Med1 nuclear foci are sensitive to 1, 6-Hexanediol and are dissolved during mitosis
Immunofluorescence staining revealed that Med1 formed numerous, well-distinguishable nuclear foci in U2OS cells (Fig. 1a), consistent with results with imaging GFP-Med1 in CRISPR knock-in mouse embryonic stem (ES) cells [13]. The specificity of this antibody against endogenous Med1 proteins was confirmed by western blot (Additional file 1: Fig. S1a). Next, we performed Med1 immunostaining in cells treated with 1, 6-Hexanediol, an aliphatic alcohol that has been frequently used to study biomolecular condensates in cells. Distinct Med1 foci became invisible in most cells after 1 min of Hexanediol treatment (Additional file 1: Fig. S1b). Med1 foci reappeared in most cells at 10 min and 30 min after Hexanediol withdrawal (Additional file 1: Fig. S1b). Because Med1 foci could be distinguished as individual fluorescence spots in a 3D image stack (Fig. 1a), we used the AirLocalize program [22] to obtain the number of Med1 foci in cells. The median number of Med1 foci per nucleus decreased from ~ 150 in untreated cells to less than 50 in Hexanediol-treated cells and increased to 200–300 in cells recovered for 10 min and 30 min (Additional file 1: Fig. S1c). We note that fluorescence intensities of Med1 foci were generally higher in recovered cells than in untreated cells (Additional file 1: Fig. S1b), which might explain the higher number of Med1 foci in recovered cells because the same intensity threshold was used for quantification.
Many well-known membraneless organelles, such as nuclear speckles, nucleoli, and Cajal bodies, are dissolved during mitosis [23,24,25]. Nonetheless, the status of Med1 nuclear foci in mitotic cells has not been described. By immunofluorescence staining, we found a more homogeneous localization of Med1 in mitotic cells than in interphase cells (Fig. 1b). The median number of Med1 foci decreased from ~ 150 in interphase cells to less than 50 in mitotic cells (Fig. 1c). Thus, Med1 nuclear foci resemble other membraneless organelles in their ability to dissolve during mitosis.
Characterization of nuclear condensates formed by Med15
The Mediator complex consists of over 30 protein subunits, many of which contain IDR sequences [26] that might contribute to phase separation. Thus, it would be interesting to study whether additional Mediator subunits might participate in the formation of nuclear condensates. In this study, we focused on a single subunit, Med15, which contains a large IDR including multiple Glutamine (Q) residues (Additional file 1: Fig. S2a). First, we found that GFP-tagged human Med15 or RFP-tagged mouse Med15 formed multiple nuclear foci in U2OS cells, respectively (Fig. 2a, Additional file 1: Fig. S2b). Consistently, immunofluorescence staining using a Med15 antibody revealed numerous nuclear foci in U2OS cells (Fig. 2b), and Med15 foci detected by immunofluorescence were colocalized with TagRFP-mMed15 (Additional file 1: Fig. S2b). We performed western blot to confirm the specificity of this antibody in detecting endogenous Med15 proteins in U2OS cells (Additional file 1: Fig. S2c). To further characterize the properties of Med15 nuclear condensates, we generated a T24 stable cell line expressing GFP-hMed15. GFP-tagged human Med15 formed multiple nuclear foci (Additional file 1: Fig. S2d) and were colocalized with Med15 foci detected by immunofluorescence (Additional file 1: Fig. S2e). Furthermore, we found that all prominent GFP-hMed15 foci were colocalized with nuclear foci formed by endogenous Med1 in this stable cell line (Fig. 2c). Therefore, we concluded that both endogenous and overexpressed Med15 formed nuclear condensates in human cells.
We attempted to determine the state of Med15 foci in mitotic cells but did not obtain conclusive results. Med15 immunofluorescence staining revealed multiple foci in mitotic U2OS cells (Additional file 1: Fig. S3a) and Med15 foci numbers per cell were higher in mitotic cells than in interphase cells (Additional file 1: Fig. S3b). In the stable T24 cell line expressing GFP-Med15, however, prominent Med15 foci observed in interphase cells were absent in most mitotic cells (Additional file 1: Fig. S3c, d). We suggest that prominent GFP-Med15 foci in the stable cell line that were colocalized with anti-Med1 (Fig. 2c) may be more consistent markers of Mediator condensates reported in previous studies [13, 14].
Next, we examined the state of Mediator condensates in U2OS cells where Med15 was depleted by RNAi. Both Med15 and Med1 protein levels appeared to be reduced in Med15 knockdown cells (Additional file 1: Fig. S4c), and Med15 mRNA level was substantially lower (Additional file 1: Fig. S4d). Anti-Med15 staining was reduced to background levels in Med15 knockdown cells (Additional file 1: Fig. S4a), confirming the specificity of this antibody in immunostaining experiments. Notably, anti-Med1 staining intensity was diminished and the numbers of Med1 nuclear foci were significantly decreased in Med15 knockdown cells (Additional file 1: Fig. S4a, b). Our results thus suggested Med15 was important for both maintaining Med1 protein level and forming Med1 nuclear foci. Additionally, GFP-hMed15 expressed in Med15 knockdown cells formed nuclear foci similarly as in control cells (Additional file 1: Fig. S4e).
Furthermore, we examined the response of GFP-Med15 nuclear foci to Hexanediol treatment by live cell imaging. Because high concentrations of Hexanediol likely introduce non-specific effects to cells [27], we tested Hexanediol concentrations lower than previously used to examine nuclear condensates in the GFP-Med15 stable cell line. Interestingly, application of 0.5% 1, 6-Hexanediol resulted in rapid and substantial decrease of fluorescence intensities of GFP-Med15 nuclear foci (Fig. 2d, Additional file 1: Fig. S5, Additional file 2: Video S1), and withdrawing Hexanediol from the growth media resulted in the reassembly of GFP-Med15 foci that plateaued in about 15 min (Fig. 2d, Additional file 1: Fig. S5, Additional file 2: Video S1). Therefore, rapid disruption/reassembly upon 1,6-Hexanediol treatment/withdrawal is a property shared between Med15 foci and Med1 foci. Our results indicated that low concentrations of Hexanediol (i.e., 0.5%) could dissolve nuclear condensates that have small sizes (such as GFP-Med15 foci).
Dynamics of Med15 foci in living cells
A characteristic feature of liquid-like nuclear condensates is the rapid exchange of their molecular components with the nucleoplasm [13, 14]. We next examined the association of Med15 with nuclear foci in living cells by fluorescence recovery after photobleaching (FRAP). In NIH3T3 cells expressing AcGFP-Med15, we observed that fluorescence intensities of nuclear foci recovered to approximately initial levels within 10 s after initial photobleaching (Fig. 3a, b). Similar results were obtained from NIH3T3 cells expressing TagRFP-Med15 (Additional file 1: Fig. S6). Furthermore, we observed fusion events and fission events of GFP-Med15 foci on the timescale of several minutes (Fig. 3c, Additional file 1: Fig. S7, Additional file 3: Video S2, Additional file 4: Video S3). Therefore, our results indicated that Med15 exchanged between nuclear condensates and the nucleoplasm at a rate comparable with that measured on Med1 [13], and suggested that Med15 molecules within these nuclear condensates were in a liquid-like phase.
DYRK3 overexpression disrupts Med1 nuclear foci and Med15 nuclear foci.
Recent studies revealed that the dual-specificity tyrosine kinase DYRK3 played a key role in dissolving multiple, but not all membraneless organelles during mitosis [9]. Moreover, overexpressing DYRK3 disrupted several nuclear organelles (such as nuclear speckle and Cajal body) in interphase cells [9]. We hypothesized that DYRK3 might play a role in the dissolution of Med1 foci during mitosis. As expected, Med1 foci were mostly dissolved in cells synchronized at mitotic stage by thymidine-nocodazole block (Additional file 1: Fig. S8a). Notably, Med1 foci reappeared in a portion of mitotic cells upon treatment with GSK626616, a small molecule inhibitor of DYRK3 (Additional file 1: Fig. S8b-d), suggesting that DYRK3 kinase activity plays a role in dissolving Med1 foci in mitotic cells.
Next, we examined the effects of DYRK3 overexpression on Med1 and Med15 nuclear foci in interphase cells. We expressed mCherry or mCherry-NLS*-DYRK3 (NLS*: SV40 nuclear localization signal) in U2OS cells and performed immunofluorescence staining against Med1. Most cells overexpressing DYRK3 showed diffuse Med1 localization in the nucleoplasm, in which the numbers of Med1 nuclear foci were substantially decreased (Fig. 4a, b, e). The same results were obtained in NIH3T3 cells (Additional file 1: Fig. S9). In the T24 cell line stably expressing GFP-Med15, most cells transfected with TagRFP-NLS*-DYRK3 lost prominent GFP-Med15 foci that were observed in untransfected interphase cells (Fig. 4c, d, f). Interestingly, dissolution of GFP-Med15 foci was affected by relative expression levels of TagRFP-NLS*-DYRK3. The mean intensity of TagRFP-NLS*-DYRK3 in lentivirus-infected cells was ~ 7 fold lower than that in transfected cells (Additional file 1: Fig. S10b), and GFP-Med15 foci were still present in most lentivirus-infected cells (Additional file 1: Fig. S10a, c). Likewise, transfected cells containing GFP-Med15 foci had significantly lower mean intensity of TagRFP-NLS*-DYRK3 compared to those without visible GFP-Med15 foci (Additional file 1: Fig. S10d). Taken together, our work revealed that Med1 foci and GFP-Med15 foci can be dissolved by overexpressing DYRK3 kinase, which provides a likely explanation for the dissolution of Med1 foci and GFP-Med15 foci in mitotic cells.
Because the Serine-rich IDR region of Med1 was shown to mediate its phase separation in vitro [13], we tested whether overexpressing DYRK3 could affect nuclear condensates formed by Med1 IDR in cells. Interestingly, when GFP-tagged Med1 IDR region (amino acid residues 948-1568) was expressed in NIH3T3 cells, it was enriched in the nucleolar regions and colocalized with Nucleophosmin (NPM1), an abundant nucleolar protein (Additional file 1: Fig. S11a). Notably, expressing TagRFP-NLS*-DYRK3 resulted in the redistribution of GFP-Med1 (948-1568) to the nucleoplasm (Additional file 1: Fig. S11b).
The Q-rich IDR and a hydrophobic amino acid region of Med15 are both required to form nuclear condensates
Next, we sought to identify the amino acid regions responsible for the formation of Med15 nuclear condensates. We generated a series of mouse Med15 truncation mutants fused to TagRFP at its C-terminus and compared their abilities to form nuclear foci (Fig. 5a). Med15 contains a KIX domain at its N-terminus, followed by a long glutamine-rich IDR (71-617) and a structured C-terminal domain that also contains its NLS (661-670). Surprisingly, Med15 (100-600) and Med15 (1-617) fragment fused to TagRFP and SV40 NLS (NLS*) were diffusely localized in the nucleus (Fig. 5b, c). Med15 (100-600) and Med15 (1-617) fused to TagRFP only were localized in the cytoplasm and formed several large aggregates (Fig. 5b) distinct from numerous small nuclear foci formed by full-length Med15 (Fig. 2a, Fig. 3c). These results suggested that the glutamine-rich IDR of Med15 was not sufficient to form condensates in the nucleus. These observations were also consistent with previous findings on several prion-like RNA-binding proteins that formed condensates in the cytoplasm while remained soluble in the nucleus [28].
Interestingly, both Med15 (1-670) and Med15 (1-680) formed multiple small nuclear foci (Fig. 5b, c) resembling those formed by full-length Med15. Because the 661-670 amino acid region is the native NLS of Med15, we examined TagRFP-NLS*-Med15 (1-660) and found that it also formed multiple nuclear foci (Fig. 5b, c). Importantly, both TagRFP-NLS*-Med15(1-660) and TagRFP-Med15(1-680) were colocalized with GFP-Med15 foci (Additional file 1: Fig. S12a). Thus, Med15 (618-660) region likely plays a role in condensate formation. Meanwhile, a C-terminal truncation of Med15 (amino acid 618-789) failed to form nuclear foci (Fig. 5b), suggesting that the N-terminal region (1-617) containing the Q-rich IDR also contributed to nuclear condensate assembly. Furthermore, we generated human Med15 truncation mutants according to the alignment between human and mouse Med15 protein sequences and observed a strong effect of the 616-659 amino acid region in condensate assembly in both wild-type cells (Additional file 1: Fig. S13a, b) and in Med15 knockdown cells (Additional file 1: Fig. S13c, d). Therefore, the mechanisms underlying nuclear condensate formation are likely conserved between mouse and human Med15 proteins.
We next sought to identify the motifs within this region that contribute to the formation of Med15 nuclear condensates. We noticed that mouse Med15 (637-660) region contained eight hydrophobic amino acid residues (Fig. 5d), raising the possibility that hydrophobic interactions may in part mediate the formation of Med15 nuclear condensates. Seven out of the eight hydrophobic amino acid residues are conserved in human Med15. To test this hypothesis, we mutated all eight hydrophobic amino acids in AcGFP-mMed15 to their hydrophilic mimics and found that the mutated protein formed visibly fewer nuclear foci than wild-type Med15 and that a lower fraction of cells showed Med15 foci (Fig. 5d, e). Taken together, although either the glutamine-rich IDR or the hydrophobic amino acid region (637-660) of Med15 was insufficient to form nuclear condensates, synergistic functions from both regions likely resulted in condensate formation.
Both IDR and C-terminal domain of Med15 contribute to phase separation in optodroplet assays
Determining the capacity of Med15 IDR or Med15 C-terminal region (618-789) in promoting phase separation in vivo would benefit from a cellular assay that can visualize condensate formation in real time. We applied the optodroplet assay to analyze how Med15 IDR, Med15 C-terminal domain, or Med1 IDR contribute to phase separation in cells. In this assay, protein domains of interest were fused to a fluorescent protein and the coding sequence of cryptochrome2 (Cry2), a blue light-sensitive protein from Arabidopsis thaliana, and the formation of optodroplets after blue light stimulation was visualized in real time [29]. First, we transiently expressed mCherry-Cry2 in NIH3T3 cells and did not observe optodroplet formation after illumination with blue light for 90 s (Fig. 6a). In contrast, mCherry-Cry2 fused to a Serine-rich IDR region of Med1 (amino acid 948-1157) formed optodroplets within 30 s of blue light stimulation (Fig. 6b), consistent with a previous study [13]. Next, we generated constructs of mCherry-Cry2 fused to NLS*-Med15 IDR (amino acid 71-617) or Med15 C-terminal region (amino acid 618-789) and examined optodroplet formation in living cells. Optodroplets formed by Med15 IDR appeared in spherical shape but were smaller in size than those formed by Med1 IDR after the same duration of blue light stimulation (Fig. 6b, c). Remarkably, Med15 C-terminal region formed optodroplets within 5 s after blue light stimulation (Fig. 6d), considerably faster than Med1 or Med15 IDR. The apparently lower efficiency of Med1 IDR in optodroplet formation (Fig. 6e) could arise from the shorter length of Med1 IDR or from our experimental conditions. Therefore, the optodroplet assay confirmed that both Med15 IDR and Med15 C-terminal region contributed to phase separation in cells.
FRAP revealed that optodroplets formed by Med1 IDR or Med15 IDR rapidly recovered with t1/2 < 10 s, and about 80% recovery was reached at 60 s after photobleaching (Additional file 1: Fig. S14a, b, d), consistent with measurement on optodroplets formed by Med1 IDR in a previous study [13]. In contrast, only about 20% FRAP recovery was observed on optodroplets formed by Med15 C-terminal region at 60 s after photobleaching (Additional file 1: Fig. S14c, d). Thus, Med15 C-terminal domain appeared to drive phase separation more efficiently than Med15 IDR or Med1 IDR in the optodroplet assay and might provide a strong adhesive force for maintaining the Mediator condensates.
Testing the effects of Hexanediol treatment in transcriptional activation of immediate early genes (IEGs) during the serum response.
Although recent studies have revealed phase separation phenomena of multiple key components of transcriptional machineries [6, 11, 15], roles of these nuclear condensates in transcriptional regulation were less well understood. We explored the roles of nuclear condensates during rapid gene activation by examining the effects of Hexanediol treatment and withdrawal on IEG expression during the serum response. IEGs respond very rapidly to a variety of cell-extrinsic and cell-intrinsic signals, including serum, growth factors, cytokines, and UV radiation [30, 31]. Given that Hexanediol treatment at high concentrations leads to inhibition of kinase and phosphatase activities [27], we compared the effects of 0.5% and 10% Hexanediol on IEG activation. We examined a few well-characterized IEGs (c-Fos, c-Jun, and Egr-1) in this study. NIH3T3 cells were analyzed in two groups. In Group I, cells were serum starved for 24 h and treated with media containing 20% serum. In Group II, cells were serum starved for 24 h, treated with 0.5% or 10% Hexanediol diluted in serum starvation media for 1 min and then stimulated with media containing 20% serum but no Hexanediol. IEG expression at distinct time points was analyzed by RT-qPCR (Fig. 7a, Additional file 1: Fig. S17a). By immunostaining, we found that Med1 and Med15 nuclear foci were both abolished after 1 min treatment with 10% Hexanediol and were restored to pre-treatment levels after 30 min serum induction (Additional file 1: Fig. S15). In a T24 cell line stably expressing GFP-Med15, Med15 foci were rapidly diminished upon 0.5% Hexanediol treatment and restored upon serum induction/Hexanediol withdrawal (Additional file 1: Fig. S16, Additional file 5: Video S4).
We found that transcriptional activation of c-Fos, c-Jun, and Egr-1 genes was significantly delayed in cells pretreated with 10% Hexanediol but minimally affected in cells pretreated with 0.5% Hexanediol. Highest levels of IEG expression were found at about 30 min after serum induction in Group I cells but instead at 60 min or 120 min after serum induction in cells pretreated with 10% Hexanediol (Additional file 1: Fig. S17b-d). However, disruption of Mediator condensates by 0.5% Hexanediol prior to serum induction (Additional file 1: Fig. S16, Additional file 5: Supplementary Video S4) did not result in a delay in IEG expression (Fig. 7b–d). Whether the presence of 0.5% Hexanediol during serum stimulation can affect IEG activation remains to be tested. Ideally, molecular reagents with improved specificity would help to better understand the functions of Mediator condensates in inducible gene expression.
Importantly, Med15 knockdown attenuated IEG activation. We found that expression of c-Fos and Egr-1 in Med15 knockdown cells after 30 min serum induction was 2–3 fold lower than wild-type U2OS cells (Additional file 1: Fig. S18a, c). Most substantial decrease in expression levels of all three IEGs upon Med15 knockdown was found at 60 min serum induction (Additional file 1: Fig. S18a-c). Thus, Med15 knockdown impairs the functions of the Mediator complex in regulating IEG activation upon serum induction.