Transcription factor mobility

使用光漂白后荧光恢复（FRAP），我们之前已经证明，即使转录持续数小时，gfp标记的糖皮质激素受体在特定启动子上的结合时间最多为60秒。类似的结果现在已经在许多其他转录因子和各种其他核蛋白中被观察到。在许多情况下，这些体内的测量结果与试管中的测量结果非常不同，试管中的测量结果通常表明核蛋白（包括转录因子）的结合要稳定得多。因此，为了了解这些蛋白质在活细胞中的功能，测量它们在染色质上的停留时间并了解这与它们在染色质上的功能之间的关系是至关重要的。对于转录因子而言，这个问题转化为转录因子的停留时间与转录因子结合的基因产生的转录物数量之间的关系。为了解决这些问题，我们一直在开发方法来测量转录因子在活细胞内染色质上的停留时间。我们最初使用光漂白后荧光恢复数据（FRAP）结合本实验的数学模型来估计转录因子在染色质上的停留时间。然而，其他小组使用类似的分析程序得到了非常不同的估计，因此我们调查了这种差异的来源。我们发现许多不同的数学模型可以拟合相同的FRAP数据，因此产生非常不同的停留时间。通过评估这些不同的建模方法，我们展示了一些模型中的错误假设如何导致在估计停留时间时出现重大错误。这使我们提出了一种更可靠的方法，通过FRAP分析来进行这些测量。为了验证我们的FRAP方案，我们开发了一种使用荧光相关光谱（FCS）的替代方法，我们证明该方法也能够测量转录因子与染色质的结合。通过比较FCS分析和FRAP分析，我们确定了FCS分析中的错误，我们能够纠正这些错误，从而在FRAP和FCS对停留时间的估计之间取得良好的一致性。FRAP和FCS的一个局限性是，它们依赖于数学模型来描述荧光强度的变化，这些变化是由至少两个潜在的过程引起的，即扩散和结合。这两个过程都不能通过FRAP或FCS直接可视化，因此关于扩散如何发生的错误假设可能导致结合估计的错误。为了更直接地评估扩散和结合是如何在细胞核中发生的，我们开发了活细胞核中转录因子单分子跟踪的方法。这种方法更容易区分扩散和结合，因为与染色质结合的单个分子比通过核质扩散的分子移动得少得多。使用这种方法，我们已经证明单分子跟踪测量的停留时间与FRAP和FCS测量的转录因子p53的停留时间接近。测量活细胞结合的三种不同方法之间的这种合理的一致性表明，我们现在可以使这些测量相当准确。这将使我们能够将注意力转向转录因子停留时间如何影响转录。为了准确估计糖皮质激素受体与启动子结合的时间，我们建立了数学模型来分析光漂白实验后荧光恢复过程中受体的扩散和结合相互作用。我们的模型预测单个糖皮质激素受体在启动子上结合的时间不到一秒。这种非常短暂的结合提出了新的问题，即转录因子如何在如此短的停留时间内组装转录复合体。同时，我们已经表明，光漂白后荧光恢复的不同分析程序可以产生不同的停留时间估计。这表明，活细胞结合参数的估计仍然存在不确定性，需要开发替代测量程序来达到共识估计。为此，我们现在开发了一个数学模型，从荧光相关光谱数据中提取结合估计，并用它来比较光漂白后荧光恢复和荧光相关光谱对同一分子获得的结合估计。我们已经表明，这两种方法是一致的，但前提是在荧光相关光谱的标准方法中进行校正。校正必须考虑到在测量过程中发生的漂白。因此，我们的交叉验证程序有助于识别其中一个程序中的错误，同时有助于提高我们对当前活细胞结合估计的信心。我们还扩展了这些活细胞结合程序，以检查活细胞内单个分子的合作相互作用。为此，我们使用了连接体组蛋白H1，并使用我们的分析程序来分析其与染色质的结合，以估计结合H1分子的比例。我们对野生型H1分子以及一系列缺乏各种关键结合域的突变体进行了这样的研究。通过比较不同突变体中结合的分子的比例，我们可以确定哪些结构域在结合过程中相互作用。简单地说，协作交互域是指当两者都存在时，绑定的分数比单独存在时绑定分数的总和高得多的域。这种方法对于研究活细胞中分子的协同结合是一种普遍有用的方法。最后，我们开发了单分子跟踪技术来监测活细胞核内转录因子的运动。我们正在使用这个程序来估计结合的停留时间转录因子通过测量多长时间，他们保持不动。这反映了它们与染色质结合的时间。初步数据表明，这些时间比荧光相关光谱和光漂白后荧光恢复测量的停留时间快2-5倍。这种差异可能是由于单分子跟踪所揭示的复杂扩散过程，而这些过程没有被纳入用于分析荧光相关光谱或光漂白后荧光恢复的模型中。

Using fluorescence recovery after photobleaching (FRAP), we have previously shown that the GFP-tagged glucocorticoid receptor is bound at a specific promoter for at most 60 seconds, even though transcription persists for several hours. Similar results have now been observed for a number of other transcription factors and for a variety of other nuclear proteins. These in vivo measurements are in many cases very different from measurements made in the test tube which typically have indicated that nuclear proteins, including transcription factors, are much more stably bound. Thus to understand how these proteins function in live cells it is critically important to measure their residence times on chromatin and see how this relates to their functions on chromatin. For the case of transcription factors, this question translates to how does the transcription factor residence time relate to the amount of transcript produced from genes to which the transcription factor binds.In order to address these questions we have been developing methods to measure residence times of transcription factors on chromatin within live cells. We initially used the data from fluorescence recovery after photobleaching data (FRAP) combined with mathematical models of this experiment to obtain estimates of transcription factor residence times on chromatin. However, other groups using similar analysis procedures obtained very different estimates, and so we investigated the source of this discrepancy. We found that many different mathematical models could fit the same FRAP data, and so yield very different residence times. By evaluating these different modeling approaches, we showed how false assumptions in some of the models led to significant errors in the estimation of residence times. This led us to propose a more robust approach to make these measurements by FRAP analysis.To validate our FRAP protocol, we developed an alternative approach using fluorescence correlation spectroscopy (FCS), which we showed is also capable of measuring binding of a transcription factor to chromatin. By comparing the FCS analysis with the FRAP analysis we identified errors in the FCS analysis that we were able to correct and so achieve good agreement between the estimates of residence times by FRAP and FCS.A limitation of both FRAP and FCS is that they rely on mathematical models to describe changes in fluorescence intensity that arise due to at least two underlying processes, diffusion and binding. Neither process can be directly visualized by FRAP or FCS, so an incorrect assumption about how diffusion occurs can lead to an error in the estimates of binding. To evaluate more directly how diffusion and binding occur in the nucleus we have developed methods for single molecule tracking of transcription factors in live cell nuclei. This approach makes it easier to distinguish diffusion from binding since single molecules bound to chromatin move much less than molecules that diffuse through the nucleoplasm. Using this approach, we have shown that the measured residence times by single molecule tracking are close to those measured by FRAP and FCS for the transcription factor p53. This reasonable agreement among three different methods for the measurement of live cell binding suggests that we can now make these measurements reasonably accurately. This will allow us to direct our attention to how transcription factor residence times affect transcription.To obtain a precise estimate of how long the glucocorticoid receptor remains bound to the promoter, we have developed mathematical models to analyze the diffusion and binding interactions of the receptor that occur during the fluorescent recovery after photobleaching experiment. Our model predicts that individual glucocorticoid receptors are bound at the promoter for less than a second. This very transient binding raises new questions about how the transcription complex can be assembled with such short residence times of the transcription factor. At the same time, we have shown that different analysis procedures for fluorescence recovery after photobleaching can yield different estimates of residence times. This shows that there are still uncertainties in the estimation of live cell binding parameters that will require developing alternate measurement procedures to arrive at consensus estimates.Towards this end, we have now developed a mathematical model to extract binding estimates from fluorescence correlation spectroscopy data, and used this to compare binding estimates obtained for the same molecule by fluorescence recovery after photobleaching and by flourescence correlation spectroscopy. We have shown that the two approaches agree, but only if a correction is made in the standard approach for fluorescence correlation spectroscopy. The correction must account for the bleaching which occurs during the measurement process. Thus, our cross validation procedure has helped to identify an error in one of the procedures, and at the same time has helped increase our confidence in our current live cell binding estimates. We have also extended these live cell binding procedures to examine cooperative interactions of a single molecule inside of a live cell. For this purpose, we used the linker histone H1 and analyzed its binding to chromatin by using our analysis procedures to estimate the fraction of bound H1 molecules. We did this for the wild type H1 molecule as well as for a series of mutants that lack various key binding domains. By comparing the fraction of molecules bound in the different mutants, we could determine which domains interact cooperatively in the binding process. In simple terms, cooperatively interacting domains are those for which a much higher fraction is bound when both are present compared to the sum of the bound fractions when either is present by itself. This procedure will be a generally useful one for investigating cooperative binding of molecules in live cells.Finally, we have developed single molecule tracking techniques to monitor transcription factor movement inside of live cell nuclei. We are using this procedure to estimate the binding residence times of transcription factors by measuring how long they remain immobile. This reflects the time that they are bound to chromatin. Preliminary data indicate that these times are somewhat faster by a factor of 2-5 than the residence times measured by fluorescence correlation spectroscopy and fluorescence recovery after photobleaching. The difference may be due to complex diffusion processes revealed by single molecule tracking which are not incorporated into the models used to analyze fluorescence correlation spectroscopy or fluorescence recovery after photobleaching.