Surprising features of nuclear receptor interaction networks revealed by live-cell single-molecule imaging
eLife Assessment
This important study provides data that challenges the standard model that binding of Type 2 Nuclear Receptors to chromatin is limited by the available pool of their common heterodimerization partner Retinoid X Receptor. The evidence supporting the conclusions is compelling, utilizing state-of-the-art single-molecule microscopy. This work will be of broad interest to cell biologists who wish to determine limiting factors in gene regulatory networks.
https://doi.org/10.7554/eLife.92979.3.sa0Important: Findings that have theoretical or practical implications beyond a single subfield
- Landmark
- Fundamental
- Important
- Valuable
- Useful
Compelling: Evidence that features methods, data and analyses more rigorous than the current state-of-the-art
- Exceptional
- Compelling
- Convincing
- Solid
- Incomplete
- Inadequate
During the peer-review process the editor and reviewers write an eLife Assessment that summarises the significance of the findings reported in the article (on a scale ranging from landmark to useful) and the strength of the evidence (on a scale ranging from exceptional to inadequate). Learn more about eLife Assessments
Abstract
Type II nuclear receptors (T2NRs) require heterodimerization with a common partner, the retinoid X receptor (RXR), to bind cognate DNA recognition sites in chromatin. Based on previous biochemical and overexpression studies, binding of T2NRs to chromatin is proposed to be regulated by competition for a limiting pool of the core RXR subunit. However, this mechanism has not yet been tested for endogenous proteins in live cells. Using single-molecule tracking (SMT) and proximity-assisted photoactivation (PAPA), we monitored interactions between endogenously tagged RXR and retinoic acid receptor (RAR) in live cells. Unexpectedly, we find that higher expression of RAR, but not RXR, increases heterodimerization and chromatin binding in U2OS cells. This surprising finding indicates the limiting factor is not RXR but likely its cadre of obligate dimer binding partners. SMT and PAPA thus provide a direct way to probe which components are functionally limiting within a complex TF interaction network providing new insights into mechanisms of gene regulation in vivo with implications for drug development targeting nuclear receptors.
Introduction
Complex intersecting regulatory networks govern critical transcriptional programs to drive various cellular processes in eukaryotes. These networks involve multiple transcription factors (TFs) binding shared cis-regulatory elements to elicit coordinated gene expression (Gerstein et al., 2012; Pan et al., 2009; Reményi et al., 2004). A distinct layer of combinatorial logic occurs at the level of specific TF-TF interactions, which can direct TFs to distinct genomic sites. Thus, fine-tuning of TF-TF interactions can modulate TF-gene interactions to orchestrate differential gene expression. Often, dimerization between different members of a protein family can generate TF-TF combinations with distinct regulatory properties resulting in functional diversity and specificity (Nandagopal et al., 2022; Puig-Barbé et al., 2023). For example, E-box TFs of the basic helix-loop-helix (bHLH) family such as MYC/MAD share a dimerization partner MAX. Several studies have shown that switching of heterocomplexes between MYC/MAX and MAD/MAX results in differential regulation of genes and cell fate (Amati et al., 1993; Bouchard et al., 2001; Hurlin and Huang, 2006; Xu et al., 2001).
Type II nuclear receptors (T2NRs) present another classic example of such a dimerization network wherein the distribution of heterodimeric species regulates gene expression (Bwayi et al., 2022; Chan and Wells, 2009; Chen et al., 2018; Wang et al., 2017). T2NRs constitute an extensive group of basic leucine zipper TFs that share a common modular structure composed of a well-conserved ligand binding domain (LBD) that mediates heterodimerization with their obligate partner retinoid X receptor (RXR) and a highly conserved DNA binding domain that recognizes consensus sequences termed direct response elements (Evans and Mangelsdorf, 2014). Unlike Type I nuclear receptors, T2NRs do not depend on ligand binding for DNA engagement. Rather, binding of ligands to the LBD of chromatin- associated T2NR heterodimers results in eviction of co-repressors and recruitment of co-activators (Evans and Mangelsdorf, 2014; McKenna and O’Malley, 2002; Figure 1A).
-
Figure 1—source data 1
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data1-v1.pdf
-
Figure 1—source data 2
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data2-v1.pdf
-
Figure 1—source data 3
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data3-v1.pdf
-
Figure 1—source data 4
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data4-v1.pdf
-
Figure 1—source data 5
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data5-v1.zip
-
Figure 1—source data 6
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data6-v1.zip
-
Figure 1—source data 7
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data7-v1.zip
-
Figure 1—source data 8
- https://cdn.elifesciences.org/articles/92979/elife-92979-fig1-data8-v1.zip
As the common obligate partner of many other TFs, MAX and RXR are thought to act as the ‘core’ for their respective dimerization networks. It has been postulated that availability of such core TFs is likely to be limited in cells, resulting in competition between the various partner TFs involved in the network (Chan and Wells, 2009; Walker et al., 2005). Indeed, early studies using purified proteins revealed that MAD and MYC compete for binding to MAX with equal affinities, and reduced complex formation was seen for either heterodimer with increasing amounts of a competing partner (Ayer et al., 1993; Baudino and Cleveland, 2001). Similarly, the T2NRs liver X receptor (LXR) and peroxisome proliferator-activated receptor (PPAR) were observed in vitro to have reduced binding to their respective response elements in the presence of competing T2NRs (Ide et al., 2003; Matsusue et al., 2006; Yoshikawa et al., 2003).
Whether such in vitro systems would capture the complexity and competitive dynamics at play in live cells has remained an unresolved issue. Moreover, to date, in vivo studies of competitive heterodimerization networks have not examined TFs at endogenous expression levels and thus may not accurately recapitulate their dynamic interactions with each other and with chromatin (Fadel et al., 2020; Grinberg et al., 2004) or do not account for expression levels of relevant TFs in individual cells due to the population averaging nature of most such studies (Rehó et al., 2023). To overcome these potential shortcomings, we employed fast single-molecule tracking (fSMT) (Boka et al., 2021; Dahal et al., 2023; Elf et al., 2007; Hansen et al., 2018; Hansen et al., 2017) and its newly developed complement, proximity-assisted photoactivation (PAPA-SMT) (Graham et al., 2022) to test the effects of varying the stoichiometry of a core TF (RXRα) and its partner TF (RARα). We find that, contrary to expectations, the core component in cancer cells is not limiting but rather is in sufficient excess to accommodate more RARα even in the presence of many other endogenous partner T2NRs.
Results
Live-cell SMT of KI Halo-tagged RAR and RXR
As an initial test case for studying T2NR interactions, we focused on the heterodimeric partners RARα and RXRα, which are endogenously expressed alongside various other T2NRs in U2OS cells, a well-established cancer cell line for SMT (Hansen et al., 2017; McSwiggen et al., 2019; Figure 1A, Appendix 2—table 1). Using CRISPR/Cas9-mediated genome editing, we generated clonal lines with homozygous knock-in (KI) of HaloTag at the N-terminus of RXRα and the C-terminus of RARα (Heckert et al., 2022b; Los et al., 2008; Figure 1—figure supplement 1A). Western blotting confirmed that RARα and RXRα were tagged appropriately and expressed at similar levels to the untagged proteins (Figure 1B), while co-immunoprecipitation (co-IP) experiments verified that Halo-tagged RARα and RXRα heterodimerize normally as expected (Figure 1—figure supplement 1B). In addition, we confirmed using luciferase assays that the RAR ligand, all-trans retinoic acid (atRA), activated retinoic acid responsive element (RARE)-driven gene expression in wild-type and homozygously edited clones, confirming the normal transactivation function of the tagged proteins (Figure 1—figure supplement 1C). Confocal live-cell imaging of cells stained with Janelia Fluor X 549 (JFX549) Halo ligand displayed normal nuclear localization of both Halo-tagged RARα and RXRα (Figure 1—figure supplement 1D).
To evaluate how RARα and RXRα explore the nuclear environment and interact with chromatin, we used fSMT with a recently developed Bayesian analysis method, SASPT (Heckert et al., 2022b), to infer the underlying distribution of diffusion coefficients within the molecular population, yielding a ‘diffusion spectrum’ (Figure 1C and D). From these diffusion spectra we can extract quantitative parameters of subpopulations (peaks) such as mean diffusion coefficients and fractional occupancy, allowing us to measure the chromatin-bound fraction (fbound) in live cells (Figure 1C and D). To benchmark our SMT measurements, we first compared the diffusion spectra of RARα-Halo and Halo-RXRα to that of H2B-Halo (which as expected is largely chromatin-bound) and Halo-3×NLS (which is mostly unbound). We observed that RARα and RXRα both exhibit a clearly separated slow diffusing population (<0.15 μm2/s) along with a faster mobile population (1–10 μm2/s) (Figure 1C). The former we classify as ‘bound’ since it represents molecules diffusing at a rate indistinguishable from that of Halo-H2B (chromatin motion). The proportion of molecules diffusing at <0.15 μm2/s is henceforth designated as fbound. In comparison with Halo H2B (fbound = 75.5 ± 0.7%) and Halo-NLS (fbound = 10 ± 0.9%), RARα and RXRα have intermediate levels of chromatin binding (fbound of 50 ± 3.0% and 38 ± 3.5%, respectively) (Figure 1D). The fbound was reproducible between three clonal cell lines of RARα (49 ± 1%, 47 ± 1%, 53 ± 1%) and RXRα (36 ± 1%, 36 ± 1%, 42 ± 1%). Finally, although recent studies have reported an increase in chromatin interaction upon agonist treatment (Brazda et al., 2014; Brazda et al., 2011; Rehó et al., 2023; Rehó et al., 2020), we did not observe a significant change in fbound of RARα and RXRα upon atRA treatment, nor did atRA treatment appear to alter the fast-moving populations of RARα and RXRα (Figure 1—figure supplement 3A and B). These results are consistent with the classic model in which dimerization and chromatin binding of T2NRs are ligand independent.
Chromatin binding of RARα and RXRα can be saturated in live cells
Chromatin binding by an individual TF within a dimerization network is predicted to be sensitive to its expression level (Klumpe et al., 2023). To test how the expression levels of RARα and RXRα affect fbound, we overexpressed Halo fusions from stably integrated transgenes in U2OS cells (Figure 2A). After confirming transgene expression by western blotting (Figure 2A), we compared the abundance of endogenous and exogenous Halo-tagged RARα and RXRα using flow cytometry (Cattoglio et al., 2019; Figure 2—figure supplement 1A). Average cellular abundance of endogenous RARα and RXRα obtained from biological replicates of each homozygous clone were similar (Figure 2B). In contrast, expression of exogenous RARα-Halo was approximately four times that of the endogenous protein, while expression of Halo-RXRα was nearly twenty times that of endogenous RXRα (Figure 2B). Chromatin binding of overexpressed RARα was reduced by approximately half compared to endogenous (fbound = 27 ± 0.72%) (Figure 2C, Figure 2—figure supplement 1B), while that of overexpressed RXRα was decreased even more dramatically to 11 ± 0.75% - a value barely above the fbound of the Halo-NLS control (Figure 2C, Figure 2—figure supplement 1B). Using nuclear fluorescence intensity as a rough proxy for protein concentration in individual cells, we observed a negative correlation in single cells between TF concentration and fbound (Figure 2—figure supplement 1C). These results imply that chromatin binding of both RARα and RXRα in U2OS cells is saturable - i.e., the total number of chromatin-bound molecules does not increase indefinitely with expression level but is in some way limited.
RARα limits chromatin binding of RXRα
We next assessed how overexpression of RARα and RXRα affects fbound of the endogenous proteins by stable integration of SNAP-tagged RARα or RXRα transgenes in Halo-KI RARα and RXRα cell lines (Figure 2D). As controls, we also stably integrated transgenes expressing SNAP-NLS or a SNAP-tagged dimerization-incompetent RARα (RARαRR) (Bourguet et al., 2000; Zhu et al., 1999; Figure 2D). We validated disruption of the RARαRR-RXRα interaction using Rosetta modeling (Shringari et al., 2020) and co-IP (Figure 2—figure supplement 2A). Using flow cytometry, fluorescent gels, and western blots, we first assessed if transgene expression alters expression of endogenous Halo-tagged RARα and RXRα (Figure 2E, Figure 2—figure supplement 2B–D). While no drastic changes in the cellular abundance of KI Halo-tagged RARα or RXRα was observed in the presence of SNAP, SNAP-RXRα, or RARαRR-SNAP proteins, the abundance of KI Halo-tagged RARα was approximately halved when RARα-SNAP was overexpressed (Figure 2E, Figure 2—figure supplement 2B). This is likely distinct from previously reported ligand-dependent RARα degradation (Tsai et al., 2023; Zhu et al., 1999) (see Discussion).
We then carried out a series of fSMT experiments to understand how fbound of the endogenous RARα or RXRα is altered when its binding partner is present in excess (Figure 2—figure supplement 4A and B). Surprisingly, we found that the fbound of endogenous RARα-Halo (47 ± 1%) was largely unchanged upon expression of RXRα-SNAP (49 ± 1%), consistent with the control SNAP (47 ± 1%) (Figure 2F, Figure 2—figure supplement 4B), implying that RARα fbound is not limited by the availability of RXR. In contrast, RARα-SNAP expression significantly decreased chromatin binding of endogenous RARα-Halo to 29 ± 5% (Figure 2F, Figure 2—figure supplement 4B). However, overexpression of mutant RARαRR-SNAP did not change the fbound of endogenous RARα (43 ± 4%) (Figure 2F, Figure 2—figure supplement 4B), suggesting that heterodimerization with RXRα is required for this effect.
In the reciprocal experiment with endogenous Halo-RXRα, the initial fbound (35 ± 1%) was barely altered by overexpression of SNAP (38 ± 1%) or RARαRR-SNAP (35 ± 1%). As expected, overexpression of SNAP-RXRα reduced the fbound of endogenous Halo-RXRα to 16 ± 2%. In contrast to RARα, however, overexpression of RARα-SNAP increased the fbound of endogenous Halo-RXRα to 56 ± 1%. This result suggests that endogenous RXRα chromatin binding is limited by the availability of RAR (Figure 2F, Figure 2—figure supplement 4B), and not due to limiting amounts of the universal dimer core partner RXR as would be expected based on current models in the literature.
While it may seem paradoxical that RAR is limiting for RXR binding, given the similar number of molecules of endogenous RARα and RXRα per cell (Figure 2B), this is likely due to the presence of other endogenous RXR paralogs (see Appendix 2—table 1 and Discussion). It is also probable that some fraction of RXR binding to chromatin arises from other T2NRs that can produce chromatin binding-competent dimers with RXR (Figure 1A, Appendix 2—table 1, and Discussion). Notwithstanding these complexities, the results of the above SMT measurements in the presence of varying amounts of partner TFs allow us to infer that endogenous RXRα is likely in excess while RXR partners are limiting in U2OS cells.
PAPA-SMT shows that RARα-RXRα dimerization correlates with chromatin binding
The above results support a model in which the total RXR pool (including all paralogs) is in stoichiometric excess of its partners in U2OS cells. To more directly monitor RXR-RAR interactions, we employed a recently developed SMT assay by our lab, PAPA (Graham et al., 2022). PAPA detects protein-protein interactions by using excitation of a ‘sender’ fluorophore with green light to reactivate a nearby ‘receiver’ fluorophore from a dark state (Graham et al., 2022; Figure 3A). As an internal control, violet light is used to induce direct reactivation (DR) of receiver fluorophores independent of their proximity to the sender (Figure 3A). See Appendix 1 for a more detailed description of the steps involved in a PAPA-SMT experiment.
We applied PAPA-SMT to cell lines expressing SNAP-tagged RARα and RXRα transgenes in the background of Halo-tagged endogenous RXRα or RARα, respectively (Figure 3B). As controls, we also imaged SNAP-3×NLS and RARαRR-SNAP, which are not expected to interact with Halo-RARα or RXRα. First, we plotted the number of single-molecule localizations reactivated by green and violet light in each cell (Figure 3B). As expected, a low level of background reactivation by green light was observed for SNAP negative control, reflecting the baseline probability that an unbound SNAP molecule will at some low level be close enough to Halo to observe PAPA (Figure 3B, orange points) (Graham et al., 2022). Violet light-induced DR and green light-induced PAPA were linearly correlated as expected, since both are proportional to the number of receiver molecules. However, a greater PAPA signal was seen for the Halo-RXRα → RARα-SNAP and RARα-Halo → SNAP-RXRα combinations than for the negative control, consistent with direct protein-protein interactions (Figure 3B). The relation between DR and PAPA was sublinear (Figure 3B, left and middle panel, see residuals of the linear fit in Figure 3—figure supplement 1B), consistent with saturation of binding to the Halo-tagged component. In contrast, the ratio of PAPA to DR for Halo-RXRα → RARαRR-SNAP was similar to the SNAP negative control, confirming that PAPA signal depends on a functional RAR-RXR interaction interface (Figure 3B, right panel).
Next, we compared the diffusion spectra of molecules reactivated by DR and PAPA (Figure 3C). For both Halo-RXRα → RARα-SNAP and RARα-Halo → SNAP-RXRα combinations, PAPA trajectories had a substantially higher chromatin-bound fraction than DR trajectories, indicating that a greater proportion of SNAP-tagged RARα/RXRα binds chromatin when it is in complex with its Halo-tagged partner. Only a slight shift in bound fraction was seen for RARαRR-SNAP (fbound,DR = 7.5 ± 0.6%; fbound,PAPA = 11.1 ± 0.8%), comparable to that seen for the SNAP negative control (fbound,DR = 7.1 ± 0.5% and fbound,PAPA = 11.2 ± 0.9% for Halo-RXRα; fbound,DR = 7.3 ± 0.4% and fbound,PAPA = 9.3±0.7 for RARα-Halo) (see Discussion), confirming that RARαRR fails to form chromatin binding-competent heterodimers with RXR.
The enrichment of chromatin-bound molecules in the PAPA-reactivated population of both RAR and RXR is consistent with heterodimerization mediating chromatin binding, that is further validated by the low fbound of the dimerization-incompetent mutant.
Discussion
According to the current consensus models for how nuclear receptor interaction networks operate, specific T2NRs members compete for a limited pool of the ‘core’ partner RXR in cells, thus establishing a competitive regulatory network (Chan and Wells, 2009; Fadel et al., 2020; Rehó et al., 2023; Wang et al., 2005; Wood, 2008; Yoshikawa et al., 2003). However, previous in vivo studies were carried out with overexpressed proteins, limiting their ability to accurately address this model of competition (Fadel et al., 2020; Rehó et al., 2023). Here, we used SMT in live cells with endogenously tagged proteins and carefully controlled protein levels of two players, RARα and RXRα, in the T2NRs dimerization network, thereby directly addressing a fundamental question - whether RAR (partner) or RXR (core) is limiting for chromatin association.
In stark contrast to the generally accepted T2NR competition model, our results reveal that in U2OS cells, formation and chromatin binding of RAR-RXR heterodimers are limited by the concentration of RAR and not the core subunit RXR (Figure 4). PAPA-SMT directly confirmed in vivo that the association with RXRα promotes chromatin binding of RARα and vice versa (Figure 3C). However, unexpectedly, overexpression of RARα increases fbound of endogenous RXRα, while the reverse is not true (Figure 2F), indicating that RXR is not limiting but is rather constrained by the availability of its binding partners. Note that even though the number of RXRα and RARα molecules is about the same and the heterodimer complex has a predicted 1:1 stoichiometry (Rastinejad, 2022; Rastinejad et al., 2000), the fbound of endogenous RARα (50%) and RXRα (38%) differed by about 10% (Figures 1D and 2B). This surprising result could be explained by expression of additional RXR isoforms (most likely RXRβ) in U2OS cells (Figure 1A, Appendix 2—table 1, Figure 1—figure supplement 2B). The fact that the fbound of overexpressed RARα decreases only twofold, even though it is in fourfold excess over endogenous RXRα, likewise suggests that there are other heterodimerization partners of RARα available (Figure 2A and B, see Appendix 2—table 1). Moreover, the fbound of overexpressed RXR is essentially the same as the NLS control, indicating this excess core partner remains nearly totally unbound (Figure 2C). This is consistent with endogenous RXR already being in excess such that any additional RXR would remain mostly monomeric and unbound to chromatin. Hence, despite not directly measuring the protein levels of all RXR isoforms, we can still deduce in U2OS cells that RXR is in excess relative to its binding partners.
Control of RAR-RXR heterodimer concentration by RAR abundance makes sense considering the observation that RAR protein levels appear to be regulated by multiple feedback mechanisms: First, we find that overexpression of RARα significantly lowers the expression of endogenous RARα (Figure 2E) implying either that RAR participates in negative autoregulation at the transcriptional level or, that an excess of RAR causes instability at the protein level. We favor the latter possibility because overexpression of RARα reduces not only the expression of endogenous RARα but also its fbound (Figure 2F), indicating a reduction in dimerization and chromatin binding. It is possible that endogenous RAR may be more readily degraded when not bound by RXR, analogous to what was recently shown for the c-MYC/MAX heterodimers (Mark et al., 2023). Second, as has been previously reported, RARα expression decreases upon ligand treatment (Ismail and Nawaz, 2005; Kopf et al., 2000; Osburn et al., 2001; Tsai et al., 2023; Zhu et al., 1999). Curiously, a 50% reduction in RARα upon addition of ligand did not affect the fbound of either RARα or RXRα (Figure 1—figure supplement 3A and B), suggesting that most chromatin binding of endogenous RXRα in U2OS cells depends on heterodimerization partners other than RARα (see Appendix 2—table 1).
In contrast to the generally accepted model, we thus envision a network of T2NR heterodimers not always driven by competition for the core TF partner (RXR). Instead, an excess of RXR may ensure independent regulation of the different T2NRs without disrupting crosstalk between them (Figure 4).
In this first study, we have not examined other T2NRs expressed in U2OS cells or whether there is a similar excess of RXR in other cell types. Therefore, we cannot rule out that competition for RXR between T2NRs occurs in some cases since different cell types express RXR and partner T2NRs in varying amounts. It seems likely that chromatin binding by any given T2NR will depend on the concentrations of all other T2NRs. It is thus important to determine how interactions within the dimerization network are perturbed upon up- or down-regulation of one or more T2NRs, especially since dysregulation of T2NR expression has been reported in several types of cancers as well as other diseases (Brabender et al., 2005; Collins-Racie et al., 2009; Frigo et al., 2021; Long and Campbell, 2015). Here, we have shown that SMT and PAPA-SMT provide one empirical way to determine which set of components in a dimerization network is stoichiometrically limiting in a given cell type, without having to measure the concentration of every T2NR or the affinity of every interaction. The basic framework we have established here could also be extended to probe the effect of ligands or small molecules on the T2NR interaction network or other dimerization networks seen in bHLH or leucine zipper family of TFs, providing useful information about critical regulatory network interactions often implicated in diseases.
Methods
Cell culture and stable cell line generation
U2OS cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) with 4.5 g/l glucose supplemented with 10% fetal bovine serum (FBS) (HyClone, Logan UT, Cat. #AE28209315), 1 mM sodium pyruvate (Thermo Fisher 11360070), L-glutamine (Sigma #G3126-100G), GlutaMAX (Thermo Fisher #35050061), and 100 U/ml penicillin-streptomycin (Thermo Fisher #15140122) at 37°C and 5% CO2. Cells were subcultured at a ratio of 1:4 to 1:10 every 2–4 days for no longer than 30 passages. Regular mycoplasma testing was performed using polymerase chain reaction (PCR). Phenol containing media was used for regular cell culture and phenol red-free DMEM (Thermo Fisher #21063029) supplemented with 10% FBS and 100 U/ml penicillin-streptomycin was used for imaging.
Stable cell lines expressing the exogenous gene products (Appendix 2—table 2) were generated by PiggyBac transposition and antibiotic selection. Gibson assembly was used to clone genes of interest into a PiggyBac vector containing a puromycin or neomycin resistant gene. Plasmids were purified by Zymo midiprep kit (Zymo D4200) and all cloning was confirmed by Sanger sequencing. Cells were transfected by nucleofection using the Lonza Cell line Nucleofector Kit V (Lonza, Basel, Switzerland, #VVCA1003) and the Amaxa Nucleofector II device. For each transfection cells were plated 1–2 days before nucleofection in a six-well plate until they reach 70–90% confluency. Cells were trypsinized, resuspended in DMEM media, and centrifuged at 200×g for 2 min before the media was aspirated. Cells were then resuspended in 100 µl Lonza transfection reagent (82 μl Kit V solution+16 μl of Supplement #VVCA1003) containing 0.4 μg of SuperPiggyBac transposon vector and 0.8 μg of donor PiggyBac plasmid and transferred to an electroporation cuvette. Cells were electroporated using program X-001 on the Amaxa Nucleofector II (Lonza). Transfected cells were cultured in DMEM growth media without any antibiotics for 24–48 hr and then selected for 10 days with 1 μg/ml puromycin (Thermo Fisher #A1113803) or 1 mg/ml neomycin (G418 Sulfate, Thermo Fisher #10131027). After selection polyclonal cell lines were maintained in the selection media containing required antibiotics.
For ligand treatment, 100 μM atRA stock was prepared by dissolving atRA powder (CAS No: 302794, Sigma-Aldrich #R2625) in dimethyl sulfoxide (DMSO) (Sigma-Aldrich #D2650) and was diluted 1:100,000 or 1:1000 in growth media to final concentration of 1 nM or 100 nM respectively. The same volume of DMSO used for 100 nM atRA treatment (0.1%) is used for the control group as a condition without atRA treatment. Cells were treated 24 hr in either atRA or DMSO alone before imaging.
Genome editing cell lines
KI cell lines were generated as previously described (Hansen et al., 2017) with some modifications. Halo-tagging of endogenous RARα was described in our previously published work (Heckert et al., 2022b) which we have further validated for the current study. For Halo-tagging we designed sgRNAs using CRISPOR web tool (Concordet and Haeussler, 2018). Since exon 1 of RXRα is very short (only 9 aa) we chose to gene edit at the start of exon 2 for a successful tagging. sgRNAs were cloned into the Cas9 plasmid (a gift from Frank Xie) under the U6 promoter (Zhang Lab) with an mVenus reporter gene under the PGK promoter. Repair vectors were cloned in a basic pUC57 backbone for N-terminal tagging and pBluescript II SK (+) (pBSKII+) backbone for C-terminal tagging, with 500 bp left and right homology arms on either side of the Halo-tag sequence. Two guide/repair for N-terminal and three guide/repair vector pairs for C-terminal were attempted; only-N-terminal clones were ultimately recovered. Each sgRNA/donor pair were transfected to approximately 1 million early passage U2OS cells, at 1:3 ratio of sgRNA/donor (total 5 μg DNA) and plated in a six-well plate. 48 hr after transfection, Venus-positive cells were FACS sorted and cultured for another 7–10 days. Then, Halo-positive cells (stained with TMR) were sorted individually into single wells of 96-well plates and cultured for another 12–14 days. Clones were expanded and genotyped using PCR. PCR was done using one primer upstream of the left homologous arm and the other primer downstream of the right homologous arm. Another PCR with either external primer paired with a corresponding internal primer located in the Halo-tag coding region was done for further validation. Homozygous clones with the correct genotype (V5-Halo-RXRα clones C10, D6, and D9) were confirmed by Sanger sequencing and western blotting.
Antibodies
The following antibodies were used for western blotting: mouse monoclonal anti-RARα [H1920] (Abcam, #ab41934) diluted at 1:400, rabbit monoclonal anti-RXRα [EPR7106] (Abcam, #ab125001) diluted at 1:500, rabbit polyclonal anti-RXRα (Proteintech, #212181AP), mouse monoclonal anti-V5 tag (Thermo Fisher, #R960-25) diluted at 1:5000, mouse monoclonal anti-FLAG [M2] (Sigma-Aldrich, #F1804) diluted at 1:5000, mouse monoclonal anti-Halo (Promega, #G9211) diluted at 1:500, mouse monoclonal anti-TBP (Abcam, #ab51841) diluted at 1:5000, rabbit polyclonal anti-Centrin 2 (Proteintech, # 158771AP) diluted at 1:1000, goat polyclonal anti-mouse IgG light chain specific (Jackson ImmunoResearch, # 115035174) diluted at 1:10,000, mouse monoclonal anti-rabbit IgG light chain specific (Jackson ImmunoResearch, #211032171) diluted at 1:10,000.
The following antibodies were used for co-IP: rabbit polyclonal anti-V5 (Abcam, # ab9116) for immunoprecipitation, mouse monoclonal anti-FLAG [M2] (Sigma-Aldrich, #F1804) diluted at 1:5000 and mouse monoclonal anti-V5 tag (Thermo Fisher, #R960-25) diluted at 1:5000 for blotting.
Western blotting
For western blots, cells growing in six-well or 10 cm plates at 80–90% confluency were scraped and pelleted in ice-cold phosphate-buffered saline (PBS) containing protease inhibitors. Cell pellets were resuspended using 500–1000 μl hypotonic buffer (100 mM NaCl, 25 mM HEPES, 1 mM MgCl2, 0.2 mM EDTA, 0.5% NP-40 alternative) containing protease inhibitors - 1× aprotinin (Sigma, #A6279, diluted 1:1000), 1 mM benzamidine (Sigma, #B6506), 0.25 mM PMSF (Sigma #11359061001), and 1× cOmplete EDTA-free Protease Inhibitor Cocktail (Sigma, #5056489001) along with 125 U/ml of benzonase (Novagen, EMD Millipore #71205-3). Resuspended cells were gently rocked at 4°C for at least 2 hr after which 5 M NaCl was added. Cells were left rocking for extra 30 min at 4°C and then centrifuged at max speed at 4°C. Supernatants were quantified using Bradford and 15–20 μg was loaded on 8% Bis-Tris SDS gel. Wet transfer to 0.45 μm nitrocellulose membrane (Thermo Fisher #45004031) was performed in a transfer buffer (15 mM Tris-HCl, 20 mM glycine, 20% methanol) for 60–80 min at 100 V, 4°C. Membranes were blocked in 10% non-fat milk in 0.1% TBS-Tween (TBS-T) for 1 hr at room temperature (RT) with agitation. Membranes were then blotted overnight with shaking at 4°C with primary antibodies diluted in TBS-T with 5% non-fat milk. After 5×5 min washes in 0.1% TBS-T, membranes were incubated at RT with HRP-conjugated light chain secondary antibodies diluted 1:10,000 in TBS-T with 5% non-fat milk, for 1 hr with agitation. After 5×5 min washes in 0.1% TBS-T, membranes were incubated for 2 min in freshly prepared Perkin Elmer LLC Western Lightning Plus-ECL, enhanced Chemiluminescence Substrate (Thermo Fisher, #509049326). Finally, membranes were imaged with Bio-Rad Chemidoc Imaging System (Bio-Rad, Model No: Universal Hood III).
Luciferase assays
pGL3-RARE luciferase, a reporter containing firefly luciferase driven by SV40 promoter with three RAREs, a gift from T Michael Underhill (Addgene plasmid 13456; addgene:13458; RRID:Addgene_13458; Hoffman et al., 2006) was used for luciferase assays. A pRL-SV40 vector (Promega #E2261) expressing Renilla luciferase with SV40 promoter was used as a control to normalize luciferase activity. Cells were plated on six-well plate at least 48 hr before being co-transfected with 200 ng pGL3-RARE-luciferase and 10 ng Renilla luciferase vector, using TransIT-2020 Transfection Reagent (Mirus Bio, #MIR5404). A day after transfection, cells were treated with 100 nM atRA or 0.1% DMSO (control). Luciferase assay was performed the next day, using Dual-Luciferase Reporter Assay System (Promega, #E1910) according to the manufacturer’s protocol, on the Glomax Luminometer (Promega). The relative luciferase activity was calculated by normalizing firefly luciferase activity to the Renilla luciferase activity to control for transfection efficiency.
Co-immunoprecipitation
For co-IP experiments, Cos7 cells were plated to 2×15 cm dishes for each condition, at 60–70% confluency. DNA Lipofectamine 3000 (Thermo Fisher #L3000015) was used to co-transfect the cells with plasmids expressing RARα-Halo-3×FLAG and RXRα-V5 or V5-Halo-RXRα and RARα-3×FLAG or RARαRR-Halo-3×FLAG and RXRα-V5. According to the manufacturer’s instructions, 500 μl Opti-MEM medium (Thermo Fisher #31985062) was combined with 20 μl P3000 reagent and the plasmids for each condition. To control for the ratio of transfected DNA to the reagents, total transfected DNA mass was kept at 10 μg for each condition using empty pBSK vector (Addgene # 212205). 500 μl Opti-MEM medium containing 20 μl Lipofectamine 3000 reagent was subsequently added to the plasmid containing mixture and incubated at RT for 15 min, after brief pipetting to mix the solutions. The mixture was then divided equally to the cells plated in 2×15 cm dishes. COS7 cells were cultured in DMEM media with 4.5 g/l glucose supplemented with 10% FBS, 1 mM sodium pyruvate, L-glutamine, GlutaMAX, and 100 U/ml penicillin-streptomycin and maintained at 37°C and 5% CO2.
Cells were collected from plates 48 hr after transfection by scraping in ice-cold 1× PBS with protease inhibitors (1× aprotinin, 0.25 mM PMSF, 1 mM benzamidine, and 1× cOmplete EDTA-free Protease Inhibitor Cocktail). Collected cells were pelleted, flash-frozen in liquid nitrogen, and stored at –80°C. On the day of co-IP experiments, cell pellets were thawed on ice, resuspended to 700 μl of cell lysis buffer (10 mM HEPES pH 7.9, 10 mM KCl, 3 mM MgCl2, 340 mM sucrose [1.16gr], and 10% glycerol) with freshly added 10% Triton X-100. Cells were rocked at 4°C for 8 min to allow lysis and centrifuged for 3 min at 3000×g. The cytoplasmic fraction was removed, and the nuclear pellets were resuspended in 1 ml hypotonic buffer (100 mM NaCl, 25 mM HEPES, 1 mM MgCl2, 0.2 mM EDTA, 0.5% NP-40 alternative) containing protease inhibitors and 1 μl benzonase. After rocking for 2–3 hr at 4°C, the salt concentration was adjusted to 0.2 M NaCl final and the lysates were rocked for another 30 min at 4°C. Supernatant was removed after centrifugation at maximum speed at 4°C for 20 min and quantified by Bradford. Typically, 1 mg of protein was diluted in 1 ml 0.2 mM co-IP buffer (200 mM NaCl, 25 mM HEPES, 1 mM MgCl2, 0.2 mM EDTA, 0.5% NP-40 alternative) with protease inhibitors and cleared for 2 hr at 4°C with magnetic Protein G Dynabeads (Thermo Fisher, #10009D) before overnight immunoprecipitation with 1 μg per 250 μg of proteins of either normal serum IgGs or specific antibodies as listed above. Some precleared lysates were kept overnight at 4°C as input. Magnetic Protein G Dynabeads were also precleared overnight with 0.5% BSA at 4°C. Next day the precleared Protein G dynabeads were added to the antibody containing samples and incubated at 4°C for 2 hr. Samples were briefly spun down and placed in a magnetic rack at 4°C for 5 min to remove the co-IP supernatant. After extensive washes with the co-IP buffer, the proteins were eluted from the beads by boiling for 10 min in 1× SDS-loading buffer and analyzed by SDS-PAGE and western blot.
Flow cytometry
Cells were grown in six-well dishes. On the day of the experiment, cells were labeled with 500 nM Halo-TMR for 30 min, followed by one quick wash with 1× PBS and 15 min wash in dye-free DMEM before trypsinization. Cells were pelleted after centrifugation, resuspended in fresh medium, filtered through 40 μm filtration unit, and placed on ice until fluorescence read out by Flow Cytometry (within 30 min). Using an LSR Fortessa (BD Biosciences) flow cytometer, live cells were gated using forward and side scattering and TMR fluorescence emission readout was filtered using 610/20 band pass filter after excitation with 561 nm laser. Mean fluorescence intensity of the samples and absolute abundance were calculated as described in Cattoglio et al., 2019, using a previously quantified Halo-CTCF cell line as a standard (Cattoglio et al., 2019).
Cell preparation and dye labeling for imaging
For confocal imaging ~50,000 cells were plated in tissue culture-treated 96-well microplate (Perkin Elmer, #6055300) a day before imaging. Cells were labeled with Halo and/SNAP for 1 hr and incubated in dye-free media for 20 min after a quick 1× PBS wash to remove any free dye. Phenol-free media containing Hoechst (1 μM) was added to the cells and incubated for 1 hr before proceeding with imaging.
For SMT, 25 mm circular No. 1.5H precision coverglass (Marienfield, Germany, 0117650) were sonicated in ethanol for 10 min, plasma cleaned then stored in 100% isopropanol until use. ~250,000 U2OS cells were plated on sonicated, and plasma cleaned coverglass placed in six-well plates. For ligand treatment, cells were treated with DMSO (control, 0.1%), 1 nM atRA or 100 nM atRA, a day after plating cells in the coverslip and a day before imaging. After ~24–48 hr, cells were incubated with 100 nM PA-JFX549 dye in regular culture media for 30 min, followed by 4×30 min incubations with dye-free culture media at 37°C. A quick 2× PBS wash was interspersed between each 30 min dye-free media incubation. After the final wash, coverslips with plated cells facing upward were transferred to Attofluor Cell Chambers (Thermo Fisher, # A7816), and phenol-free media added for imaging. For Halo-tagged RARα and RXRα homozygous clones that were stably integrated with SNAP-tagged RARα, RXRα, or control transgenes under EF1α promoter; cells were double-labeled with 100 nM HaloTag ligand (HTL) PA-JFX549 and SNAP tag ligand (STL) 50 nM SF-650 simultaneously. The 30 min labeling step was followed by the same wash steps to remove free dye as described above, before transferring the coverslips to Attofluor Cell Chambers and imaging.
For PAPA-SMT experiments, polyclonal U2OS cells stably expressing SNAP-tagged RARα or RXRα under EF1α promoter within the endogenous Halo-tagged RARα and RXRα homozygous clones were used. ~400,000 polyclonal cells were plated in glass-bottom dishes (MatTEK P35G-1.5–20-C) and stained overnight with 50 nM JFX549 HTL and 5 nM JFX650 STL in phenol red-free DMEM (Thermo Fisher #21063029). Next day, cells were briefly rinsed twice with 1× PBS, incubated twice for 30 min in phenol red-free DMEM to remove free dye, and exchanged into fresh phenol red-free medium more before imaging.
Confocal imaging
For confocal imaging, endogenously Halo-tagged cells were labeled with Halo ligand JFX549 (100 nM) and Hoechst (1.6 μM); endogenously Halo-tagged cells overexpressing SNAP-tagged proteins were labeled with Halo ligand JFX549 (100 nM), SNAP ligand SF650 (25 nM), and Hoechst (1.6 μM). Imaging was performed at the UC Berkeley High Throughput Screening facility on a Perkin Elmer Opera Phenix equipped with 37°C and 5% CO2, using a built-in ×40 water immersion objective.
Live-cell SMT
All SMT experiments were performed using a custom-built microscope as previously described (Hansen et al., 2017). Briefly, a Nikon TI microscope was equipped with a ×100/NA 1.49 oil immersion total internal reflection fluorescence (TIRF) objective (Nikon apochromat CFI Apo TIRF 100× Oil), a motorized mirror, a perfect Focus system, an EM-CCD camera (Andor iXon Ultra 897), laser launch with 405 nm (140 mW, OBIS, Coherent), 488 nm, 561 nm, and 639 nm (1W, Genesis Coherent) laser lines, an incubation chamber maintaining a humidified atmosphere with 5% CO2 at 37°C. All microscope, camera, and hardware components were controlled through the NIS-Elements software (Nikon).
For tracking endogenous as well as overexpressed Halo-tagged proteins, PA-JFX549 labeled cells were excited with 561 nm laser at 2.3 kW/cm2 with single band-pass emission filter (Semrock 593/40 nm for PA-JFX549). All imaging was performed using highly inclined optical sheet (HILO) illumination (Tokunaga et al., 2008). Low laser power (2–3%) was used to locate and focus cell nuclei. Region of interest (ROI) of random size but with maximum possible area was selected to fit into the interior of the nuclei. Before tracking, 100 frames of continuous illumination with 549 laser (laser power 5%) at 80 ms per frame was recorded to later calculate mean intensity of each nuclei. After partial pre-bleaching (only for overexpressed Halo-tagged proteins), movies were taken with 1 ms pulses of full power of 561 nm illumination at the beginning of frame interval with camera exposure of 7 ms/frame, while the 405 nm laser (67 W/cm2) was pulsed during ~447 μs camera transition time. Total of 30,000 frames were collected. 405 nm intensity was manually tuned to maintain low density (1–5 molecules fluorescent particles per frame).
For tracking experiments with endogenous Halo-tagged proteins in the presence of transgenic SNAP protein products, cells were dual-labeled with both PA-JFX549 and STL-SF650. Before tracking cells were excited with a 639 nm laser at 88 W/cm2 and single band-pass emission filter set to Semrock 676/37 nm. Low laser power (2–3%) was used to locate and focus cell nuclei. As before, ROI of random size but with maximum possible area was selected to fit into the interior of the nuclei. The emission filter was switched to Semrock 593/40 nm band-pass filter for PA-JFX549, keeping the TIRF angle, stage XYZ position, and ROI the same. Cells were then excited with 549 nm and single molecules were tracked for 30,000 frames as described above.
Note, we estimated that to calculate fbound of RARα and RXRα with minimal bias and variation, we need to analyze at least 10,000 pooled trajectories from n≥20 cells (Figure 1—figure supplement 1E). Therefore, all analyses of fSMT to calculate fbound or fbound % presented in this paper are done accordingly. At least 8–10 movies were collected for each sample as one technical replicate on a given day. Two technical replicates on 2 separate days were collected to produce the reported results. As observed previously for SMT experiments (McSwiggen et al., 2024), the variance within collected data was largely due to cell-to-cell variance (Figure 1—figure supplement 1E).
PAPA-SMT
PAPA-SMT imaging was performed using the same custom-built microscope described above using an automated system (Graham et al., 2024; Walther et al., 2024). Briefly, the microscope described above was programmed using Python and NIS Elements macro language code to raster in a grid over the coverslip surface, acquire an image in the JFX650 channel, identify cell nuclei, reposition the stage to center it on a target nucleus, resize the imaging ROI to fit the nucleus, pre-bleach JFX650 with red light (639 nm) for 5 s, and finally perform a PAPA-SMT illumination sequence. The illumination sequence consisted of 5 cycles of the following, at a frame rate of 7.48 ms/frame:
Bleaching: 200 frames of red light (639 nm), not recorded
Imaging: 30 frames of red light, one 2 ms stroboscopic pulse per frame, recorded
PAPA: 5 frames of green light (561 nm), not recorded
Imaging: 30 frames of red light, one 2 ms stroboscopic pulse per frame, recorded
Bleaching: 200 frames of red light, not recorded
Imaging: 30 frames of red light, one 2 ms stroboscopic pulse per frame, recorded
DR: 1 frames of violet light (405 nm), one 1 ms stroboscopic pulse, not recorded
Imaging: 30 frames of red light, one 2 ms stroboscopic pulse per frame, recorded
Laser power densities were approximately 67 W/cm2 for 405 nm, 190 W/cm2 for 561 nm, and 2.3 kW/cm2 for 639 nm. This illumination sequence includes non-stroboscopic illumination periods to bleach fluorophores or place them back in the dark state (steps 1 and 5), and it records only the frames immediately before and after the PAPA and DR pulses. This is done to decrease file sizes and speed up subsequent analysis compared to our previous protocol (Graham et al., 2022).
SMT and SMT-PAPA data processing and analysis
All SMT movies were processed using the open-source software package quot (Heckert and Fan, 2022). For SMT dataset quot package was run on each collected SMT movie using the following settings: [filter] start = 0; method = ‘identity’; chunk_size = 100; [detect] method = ‘llr’; k=1.0; w=15, t=18; [localize] method = ‘ls_int_gaussian’, window size = 9; sigma = 1.2; ridge = 0.0001; max_iter = 20; damp = 0.3; camera_gain = 109.0; camera_bg = 470.0; [track] method = ‘euclidean’; pixel_size_μm=0.160; frame interval = 0.00748; search radius = 1; max_blinks = 0; min_IO = 0; scale - 7.0. The first 500 or 1000 frames of each movie were removed due to high localization density. To confirm that all movies were sufficiently sparse to avoid misconnections, a maximum number of 5 localizations per frame was maintained (although most frames had 1 or less). To infer the distribution of diffusion coefficients from experimentally observed trajectories, we used the state array method, which is publicly available at https://github.com/alecheckert/saspt (Heckert et al., 2022a; Heckert et al., 2022b). The following settings were used: likelihood_type = ‘rbme’; pixel_size_μm=0.16; frame_interval = 0.00747; focal depth = 0.7; start_frame = 500 or 1000 frames. RBME likelihood for individual cells and occupations as the mean of the posterior distribution over state occupations marginalized on diffusion coefficient is reported. For SMT movies in Figure 2—figure supplement 1C, mean intensity of each cell was calculated using the mean gray value tool in FIJI. Mean gray value of each frame (total 100 frames) of the pre-bleaching 80 ms movie was measured. Average mean gray value over the 100 frames is reported as mean intensity for each cell.
For PAPA-SMT quot package was run on each collected SMT movie using the following settings: [filter] start = 0; method = ‘identity’; chunk_size = 100; [detect] method = ‘llr’; k=1.2; w=15, t=18; [localize] method = ‘ls_int_gaussian’, window size = 9; sigma = 1.2; ridge = 0.0001; max_iter = 10; damp = 0.3; camera_gain = 109.0; camera_bg = 470.0; [track] method = ‘conservative’; pixel_size_μm=0.160; frame interval = 0.00748; search radius = 1; max_blinks = 0; min_IO = 0; scale - 7.0. Trajectories were sorted based on whether they occurred before or after green (561 nm; PAPA) or violet (405 nm; DR) reactivation pulses, and state array analysis (Heckert et al., 2022b) was applied to each set of trajectories. For Figure 3C, reactivation by PAPA and DR for each cell were quantified by calculating the difference in total number of localizations before and after reactivation pulses.
Statistical analysis
To derive a measure of error for SMT and SMT-PAPA, we performed bootstrapping analysis on all SMT datasets. For each dataset, a random sample of size n, where n is the total number of cells in the dataset, was drawn 100 times. The mean and standard deviation (stdev) from this analysis is reported.
Two-tailed p-values were calculated based on a normal distribution (Scipy function, scipy.stats.norm.sf) with mean equal to the difference between sample means and variance equal to the sum of the variances from bootstrap sampling (for SMT and SMT-PAPA) or biological replicates (for absolute abundance calculation of molecules from flow cytometry assay). The Šidák correction was applied to correct for multiple hypothesis testing within each experiment.
Appendix 1
Brief overview of the steps involved in a PAPA-SMT experiment
A PAPA-SMT (Graham et al., 2022) experiment involves several steps (Figure 3A): (1) A SNAP-tagged target protein is labeled with a ‘receiver’ fluorophore such as JFX650, while a Halo-tagged partner protein is labeled with a ‘sender’ fluorophore such as JFX549. (2) The JFX650 receiver is placed in a dark state by illumination with intense red (639 nm) light. (3) A pulse of green (561 nm) light is used to excite the JFX549 sender, reactivating nearby dark-state JFX650 molecules. Because association of the two proteins brings the sender and receiver together, this step selectively reactivates complexes of the two proteins. (4) After switching off the green light, red light is used to image reactivated receiver molecules. (5) As an internal control, a violet (405 nm) light pulse is used to reactivate the receiver by ‘DR’, independent of proximity with the sender (Heilemann et al., 2008). (6) After switching off the violet light, reactivated molecules are again imaged with red light. Steps 3–6 are repeated for multiple rounds, and SMT trajectories occurring after green and violet pulses are analyzed separately using SASPT. Trajectories occurring after a green pulse (PAPA trajectories) are enriched for complexes between the SNAP- and Halo-tagged proteins, while trajectories occurring after a violet pulse (DR trajectories) approximate a random sample from the whole population.
Appendix 2
Data availability
Original western blots included in this study are available in accompanying source data zip files. SMT raw data for Figure 1-3 are accessible through https://doi.org/10.5281/zenodo.14009060.
-
ZenodoSurprising Features of Nuclear Receptor Interaction Networks Revealed by Live Cell Single Molecule Imaging [Data set].https://doi.org/10.5281/zenodo.14009060
References
-
The Max network gone madMolecular and Cellular Biology 21:691–702.https://doi.org/10.1128/MCB.21.3.691-702.2001
-
CRISPOR: intuitive guide selection for CRISPR/Cas9 genome editing experiments and screensNucleic Acids Research 46:W242–W245.https://doi.org/10.1093/nar/gky354
-
Single-molecule tracking (SMT): a window into live-cell transcription biochemistryBiochemical Society Transactions 51:557–569.https://doi.org/10.1042/BST20221242
-
Agonist binding directs dynamic competition among nuclear receptors for heterodimerization with retinoid X receptorThe Journal of Biological Chemistry 295:10045–10061.https://doi.org/10.1074/jbc.RA119.011614
-
Nuclear receptors: from molecular mechanisms to therapeuticsEssays in Biochemistry 65:e00222.https://doi.org/10.1210/er.2018-00222
-
Visualization of Myc/Max/Mad family dimers and the competition for dimerization in living cellsMolecular and Cellular Biology 24:4294–4308.https://doi.org/10.1128/MCB.24.10.4294-4308.2004
-
BMP action in skeletogenesis involves attenuation of retinoid signalingThe Journal of Cell Biology 174:101–113.https://doi.org/10.1083/jcb.200604150
-
The MAX-interacting transcription factor networkSeminars in Cancer Biology 16:265–274.https://doi.org/10.1016/j.semcancer.2006.07.009
-
Cross-talk between peroxisome proliferator-activated receptor (PPAR) alpha and liver X receptor (LXR) in nutritional regulation of fatty acid metabolism. II. LXRs suppress lipid degradation gene promoters through inhibition of PPAR signalingMolecular Endocrinology 17:1255–1267.https://doi.org/10.1210/me.2002-0191
-
Pan-cancer analyses of the nuclear receptor superfamilyNuclear Receptor Research 2:101182.https://doi.org/10.11131/2015/101182
-
Halotag: a novel protein labeling technology for cell imaging and protein analysisACS Chemical Biology 3:373–382.https://doi.org/10.1021/cb800025k
-
Ligand-activated PPARbeta efficiently represses the induction of LXR-dependent promoter activity through competition with RXRMolecular and Cellular Endocrinology 256:23–33.https://doi.org/10.1016/j.mce.2006.05.005
-
Ligand-dependent degradation of retinoid X receptors does not require transcriptional activity or coactivator interactionsMolecular and Cellular Biology 21:4909–4918.https://doi.org/10.1128/MCB.21.15.4909-4918.2001
-
How do transcription factors select specific binding sites in the genome?Nature Structural & Molecular Biology 16:1118–1120.https://doi.org/10.1038/nsmb1109-1118
-
Expression Atlas update: from tissues to single cellsNucleic Acids Research 48:D77–D83.https://doi.org/10.1093/nar/gkz947
-
Integrated view of baseline protein expression in human tissuesJournal of Proteome Research 22:729–742.https://doi.org/10.1021/acs.jproteome.2c00406
-
Retinoic acid receptor structures: the journey from single domains to full-length complexJournal of Molecular Endocrinology 69:T25–T36.https://doi.org/10.1530/JME-22-0113
-
Agonist-controlled competition of RAR and VDR nuclear receptors for heterodimerization with RXR is manifested in their DNA bindingThe Journal of Biological Chemistry 299:102896.https://doi.org/10.1016/j.jbc.2023.102896
-
Combinatorial control of gene expressionNature Structural & Molecular Biology 11:812–815.https://doi.org/10.1038/nsmb820
-
A subcellular map of the human proteomeScience 356:eaal3321.https://doi.org/10.1126/science.aal3321
-
Mnt-max to myc-max complex switching regulates cell cycle entryThe Journal of Cell Biology 169:405–413.https://doi.org/10.1083/jcb.200411013
-
Vitamin D and adipogenesis: new molecular insightsNutrition Reviews 66:40–46.https://doi.org/10.1111/j.1753-4887.2007.00004.x
-
Cross-talk between peroxisome proliferator-activated receptor (PPAR) α and liver X receptor (LXR) in nutritional regulation of fatty acid metabolism. I. PPARS suppress sterol regulatory element binding protein-1c promoter through inhibition of LXR signaliMolecular Endocrinology 17:1255–1267.https://doi.org/10.1210/me.2002-0190
Article and author information
Author details
Funding
National Institutes of Health (U54-CA231641-01659)
- Xavier Darzacq
Howard Hughes Medical Institute
- Robert Tjian
Jane Coffin Childs Memorial Fund for Medical Research
- Thomas GW Graham
National Institutes of Health (T32 GM098218)
- Alec Heckert
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
We would like to thank all members of the Tjian-Darzacq lab for helpful discussions and suggestions over the years, in particular Claudio Cattaglio for advice on biochemistry experiments and Vinson Fan for suggestions on SMT data analysis. We are grateful to John J Ferrie for aiding with Rosetta modeling as well as both him and Jonathan P Karr for critical reading of the manuscript. We also thank Luke Lavis for providing fluorescent HaloTag ligands, CRL Flow Cytometry Facility for use of their instruments, the QB3 High Throughput Screening Facility for providing access to the Opera Phenix automated confocal microscope. This work was supported by the NIH grant U54-CA231641-01659 (to XD) and the Howard Hughes Medical Institute (to RT). TG was supported by a postdoctoral fellowship from the Jane Coffin Childs for Medical Research. AH was supported by the NIH Stem Cell Biological Engineering predoctoral fellowship T32 GM098218.
Version history
- Preprint posted:
- Sent for peer review:
- Reviewed Preprint version 1:
- Reviewed Preprint version 2:
- Version of Record published:
Cite all versions
You can cite all versions using the DOI https://doi.org/10.7554/eLife.92979. This DOI represents all versions, and will always resolve to the latest one.
Copyright
© 2023, Dahal et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 526
- views
-
- 14
- downloads
-
- 2
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Cell Biology
- Chromosomes and Gene Expression
During oncogene-induced senescence there are striking changes in the organisation of heterochromatin in the nucleus. This is accompanied by activation of a pro-inflammatory gene expression programme – the senescence-associated secretory phenotype (SASP) – driven by transcription factors such as NF-κB. The relationship between heterochromatin re-organisation and the SASP has been unclear. Here, we show that TPR, a protein of the nuclear pore complex basket required for heterochromatin re-organisation during senescence, is also required for the very early activation of NF-κB signalling during the stress-response phase of oncogene-induced senescence. This is prior to activation of the SASP and occurs without affecting NF-κB nuclear import. We show that TPR is required for the activation of innate immune signalling at these early stages of senescence and we link this to the formation of heterochromatin-enriched cytoplasmic chromatin fragments thought to bleb off from the nuclear periphery. We show that HMGA1 is also required for cytoplasmic chromatin fragment formation. Together these data suggest that re-organisation of heterochromatin is involved in altered structural integrity of the nuclear periphery during senescence, and that this can lead to activation of cytoplasmic nucleic acid sensing, NF-κB signalling, and activation of the SASP.
-
- Chromosomes and Gene Expression
- Evolutionary Biology
Gene regulation is essential for life and controlled by regulatory DNA. Mutations can modify the activity of regulatory DNA, and also create new regulatory DNA, a process called regulatory emergence. Non-regulatory and regulatory DNA contain motifs to which transcription factors may bind. In prokaryotes, gene expression requires a stretch of DNA called a promoter, which contains two motifs called –10 and –35 boxes. However, these motifs may occur in both promoters and non-promoter DNA in multiple copies. They have been implicated in some studies to improve promoter activity, and in others to repress it. Here, we ask whether the presence of such motifs in different genetic sequences influences promoter evolution and emergence. To understand whether and how promoter motifs influence promoter emergence and evolution, we start from 50 ‘promoter islands’, DNA sequences enriched with –10 and –35 boxes. We mutagenize these starting ‘parent’ sequences, and measure gene expression driven by 240,000 of the resulting mutants. We find that the probability that mutations create an active promoter varies more than 200-fold, and is not correlated with the number of promoter motifs. For parent sequences without promoter activity, mutations created over 1500 new –10 and –35 boxes at unique positions in the library, but only ~0.3% of these resulted in de-novo promoter activity. Only ~13% of all –10 and –35 boxes contribute to de-novo promoter activity. For parent sequences with promoter activity, mutations created new –10 and –35 boxes in 11 specific positions that partially overlap with preexisting ones to modulate expression. We also find that –10 and –35 boxes do not repress promoter activity. Overall, our work demonstrates how promoter motifs influence promoter emergence and evolution. It has implications for predicting and understanding regulatory evolution, de novo genes, and phenotypic evolution.