Photo-addressable microwell devices for rapid functional screening and isolation of pathogen inhibitors from bacterial strain libraries

Abstract

Discovery of new strains of bacteria that inhibit pathogen growth can facilitate improvements in biocontrol and probiotic strategies. Traditional, plate-based co-culture approaches that probe microbial interactions can impede this discovery as these methods are inherently low-throughput, labor-intensive, and qualitative. We report a second-generation, photo-addressable microwell device, developed to iteratively screen interactions between candidate biocontrol agents existing in bacterial strain libraries and pathogens under increasing pathogen pressure. Microwells (0.6 pl volume) provide unique co-culture sites between library strains and pathogens at controlled cellular ratios. During sequential screening iterations, library strains are challenged against increasing numbers of pathogens to quantitatively identify microwells containing strains inhibiting the highest numbers of pathogens. Ring-patterned 365 nm light is then used to ablate a photodegradable hydrogel membrane and sequentially release inhibitory strains from the device for recovery. Pathogen inhibition with each recovered strain is validated, followed by whole genome sequencing. To demonstrate the rapid nature of this approach, the device was used to screen a 293-membered biovar 1 agrobacterial strain library for strains inhibitory to the plant pathogen Agrobacterium tumefaciens sp. 15955. One iterative screen revealed nine new inhibitory strains. For comparison, plate-based methods did not uncover any inhibitory strains from the library (n = 30 plates). The novel pathogen-challenge screening mode developed here enables rapid selection and recovery of strains that effectively suppress pathogen growth from bacterial strain libraries, expanding this microwell technology platform toward rapid, cost-effective, and scalable screening for probiotics, biocontrol agents, and inhibitory molecules that can protect against known or emerging pathogens.

FIG. 1.

(a) Microwell recovery arrays (MRA) contain a total of 11 025 co-culture sites available to screen a bacteria strain library (here a 293-member collection of biovar 1 agrobacteria recovered from the rhizosphere of Helianthus annus plants) for antagonistic interactions with a fluorescently labeled test pathogen (here A. tumefaciens sp. 15955 expressing GFP). (b) In the first iteration, unique library strains (denoted by varied shades of red) are combined with the pathogen at an equivalent pathogen-to-strain library cellular ratio, multiple library strains and pathogens may exist in each well. Wells where pathogen growth is significantly diminished are identified as outlier wells (red). (c) Library strains can be cultured against greater numbers of pathogen cells by diluting the strain library into a background of pathogen cells, thereby increasing the pathogen-to-library strain ratio and reducing the number of outlier wells. (d) At a high challenge, library strains are seeded against an excess number of pathogens, enabling the identification of rare outlier wells. These wells contain the strains with the highest levels of pathogen antagonism and are exposed to patterned light to remove a photodegradable polymer membrane, releasing the antagonistic strains from the device for sequencing and further characterization. The figure was prepared with assistance from BioRender software.

FIG. 2.

(a) TLFM images of a sample 15 × 15 array of microwells at 0 and 24 h for 15955-GFP monoculture control and 15595-GFP / strain library co-cultures at increasing 15955-GFP to strain library seeding ratios. White arrows denote wells where 15955-GFP growth appeared diminished. (b) Averaged 24h growth trajectories and (c) averaged growth rates from a representative 900 microwell area of the array. (d) Percent of wells with diminished 15955-GFP growth under each culture condition. Wilcoxon’s two-sample tests were conducted to compare the percent of antagonistic wells with the 15955-GFP monoculture. The percent of wells with diminished 15955-GFP growth in the 1:1 and 10:1 co-culture ratios were significantly higher compared to that in the 100:1 co-culture ratio and hence the 15955-GFP monoculture (*, p-value < 0.01). Data in (b)–(d) were generated from screening at each seeding condition n = 4 times.

FIG. 3.

(a) 20× fluorescent and brightfield images of target wells during photoextraction of antagonistic library strains from a target well in the 100:1 15955-GFP:strain library screening step. The yellow arrow highlights the region where the photodegradable membrane was removed during light exposure. (b) Maximum likelihood phylogenetic tree based on 2024 single-copy orthogroups of the recovered agrobacterial library strains (tip labels color-coded based on source well) and select representatives of genomospecies 1 (G1), G2, G3, G4, G5, G7, G8, G9, G13, and G21 using the JTT + F + Γ substitution model and 100 bootstrap replicates. A. larrymoorei strain FPH AR2 and strain ATCC 5159 were used as outgroup taxa. Nodes with corresponding bootstrap percentages were labeled and labels were omitted from nodes with less than 90% bootstrap support.

FIG. 4.

(a) Representative15955-GFP growth trajectories after inoculation into media conditioned by each individual antagonistic library strain recovered from the MRA screen, or into the unconditioned media (UCM) control. Markers represent averaged OD₆₀₀ readings at 30 min time intervals. The black line with no growth represents conditioned media not inoculated with 15955-GFP, which serves as a negative control to verify that no strains or contaminating microbes were present in the conditioned media. (b) 15955-GFP growth rates and (c) carrying capacities corresponding to each growth trajectory. All growth experiments occurred at 28 °C (n = 3 independent experiments). 15955 growth metrics in conditioned media were compared 15955 growth metrics in UCM in (b) and (c), all showed statistically significant differences (Wilcoxon two-sample test, p < 0.01).

FIG. 5.

Maximum likelihood phylogenetic tree of 1721 single-copy orthogroups of the nine antagonistic library strains recovered in the MRA screen (tip label colors correspond to the well from which the strain was extracted from) from the 293-strain library input into MRA (unlabeled tips) using the JTT + F + Γ substitution model and 100 bootstrap replicates. A. larrymoorei strain FPH AR2 and strain ATCC 5159 were used as outgroup taxa.