Fig 1.
The signals reported by the cytometer.
As a particle enters the laser beam, an electric signal (pulse) is generated which reaches its maximum when the particle is in the middle of the beam and trails off as the particle leaves the beam. Each pulse with height over a certain threshold is recorded and three quantities are reported: height, area, and width of the pulse.
Fig 2.
Mixture model fitting of the scatter signals.
The panels show different two-dimensional projections of the full 4D distribution of heights (H) and widths (W) of forward- (FSC) and side-scatter (SSC) measurements for 5 × 104 events obtained from E. coli cells growing in M9 minimal media with lactose. The ellipses show the contour of the fitted multivariate Gaussian distribution, one standard deviation away in each principal direction. Note that the color indicates the local density of points.
Fig 3.
Correlation between cell size and fluorescence measurements for microscopy and cytometer measurements.
Each panel shows measured GFP fluorescence (vertical axis) and cell size estimates (horizontal axis) of cells growing in M9 minimal media with lactose. The top 2 panels show microscopy measurements from a microfluidic device [47]. The lower 4 panels show fluorescence measurements as a function of size estimates based on forward- (middle 2 panels) or side-scatter (bottom 2 panels) measurements in the flow cytometer (FCM). The squared Pearson correlations between fluorescence and size measurements are indicated in each panel. Note that the color indicates the density of points. The white dots show median values of equally spaced bins along the horizontal axis.
Fig 4.
Autofluorescence measurements.
Each panel shows the measured mean fluorescence (left 4 panels) and variance in fluorescence (right 4 panels) on one day, with each bar indicating the measured value and error bar for one replicate. Two different strains were used (indicated in red and blue) and each was measured in triplicate on each day. The black line and grey bar indicate the estimated averages μd and corresponding error-bars σd for each day d. Note that well G6 on 20/12/2016 appears to be an outlier, possibly due to contamination of the well, which was excluded from the analysis.
Fig 5.
Estimated mean expression levels of different promoters as estimated by FCM and microscopy.
After correcting for autofluorescence, mean fluorescence levels of different promoters (colors) are perfectly linearly correlated between microscopy and FCM measurements, over the entire range of expression levels. The scales of the axes are in natural log and the error bars show the standard error of the mean. Note that the slope of the black line is 1.
Fig 6.
Difference in CV2 between the FCM and microscopy measurements shows FCM measurements contain substantial shot noise.
Top: Difference between the CV2 as measured by the FCM and the microscopy setup for different transcriptional reporters of E. coli promoters (colored points). Both axes are shown on a logarithmic scale. The difference in CV2 scales inversely with mean expression. Bottom: The observed CV2 of calibration beads of three different intensities also decreases as the inverse of mean intensity and this dependence can be well modeled by shot noise (black line), as given by Eq (3).
Fig 7.
Comparison of CV2 from FCM and microscope measurements after correcting for autofluorescence and shot noise.
Absolute difference of the CV2 of different transcriptional reporters of native and synthetic E. coli promoters as estimated from FCM and microscope measurements. The black transparent dots use uncorrected FCM measurements and reproduce Fig 6 in linear scale, while the colored dots are obtained when using the CV2 that are corrected for the FCM shot noise.
Fig 8.
Correlation of CV2 in the FCM and microscope measurements before and after correcting for autofluorescence and shot noise.
Top: The CV2 of the raw FCM fluorescence measurements is consistently higher than the CV2 of fluorescence in the microscope measurements, and there is little correlation between the two. Bottom: Once the FCM measurements are corrected for autofluorescence and shot noise, there is now a good agreement between the CV2 as estimated by FCM and microscopy. Measurements for different promoters are indicated by different colors (see legend) and different points of the same color represent replicate FCM measurements. Only promoters expressing more than exp(4) above the background are shown and the black line in both plots is a line with slope 1 and intercept 0.
Fig 9.
Average forward- and side-scatter of cells show approximate power-law dependence on average cell size.
Each panels shows the average of the logarithm of one of the four scattering signals, i.e. height or width of either forward- (FSC) or side-scatter (SSC), as a function of the average logarithm of cell area for E. coli cells growing in different media (M9 + glucose, glycerol or lactose; MOPS + glucose, see legend) as measured by microscopy [31]. The error bars represent the standard errors of the mean over replicate experiments.
Fig 10.
The CV2 of the scattering distribution is affected by shot noise.
Top panel: The CV2 of the cell areas as a function of mean cell area across growth conditions, as measured by microscopy. Bottom panel: The CV2 of the height of the forward scattering signal as function of the mean height of forward scattering across growth conditions, as measured by FCM. In both panels the colors corresponds to different growth media as indicated in the legend (M9 + glucose, M9 + glycerol or M9 + lactose, and MOPS + glucose).