The similarity of population distributions in habitats in the same device could potentially be caused by a coupling between habitats (e.g., diffusion through the PDMS layer which seals the devices), an identical response of the bacteria to device-wide gradients (e.g., of oxygen or temperature) or by other extrinsic variation. We tested for these possibilities using two sets of experiments. First, we used a type-4 device that consists of two habitats separated by 1.2 mm, which are inoculated in reverse order (red from the left in habitat 1 and from the right in habitat 2, Additional file

10B). The patterns in these two habitats were PCI-34051 similar to each other (d = 0.28, Additional file 10A), suggesting that spatial proximity is not a necessity for obtaining similar population distributions in replicate habitats. Secondly, we used devices of type-5 consisting of four parallel habitats, which were inoculated from two sets of initial cultures such that neighboring habitats were

colonized by different cultures (see Methods and Additional files 11 and 12). We found that neighboring habitats inoculated from different initial cultures do not become Crenolanib more similar due to their proximity to each other, with a LY3023414 median difference between patterns in habitats located on the same device, but inoculated from different cultures, of d different  = 0.32 (median, 25%-75% quartiles = 0.27-0.42), which is similar to the observed value of the difference between patterns in habitats

located on separate type-1 and 2 devices, which were inoculated from different cultures, of d different  = 0.38 (median, 25%-75% quartiles = 0.37-0.40; p = 0.32, Wilcoxon rank sum test, N = 8 for selleck inhibitor type-5 devices, N = 10 for type-1 and 2 devices combined, Additional file 9C). This demonstrates that population distributions in neighboring habitats that were inoculated from the same initial cultures are not similar just because of their location next to each other on the same device. For the type-5 devices the difference between habitats inoculated from different initial cultures is calculated by comparing habitats on the same device, while for the type-1 and 2 devices this difference is calculated by comparing habitats located on different devices. To make sure that the calculated values are comparable, we also calculated the difference between habitats located on different devices (and thus inoculated with different cultures) for the type-5 devices. Here we find a median difference of d different  = 0.38 (25%-75% quartiles = 0.37-0.39) which is similar to that of the type-1 and 2 devices (d different  = 0.38 median, 25%-75% quartiles = 0.37-0.40; p = 0.9, Wilcoxon rank sum test), indicating that the calculated values for the differences between population distributions are comparable between the type-5 and the type-1 and 2 devices.