathogenesis^{1, 2} Intestinal inflammation, such as inflammatory bowel disease (IBD), is attributed to dysbiosis and the mucosal immune system³. The disease is characterized by enrichment of motile flagellated bacteria resident in the microbiome and its encroachment into the inner mucus layer and the intestinal epithelial cells (IEC)^4–6. However, despite cues of the molecular mechanisms of flagella during intestinal health and disease^7,8–10,11, the functional importance and consequence of bacterial motility in a microbial consortium is unknown.

Swimming and swarming are the two primary and common forms of bacterial motility¹². Swarming, driven by flagella, is a distinct process in certain groups of bacteria characterized by collective and rapid movement across a surface^{12, 13}. This process, in contrast with swimming in liquid, offers bacteria a competitive advantage in occupying specific niches (e.g., seeding colonization)¹⁴; however, the cost-benefits to bacteria^{15, 16} and consequences to its host or the environment remain primarily unknown¹⁷.

We hypothesized that bacterial swarming is a necessary adaptation to a noxious environment in a host such as bacteria within inflamed or stressed intestines. Since prototypical swarming bacteria (e.g., Proteus mirabilis, Pseudomonas aeruginosa) are associated with virulence^{17, 18}, we surmised that bacterial swarming might be well represented in colonoscopy aspirates from humans with bacterial virulence-associated pathologies (e.g., intestinal inflammation)¹⁹. This study aims to determine the occurrence and consequence of bacterial swarming in humans and in the animal kingdom, in the context of a stressed and non-stressed intestinal environment. In addition, we aim to uncover potential mechanisms by which swarming bacteria interact with the host.

Materials and Methods

Isolation of bacterial swarmers from feces and colonoscopic aspirates.

Patients diagnosed with either inflammatory bowel disease (Crohn’s disease or Ulcerative colitis) or undergoing routine screening colonoscopy for colorectal polyps/cancer or required a colonoscopy as part of their medical management of any gastrointestinal disorder as clinically indicated, were recruited for the study. Sixty-three (63) patients who consented to participate in a colonoscopy aspirate or fecal collection study was approved by the Institutional Review Board (IRB) (#2015–4465; #2009–446; #2007–554). Bacterial swarmers were isolated on Luria Bertani (LB) swarming agar medium containing 5 g/L agar with some modifications to an established method²⁰. To isolate a singular dominant swarmer from a polymicrobial mix of bacteria (such as feces), we initially focused on developing an assay to isolate swarmers using known polymicrobial mixed cultures of bacteria. Single bacterial species (up to seven strains belonging to different taxa) grown in LB [OD₆₀₀ of 1.0–1.3] were mixed in a 1:1 ratio and, 5 μL of this mix was spotted on 0.5% agar plates. Following air drying at room temperature, the plates were incubated at 37°C, 40% RH (relative humidity) for 10 hours. Bacterial swarm front was swabbed using a sterile tooth-pick from the edge of swarming colony at different locations (see arrows, Fig. S1) and after restreaking on separate agar plates and scaled by growth in LB, the bacteria in the samples were identified using Matrix Assisted Laser Desorption and Ionization-Time of Flight (MALDI-TOF). Swarmers present in the fecal or colonoscopic samples were isolated and determined using an identical approach. Fecal pellets and/or colonoscopy aspirates from the clinic and/or feces of mice and pigs were collected in sterile tubes, and freshly homogenized in PBS for swarming assays. Most bacterial swarmers were detected within the first 48–72 h from incubation. Dominant swarmers from the edge of the colony were identified using MALDI-TOF. Once identified, cells from the same aliquot were plated on to 1.5% LB agar and serially passaged from a single colony to obtain a pure culture of the strain. Details of the procedure are presented in Supplementary Materials.

Characterization of the bacterial strains.

Swarming ability of a single bacterial species using a pure culture of Enterobacter sp. SM1 and its isogenic mutant, Enterobacter sp. SM3 and its transposon mutants, Serratia marcescens Db10 and JESM267, clinical isolate of Serratia marcescens, Bacillus subtilis 3610 and its isogenic mutant DS215 was always determined on LB swarming agar at 37°C and 40% RH prior to any experiments using these strains. B. subtilis 3610 and its isogenic mutant were compared on LB swarming agar containing 0.7% agar²¹. In order to capture real time swarming motility, a temperature and humidity-controlled incubator equipped with time lapse photography was built and swarming area was calculated using a python-based script (Nature Protocol Exchange for detailed protocol, doi:10.21203/rs.2.9946/v1). Growth kinetics was observed in LB broth, while swimming potential of the strains were assessed in freshly grown cultures (OD₆₀₀ ~ 0.3) or 0.3% LB agar. Surfactin synthesis was determined using Blood agar hemolysis²², drop-collapse²³ and drop-counting assay²⁴. Swarming on mucosal surface was demonstrated using a colon tissue from mice that was treated with 3% DSS via a mucosal race experiment. Details of the techniques are presented in Supplementary Materials.

Mouse model studies.

Four to six-week old female C57BL/6 mice (Jackson Laboratories, Bar Harbor, ME; # 000664) were co-housed for acclimatization at the vivarium for 2 weeks prior to randomization by coin toss as previously described²⁵. Five-week old germ-free (GF) wildtype (WT) C57BL/6 mice were transferred to specific pathogen free (SPF) conditions²⁶ during experimentation (GF/SPF). Acute colitis was induced by administering 3% (w/v) DSS (MP Biomedicals, Cat. No. 160110). To determine the effect of swarming and swarming deficient strains during colitis, WT mice were orally gavaged with 100 μL (~ 4×10⁹ CFU/mL) test bacteria or LB as vehicle, daily for 9–12 days until the weight of vehicle group dropped > 20%. Swarming-deficient strains were generated either using recombineering and PCR Ligation mutagenesis approach^{27, 28} or mariner-based transposon mutagenesis²⁹. GF/SPF mice were gavaged with SM3 or LB and treated for 7 days, when most mice had > 10% weight drop. Daily gavage of bacterial strains required use of unwashed bacterial strains grown in fresh LB (OD₆₀₀ ~ 1.0). To determine the healing effect of SM3 in colitis, C57BL/6 mice were administered 3% DSS in drinking water for 7 days (when most mice had a weight loss > 10% of their pre-DSS exposure weight). Subsequently, mice received animal facility drinking water without DSS and were further randomized by coin-toss to a treatment group delivered 4×10⁹ CFU/mL of bacterial cells or LB by oral gavage for 5 days. At the end of the experiment, mice were euthanized using isoflurane anesthesia or CO₂ asphyxiation and intestines harvested for Hematoxylin-Eosin (H&E) staining and histopathology. Lipocalin (LCN2) assay was performed using Mouse Lcn2/NGAL Duoset ELISA kit (R&D System, DY1857).

The role of TLR5 was assessed in chronic colitis model of TLR5KO mice administered anti-IL-10R monoclonal antibody³⁰. Mice were orally gavaged with SM1 or SM3 every third day from Day 1 onwards. Histology scoring for inflammatory damage was performed according to published criteria for colonic inflammation as a consequence of cytokine imbalance⁴.

Fecal microbiome profiling.

16S rRNA meta-analyses of the fecal samples from mice were conducted at Wright Labs, LLC. DNA was isolated from feces using a Qiagen DNeasy Powersoil DNA Isolation kit following the manufacturer’s instructions (Qiagen, Frederick, MD). The 16S rRNA gene was amplified using Illumina iTag Polymerase Chain Reactions (PCR)³¹, pooled, gel purified at ~400bp and multiplexed with other pure libraries to form a sequencing library normalized to the final concentration of library observed within each sample. The sequencing library was sequenced using an Illumina MiSeq V2 500 cycle kit cassette with 16S rRNA library sequencing primers set for 250bp paired-end reads at Laragen Inc (Culver City, CA). The paired-end sequences were merged with a minimum overlap of 200 bases, trimmed at a length of 251 bp, and quality filtered at an expected error of less than 0.5% using USEARCH³², analyzed using the QIIME 1.9.1^{33, 34} and assigned operational taxonomic units (OTU) using UPARSE at 97% identity³⁵. The taxonomy was assigned using the Greengenes 16S rRNA gene database (13.5 release)³⁶.

In vitro co-culture assay using M. intestinale.

A broth-based or swarm plate-based co-culture assay was designed to identify possible interaction between SM3 and Muribaculum intestinale (DSM 28989). Early exponential phase cells (OD₆₀₀ 0.5–0.6) grown in chopped meat carbohydrate broth, PR II (BD, BBL) (CMCB-PRII) in an anaerobic chamber, at 37°C (O₂ = 2–3%) were used to establish the assay (Fig. S9C). For broth-based assay, M. intestinale was grown with fresh cells of SM3/SM3_18/SM1 in a Hungate tube and cells collected at different time points (21/24, 36, 45/48 hrs) for DNA extraction. For swarm-plate based assay, M. intestinale grown in CMCB-PRII was transferred into a bore-well at the center of a swarming plate on which SM3/SM3_18/SM1 swarmed. Plates were incubated at different conditions for 64 hours (aerobic, sealed or anaerobic) at 37°C. For sealed condition, plates were taped carefully using parafilm to maintain anaerobiosis throughout the experiment. For Divided/Sealed condition (please see caricature in Fig. 4), a small Petri dish was placed inside a big Petri dish, both containing swarming agar. The bore well containing M. intestinale was stationed in the small Petri dish, while the swarming or less swarming strains were spotted on agar present in the big Petri dish. This allowed physical separation of M. intestinale from the swarming bacteria, nevertheless maintaining an anaerobic condition in this sealed system. DNA was extracted and qPCR analysis was performed using equal volume of each diluted DNA sample and M. intestinale specific primers (Supplementary Table 1). Details of the technique is presented in Supplementary Materials.

Figure 4 |. Effect on S24–7 levels in the presence of SM3 and the insufficient (or inefficient) swarming variants in vivo and in vitro.

(A) 8-week old mice (n ≥ 5 per treatment group) were exposed to DSS water and treated with SM3_18, SM3_24, and SM1 by oral gavage for 12 days. Total DNA was extracted from feces collected on day 0 and day 12, processed and assessed using qPCR. Five (5) ng of total DNA in conjunction with S24–7 specific primers were used to quantify bacterial copy numbers. In each assay, DNA copy number/μL was calculated based on an internal standard curve. b-c, In vitro co-culture assay using M. intestinale cells grown in Chopped meat medium under anaerobic condition until early log phase (OD₆₀₀ ≈ 0.5) were used. (B) Fold change DNA copy number/μL relative to M. intestinale monoculture. In broth-based assay, 2 μL of early log phase cells of SM3, SM3_18, or SM1 was added to M. intestinale cells and mixed cells or monoculture of M. intestinale was collected at regular intervals (21–24, 36, 45–48 hrs). (C) In a swarming-plate based assay, early log phase M. intestinale was transferred in the bore-well and SM3, SM3_18, or SM1 was allowed to swarm either under aerobic or sealed condition at 37°C and RH ≈ 50%. Plates were sealed using parafilm to create and maintain anaerobiosis due to the act of swarming. M. intestinale grown under anaerobic condition was used as a positive control. In Divided/Sealed condition, swarming region was physically separated from the bore-well containing M. intestinale and sealed using parafilm. Closed boxes represent incubation in an anaerobic chamber. DNA extracted from equal volume of culture and resuspended in equal volume of TE buffer was used for qPCR in conjunction with M. intestinale specific primers. A, Data represented as mean and 95% CI, and significance tested using paired t-test. B-C, Data represented as mean (± SD) (n = 2 independent experiments and 2 technical replicates for each).

Statistical analysis.

P-values for statistical tests involving the experimental data were obtained by using appropriate parametric or non-parametric methods, as indicated in the figure legends. 95% confidence intervals (CI) were obtained for the relevant parameters. Normality (Gaussian distribution) for the data was not assumed, to begin with. For each dataset, normality was tested. If there was not much evidence in favor of normality, then suitable transformations (e.g., log normal transformation) were considered to discern whether the transformed data fit a Gaussian distribution. All statistical tests, except where otherwise indicated, were performed with Graph Pad Prism v.8.2.0; All plots depict the mean and 95% CI (except where otherwise indicated) for the relevant parameters.

Results

Presence of bacterial swarmers is a feature of a stressed intestine.

To test the relationship between fecal abundance of bacteria with swarming potential and human health, we developed a fecal agar-based modified polymicrobial swarming assay²⁰. We obtained colonoscopy aspirates from individuals with a progressive illness (inflammatory bowel disease - Crohn’s and ulcerative colitis and other common forms of intestinal stress like intestinal polyps^37,38, as well as age and gender-matched controls (those without a clinically active illness)). Within our sampling pool, bacterial collective spreading on soft agar was over-represented in cases with overt or clinically active intestinal stress (Fig. 1A–B). As a preliminary assessment, we judged bacterial swarmers’ presence in feces by the bacterial spread with a surfactant layer on soft-agar. Swarmers were isolated, identified by MALDI-TOF, and validated for their swarming motility (Table 1).

Figure 1 |. Effect of intestinal inflammation on bacterial swarming.

(A-C) Human colonoscopy aspirates (n = 45 intestinal disease; n = 25 non-disease) were spotted on 0.5% agar plates and the swarming assay performed. (A) Colonoscopic washes were obtained from individuals with active intestinal disease and matched controls. Swarming assays performed using aspirates were binned by disease as defined both clinically and by intestinal histopathology, where available. (B) Clinical demographics are described for the disease and non-disease population. (C) Swarming assays’ clinical test characteristics. (D) Swarming assays (72h) of fecal samples collected from pigs with and without IBD. Swarming scores - 0: no swarming, 1: swarming within 72h, 2: swarming within 48h, 3: swarming within 24h or less (Control: n = 6; IBD: n = 7). (E) C57BL/6 mice (8-week old) were exposed to water or DSS water for 7 days (n = 4 per group). Fecal samples of control group (above red line) and DSS group (below red line) were collected for swarming assay. Swarming plates were scanned at 12, 24, and 48 hours. (F) C57BL/6 mice (8-week old) were exposed with water or DSS water for 12 days (n = 8 per group). Fecal samples were collected for DNA extraction and SM1/SM3-specific PCR analysis was performed, and DNA copy number ascertained. Data represented as mean and 95% CI, significance tested using Fisher’s Exact test.

Table 1.

Bacterial Strains isolated and used in this study

Bacterial strains identified from luminal contents and isolated on swarming agar*
Strain Isolated	Swarming	Source
Escherichia coli #	+	Human IBD
Escherichia coli #	+	Human IBD
Escherichia coli	+	Human anal fistula
Klebsiella pneumoniae	+†	Human IBD
Klebsiella pneumoniae	− ‡	Healthy Human
Citrobacter koseri	+	Human IBD
Morganella morganii	− §	Human IBD
Serratia marcescens	+	Human adenomatous polyp
Proteus mirabilis	+ ‖	Mouse colitis
Proteus mirabilis	+ ¶	Mouse colitis
Enterobacter sp. #	+	Mouse (DSS colitis)
Enterobacter sp. #	+	Mouse (TNBS colitis)
Bacterial strains used in this study
Organism	Description	Reference
Enterobacter sp. SM1	A clinical isolate from feces of normal mice.	This study
ΔmotA SM1	A flagella motor function abrogated mutant of SM1, motA::kan	This study
ΔflhE SM1	A flagella associated gene involved in swarming, flhE::FRT:Kan:FRT	This study
HS2B SM1	A hyperswarming variant of SM1 generated by serial passage on swarming agar.	This study
Enterobacter sp. SM3	A clinical isolate from feces of DSS-colitis mice.	This study
SM3_18	A transposon mutant of SM3, putative aerobactin synthesis gene iucB::Tn::kan	This study
SM3_24	A transposon mutant of SM3, putative isocitrate/isopropylmalate dehydrogenase/ADP-ribose pyrophosphate gene::Tn::kan	This study
Serratia marcescens	A clinical isolate from human adenomatous polyp.	This study
Bacillus subtilis
3610	A wild-type isolate.	$Kearns   Losick, 20039
DS215	A swarming defective mutant of 3610, swrA::tet	$Kearns et.al10
Serratia marcescens
Db10	A wild-type isolate.	$Pradel et.al11
JESM267	A serrawettin W2 defective mutant of Db10, swrA::miniTn5-Sm	$Pradel et.al11
Salmonella enterica serovar Typhimurium
ATCC 14028	A wild-type isolate.
ΔfliL	A swarming deficient mutant of S. enterica, fliL: :FRT.	This study
Muribaculum intestinale YL27	A strict anaerobe isolated from cecal content of mice.	Lagkouvardos et al.42

^*Human or mouse feces was subject to the swarming assay and any swarm colony detected within 24 h was swabbed for strain identification. In addition, delayed swarmers were classified as negative but their swarm edge also yielded single species

^†Feces from patient with clinically controlled Crohn’s disease with moderate surfactant edge detected at 74 h

^‡Classified as non-swarmer, however, a very minimal surfactant edge present at 24h and no progression thereafter

^§Feces from patient with clinically controlled Crohn’s disease with surfactant edge detected at 48h

^{‖, ¶}Mouse model: Msh2/-loxPTgfbr2 loxp Villin-cre

^#Also confirmed using Illumina Sequencing (PacBio)

^$References cited in the Supplementary Materials.

In this pilot evaluation, the specificity and positive predictive value of the test for disease as defined was approximately 88 and 89%, respectively. In comparison, the test’s sensitivity and the negative predictive value was only approximately 56 and 52%, respectively (Fig. 1C). Similarly, feces collected from a limited sample size of pigs with active inflammatory bowel disease also showed an increased qualitative trend of collective spreading and swarming compared to control pigs (Fig. 1D).

Novel Enterobacter swarming strains were isolated from mouse feces.

Next, we focused on isolating endogenous swarming bacteria residing in rodents and humans. An initial approach was to determine if a single dominant swarming species could always be isolated from a polymicrobial culture (e.g., mammalian feces). In our competitive swarming assay, a mix of different pure bacterial cultures gave rise to a single bacterial species populating the leading edge of the swarm colony on agar (Fig. S1A–B). Similarly, swarming assays using the pooled mouse or individual human feces yielded single species of a dominant swarmer as identified by MALDI-TOF (Table 1; Fig. S1B). To test whether swarming bacteria are also present in preclinical models, we screened feces of mice exposed to dextran sulfate sodium (DSS) that caused acute colonic inflammation^{8, 39}. Swarmers (in feces) were uniformly absent in water exposed mice, while present in DSS exposed mice (Fig. 1E). In a single experiment, we found “nearly identical isolates” [>99% identical, one contig of 5,107,194 bp (NCBI BioProject PRJNA558971)] from two different mouse fecal specimens - Enterobacter sp. SM1 from mice exposed to water and Enterobacter sp. SM3 from mice exposed to DSS (Fig. S2A). SM3 swarmed significantly faster compared to SM1. A quantitative PCR sequencing-based approach to accurately identify SM1 or SM3 like bacteria in feces showed a significant increase in its abundance during the evolution of DSS-induced colitis (>10,000 DNA copy number /μL) than water only group (Fig. 1F). Taken together, using an agar-based assay, we were able to isolate nearly identical strains from control and mouse with colitis that exhibit striking differences in their swarming potential.

Swarming Enterobacter sp. SM3 abrogates intestinal inflammation in a mouse model of colitis.

To determine the functional consequence of bacterial swarmers in the host, we administered the “near-identical” swarming competent SM1 or SM3 strains to mice with DSS-induced colitis. In comparison with SM1, SM3 is a hyperswarmer (Fig. S3A; Video S1), but both strains possess the same swim speed (Fig. S3B–C), surfactant production (Fig. S3D), and growth rate (Fig. S3E). In contrast to that observed with SM1, SM3 significantly protected mice from intestinal inflammation (Fig. 2A–F). Comparison of clinical parameters showed that SM3 significantly protected from body weight loss (Fig. 2A), increased colon length (Fig. 2B), reduced the colonic inflammation score (Fig. 2D), and had reduced expression of pro-inflammatory mediators compared to vehicle-treated colitic mice (Fig. 2E–F). To test the mucosal healing capacity of swarming bacteria, we administered strains SM1 and SM3 to mice during the recovery phase of DSS exposure⁴⁰. When compared to the vehicle, SM3 significantly improved weight gain and colon length with reduced total inflammation and fibrosis at the microscopic level (Fig. S4). In mice exposed to DSS, SM3, but not the swarming deficient mutants (SM3_18 and SM3_24), showed significant protection against weight loss (Fig. 2G), colon length (Fig. 2H), and inflammation (Fig. 2I). SM3 and its isogenic transposon mutants (SM3_18 and SM3_24) only differed in swarming potential (Fig. S3H), but not swimming speed (Fig. S3I–J), surfactant production (Fig. S3K), or growth rate (Fig. S3L).

Figure 2 |. Effects of Enterobacter sp. SM strains on DSS induced colitis in C57BL/6 mice.

(A-F) 8-week old mice were exposed to DSS water and treated with vehicle (LB), SM1 or SM3 by oral gavage for 10 days. A-B indicates weight loss (A) and colon length (B) (n = 21 per treatment group). (C) Representative images (100x magnification) of H&E stained colonic section treated with vehicle (left), SM1 (middle) and SM3 (right). (D) Inflammation score (n = 21 per treatment group). (E-F) In a separate experiment, myeloperoxidase (MPO) enzyme activity was determined (n= 3, each in duplicate) (E). Colon total RNA (n = 4) was isolated and reverse transcribed to cDNA. RT-qPCR data show fold induction of mRNA (TNFα, IL10, TNFR2, IL6). PCR was repeated in quadruplicate. The expression was normalized to internal control, TBP. The entire experiment was repeated n = 2 for reproducibility (F). (G-I) C57BL/6 mice (8-week old) were exposed to DSS water and administered vehicle (LB), SM3, or its mutants (SM3_18 or SM3_24) for 10 days. G-I indicates weight loss (G), colon length (H) and inflammation score (I) (n = 10 per treatment group). Unless otherwise noted, data are represented as mean and 95% CI, and significance tested using one-way ANOVA followed by Tukey’s post hoc test. B, data represented as median and interquartile range, and significance tested using Kruskal-Wallis followed by Dunn’s multiple comparisons test. ns, not significant. H&E, Hematoxylin and Eosin; TBP, TATA-Box Binding Protein; CI, Confidence Interval.

As a second model of colitis, we used a TLR5KO IL-10R neutralization-induced colitis model of mice. SM3 also significantly protected from body weight loss (Fig. S5A), reduced spleen and colon weight (Fig. S5B–C), increased the cecum weight (Fig. S5D), reduced serum KC level and lipocalin level (Fig. S5E–F), reduced levels of fecal lipocalin (Fig. S5G), reduced myeloperoxidase activity (Fig. S5H), and had reduced the colonic inflammation score (Fig. S5I), when compared to the SM1.

To determine if the bacterial abundance differed among mice exposed to SM3 or its mutants, we assessed the abundance of the strains in feces using qPCR. Using the DSS colitis model, on day 4, the levels of SM1, SM3 and its mutants present in feces were not significantly different (Fig. S6A–C, F). We chose to enumerate bacterial levels in feces on day 4 due to the equivalent pathological conditions of mice, as defined by weight change, when treated with different strains. To identify if the loss of protection by SM3_18 could be related to slightly higher levels of its presence compared to SM3, although not significant, we performed a dose attenuation study, which demonstrated non-significant changes in either weight loss (Fig. S6G) or lipocalin levels (Fig. S6H). In accordance with these results, a diverse set of commensal swarmers (Bacillus subtilis 3610 and Serratia marcescens Db10) and a clinical strain of S. marcescens (isolated from the surface washing of a human dysplastic polyp) exhibited protection against DSS induced inflammation in mice (Supplementary text and Fig. S7 & S8). Together, these data implicate or associate SM3 with swarming properties, as opposed to swarming-deficient strains, with anti-inflammatory activity.

SM3 mediated abrogation of intestinal stress is microbiome dependent.

We used germ-free mice (GF/SPF) exposed to DSS and treated them with SM3 to determine if the anti-inflammatory role of SM3 depends on the conventional intestinal microbiome composition. This strain was unable to abrogate intestinal inflammation in GF/SPF mice (Fig. 3A). We analyzed fecal samples of colitic mice (conventional and GF/SPF) with SM3 administered using 16S rRNA gene profiling. In contrast to GF/SPF mice, conventional mice feces showed specific enrichment of anaerobes belonging to the family S24–7 and Lactobacillaceae within SM3 treated mice when compared to vehicle mice (Fig. 3B). Specifically, in conventional mice, we found a significant increase in the abundance of S24–7 with SM3 gavage compared to vehicle in DSS exposed mice (Fig. 3C). However, quantitative PCR analysis of the levels of S24–7 in the feces of DSS-induced colitis mice gavaged with SM1 or SM3_18 or SM3_24, which did not exhibit protection from intestinal inflammation, was significantly reduced (Fig. 4A). In mice not exposed to DSS, the levels of S24–7 bacteria remain stable in the SM3 treated group compared with the untreated group (Fig. 3C). We observed that enriched S24–7 negatively co-occurred within DSS exposed conventional mice with pathogenic taxa such as the Peptostreptococcaceae and Enterobacteriaceae (Fig. 3D). Together, these data suggest that protection from intestinal inflammation by SM3 is associated with the presence of a beneficial S24–7 group of bacteria⁴¹.

Figure 3 |. Effects of SM3 on the intestinal microbiota of GF/SPF and conventional mice.

(A) C57BL/6 GF/SPF mice (5-week old) were exposed to DSS water and treated with vehicle (LB) or SM3 for 6 days. A indicates weight loss (left), colon length (middle), and inflammation score (right) (n = 10 per treatment group). (B) Linear discriminant analysis Effect Size (LEfSe) plot of taxonomic biomarkers identified using feces of SM3 treated conventional (n = 10) (upper) and GF/SPF (n = 10) (lower) colitic mice on day 12 and day 6, respectively, as compared to vehicle (n = 10). Green and red bars indicate bacterial enrichment within SM3 treated and vehicle group respectively. All taxa that yielded an LDA score >3.0 are presented. (C) Relative abundance of S24–7 in the feces from DSS (lower) and control (upper) mice treated with SM3 or vehicle (n = 8 per treatment group). Linear regression line was fit to show the trend of the change (dotted lines show the 95% confidence bands). The slope of the SM3 treated group is similar to vehicle in water control group (P = 0.783), but significantly different in DSS group (P = 0.018). (D) Co-occurrence network plot showing strong positive and negative correlations between OTU abundances. All networks were generated with CoNet and visualized in Cytoscape. Processing was applied to the dataset with CoNet. Input filtering constrained the minimum occurrence of OTUs and considered only those present in at least 50% of samples. Standardization normalized dataset columns. Networks were constructed using Spearman’s correlation methods with threshold setting at 0.9, Bray Curtis dissimilarity at the automatic threshold setting, and Kullback-Leibler dissimilarity at the automatic threshold setting; the edge selection parameter was set to 30 for the strongest positive and negative correlations. Randomization steps included permutations and bootstraps with filtering of unstable edges and Benjamini-Hochberg procedure with a P-value of 0.05. Node clusters with less than or equal to three edges were not shown in the final network. Edge coloration indicates copresence in green or mutual exclusion in red. Nodes were colored by taxonomic phylum and labeled by the highest taxonomic ranking available. Unless otherwise noted, data are represented as mean and 95% CI, and significance tested using a two-tailed Student’s t-test. OTU, Operational Taxonomic Unit; GF/SPF, Germ-Free mice transferred to specific pathogen free conditions.

Enterobacter sp. SM3 promotes growth of Muribaculum intestinale in vitro.

A recent study has reported the first cultured bacterium Muribaculum intestinale (DSM 28989) belonging to Bacteriodales S24–7 family⁴². We used this strain to delineate any potential interspecies interaction with SM3 using an in vitro co-culture assay system. However, in precedence, we assessed if the strain M. intestinale shared sequence homology to any of the S24–7 taxa identified in our fecal 16S rDNA profile. OTU_5, which we found in the highest abundance among all other OTU’s representing S24–7 taxa, exhibited > 96% identity to M. intestinale (Fig. S9). Hence, we performed a broth-based co-culture assay using this strain and SM3 or SM1 or SM3_18. Interestingly, the proportion of M. intestinale during co-culture was higher than its monoculture at any tested time point. SM3 and the partially swarming deficient strains, SM1 and SM3_18, had a two-four (2–4) fold increase in DNA copy number/μL when analyzed by qPCR using S24–7 specific primers (Fig. 4B).

We also designed and developed a plate-based co-culture assay to compare the effects of swarming bacteria SM3 and swarming-deficient variants, SM1 or SM3_18, on the growth of M. intestinale. In this assay, swarming plates harbored a central bore well containing M. intestinale that guarantees a direct or indirect interaction with the spreading bacteria on agar of the same plate. We sealed the plates so that the act of swarming generated an anaerobic environment suitable for the growth of M. intestinale. At 64 hours, in congruence with the broth co-culture assay results, we observed an increase in M. intestinale counts with SM3, SM3_18, and SM1 (Fig. 4C). To better understand the observed increase in M. intestinale levels, we developed a separate plate-based co-culture assay. In this assay, we physically separated the swarming region from the central borewell containing M. intestinale to prevent any direct or indirect interaction with the swarming bacterium. In this system, as the bacteria swarmed on the agar surface over 64 h, oxygen levels were reduced. M. intestinale showed no growth under the conditions tested (Fig. 4C, Divided/Sealed). Overall, our results suggest that both planktonic and swarming cells of SM3, SM1, or SM3_18, when co-cultured in vitro, can promote the growth of the S24–7 family (M. intestinale), independent of reduced oxygen concentrations in the environment. Coincidently, the development of significantly reduced oxygen concentrations in the environment is also observed in vivo but only with SM3 and not SM1 or the other SM3 mutant bacteria (Supplementary text and Fig. S10). Our results suggest that SM3 proximity to M. intestinale is necessary to induce the growth of the latter species.

In this context, to understand if SM3 does swarm in vivo, we searched for the presence of transcriptomic markers in the feces that can be linked to SM3 swarming physiology. In agar-based studies of RNA sequencing of SM3 obtained from the edge of a swarming colony versus the pre-swarming colony at the center, a singular pathway was significantly upregulated – the lipid A biosynthetic pathway (fold change three-fold, q value = 0.0376). Meta-transcriptomic analysis of feces from SM3 treated DSS-induced colitic mice identified a steady increase of the lauroyl acyltransferase transcript involved in Lipid A biosynthesis Day 4 and Day 12 when compared to Day 0 (Fig. S14A). However, heat-killed SM3 treatment showed a reduction in transcript abundance by Day 12. Other genes are known to be associated with swarming. The sigma factor FliA and nitrate reductase NarH were also enriched in SM3 versus heat-killed SM3 gavaged colitic mice (Fig. S14B) (Supplementary text). Normal mice gavaged with SM3, or heat-killed SM3 did not show enrichment of these genes. Also, an ex vivo race assay on a colitic surface demonstrates the potential of bacterial swarming in vivo during colitis (Fig. S13 & S14, Video S2–S5). Collectively, our data provides multiple lines of indirect evidence suggesting that bacterial swarming is a likely phenomenon in vivo, and a motility form that is necessary for the induction of M. intestinale growth.

Discussion

Our study finds that intestinal inflammation itself promotes a protective niche that facilitates enrichment of bacterial swarmers. Despite the caveat that our approach might preclude the selection of swarmers that do not produce surfactant¹², these pilot data indicate that collective spreading and swarming is a specific feature and potentially a biomarker of an intestinal pathology, as defined by harboring active intestinal inflammation or polyps. Surprisingly, however, these bacterial swarmers, when dosed in sufficient abundance, abrogate intestinal inflammation in mice. Unfortunately, a limitation of our clinical study was that we did not analyze details of patient clinical history (e.g., use of medications) to determine their relationship with the presence or quantitative performance of bacterial swarmers in feces.

We focused on a newly isolated bacterium, Enterobacter sp. SM3, which is resident to the intestinal microflora of mice. In vivo, SM3, but not SM1, or SM3 swarming deficient mutants (poor swarmers), influenced the specific enrichment of the S24–7 group of bacteria. Notably, the family of S24–7 (Muribaculaceae) is known to repair barrier function in inflamed mice intestines^41,43. However, the in vitro co-culture experiment proved that a close interaction between SM3 and the S24–7 group of bacteria is essential for its enrichment. Thus, we hypothesized that it is the relative hyperswarming activity of SM3 (but not the weak swarming SM1 or SM3 mutants) that may facilitate close interaction with the S24–7 group of bacteria in vivo. Further support of this hypothesis comes from the bacteria’s ability to swarm on a mucosal surface afflicted by colitis but not on the normal mucosal surface (ex vivo mucosal race assay), and from the meta-transcriptomic mining of swarming associated genes in colitic mice administered SM3 (but not its heat-killed counterpart). The present mechanism implicates swarming SM3 in enhancing S24–7 (Muribaculaceae) which then suppresses host inflammation. Nevertheless, we do not exclude other direct or indirect effects of the swarming SM3 on mucosal inflammation and healing. However, if present, it would assist in suppressing host inflammation in conjunction with enrichment of the S24–7 group of bacteria in the gut.

Swarming bacteria secrete surfactants, such as surfactin, that facilitate motility on a solid surface¹². Surfactin is known to attenuate TNBS induced colitis, possibly by differentially regulating anti-inflammatory and pro-inflammatory cytokines⁴⁴. However, none of the isogenic pairs showed significant difference in surfactant production, suggesting that the observed protection was not due to secreted surfactin (Supplementary text). Transpositions in SM3_18 and SM3_24 were found to be located within the putative structural genes encoding N6-hydroxylysine O-acetyltransferase or aerobactin synthesis protein (iucB) and isocitrate/isopropylmalate dehydrogenase/ADP-ribose pyrophosphate of COG1058 family, respectively (Supplementary text). Nevertheless, transposon integration in SM3_18 led to a polar insertion that will only disrupt the expression of downstream genes iucC, iucD and iutA located within the operon, hampering aerobactin synthesis only. Fundamentally, genes iucD and iutA aid in iron acquisition in bacteria during nutrient limiting condition⁴⁵. A single study has also shown the dependence of bacterial phytopathogen Pantoea stewartii swarming on aerobactin synthesis⁴⁶. However, in this context, we did not observe any changes in fecal iron levels in colitic mice exposed to either wild-type SM3 or SM3_18 (Supplementary text), negating the iron effects in the suppression of inflammation.

In a germ-free/SPF condition, the loss of protection by SM3 allows us to speculate the role of a full spectrum intestinal microbiome in the observed effect. Oral gavage of SM3 in conventional colitic mice enriched beneficial anaerobes. As intestinal inflammation creates a shift from anoxic to oxic⁴⁷ (Fig. S10A), it was unexpected to find enrichment of obligate anaerobes such as Bacteriodales S24–7 in SM3 treated mice. We observed SM3 fed colitic mice had significantly lower oxygen concentration than the colitic mice treated with swarming deficient variants. We conjectured the possible role of swarming movement of SM3, if occurring in vivo, in reducing oxygen concentration as also observed in vitro. It was further corroborated by the increase in anaerobic taxa in the feces of GF/SPF mice treated with SM3 (Fig. S11).

Nevertheless, a steady increase of S24–7 specific OTU’s in SM3 treated DSS-colitic mice pointed towards a potential mechanism underlying the observed protection. Hence, we designed a broth and plate-based co-culture assay to identify possible interactions between SM3 and the first cultured bacterium belonging to the S24–7 family, M. intestinale. Both SM3 and the less swarming variants promoted the growth of M. intestinale in co-culture assay. However, linking this observation with a decrease in the levels of S24–7 in the fecal DNA obtained from SM1, SM3_18, and SM3_24 led us to speculate the essential role of swarming by SM3 in exhibiting protection. We conjectured that in addition to an anaerobic environment generated by the act of swarming on the agar plate, all the tested strains either required a direct cell-cell contact or produced a secretome, which promoted the growth of M. intestinale. This was further validated by the negative outcome by design of a plate-based assay that allowed physical separation of swarming SM3 from M. intestinale and created an anaerobic condition in the system suitable for the growth of M. intestinale.

Based on the evidence of swarming on a mucosal surface, we conclude that swarming of SM3 in vivo may facilitate close spatial interaction with the S24–7 group of bacteria. SM3 may aid in re-establishing hypoxia and, consequently, creating an optimal condition for the enrichment of S24–7 and other anaerobes in a specific microenvironment. In summary, our work demonstrates the unique and unprecedented role that bacterial swarmers play in intestinal homeostasis. We find the potential for a new personalized “probiotic” approach stemming from the ability to isolate and bank swarming microbes during colitic episodes.

What you need to know Background and Context.

Bacterial swarming is defined as collective movement of cells on a surface. As opposed to biofilms, bacterial swarming has rarely been associated with host pathophysiology.

New Findings

Presence of bacterial swarmers is a feature of a stressed intestine. Bacterial commensal swarmers can protect from intestinal inflammation, when present in high abundance, in a microbiome dependent manner. A novel swarming bacterium Enterobacter sp. SM3 can enrich S24–7 group of bacteria, associated with IBD remission.

Limitations

This study lacks direct evidence of in vivo bacterial swarming.

Impact

This study encourages isolation and banking of bacterial swarmers as a potential personalized probiotic approach.