Abstract

   Transcriptional coactivators Yes-associated protein (YAP) and
   transcriptional coactivator with PDZ-binding motif (TAZ) play key roles
   in cancers through transcriptional outputs. However, their
   transactivation mechanisms remain unclear, and effective targeting
   strategies are lacking. Here, we show that YAP/TAZ possess a
   hydrophobic transactivation domain (TAD). TAD knockout prevents tumor
   establishment due to growth defects and enhances immune attack.
   Mechanistically, TADs facilitate preinitiation complex (PIC) assembly
   by recruiting the TATA-binding protein-associated factor 4
   (TAF4)-dependent TFIID complex and enhance RNA polymerase II (Pol II)
   elongation through mediator complex subunit 15 (MED15)-dependent
   mediator recruitment for the expressions of
   oncogenic/immune-suppressive programs. The synthesized peptide TJ-M11
   selectively disrupts TAD interactions with MED15 and TAF4, suppressing
   tumor growth and sensitizing tumors to immunotherapy. Our findings
   demonstrate that YAP/TAZ TADs exhibit dual functions in PIC assembly
   and Pol II elongation via hydrophobic interactions, which represent
   actionable targets for cancer therapy and combination immunotherapy.

   Subject terms: Oncogenes, Transcription
     __________________________________________________________________

   The mechanism underlying YAP/TAZ transactivation remains to be
   explored. Here the authors report that the hydrophobic transcriptional
   activation domains of YAP/TAZ exert dual functions on transcriptional
   initiation and elongation through interactions with TAF4 and MED15,
   which could be targeted for anti-cancer therapy.

Introduction

   YAP/TAZ, a pair of paralogs of potent transcriptional coactivators,
   function downstream of the Hippo pathway in cancer initiation,
   progression, and immunotherapy resistance by promoting the
   transcription of target genes transcription^[44]1–[45]7. Compared to
   their functions in covalent and noncovalent epigenetic
   regulation^[46]8–[47]11, the mechanism and in vivo function of YAP/TAZ
   to initiate transcription is not well known. Like other transcription
   events, the specific binding of YAP/TAZ to DNA through TEA domain
   (TEAD) family transcription factors (TFs) has been well documented,
   whereas how their intrinsically disordered regions (IDRs) of TAD
   promoting transactivation has been unresolved questions^[48]12. There
   are over 2000 TFs with hundreds of TADs in human genome^[49]13,[50]14
   and these TADs exhibited obvious primary sequences variation and
   functional conservation through interacting with only a few cofactors
   for driving transcription. The mechanism underlying TAD and cofactors
   interactions driving transcriptions needs to be addressed^[51]15. Thus,
   dissecting the structure and function of YAP/TAZ-TADs will facilitate
   the understanding of the transcription output of YAP/TAZ as well as
   provide potential targeting strategies.

   YAP/TAZ are upregulated in numerous human cancers and exhibit strong
   multipotency in autonomous tumor cell growth, non-autonomous immune
   suppression, microenvironment regulation, and immunotherapy
   resistance^[52]1,[53]6,[54]16. Because of the key function of YAP/TAZ
   in solid tumors, targeting them for clinical applications is
   striking^[55]17. Previous studies have focused mainly on the
   interaction between TEAD and YAP^[56]18. Several strategies have been
   developed to identify small molecules and peptides that block TEAD-YAP
   interactions^[57]19,[58]20. However, to date, no drugs that inhibit YAP
   have been clinically used because of their limited activity or
   cytotoxicity. The prevention of transcriptional activation by small
   peptides or chemicals has recently shown good efficacy against some key
   onco-transcription factors^[59]21,[60]22. Thus, targeting the
   interactions between intrinsically disordered TFs and cofactors is a
   potential field for drug development. However, challenges remain due to
   the dynamic interface of these interactions^[61]23.

   Here, we show the structure and function of YAP and TAZ transactivation
   domains. They exhibit dual functions during transcriptional activation
   including initiation and elongation. TADs assemble the pre-initiation
   complex for Pol II loading by interacting with TAF4 and promote Pol II
   efficient elongations via recruiting the mediator complex. Both TFIID
   and the Mediator complex bind to transcriptional activation domains
   (TADs) through hydrophobic interactions involving an α-helix and the
   linear FLTWL motif. This interaction promotes the folding of the IDR,
   facilitating the formation of transcriptional condensates via
   co-condensation. Knockout of TADs of YAP/TAZ leads to tumor growth
   defects and enhances immune response due to reducing oncogenic and
   immunosuppressive transcriptional program expression. Dual targeting of
   the interaction between the TADs of YAP/TAZ and the cofactors TAF4 and
   MED15 by small peptides mimicking FLTWL inhibits their transactivation
   activity. This suppresses tumor growth and enhances the efficacy of
   immunotherapy, supporting the strategy of targeting the transactivation
   domains of YAP/TAZ for cancer therapy.

Results

Structural characterization of TADs of YAP and TAZ

   Our previous studies identified a 55-amino acid TAD at the C-terminus
   of YAP^[62]24. To determine whether TAZ has a similar TAD, we performed
   protein sequence alignments and evaluated their activities using
   Gal4-UAS-luciferase reporter assays (Fig. [63]1a). Similarly, the final
   35 residues (366-400) at the C-terminus of TAZ could initiate reporter
   gene transcription, although with only one-third of the activity seen
   in YAP-TAD (Fig. [64]1b). Notably, both TADs were composed of over 50%
   hydrophobic and ~20% acidic amino acids (Fig. [65]1c). Interestingly,
   at the termini of these TADs is a five-residue linear hydrophobic motif
   FLTWL, previously identified as the PDZ binding domain^[66]25.
   Hydrophilic replacement with four serine residues (S) disrupted the
   ability to drive the transcription of cellular target genes
   (Fig. [67]1d).

Fig. 1. Structural characterization of TADs of YAP and TAZ.

   [68]Fig. 1
   [69]Open in a new tab

   a Diagram of the conserved domains of YAP/TAZ and the
   Gal4-UAS-Luciferase reporter system for the TAD activity test. (Pro:
   Proline, CC: Coiled coil domain, WW: WW domain, sIDR: small Intrinsic
   Disorder Region. TAD: Transactivation Domain, 40×UAS: 40 repeats of
   Upstream Activating Sequence). b Luciferase activity of HEK293T cells
   co-transfected with the indicated plasmids and 40× UAS-luciferase
   reporters (n = 3 biological replicates). c Sequence alignment of YAP
   and TAZ TADs. Acidic amino acids are shown in red, and hydrophobic
   amino acids are shown in green. d RT-qPCR shows YAP target genes in
   HEK293T cell transfected with YAP, YAPΔTAD, or F500L501W503L504
   mutating to S (n = 3 biological replicates). e Assignments of backbone
   amide groups of YAP-TAD and NMR structure ensemble of YAP-TAD
   consisting of the 20 lowest-energy models. The α-helix is shown in pink
   and the intrinsically disordered region in gray. f Assignments of
   backbone amide groups of TAZ-TAD and NMR structure ensemble of TAZ-TAD
   consisting of the 20 lowest-energy models. g Predicted structural model
   of the α-helix of YAP-TAD by Alphafold2. Hydrophilic (blue) and
   hydrophobic (red) residues are indicated. h RT-qPCR analysis of YAP
   target genes in HEK293T cells transfected with YAP, YAPΔTAD,
   L483M486L490 to S, I482M486V489 to S, or N484S488 to L (n = 3
   biological replicates). Data are presented as mean ± standard error of
   the mean (SEM). Statistical analyses were performed using one-way ANOVA
   (d, h).

   To dissect the structures of both TADs, we performed nuclear magnetic
   resonance (NMR) analysis (Fig. [70]1e, [71]f) on highly purified
   protein samples expressed and purified from E. coli (Supplementary
   Fig. [72]1a, [73]b). The solution structure of YAP-TAD has been
   resolved and submitted to Protein Data Bank (PDB) under accession
   number 8WRG. Our results revealed that the TADs were predominantly
   composed of IDRs (Fig. [74]1e, f and Supplementary Fig. [75]1c–f).
   Compared to the TAD of TAZ, the TAD of YAP exhibited a short
   amphipathic α-helix spanning seven amino acids, with Alphafold2
   predictions suggesting a potential extension of up to 17 residues
   (Fig. [76]1g). Interestingly, substituting residues asparagine (N) and
   serine (S) on the hydrophilic surface with hydrophobic residue leucine
   (L) enhanced transcriptional activation. Conversely, replacing
   hydrophobic residues on the hydrophobic surface with serine (S) reduced
   transcriptional activity (Fig. [77]1h and Supplementary Fig. [78]1g).
   These results indicate that YAP possesses a hydrophobic TAD, and the
   interactions between the amphipathic helix and the FLTWL motif with TFs
   are crucial for YAP over-expression induced gene activation.

The TAD of YAP initiates immunosuppressive transcriptional program

   Next, we evaluated the functional significance of TADs of YAP/TAZ in
   transcriptional promotion and tumor growth using a
   Kras^LSLG12D;Trp53^f/f mouse lung tumor-derived cell line (TDCL)^[79]26
   (Supplementary Fig. [80]2a). To this end, we knocked out the TAD of YAP
   in TAZ ^-/- background by generating a frameshift near the start
   residue of the TAD of YAP using the clustered regularly interspaced
   short palindromic repeats (CRISPR-Cas9) assay (Supplementary
   Fig. [81]2b). The loss of TAD or FLTWL in YAP did not affect its
   stability or nuclear localization (Supplementary Fig. [82]2c, d). To
   test whether TADs are essential for tumor cell growth, we examined the
   growth of TDCL with TADs depletion using plate cloning and allograft
   assays in C57BL/6J mice. The results showed that knockout of TADs
   partially impaired tumor cell growth in vitro (Supplementary
   Fig. [83]2e). Strikingly, tumors with TAD depletion could not be
   established in immune-competent animals (Supplementary Fig. [84]2f),
   suggesting a strong immune attack from the hosts when TADs are
   depleted.

   To evaluate the contribution of TADs of YAP/TAZ in immunosuppression,
   we transplanted tumors with TADs depletion in immunodeficient NCG
   (NOD/ShiLtJGpt-Prkdc^em26Cd52Il2rg^em26Cd22/Gpt) mice and compared
   their growth in C57BL/6 J mice. Obviously, the tumors in
   immunodeficient mice can grow bigger than those in C57BL/6J mice
   (Fig. [85]2a). To assess tumor-specific immune responses, we examined
   CD8^+ T cells specific to ovalbumin (OVA) using H-2K(b) OVA tetramer
   antibodies. Initially, we transfected wild-type or TAD-deficient TDCL
   cells with a lentiviral vector expressing OVA fused with
   nano-luciferase (nLuc). Comparable intracellular luciferase activity
   confirmed similar OVA expression levels between the two cell lines
   (Supplementary Fig. [86]2g). Subsequently, we subcutaneously implanted
   these cells into C57BL/6 J mice. Our results demonstrated that tumors
   lacking TAD exhibited increased infiltration of OVA-tetramer-positive
   CD8^+ T cells (Fig. [87]2b, c). Moreover, infiltrating CD8^+ T cells
   from TAD-depleted tumors displayed higher fraction and enhanced
   cytotoxic activity (Fig. [88]2d, e and Supplementary Fig. [89]2h),
   suggesting a role for YAP-TAD in mediating immunosuppression within the
   tumor microenvironment. Importantly, depletion of CD8^+ T cells using
   anti-CD8 antibody in C57BL/6 mice restored tumor growth in TAD-depleted
   tumors (Supplementary Fig. [90]2i, j). To exclude potential effects of
   Cas9 neoantigens, we restored YAP expression in TAD-deficient cells.
   Our results showed that restorage of YAP expression reversed both the
   tumor growth defects and target gene downregulations due to TAD
   deficiency (Supplementary Fig. [91]2j, k). Collectively, these findings
   highlight the critical role of CD8^+ T cell-mediated antitumor immunity
   in the context of TAD deficiency.

Fig. 2. The TAD of YAP initiates immunosuppressive transcriptional program.

   [92]Fig. 2
   [93]Open in a new tab

   a Tumor growth of TDCL WT, YapΔTAD Taz^-/-#1, and YapΔTAD Taz^-/-#2
   cells injected subcutaneously to C57BL/6J mice or NCG mice (n = 5 mice
   per group). b Tetramer experimental design of tumor-specific T cells
   against OVA^+ cancer cells. c Flow cytometry analysis of OVA-tetramer^+
   in CD8 T cells of TDCL WT-OVA, YapΔTAD Taz^-/- OVA tumors in C57BL/6J
   mice (n = 5 mice per group). d Flow cytometry analysis of CD8^+ T cells
   in CD45^+ cells, GZMB^+ cells, and IFN-γ^+ cells of TDCL WT, YapΔTAD
   Taz^-/- tumors in C57BL/6J mice (n = 6 mice per group). e
   Immunofluorescence images of TDCL WT, YapΔTAD Taz^-/- tumors were
   stained for CD8 (red), and macrophage markers CD206 (red), CD163 (red)
   and F4/80 (green). The value was calculated by counting total CD8^+,
   CD206^+, or CD163^+ cells / tumor area (n = 5 per group). f GSEA
   analysis of YAP signature genes from RNA-seq data of TDCL YapΔTAD
   Taz^-/- and WT Cells. YAP signature genes set was obtained from the
   Molecular Signatures Database (MsigDB). g Heatmap showed the YAP target
   genes and immune suppressive genes from RNA-seq. h RT-qPCR analysis of
   gene expression in indicated genotypes of TDCL cells (n = 3 biological
   replicates). Data are presented as mean ± SEM. Statistical analysis was
   performed using two-way ANOVA (a), 2-tailed Student’s t test (c–e),
   one-way ANOVA (h).

   To identify YAP-TAD target genes involved in immunosuppression, we
   performed RNA sequencing followed by Gene Set Enrichment Analysis
   (GSEA). Our results showed significant downregulation of classic
   YAP/TAZ target genes, highlighting the essential role of YAP-TAD in
   transcriptional regulation (Fig. [94]2f). Additionally, we identified
   reduced expression of both known and novel immunosuppressive genes,
   including Tgfb2, Il33, and Itgb2, which play key roles in
   immunosuppression (Fig. [95]2g, Supplementary Fig. [96]2l).

   To validate these genes as direct YAP-TAD targets, we conducted RT-qPCR
   in TAD-depleted and WT cells, confirming their reduced expression upon
   TAD depletion (Fig. [97]2h, Supplementary Fig. [98]2m). Notably, Csf1
   and Il33 are known to promote macrophage recruitment and M2
   polarization, contributing to tumor immunosuppression^[99]27–[100]29.
   Immunofluorescence (IF) staining for CD206 and CD163 further revealed
   that YAP-TAD depletion significantly decreased M2 macrophage
   infiltration, which in turn enhanced CD8^+ T cell infiltration and
   cytotoxicity (Fig. [101]2e).

   In summary, our findings demonstrate that the TADs of YAP/TAZ are
   essential for driving oncogenic and immunosuppressive transcriptional
   programs in tumorigenesis.

TADs of YAP/TAZ initiate PIC assembly and Pol II loading through recruiting
TAF4

   To explore the detail molecular mechanism underlying transcriptional
   activation by the TADs of YAP/TAZ, we first performed the spatial
   interactome of YAP TAD using a Proximity-dependent Biotin
   Identification assay (BioID) (Supplementary Fig. [102]3a). Enrichment
   analysis of the mass spectrometry profile identified several
   transcriptional components, including TFIID, the Mediator complex, and
   transcription coactivators (Fig. [103]3a and Supplementary
   Fig. [104]3b), suggesting the involvement of TADs in transcription
   initiation via interactions with these transcriptional cofactors.

Fig. 3. TADs of YAP/TAZ initiate PIC assembly and Pol II loading through
recruiting TAF4.

   [105]Fig. 3
   [106]Open in a new tab

   a BirA^*-Gal4 and BirA^*-Gal4-YAP-TAD were expressed with 40×UAS in
   HEK293T cells, and the cell lysates were subjected to BioID-MS. b
   HA-YAP or YAPΔTAD was co-expressed with Flag-TAF4-v and subjected to
   coimmunoprecipitation assay (n = 3 independent experiments) (TAF4-v:
   TAF4 splice variant). c Flag-YAP was co-expressed with HA-TAF4
   M[373-837] or HA-TAF4 C[837-1086] and subjected to
   coimmunoprecipitation assay (n = 3 independent experiments). d Analysis
   for the CSP in ^1H-^15N-HSQC of ^15N-TAFH upon titration with YAP-TAD.
   For (d, f), CSP represents the difference in chemical shift of the peak
   position before and after titration. The detailed calculation method
   can be found in the Methods section. e The residues with significant
   CSP in (d) are mapped on the AlphaFold-predicted TAFH structure. f
   Analysis for CSP in ^1H-^15N-HSQC of ^15N-labeled YAP-TAD upon
   titrations with TAFH. g The residues with significant CSP in (f) are
   mapped on the YAP-TAD structure (red). h The CSP analysis for NMR
   titration of ^15N-TAFH with the pentapeptide FLTWL. i The residues with
   significant CSP in (h) are mapped on the Alphafold-predicted TAFH
   structure (blue). j A heatmap of YAP target genes and immunosuppressive
   genes in NCI-H1299 siNC or siTAF4 cells, derived from RNA-seq. k Upper
   panel: YAP/TAZ siRNA CUT&Tag analysis Pol II occupancy of YAP target
   genes (n = 269) in NCI-H1299 cells transfected with siNC or siYAP/TAZ
   (siYT). Lower panel: Metagene analysis showing total Pol II occupancy
   of YAP target genes (n = 242) in NCI-H1299 PLKO.1 or shTAF4 cells (TSS
   transcription start sites, TES transcription end sites). l
   Representative genome browser tracks of Pol II CUT&Tag in NCI-H1299
   siYT or shTAF4 cells for YAP target genes. The x-axis indicates the
   chromosome position, and the y-axis represents normalized read density
   in reads per million.

   To further confirm these interactions, we initially performed
   co-immunoprecipitation (Co-IP) assays to investigate the interactions
   between YAP and members of the TFIID family, which scored highly in the
   BioID assay and are key regulators of promoter occupancy. Our results
   showed that YAP physically interacts with TAF4 but not with TAF6, TAF7,
   or TAF9a (Supplementary Fig. [107]3c, d). Notably, depletion of the TAD
   of YAP or FLTWL motif significantly weakened these interactions
   (Fig. [108]3b and Supplementary Fig. [109]3e). Similarly, depletion the
   TAD of TAZ reduced its interaction with TAF4 (Supplementary
   Fig. [110]3f), aligning with previous transcriptional activity analysis
   (Fig. [111]1d). Further Co-IP mapping revealed that the middle portion
   of TAF4 is crucial for interacting with YAP (Fig. [112]3c).

   Prompted by the direct interactions observed in cells between the TAD
   of YAP and the middle fragment of TAF4, we employed NMR techniques to
   gain a more comprehensive and refined understanding of the binding
   interface between the two proteins. By dissecting the middle part of
   TAF4 into M1 (residues 373-582), M2 (residues 679-837), and the highly
   conserved TAF4 homology (TAFH) domain (residues 582−678)^[113]30
   (Fig. [114]3c), we expressed and purified the recombinant TAF4 M1,
   TAFH, M2, and TAF4[582-837] (i.e., TAFH + M2) proteins (Supplementary
   Fig. [115]3g–j) for biophysical characterization. Follow-up NMR
   titration assays revealed that TAF4 M1 and M2 were not observed to
   significantly interact with YAP-TAD (Supplementary Fig. [116]4a, c),
   but the binding affinity between TAF4[582-837] and YAP-TAD is similar
   to that of TAFH and YAP-TAD (Supplementary Fig. [117]4b and
   Supplementary Fig. [118]5a), which narrowed the binding site of TAF4 to
   the TAFH domain. We subsequently obtained most of the backbone
   resonance assignments for TAFH, enabling further biophysical
   characterization. The ^1H-^15N-HSQC spectrum for TAFH exhibits
   heterogeneously distributed peak intensities, resulting in the
   inability to assign the backbone resonances for certain residues in α1
   (Supplementary Fig. [119]5b). These observations, possibly caused by
   severe peak broadening, reflect the masked conformational dynamics of
   the putative hydrophobic pocket enclosed by these residues.

   Based on the backbone assignments for TAFH, we investigated the binding
   interface of TAFH involved in the YAP-TAD and TAFH interactions using
   NMR titration experiments. Observations from the overlaid ^1H-^15N-HSQC
   spectra of ^15N-TAFH in the presence or absence of YAP-TAD showed that
   peaks of certain TAFH residues experienced chemical shift perturbations
   (CSPs) upon YAP-TAD binding (Fig. [120]3d, e, and Supplementary
   Fig. [121]6a), indicating near μM-level binding affinity. These
   residues especially hydrophobic residues like L603, L653, L657, Y642,
   L606, and F599, were primarily scattered around the α1/α3/α4 helices,
   enclosing a hydrophobic pocket (Fig. [122]3e). The structural model of
   TAFH was predicted by AlphaFold2. In a reverse series of NMR titration
   assays, in which ^15N-labeled YAP-TAD was titrated with TAFH, both the
   C-terminal FLTWL motif and part of the helix of YAP-TAD displayed major
   chemical shift perturbations, indicating their deep involvement in
   TAD-TAFH interactions (Fig. [123]3f, g). Notably, the results of the
   two titrations indicated the involvement of the hydrophobic regions in
   both YAP-TAD and TAFH. Considering the similarity in hydrophobicity
   between the helix and the FLTWL of YAP-TAD, we conducted two additional
   NMR titration experiments to examine the binding properties of the
   standalone peptide fragments of these two regions toward TAFH.
   Interestingly, we observed that both fragments bound to almost the same
   hydrophobic pocket of TAFH as full-length YAP-TAD (Fig. [124]3h, i and
   Supplementary Fig. [125]6b, c). Collectively, NMR analysis revealed
   that the YAP-TAF4 interaction was heavily dependent on hydrophobic
   interactions, specifically involving a hydrophobic pocket on TAFH
   domain of TAF4 and both the  α-helical region and FLTWL motif on
   YAP-TAD.

   To further investigate the functional implications of the
   protein–protein interaction (PPI) between TAF4 and YAP-TAD in
   YAP-driven transcription, we measured the mRNA levels of target genes
   by RT-qPCR following TAF4 knockdown, both with and without YAP
   overexpression. TAF4 knockdown consistently reduced the classic target
   genes (CTGF, CYR61, and ANKRD1) expression in both conditions
   (Supplementary Fig. [126]7a, b). Moreover, overexpression of YAP did
   not reverse the reduction in YAP target gene expression by TAF4
   knockdown (Supplementary Fig. [127]7c). To assess the broader impact of
   TAF4 on YAP/TAZ genome-wide transcriptional activity, we conducted
   transcriptome profiling and observed the downregulation of YAP typical
   target genes ANKRD1, CYR61, MYC and the immunosuppressive genes CSF1,
   IL33, and ITGB2. (Fig. [128]3j and Supplementary Fig. [129]7d).

   To determine whether the downregulation of target gene expression by
   TAF4 knockdown was associated with reduced Pol II loading on the
   transcription start site (TSS) regions of YAP/TAZ target genes, we
   performed cleavage under targets and tagmentation (CUT&Tag) analyses of
   Pol II occupancy in TAF4-or YAP/TAZ-knockdown cells. Indeed, as
   expected, TAF4 depletion led to a genome-wide reduction in Pol II
   enrichment at both the TSS and gene body, supporting the role of TAF4
   as a broadly acting transcription initiation factor belonging to the
   TFIID family (Fig. [130]3k down, and Supplementary Fig. [131]7e, f). In
   contrast, siYAP/TAZ had a smaller impact on Pol II, with the most
   significant changes observed in Pol II occupancy at YAP target genes,
   indicating that YAP primarily regulates its specific target genes by
   interaction with TAF4 (Fig. [132]3k up and Supplementary
   Fig. [133]7g–j). For instance, Pol II occupancy at the typical target
   loci—CYR61, CTGF, MYC, AMOTL2, and CD155—was reduced in TAF4- or
   YAP/TAZ-knockdown cells (Fig. [134]3l and Supplementary Fig. [135]7k).

   In summary, the TADs of YAP/TAZ initiate Pol II loading for target gene
   transcription through TAF4 dependent Pol II loading, where the TAD-TAF4
   interaction is heavily reliant on hydrophobic interactions,
   specifically involving a hydrophobic pocket on TAF4 and the helix, and
   FLTWL on YAP-TAD.

TADs of YAP/TAZ recruit MED15 for Pol II efficient elongations

   In addition to TAF4, our interactome results showed potential
   interactions of mediator components, including MED1, MED15, and MED23,
   associated with the TAD of YAP (Fig. [136]3a). Previous studies have
   suggested a role for the mediator complex in facilitating Pol II
   release from promoters at YAP target loci^[137]31. This prompted us to
   explore whether the TAD of YAP directly recruits the mediator. Through
   an immune co-precipitation assay, we observed that YAP interacted with
   MED15, but not with MED23 or MED24 (Fig. [138]4a, b). Further
   examination of the interaction regions of MED15 and YAP revealed that
   the deletion of TAD completely abolished their interactions
   (Supplementary Fig. [139]8a). To explore the function of MED15 in YAP
   transcriptional activation, we performed a BioID assay to determine its
   spatial interactome (Supplementary Fig. [140]8b). As expected, we
   identified more than ten mediator components and numerous Pol II
   elongation factors (Supplementary Fig. [141]8c, d), suggesting that the
   mediator complex recruited elongation factors for YAP target gene
   elongation. Co-IP assays demonstrated that MED15 interacts with
   elongation factors such as PAF1, SPT5, and CDK9 (Fig. [142]4c). These
   interactions prompted us to examine the function of MED15 in Pol II
   elongation at the YAP target loci using the CUT&Tag assay. Knockdown of
   MED15 resulted in the downregulation of target genes (Supplementary
   Fig. [143]8e), and reduced Ki67 staining due to impaired growth
   (Supplementary Fig. [144]8f, g). Moreover, reconstitution of YAP/TAZ
   did not restore the YAP target gene reduction by MED15 depletion,
   suggesting that MED15 functions downstream of YAP/TAZ (Supplementary
   Fig. [145]8h). CUT&Tag analysis revealed that MED15 knockdown increased
   both the TSS and gene body occupancy of Pol II in several YAP/TAZ
   target genes, including CTGF, CYR61, AMOTL2, and FOSL1 (Fig. [146]4d,
   e). However, increased Pol II loading at target loci led to decreased
   target gene expression, reminiscent of previous findings, wherein the
   loss of PAF1 and SPT5 resulted in inefficient Pol II
   elongation^[147]32. To explore whether MED15 depletion leads to
   inefficient elongation, we examined premature termination events in the
   classic target gene CTGF by comparing the levels of nascent RNA and
   mRNA using RT-qPCR. The results revealed an increased ratio of nascent
   RNA to mRNA, indicating the inefficient elongation of Pol II on the
   target gene bodies (Fig. [148]4f). The above results implicate an
   MED15-involving YAP transcriptional activation process, necessitating a
   deeper understanding of the underlying mechanisms.

Fig. 4. TADs of YAP/TAZ recruit MED15 for Pol II efficient elongations.

   [149]Fig. 4
   [150]Open in a new tab

   a HA-YAP was co-expressed with Flag-MED15/MED23/MED24 and subjected to
   coimmunoprecipitation assay (n = 3 independent experiments). b
   Endogenous Co-IP of MED15 and YAP in NCI-H1299 cells (n = 3 independent
   experiments). c Endogenous Co-IP of PAF1, SPT5, CDK9 and MED15 in
   NCI-H1299 cells (n = 3 independent experiments). d Metagene profile of
   Pol II CUT&Tag reads of YAP/TAZ target genes (n = 150). e
   Representative genome browser tracks of Pol II CUT&Tag in NCI-H1299
   PLKO.1 or shMED15 cells. f RT-qPCR was used to detect both mature mRNA
   and nascent RNA levels of CTGF in NCI-H1299 PLKO.1 or shMED15 cells
   (n = 3 biological replicates). For nascent RNA detection, the reverse
   transcription primer was designed to target the first intron of CTGF,
   87 nucleotides downstream of the transcription start site (TSS).
   Expression levels were normalized to GAPDH. g Heatmap shows YAP target
   genes of NCI-H1299 MED15 KO or WT cells. h GSEA analysis of YAP
   signature genes from RNA-seq data of NCI-H1299 WT and MED15 knockout
   cells. i Tumor growth of NCI-H1299 WT or MED15 KO cells in xenograft
   models (n = 5 mice). j Model depicting YAP/TAZ-TAD promotion of PIC
   assembly through TAF4 and Pol II elongation via MED15. Data are
   presented as mean ± SEM. Statistical analysis was performed using
   2-tailed Student’s t test (f), and two-way ANOVA (i).

   To evaluate the function of MED15 in YAP-associated transcription,
   tumor cell growth, and tumorigenesis in vivo, we established MED15
   knockout cell lines in NCI-H1299 cells in a CRISPR-Cas9 dependent
   manner (Supplementary Fig. [151]8i). Initially, we compared the
   transcriptomes of MED15 KO and WT cells. The profiles revealed
   significant downregulation of several YAP target genes associated with
   cancer initiation and the progression of immunosuppressive genes
   (Fig. [152]4g). Moreover, GSEA analysis revealed that MED15 depletion
   downregulated YAP signature genes (Fig. [153]4h). To evaluate the
   function of MED15 in cell growth in vivo, we performed xenograft assays
   and observed that the knockout of MED15 robustly reduced the growth
   rate of tumor cells (Fig. [154]4i and Supplementary Fig. [155]8j, k).
   Additionally, restoring MED15 expression in NCI-H1299 MED15^-/- cells,
   rescued tumor growth defects (Supplementary Fig. [156]8l, m).
   Collectively, these findings indicate that MED15 is indispensable for
   the oncogenic transcriptional program regulated by YAP and for tumor
   cell growth, likely due to its role in transcriptional elongation.

   Taken together, TADs of YAP/TAZ have dual functions in initiation and
   elongation by recruiting TAF4 and Mediator 15 (Fig. [157]4j).

The hydrophobic interaction of YAP-TAD and MED15 boosts transcriptional hub
formation by co-condensations

   Multiple reports support the notion that YAP/TAZ functions as a
   transcriptional hub through coiled-coil (CC) domain-mediated phase
   separation facilitated by hydrophobic interactions. Key transcription
   regulators are recruited to the condensates for efficient transcription
   in a spatially partitioned manner^[158]24,[159]33–[160]35. This
   prompted us to test whether MED15 was recruited to YAP condensates.
   Interestingly, in contrast to other mediator components, MED15 was
   recruited as well as boosting YAP condensation, including increased
   condensate size and number when co-expressed with YAP (Fig. [161]5a and
   Supplementary Fig. [162]9a, b). To examine whether co-condensation and
   enhancement occurred in vitro, we synthesized and purified YAP and
   MED15 (Supplementary Fig. [163]9c). Upon mixing them in a physiological
   concentration of NaCl, we observed similar co-condensation and
   enhancement of YAP droplet formation (Fig. [164]5b, c). We further
   identified the region involved in the co-condensation of YAP and MED15
   by co-expressing different regions of MED15 and YAP, which revealed
   that the middle fragment of MED15 is required for co-condensation
   (Fig. [165]5d). Previous studies in yeast have identified several
   activation-binding domains (ABDs) in the middle region of
   MED15^[166]36,[167]37. Multiple sequence alignment revealed three
   relatively conserved ABDs (ABD1-3) in human MED15. Individual
   expression of ABDs with YAP demonstrated obvious co-condensation in the
   nucleus, with ABD3 showing particularly pronounced effects
   (Fig. [168]5e). These results suggested that the hydrophobic
   interactions of MED15-YAP acted as another driving forces for YAP
   condensation paralleled to CC mediated hydrophobic interactions. To
   test this hypothesis, we examined whether the condensation defect of
   YAP^S127A△CC could be restored by co-expression with MED15.
   Surprisingly, MED15 co-expression with YAP^S127A△CC restored its phase
   separation, subsequently rescuing its transcriptional activity defects
   (Fig. [169]5f). To test whether co-condensation drives transcription in
   the droplets, we fused the TAD of YAP/TAZ with rTetR, co-expressed it
   with PP7 coat protein (PCP) -3×mCherry and TetO-PP7, and visualized and
   monitored the levels of nascent RNA by microscopy^[170]38. We observed
   increased transcription in the condensates after doxycycline treatment
   (Fig. [171]5g).

Fig. 5. The hydrophobic interaction of YAP-TAD and MED15 boosts
transcriptional hub formation by co-condensations.

   [172]Fig. 5
   [173]Open in a new tab

   a Confocal sections of YAP^S127A -mEGFP expressed with or without
   Flag-MED15, and scatter plot showed the puncta diameter (n = 24 cells
   per group). b Fluorescence images of in vitro His-YAP-mEGFP (green) and
   His-MED15-CY3 (red) were mixed with the indicated module concentration
   in 10% PEG solution. c Plot shows mean droplet size. DIC images of
   His-YAP-mEGFP and His-MED15-CY3 mixture (1:1 at 5 µM, 10% PEG) shows
   numerous droplets at room temperature in a coverslip chamber. d
   mEGFP-YAP^S127A was expressed with Flag-MED15-N, Flag-MED15-M, and
   Flag-MED15-C. e Schematic representation of ABDs of MED15.
   mEGFP-YAP^S127A was expressed with MED15-ABD1, ABD2, Linker, and ABD3.
   f Images of Flag-MED15 co-expressed with mEGFP- YAP^S127A or mEGFP
   YAP^S127AΔCC, histogram shows the ratio of cells with puncta (n = 100
   cells per group). RT-qPCR analysis of YAP target genes in cells
   transfected with Flag-MED15 (1 μg), mEGFP- YAP^S127A (200 ng) or mEGFP-
   YAP^S127AΔCC (200 ng) (n = 3 biological replicates). g Schematic of
   visualizing transcriptional activation in YAP/TAZ-TAD and MED15
   condensates. Arrows indicate the nascent RNA signals (bound by
   PCP-3×mCherry) colocalizing with TAD/MED15 condensates
   (rTetR-BFP-YAP/TAZ-TAD and mEGFP-MED15) after addition of doxycycline
   (0.01 mg/ml). h Schematic of the hydrophobic interaction of YAP and
   MED15 forming a transcriptional hub to regulate the transcription
   initiation and elongation. Data are presented as mean ± SEM.
   Statistical analysis was performed using 2-tailed Student’s t test (a),
   one-way ANOVA (c, f). Scale bars: 10 μm (a, f), 5 μm (b, c, d, e, g).

   These results indicated that the hydrophobic interactions of TAD-ABD3
   enhanced YAP condensation parallel to CC-mediated hydrophobic
   interactions (Fig. [174]5h).

The mechanism of interaction between TADs and MED15

   To further understand elongation, we investigated the molecular
   mechanisms underlying the interactions between TADs and MED15. By
   performing multiple sequence alignments, we identified three relatively
   conserved ABDs (ABD1-3) in human MED15 (Fig. [175]5e). This result
   prompted us to examine whether YAP-TAD binds to MED15-ABDs in humans
   similarly to the fuzzy interactions observed in yeast. We prepared
   protein samples from MED15 and its ABDs (Supplementary Fig. [176]10a–e)
   and performed hydrogen-deuterium exchange (HDX) detected by mass
   spectrometry (MS) (HDX-MS) assays to analyze the binding sites on YAP
   and MED15. Unlike the fuzzy interactions in yeast, the results
   indicated that the TAD domain of YAP and the ABD3 domain of MED15 were
   the primary binding sites responsible for the YAP-MED15 interaction
   (Fig. [177]6a; Supplementary Fig. [178]10f–h, Supplementary
   Fig. [179]11a, b; Supplementary Fig. [180]12a, b; Supplementary
   Fig. [181]13a). Follow-up NMR titration assays confirmed that ABD1 and
   ABD2 did not bind YAP-TAD (Supplementary Fig. [182]13b, c). Consistent
   with our expectations, surface plasmon resonance (SPR) characterization
   revealed a substantial interaction between YAP-TAD and MED15
   (particularly its ABD3 domain), evidenced by comparable dissociation
   constants (Kd = 27.5 ± 14.8 nM for MED15 and YAP, Kd = 160.2 ± 54.3 nM
   for YAP-TAD and ABD3; Supplementary Fig. [183]14a–c). These in vitro
   biophysical assays aligned with findings from co-condensation
   experiments for individual ABDs, prompting us to focus on ABD3-TAD
   interactions in subsequent studies.

Fig. 6. The mechanism of interaction between TADs and MED15.

   [184]Fig. 6
   [185]Open in a new tab

   a HDX-MS characterization for identification of YAP-MED15 binding
   sites. Deuterium uptake plots of MED15 peptides (Region I: AA 492-511,
   +4 charge; Region II: AA 531-551, +5 charge) and YAP peptide (AA
   491-502, +3) were measured in the presence and absence of their binding
   partners. Data are plotted as percent deuterium uptake versus time
   (logarithmic scale). Red and green plots represent the unbound and
   bound states, respectively (n = 3 technical replicates). b NMR
   structure ensemble of MED15-ABD3 consisting of 20 lowest-energy models.
   c Analysis for the CSP of ^15N-YAP-TAD upon titration with ABD3. d The
   residues with significant CSP during NMR titration in (c) are mapped on
   the YAP-TAD structure in red. e Predicted helix formation propensity
   for YAP-TAD with or without ABD3 using TALOS-N webserver. f Analysis
   for the change in CSP of ^15N-ABD3 upon titration with YAP-TAD. g The
   residues with significant CSP change during NMR titration in (f) are
   mapped on the MED15-ABD3 structure in blue. h Analysis for the change
   in CSP of ^15N-ABD3 upon titration with the pentapeptide FLTWL. i The
   residues with significant CSP change in (h) during NMR titration
   experiment are mapped on the MED15-ABD3 structure in blue. Data are
   presented as mean ± SD.

   To acquire detailed properties of the TAD-MED15 interaction, we
   assigned most of the backbone and side-chain chemical shifts of a
   standalone truncation of the ABD3 domain of MED15 (residues 453-561 of
   MED15 isoform 2, named MED15-ABD3) using NMR spectroscopy. We next
   determined the solution structure of MED15-ABD3 under PDB accession
   number 8J9A, which consists of a long N-terminal loop and three
   α-helices (designated as α1, α2, and α3) (Fig. [186]6b and
   Supplementary Fig. [187]15a, b). Subsequently, we conducted 2D HSQC
   spectrum-based NMR titration experiments, using the CSPs observed
   during the titration of MED15-ABD3 into ^15N-YAP-TAD to identify the
   binding sites on YAP-TAD for the YAP-ABD3 interaction (Fig. [188]6c and
   Supplementary Fig. [189]16a). Notably, certain hydrophobic residues in
   the helix and C-terminal FLTWL of YAP-TAD exhibited significant
   chemical shift perturbations, indicating their involvement in the
   binding of MED15-ABD3 (Fig. [190]6c, d). Interestingly, ABD3 binding
   induced the elongation of the α-helix of YAP-TAD (Fig. [191]6e). The
   extension of the secondary structural elements of the initially
   disordered region of YAP-TAD depicts the flexible and binding-ready
   nature of this dual-targeting transcriptional coactivator, reflecting
   its versatile use in various situations regarding the transcription
   processes.

   Backbone assignments and the MED15-ABD3 structure enabled us to perform
   reverse NMR titrations where serial concentrations of YAP-TAD were
   titrated into ^15N-MED15-ABD3. These experiments revealed that residues
   involved in binding were scattered mainly on α2 and α3 helices, with
   the majority located at the helix-helix interfaces, forming a
   clustering pattern of a line-like continuum (Fig. [192]6f, g and
   Supplementary Fig. [193]16b). Similarly, titration involving
   ^15N-TAZ-TAD with MED15-ABD3, as well as its reverse titration, showed
   a similar binding pattern to interactions between YAP-TAD and
   MED15-ABD3 (Supplementary Fig. [194]16c–f). Based on these findings on
   the TAD-ABD3 interaction and TAD-TAF4 binding sites, we investigated
   the roles played by the helix and FLTWL motif of YAP-TAD in the
   TAD-ABD3 interaction by titrating peptides containing each specific
   fragment into ^15N-MED15-ABD3. The FLTWL peptide had a binding
   interface analogous to YAP-TAD on MED15-ABD3, with most binding
   residues being hydrophobic and positioned at the aforementioned
   helix-helix junction interface (Fig. [195]6h, i and Supplementary
   Fig. [196]17a). Contrary to the ^15N-YAP-TAD titration results
   (Fig. [197]6c, d), the α-helix-based peptide showed reduced affinity
   for binding ABD3 in a standalone state, characterized by hardly
   observed CSPs or peak intensity changes (Supplementary Fig. [198]17b).
   This observation suggests a weaker α-helix-ABD3 interaction compared to
   α-helix-TAF4 binding (Supplementary Fig. [199]6c and Supplementary
   Fig. [200]17b), potentially reflecting differentiated binding
   mechanisms and preferences for the two target proteins. In summary, the