Table 1.
Colocalization coefficients.
Figure 1.
New colocalization algorithm, description.
A) Two-dimensional fluorescence histogram (left panel) of the images on the right. The pixel distribution is highlighted in jet colour map. Pixels displaying fluorescence only in the green or red channel lie on the x or y axis, respectively, pixels displaying fluorescence in both channels are placed along the diagonal, pixels with background signal are accumulated at the origin of the histogram. Four representative rounds of pixel classification are highlighted in pink colours (rounds 3, 15, 30 and 45). On the right, 256×256 pixel sized images with 15 “green channel” and 14 “red channel” objects (circles) with random intensities varying between 40 and 160. A definite number of objects were deliberately set as colocalized (True colocalization). Jet colour bar, number of pixels. Bins, intervals of fluorescence. B) Four representative rounds of our colocalization algorithm. Upper panels, pixel populations that contribute positively to the r×R coefficient; middle panels, pixel populations that contribute negatively to the r×R coefficient and lower panels, the remaining population of pixels. C) Graph bar representing the amount of pixels in the masks shown on B). As the algorithm proceeds with the detection and classification, the positive and negative populations increase, as the remaining population decreases.
Figure 2.
Performance of the new colocalization algorithm in simulated images.
The accuracy of the algorithm was tested in a variety of simulated images with different colocalization extent. A) 160×160 pixel sized images were generated with different number of pixel-sized objects either located at random positions or in the same position (colocalized). Intensity distribution varied between 30 and 155 for both channels. Colocalized pixels/objects had identical non-zero intensities. Noise with intensities distributed uniformly between 10 and 30 was added to both channels. The density of each channel (percentage of the image covered by objects) varied and ranged from 5% to 30% for each green or red channel. Simulations covering the full range of possible colocalization were performed, from no colocalization to all green objects colocalized. The performance of our algorithm is plotted in black circles. Comparatively, it is plotted the performance of an algorithm written after Costes et al. [3] (red asterisks). Pearsońs correlation coefficient for each image pair is plotted multiplied by a factor of 100 (green dots). B) Example of a pair of images generated for this simulation.
Figure 3.
Performance of the new colocalization algorithm in biological samples.
The accuracy of the algorithm was tested in a variety of biological images with a previously known colocalization extent. A, B and C are examples of whole, partial or no colocalization, respectively. Whole colocalization is achieved by labelling NIH cells with both Phalloidin-FITC and Phalloidin-Rhodamine. Partial colocalization is studied by transfection of TrkA-CFP and p75-YFP receptors into PC12 cells (Iacaruso et al., in preparation). Absence of colocalization is achieved by transfection of p75-YFP into HeLa cells and subsequent labelling with MitoTracker Deep Red. Green and red fluorescence channels and the merge of both channels are shown in every case. A mask enclosing the colocalized population of pixels detected with our algorithm is shown (middle left panel). Comparatively, the mask enclosing the colocalized population of pixels detected by the algorithm written after Costes et al. [3] (lower left panel). Additionally, the m1 and m2 maps generated with ours and comparatively, Costeś colocalization masks. On the right, two dimensional fluorescence histograms for each pair of images. Note the disposition of the pixels along the diagonal when colocalization is complete (A), or their segregation towards the x and y axis upon lack of colocalization (C). Bar, 10 µm. Jet colour bar, contribution of each pixel to the m1 or m2 coefficients (i.e., colocalized pixel intensity/whole image intensity). Bins, intervals of fluorescence intensity.
Figure 4.
Image deconvolution: validation for the study of kinase translocation into mitochondria.
HeLa cells co-transfected with mito-DsRed and either A) Akt1-T308A-GFP, B) Akt1-GFP, or C) Akt1-S473A-GFP were analysed by confocal microscopy and images were deconvolved with Huygens Deconvolution Software (Scientific Volume Imaging). A confocal plane of the deconvolved green and red channels is shown individually or merged (upper panels). Colocalization between the images was assessed by our colocalization algorithm and the m1 and m2 maps were generated utilizing the derived colocalization mask (lower panels). The fluorescence intensity profile across the arrow for both green and red channels is shown in the graph. Black arrows indicate the decrease in GFP inside mitochondrial regions. Bar, 10 µm. Jet colour bar, contribution of each pixel to the m1 or m2 coefficients (i.e., colocalized pixel intensity/whole image intensity). On the right, Simulated Fluorescence Process (SFP) volume. SFP volume is generated with an algorithm in which the data is taken as a distribution of fluorescent dye. By modelling a physical light/matter interaction process an image is computed showing the data as it would have appeared in reality when viewed under these conditions.
Figure 5.
Performance of the new colocalization algorithm in the study of protein translocation kinetics.
NIH/3T3 cells transfected with Akt1-GFP and stained with MitoTracker Deep Red were 24 h serum starved and subsequently stimulated with 10% fetal calf serum. Fluorescence intensity of both green (GFP) and red (Mitotracker) channels was followed for 20 min in a confocal microscope. The change in GFP fluorescence intensity after serum stimulation was analyzed in mitochondria by generating a colocalization mask with our colocalization algorithm or comparatively, generating a mask using MitoTracker fluorescence intensity [20], [22]. The graphs show the redistribution of Akt1-GFP in mitochondria assessed by computing the m1 coefficient either with our algorithm or MitoTracker fluorescence generated masks for each of the 3 confocal planes analysed. A series of merged images and m1 maps of representative time points after serum stimulation is shown on the left. Bar, 10 µm. Jet colour bar, contribution of each pixel to the m1 coefficient (i.e., colocalized pixel intensity/whole image intensity).
Figure 6.
MEK1 is present in mitochondria of HeLa cells.
A) HeLa cells transfected with mito-YFP and DsRed-MEK1 either 24 h serum starved or continuously grown in fetal calf serum were analysed by confocal microscopy and images were deconvolved with the Huygens Deconvolution Software (Scientific Volume Imaging). Deconvolved green and red channels of a confocal plane are shown individually or merged (upper panels). The Pearsońs correlation coefficient and Manders Overlap coefficient maps (lower panels) are also shown. The fluorescence intensity profile across the arrow for both green and red channels is shown in the graph. Blue arrows indicate an increase in MEK fluorescence intensity in mitochondrial areas. Bar, 10 µm. Jet colour bar, contribution of each pixel to the m1 coefficient (i.e., colocalized pixel intensity/whole image intensity). B) Pearsońs correlation coefficient (r) and Manders overlap coefficient (R) were estimated for the image enclosed in the colocalization mask generated by our algorithm, for both serum starved and serum grown cells. R was significantly higher for cells continuously grown in serum, which suggests that MEK accumulates in mitochondria in this condition (p = 0.055, n = 6, Students t test). C) Significance of r and R was determined by comparison with those r and R values obtained when one of the images was repeatedly scrambled. The graph shows the empirical distribution of r and R values for independent (scrambled) images. Red line, R or r distribution adjusted to a normal fit. The R and r values obtained for the original pair of images is far beyond the probability distribution of random r or R, indicated by the black arrows.
Figure 7.
Kinetics of MEK1 translocation into mitochondria.
A) HeLa cells transfected with DsRed-MEK1 and stained with MitoTracker Green were 24 h serum starved and subsequently stimulated with 10% fetal calf serum. Fluorescence intensity of both green (MitoTracker) and red (MEK) channels was followed for 20 min in a confocal microscope. The change in DsRed fluorescence intensity after serum stimulation was analyzed in mitochondria by generating a colocalization mask with our algorithm and subsequently generating the m1 map and estimating m1 coefficient value. A series of merged images and m1 maps of representative time points after serum stimulation is shown on the left. For comparative purposes, MEK redistribution was evaluated in time prior to serum stimulation. A series of representative merged and m1 images is shown on the right. Bar, 10 µm. Jet colour bar, contribution of each pixel to the m1 coefficient (i.e., colocalized pixel intensity/whole image intensity). B) Changes in DsRed-MEK1 fluorescence intensity prior (right) or after (left) serum stimulation in mitochondria assessed by computing the m1 coefficient using the colocalization mask generated with our algorithm for each of the 5 confocal planes analysed.
Figure 8.
Intracellular redistribution of MEK and ERK.
Scheme showing the new perspective of the redistribution of MEK and ERK inside the cell. As mitochondria are the source of ATP, MEK and ERK would go to the mitochondrial surroundings where phosphorylation might occur and afterwards ERK would translocate into the nuclei. There, ERK interacts with transcription factors and the transcription machinery to enhance RNA synthesis. The scheme was taken and modified from Galli et al. [20].