Fig. 6. Adaptation of the iFFL output level to DNA copy number variation.

a Genetic diagram of the iFFL with hEF1a promoters, showing DNA as the input. No resource competitor was added in this experiment. The output is mNeonGreen. The constitutive TX marker (TagBFP) reports the plasmid dosage delivered to each cell. b Top row: TX marker vs output levels for each sample, overlaid with fits of the iFFL model. For the UR samples, the output is proportional to the TX marker, so we fit with a simple linear formula: output = m ⋅ TX marker. The CV(RMSE) is the root-mean-square error between the model and non-binned data, normalized by the mean of the data (log₁₀-transformed first since the cell-to-cell variance is approximately log-normally distributed). To facilitate better comparability among plots, each bin was sub-sampled with the same number of cells (n = 3000). Bottom row: histograms of the output levels for cells within each color-coded bin (as indicated in the scatters). Data are representative and taken from the first of three experimental repeats. c Correlation between the range of DNA copy numbers over which the output of an iFFL variant is adapted and the number of uORFs in the 5′ UTR of CasE’s transcription unit. The adaptation range is defined as the largest sum of the log widths of contiguous adapted bins in a sample (individual bins shown in Supplementary Fig. 36). Individual experimental repeats are shown separately. d Median expression over time for UR and iFFL variants (including a 5′ UTR-targeted miRNA-based iFFL for comparison— see Supplementary Fig. 38). The absolute accumulated change is the sum of the absolute values of the log₂ changes in median expression between time points, summed from 12 to 120 h. The results are from populations of cells gated positive for either output or TX marker. The iFFL variant with 0x-uORFs is omitted from panels (c, d) because its output level is nearly undetectable. Median values and fit parameters are provided in Source Data.