Schematic illustration of the photoselection platform.

A: Automated pipeline for iterative photoselection. After selecting trenches for enrichment, cells are imaged to compute growth rates, which are then used to generate projection patterns that selectively kills cells. This process can be iterated to further refine the selected population. B: Hardware outline: Light from a high-power UV LED travels through a fly-eye lens for improved uniformity. A prism positioned directly in front of the digital micromirror device (DMD) ensures that light converges correctly onto the DMD, maximising the area utilised, and ensuring the correct angle of light for reflection into the rest of the optics system. The DMD is an array of micromirrors, which can be individually tilted between two positions (on/off) to dictate the pattern of light projected onto the sample. Patterned UV light travels via three further lenses and is reflected by the dichroic onto the sample, where it damages illuminated cells and prevents further division. When imaging the sample, emmission wavelength light travels back through the dichroic, is further refined by the filter wheel, and subsequently detected by the high-speed camera. L = lens, TL = tube lens, CL = collimating lens, OF = Optical Filter, A = aperture, PD =photodiode. C Overview of hardware peripherals controlled by the photoselection platform.

Characterisation of photoselection with UV and red light with the photosensitiser methylene blue (MB).

A: Segmentation of cells in the mother machine by DeLTA, allowing for identification and localisation of cells for UV projection. B: Targeting of alternating trenches with UV light projection ≈ 8 W cm6(90 % maximum deliverable power) for 300 s (left), and subsequent division rates for each cell (right). Targeted cells division rates highlighted in blue, untargeted in grey. C: Impact of increasing intensities of UV light on alternating targeted and untargeted cells. Relative on:off target killing (higher is better) using D: UV, and E: MB (2 μM) with 645 nm red light.

Characterisation of off-target effects.

A: Experiment to explore local off-target effects, in which an increasingly large strip of UV light was projected down the centre of a sample, and the division rates of the immediate neighbours measured. B: Schematic of an experiment to explore global off-target effects using the excitation wavelength of GFP. An on-target measurement is from illumination of that cell only. An off-target measurement is illumination of an area away from that cell, but subsequent measurement of any excitation that is a result of off-target light. C: Demonstration that as the proportion of trenches illuminated increases, the off-target illumination also increases. D: Quantification of the relationship between targeted area size and off-target light. Gradient (m) w.r.t proportion cell trenches targeted = 0.26.

Enrichment of mCherry cells from a mixed mCherry/GFP population by photoselection.

A: Identification of GFP and mCherry cells was carried out by imaging with a 450 nm LED (GFP) and 565 nm LED (mCherry), segmenting the images with DeLTA and comparing the brightness in each channel. B: The resulting classifications were converted into a projected image. Unidentified or mixed trenches were projected on by default. C: Length of the mother cell in each trench can be used to measure divisions (at which point the cell length sharply decreases, as indicated by dashed red line). When UV projection occurs at time point 40 (dashed blue line), GFP cell length becomes static, indicating no more division, while mCherry cells continue to divide. D: Comparison of the population composition within the mother machine device (pre-enrichment), to the population composition flowing out of the device after UV (post-enrichment).

Simulating multiple rounds of targeted selection.

A: Illustrating the two steps of each selection cycle, noting that off-target effects decrease with each round. B: Survival probability fit from UV dosage experiments (Supplementary UV Selection Model). One unit of UV dose corresponds to ≈ 8 W cm−2 for 300 s. C: Visualising three rounds of simulated selection, including population composition and off-target ratio results from each round. D: Population composition and E: untargeted cell survival over 10 rounds with different UV intensities.