Figure 3.

GFPi predictably quantifies RAG activity. The 12- and 23-consensus RSS GFPi was co-transfected with serial 2-fold titration of CMV-RAG1/2 constructs (from 1.6/1.4 to 0.2/0.175 μg, final dilution 1/16th of the initial plasmid DNA amount). (A) Representative plots of each RAG expression condition. (B) Recombination efficiency defined by% of GFP⁺ cells inside an RFP⁺ gate (test – background) with means and standard deviations of at least three replicates. (C) Relationship between the number of recombined plasmids r in cells containing n total plasmids (left) and of the corresponding GFP and RFP values (right). The theoretical model assumes r is Poisson distributed with mean ρn, where ρ is the recombination coefficient. Left panel: the dots represent two random realizations of the model for ρ = 0.1 (black) and ρ = 0.001 (gray), while the lines represent the mean values. Right panel: the dots represent the logarithm of the RFP (x) and GFP (y) signals corresponding to the number of plasmids in the left panel, assuming that autofluorescence, amplification, variation and measurement noise result in bivariate normal-distributed around log-mean values (curves). (D) Estimation of apparent recombination coefficient ρ_a in GFPi assays by obtaining the frequency of cells with zero recombination events in binned RFP data (left and center). Left panel: experimental bivariate GFP vs. RFP reporter log-intensities measured in cells transfected with GFPi alone (without RAG) were fitted by the model log(y) = log(x₀ + bx), by least square minimization and minimization of residuals mean and trend. The dashed curve represents the best fit and the full black curve represents the best fit added by four times standard deviation of the residuals. Right panel: experimental bivariate GFP vs. RFP log-intensity measurements in cells co-transfected with RAG were binned according RFP log-intensity and the frequency of GFP⁻ cells in each bin scored (i.e., with GFP values below the black curve). (E) Fitting the model P(r = 0) = exp(−ρ_a(x − x₀)) to the experimental data. The frequency of GFP⁻ cells was plotted vs. the mean RFP log-intensity per bin for the cells with serial dilutions of CMV-RAG1/2 constructs (A), and fitted by the model. (F) The relative and apparent recombination coefficients (respectively ρ and ρ_a) in RAG titration experiments are proportional to the expected RAG activity. The estimates of ρ_a obtained by model fitting (as in (D) were normalized such that the relative recombination coefficient ρ at the highest titer is 1. Both sets of value are plotted as a function of the CMV-RAG1/2 titers, ρ_a = 0.0007 + 0.028[CMV-R1R2], ρ = 0.02 + 0.93[CMV-R1R2]. (G) The frequency of GFP⁺ cells inside the RFP⁺ gate (as in B) is proportional to the logarithm of the relative recombination coefficient ρ.