Extended Data Fig. 1: Microscopic analysis of GFP distribution.

We tested three circuit variants: (A) the variant circuit encoding a perfectly matching spacer and constitutively expressing GFP-ssrA (circuit used in Fig. 2); (B) the variant circuit encoding a perfectly matching spacer and expressing GFP under pLac (circuit used in Extended Data Fig. 2); and (C) the variant circuit encoding a mismatching spacer expressing GFP under pLac (circuit used in Fig. 3). Overnight culture carrying each of the three circuits were first primed by diluting 1/100-fold in LB + 25 µg/mL Cm for 3 hours (37 °C, 225 rpm) in 50 mL Erlenmeyer flasks. Then, the cells were diluted another 1/10-fold and distributed into 4 types of media: (1) LB + 25 µg/mL Cm, (2) LB + 25 µg/mL Cm + 100 ng/mL ATc, (3) LB + 25 µg/mL Cm + 50 µg/mL Kan, and (4) LB + 25 µg/mL Cm + 50 µg/mL Kan + 100 ng/mL ATc in a 2-mL deep well plate. After another 3 hours of incubation (37 °C, 700 rpm), 1.5 µl of each sample was loaded on glass sides and was covered with cover slips. Cells were imaged on a microscope (Keyence, BZ-X800). At least five frames with moderate cell density were taken for each sample type. The acquired images were analyzed using standard Python packages (Numpy (version 1.25.2), Pandas (version 2.0.3), and Skimage (version 0.23.2)). The GFP values were log transformed before plotting the histogram for clearer visualization. All circuits showed a bimodality when the CRISPR/Cas9 interference was induced. When the spacer is perfectly matching, the bimodality is stronger, especially when Cas9-mediated cutting is not induced, suggesting a substantial basal level cutting activity.