Enterobacter swarms spread most across agar when the surface was tilted to an intermediate angle, according to a laboratory preprint. Colonies were largest at 45° and 75°, while those at 150° and 180° were smallest. Late-stage growth followed the same non-monotonic pattern, rising and falling rather than changing steadily with tilt.
The angle also mattered in which direction the surface faced. For matched angles, up-facing acute setups tended to expand more than down-facing obtuse ones: 30° compared with 150°, 45° with 135°, and 75° with 105°. The flat reference configurations were 0° and 180°.
A controlled tilt test
The researchers placed Enterobacter sp. SM3 droplets on semi-solid 0.5% LB agar. They let the colonies grow horizontally for three hours at 30 °C and 90% relative humidity, then transferred the plates to eight orientations and imaged them hourly from three to nine hours.
The primary quantitative measurements used 15 independent plates at each inclination. For edge motion, repeated recordings were averaged within each plate-sector, and sectors with unavailable data were not filled in. The statistical comparisons used eight-angle Kruskal-Wallis tests; models for tilted conditions accounted for plate clustering and tested angle, sector, and their interaction. Regional slope estimates used 95% percentile intervals from 10,000 plate-level bootstrap resamples, with false-discovery adjustment across metric families.
The edge became more directional
The colonies' shape changed along with their size. At nine hours, range-over-mean anisotropy, a measure of how unevenly a colony expanded around its edge, was about 0.1 in the flat references but reached 0.901 ± 0.170 at 75°. Downhill bias was about 1.46 to 1.52 from 75° to 135°, while flat references stayed near one. Late expansion favored the downhill front.
At 75°, the tip-localization index reached 2.16 ± 0.33, compared with values near one at 0° and 180°. The measure captured a concentration of late growth at the leading edge, giving some inclined colonies a comet-like downhill tip.
An edge-resolved particle image velocimetry analysis, or PIV, provided a view of collective motion rather than just the final outline. Tilted plates generally showed lower mean activity, and mean speed varied non-monotonically with angle. At every tilted inclination, speed fell from the upper sector to the leading tip.
The spatial organization of that motion also shifted across the colony. The leading tip had a shorter velocity-correlation length, meaning motion stayed similar over a shorter distance, than the upper sector: 3.66 ± 0.13 micrometres versus 3.99 ± 0.13 micrometres. Vortex cores were smaller at the tip, 7.15 micrometres versus 7.47 micrometres, while vortex density was higher, 1.93 versus 1.71 per 1,000 square micrometres.
A model separates gravity's effects
To interpret the angular pattern, the authors extended a two-phase thin-film model of swarming. In plain terms, the model treated the swarm as interacting layers and separated gravity into a component along the agar and a component pressing into the surface. It then represented different transport, pressure, and water-exchange pathways for those components.
The full simulation and the experiment both peaked at 75° for final area and late growth. When the model kept only the surface-normal component, it produced weak separation and weak anisotropy. Keeping only the surface-tangential component produced strong, time-dependent separation and anisotropy, and matched supplementary-angle pairs collapsed onto one another in that version. The contrast shows that the model assigns different effects to the two components.
That interpretation has a clear boundary. The reduced model was not calibrated to absolute experimental length or time scales, so the comparison is limited to relative angular-response patterns. It cannot by itself supply an exact prediction for colony area, elapsed time, or microscopic velocity.
The result has a narrow reach
The primary quantitative experiment remained anchored to 15 plates per angle, and repeated recordings were averaged within each plate-sector. In other words, the main replication unit was the plate, not every recording.
The paper is an arXiv version 2 preprint dated 3 October 2026. The authors report support from the National Natural Science Foundation of China, grant 12202275, and declare no competing interests.
Paper data and sources
Original title: Inclined surfaces bias bacterial swarming expansion and edge organization
Authors: Pengxi Gu, Zixiang Lin, Hui Ren et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-10-01
DOI: Not available
Original paper · Full text