Supplementary:
FIG. S1.Analysesofthe BLM diffusiondynamicsinnucleoplasm.
We bleach a 1μm diameter spot in the nucleoplasmA) A series of intensity measurements surrounding the initial bleached spot are fitted to a Gaussian profile. The data shows good agreement with the fit, which thus describes our bleached volume. The initial bleaching profile I0(r) is used in the 3 dynamical models (See methods for details). B) Fitting recovery of fluorescence after photobleaching using the 3 models: Reaction, Diffusion and Reaction-Diffusion. From the AIC model selection we get that the diffusion model is the superior model, follow by the Reaction-Diffusion model (p=0.07) and the reaction model is very inferior (). The effective diffusion coefficient for BLM is 1.34 . The ratio between this effective diffusion coefficient and the theoretical diffusion coefficient based on shape and mass lets us estimate the fraction of bound protein at a given time (see methods). For both WRN and BLM, roughly 90% of proteins are bound at any given time.
FIG. S2.Analysesofthe BLM diffusiondynamicsin nucleoli. We bleach a 1μm diameter spot in a singlenucleoli A) A series of intensity measurements surrounding the initial bleached spot are fitted to a Gaussian profile. The data shows good agreement with the fit, which thus describes our bleached volume. The initial bleaching profile I0(r) is used in the 3 dynamical models (See methods for details). B) Fitting recovery of fluorescence after photobleaching using the 3 models: Reaction, Diffusion and Reaction-Diffusion. From the AIC model selection we get that the diffusion model is the superior model, follow by the Reaction-Diffusion model (p=0.07) and the reaction model is very inferior (). The effective diffusion coefficient for BLM is 0.13. The ratio between this effective diffusion coefficient and the theoretical diffusion coefficient based on shape and mass lets us estimate the fraction of bound protein at a given time (see methods). For both WRN and BLM, roughly 90% of proteins are bound at any given time.
WRNandBLMdynamicsatdamagesitesoccur ontwoseparatetimescales
We performed FRAP measurements at sites of DNA damage, by first using a high power laser beam (1.8μW) to induce damage, and two hours later using a low power laser beam (0.6μW) to photobleach 1µm at the damage site and measure recovery (see Materials and Methods). Control experiments were performed to show that this low laser power did not induce additional recruitment (see Figure S4). As illustrated in Figure S??and Figure S3, none of the three models fit the data well. This is also seen from the AICc which suggests Diffusion to be the better model, but the p-values are very high for both the Reaction-Diffusion (p=0.67) and Reaction (p=0.45) models.
Two peculiarities are evident from the data presented in Figure ??C. First, not even the full R-D model provides a good fit for the measured recovery dynamics (in particular see the early time points in the inset of Figure ??C). This indicates that the dynamics governing the binding/unbinding of WRN and BLM at damaged sites is a result of complex interactions, possibly involving binding sites of different affinities. Secondly, the dynamics of WRN and BLM at these sites showed a distinctive feature in that only ~85% of the signal is recovered after photobleaching. This lack of recovery means that a ‘mobile’ fraction of WRN and BLM (~85%) exists alongside a smaller (~15%) fraction that is immobile at least on a timescale of minutes. This immobile fraction could represent proteins actively involved in DNA repair. A further observation is that the accumulation of WRN and BLM occurs on a timescale that takes several hours to saturate (see Figure ??A). Since the damage is created immediately after laser irradiation and our FRAP measurements show that diffusion to the damaged site is completed within minutes. This slow accumulation cannot be explained by WRN and BLM diffusing and binding directly to DNA damage. We believe that the simplest model to account for the accumulation of WRN and BLM at the damage site (see Figure 4B) is a conversion process of potential binding sites, , generated by the initial irradiation, into active binding sites for WRN and BLM. Within our model the potential binding sites are converted into active binding sites for WRN/BLM (which we denote B) at a constant rate, followed by binding of WRN/BLM. Because we show that WRN/BLM diffuse to binding sites within a minute, the “find” step is negligible for accumulation occurring over several hours (see Figure 4). The accumulation of WRN and BLM at the damage site is then limited by the rate of newly synthesized binding sites:
Here m is simply a factor to scale the saturation level. The enzymatic rates are and corresponding to half lives for the conversion of potential to actual binding sites of 20 minutes and 40 minutes respectively.
It is important to note that we found the accumulation kinetics to be much slower than previously observed, where the accumulation was reached within a few minutes[23]. This could be due to the frequency of imaging since imaging creates photo bleaching leading to a decrease in Intensity at the damage site. It could also be due to the difference in wavelength used to create the DNA damage. We were careful to keep the imaging frequency as low as possible, in order for the imaging to be the least invasive.
FIGS3: At the site of DNA damage WRN has at least two distinct binding modes. We preformed the FRAP measurements after the accumulation at the damage sites had saturated, which is roughly 2 hours (see figure 4). A)In cells expressingEGFP-WRN, a 1μm spot is bleached in the DSB, as indicated by the orange arrow.B)To identify the shape of the initial loss of intensity, I0(r), aGaussian profile (blue line) is fitted to the measurements of theinitialintensity at a given radius from the center of the bleached region (circles). C) Fluorescence recovery curves (circles) and corresponding best fitsusing: Reaction (solid green line), Diffusion (dashed blue line) and Reaction-Diffusion (solid black line) models.The inset is a zoom-in of the first five seconds. Data points show the averages offive cells and the error bars representthe standard deviations.
FIG. S4.Analysesofthe BLM dynamics at damage site.A) The initial bleaching profile is fitted to a Gaussian profile. The data shows good agreement with the fit. The initial bleaching profile I0(r) is used in the 3 dynamical models, see methods. B) Fitting the fluorescent recovery to the 3 models: Reaction, diffusion and reaction-diffusion, we see that none of the models give a very good absolute fit to the data which is also seen in the model selection criterion where the diffusion model is only marginally preferred to the two other models, R-D model (p=0.26) and reaction (p=0.17), slightly better than WRN but still not good. Even though our models don’t fit the data very well, we find that both WRN and BLM have a fraction of ~15% that are bound very strongly to the damage, which means a very slow off-rate for a subset of the proteins bound at the damage site, indicating at least 2 very distinct binding dynamics at the damage site. Data is from 5 cells and the error bars are the standard deviations.
FIG S5: Accumulation of WRN and BLM at sites of DNA damage occurs on a much slowertimescalethan the diffusion dynamics. A) Tracking the fluorescent signal when DNA damage sites are induced by micropoint irradiation reveals that the accumulation of WRN and BLM at these sites takes hours to reach saturation. Data is averaged from 21 (WRN) and 9 (BLM) cells, normalizing to the highest observed intensity for each cell. Error bars show standard deviation. B) Suggested model for accumulation at the damage site, which assumes that diffusion occurs much faster, compared to a rate-limiting generation of binding sites. Hence accumulation at the damage site is not limited by the time it takes the proteins to find the damage, but by reaction steps at the site. Fitting the model in B) to the data points in A)predicts half-lives of the limiting reaction to be: and
No recruitment for low laser power (0.6μW)
Fig S6. No double stranded breaks were created with the low laser power(0.6μW). As a control to see if the FRAP experiments induced double stranded breaks we bleached cells and monitored them for roughly two and a half minute, but no recruitment was seen.
Fig S7.Immunofluorescence staining for endogenous WRN shows that it localizes to nucleoli.
Fig S8. Imaging frequency can create artificial saturation of DNA damage recruitment signal. Two cells were targeted with high laser power to induce DNA damage. With high imaging frequency ~1 image/second, the recruitment of WRN to DNA damage seemed to saturate in about 1 minute. For the Low imaging frequency 1 image/minute WRN was not saturated after 10 minutes. Note the difference of timescale on the x-axis
Measured Diff Coeff./ Theoretical Diff. Coeff.
/ References
RAD54 / 14 / 14.6 / Hamster ovary cell (Essers 2002)
PCNA / 13 / 18.3 / Hamster ovarycell
(Essers et al. 2005)
RAD52 / 8 / 14.6 / Hamster ovary cell (Essers 2002)
RAD51 / 7 / 17.5 / Hamster ovary cell (Essers 2002)
NBS1 / 3 / 14.6 / Human U2OS
(Lukas et al. 2004)
MDC1 / 2 / 11.7 / Human U2OS
(Lukas et al. 2004)
Ku70 / 0.35 / 15.2 / HeLa cells and B cells
(Rodgers et al. 2002)
Ku86 / 0.35 / 14.7 / HeLa cells and B cells
(Rodgers et al. 2002)
WRN / 1.6 / 12.3 / U2OS This study
BLM / 1.34 / 13.4 / U2OS This study
Table S1: Summary of measured effective and theoretical diffusion coefficients. The theoretical diffusion coefficients have been corrected for shape factors but for some proteins there is still a significant difference between the theoretical and the measured diffusion coefficients..
WRN / BLMNucleoplasm / Models / D / R / R-D / D / R / R-D
AICc / -55.7 / -40.1 / -50.7 / -41.3 / -25.3 / -35.9
BIC / -316.1 / -227.6 / -309.8 / -194.0 / -163.9 / -188.1
<RSS> / 4･10-5 / 1.4･10-3 / 4･10-5 / 6･10-4 / 2.5･10-3 / 6･10-4
PAIC / 4.1･10-4 / 0.08 / 3.4･10-4 / 0.07
PBIC / 6･10-20 / 0.04 / 2.7･10-7 / 0.05
Nucleoli / AICc / -58.5 / -37.6 / -53.3 / -44.6 / -37.1 / -39.4
BIC / -254.3 / -202.9 / -248.2 / -213.4 / -200.9 / -207.4
<RSS> / 1･10-4 / 1･10-3 / 1･10-4 / 7･10-4 / 1.1･10-3 / 7･10-4
PAiC / 2.9･10-5 / 0.07 / 0.02 / 0.07
PBIC / 6･10-12 / 0.05 / 1.9･10-3 / 0.05
Damage / AICc / -31.6 / -30.8 / -30 / -34.3 / -30.8 / -31.6
BIC / -181.2 / -180.3 / -183.8 / -193.6 / -188.4 / -192.8
<RSS> / 2･10-3 / 1.8･10-3 / 1.3･10-3 / 1.8･10-3 / 2･10-3 / 1.4･10-3
PAiC / 0.67 / 0.45 / 0.17 / 0.26
PBiC / 0.27 / 0.174 / 0.07 / 0.67
Table S2:Model Selection. The Akaike Information Criteria (AICc), Akaike probability (p) and the average residuals sum of squares (<RSS>) are shown for respectively Diffusion (D), Reaction(R), and full Reaction –Diffusion (R-D) models. The values for preferred models are shown in blue.
WRN / BLMNucleoplasm / Diffusion:
/ Diffusion:
Nucleoli / Diffusion:
/ Diffusion:
Damage / None of the models fit the data well / None of the models fit the data well
Table S3: The parameters of the best fits are shown for the models selected by AICc. The confidence intervals report the range of the parameters that with probability, p>=0.95, describe the data equally well as the best fitting parameter. For FRAP at the damage site the models do not seem to fit the data very well and we have therefore not selected any of the models.