Cell Shape and Population Migration Are Distinct Steps of Proteus mirabilis Swarming That Are Decoupled on High-Percentage Agar

How surfaces influence cell size, cell-cell interactions, and population migration for robust swarmers like P. mirabilis is not fully understood. Here, we have elucidated how cells change length along a spectrum of sizes that positively correlates with increases in agar concentration, regardless of population migration. Single-cell phenotypes can be decoupled from collective population migration simply by increasing agar concentration. A cell’s lipopolysaccharides function to broaden the range of agar conditions under which cell elongation and single-cell motility remain coupled with population migration. In eukaryotes, the physical environment, such as a surface matrix, can impact cell development, shape, and migration. These findings support the idea that rigid surfaces similarly act on swarming bacteria to impact cell shape, single-cell motility, and collective population migration.

ABSTRACT

Swarming on rigid surfaces requires movement of cells as individuals and as a group of cells. For the bacterium Proteus mirabilis, an individual cell can respond to a rigid surface by elongating and migrating over micrometer-scale distances. Cells can form groups of transiently aligned cells, and the collective population is capable of migrating over centimeter-scale distances. To address how P. mirabilis populations swarm on rigid surfaces, we asked whether cell elongation and single-cell motility are coupled to population migration. We first measured the relationship between agar concentration (a proxy for surface rigidity), single-cell phenotypes, and swarm colony phenotypes. We find that cell elongation and single-cell motility are coupled with population migration on low-percentage hard agar (1% to 2.5%) and become decoupled on high-percentage hard agar (>2.5%). Next, we evaluate how disruptions in lipopolysaccharide (LPS), specifically the O-antigen components, affect responses to hard agar. We find that LPS is not essential for elongation and motility of individual cells, as predicted, and instead functions to broaden the range of agar concentrations on which cell elongation and motility are coupled with population migration. These findings demonstrate that cell elongation and motility are coupled with population migration under a permissive range of surface conditions; increasing agar concentration is sufficient to decouple these behaviors. Since swarm colonies cover greater distances when these steps are coupled than when they are not, these findings suggest that collective interactions among P. mirabilis cells might be emerging as a colony expands outwards on rigid surfaces.

IMPORTANCE How surfaces influence cell size, cell-cell interactions, and population migration for robust swarmers like P. mirabilis is not fully understood. Here, we have elucidated how cells change length along a spectrum of sizes that positively correlates with increases in agar concentration, regardless of population migration. Single-cell phenotypes can be decoupled from collective population migration simply by increasing agar concentration. A cell’s lipopolysaccharides function to broaden the range of agar conditions under which cell elongation and single-cell motility remain coupled with population migration. In eukaryotes, the physical environment, such as a surface matrix, can impact cell development, shape, and migration. These findings support the idea that rigid surfaces similarly act on swarming bacteria to impact cell shape, single-cell motility, and collective population migration.

INTRODUCTION

Eukaryotic and bacterial cells can change cell morphology and collective behaviors in response to interactions with the physical environment. This broad phenomenon can be readily observed in swarming bacteria. Many bacteria move on top of surfaces with the help of external appendages termed flagella, as extensively reviewed in several articles (see, e.g., references 1 and 2). The transition from swimming through liquid to swarming on a surface coincides with changes in part triggered by the sensation of increased viscosity, which is at least partially sensed by the flagella (1, 3–10). In contrast, robust swarming bacteria, such as the human opportunistic pathogens Proteus mirabilis and Vibrio parahaemolyticus, exhibit more-dramatic phenotypic and behavioral responses upon transition to a swarm-permissible surface. Cells are short in liquid and in low-percentage (0.3%) agar; however, bacterial cells can elongate up to 40-fold and move as a collective swarm on low-wetness high-percentage (0.75% to 2.5%) agar. Swarm motility is hypothesized to be coupled with cell elongation, especially during movement on high-percentage agar, partially because increased cell length would allow for the accommodation of additional flagella on a cell’s surface (11).

We used P. mirabilis as a model system to interrogate how the surface environment impacts swarming, specifically with consideration to cell elongation, single-cell motility, and population-wide migration for a robust swarmer. In low-percentage swim-permissive agar, P. mirabilis cells are approximately 2-μm-long, rigid, and rod-shaped cells that move independently from one another. Upon contact with a rigid surface, cells elongate into flexible hyperflagellated swarmer cells (12–14). Many genetic controls for swarmer cell elongation have previously been elucidated (reviewed in references 15, to ,17). To facilitate cooperative swarm motility, elongated P. mirabilis swarmer cells can bundle their flagella (18); groups of cells can form clusters that aggregate and disperse dynamically. Swarmer cells will divide into rigid 2-μm-long non-swarm-motile cells to restart the swarm developmental cycle. Iterative cycles of cell elongation, population migration, and division contribute to the swarm colony’s appearance as a bullseye pattern.

Several factors affect population migration, such as surface tension, agar wetness, osmotic pressure, nutrient availability, membrane stress, and lipopolysaccharide (LPS) biosynthesis and structure (19–26). Of these, LPS, which comprises the outer leaflet of the outer membrane of Gram-negative bacteria, is particularly interesting for several reasons. First, P. mirabilis cells undergo a drastic remodeling of the LPS-associated cell envelope components during the transition between swimming and swarming motility (27–29). Second, LPS is reportedly required for initial swarmer cell differentiation and elongation. Disruptions of P mirabilis LPS biosynthesis genes, including the O-antigen ligase gene waaL (24, 30) and the sugar-modifying enzyme gene ugd (22, 31, 32), activate cell envelope stress-associated pathways that in turn downregulate pathways associated with surface sensing, cell elongation, and population migration. These results led to a model in which LPS might contribute to surface sensing by P. mirabilis. Finally, LPS is required for swarm motility in Escherichia coli (33) and Salmonella enterica (21, 34), though E. coli, Bacillus subtilis, and S. enterica generally swarm on low-percentage agar (<0.7%) or high-wetness Eiken agar. LPS reportedly promotes swarm motility by serving as an osmolarity agent to draw moisture from an agar gel environment, which serves to facilitate swarming as a collective population (21, 34–41).

We designed this study to address how bacteria swarm on rigid surfaces by answering the question of whether single-cell behaviors are coupled with population migration across hard-agar surfaces and, if so, how. We have determined that P. mirabilis cells adopt a continuous and increasing gradient of cell lengths and cell-cell interactions in response to increasing densities of agar. Under swarm-permissive conditions, collective population migration is coupled with these single-cell behaviors, resulting in swarm colonies that occupy centimeter-scale distances. These single-cell behaviors and population migration become decoupled at agar densities above an inflection point (∼2.5% LB agar), resulting in swarm colonies that expand to a lesser extent or not at all. We found that loss of LPS-linked polysaccharides promotes greater single-cell elongation and reduces the range of conditions permissive for swarm colony expansion. Further, we show that comingling with the wild type does not rescue the population migration defects of LPS-deficient cells. These findings demonstrate that LPS functions to modify the immediately local environment around each cell to promote elongation and motility. In this way, LPS indirectly contributes to collective population migration. These findings illustrate that cell elongation, single-cell motility, and population migration are separate contributions that together allow for centimeter-scale swarm colonies to emerge.

RESULTS

Swarmer cell size, cell-cell interactions, and colony structure shift in response to agar concentration.

We set out to examine how P. mirabilis populations of strain BB2000 respond to rigid surfaces. We chose to alter the single surface variable of agar percentage to determine whether the visible structure of swarm colonies, cell shape, or cell-cell interactions responded to changes in surface properties. P. mirabilis swarms are typically grown on nutrient-rich medium, e.g., Lennox lysogeny broth (LB) or CM55 containing 1.5% agar. Increasing agar concentration up to 2.5% is known to narrow swarm colony terraces; swarm colony expansion is inhibited beyond 2.5% agar (38).

Using LB medium, we altered the agar concentrations, ranging from 0.75% agar, on top of which cells move, to 4% agar on which cells form isolated colonies on top of the surface. We inoculated wild-type strain BB2000 cells at a single point on LB plates, followed by growth at 37°C overnight. Populations grew to form unstructured thin films on top of the 0.75% agar plates (Fig. 1A). This visible structure is distinct from the stereotypical haze of a swimming population grown in 0.3% LB agar (42). Terraces were visible between 1% agar and 2.5% agar and narrowed as the agar concentration increased (Fig. 1A; see also Fig. SF1 in the supplemental material). On 1 and 1.25% agar, populations formed semistructured swarm colonies in which visible terraces were broad and poorly defined (Fig. 1A). On 1.5%, 1.75%, and 2% agar, populations formed distinctly terraced swarm colonies with a terrace width inversely correlated with agar concentration (Fig. 1A and SF1). On 2.5% agar, the terrace rings became indistinct, lending the colony a mucoid appearance (Fig. 1A). The populations did not expand beyond the inoculation point when the agar concentration was 4% (Fig. 1A). Altogether, the formation of visible terraced structures, as well as the narrowing of terrace width, emerged at 1% LB agar and increased in frequency as the agar concentration increased.

FIG 1.

Swarmer cell length and visible swarm colony structure scale with agar concentration. (A) Representative images of populations inoculated onto LB medium containing 0.75% to 4.0% agar and grown overnight at 37°C on a 10-cm-diameter petri dish. Visible swarm colony structure was categorized as unstructured, semistructured, structured, or restricted based on a comparison of terrace definition and width relative to wild-type populations grown on 1.5% LB agar. (B) Representative images of cells encoding a chromosomal Venus reporter for fliA expression. Cells were inoculated onto 1-mm-thick LB medium pads containing indicated concentrations of agar, grown at 37°C, and then imaged using epifluorescence microscopy. Images of swarming populations (on up to 2.5% agar) were taken at the leading edges of swarms, while images of nonswarming populations (on 4% agar) were taken at the edges of the inoculum. Images are phase (top) and fluorescence (bottom). Fluorescence images were subjected to background subtraction. The corresponding macroscopic swarm colony structure characteristics as described in panel A are indicated above the images. Scale bar = 10 μm. (C). Swarmer cell size was measured under the listed medium conditions. Wild-type cells were comixed in a 10:1 ratio with wild-type cells carrying the fliA reporter strain. Data are shown in box-and-whisker plots. The box denotes median, 25th, and 75th percentiles. Whisker boundary, 1st and 99th percentiles. Individual data points represent values beyond the 1 to 99 percentile range. The numbers of measured cells for each agar percentage, sequentially from 0.75% to 2.5% agar, are as follows: 2,446, 2,894, 1,419, 1,502, 1,262, 1,264, and 913. Histograms of these data can be found in Fig. SF1.

We next considered whether shifts toward terraced structures and narrower swarm rings with increasing agar concentration resulted from changes in single-cell morphology or behavior. We used epifluorescence microscopy to examine individual cells and followed the activation of the swarm developmental cycle by monitoring a Venus reporter for fliA expression; this reporter was integrated onto the chromosome. The fliA gene encodes the flagellar sigma-28 factor and is positively regulated by FlhD₄C₂ (43). This reporter strain has no discernible differences in swarm behavior from the unlabeled parent strain, and fluorescence associated with the fliA reporter is highly exhibited when cells elongate into flexible motile swarmer cells (14, 42). Such swarmer cells often migrate across the surface as part of a cluster of cells (Fig. 1B). The reporter strain was grown under agar conditions ranging from 0.75% to 2.5% LB agar. We observed that the length of each swarmer cell and the numbers of cells per group increased as the agar concentration increased (Fig. 1B and SF1). Cells did not consistently align into groups of cells when grown on 0.75% to 1% LB agar (Fig. 1B). However, groups of cells visibly aligned with one another, and the number of cells within each raft increased as the agar concentration increased from 1.25% to 2.5% (Fig. 1B). Cells are motile under each agar condition (see https://www.youtube.com/playlist?list=PLDTC8EPnqPdtZmPVlcKGCm29gMhNz3n3a).

To quantify swarmer cell size, we mixed unlabeled wild-type cells with cells carrying the fluorescent fliA reporter at a 10:1 ratio, allowed populations to initiate swarm motility, and imaged across the swarm front. The area of the fluorescence signal associated with individual swarmer cells was measured (Table 1). Cells grown on 0.75% and 1% LB agar had median cell areas of 4.0 and 4.1 μm², respectively (Fig. 1C). Cells became increasingly elongated as the agar concentration increased from 1.25% to 2.5% LB agar (Fig. 1C). At 1.5% LB agar, cells had a median cell area of 7.8 μm², while cells on 2.5% LB agar were longer, with a median cell area of 18.2 μm². The distribution of cell sizes also increased as agar concentration increased, indicating that populations were becoming increasingly heterogeneous in length (Table 1 and Fig. SF1). Altogether, these findings show that single-cell phenotypes (e.g., increased cell size, heterogeneity of sizes within the population, and cluster size) correlate with a narrowing of terraces and a reduction in swarm colony expansion.

TABLE 1.

Cell measurements for BB2000

Parametera	Cell measurement (μm2) by % LB agar:
	0.75 (n = 2,446)	1 (n = 2,894)	1.25 (n = 1,419)	1.50 (n = 1,502)	1.75 (n = 1,262)	2 (n = 1,264)	2.50 (n = 913)
Minimum	0.69	0.63	0.61	1.15	0.87	0.53	0.54
25th percentile	3.06	3.12	4.79	6.128	6.36	8.96	12.38
Median	4.015	4.1	6.05	7.79	8.735	11.87	18.21
75th percentile	5.16	5.373	7.45	10.1	11.87	15.08	24.23
Maximum	13.9	14.38	18.83	27.8	34.4	37.59	68.7
Mean	4.152	4.357	6.262	8.472	9.38	12.53	18.84
SD	1.601	1.719	2.219	3.535	4.325	5.118	9.146
SE	0.032	0.0320	0.059	0.091	0.122	0.144	0.303
Lower 95% CI	4.088	4.294	6.146	8.293	9.142	12.24	18.24
Upper 95% CI	4.215	4.42	6.378	8.651	9.619	12.81	19.43

^aCI, confidence interval.

Perturbing LPS structures uncouple swarmer cell elongation from swarm colony expansion.

Further, we found that swarmer cell elongation and single-cell motility could be decoupled from swarm colony expansion and, therefore, population migration. Specifically, motile elongated cells were apparent on the periphery and within the interior of colonies grown on 4% LB agar (Fig. 1B), even though swarm colony expansion was not visible (Fig. 1A). A phenotype similar to that of the 4% LB agar was previously reported in a P. mirabilis mutant strain, observed on 2% agar, that was incapable of producing a strain-specific capsular polysaccharide (CPS); the disrupted gene was named cmfA for “colony migration factor” (44). We reasoned that a subset of previously identified nonswarming mutant strains might also be capable of elongating into swarmer cells. Such hypothetical mutant strains would fail to display swarm colony expansion on standard medium but would differentiate into swarmer cells. These strains would provide additional factors necessary for population migration.

We reanalyzed nonswarming mutant strains that arose from a library of ∼13,000 unmapped mutant strains, which were generated using transposon mutagenesis of strain BB2000 (this study and reference 45). We identified three mutant strains that did not expand on swarm-permissive CM55 (1.5%) agar and contained elongated cells in swarm colonies (Fig. SF2). We mapped the disruptions in the isolated stains using a modified inverse PCR protocol, followed by Sanger sequencing. Mutations were found in genes BB2000_3203, BB2000_3207, and BB2000_3208. Strain KEL21 contains a disruption in the gene BB2000_3203, which encodes a putative O-antigen acetylase. Strain KEL22 contains a disruption in the gene BB2000_3208, which encodes a predicted NAD-dependent epimerase/dehydratase. BB2000_3207 encodes a putative EpsG family protein that is likely a glycosyltransferase. All three genes are in a gene cluster between secB and cpxA (Fig. SF2C). While the gene sequences are not conserved between strains, the predicted functions of the encoded peptides are conserved. Another gene, BB2000_3206, encodes a predicted protein that has 21% pairwise identity to CmfA and is a putative glycosyltransferase family 4 protein. This locus is reportedly associated with the LPS biosynthesis pathway, specifically, the production of O antigen, and varies in the number and composition of genes among wild-type P. mirabilis strains (39, 46). We opted to proceed with mutant strains KEL21 and KEL22.

We next assessed whether the mutations in BB2000_3203 and BB2000_3208 disrupted the LPS structures. We extracted LPS from swarmer cells of each mutant strain and of BB2000 using an iNtRON Biotechnology LPS extraction kit. The resulting fractions were analyzed using SDS-PAGE and a modified silver stain protocol (47). Comparably less LPS staining and nearly complete loss of the O-antigen ladder were repeatedly apparent in samples extracted from the strains KEL21 and KEL22 compared to wild-type strain BB2000 (Fig. 2). In contrast, no differences in LPS were reportedly found between the cmfA-deficient strain and its wild-type parent (44). Therefore, we concluded that the KEL mutant strains are deficient in the production of LPS, and likely, O antigen.

FIG 2.

Strains KEL21 and KEL22 have disrupted LPS structures with apparent loss of O antigen. LPS was independently extracted from surface-grown populations of the wild type, strain KEL21, and strain KEL22. These extracts were run on an SDS-PAGE gel and visualized with silver stain for qualitative comparisons. LPS extractions for the mutant strains were loaded in excess relative to the wild type for increased visualization (5-fold for KEL21 and 2.5-fold for KEL22). Predicted sizes for lipid A core and O-antigen ladder are labeled on left. The image is representative of the results from three biological repeats.

To determine whether the LPS deficiencies correlated with a disruption of O antigen, we examined the susceptibilities of the wild-type and mutant strains to polymyxin B on swarm-permissive surfaces using antibacterial halo assays. Wild-type P. mirabilis is resistant to polymyxin B; however, a loss of LPS results in sensitivity to this antimicrobial peptide (23, 48). We performed halo assays on the wild type, strain KEL21, and strain KEL22. The MIC of the wild type was a saturated solution of 50 mg/ml polymyxin B; however, both mutant strains were inhibited on surfaces at a concentration of 500 μg/ml. Strains KEL21 and KEL22 exhibited a 100-fold increased sensitivity to polymyxin B than the wild type. Therefore, we concluded that these mutant strains KEL21 and KEL22 contained deficiencies in LPS-associated structures.

LPS-defective cells respond in cell shape and colony expansion at a lower range of agar concentrations than the wild type.

We hypothesized that the LPS-associated polysaccharides are a factor for initiating swarm colony expansion after cell elongation. We subjected strains KEL21 and KEL22 to analyses of single-cell and swarm colony morphology similar to those used for the wild type. For single-cell analysis, we constructed derived strains of KEL21 and KEL22 that encoded the chromosomal reporter for fliA expression; these cells were observed in clonal swarms and in mixed swarms with the unlabeled parent strains for size measurements (Fig. 3 and Table 2). We evaluated a range of LB agar concentrations from 0.75% to 1.5%. At 0.75% LB agar, these populations covered the surface of the plate without apparent terrace structures (Fig. 3A and D). KEL21 and KEL22 cells on 0.75% LB agar were short and rigid, with median cell areas of 5.4 and 5.7 μm², respectively (Fig. 3). Areas of ∼5 μm² are comparable in size to wild-type cells grown on 1.25% LB agar. Swarm colonies of both mutant strains exhibited visible terraces on LB agar concentrations of 1% and 1.25% (Fig. 3); these terraces were visibly similar to the wild type grown on 1.5% and 2.5% LB agar (Fig. 1). Flexible elongated cells, aligned into clusters, were readily visible within swarm colonies of the mutant strains. The median cell areas of strains KEL21 and KEL22 were approximately 10 μm² and 13 μm², respectively (Fig. 3). This range of bacterial cell sizes was roughly equivalent to that of the wild type on 2% LB agar. At 1.5% LB agar, populations of each mutant strain were constrained to the site of inoculation (Fig. 3).

FIG 3.

LPS-defective cells swarm and elongate more extensively than the wild type on lower-percentage agar. (A and D) Representative images of strains KEL21 (A) and KEL22 (D) inoculated onto LB medium containing 0.75% to 1.5% agar and grown overnight at 37°C on a 10-cm-diameter petri dish. Colony structure was categorized as unstructured, semistructured, structured, or restricted as denoted in the Fig. 1 legend. Colonies with no swarm colony expansion are also indicated. (B and E) Representative images of strains KEL21 (B) and KEL22 (E) encoding a chromosomal Venus reporter for fliA expression. Cells were inoculated onto 1-mm-thick LB medium pads containing 0.75% to 1.5% agar, grown at 37°C, and imaged using epifluorescence microscopy. Images of swarming populations (on up to 1.25% agar) were taken at the leading edges of swarms, while images of nonswarming populations (on 1.5% agar) were taken at the inoculum edges. Images are phase (top) and fluorescence (bottom). Fluorescence images were subjected to background subtraction. Scale bars = 10 μm. Swarmer cell size on medium containing various concentrations of agar. (C and F) KEL21 (C) and KEL22 (F) cells were comixed in a 10:1 ratio with corresponding cells carrying the fliA reporter strain and swarmed and analyzed and represented as described in Fig. 1C. The numbers of measured cells for each agar percentage, sequentially from 0.75% to 1.25%, are as follows: for KEL21, 2,344, 795, and 823; for KEL22, 1,351, 757, and 796. Data for the wild type at 1.25% agar (1,419 measured cells) is duplicated from Fig. 1C and colored red for visual comparison. Images are representative of the results from three biological repeats.

TABLE 2.

Cell measurements for LPS-deficient mutant strains

Parametera	Cell measurement (μm2) by % LB agar for strain:
	KEL21			KEL22
	0.75 (n = 2,344)	1 (n = 795)	1.25 (n = 823)	0.75 (n = 1,351)	1 (n = 757)	1.25 (n = 796)
Minimum	1.14	1.11	1.64	0.87	1.43	0.86
25th percentile	4.14	7.76	7.94	3.84	6.375	6.088
Median	5.42	11.01	13.36	5.65	9.73	12.76
75th percentile	7.078	14.17	19.47	7.47	14.49	21.02
Maximum	19.7	29.76	42.23	22.54	35.57	68.91
Mean	5.781	11.34	14.29	5.989	10.81	14.57
SD	2.316	4.96	8.001	2.797	5.755	9.696
SE	0.048	0.176	0.279	0.0761	0.209	0.344
Lower 95% CI	5.687	11	13.74	5.84	10.4	13.9
Upper 95% CI	5.875	11.69	14.84	6.138	11.22	15.25

^aCI, confidence interval.

The distributions of cell sizes within population of the mutant strains scaled with agar concentration and were broader than those of wild-type cell sizes at equivalent agar concentrations (Tables 1 and 2). Fluorescence due to the fliA reporter was evident in longer cells of each strain, indicating that expression of FlhD₄C₂-controlled genes was occurring and suggestive of cells entering into the swarm developmental cycle even though swarm colony expansion was not visible. Individual cells of strain KEL21 and KEL22 were capable of moving on swarm-restrictive 1.5% LB agar (see https://www.youtube.com/playlist?list=PLDTC8EPnqPdtZmPVlcKGCm29gMhNz3n3a). Thus, single-cell size and cluster sizes correlated with terrace formations and the width of the rings in the LPS-deficient populations (Fig. 3), similar to the wild-type populations (Fig. 1). Taken together, we concluded that the LPS-deficient strains were not disrupted in the ability to elongate into swarmer cells or for single-cell motility. Instead, we hypothesized that at a lower range of LB agar concentrations, the LPS-deficient populations experienced a decoupling of cell elongation and single-cell motility from the collective motility necessary to achieve centimeter-scale swarm colony expansion.

LPS-associated activity functions locally at individual cells to promote swarm colony expansion.

The LPS deficiencies in KEL21 and KEL22 appeared to inhibit the population from outward expansion on agar concentrations of 1.5% and above. We considered whether LPS, or LPS-associated polysaccharides, might function as a shared swarm-promoting factor among neighboring cells in a swarm. We first examined whether a surfactant could rescue the population migration defect of the LPS-deficient strains. We tested whether the addition of the anionic surfactant Tween 20 impacted swarm colony expansion of the wild type and/or those of the mutant strains. We incorporated increasing amounts of Tween 20 (0% to 0.1%) uniformly into CM55 agar and then inoculated the wild type and strains KEL21 and KEL22 (Fig. SF3). We evaluated the swarm colony radius after growth overnight at 37°C. Strains KEL21 and KEL22 exhibited increased swarm colony expansion on plates containing Tween 20, and the wild type displayed increased terrace widths (Fig. SF3). Therefore, an external surfactant can rescue swarm colony expansion of the LPS-deficient strains.

We next examined whether wild-type cells could rescue the population migration of the LPS-deficient strains in trans; this would ostensibly be achieved through the sharing of the wild-type LPS. To employ differential measurement of swarm colony expansion (45), we engineered constitutive expression of carbenicillin resistance (Cb^r) onto the chromosome of a wild-type population (wild-type::Cb^r) and similarly engineered resistance to streptomycin (Sm^r) onto the chromosome of the wild-type (wild-type::Sm^r), KEL21 (KEL21::Sm^r), and KEL22 (KEL22::Sm^r) populations. We mixed the wild-type::Cb^r and the streptomycin-resistant wild-type or mutant strain populations at 1:1 (Fig. 4A) or at 1:10 (Fig. SF4) initial ratios. The mixtures were inoculated onto 1%, 1.25%, and 1.5% LB agar and onto CM55 (1.5%) agar and then grown for 20 h at 37°C. The swarm colony was sampled using a 48-pronged pinner and then plated in duplicate onto non-swarm-permissive selective medium containing either 100 μg ml⁻¹ carbenicillin or 25 μg ml⁻¹ streptomycin to provide a granular spatial measurement for swarm colony expansion. On CM55 and 1.5% LB agar, which are swarm permissive for the wild-type strain and swarm restrictive for the LPS-defective strain, we found that populations of the KEL21 and KEL22 strains were restricted close to the center, while the wild-type population was at the leading edge regardless of initial ratio (Fig. 4A). Under the 1% LB agar conditions where all populations can swarm outwards when alone, the LPS-deficient strains were found near or at the leading edge along with the wild-type cells (Fig. 4A). An intermediate migration distance was observed for 1.25% LB agar (Fig. 4A). We do not think that the restriction of LPS-deficient cells within a mixed swarm on CM55 agar was due to competition between wild-type and LPS-deficient populations for two reasons. First, altering the starting ratio to have 10 LPS-deficient cells to each wild-type cell did not impact the final result. Second, LPS-deficient cells and wild-type cells were each found on the outer edges of the swarm colony on 1% LB agar. We concluded that since the wild type did not rescue the mutant strain for swarming in trans, LPS did not function as a broadly shared factor in these swarm colonies.

FIG 4.

The swarm motility of LPS-deficient cells is not rescued by the wild type. (A) The wild type constitutively producing carbenicillin resistance (Cb^r) was mixed at a 1:1 ratio with either the wild-type, KEL21, or KEL22 strain, each of which was constitutively producing streptomycin resistance (Sm^r). We measured the distance that each strain migrated within the mixed swarm colony. (B to D) Wild-type cells constitutively producing RFP were mixed at a 1:1 ratio with either wild-type (B), KEL21-derived (C), or KEL22-derived (D) strains carrying the chromosomal Venus reporter for fliA expression. Mixtures were swarmed on CM55 agar pads at 37°C. Phase images (top) and false-colored merge of fluorescence images (bottom) were taken once the first swarm ring formed. Images are taken scanning horizontally from the edge of the inoculum (white haze, far left image) to the leading front of the expanding swarm colony (far right image). RFP-producing cells are false colored magenta; Venus-producing cells are false colored green. Fluorescence images were subjected to background subtraction. Scale bars = 50 μm. Images are representative of the results from three biological repeats.

We next considered whether the lack of rescue for swarm colony expansion could instead be caused by an inability of wild-type and LPS-deficient cells to closely associate. We mixed wild-type cells constitutively producing red fluorescent protein (RFP) at a 1:1 ratio with either the wild-type, KEL21, or KEL22 strain that was engineered to encode the chromosomal Venus reporter for fliA expression and analyzed populations using epifluorescence microscopy (Fig. 4 and SF4). The mixed populations were inoculated onto CM55 agar and permitted to grow at 37°C. Upon emergence of the first swarm ring, swarm populations were imaged continuously in nonoverlapping frames from the inoculation point to the leading edge of the actively migrating swarm (∼2 to 5 mm) (Fig. 4B). Cells swarmed equally together throughout the first swarm ring for the mixed wild-type strains (Fig. 4B). In contrast, cells producing Venus were restricted closer to the center than were RFP-expressing wild-type cells for the mixed populations containing KEL21- or KEL22-derived strains (Fig. 4C and D). Within cell clusters, swarming cells of all populations mixed, indicating that LPS-deficient cells could closely interact with wild-type cells. Similar results were obtained when the wild type was coswarmed at a 1:10 ratio with wild-type, KEL21-derived, and KEL22-derived strains (Fig. SF4). Wild-type cells could not rescue the swarm motility of individual LPS-deficient cells even when physically aligned in an actively expanding swarm (Fig. 4). Therefore, defects associated with loss of wild-type LPS-linked structures appear to be constrained to individual cells. We concluded that the LPS defects in strains KEL21 and KEL22 inhibited collective population migration as a result of impacts on cell elongation and single-cell motility.

DISCUSSION

The ability of bacteria to tune cell morphology and behavior likely allows cells to survive changing environments and may contribute to pathogenicity. P. mirabilis, an opportunistic pathogen, can transverse and colonize catheters (49), and the swarm-promoting regulatory complex FlhD₄C₂ regulates both swarmer cell elongation and the expression of some virulence genes (14, 50, 51). However, gaps remain in our understanding of the full suite of interactions among environmental conditions, structural components of the cell (flagella and the cell surface), single-cell phenotypes (cell shape and single-cell motility), and collective behaviors (population-wide swarm motility). Here, we examine how rigid surfaces influence how this bacterium swarms, both as individual cells and as a population. Altogether, these findings demonstrate that cell elongation and single-cell motility, as well as the apparent size of cell clusters, are coupled with population migration on low-percentage hard-agar concentrations (1% to 2.5%) and are decoupled at high-percentage hard agar concentrations (>2.5%) (Fig. 5). We also find that as cell length increases, the width of each terrace in the swarm colony decreases, though the cause of this correlation remains unknown. These data indicate that LPS-linked structures on each cell function to widen the range of agar concentrations for which collective migration is permissive, likely by altering the interactions between individual cells (Fig. 5).

FIG 5.

P. mirabilis populations occupy a spectrum of cell shapes and cell-to-cell interactions, depending on agar concentration. (A) We propose that P. mirabilis populations have a gradient of responses to rigid surfaces. On such surfaces, e.g., >1% LB, barriers to motility are likely related to local moisture, friction, and tension (26). On lower-percentage hard agar, heterogeneous cell elongation, cell-cell alignment, and single-cell motility are coupled with population migration as agar concentration increases. On higher-percentage hard agar, such as those previously considered nonswarming, cells still elongate and display single-cell motility; however, these individual behaviors become decoupled from population migration. (B) We propose that LPS-associate structures, such as O antigen, allow cells to broaden the range of agar conditions under which single-cell behaviors (elongation and motility) remain coupled to population migration. We observed that populations of LPS-deficient cells form swarm colonies with narrower terraces than those of the wild type, which could be explained by decreased outward migration, or alternatively, a shorter developmental cycle.

These findings highlight several factors that could be incorporated into existing models for surface adaptation, for single-cell motility on surfaces, and for collective migration. An increase in cell size when swarming on a surface is stereotypical for P. mirabilis cells. However, we found that the distribution of lengths, as well as the maximum absolute length, for P. mirabilis cells increases in response to increasing agar concentration (Fig. 5A). Beyond a specific point (∼2% LB agar), as the heterogeneity of cell lengths increased, the swarm expansion of the population decreased. Therefore, cell size increases may be a basic principle for surface adaptation on harder surfaces; however, simple cell elongation is not sufficient to ensure population-wide expansion.

These results suggest that population migration resulting in colony expansion requires cell elongation and single-cell motility, as well as collective behaviors, such as interactions between clusters of cells. Be’er and colleagues have shown that for B. subtilis, the cell’s aspect ratio correlates to speed and can contribute to the entire population’s migration (52). Further, Jeckel and colleagues recently reported that swarm colony expansion (i.e., population-wide migration) of B. subtilis on 0.5% agar is largely reflective of growth dynamics; cell-cell interactions are the dominant influence on single-cell motility (53). The P. mirabilis results on lower-percentage hard agar (from 1% to 2.5% LB agar) are congruent with these B. subtilis models. We observed that cell elongation and single-cell motility (on the hundreds-of-micrometer scale) are coupled with collective behaviors within the range of agar percentages. Under these conditions, elongated cells display freedom of migration in 360°, can flexibly bend in several directions (including making a U-shaped cell), and often align with adjacent cells. Cells migrate most fluidly when in a cluster with other cells. However, these clusters are transient, as cells come together and disperse continuously (see https://www.youtube.com/playlist?list=PLDTC8EPnqPdtZmPVlcKGCm29gMhNz3n3a). This combination of single-cell motility and collective behaviors allows for the entire population to achieve visible centimeter-wide occupation of a surface. However, these single-cell behaviors decouple from population migration on higher-percentage-agar surfaces (>2.5% LB agar). We hypothesize that this is potentially due to an inability of cells to migrate freely or of clusters to be transient in size and formation (see https://www.youtube.com/playlist?list=PLDTC8EPnqPdtZmPVlcKGCm29gMhNz3n3a). Tracking cells and interactions between cells should help span this gap between single-cell behaviors and population migration as a whole.

The molecular and physical mechanisms driving bacterial cell elongation on surfaces remain to be fully defined. Partridge and Harshey synthesized data from many robust and temperate swarming bacteria to propose that water, friction, and tension are the major barriers for motility that bacteria face when on a rigid surface; bacteria can respond by increasing cell length (26). These results with the LPS-deficient strains point to potential feedback between cell size, flagella, and surface tension. For P. mirabilis, disruptions in LPS biosynthesis pathways were previously reported to prevent surface-associated cell elongation and to repress flhDC via an Rcs-mediated membrane stress response, which inhibits the expression of flagellum- and swarmer-associated genes, such as fliA (22, 24, 30–32). However, cells and populations of strains KEL21 and KEL22 swarm and elongate into swarmer cells on rigid surfaces, and the fliA gene, which is an flhDC-activated and rcsB-inhibited gene, is expressed. The cell elongation with reduced population migration of the KEL strains is similar to a deficient strain; however, the published cmfA-deficient strain reported no defect in LPS composition (44). We did not measure the production of colony migration factor, which is encoded by cmfA, because the composition of the polysaccharide has yet to be determined in BB2000. Given these data, we propose that KEL21 and KEL22 populations exhibit a loss of O antigen rather than a general loss in LPS, which may explain the less severe defects than those in previously studied LPS-deficient mutants. Significant additional research is needed to characterize the LPS-linked polysaccharides of BB2000 and of the KEL mutant strains to interrogate this proposal.

The phenotypes of the KEL strains are congruent with current models for the function of LPS during swarming on a rigid agar surface. Partridge and Harshey propose a synthesized model in which LPS could allow bacteria to extract water from the agar surface and could also possibly reduce friction due to the surface (26). LPS is just one of many factors needed for swarming (26). We hypothesize that KEL21 and KEL22 cells, which likely lack the O-antigen extension of LPS, are less able to extract water from the agar surface and are then sensitized to potential factors preventing migration. Increasing cell length could be a compensatory measure to promote single-cell motility; the number of flagella produced on each cell might increase as cell length increases. Quantitative analysis of flagellar composition per cell at the transition to active motility will be needed to fully interrogate this hypothesis. We further hypothesize that a feedback between sensing of physical restraints on motion and cell size exists, resulting in cells increasing in length until motility is achieved. Rather and colleagues have previously pointed to such a possibility when characterizing the role of MinC in swarmer cell elongation (54). They suggested that inhibition of cell division (which drives swarmer cell elongation) and flhDC-regulated swarmer cell motility might not be codependent (54). However, the molecular and physical mechanisms of such potential feedback between individual cell size, flagella, and surface tension remain to be resolved.

While these findings show that LPS-linked polysaccharides contribute to elongation and motility of individual P. mirabilis cells on rigid surfaces, we also found that these LPS-linked polysaccharides do not appear to directly impact the ability of groups of cells to move. This stands in contrast to one model in which LPS is thought to become detached from individual cells and then act as a diffusive surfactant to help other cells move on the surface (35, 39); studies supporting this model were typically performed with swarming on <0.7% LB agar. An alternative model proposes that surface-motile cells migrate along tracks composed of furrows in the medium lined with extracellular material or surfactants, e.g., in Pseudomonas aeruginosa (55, 56). While P. mirabilis swarm colonies can produce extracellular slime trails that appear to spatially coordinate swarming cells to promote swarm colony expansion (57), the findings with the KEL21 and KEL22 mutant strains do not support these models for P. mirabilis migration on hard agar. Cells of the LPS-deficient mutant strains were not rescued in swarm colony expansion by comingling with wild-type cells under various conditions. Yet, the motility of the mutant strains can be partially rescued by the addition of a nonionic surfactant. These results indicate that wild-type cells do not produce sufficient shared LPS, tracks, or surfactants to complement the single-cell motility or population migration of the KEL mutant strains. We favor the explanation that population migration ultimately relies on a summation, or collective, of individual cell behaviors.

A balance between cell length and flexibility as well as interactions between neighboring cells are likely factors in collective motion. These findings show that P. mirabilis is able to occupy several distinct states along a spectrum of possible intersections between single-cell shape/behaviors, group interactions, and population migration. Therefore, P. mirabilis could be potentially used as an experimental model for the quantitative characterizations necessary for deconvolving emergent behaviors.

MATERIALS AND METHODS

Media.

Liquid cultures were grown in Lennox lysogeny broth (LB). Colonies were grown using LB containing Bacto agar, from 0.75% to 4% agar, or using CM55 agar (Oxoid, Hampshire, UK) (1.5% agar) for motility assays, as indicated in the figures. LSW⁻ agar (58) was used for plating nonmotile colonies. For swarm assays, overnight cultures were normalized to an optical density at 600 nm of 1.0, and the culture was then inoculated with an inoculation needle onto the medium before growth at 37°C. The antibiotics used were 100 μg ml⁻¹ tetracycline, 50 μg ml⁻¹ chloramphenicol, 100 μg ml⁻¹ carbenicillin, and 35 μg ml⁻¹ kanamycin.

Strains and plasmids.

Wild-type strain P. mirabilis strain BB2000 (58) and an RFP-producing strain (59) were previously described. The mutant strains KEL21 (BB2000 with a Tn-Cm^r disruption of BB2000_3203) and KEL22 (with a Tn-Cm^r disruption of BB2000_3208) were constructed in this study. Each strain was generated through mutagenesis of strain BB2000 with mini-Tn5 carrying a chloramphenicol resistance marker within the transposable element and a carbenicillin resistance marker on the vector backbone, as previously described (58). The locations of transposon insertions in strains KEL21 and KEL22 were determined through inverse PCR following the previously described protocol for P. mirabilis (45). Briefly, genomic DNA was extracted from each strain through phenol-chloroform extraction. Genomic DNA was digested with HhaI, and the resultant fractions were ligated with T4 ligase overnight. Transposon-specific primers coupled with random primers were used to PCR amplify regions flanking 5′ and 3′ of the insertion site. The resultant PCR products were purified, subjected to Sanger sequencing with transposon-specific primers by Genewiz, Inc. (Cambridge, MA), and mapped to the BB2000 genomic sequence. Protein sequences were analyzed using HMMER (phmmer) (60) and the UniProtKB database (61).

Strain construction was performed as described in reference 66. For all strains, expression plasmids were introduced into P. mirabilis via conjugation with E. coli SM10 λpir (58). Resultant strains were confirmed by PCR of the targeted region. For construction of the fliA reporter strain, a gBlock (Integrated DNA Technologies, Skokie, IL) encoding the last 500 bp of fliA (P. mirabilis BB2000, accession number CP004022; nucleotides [nt] 1856328 to 1856828), a ribosomal binding site (AGGAGG), a modified variant of the Venus fluorescent protein (63), and 500 bp downstream of fliA (P. mirabilis BB2000, accession number CP004022; nt 1855828 to 1856328) was inserted into pKNG101 (62) at the ApaI and XbaI sites. For construction of the streptomycin-resistant strains, a constitutive mKate2 expression construct flanked by regions homologous to rluA (BB2000_2998) and BB2000_2999 was generated using a gBlock (Integrated DNA Technologies) and inserted into pKNG101 (62) at the ApaI and SpeI sites. For generation of the carbenicillin-resistant strains, a constitutive gfpmut2 construct flanked by regions homologous to rluA (BB2000_2998) and BB2000_2999 was generated by splice overlap extension (SOE) PCR (64) and inserted into pRE107 (65) at the KpnI and XmaI sites. The resultant strains were maintained as merodiploids. All plasmids were confirmed by Sanger sequencing (Genewiz, South Plainfield, NJ).

Antibiotic susceptibility assay.

Cultures were top-spread on LSW⁻ medium and allowed to sit at room temperature until the surface appeared dry (approximately 2 h). Then, 6-mm sterile filter disks were placed onto plates and soaked with 10 μl of polymyxin B (Sigma-Aldrich, St. Louis, MO) solution. A water-alone control was included. Once filter disks dried (approximately 2 h), plates were incubated at 37°C overnight and imaged. The MICs were determined based on the concentration of polymyxin B that produced a clearing in the bacterial lawn.

LPS extraction and analysis.

Cells were grown overnight at 37°C on CM55 agar, and the resultant swarms were harvested with LB broth, as previously described (42). LPS was extracted from cells using an LPS extraction kit, according to manufacturer’s instructions (iNtRON Biotechnology, Inc., Seongnam, Gyeonggi, South Korea). Extracts were resuspended in 10 mM Tris (pH 8.0) buffer and run on a 12% SDS-PAGE gel. Gels were stained using silver stain (47).

Microscopy.

Microscopy was performed as previously described (66). Briefly, agar pads were inoculated from overnight stationary cultures and incubated at 37°C in a modified humidity chamber. Pads were imaged using a Leica DM5500B microscope (Leica Microsystems, Buffalo Grove, IL) and a CoolSNAP HQ2-cooled charge-coupled-device (CCD) camera (Photometrics, Tucson, AZ). MetaMorph version 7.8.0.0 (Molecular Devices, Sunnyvale, CA) was used for image acquisition. Images were analyzed using FIJI (67). Images were subjected to background subtraction equally across the entire image. Venus (maximum excitation, 515 nm; maximum emission, 528 nm) was visualized using a green fluorescent protein (GFP) ET filter cube (excitation, 470/40 nm; emission, 525/50 nm; Leica Microsystems).

Quantifying swarmer cell size.

Mixed swarms of fliA reporter-encoding and -nonencoding populations were mixed at the indicated ratios and allowed to swarm on LB pads containing the indicated amount of agar at 37°C until the first swarm ring was visible by eye. Swarms were imaged using phase and epifluorescence microscopy. We utilized MATLAB and Image Processing Toolbox Release 2018a (The MathWorks, Inc., Natick, MA [68]) to assemble a pipeline for analyzing swarmer cell size from microscopy images. The source for basic structure of the segmentation code, specifically, the section that thresholds the image, was based on reference 69. The source for utilizing other algorithms used in code as well as the assistance in implementing was The MathWorks, Inc. Image Processing Toolbox reference (r2018a) (68). Data were filtered by eye to only include intact individual cells in the analysis.

Mixed swarm assays.

Overnight cultures were normalized to an optical density at 600 nm of 1. For 1:1 and 1:10 mixed populations, samples were prepared at the appropriate ratio from the normalized cultures. Topological mapping was performed as detailed in reference 45. Mixed populations were inoculated onto CM55 agar or 1.5%, 1.25%, or 1% LB agar and then allowed to swarm at 37°C overnight. Once the swarm reached the plate’s edge, a 48-pronged replicator device was used to sample regions of the swarm and then replica plated it onto two independent plates, with one containing streptomycin and the other containing carbenicillin. These antibiotic-containing plates were grown overnight at 37°C and imaged with a Canon EOS 60D camera. On the antibiotic plates, the greatest distance between two spots of saturated bacterial growth was recorded; this value is the reported swarm diameter. For microscopy, mixed cultures were allowed to swarm on CM55 agar plates until the first swarm ring was visible by eye. Plates were then imaged at ×40 magnification, as described above. Images were taken continuously in nonoverlapping frames from the inoculation point to the leading edge of the actively migrating swarm.