Figure 1.
Images used for quantification of the GFP-p65 relocation dynamics.
(A) Low magnification (20×) image taken at time 0 in the GFP channel showing a representative field from a time-lapse acquisition experiment. GFP-p65 is mostly localized in the cytoplasm of unstimulated cells. Color-map and scale are shown on the right. (B) HOE channel acquisition of the same field as in A. Nuclei counterstained with the nuclear dye Hoechst 33342 appear as bright blue spots. (C) The GFP image as in A is shown using an adjusted color-map (on the right) The scale is extremely compressed for high GFP intensities (white for values ranging from 6.000 to 20.000) but expanded for values that approximate the background (orange). With these settings, variations in the background intensities can be appreciated; it is reasonable to presume that spatial variation exists for the GFP-p65 signal as well. The background is roughly of one order of magnitude less, but it is high enough to be taken into account, most importantly for the quantification of very low signals as for the nuclear non-zero basal level in unstimulated cells.
Figure 2.
Workflow of the automated method for the quantification of nuclear-to-cytoplasmic dynamics of GFP-p65.
For each frame we generate two images, a HOE image of the stained nuclei (A) and a GFP image of GFP-p65 (B). (C) The HOE image shown in A (pseudocolors) is divided in tiles and by a K-means algorithm we determine a threshold used for an approximate detection of the nuclei. (D) The approximate position of the cells – derived from the approximate nuclei position – is taken into account for the estimation of the background of the images, which is spatially inhomogeneous. (E) The segmentation of the nuclei is refined by computing the average brightness of each nucleus that is used to determine a nucleus-specific threshold (the colorcode is based on the incremental identification of nuclei). (F) An algorithm is used to reconstruct the background for the full image. Of note, the background clearly varies in space. (G) Using the local background estimation we identify the boundaries of each cell. The nuclear to total ratio (NT) of the GFP-p65 protein is then calculated using the segmentation of nucleus and cytoplasm of each cell as described in the Results.
Figure 3.
Non-linear relationship between Nuclear to Total Ratio and Nuclear to Cytoplasmatic Ratios.
Plot of the values of the Nuclear to Cytoplasmic ratio, NC (red), and Nuclear to Cytoplasmic mean Intensity ratio of NF-κB, NCI (green), which would be obtained for all the possible values of the Nuclear to Total ratio of NF-κB, NT (also plotted, in blue). The relationship between the magnitudes is clearly nonlinear, note that NC and NCI diverge as NT gets close to 1. Details on how these magnitudes can be related are given in Document S2.
Figure 4.
MEFs have non-zero nuclear basal levels of NF-κB.
(A) A time series of a cell exposed to 10 ng/ml TNF-α; the insets show the nuclear and cytoplasmic segmentations at different times. (B) Distribution of the initial NT values in unstimulated cells (blue) and maximal NT values in cells stimulated with 10 ng/ml TNF-α (red): the blue histogram indicates that most unstimulated cells have low but non-zero levels of nuclear NF-κB. (C) Image at high magnification (63×) in the GFP channel of an optical slice from an unstimulated cell. The part corresponding to the nucleus shows fluorescence values clearly higher than the background. Nucleoli are the small black areas indicated by arrows. This is representative of cells for which the basal levels of NF-κB are well above the background. (D) Image of the same z-slice as in C acquired in the HOE channel confirms that we are observing an optical slice of the nucleus.
Figure 5.
MEFs responses to TNF-α stimulation.
(A) Graphical representation of a NT time series. A sequence of relative minimum (red ‘x’), maximum (blue ‘x’) and minimum identify a peak. The two quantities θL and θR are used to determine whether the second peak is significant. Other descriptors of the cell response to the stimulus are reported: timing of the response (Tresponse), the response value NT(Tresponse,) and the area under the response peak, defined as the area of the peak confined by its two local minima (shaded in orange). (B)The graph plots the probability of having a short-term fluctuation as a function of their θ value, θfluc. Overall, short-term fluctuations have an average value of θfluc close to 0.012. From over more than 104 short term fluctuations we find none with θ≥0.06. (C) Significant peaks (maximum and minima denoted by red and blue ‘x’ respectively) have θ≥0.06, other fluctuations are considered noise (black ‘x’). (D) Time series obtained with 10 ng/ml TNF-α and (E) 1 ng/ml TNF-α. Blue symbols represent the maximum of the first significant peak observed for each cell in the first 2 hours and denote a responding cell. The black line denotes the average of the time series, which looks strongly damped due to the asynchrony and heterogeneity of the dynamics at single cell level, plotted in green. (F) Fraction of responding cells as a function of TNF-α concentration. (G) Timing of the maximum of the response peak after stimulation as a function of TNF-α concentration. (H) The maximal value of the response peak increases with TNF-α doses and appears to plateau at 10 ng/ml. (I) The area under the response peak is remarkably constant with increasing TNF-α doses. In panels from (F) to (I), each symbol corresponds to an independent experiment; error bars represent standard deviation.
Table 1.
Parameters of the response.
Figure 6.
Spontaneous NF-κB dynamics in the absence of stimuli.
(A) Three NT time series corresponding to three unstimulated cells showing spontaneous activity. The insets show the cells corresponding to the traces (same color code) at time zero and at the maximum of the most conspicuous peaks. (B) Distribution of the peak values observed in unstimulated cells (out of 300 cells tracked, see Table S1). (C) The distribution of the timing of the peaks observed for unstimulated cells (blue) and for cells stimulated with 10 ng/ml TNF-α (red).The p-values are for the null hypothesis “the peaks are evenly distributed (chi-square test with 8 degrees of freedom).
Figure 7.
Clustering analysis of the short-term dynamics of NF-κB.
Clustering of the NT traces for the first 1.5 hours of cells exposed to 10 ng/ml TNF-α, using a K-means algorithm. The individual traces (450) are in cyan and the centroids of the cluster are plotted in blue.
Figure 8.
Heterogeneous dynamics of NF-κB.
(A) Fraction of cells exposed to 10 ng/ml TNF-α that show at least one (blue), two (red) and three (green) peaks as a function of the threshold θ. Each kind of marker is representative of a single experiment. The light black dotted line indicates the threshold above which a peak is significant using our criterion. For θ≥0.06 the fraction of cells having at least two peaks (oscillating cells) is close to 0.8. (B) Examples of oscillating and (C) non-oscillating cells for 10 ng/ml TNF-α. (D) Fraction of cells exposed to 1 ng/ml TNF-α having one (blue), two (red) and three (green) peaks as a function of the threshold θ. (E) Examples of oscillating and (F) non-oscillating cells for 1 ng/ml of TNF-α. (G) Fraction of oscillating cells (with at least two peaks with θ>0.06 in the first 5 hours) for different concentrations. Each marker represents results from an independent experiment. (H) Histogram of the time interval between the first two peaks for TNF-α 10 ng/ml; the mode is close to 90 minutes. (I) The time interval between the second and the first peak (blue) and the third and the second peak (red) is remarkably constant and independent from the dose of TNF-α. (J) Ratio of the values of the second and first peak (blue) and of the third and first peak (red) for oscillating cells. The values are close to 1, indicating that oscillatory peaks are similar in height. (K) Ratio of the second and first peak values θ2 and θ1 (blue) and of the third and first peak values θ3 and θ1 (red) for oscillating cells. Second and third peaks tend to have smaller peak values.
Table 2.
Parameters of the oscillations.
Figure 9.
Schematic representation of the oscillatory features of NF-κB upon decreasing TNF-α stimulation (from 10 ng/ml, red curve, to 0 ng/ml, gray curve).
On the x-axis (time), red numbers and ticks mark the mean period of oscillations that is specific for each stimulation, in minutes. On the y-axis, the mean NT value specific for each stimulation is indicated in red. Numbers over the gray lines indicate the mean interval in minutes between two consecutive peaks (T2-T1 and T3-T2 as reported in Table S1). On the right side, mean values of responding and oscillating cells are reported for each condition. The mean area values are also indicated. It is possible to appreciate that the period is constant and peaks have a minimal variation in height.