Mediator Tail Subunits Hierarchically Couple Transcriptional Condensates to Gene Activation and Genome Organization

Cells respond to acute environmental stress by rapidly reorganizing transcriptional machinery and genome architecture, yet how these processes are mechanistically integrated remain poorly understood. We find that Mediator Tail subunits function in a hierarchical fashion to coordinate transcription factor condensate assembly, three- dimensional genome organization and transcriptional output during the heat shock response (HSR) in Saccharomyces cerevisiae. We identify the Mediator Tail triad— Med2, Med3, and Med15—as exceptionally enriched in intrinsically disordered regions and possessing strong intrinsic liquid–liquid phase separation potential, in contrast to the more structured Tail subunits Med5 and Med16. Live-cell imaging reveals that this IDR-rich triad is critically required for thermal stress-induced Heat Shock Factor 1 (Hsf1) condensate formation, HSR gene coalescence and robust transcriptional induction. Mechanistically, Med15 executes these functions through its activator-binding domains, with the IDR-rich ABD2 playing a dominant role in stabilizing Hsf1, Mediator, and RNA polymerase II (Pol II) occupancy at HSR loci, while the C-terminal IDRs of Med2 and Med3 provide critical interaction platforms that couple condensate formation to genome organization. Strikingly, Med16 defines a parallel regulatory axis: although dispensable for Hsf1 condensate nucleation, Med16 is required for Mediator and Pol II condensate formation and for HSR gene coalescence, revealing that transcription factor clustering and productive transcriptional condensates are mechanistically separable. Finally, Med5 plays a minor yet detectable role in HSR transcription and gene coalescence. Together, our findings establish a modular and hierarchical organization of the Mediator Tail that integrates phase separation, transcriptional condensate composition, and 3D genome architecture to drive rapid stress-induced gene activation. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -3-

Mediator is an evolutionarily conserved co-activator of Pol II transcription that plays a central and integrative role in gene expression. Its dynamic and modular architecture, comprising the Head, Middle, and Tail modules (Fig. 1a), enables Mediator to interact with gene-specific transcription factors (TFs), Pol II, general transcription factors such as TFIID and TFIIH, and co-activator complexes including SAGA1-4. Structural and biochemical studies have demonstrated that Mediator undergoes extensive conformational rearrangements that facilitate high-affinity Pol II engagement and transcriptional activation5,6. During pre-initiation complex (PIC) assembly and transcription initiation, Mediator regulates multiple dynamic aspects of transcription, including TF function and multivalent activator interactions, TF residence time on DNA, transcriptional bursting and reinitiation, Pol II distribution across genes, and phosphorylation of the Pol II C-terminal domain (CTD)1-4. Beyond its role in initiation, Mediator coordinates post-initiation transcriptional processes, contributing to transcription elongation, termination, and co-transcriptional mRNA splicing and nuclear export7 (reviewed in 3,4,8). In addition to these transcriptional functions, Mediator interfaces with chromatin regulatory pathways, influencing histone and DNA methylation and contributing to epigenetic transcriptional memory 3,4,9. Beyond transcription, Mediator plays a pivotal regulatory role in genome organization by orchestrating short- and long-range chromatin interactions through partnerships with diverse factors, including chromatin architectural proteins and remodelers4,8. By interacting with cohesin, Mediator promotes enhancer-promoter interactions and higher- order chromatin folding to regulate global chromatin structure10-12. These observations, primarily made in mammalian systems, reflect developmental processes. Less well understood is the role Mediator might play in the dynamic 3D genome restructuring that occurs in response to environmental stress 13-17. The robust heat shock response (HSR) system in budding yeast provides a powerful model to uncover the basic mechanisms of genome organization and gene regulation. The HSR is a conserved transcriptional program that safeguards eukaryotic cells against proteotoxic stress and is orchestrated by the transcription factor, Heat Shock Factor 1 (Hsf1)18,19. Upon activation by stressors such as heat, ethanol, or oxidative stress, Hsf1 undergoes trimerization, binds to heat shock elements (HSEs) upstream of .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -4- HSR genes, and induces expression of molecular chaperones and cytoprotective proteins18,20-22. This regulatory mechanism is evolutionarily conserved, as evidenced by the functional interchangeability of human HSF1 and its yeast counterpart23. Reversible formation of biomolecular condensates, a process that frequently involves liquid-liquid phase separation (LLPS), has emerged as a key mechanism regulating the cellular stress response from yeast to mammals15,16,24 (reviewed in 19,25). This process is primarily driven by intrinsically disordered regions (IDRs). Notably, Mediator contains extensive IDRs in many subunits, consistent with its striking conformational flexibility 26,27. Previous work has suggested that dynamic IDR-IDR interactions between TFs and Mediator can drive the formation of phase-separated condensates28,29, yet roles played by such IDR-driven interactions in fostering recruitment of regulatory factors, gene- specific transcription, and 3D genome organization remain unclear. Recent studies have shown that Hsf1 forms reversible nuclear condensates in response to acute thermal stress and that these localize to HSR gene loci in both human and yeast 16,24. In Saccharomyces cerevisiae, these heat-inducible transcriptional condensates drive HSR-HSR gene interactions across and between chromosomes14-16. In addition to driving profound reorganization of genome structure, such condensates concentrate components of the transcriptional machinery, including Hsf1, Pol II, and Mediator16,31,33. However, whether factors beyond Hsf1 drive the formation of transcriptional condensates and concomitant 3D genome reorganization, and the mechanistic basis by which they might do so, are largely unknown. Using the HSR system in budding yeast, we previously showed that the Mediator Tail triad, comprised of Med2-Med3-Med15, can be independently recruited to HSR genes through interactions with Hsf1, which uses its dual activation domains to physically interact with Med15 34,35. However, the role of Mediator in heat shock-induced transcriptional condensate formation and HSR gene clustering has remained unclear. Here, we demonstrate that subunits within the Tail triad are critically required for heat shock-induced formation of transcriptional condensates, HSR gene activation, and concomitant 3D genome restructuring, and principally act through their intrinsically disordered regions. A fourth, highly structured subunit, Med16, defines a parallel regulatory axis: although dispensable for Hsf1 condensate nucleation, Med16 is .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -5- required for Mediator and Pol II condensate formation and for HSR gene coalescence, revealing that transcription factor clustering and productive transcriptional condensates are mechanistically separable. Med5, another highly structured Tail subunit, plays comparatively minor roles in stress-induced genome restructuring and HSR gene expression. Our findings establish a modular and hierarchical organization of the Mediator Tail that integrates phase separation, transcriptional condensate composition, and 3D genome architecture to drive rapid stress-induced gene activation.

Mediator Tail triad subunits exhibit robust intrinsic LLPS potential The formation of biomolecular condensates is often driven by LLPS and is dictated by IDR abundance, amino acid composition, and sequence patterning36-39. We therefore systematically evaluated Mediator Tail subunits for IDR content, underlying amino acid composition, and intrinsic LLPS potential. IDR abundance and LLPS propensity were quantified using complementary sequence-based and machine-learning approaches for predicting intrinsic disorder and droplet-forming potential40-44. This analysis revealed that Med2, Med3, and Med15 exhibit a high degree of disorder, with >70% of each polypeptide predicted to be disordered (Fig. 1b). Med2 and Med3 contain extended IDRs, whereas Med15 features multiple IDRs that punctuate its sequence end-to-end (Fig. 1c). In contrast, Med5 and Med16 contain few regions predicted to be disordered. To evaluate the phase-separation potential of Mediator Tail subunits, we used PhasePred to predict both self-assembling (PS-Self) and partner-dependent (PS-Part) propensities. Med2, Med3, and Med15 emerged as dominant candidates, exhibiting markedly elevated PS-Self and PS-Part scores and the highest predicted phase- separation propensity among all Mediator subunits (Figs. 1d and S1 a, b). Notably, PS- Part scores strongly correlated with IDR content (R² = 0.9515), implicating IDRs within the Tail triad as major determinants of phase separation (Fig. 1e). Consistent with these findings, FuzDrop analysis also predicted elevated LLPS propensity for Med2, Med3, and Med15 (Fig. 1f). AlphaFold3 structural predictions 45 likewise suggested extensive unstructured regions in these subunits, in contrast to the more ordered architectures of .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -6- Med5 and Med16 (Fig. 1g). Together, these data indicate that the Mediator Tail triad is characterized by pronounced intrinsic disorder and a strong propensity for phase separation. Because amino acid composition is a critical determinant of IDR-driven phase separation37, we next analyzed the residue composition of Mediator Tail subunits. The Tail triad was markedly enriched in asparagine (N) and glutamine (Q) residues, with Med2 showing preferential enrichment for asparagine, Med15 for glutamine, and Med3 for both. By contrast, Med5 and Med16 exhibited minimal enrichment in these residues (Fig. S1c, d). Notably, N and Q residues were largely spatially aligned with IDRs within the Tail triad (Fig. S1e), and their combined abundance was significantly higher than that observed in other Tail subunits. Collectively, these features indicate that the Mediator Tail triad harbors intrinsic sequence and structural properties consistent with robust liquid–liquid phase separation potential. The Mediator Tail triad regulates Hsf1 condensate formation in response to heat shock HSF1, the master regulator of the heat shock response, undergoes LLPS to form nuclear condensates at HSR gene loci in both yeast and human cells 16,24 and in yeast drives the 3D reorganization of HSR genes15,16,31,46. Recently, evidence has been reported for HSF1-mediated 3D reorganization of HSR genes in mouse embryonic fibroblasts47. Although Mediator and Pol II colocalize with Hsf1 following exposure to acute thermal stress16,31,33, the contribution of individual Mediator subunits remains poorly defined. To investigate the role of the Mediator Tail in the HSR, we generated haploid yeast strains harboring individual gene deletions (med2Δ, med3Δ, med15Δ, med5Δ, med16Δ) and assessed growth phenotypes at 24°, 30°, and 37°C. Except for med5Δ, all mutants exhibited sensitivity to the elevated temperature (37°C) (Fig. S2), implicating essential roles for Med2, Med3, Med15, and Med16 in the HSR. Dynamic IDR-IDR interactions of transcriptional machinery drive phase-separated condensates, crucial for transcriptional regulation (reviewed in 36,48,49). Optimal levels of TF IDR-IDR interactions facilitate condensate formation that often correlates with transcriptional induction28,29,48-50 (although see ref.51). Mammalian Med1 plays an essential role in condensate formation in mammalian cells, facilitating the .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -7- compartmentalization of Pol II and positive regulators and activating transcription by concentrating transcriptional machinery via its large IDR28,29,39,52. To test the role of Mediator Tail subunits in HS-induced Hsf1 condensate formation, we employed live-cell imaging of wild-type (WT) and individual Tail subunit deletion strains expressing Hsf1- mNeonGreen (Hsf1-mNG). Consistent with previous observations15,16,31,33,46, Hsf1-mNG rapidly condensed in WT cells into several discrete intranuclear puncta in response to a 2.5-minute, 25° to 39°C thermal upshift. Such puncta were dynamic and began to resolve as early as 15 min (Fig. 2a). Deletion of any component of the Mediator Tail triad markedly suppressed Hsf1 condensate formation, as reflected by both the number of puncta per cell and the fraction of cells containing puncta, although the kinetics of formation and resolution were similar to those observed in WT cells (Figs. 2a,c; S3i). In contrast, deletion of MED5 or MED16 did not reduce either the kinetics or abundance of intranuclear Hsf1 condensates. Notably, Hsf1-mNG condensates persisted longer in med16Δ cells than in WT cells (Fig. 2c). Consistent with these findings, conditional depletion of Med16 using the auxin-inducible degron (mAID) system did not suppress Hsf1 clustering but instead prolonged condensate persistence, whereas AID-mediated Med15 depletion abolished Hsf1 condensation (Figs. 2b,d; S3h,j). Auxin treatment efficiently reduced Med15 and Med16 protein levels (~99% and ~90%, respectively) within 30 minutes, without affecting cell viability over the time course analyzed (Fig. S3a–f). Together, these results demonstrate that the Mediator Tail triad, which is enriched in IDRs and exhibits strong LLPS propensity, contributes to the HSR by stimulating the formation of Hsf1 condensates. By contrast, Med5 and Med16, both of which lack pronounced IDRs and LLPS propensity (Fig. 1), are dispensable for Hsf1 nucleation. Both IDR- containing and structured Mediator Tail subunits drive HS-induced HSR gene interactions Heat shock–induced transcriptional condensates containing Hsf1, Pol II, and Mediator are proposed to drive the spatial reorganization of HSR genes, with evidence supporting a functional link between Hsf1 condensate formation and intergenic HSR gene interactions 16,31. To gain further insight into the relationship between condensate assembly and HSR gene coalescence, we analyzed the contribution of Mediator Tail .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -8- subunits using a highly sensitive version of chromosome conformation capture, TaqI-3C 53, in both gene deletion and conditional depletion strains (strategy summarized in Fig. 3a). This analysis included the IDR-rich subunits Med2, Med3, and Med15, as well as the more structured subunits Med5 and Med16. In WT cells, brief heat shock (5 min) induced robust cis- and trans-intergenic interactions among HSR genes (Figs. 3b and S4). Deletion of MED2, MED3 or MED15, each of which disrupts Hsf1 condensate formation, resulted in a pronounced loss of HSR intergenic contacts. Notably, deletion of MED16 caused the most severe defect in HSR gene coalescence despite leaving Hsf1 condensates intact. Deletion of MED5 had a more variable and less impactful effect on both cis- and trans-interactions (Figs. 3b and S4). To validate the most striking phenotypes, we conditionally depleted Med15 or Med16 using AID as above. Like med15Δ, conditional depletion of Med15 severely disrupted both cis- and trans- HSR intergenic interactions (Figs. 3b, S5a). Med16-mAID produced a similar but weaker defect than med16Δ that may reflect the incomplete depletion of Med16 protein (Figs. 3b, S3d-e, S5b). We next examined intra-locus HSR interactions, including enhancer–promoter (E-P), promoter–coding region, and promoter–terminator (gene-looping) contacts. Except for Med5, all Mediator Tail subunits were required for efficient intragenic interactions in acute HS cells (Figs. 3c, S4c). Conditional depletion of Med16 yielded comparable

to med16∆ (Fig. S5b). The similar impact of gene deletion and conditional protein depletion of Med15 and Med16 argues against the effects observed having arisen from either extragenic suppression or auxin-induced effects. Together, these orthologous approaches argue for a dual role for IDR-containing Med15 in promoting Hsf1 clustering and HSR intergenic interactions and a singular role for Med16 in promoting HS-induced 3D genome restructuring. Mediator Tail triad subunits drive robust HSR gene transcription TF condensates have been proposed as a key mechanism for transcriptional regulation 28,29,32,54. In yeast, HSR gene transcription is temporally correlated with HS- induced Hsf1 condensation15,16,31,33,46. To determine whether Mediator Tail subunits implicated in Hsf1-condensate formation and HSR gene coalescence are required for HS-induced HSR transcription, we measured mRNA levels of representative HSR .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -9- genes (HSP104, SSA4, and BTN2) during a HS time course using RT–qPCR. Loss of Tail triad subunits Med2, Med3, and Med15 severely impaired HSR gene transcription, with Med15 exhibiting the most pronounced defect (Figs. 3d, S6). These transcriptional defects parallel the disruptions observed in Hsf1 condensate formation and HSR gene interactions. In contrast, deletion of Med16, which strongly disrupts HSR gene interactions but not Hsf1 condensates, caused only a moderate reduction in HSR transcription. Finally, deletion of Med5, which does not affect the number of Hsf1 condensates per cell and has a variable impact on HSR intergenic interactions, had a modest yet detectable effect on HSR gene transcription. Together, these data indicate that the Tail triad subunits are essential for robust HSR transcription, that Med16 plays a moderate but supportive role, and that Med5 plays a minor role. Med15 activator-binding domains govern the HSR through regulation of Hsf1, Mediator, and Pol II occupancy Med15 is a well-characterized Mediator subunit that regulates gene expression through interactions with transcription factors across eukaryotes, from yeast to human34,55-58. Yeast Med15 contains three N-terminal activator-binding domains (ABDs 1-3) comprised of extensive regions predicted to be intrinsically disordered (Figs. 4a, S7) and that are essential for TF binding 56,59,60. Through these domains, Med15 engages TFs via a “fuzzy” protein interface57,58,60,61. Med15 additionally harbors an evolutionarily conserved, N-terminal TF-interacting KIX domain55,56 and a C-terminal Mediator- associated domain (MAD) that mediates interactions with the remainder of the Mediator complex, TFIIE and TFIIH5,62. As a core component of the Mediator Tail, Med15 has been shown to be essential for the HSR34,35, a notion confirmed and extended by observations described above (Figs. 2, 3, S4-S6). To determine whether Med15 mediates these HSR functions through its ABDs, we generated haploid strains carrying individual genomic deletions of each domain59 (see Methods). Deletion of ABD2 (Δabd2, residues 277–404) caused an increasingly severe growth defect as temperature was elevated, although not as severe as that conferred by med15Δ (Fig. S8a-c), whereas the remaining ABD deletions had little detectable effect. Significantly, none of the domain deletions altered overall Med15 protein expression levels (Fig. S8d). .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -10- We next assessed whether the ABD deletions affected Hsf1 condensate formation. Except for Δabd2, none of the deletions significantly altered the fraction of cells with Hsf1 or kinetics of formation over the heat shock time course (Fig. 4b). Quantification of Hsf1 puncta per cell revealed a substantial reduction in the Δabd2 mutant, a moderate reduction in Δabd3, and no appreciable effect in either the Δkix or Δabd1 strains (Fig. 4c). We then examined the contribution of Med15 ABDs to intergenic interactions at HSR genes by 3C. All ABD deletions impaired intrachromosomal (cis) interactions, whereas only Δabd2 consistently disrupted interchromosomal (trans) interactions and intragenic contacts (Figs. 4d and S9a). Moreover, analysis of intragenic architecture showed that ABD2 is required for both enhancer–promoter (E–P) and promoter– terminator (gene-looping) interactions. In contrast, deletion of the KIX domain, ABD1, or ABD3 resulted in only moderate reductions in gene-looping interactions, while leaving E–P interactions largely intact. Together, these findings indicate that all Med15 ABDs contribute to the 3D genome topology of HSR genes, with ABD2 playing the predominant role. To determine how Med15 contributes to the HSR through its ABDs, we assessed the occupancy of key HSR transcriptional components, including Hsf1, Pol II (Rpb1), and the Mediator Head (Med17) and Tail (Med3) subunits, using ChIP at representative HSR gene loci. All Med15 ABDs supported recruitment of these factors, with ABD2 typically playing the most important role (Figs. 4e, S10). It is notable that Med15, acting through each of its activator binding domains, enhances HS-induced Hsf1 recruitment in addition to that of Pol II and Mediator. This observation is consistent with previous reports suggesting that Med15 can facilitate (or stabilize) the binding of gene-specific transcription factors to their cognate sites in the yeast genome 34,63. We further evaluated the role of ABD2 in driving HSR gene expression and found that its deletion severely reduced mRNA abundance of HSR genes (Figs. 4f, S9b). Together, these

indicate that Med15 ABDs are critical for the HSR, functioning to enhance or stabilize transcriptional machinery at HSR genes in an HS-dependent manner, with ABD2 playing the most important role. The C-terminal IDRs of Med2 and Med3 regulate the HSR by modulating HSF1 condensation, genome topology and factor occupancy .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -11- Previous studies have demonstrated that Hsf1 is required for the HS-induced formation of transcriptional condensates, 3D genome restructuring, and HSR gene activation in acutely stressed cells15,16,31,46. Consistent with this framework, our results show that each component of the Mediator Tail triad—Med2, Med3, and Med15—is essential for a robust HSR, as loss of any individual subunit severely compromises the response. These findings support a model in which the Tail triad subunits function cooperatively as an integrated triad rather than as independent factors5,64-66. Furthermore, Med15 contributes to the HSR through its activator-binding domains (ABDs), with ABD2 playing a predominant role (Figs. 4, S9, S10). To define the IDR-mediated interactions that underlie Tail triad assembly, we generated intermolecular interaction maps using the FINCHES online tool67 . These analyses predicted strong interactions between Med15 ABDs 1–3 and the C-terminal IDRs of Med2 (residues 290–400) and Med3 (residues 301–374) (Fig. S11). In contrast, the same ABDs showed only weak to moderate predicted interactions with Med5 or Med16. To determine whether Med2 and Med3 contribute to the HSR through their C-terminal IDRs, we used CRISPR/Cas9 to generate genomic deletions of these glutamine- and asparagine-enriched regions (see Fig. 5a). The med2-ΔIDR mutant grew poorly at elevated temperature while med3-ΔIDR evinced no growth defect (Fig. S12a). Notably, deletion of the C-terminal IDRs did not alter the protein expression levels of either med2-ΔIDR or med3-ΔIDR (Fig. S12b, c). We next examined the impact of these deletions on Hsf1 condensate dynamics during a heat shock time course. Both med2-ΔIDR and med3-ΔIDR mutants exhibited defects in Hsf1 condensate formation throughout the HS time course, paralleling the more pronounced phenotypes of their corresponding gene deletions (Fig. 5b). At 10 min HS, both med2-ΔIDR and med3-ΔIDR mutants displayed significant reductions in the number of Hsf1 condensates relative to WT. However, the defects were less pronounced than those observed in the corresponding gene deletion strains (Fig. 5c). Next, we assessed the role of the Med2 and Med3 C-terminal IDRs in driving HSR gene interactions using the 3C method at the 5-min HS time point. Both med2-ΔIDR and med3-ΔIDR exhibited severe defects in HSR gene coalescence, disrupting cis-, trans-, and intragenic interactions (Figs. 5d and S13a). These defects closely mirrored the .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -12- phenotypes of the gene deletion mutants, indicating that Med2 and Med3 play a vital role in driving HSR gene coalescence through their C-terminal IDRs. To further define how the Med2 C-terminal IDR influences Hsf1 condensate formation and HSR gene interactions, we assessed the occupancy of Hsf1, Pol II (Rpb1), and Med15 at representative HSR genes under NHS and HS conditions. Deletion of MED2 markedly reduced recruitment of all three factors, and the med2-ΔIDR produced a comparable reduction of Rpb1 and Med15 at each locus examined (Figs. 5e, S13b–d). We next assessed the contribution of the Med2 and Med3 C-terminal IDRs to HSR gene transcription during the heat shock time course by RT–qPCR. Deletion of the Med2 C- terminal IDR (med2-ΔIDR) markedly impaired induction of HSR transcripts, including HSP104 and SSA4, phenocopying the MED2 deletion, whereas med3-ΔIDR caused a more modest transcriptional defect relative to med3Δ (Figs. 5f, S14). Together, these

suggest that Med2 and Med3 regulate the HSR through their C-terminal IDRs, with Med2 playing a dominant role in coordinating Hsf1 condensation, HSR gene coalescence, and transcriptional activation. Med16 Is required for the formation of functional transcriptional condensates Observations described above support a model in which Tail triad–mediated multivalent interactions integrate condensate assembly with higher-order chromatin organization to drive robust transcription. By contrast, med16Δ profoundly disrupted HSR gene coalescence without affecting Hsf1 condensate formation (Figs. 2, 3, S4), consistent with the notion that genome reorganization occurs downstream of TF clustering16,46. To determine how Hsf1 condensates in the med16Δ mutant differ from those in WT cells, we performed live cell colocalization analyses to define the contribution of individual Mediator subunits to transcriptional condensate composition. Heat shock-induced WT cells exhibited robust recruitment of Mediator and Pol II into Hsf1 condensates, resulting in dynamic transcriptional assemblies. Deletion of MED5 had little effect on condensate formation, whereas loss of either MED2 or MED3 severely disrupted Hsf1, Mediator, and Pol II condensates (Fig. 6a-d, g-j). Although med5Δ did not alter the fraction of cells containing Hsf1, Mediator (Med15), or Pol II (Rpb3) condensates, it modestly reduced the number of Mediator puncta per cell (Fig. 6d). .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -13- In contrast, MED16 deletion did not impair Hsf1 condensate formation, consistent with earlier observations (Fig. 2), but markedly reduced the fraction of cells containing Mediator and Pol II condensates (Fig. 6a-c, g-i), as well as the proportion of Hsf1 puncta colocalized with Mediator or Pol II (Fig. 6e, f, k, l). While the number of Hsf1 puncta per cell was unchanged in med16Δ cells, Mediator and Pol II puncta were substantially decreased (Fig. 6d, j), indicating that Med16 is required for efficient assembly of Mediator and Pol II condensates and for stable incorporation of transcriptional machinery into Hsf1 condensates. Notably, recruitment of Hsf1, Pol II, and Med15 to HSR genes was comparable between heat-shocked WT and med16Δ cells (Fig. S15), indicating that Med16 primarily regulates condensate organization rather than factor recruitment. Together, these findings support a hierarchical model in which Hsf1 condensate formation depends on the Mediator Tail triad, whereas productive transcriptional condensates, HSR gene coalescence, and robust transcriptional output additionally require Med16.

While Mediator has been shown to be a central component of transcriptional condensates in both lower and higher eukaryotes, how its individual subunits shape condensate organization, genome architecture, and transcription remains unclear. Here, we have investigated this question in Saccharomyces cerevisiae and identified a hierarchical organization of the Mediator Tail module that links IDR-driven condensate formation to higher-order chromatin architecture and robust transcription during the HSR. Mediator Tail triad subunits deploy their IDRs to drive Hsf1 condensate formation, 3D genome restructuring and HSR gene activation in acutely stressed cells Our computational and genetic analyses identify Med2, Med3, and Med15 as an IDR- rich (Fig. 1), phase-separation–competent Tail triad that is essential for the formation of Hsf1 condensates (Fig. 2). These subunits exhibit extensive intrinsic disorder, enrichment of glutamine/asparagine residues, and high LLPS propensity—sequence features previously shown to promote condensate assembly through multivalent, weak interactions 37,38 . We found that loss of any Tail triad component markedly suppressed Hsf1 condensate formation and compromised HSR gene coalescence and transcription .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -14- (Figs. 2, 3). Previous studies have shown that Med2-Med3-Med15 can physically interact with gene-specific transcription factors and be directly recruited to UAS regions34,35,64-66. Here, we have extended these findings by demonstrating that the IDRs within the Tail triad are functionally essential and promote Hsf1 condensate formation during the HSR. Notably, deletion or depletion of Med16, which disrupts the physical linkage between the Tail triad and core Mediator35,66,70, does not abolish Hsf1 condensate formation (Fig. 2). Instead, the released IDR-rich Tail triad is sufficient to interact with Hsf1 (suggested by Hsf1 ChIP analysis of med16∆ vs. WT; Figs. S10 and S15), likely through multivalent IDR–IDR interactions, to enable heat shock–induced transcriptional activation of HSR genes (Figs. 3d and S6). These findings, together with the observation that deletion of either the ABD2 domain of Med15 or the C-terminal IDRs of Med2 / Med3 profoundly affects Hsf1 condensate formation, 3D genome architecture, and HSR gene activation (Fig. 5), highlight the ability of the Tail triad to drive the three pillars of the heat shock transcriptional response, and do so via its IDRs. The graded phenotypes observed upon IDR deletion—less severe than complete subunit loss—suggest that these IDRs primarily tune condensate strength and transcriptional efficiency rather than serving as absolute on/off switches. Hsf1 condensate formation and genome restructuring can be mechanistically uncoupled An important implication of this work is that transcriptional condensate formation and 3D genome organization are mechanistically separable. Deletion of any of the three IDR- rich Mediator tail triad subunits disrupts both Hsf1 condensate formation and HSR gene coalescence, consistent with a condensate-dependent mode of genome organization16,31. In contrast, loss of Med16 selectively abolishes intergenic and intragenic HSR gene interactions without impairing Hsf1 condensate assembly (Figs. 2 and 3). In this regard, the role of Med16 resembles that of two nuclear basket proteins, Mlp1 and Nup2, that are critical for driving HS-induced HSR gene coalescence yet dispensable for Hsf1 condensate formation33. In addition, our data support a hierarchical regulatory model in which Hsf1–Mediator Tail triad interactions nucleate Hsf1 condensates through IDR-mediated assembly, while additional Mediator-dependent mechanisms orchestrated by the triad and Med16 stabilize Pol II- and Mediator-enriched transcriptional complexes and promote long- .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -15- range chromatin contacts across HSR genes (model presented in Fig. 7). This functional separation is conceptually aligned with recent studies demonstrating that condensate formation and downstream transcriptional processes, including productive transcription and genome restructuring, can be genetically and mechanistically uncoupled16,31,33,46 . Together, these findings establish Hsf1 condensate nucleation and genome organization as distinct yet coordinated layers of transcriptional regulation during the HSR. Med15 ABD2 integrates TF binding, condensate assembly, gene coalescence and transcription Med15 has emerged as a central integrator of HSR regulation. Its activator-binding domains engage transcription factors through dynamic, “fuzzy” interfaces that have been shown to promote LLPS and transcriptional activation 57,58,60,61. Our domain- specific analyses reveal that the intrinsically disordered ABD2 region plays a uniquely essential role, being required for Hsf1 condensate formation, HSR gene coalescence, recruitment of Mediator and Pol II, and robust transcriptional activation (Figs. 4 and S9). Mechanistically, ABD2 enhances the occupancy of Hsf1, Mediator, and Pol II at HSR loci, thereby directly coupling condensate assembly to transcriptional output. These findings extend prior work demonstrating that Med15 mediates transcription through multivalent interactions with TFs 28,34,56,58,59,63 by directly linking a specific Med15 IDR to condensate biology and genome organization. In contrast, other ABDs contribute more modestly, primarily influencing intrachromosomal architecture and factor occupancy, highlighting functional specialization among Med15 interaction surfaces. Whether the dominant role played by ADB2 is due to its being the primary target of Hsf1 was not addressed here. Other yeast and mammalian TFs have been observed to interact with the four ABDs in a “fuzzy” manner 57,58,60 or primarily with the KIX domain55. Future work will address the question of which Med15 domain(s) Hsf1 physically contacts. Med16 promotes HSR gene coalescence by facilitating the formation of productive transcriptional condensates Our live cell colocalization analyses uncover a distinct and essential role for Med16 in organizing the transcriptional machinery during the HSR. Med16 is specifically required for the formation and/or stabilization of Mediator and Pol II condensates and their .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -16- incorporation into Hsf1 condensates (Fig. 6), while being dispensable for Hsf1 condensate assembly (Fig. 2). In the Med16 mutant, Hsf1 puncta form; however, these condensates fail to efficiently recruit or retain Mediator and Pol II, and this correlates with defective higher-order genome organization and reduced transcriptional output. These observations indicate that Hsf1 condensate nucleation alone is insufficient to drive HSR gene coalescence and is consistent with the findings of others that implicate Pol II condensation as a key driver of cis- and trans- HSR intergenic interactions in HS cells31. Together, these results position Med16 as a critical organizer of transcriptional machinery, acting downstream of Hsf1 condensate formation to enable genome organization and robust transcription during stress. A hierarchical model for Mediator-driven transcriptional regulation Together, our findings support a hierarchical and modular model of Mediator function during the HSR. In this framework, Hsf1 cooperates with the intrinsically disordered, phase-separation-prone Mediator Tail triad to nucleate transcriptional condensates through multivalent IDR-mediated interactions, consistent with current models of condensate-driven transcriptional activation 28,29,39,48,71. Within this triad, Med15— particularly its intrinsically disordered ABD2 region— acts as a central integrator, linking transcription factor binding to condensate assembly and transcriptional activation. The C-terminal IDRs of Med2 and Med3 further reinforce condensate stability and enhance recruitment of Mediator and Pol II, thereby promoting robust transcriptional output. In parallel, the structured Mediator subunit Med16 functions independently of Hsf1 condensate nucleation to organize Mediator–Pol II assemblies and facilitate higher- order chromatin interactions required for productive transcription. Med5 plays a detectable, albeit comparatively minor, role in promoting the formation of Hsf1 condensates, stimulating 3D genome restructuring, and contributing to HS-induced transcriptional activation (see Figs. 7 and S16 for a schematic model and graphic summary). This division of labor supports a model in which condensate nucleation, transcriptional machinery stabilization, and genome organization represent distinct yet coordinated regulatory layers. Given the evolutionary conservation of Mediator architecture 1,4,3,5 and the widespread role of IDR-driven transcriptional regulation, these principles are likely to extend beyond the yeast HSR to diverse stress responses, .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -17- enhancer-driven gene programs, and pathological transcriptional states in higher eukaryotes. We note that a recent study observed no effect on heat shock-induced Hsf1 condensation when Med15 was conditionally depleted using the anchor-away (AA) technique31. A key difference between the AA approach and ours is that anchoring Med15 in the cytoplasm depletes the nucleus of not only Med15 but also potentially other factors (including other Mediator subunits35), in marked contrast to our subunit- specific approach, which does not (Fig. S3g). These contrasting observations imply that the Triad serves to counter an activity within one or more nuclear proteins that suppresses inappropriate Hsf1 condensation, an idea that will be tested in future work.

and future directions While our findings establish a central role for Mediator Tail subunits in coordinating transcriptional condensate formation, genome organization, and transcription during the HSR, several important questions remain. First, although computational analyses and genetic perturbations strongly implicate LLPS–driven mechanisms, direct biochemical reconstitution of the Mediator Tail and interactions of individual subunits with Hsf1 will be necessary to quantitatively define the properties of the condensates, constituent stoichiometries, and dynamic behavior. Second, our study focuses on an acute stress- responsive transcriptional program in yeast; whether the modular and hierarchical principles uncovered here extend to developmental enhancers, super-enhancers, or disease-associated regulatory elements in higher eukaryotes remains to be determined. Finally, although Med16 emerges as a key organizer of the transcriptional machinery and higher-order genome architecture, independent of Hsf1 condensate nucleation, the specific molecular interfaces and cofactors that mediate these functions remain unknown. One such candidate is Pol II, as cryo-EM analysis indicates that Med16 interacts with Rpb1 and Rpb9 in reconstituted Mediator-PIC complexes, an interaction that does not directly involve any other Tail subunit (S. Nagai and R.D. Kornberg, personal communication). Future studies integrating in vitro phase separation assays, high-resolution live-cell imaging—including single- particle tracking— and genome-wide chromatin conformation approaches such as Hi-C will be essential to elucidate how .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -18- Mediator integrates condensate biology with 3D genome organization to regulate transcriptional output across diverse cellular contexts.

We thank David Pincus and Surabhi Chowdhary for critical reading of the manuscript; Amoldeep S. Kainth for Med15 schematics; Steve Hahn, Tom Ellis and Surabhi Chowdhary for plasmid constructs; Kelly Tatchell for yeast knockout strains; and Shigeki Nagai and Roger D. Kornberg for sharing unpublished information. This work was supported by NIH R01GM13988 and an LSUHSC intramural grant awarded to DSG and Ike Muslow post- and pre-doctoral fellowships awarded to GM, SM and LSR. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -19-

1. Malik, S., and Roeder, R.G. (2010). The metazoan Mediator co-activator complex as an integrative hub for transcriptional regulation. Nat Rev Genet 11, 761–772. 10.1038/nrg2901. 2. Allen, B.L., and Taatjes, D.J. (2015). The Mediator complex: a central integrator of transcription. Nat Rev Mol Cell Biol 16, 155–166. 10.1038/nrm3951. 3. Soutourina, J. (2018). Transcription regulation by the Mediator complex. Nat Rev Mol Cell Biol 19, 262–274. 10.1038/nrm.2017.115. 4. Richter, W.F ., Nayak, S., Iwasa, J., and Taatjes, D.J. (2022). The Mediator complex as a master regulator of transcription by RNA polymerase II. Nat Rev Mol Cell Biol 23, 732–749. 10.1038/s41580-022-00498-3. 5. Robinson, P .J., Trnka, M.J., Pellarin, R., Greenberg, C.H., Bushnell, D.A., Davis, R., Burlingame, A.L., Sali, A., and Kornberg, R.D. (2015). Molecular architecture of the yeast Mediator complex. Elife 4. 10.7554/eLife.08719. 6. Rengachari, S., Schilbach, S., Aibara, S., Dienemann, C., and Cramer, P . (2021). Structure of the human Mediator-RNA polymerase II pre-initiation complex. Nature 594, 129–133. 10.1038/s41586-021-03555-7. 7. Kremer, S.B., Kim, S., Jeon, J.O., Moustafa, Y .W., Chen, A., Zhao, J., and Gross, D.S. (2012). Role of Mediator in regulating Pol II elongation and nucleosome displacement in Saccharomyces cerevisiae. Genetics 191, 95–106. 10.1534/genetics.111.135806. 8. Maalouf, C.A., Alberti, A., and Soutourina, J. (2024). Mediator complex in transcription regulation and DNA repair: Relevance for human diseases. DNA Repair (Amst) 141, 103714. 10.1016/j.dnarep.2024.103714. 9. Yin, J.W., and Wang, G. (2014). The Mediator complex: a master coordinator of transcription and cell lineage development. Development 141, 977–987. 10.1242/dev.098392. 10. Kagey, M.H., Newman, J.J., Bilodeau, S., Zhan, Y ., Orlando, D.A., van Berkum, N.L., Ebmeier, C.C., Goossens, J., Rahl, P .B., Levine, S.S., et al. (2010). Mediator and cohesin connect gene expression and chromatin architecture. Nature 467, 430–435. 10.1038/nature09380. 11. Haarhuis, J.H.I., van der Weide, R.H., Blomen, V .A., Flach, K.D., Teunissen, H., Willems, L., Brummelkamp, T.R., Rowland, B.D., and de Wit, E. (2022). A Mediator- cohesin axis controls heterochromatin domain formation. Nat Commun 13, 754. 10.1038/s41467-022-28377-7. 12. Ramasamy, S., Aljahani, A., Karpinska, M.A., Cao, T.B.N., Velychko, T., Cruz, J.N., Lidschreiber, M., and Oudelaar, A.M. (2023). The Mediator complex regulates enhancer-promoter interactions. Nat Struct Mol Biol 30, 991–1000. 10.1038/s41594- 023-01027-2. 13. Brickner, D.G., Ahmed, S., Meldi, L., Thompson, A., Light, W., Young, M., Hickman, T.L., Chu, F ., Fabre, E., and Brickner, J.H. (2012). Transcription factor binding to a DNA zip code controls interchromosomal clustering at the nuclear periphery. Dev Cell 22, 1234–1246. 10.1016/j.devcel.2012.03.012. 14. Chowdhary, S., Kainth, A.S., and Gross, D.S. (2017). Heat Shock Protein Genes Undergo Dynamic Alteration in Their Three-Dimensional Structure and Genome .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -20- Organization in Response to Thermal Stress. Mol Cell Biol 37. 10.1128/MCB.00292- 17. 15. Chowdhary, S., Kainth, A.S., Pincus, D., and Gross, D.S. (2019). Heat Shock Factor 1 Drives Intergenic Association of Its Target Gene Loci upon Heat Shock. Cell Rep 26, 18–28 e15. 10.1016/j.celrep.2018.12.034. 16. Chowdhary, S., Kainth, A.S., Paracha, S., Gross, D.S., and Pincus, D. (2022). Inducible transcriptional condensates drive 3D genome reorganization in the heat shock response. Mol Cell 82, 4386–4399 e4387. 10.1016/j.molcel.2022.10.013. 17. Lee, J., Simpson, L., Li, Y ., Becker, S., Zou, F ., Zhang, X., and Bai, L. (2024). Transcription factor condensates, 3D clustering, and gene expression enhancement of the MET regulon. Elife 13. 10.7554/eLife.96028. 18. Verghese, J., Abrams, J., Wang, Y ., and Morano, K.A. (2012). Biology of the heat shock response and protein chaperones: budding yeast (Saccharomyces cerevisiae) as a model system. Microbiol Mol Biol Rev 76, 115–158. 10.1128/MMBR.05018-11. 19. Dea, A., and Pincus, D. (2024). The Heat Shock Response as a Condensate Cascade. J Mol Biol 436, 168642. 10.1016/j.jmb.2024.168642. 20. Sorger, P .K., and Nelson, H.C. (1989). Trimerization of a yeast transcriptional activator via a coiled-coil motif. Cell 59, 807–813. 10.1016/0092-8674(89)90604-1. 21. Mahat, D.B., Salamanca, H.H., Duarte, F .M., Danko, C.G., and Lis, J.T. (2016). Mammalian Heat Shock Response and Mechanisms Underlying Its Genome-wide Transcriptional Regulation. Mol Cell 62, 63–78. 10.1016/j.molcel.2016.02.025. 22. Pincus, D., Anandhakumar, J., Thiru, P ., Guertin, M.J., Erkine, A.M., and Gross, D.S. (2018). Genetic and epigenetic determinants establish a continuum of Hsf1 occupancy and activity across the yeast genome. Mol Biol Cell 29, 3168–3182. 10.1091/mbc.E18-06-0353. 23. Liu, X.D., Liu, P .C., Santoro, N., and Thiele, D.J. (1997). Conservation of a stress response: human heat shock transcription factors functionally substitute for yeast HSF . EMBO J 16, 6466–6477. 10.1093/emboj/16.21.6466. 24. Zhang, H., Shao, S., Zeng, Y ., Wang, X., Qin, Y ., Ren, Q., Xiang, S., Wang, Y ., Xiao, J., and Sun, Y . (2022). Reversible phase separation of HSF1 is required for an acute transcriptional response during heat shock. Nat Cell Biol 24, 340–352. 10.1038/s41556-022-00846-7. 25. Mohajan, S., and Gross, D.S. (2025). Transcriptional condensates and the nuclear pore complex regulate gene expression and 3D genome architecture in response to stress. Biochem Soc Trans 53, 1295–1309. 10.1042/BST20253086. 26. Toth-Petroczy, A., Oldfield, C.J., Simon, I., Takagi, Y ., Dunker, A.K., Uversky, V .N., and Fuxreiter, M. (2008). Malleable machines in transcription regulation: the mediator complex. PLoS Comput Biol 4, e1000243. 10.1371/journal.pcbi.1000243. 27. Nagulapalli, M., Maji, S., Dwivedi, N., Dahiya, P ., and Thakur, J.K. (2016). Evolution of disorder in Mediator complex and its functional relevance. Nucleic Acids Res 44, 1591–1612. 10.1093/nar/gkv1135. 28. Boija, A., Klein, I.A., Sabari, B.R., Dall'Agnese, A., Coffey, E.L., Zamudio, A. V ., Li, C.H., Shrinivas, K., Manteiga, J.C., Hannett, N.M., et al. (2018). Transcription Factors Activate Genes through the Phase-Separation Capacity of Their Activation Domains. Cell 175, 1842–1855 e1816. 10.1016/j.cell.2018.10.042. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -21- 29. Sabari, B.R., Dall'Agnese, A., Boija, A., Klein, I.A., Coffey, E.L., Shrinivas, K., Abraham, B.J., Hannett, N.M., Zamudio, A. V ., Manteiga, J.C., et al. (2018). Coactivator condensation at super-enhancers links phase separation and gene control. Science 361. 10.1126/science.aar3958. 30. Brodsky, S., Jana, T., Mittelman, K., Chapal, M., Kumar, D.K., Carmi, M., and Barkai, N. (2020). Intrinsically Disordered Regions Direct Transcription Factor In Vivo Binding Specificity. Mol Cell 79, 459–471 e454. 10.1016/j.molcel.2020.05.032. 31. Chowdhary, S., Paracha, S., Dyer, L., and Pincus, D. (2025). Emergent 3D genome reorganization from the stepwise assembly of transcriptional condensates. bioRxiv. 10.1101/2025.02.23.639564. 32. Hnisz, D., Shrinivas, K., Young, R.A., Chakraborty, A.K., and Sharp, P .A. (2017). A Phase Separation Model for Transcriptional Control. Cell 169, 13–23. 10.1016/j.cell.2017.02.007. 33. Mohajan, S., Rubio, L.S., and Gross, D.S. (2025). Nuclear basket proteins Nup2 and Mlp1 drive heat shock-induced 3D genome restructuring downstream of transcriptional activation. J Biol Chem 301, 110568. 10.1016/j.jbc.2025.110568. 34. Kim, S., and Gross, D.S. (2013). Mediator recruitment to heat shock genes requires dual Hsf1 activation domains and mediator tail subunits Med15 and Med16. J Biol Chem 288, 12197–12213. 10.1074/jbc.M112.449553. 35. Anandhakumar, J., Moustafa, Y .W., Chowdhary, S., Kainth, A.S., and Gross, D.S. (2016). Evidence for Multiple Mediator Complexes in Yeast Independently Recruited by Activated Heat Shock Factor. Mol Cell Biol 36, 1943–1960. 10.1128/MCB.00005- 16. 36. Banani, S.F ., Lee, H.O., Hyman, A.A., and Rosen, M.K. (2017). Biomolecular condensates: organizers of cellular biochemistry. Nat Rev Mol Cell Biol 18, 285–298. 10.1038/nrm.2017.7. 37. Wang, J., Choi, J.M., Holehouse, A.S., Lee, H.O., Zhang, X., Jahnel, M., Maharana, S., Lemaitre, R., Pozniakovsky, A., Drechsel, D., et al. (2018). A Molecular Grammar Governing the Driving Forces for Phase Separation of Prion-like RNA Binding Proteins. Cell 174, 688–699 e616. 10.1016/j.cell.2018.06.006. 38. Martin, E.W., Holehouse, A.S., Peran, I., Farag, M., Incicco, J.J., Bremer, A., Grace, C.R., Soranno, A., Pappu, R.V ., and Mittag, T. (2020). Valence and patterning of aromatic residues determine the phase behavior of prion-like domains. Science 367, 694–699. 10.1126/science.aaw8653. 39. Lyons, H., Veettil, R.T., Pradhan, P ., Fornero, C., De La Cruz, N., Ito, K., Eppert, M., Roeder, R.G., and Sabari, B.R. (2023). Functional partitioning of transcriptional regulators by patterned charge blocks. Cell 186, 327–345 e328. 10.1016/j.cell.2022.12.013. 40. Wilkins, M.R., Gasteiger, E., Bairoch, A., Sanchez, J.C., Williams, K.L., Appel, R.D., and Hochstrasser, D.F . (1999). Protein identification and analysis tools in the ExPASy server. Methods Mol Biol 112, 531–552. 10.1385/1-59259-584-7:531. 41. Xue, B., Dunbrack, R.L., Williams, R.W., Dunker, A.K., and Uversky, V .N. (2010). PONDR-FIT: a meta-predictor of intrinsically disordered amino acids. Biochim Biophys Acta 1804, 996–1010. 10.1016/j.bbapap.2010.01.011. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -22- 42. Hardenberg, M., Horvath, A., Ambrus, V ., Fuxreiter, M., and Vendruscolo, M. (2020). Widespread occurrence of the droplet state of proteins in the human proteome. Proc Natl Acad Sci U S A 117, 33254–33262. 10.1073/pnas.2007670117. 43. Erdos, G., Pajkos, M., and Dosztanyi, Z. (2021). IUPred3: prediction of protein disorder enhanced with unambiguous experimental annotation and visualization of evolutionary conservation. Nucleic Acids Res 49, W297–W303. 10.1093/nar/gkab408. 44. Chen, Z., Hou, C., Wang, L., Yu, C., Chen, T., Shen, B., Hou, Y ., Li, P ., and Li, T. (2022). Screening membraneless organelle participants with machine-learning models that integrate multimodal features. Proc Natl Acad Sci U S A 119, e2115369119. 10.1073/pnas.2115369119. 45. Abramson, J., Adler, J., Dunger, J., Evans, R., Green, T., Pritzel, A., Ronneberger, O., Willmore, L., Ballard, A.J., Bambrick, J., et al. (2024). Addendum: Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 636, E4. 10.1038/s41586-024-08416-7. 46. Rubio, L.S., Mohajan, S., and Gross, D.S. (2024). Heat Shock Factor 1 forms nuclear condensates and restructures the yeast genome before activating target genes. Elife 12. ARTN RP9246410.7554/eLife.92464. 47. Pataki, E., and Gerst, J.E. (2025). Conservation of mRNA operon formation in control of the heat shock response in mammalian cells. Sci Adv 11, eadu0315. 10.1126/sciadv.adu0315. 48. Sabari, B.R. (2020). Biomolecular Condensates and Gene Activation in Development and Disease. Dev Cell 55, 84–96. 10.1016/j.devcel.2020.09.005. 49. Pei, G., Lyons, H., Li, P ., and Sabari, B.R. (2025). Transcription regulation by biomolecular condensates. Nat Rev Mol Cell Biol 26, 213–236. 10.1038/s41580- 024-00789-x. 50. Nair, S.J., Yang, L., Meluzzi, D., Oh, S., Yang, F ., Friedman, M.J., Wang, S., Suter, T., Alshareedah, I., Gamliel, A., et al. (2019). Phase separation of ligand-activated enhancers licenses cooperative chromosomal enhancer assembly. Nat Struct Mol Biol 26, 193–203. 10.1038/s41594-019-0190-5. 51. Palacio, M., and Taatjes, D.J. (2025). Real-time visualization of reconstituted transcription reveals RNAPII activation mechanisms at single promoters. Cell Rep 44, 116251. 10.1016/j.celrep.2025.116251. 52. De La Cruz, N., Pradhan, P ., Veettil, R.T., Conti, B.A., Oppikofer, M., and Sabari, B.R. (2024). Disorder-mediated interactions target proteins to specific condensates. Mol Cell 84, 3497–3512 e3499. 10.1016/j.molcel.2024.08.017. 53. Chowdhary, S., Kainth, A.S., and Gross, D.S. (2020). Chromosome conformation capture that detects novel cis- and trans-interactions in budding yeast. Methods 170, 4–16. 10.1016/j.ymeth.2019.06.023. 54. Chong, S., Dugast-Darzacq, C., Liu, Z., Dong, P ., Dailey, G.M., Cattoglio, C., Heckert, A., Banala, S., Lavis, L., Darzacq, X., and Tjian, R. (2018). Imaging dynamic and selective low-complexity domain interactions that control gene transcription. Science 361. 10.1126/science.aar2555. 55. Yang, F ., Vought, B.W., Satterlee, J.S., Walker, A.K., Jim Sun, Z. Y ., Watts, J.L., DeBeaumont, R., Saito, R.M., Hyberts, S.G., Yang, S., et al. (2006). An ARC/Mediator .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -23- subunit required for SREBP control of cholesterol and lipid homeostasis. Nature 442, 700–704. 10.1038/nature04942. 56. Jedidi, I., Zhang, F ., Qiu, H., Stahl, S.J., Palmer, I., Kaufman, J.D., Nadaud, P .S., Mukherjee, S., Wingfield, P .T., Jaroniec, C.P ., and Hinnebusch, A.G. (2010). Activator Gcn4 employs multiple segments of Med15/Gal11, including the KIX domain, to recruit mediator to target genes in vivo. J Biol Chem 285, 2438–2455. 10.1074/jbc.M109.071589. 57. Tuttle, L.M., Pacheco, D., Warfield, L., Luo, J., Ranish, J., Hahn, S., and Klevit, R.E. (2018). Gcn4-Mediator Specificity Is Mediated by a Large and Dynamic Fuzzy Protein-Protein Complex. Cell Rep 22, 3251–3264. 10.1016/j.celrep.2018.02.097. 58. Tuttle, L.M., Pacheco, D., Warfield, L., Wilburn, D.B., Hahn, S., and Klevit, R.E. (2021). Mediator subunit Med15 dictates the conserved "fuzzy" binding mechanism of yeast transcription activators Gal4 and Gcn4. Nat Commun 12, 2220. 10.1038/s41467-021-22441-4. 59. Herbig, E., Warfield, L., Fish, L., Fishburn, J., Knutson, B.A., Moorefield, B., Pacheco, D., and Hahn, S. (2010). Mechanism of Mediator recruitment by tandem Gcn4 activation domains and three Gal11 activator-binding domains. Mol Cell Biol 30, 2376–2390. 10.1128/MCB.01046-09. 60. Warfield, L., Tuttle, L.M., Pacheco, D., Klevit, R.E., and Hahn, S. (2014). A sequence- specific transcription activator motif and powerful synthetic variants that bind Mediator using a fuzzy protein interface. Proc Natl Acad Sci U S A 111, E3506–3513. 10.1073/pnas.1412088111. 61. Brzovic, P .S., Heikaus, C.C., Kisselev, L., Vernon, R., Herbig, E., Pacheco, D., Warfield, L., Littlefield, P ., Baker, D., Klevit, R.E., and Hahn, S. (2011). The acidic transcription activator Gcn4 binds the mediator subunit Gal11/Med15 using a simple protein interface forming a fuzzy complex. Mol Cell 44, 942–953. 10.1016/j.molcel.2011.11.008. 62. Sakurai, H., Kim, Y .J., Ohishi, T., Kornberg, R.D., and Fukasawa, T. (1996). The yeast GAL11 protein binds to the transcription factor IIE through GAL11 regions essential for its in vivo function. Proc Natl Acad Sci U S A 93, 9488–9492. 10.1073/pnas.93.18.9488. 63. Mindel, V ., Brodsky, S., Yung, H., Manadre, W., and Barkai, N. (2024). Revisiting the model for coactivator recruitment: Med15 can select its target sites independent of promoter-bound transcription factors. Nucleic Acids Res 52, 12093–12111. 10.1093/nar/gkae718. 64. Zhang, F ., Sumibcay, L., Hinnebusch, A.G., and Swanson, M.J. (2004). A triad of subunits from the Gal11/tail domain of Srb mediator is an in vivo target of transcriptional activator Gcn4p. Mol Cell Biol 24, 6871–6886. 10.1128/MCB.24.15.6871-6886.2004. 65. Ansari, S.A., Ganapathi, M., Benschop, J.J., Holstege, F .C., Wade, J.T., and Morse, R.H. (2012). Distinct role of Mediator tail module in regulation of SAGA-dependent, TATA-containing genes in yeast. EMBO J 31, 44–57. 10.1038/emboj.2011.362. 66. Saleh, M.M., Jeronimo, C., Robert, F ., and Zentner, G.E. (2021). Connection of core and tail Mediator modules restrains transcription from TFIID-dependent promoters. PLoS Genet 17, e1009529. 10.1371/journal.pgen.1009529. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -24- 67. Lotthammer, J.M., Ginell, G.M., Griffith, D., Emenecker, R.J., and Holehouse, A.S. (2024). Direct prediction of intrinsically disordered protein conformational properties from sequence. Nat Methods 21, 465–476. 10.1038/s41592-023-02159- 5. 68. Cho, W.K., Spille, J.H., Hecht, M., Lee, C., Li, C., Grube, V ., and Cisse, II (2018). Mediator and RNA polymerase II clusters associate in transcription-dependent condensates. Science 361, 412–415. 10.1126/science.aar4199. 69. Cramer, P . (2019). Organization and regulation of gene transcription. Nature 573, 45– 54. 10.1038/s41586-019-1517-4. 70. Galdieri, L., Desai, P ., and Vancura, A. (2012). Facilitated assembly of the preinitiation complex by separated tail and head/middle modules of the mediator. J Mol Biol 415, 464–474. 10.1016/j.jmb.2011.11.020. 71. Jonas, F ., Navon, Y ., and Barkai, N. (2025). Intrinsically disordered regions as facilitators of the transcription factor target search. Nat Rev Genet 26, 424–435. 10.1038/s41576-025-00816-3. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -25-

Yeast Strain Construction Saccharomyces cerevisiae Mediator subunit deletion mutants (med2Δ, med3Δ, med5Δ, med15Δ and med16Δ) and MED15 domain deletion mutants were generated through homologous recombination using W303-1A and LRY037 as recipient strains1. PCR amplification of med2Δ::KAN-MX and med15Δ::KAN-MX was performed using chimeric primers of the respective genes and plasmid pFA-6a-KAN-MX6 as template. In contrast, med3Δ, med5Δ and med16Δ amplicons were amplified from genomic DNA isolated from the respective strains contained in the Res-Gen deletion collection. MED15 domain deletion mutations were amplified from plasmid constructs2 (generous gift of Steve Hahn, Fred Hutchinson Cancer Center; See Suppl. Table 2) and subsequently transformed into either W303-1A or LRY037. Med15-13xMyc expressing strains were generated by homologous recombination with the Med15 C-terminal region PCR- amplified from ASK215 genomic DNA3. MED3–3×FLAG strains were generated by PCR amplification of the 3×FLAG cassette from pFA6a-3×FLAG using MED3-specific chimeric primers, followed by transformation into ASK218. MED15 domain deletions were subsequently introduced into this background. Degron-containing strains were generated by PCR amplification of the mini-AID*–9xMyc cassette from pHyg-AID*–9xMyc, pKan-AID*–9xMyc, or pNat-AID*–9xMyc (ref. 4) using chimeric primers homologous to the C-terminus of the target gene, followed by transformation into LRY016-derived strains5. For microscopy experiments, RPB3- mCherry::hphMX6 and MED15-mCherry::hphMX6 cassettes were amplified from SCY002 and SCY001, respectively3, and introduced into LRY037 to generate LRY039 and LRY040. Corresponding Mediator subunit deletion mutants (med2Δ, med3Δ, med5Δ, med15Δ and med16Δ) were subsequently constructed using LRY039 and LRY040 as parental strains. Yeast strains, plasmids and primers used in this study are listed in Suppl. Tables 1–3. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -26- CRISPR Editing The intrinsically disordered regions (IDRs) of Med2 (residues 290-400) and Med3 (residues 301-374) were deleted using a CRISPR–Cas9 strategy following protocols developed by the Tom Ellis lab, Imperial College London6. Single-guide RNAs (sgRNAs) targeting the Med2 and Med3 IDRs were designed using the Benchling sgRNA design tool. Forward and reverse complementary sgRNA oligonucleotides were phosphorylated, annealed to generate double-stranded DNA, and cloned into the guide RNA assembly vector pWS082 using BsmBI-based Golden Gate assembly. Correct sgRNA insertion disrupts the GFP cassette on the vector backbone, enabling selection of non-fluorescent colonies. Purified sgRNA plasmids, verified by Nanopore sequencing, were digested with EcoRV and gel purified. The Cas9–sgRNA repair vector pWS173 was linearized by BsmBI digestion and gel-purified prior to yeast transformation. Targeted deletion of the Med2 and Med3 IDRs was achieved by transforming strains W303-1A, LRY039 and LRY040 with a mixture of linearized Cas9–sgRNA repair vector, purified sgRNA-encoding DNA specific to Med2 or Med3, PCR-amplified donor DNA containing 100 bp homology arms upstream and downstream of the deleted IDR region, and salmon sperm DNA at a ratio of 100 ng: 200 ng: 3 μg: 50 μg, respectively. Pertinent sgRNA sequences, donor templates, primers, and plasmids are provided in Suppl. Tables 2 and 3. Yeast Culture and Heat Shock Treatment For molecular biology experiments, cells were cultivated in rich YPDA medium (1% yeast extract, 2% peptone, 2% dextrose, and 20 mg/L adenine) at 30°C until reaching mid-logarithmic growth (OD600 = 0.6-0.8). Thermal stress (heat shock (HS)) was applied by rapidly elevating the culture temperature from 25° to 39°C through addition of an equal volume of 55°C YPDA medium, followed by incubation at 39°C for designated time points. The non-heat-shock (NHS) samples were mixed with an equal volume of 25°C YPDA and maintained at 25°C. Samples were kept at their respective temperatures using a water bath with constant shaking. For experiments involving auxin-induced degradation, a stock of Indole-3-acetic acid (IAA, Sigma-Aldrich) was prepared in 100% ethanol to a final concentration of 100 mg/mL (570 mM). Cultures received IAA at a working concentration of 1 mM; incubation .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -27- was performed at the time points indicated in the figures and text. To optimize degradation, samples were kept at 25°C for up to 1 hr. Heat shock experiments were performed by incubating cells for a predetermined length of time with IAA, followed by a thermal upshift to 39°C by adding 55°C media supplemented with 1 mM IAA to maintain a constant IAA concentration. The NHS sample was used as a control; for this, an equivalent volume of 25°C medium containing 1 mM IAA was combined with the culture. Growth Assays For growth kinetics analysis, a single colony from either a wild-type or mutant strain was inoculated into 10 ml of sterile yeast peptone dextrose adenine (YPDA) medium and incubated for 16 hours at 30°C with orbital shaking to attain mid-exponential phase. Subsequently, cultures were diluted to OD600 = 0.2 and incubated at either 30°C or 39°C with constant orbital shaking (200 rpm) to assess temperature-dependent growth. OD600 measurements were taken at 1-h intervals for at least 8 h. Experiments were performed in biological duplicates, and mean OD600 values were calculated to construct growth curves illustrating the temporal relationship between optical density and time. Spot Dilution Assays Spot dilution assays were performed as previously described5. In brief, overnight cultures were standardized to OD600 = 0.5, and 10-fold or 5-fold serial dilutions were prepared. Aliquots (~4 μl) of each dilution were spotted onto YPDA plates using a 64- prong stainless steel replicator. Plates were incubated for 48 hours at 30°C or 72 hours at 24°C and 39°C to assess cell growth and viability. Chromatin Immunoprecipitation (ChIP) ChIP assays were performed as described previously7. Briefly, 50 ml of mid-log-phase cultures (OD600 = 0.6-0.8) were treated with 1% formaldehyde for cross-linking after heat shock at 39°C for the indicated time points. Unreacted formaldehyde was quenched with 2.5 M glycine (final concentration, 375 mM). Cells were pelleted, lysed, sonicated, and clarified. Chromatin lysates (500 μL aliquots from a total volume of 2000 μL) were then immunoprecipitated with 1.5 μL anti-Hsf1 antiserum8, 1.5 μL of anti-Rpb1 antiserum9 or 2.5 μL anti-Myc mAb (9E10; Santa Cruz Biotechnology, Inc.) for 16 h at 4°C with gentle rotation. Antibody-bound chromatin was captured using Protein A .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -28- Sepharose beads (GE Healthcare) by incubating 16 h at 4°C. DNA was subsequently purified using phenol-chloroform extraction. Locus-specific primers (Suppl. Table 4) were used to quantify ChIP DNA by quantitative PCR (qPCR) on a CFX384 Touch Real Time PCR System (Bio-Rad). Data was normalized to 20% input DNA, and the percentage of input was calculated accordingly. Fluorescence Microscopy and Image Analysis Hsf1-mNeonGreen (Hsf1-mNG) puncta imaging was performed as previously described1. Cells were grown in synthetic complete dextrose (SDC) medium supplemented with 0.1 mg ml⁻¹ adenine at 30°C to early log phase, then adhered to a VaHEAT substrate (Interherence GmbH, Germany) coated with 100 µg ml⁻¹ Concanavalin A (ConA; Sigma-Aldrich) in ddH₂O. After removal of unbound cells and medium, fresh SDC medium was added, and the chamber was sealed with a coverslip to minimize evaporation. For mAID strains, mid-log phase cells were pretreated with either vehicle control or auxin for 1 h prior to attachment onto ConA-coated VaHEAT substrates, followed by imaging at 25° and 39°C. Imaging was performed using an Olympus Yokogawa CSU-W1 spinning disk confocal system equipped with Hamamatsu Fusion sCMOS cameras and controlled by cellSens Dimension software. Z-stacks of 11 planes (0.5 µm spacing) were acquired at the indicated time points using 488 nm excitation (10% laser power, 200 ms exposure) for mNeonGreen-tagged proteins and 561 nm excitation (15% laser power, 200 ms exposure) for mCherry-tagged proteins. Hsf1–mNG puncta were quantified as described previously1. The percentage of cells containing Hsf1–mNG, Med15–mCherry, or Rpb3–mCherry foci (Figs. 2a,b, 6a,g) was determined using the “Cells” module in Imaris v10.2.0. This analysis included cells exhibiting ≥2 foci. During Imaris segmentation, the nuclear diameter and spot diameter were set to 2 µm and 0.48 µm, respectively10. For Hsf1/Med15 and Hsf1/Rpb3 colocalization analysis (Fig. 6 e, f, k, l), foci were detected using the Imaris “Spots” module with a spot diameter of 0.48 µm. Colocalization was assessed by measuring the shortest volume-to-volume distance between spots, and foci were considered colocalized when this distance was ≤0.45 µm. Image cropping and 3D reconstruction were performed using FIJI/ImageJ (v1.53t) 11. All quantified data were plotted using GraphPad Prism v10. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -29- Taq I Chromosome Conformation Capture (Taq I - 3C) Taq I - 3C was performed as previously described5,12. Briefly, samples were crosslinked with 1% formaldehyde (15 min), quenched with 0.363 M glycine (10 min), and lysed for 40 min using glass beads in FA Lysis Buffer. Lysates were aliquoted and subjected to Taq I digestion (20 U/μL, 7 h, 60°C) and Quick T4 DNA ligation (10,000 cohesive units, 2 h, 25°C). Samples were then treated with RNase A (30 ng/μL, 40 min, 37°C) and Proteinase K (70 ng/μL, 16 h, 65°C), followed by phenol-chloroform extraction, ethanol precipitation, and qPCR quantification using locus-specific primers (Suppl. Table 5). Reverse Transcription-qPCR (RT-qPCR) RT-qPCR was performed as described5 using 25 mL cell culture aliquots. Cells were metabolically arrested and lysed using glass beads and acidic phenol-chloroform, followed by RNA extraction using acidic phenol-chloroform. Residual DNA was removed (Turbo DNA-free kit), and cDNA was synthesized (High-Capacity cDNA Reverse Transcription Kit, both from Applied Biosystems). cDNA was quantified on a CFX384 System using gene-specific primers (Suppl. Table 6) and a standard curve generated from genomic DNA. Quantified DNA was normalized to SCR1. Immunoblot Analysis Immunoblot analysis was performed as previously described5. Briefly, log-phase cells were metabolically arrested with 20 mM sodium azide for 30 seconds, lysed, and proteins precipitated with trichloroacetic acid. Protein pellets were dried out and resuspended in Thorner buffer (8 M urea, 5% SDS, 40 mM Tris-HCl pH 6.8, 0.1 mM EDTA, 0.4 mg/mL Bromophenol Blue, 1% 2-Mercaptoethanol). Samples were neutralized with 2 M Tris base, incubated at 42°C for 15 minutes, and separated on SDS 4-20% polyacrylamide TGX precast gels (Bio-Rad). Proteins were transferred to PVDF membranes (Amersham, 0.2 μm) (or nitrocellulose, Bio-Rad, for AID experiments), blocked with 5% nonfat milk in TBST (10 mM Tris-HCl, pH 8, 150 mM NaCl, 0.05% Tween20) overnight at 4°C, and probed with primary antibodies (Myc, 1:1000, Santa Cruz sc-40; H3, 1:1000, Abcam ab1719; Pgk1, 1:10,000, ThermoFisher 459250) and secondary HRP-conjugated antibodies (1:5000, Santa Cruz) for 1 h. Protein bands were detected using Enhanced Chemiluminescence (ECL) reagent (Thermo Scientific 34580). .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -30- Intrinsic Disorder, Structural Confidence and Phase Separation Analyses Protein sequences were retrieved in FASTA format and analyzed using complementary sequence-based and machine-learning approaches to quantify intrinsic disorder and liquid–liquid phase separation (LLPS) propensity. Intrinsic disorder was predicted using IUPred3 (long mode)13 and PONDR (VSL2)14 to generate per-residue disorder probabilities; residues with scores ≥0.5 were classified as disordered. The fraction of disordered residues per protein was calculated to determine the intrinsic disorder region (IDR) abundance. LLPS propensity was evaluated using FuzDrop (probability of spontaneous LLPS, pLLPS) 15 and PhasePred16, extracting PS-self and PS-part scores for each protein. All the graphs were generated using GraphPad Prism (v10). Structural confidence profiles were obtained from AlphaFold317 models, and per-residue predicted Local Distance Difference Test (pLDDT) scores were used as an orthogonal measure of structural order, with regions of pLDDT <70 considered indicative of structural flexibility/disorder. Intra-molecular interaction maps were generated using the FINCHES online18 tool (Mpipi-GG method; window size = 31). Amino acid composition was calculated using ExPASy ProtParam19. Default parameters were used for all analyses unless otherwise specified. Statistical Analysis Statistical analyses were performed using GraphPad Prism, as specified in the figure legends. Differences between individual means were assessed using two-tailed, unpaired t tests assuming equal variance. For violin plot analyses, statistical significance was determined by one-way ANOVA.

1. Rubio, L.S., Mohajan, S., and Gross, D.S. (2024). Heat Shock Factor 1 forms nuclear condensates and restructures the yeast genome before activating target genes. Elife 12. ARTN RP9246410.7554/eLife.92464. 2. Herbig, E., Warfield, L., Fish, L., Fishburn, J., Knutson, B.A., Moorefield, B., Pacheco, D., and Hahn, S. (2010). Mechanism of Mediator recruitment by tandem Gcn4 activation domains and three Gal11 activator-binding domains. Mol Cell Biol 30, 2376–2390. 10.1128/MCB.01046-09. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -31- 3. Chowdhary, S., Kainth, A.S., Paracha, S., Gross, D.S., and Pincus, D. (2022). Inducible transcriptional condensates drive 3D genome reorganization in the heat shock response. Mol Cell 82, 4386–4399 e4387. 10.1016/j.molcel.2022.10.013. 4. Morawska, M., and Ulrich, H.D. (2013). An expanded tool kit for the auxin-inducible degron system in budding yeast. Yeast 30, 341–351. 10.1002/yea.2967. 5. Rubio, L.S., and Gross, D.S. (2023). Dynamic coalescence of yeast Heat Shock Protein genes bypasses the requirement for actin. Genetics 223. 10.1093/genetics/iyad006. 6. Shaw, W.M., Yamauchi, H., Mead, J., Gowers, G.F ., Bell, D.J., Oling, D., Larsson, N., Wigglesworth, M., Ladds, G., and Ellis, T. (2019). Engineering a Model Cell for Rational Tuning of GPCR Signaling. Cell 177, 782–796 e727. 10.1016/j.cell.2019.02.023. 7. Chowdhary, S., Kainth, A.S., Pincus, D., and Gross, D.S. (2019). Heat Shock Factor 1 Drives Intergenic Association of Its Target Gene Loci upon Heat Shock. Cell Rep 26, 18–28 e15. 10.1016/j.celrep.2018.12.034. 8. Venturi, C.B., Erkine, A.M., and Gross, D.S. (2000). Cell cycle-dependent binding of yeast heat shock factor to nucleosomes. Mol Cell Biol 20, 6435–6448. 10.1128/MCB.20.17.6435-6448.2000. 9. Zhao, J., Herrera-Diaz, J., and Gross, D.S. (2005). Domain-wide displacement of histones by activated heat shock factor occurs independently of Swi/Snf and is not correlated with RNA polymerase II density. Mol Cell Biol 25, 8985–8999. 10.1128/MCB.25.20.8985-8999.2005. 10. Mohajan, S., Rubio, L.S., and Gross, D.S. (2025). Nuclear basket proteins Nup2 and Mlp1 drive heat shock-induced 3D genome restructuring downstream of transcriptional activation. J Biol Chem 301, 110568. 10.1016/j.jbc.2025.110568. 11. Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V ., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B., et al. (2012). Fiji: an open-source platform for biological-image analysis. Nat Methods 9, 676–682. 10.1038/nmeth.2019. 12. Chowdhary, S., Kainth, A.S., and Gross, D.S. (2020). Chromosome conformation capture that detects novel cis- and trans-interactions in budding yeast. Methods 170, 4–16. 10.1016/j.ymeth.2019.06.023. 13. Erdos, G., Pajkos, M., and Dosztanyi, Z. (2021). IUPred3: prediction of protein disorder enhanced with unambiguous experimental annotation and visualization of evolutionary conservation. Nucleic Acids Res 49, W297–W303. 10.1093/nar/gkab408. 14. Xue, B., Dunbrack, R.L., Williams, R.W., Dunker, A.K., and Uversky, V .N. (2010). PONDR-FIT: a meta-predictor of intrinsically disordered amino acids. Biochim Biophys Acta 1804, 996–1010. 10.1016/j.bbapap.2010.01.011. 15. Hardenberg, M., Horvath, A., Ambrus, V ., Fuxreiter, M., and Vendruscolo, M. (2020). Widespread occurrence of the droplet state of proteins in the human proteome. Proc Natl Acad Sci U S A 117, 33254–33262. 10.1073/pnas.2007670117. 16. Chen, Z., Hou, C., Wang, L., Yu, C., Chen, T., Shen, B., Hou, Y ., Li, P ., and Li, T. (2022). Screening membraneless organelle participants with machine-learning models that integrate multimodal features. Proc Natl Acad Sci U S A 119, e2115369119. 10.1073/pnas.2115369119. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Male et al -32- 17. Abramson, J., Adler, J., Dunger, J., Evans, R., Green, T., Pritzel, A., Ronneberger, O., Willmore, L., Ballard, A.J., Bambrick, J., et al. (2024). Addendum: Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 636, E4. 10.1038/s41586-024-08416-7. 18. Lotthammer, J.M., Ginell, G.M., Griffith, D., Emenecker, R.J., and Holehouse, A.S. (2024). Direct prediction of intrinsically disordered protein conformational properties from sequence. Nat Methods 21, 465–476. 10.1038/s41592-023-02159- 5. 19. Wilkins, M.R., Gasteiger, E., Bairoch, A., Sanchez, J.C., Williams, K.L., Appel, R.D., and Hochstrasser, D.F . (1999). Protein identification and analysis tools in the ExPASy server. Methods Mol Biol 112, 531–552. 10.1385/1-59259-584-7:531. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint a c Med2 Med3 Med15 Med5 Med16 0 1 e f d b Order Disorder Med2Med3Med15Med5Med16 0 25 50 75 100 PONDR score Med2Med3Med15Med5Med16 0 25 50 75 100 IUPred3 score FuzDrop prediction g Residue position 0 200 400 600 800 1000 0.0 0.5 1.0 Med15 0 100 200 300 400 0.0 0.5 1.0 Med3 0 200 400 600 800 1000 0.0 0.5 1.0 Med5 0 200 400 600 800 1000 0.0 0.5 1.0 Med16 Disorder score 0 100 200 300 400 0.0 0.5 1.0 Med2 PONDR IUPred3 Med3 Med5 Med15 Med16 Med2 Tail module Very high (pLDDT >90) Confident (90 > pLDDT > 70) Low (70 > pLDDT > 50) Very low (pLDDT < 50) Figure 1 Order Disorder % of Polypeptide% of Polypeptide 0 0.5 1.0 1.5 0 0.2 0.4 0.6 0.8 1.0 Med2 Med3 Med15 Med5 Med16 Triad PS-part score PS-self score Phase separation tendency 0 20 40 60 80 100 0 0.2 0.4 0.6 0.8 1.0 Med2 Med3 Med15 Med5 Med16 R2=0.9515 PS-part score % of IDR Phase separation correlation .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Figure 1 | Mediator Tail subunits Med2, Med3, and Med15 exhibit high intrinsic disorder and a strong predicted propensity for liquid–liquid phase separation. a, Schematic of the Saccharomyces cerevisiae Mediator complex illustrating its modular organization and subunit composition (adapted from Philip et al., eLife, 2015 (ref.5). b, Intrinsic disorder predictions for the indicated Mediator Tail subunits based on PONDR and IUPred3 analyses; residues with prediction scores ≥0.5 are classified as disordered. c, Comparative disorder prediction profiles of the indicated Mediator Tail subunits generated using IUPred3 (magenta) and PONDR VSL2 (blue). d, Phase separation propensity of Mediator Tail subunits predicted by PhaSePred, shown as PS-part and PS-self scores; the Mediator Tail subunit triad (Med2, Med3, and Med15) is highlighted (dotted box). e, Correlation between PS-part scores (PhaSePred) and the proportion of intrinsically disordered residues (PONDR) across the indicated Mediator subunits. f, Heatmap of predicted liquid–liquid phase separation probabilities for Mediator Tail subunits calculated using FuzDrop, with darker shading indicating higher propensity. g, Predicted structures of individual Mediator Tail subunits and the assembled Tail module generated using AlphaFold3. pLDDT, Predicted Local Distance Difference Test. . .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint a dc Imaris 3D- models Hsf1- mNeonGreen HS time points 0’ 5’ 10’ 60’ Hsf1- mNeonGreen HS time points Imaris 3D- models 0’ 5’ 10’ 60’ med16∆ Hsf1- mNeonGreen HS time points Imaris 3D- models b 0’ 5’ 10’ 60’ WT 0’ 5’ 10’ 60’ med2∆ med15∆ med3∆ 0’ 5’ 10’ 60’ med5∆ 0’ 5’ 10’ 60’ Figure 2 WT Med15-mAID Med16-mAID MED15-mAID MED16-mAID 0’ 5’ 10’ 30’ HS time points WT med2∆ med3∆ med5∆ med15∆ med16∆ 0 2.5 5 10 15 30 60 0 20 40 60 80 100 % of cells with Hsf1-puncta 0 5 10 15 30 0 20 40 60 80 100 HS time points (in min) 0’ 5’ 10’ 30’ Hsf1- mNeonGreen Imaris 3D- models HS time points (in min) Gene Deletions Conditional Depletions .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Figure 2 | Mediator Tail subunits regulate Hsf1 condensate kinetics during heat shock. a, Representative fluorescence micrographs and corresponding three-dimensional reconstructions (Imaris v10.2.0, Cells module) of Hsf1 condensates in the indicated wild-type (WT) and deletion strains during a 39°C heat shock time course. 0 min, cells maintained at 25°C. Scale bars, 1 µm; for Med15-mAID and Med16-mAID, 2 µm. Scale bar for Imaris-rendered 3D models, 0.5 µm. b, As in a, except cells expressing the indicated mAID-tagged proteins were imaged. Prior to initiating heat shock, cells were treated with 1 mM IAA for 60 min at 25°C. Scale bars, 2 µm. c,d, Quantification of the fraction of cells containing two or more Hsf1 condensates (n ≥ 2) during heat shock. Data represent the mean of two independent biological replicates; error bars indicate s.d. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint WT Med15-mAID Med16-mAID 0 20 40 60 80 100 120 ✱✱✱ ✱ WT med5Δ med2Δ med3Δ med15Δ med16Δ 0 20 40 60 80 100 120 ✱✱✱ ns a b c UAS-Promoter Intragenic interactions Interchromosomal interactions Intrachromosomal interactions WT med5Δ med2Δ med3Δ med15Δ med16Δ 0 20 40 60 80 100 120 ✱✱✱ WT med5Δ med16Δ med2Δ med3Δ med15Δ d MED Gene deletions No protein MED mAID MED mAID +Auxin Protein degradation mRNA levels 0 20 0.0 0.2 0.4 0.6 ns ** *** *** ns ** Figure 3 Normalized Interaction Frequency HSP104 F+1550/ SSA2 F+1368 WT Med15-mAID Med16-mAID0 20 40 60 80 100 120 ✱✱✱ HSP104 F+1550/ SSA2 F+1368 HSP104 F+1550/ HSP82 F+740 Normalized Interaction Frequency HSP104 F+1550/ HSP82 F+740 SSA4 F-268/ F+198 Promoter-Terminator WT med5 Δ med2 Δ med3 Δ med15Δmed16Δ 0 20 40 60 80 100 120 ✱✱✱ ** SSA4 F+198/ F+2255 WT med5Δmed2Δmed3Δmed15Δmed16Δ 0 20 40 60 80 100 120 ✱✱✱ ns Normalized Interaction Frequency Relative mRNA levels SSA4 HS time points (in min) .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Figure 3 | Mediator Tail subunits drive heat-shock-induced HSR gene interactions and promote HSR mRNA abundance. a, Schematic overview of the orthogonal experimental strategies—gene deletion and conditional depletion—used to assess the role of Mediator Tail subunits in Heat Shock Response (HSR) gene interactions. b, c, Normalized interaction frequencies of HSR genes in the indicated strains subjected to a 5 min heat shock at 39°C. b shows representative cis- and trans- intergenic interactions, and c shows intragenic interactions. Pairwise tests used forward (F; sense-strand) primers positioned nearby the indicated Taq I site; primer locations are illustrated in Fig. S1 of Mohajan et al (ref. 33). Data are presented as mean ± s.d.; N = 2-8 independent samples (as indicated). Statistical significance relative to WT was determined using an unpaired, two-tailed t-test (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant). Location of gene loci: HSP104 and SSA2, Chr XII; HSP82, Chr XVI; HSP12, Chr VI; SSA4, Chr V. d, Relative SSA4 mRNA abundance measured by RT–qPCR. Data represent mean ± s.d. from the indicated number of independent biological replicates. Statistical significance relative to WT was assessed as above. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint a b 0 2.5 5 10 15 30 60 0 20 40 60 80 100 WT med15∆ ∆kix ∆abd1 ∆abd2 ∆abd3 Intrachromosomal interactions Interchromosomal interactions WT Δkix Δabd1Δabd2Δabd3 med15Δ 0 20 40 60 80 100 120 140 ns ns *** *** ns UAS-Promoter Intragenic interactions WT Δkix Δabd1Δabd2Δabd3 med15Δ 0 20 40 60 80 100 120 ns ns ** ** ns d WT Δkix Δabd1 Δabd3 Δabd2 med15Δ 0 2 4 6 8 * *** ns ns *** Disordered Ordered 0 0.5 1 1 101 201 301 401 501 601 701 801 901 1001 13 80 116 255 277 404 418 696 Med15 1081 Disorder score WT ∆kix ∆abd1 ∆abd2 ∆abd3 med15∆* mRNA levels c e f 0 5 0 10 20 30 40 ns * ns *** *** * * ** 0 5 0.0 0.5 1.0 1.5 2.0 ns * * * 0 5 0.0 0.5 1.0 1.5 ns * p=0.091 p=0.058 Chromatin Immunoprecipitation 0 5 0 2 4 6 8 10 ns ** ** *** *** ** WT Δkix Δabd1Δabd2Δabd3 med15Δ 0 25 50 75 100 125 150 ns ✱✱✱ Normalized Interaction Frequency Figure 4 % of cells with Hsf1-puncta HS time points (in min) 10 min HS Hsf1-puncta per cell UBI4 F+524/ SSA2 F+1368 SSA4 F-268/ SSA4 F+198HSP104 F+1550/ HSP12 F-47 % of Input Hsf1 at HSP104 UAS % of Input Relative mRNA levels 0 20 0.00 0.05 0.10 0.15 * **ns 0 20 0.00 0.05 0.10 0.15 0.20 ** ns ** Pol II at HSP104 ORF Med17 at HSP104 UAS Med3 at HSP104 UAS HSP104 SSA4 Relative mRNA levels HS time points KIX ABD1 ABD2 ABD3 min min .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Figure 4 | The Med15 ABD2 domain is required for Hsf1 condensate formation and heat-shock-induced HSR gene interactions. a, Schematic of Med15 domain architecture and deletion constructs, with the domains deleted indicated. Predicted intrinsic disorder across Med15 was assessed by IUPred, with amino acid positions shown. b, Quantification of cells containing two or more Hsf1 condensates in WT and Med15 domain deletion strains during heat shock using Imaris v10.1.0 (Cells module). Data represent the mean of two independent biological replicates; error bars indicate s.d. c, Violin plot showing the distribution of Hsf1 puncta per cell in WT and med15 mutants after 10 min of heat shock, quantified as in b. Horizontal lines indicate median values; dashed lines indicate quartiles. Statistical significance relative to WT was assessed by one-way ANOVA (*, P < 0.05; **, P < 0.01; ***, P < 0.001). d, Normalized interaction frequencies of Heat Shock Response (HSR) genes measured by Taq I - 3C in the indicated strains after 5 min HS, showing intrachromosomal (cis), interchromosomal (trans), and intergenic interactions. Data represent mean ± s.d. from two independent biological replicates. Statistical significance relative to WT was determined using unpaired two-tailed t-test (*, P < 0.05; **, P < 0.01; ***, P < 0.001). e, ChIP analysis of Hsf1, Pol II large subunit Rpb1, and Mediator subunits Med17 (Head) and Med3 (Tail) at the HSP104 locus in the indicated strains. Data represent mean ± s.d. from two independent biological replicates. Statistical significance relative to WT was assessed as in d. f, Relative mRNA levels of HSP104 and SSA4 measured by RT–qPCR at the indicated heat-shock time points in the indicated strains. Data represent mean ± s.d. from two independent biological replicates. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint a b d Intrachromosomal interactions Interchromosomal interactions UAS-Promoter Intragenic interactions WT med2 Δ med2- ΔIDR med3 Δ med3- ΔIDR 0 20 40 60 80 100 120 140 ns ns ✱ WT med2 Δ med2- ΔIDR med3 Δ med3- ΔIDR 0 20 40 60 80 100 120 ns ns ✱✱✱ WT med2 Δ med2- ΔIDR med3 Δ med3- ΔIDR 0 20 40 60 80 100 120 ns ns ✱✱✱ 0 5 10 15 30 60 0 20 40 60 80 100 WT med2∆ med2-∆IDR med3∆ med3-∆IDR c WT med2Δ med2-ΔIDR med3Δ med3-ΔIDR 0 2 4 6 8 ✱✱✱ ✱ ✱ f e 0 5 0 2 4 6 8 ns p=0.067 p=0.068 0 5 0.0 0.5 1.0 1.5 2.0 ns ** ** Chromatin immunoprecipitation WT med2Δ med2-ΔIDR 0 20 0.0 0.1 0.2 0.3 0.4 ns * * ** ns 0 20 0.0 0.2 0.4 0.6 WT med2Δ med2-∆IDR med3Δ med3-ΔI DR ns ** ** *** * mRNA levels HS time points Figure 5 % of cells with Hsf1-puncta HS time points (in min) Hsf1-puncta per cell HSP104 F+1550/ SSA4 F+198 Normalized Interaction Frequency HSP104 F+1550/ SSA2 F+1368 SSA4 F-268/ SSA4 F+198 0 5 0 10 20 30 40 ns ** * Hsf1 at SSA4 UAS Pol II at SSA4 ORF Med15 at HSP104 UAS % of InputRelative mRNA levels 0 100 200 300 400 0.0 0.5 1.0 290-400 1 105 431269 381140 N-enrichedCharged Med2 0 100 200 300 400 0.0 0.5 1.0 301-374 Disorder score Med3 HSP104 SSA4 1 225 304 374236 Q, N-enrichedQA 398 min min *** 10 minutes HS .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Figure 5 | Med2 and Med3 C-terminal IDRs regulate the heat-shock response by modulating Hsf1 clustering, genome topology, and Hsf1, Mediator and Pol II occupancy. a, Domain architecture of Med2 and Med3 with accompanying IUPred3 analysis. Dashed lines indicate locations of CRISPR-Cas9–targeted deletions of the C- terminal IDRs. b, Hsf1 puncta formation in WT and Med2 or Med3 gene / domain deletion strains during a heat shock time course. c, Violin plot showing the distribution of Hsf1 puncta per cell after a 10 min heat shock, quantified using Imaris. Solid horizontal lines indicate median values. Discontinuous lines depict quartiles. Statistical significance relative to WT was determined by one-way ANOVA. *, P < 0.05; **, P < 0.01; ***P < 0.001. d, Normalized interaction frequencies of Heat Shock Response (HSR) genes measured by Taq I - 3C in the indicated strains after 5 min HS, showing intrachromosomal (cis), interchromosomal (trans), and intergenic interactions. Data represent mean ± s.d. (n = 2). Statistical significance relative to WT was determined using an unpaired two-tailed t-test (*, P < 0.05; **, P < 0.01; ***P < 0.001). e, ChIP analysis of Hsf1, Rpb1 and Med15 at representative HSR genes in the indicated strains. Data represent mean ±s.d. (n = 2). f, Relative HSP104 and SSA4 mRNA levels measured by RT–qPCR in the indicated strains following heat shock. Data represent mean ± s.d. from independent biological replicates (n=2). Statistical significance relative to WT was determined using an unpaired two-tailed t-test (*, P < 0.05; **, P < 0.01; ***P < 0.001). .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint a Figure 6b c WT med3Δ med5Δ med16Δ 0 20 40 60 80 100 Hsf1 Med15 ns ns ns ✱ WT med5Δ med16Δ 0 10 20 30 40 50 60 70 ✱ ns WTmed3∆med5∆med16∆ 10’ HS Hsf1 Med15 Merge WT med2Δ med5Δ med16Δ 0 20 40 60 80 100 Hsf1 Rpb3 ns ns ns ✱✱ WT med5Δ med16Δ 0 10 20 30 40 50 60 70 ✱✱ ns WT med5Δ med16Δ 0 2 4 6 ns *** WT med5Δ med16Δ 0 2 4 6 *** *** d e f g h i j 10’ HS Hsf1 Med15 Merge 10’ HS Hsf1 Rpb3 Merge Merge 10’ HS Hsf1 Rpb3 med16∆ WTmed2∆med5∆ k l WT med3Δ med5Δ med16Δ 0 2 4 6 8 10 12 Hsf1 Med15 ns ns ✱✱✱ ✱✱✱ WT med2Δ med5Δ med16Δ 0 2 4 6 8 10 Hsf1 Rbp3 ns ns ns ✱✱✱ Puncta per cell % of colocalized Hsf1 puncta % of cells with puncta shortest distance (µm) Hsf1 foci to Med15 foci Puncta per cell % of cells with puncta % of colocalized Hsf1 puncta shortest distance (µm) Hsf1 foci to Rpb3 foci .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Figure 6 | Med16 is required for HS-induced formation of Mediator and Pol II condensates. a, g, Representative live-cell fluorescence images of the indicated strains expressing Hsf1–mNG, Med15–mCherry and Rpb3–mCherry, marking Hsf1, Mediator, and Pol II, respectively, after 10 min HS. Scale bars, 2 µm. b, h, Representative 3D reconstructions (Imaris Spots module) corresponding to a and g, showing Hsf1–Med15 (b) and Hsf1–Pol II (h) condensates in WT, med5Δ, and med16Δ cells after 10 min HS. Foci were detected using a fixed diameter of 0.48 µm. Light shading denotes Hsf1 puncta colocalized with Med15 (b) or Rpb3 (h). Scale bars, 0.5 µm. c, i, Percentage of cells containing ≥2 Hsf1–Mediator (c) or Hsf1–Pol II (i) co- condensates after 10 min HS. Data represent mean ± s.d. from two independent biological replicates. d, j, Violin plots showing the number of Hsf1, Mediator (d), and Hsf1, Rpb3 (j) puncta per cell after 10 min of heat shock, quantified using Imaris Cells module. Data from two independent biological replicates are shown. Horizontal lines indicate median values. e, k, Fraction of Hsf1 puncta colocalized with Med15 (e) or Pol II (k), quantified from 3D reconstructions using Imaris Spots module. Approximately 100 cells per strain were analyzed across two independent biological replicates. Data are mean ± s.d. f, l, Violin plots showing the distribution of the shortest volume-to-volume distances between Hsf1 and Med15 foci (f) or Hsf1 and Rpb3 foci (l) after 10 min HS in the indicated strains. Center lines indicate medians. Statistical significance was determined using a nonparametric two-tailed test for comparisons in c, e, i, k and one-way ANOVA for d, f, j, l. *, P < 0.05, **, P < 0.01, ***, P < 0.001. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint HS Hsf1 Pol II Mediator HSR Functional condensates Med2/ Med3/ Med15 (IDR enriched) Med16 Med5 Moderate transcription Robust transcription HSR gene coalescence Condensate assembly HSR Hierarchical Roles Formation of Clusters Figure 7 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint Figure 7 | Hierarchical model of Mediator Tail subunit regulation of the heat shock response. Upon acute heat shock (HS), Hsf1 trimerizes and, together with Mediator, Pol II and associated factors, is recruited to the upstream regulatory regions of HSR genes, activating their transcription. Hsf1, Mediator, and Pol II rapidly assemble into discrete intranuclear clusters and do so in a kinetically indistinguishable manner. Genetic analysis suggests that these subsequently mature into functional condensates that promote intergenic interactions among HSR gene loci, accompanied by enhanced levels of transcription. Within this framework, Mediator Tail subunits act hierarchically: the IDR–rich subunits Med2, Med3, and Med15 are essential for all steps; Med16 facilitates transcriptional condensate maturation and HSR gene activation and is essential for the global 3D restructuring of HSR genes; Med5 modestly enhances HSR interactions and HSR transcriptional output. Large multicolored arrows indicate the steps at which the indicated Tail subunits contribute. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 22, 2026. ; https://doi.org/10.64898/2026.04.21.719956doi: bioRxiv preprint