The kinetic landscape of human transcription factors
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (ASAPbio crowd review)
Abstract
Cell-to-cell variability is shaped by transcription dynamics because genes are transcribed in bursts interspersed with inactive periods. The stochasticity of bursting means that genes transcribed in rare bursts exhibit more heterogeneity at the single cell level than genes that burst often 1, 2 . Transcription starts with the binding of Transcription Factors (TFs) to specific sequence motifs where they recruit the transcription machinery 3 . In some systems, individual TF binding events temporally correlate with the firing of transcriptional bursts, defining the target gene’s frequency and duration 4–6 . However, in the absence of methods that assess the impact of different TFs on transcription dynamics at the same genetic loci, it remains unclear whether DNA binding kinetics are the sole determinant of bursting. Here we develop an imaging-based synthetic recruitment assay, CRISPRburst, and measure how 92 human TFs impact bursting kinetics. We show that TFs recruited to chromatin under identical conditions generate diverse bursting signatures, some TFs increasing the probability of the gene turning on while others increase the number of mRNA molecules transcribed per burst. We find that the association of TFs with specific protein partners determines their bursting output, and train a model to predict the kinetic signatures of all human TFs. These kinetic signatures can be used as a TF classification system complementary to existing families based on DNA binding domains. Additionally, kinetic signatures provide a rational framework to design synthetic activators, model transcription regulation, and understand expression heterogeneity.
Article activity feed
-
Review coordinated via ASAPbio’s crowd preprint review
This review reflects comments and contributions by Ruchika Bajaj, Sree Rama Chaitanya Sridhara and Wei Chen. Review synthesized by Bianca Melo Trovò.
Genetic transcription happens through individual Transcription Factors (TFs) whose binding events can, in some systems, temporally correlate with the stochastic firing of transcriptional bursts. The determinant of bursting is however unclear, specially whether the DNA binding kinetics solely contributes to that. The study develops an imaging-based synthetic recruitment assay called CRISPRburst in order to measure the TFs impact on bursting kinetics. The authors find that the association of TFs with specific protein partners determines their bursting output, and train a model to predict the kinetic signatures of all human TFs.
Major …
Review coordinated via ASAPbio’s crowd preprint review
This review reflects comments and contributions by Ruchika Bajaj, Sree Rama Chaitanya Sridhara and Wei Chen. Review synthesized by Bianca Melo Trovò.
Genetic transcription happens through individual Transcription Factors (TFs) whose binding events can, in some systems, temporally correlate with the stochastic firing of transcriptional bursts. The determinant of bursting is however unclear, specially whether the DNA binding kinetics solely contributes to that. The study develops an imaging-based synthetic recruitment assay called CRISPRburst in order to measure the TFs impact on bursting kinetics. The authors find that the association of TFs with specific protein partners determines their bursting output, and train a model to predict the kinetic signatures of all human TFs.
Major comments
The manuscript reports that “the maximal intensity per transcription site (TS) is likely limited by physical constraints of the transcription machinery as a limited number of RNA polymerase molecules can be loaded per gene due to polymerase velocity and spacing”. It is recommended to describe how this limitation correlates with the value of active fraction, or could be part of further analysis of this functional data.
‘Functional characterization of TFs using an imaging-based synthetic recruitment assay’ section: “If the frequency and duration of active periods were solely defined by TF binding” [...] “TFs recruited via dCas9 would all exhibit similar active fractions”. This prediction appears to rely on the assumption that the binding rate is the same for all TFs, which is usually not the case.
‘Functional characterization of TFs using an imaging-based synthetic recruitment assay’ section: Given that the TFs that do not bind to the LTR also show high correlation, it is unclear how the correlation for the 6 factors that directly bind LTR justifies that dCas9 recruits TFs in a similar way to the physiological conditions. What is the explanation for the high correlation coefficient for the TFs that do not bind LTR? There is a question as to whether the dCas9 system represents the physiological conditions because the DNA binding kinetics for each TFs are distinct, and different from that for PYL1 binding to ABI1. It would be expected that those different DNA-binding kinetics also contribute to the frequency, duration, or intensity of bursting. Some clarification could be provided around this point.
‘Interactions with co-activators are more predictive of TF kinetic specificity than IDR features’ section “This model was unable to classify TFs into kinetic classes (Figure 3B, right), demonstrating that TF-cofactor interactions play a greater role in specifying kinetic function than IDR sequence content”: Given that TFs interact with cofactors through their transactivation domains, which are intrinsically disordered, why do the TF-cofactor interactions not lead to correlation between IDRs and the kinetic function? Could the protein-protein interactions besides IDR-cofactor (e.g. cofactor-cofactor interactions) play a role in the kinetic function? Do the cofactors cluster into the different kinetic function groups?
Minor comments
Introduction ‘differ in features typically used to classify TFs, such as DNA binding domain homology’: it may be worth making a mention in the introduction to what other binding partners TFs interact with.
First paragraph of results ‘CRISPRburst, an inducible dCas9-mediated recruitment platform to study transcription kinetics’: What is the binding strength of PYL1 to ABI1? How does that compare to the typical TF-DNA binding strength?
Figure 1C: “3) Live cells are imaged 16 h post-recruitment.” This is the end time point. Are there time-dependent data available?
Figure 1 F, G: The error bars are high. Can further information be provided in the legend on how these error bars were calculated (biological vs technical replicates)?
Figure 1, ‘An average of 220 cells were analyzed per TF’ Does this imply that 220 transcription sites were scored? Considering each imaged cell has single integration of the reporter gene?
‘In total, the LTR-MS2 cell line stably expresses 1) the LTR-MBS reporter gene’: Is there information on where in the genome the reporter gene is integrated? And does it impact the transcription bursts? (considering the role of (epi)genetics in the transcriptional outcome as rightly reinforced by the data related to Fig.4).
Functional characterization of TFs using an imaging-based synthetic recruitment assay: Please provide a description for the Krüppel associated box.
“Upon recruitment, 28 TFs generate an increase in reporter active fraction”. It would be helpful to provide further clarification on how the reporter active fraction is defined and how the criteria "ratio > 1.30" was determined. A mathematical equation may also aid the description.
‘0.64 to 3.04 for active fraction and 0.68 to 1.64 for intensity (Figure 1F-G, S1E) ‘: It may be helpful to divide the active fraction (0.64 to 3.04) into different categories, for example, 3.04 - 2.5, 2.5-2.0 etc. to check whether these categories are correlated to function.
Regarding intrinsically disordered regions (IDRs) in the Results section ‘Bursting kinetics define distinct TF classes’: Can further clarification be provided in the main text for the meaning of cumulatively longer IDRs.
“these findings suggest that while the biophysical properties of IDRs may tune the amplitude of TFs’ effects, they likely do not solely encode TF kinetic specialization”: does this include post-translational modifications? If so, are there any relevant examples or illustrations?
In the section ‘TF families exhibit broad kinetic diversity’ section, “the family-defining KRAB domain does not bind DNA but recruits cofactors, consistent with the idea that DNA binding domains provide little information on kinetic specialization (Figure S6B)”. It may be relevant to discuss potential solutions to these issues in the Discussion section.
Discussion section “Our study centered on the simple HIV promoter thus provides a robust conceptual framework to investigate more complex systems, e.g. how TFs synergize with one another, interact with core promoter motifs, or communicate to promoters from distal enhancers”: all the future directions mentioned here are very relevant and exciting. Could the discussion of these items be expanded e.g., how do developmental cues drive TF kinetics or bursts?
Methods section: Are there any anomalies observed in the subcellular localization of the TFs when tagged with PYL1 or under the ABA treatment?
Comments on reporting
The manuscript reports a partial least-squares multivariate regression model in which a predictive weight to each possible interactor was assigned. Can further description and a related equation be provided for this model?
Fig. 3: Can further context be provided for the choice of SEM instead of SD which may provide a better representation of data variability?
-
-