Transient proliferation by reversible YAP and mitogen control of the cyclin D1/p27 ratio
Katherine R. Ferrick, Samsara W. Upadhya, Yilin Fan, Nalin Ratnayeke, Mary N. Teruel, Tobias Meyer
TL;DR
The study explores how YAP signaling controls cell proliferation during tissue repair by regulating the cyclin D1/p27 ratio, balancing growth and cancer prevention.
Contribution
The paper reveals that YAP integrates mitogen and contact signals to regulate the cyclin D1/p27 ratio, enabling transient proliferation without cancer initiation.
Findings
YAP increases the cyclin D1/p27 ratio by enhancing mitogen signaling through receptor upregulation.
Contact inhibition reverses YAP activity, suppressing mitogen signaling and the cyclin D1/p27 ratio.
Quantitative imaging shows how signaling pathways coordinate cell cycle entry and exit in epithelial tissues.
Abstract
Hippo-YAP signaling orchestrates transient proliferation during tissue repair and is therefore an attractive target in regenerative medicine. However, it is unclear how YAP integrates mitogen and contact signals to start and stop proliferation. Here we show that reduced contact inhibition, increased mitogen signaling, and YAP-TEAD activation converge on increasing the nuclear cyclin D1/p27 protein ratio during early G1 phase, towards a threshold ratio that dictates whether individual cells enter or exit the cell cycle. YAP increases this ratio in concert with inducing mitogen signaling, by increasing EGFR and other receptors that signal primarily through ERK. After a delay, contact inhibition suppresses YAP activity, which gradually downregulates mitogen signaling and the cyclin D1/p27 ratio. Thus, critical for regeneration without cancer initiation, robust proliferation responses…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —https://doi.org/10.13039/100000057U.S. Department of Health & Human Services | NIH | National Institute of General Medical Sciences (NIGMS)
- —https://doi.org/10.13039/100000001National Science Foundation (NSF)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsHippo pathway signaling and YAP/TAZ · Cancer-related Molecular Pathways · RNA Research and Splicing
Introduction
The Hippo pathway functions as a potent and rapidly reversible restrictor of the TEAD (TEA-domain transcription factor) cofactors YAP (Yes-associated protein) and TAZ (transcriptional coactivator with PDZ binding motif) through multisite phosphorylation and cytoplasmic sequestration. Acute chemical and physical injuries inactivate the Hippo pathway, resulting in transient activation of the cofactors YAP/TAZ to increase proliferation, induce metabolic changes, and promote matrix remodeling^1,2^. YAP nuclear translocation initiates transcription through binding to TEAD and other nuclear transcription factors. Though YAP and its paralog TAZ are dispensable for the basal proliferation of several tissues, they are crucial for regeneration and repair following injury in tissues such as skin, liver, and kidney^1^. Inhibitors of Hippo pathway proteins are therefore being developed for clinical use in tissue regeneration^3,4^. However, the diversity of signaling processes controlled by YAP has made it difficult to resolve the molecular mechanism of the transient proliferation response using animal models.
Hippo signaling is kept active in epithelial sheets and other tissues by cell-cell contacts, such as adherens junctions, and is inactivated by secreted factors and mechanical stress during tissue repair^5^. Adapter proteins like Merlin (NF2) contribute to contact inhibition signaling by recruiting and facilitating the activation of the Hippo pathway kinases LATS1/2 (Large Tumor Suppressor 1/2). Subsequently, LATS1/2 suppress YAP/TAZ-TEAD activity through multisite phosphorylation^6^. Loss of Merlin or LATS activity leads to dramatic increases in the size of tissues and organs, and eventually tumorigenesis^7,8^, demonstrating the importance of restraining YAP activation.
Despite the well-known role of YAP in promoting proliferation, the mechanism by which YAP integrates with other signaling pathways to regulate both cell cycle entry and exit is less clear. For example, constitutive YAP activation cooperates with AP1 and E2F transcription factors^9,10^ and increases expression of a broad array of cell cycle genes, prominently replication factors and regulators of G2/M progression^9,11^. YAP also induces secretion of various ligands and signaling regulators^12^. Furthermore, mitogen signaling and contact inhibition pathways control proliferation^13,14^ in parallel and in crosstalk with YAP activity. However, such networks must impinge on inactivation of retinoblastoma protein (Rb), the suppressor of the cell cycle transcription factor E2F, to trigger cell cycle entry. As persistent YAP activation induces organ hyperplasia and cancer, YAP-TEAD must be robustly inactivated after its initial induction, ensuring a transient proliferative state. Thus, to understand tissue repair, it is crucial to determine how YAP, mitogens, and contact inhibition control Rb phosphorylation to first promote and then suppress cell cycle entry.
Here, we use multiplexed immunofluorescence to examine YAP-regulated proliferation in thousands of cells within an epithelial monolayer. We show that over a wide range of stimuli, mitogen signaling, contact inhibition, and YAP-activation converge on the cyclin D1/p27 ratio during G1 phase, which dictates phosphorylation of Rb, and whether cells enter the cell cycle. Persistent YAP activation increases the levels of cyclin D1 as well as EGFR and other receptors, which in turn signal primarily through MEK-ERK to increase the cyclin D1/p27 ratio. Following acute YAP inhibition, we find that EGFR signaling and the cyclin D1/p27 ratio lower gradually, so that cell cycle exit occurs after a delay. Finally, we show that YAP signaling can be robustly terminated: even a strong increase in YAP activity generated by ablation of the upstream Hippo component Merlin could be reversed by contact inhibition of YAP, but only at higher local cell density. Together, our study identifies a YAP-controlled balancing mechanism between contact inhibition and receptor signaling that transiently induces the cyclin D1/p27 ratio. The delay between initiation and inhibition ensures that the YAP-induced proliferation response is prolonged to allow for accumulating sufficient cyclin D1 to overcome CDK inhibitors, while the robust YAP inactivation by increased cell density ensures that the proliferation response is inherently short-lived.
Results
Contact inhibition of YAP opposes mitogen stimulation in a balance mechanism controlling cell proliferation
Even within epithelial monolayers subject to the same mitogen stimuli, there is considerable heterogeneity in cell cycle signaling^13,15^. We therefore used quantitative, single-cell immunofluorescence analyses to determine how YAP activation integrates with mitogen and contact inhibition to control cell cycle entry. As two indicators of cell cycle entry, we first measured phosphorylation of retinoblastoma protein (p-Rb, S807/811), which is a bimodal marker of cyclin-CDK activity^16^ that is low in arrested cells, phosphorylated in G1, and remains phosphorylated through mitosis (Fig. 1a). Second, we measured incorporation of 5-Ethynyl-2’-deoxyuridine (EdU) to mark cells that entered S phase and replicate their DNA. These two metrics are therefore closely correlated (Supplementary Fig. S1a). To quantify local cell density in imaging sites, we subdivided nuclei into a 10 × 10 grid ( ~ 75 × 75 μm) and measured local density for each cell based on the number of cells in its block (Fig. 1b and Supplementary Fig. S1b, see also “Methods”). We pooled together hundreds of density blocks and measured the fraction of proliferating cells across bins of increasing local density. We used hTERT-immortalized retinal epithelial cells (RPE1) as our primary model of density-dependent proliferation, as they reversibly arrest with serum starvation and rapidly respond to local density changes^15^. RPE1 exhibited locally graded levels of proliferation within 48 h of seeding through this block-based metric, as well as an alternate metric quantifying neighboring cells within a radius of 50 μm (Supplementary Fig. S1c, d).Fig. 1. Contact inhibition of YAP opposes mitogen stimulation in a balance mechanism that controls proliferation.a Top: diagram of cell cycle phases: EdU incorporation marks S-phase, Rb hyperphosphorylation (p-Rb, S807/811) cycling cells. Bottom: sample immunofluorescence (IF) images of RPE1 cells: Hoechst for DNA (white), EdU for S-phase cells (green), p-Rb for cycling cells (blue), where magenta outline is the automated nuclear mask (scale bar = 10 μm). b Top: representation of local cell density metric, quantified by subdividing field of view into 10 × 10 grids and counting the centroids of cell nuclei within each block (see also “Methods”). Bottom: sample histogram distribution of local cell densities within a field of view. c Percent p-Rb of wildtype (WT) and constitutively-active YAP (YAP5SA, CA-YAP) RPE1 cells in each local density bin, for increasing serum levels. The percent was calculated as the fraction of p-Rb-positive cells from local cell densities (relative units, R.U.) pooled across fields of view within a well (25-36 sites/well) and technical well replicates (2-3/biological replicate), n > 20,000 cells/condition. d Mean percent EdU-positive for WT and CA-YAP cells (treated as in (c)). N = 3 independent experiments. Student’s t test (EdU): p = 1.01 × 10^−5^ (0.1% serum), 8.13 × 10^−5^ (1% serum), 7.13 × 10^−4^ (10% serum). e Median nuclear/cytoplasmic (n/c) YAP ratio per local density bin (left) and percent p-Rb calculated in cells binned by YAP ratio (right) for WT cells treated as in (c), n > 20,000 cells/condition. Shaded error bars are 95% confidence intervals. f Top: diagram of inhibitors against LATS1/2 (0.5 µM, TDI-011536) and TEAD (1 µM, GNE-7883). Middle: schematic of TEAD reporter construct driving expression of mScarlet-NLS-PEST. Bottom: representative images of TEAD reporter in cells treated with DMSO or inhibitors for 24 h. Scale bar = 50 µm. g Mean TEAD reporter activity (real fluorescent units, RFU) over time (left) or aligned by time of mitosis (right) in cycling RPE1 treated as in (f). To select for cells that received drug during the same window within the cell cycle, cells were gated for receiving drug 6–9 h prior to mitosis. Shaded error bars are ±SEM. N > 500 cells/condition, data are representative of 2 independent experiments. h Representative images of YAP staining in WT cells treated with DMSO, LATSi, or TEADi (as in (c)) in starvation or serum conditions. Scale bar = 50 µm. Percent p-Rb per local density bin for RPE1 (i) and MCF10A (j) cells treated with inhibitors (24 h) after pre-conditioning in serum-free or serum conditions. k Percent EdU-positive cells for RPE1 (left) and MCF10A (right) treated as in (i, j). Ordinary one-way ANOVA with Dunnett’s multiple comparisons (to DMSO): RPE1, p = 0.0028 (LATSi, 0.1%), 0.84 (TEADi, 0.1%), 0.037 (LATSi, 1%), 0.66 (TEADi, 1%), 0.027 (LATSi, 10%), 0.47 (TEADi, 10%); MCF10A, p = 0.38 (LATSi, 0%), 0.95 (TEADi, 0%), 0.020 (LATSi, 0.5%), 0.80 (TEADi, 0.5%), 0.0024 (LATSi, 5%), 0.17 (TEADi, 5%). All binned plots have n > 100 cells/bin, >20,000 cells/condition, and are representative of ≥3 independent experiments. RPE1 cells were seeded for confluency at 8000–12,000 cells per 96-well (~25,000–40,000 cells cm^−2^). MCF10A cells were seeded densely according to prior reports of YAP suppression at 24,000 cells/well^25^ (~80,000 cells cm^−2^).
We first asked how persistent YAP activity changes the mitogen dependence of proliferation by stably expressing a doxycycline (DOX)-inducible, constitutively active YAP variant (5 regulatory serines mutated to alanine^17^, CA-YAP). As leaky expression of CA-YAP (without DOX treatment) was sufficient to increase proliferation, we did not use DOX induction unless otherwise stated. Using increasing concentrations of serum (fetal bovine), we found that the fraction of proliferating cells was higher in cells expressing CA-YAP compared to wild type (WT) cells at all mitogen levels and at similar final cell densities (Supplementary Fig. S1e).
Strikingly, the density dependence of proliferation was abolished in CA-YAP cells regardless of mitogen levels (Fig. 1c), demonstrating that contact inhibition of proliferation requires contact inhibition of YAP activity. CA-YAP cells also retained mitogen sensitivity since the percentage of proliferating cells increased with serum (Fig. 1d). Thus, the inverse regulation between mitogens and cell density represents an intuitive balancing mechanism: the inhibition of proliferation at a given cell density can be overcome by increasing mitogen signaling, and the increase in proliferation from mitogen signaling can be antagonized once cell division has increased local cell density. However, persistent activation of YAP breaks the balance and makes proliferation independent of cell density.
We next tested whether contact inhibition and mitogen signaling regulate endogenous YAP since RPE1 cells have not been extensively used in previous Hippo pathway studies. As the ratio of the nuclear to cytoplasmic (n/c) YAP and the absolute nuclear level of YAP are both relative measures of YAP activity^18^ and reported to depend on mitogens^19,20^, we measured both parameters as a function of local cell density (Fig. 1e and Supplementary Fig. S1f). We observed that the YAP ratio reduced with density across all serum conditions, and that serum deprivation also reduced total YAP levels. To assess how endogenous YAP contributes to proliferation, we measured the percentage of proliferating cells in bins of increasing YAP. With serum, there was a modest increase in proliferation with YAP activity, but unexpectedly, by either metric, increasing serum levels increased proliferation for cells with identical YAP levels (Fig. 1e). Thus, while YAP activation and mitogen stimuli synergistically increase proliferation, we find that mitogens can also increase proliferation without increasing endogenous YAP levels or localization in RPE1 cells.
As constitutive YAP activation can lead to cellular reprogramming over time, we next determined how acute increases and decreases in YAP activity regulate proliferation using a LATS1/2 inhibitor^3,21^(LATSi) and a pan-TEAD inhibitor^22^ (TEADi), respectively (Fig. 1f). To assess the kinetics with which LATS inhibition and TEAD inhibition affect YAP activity, we adapted a transcriptional promoter with TEAD binding sites (8x-GTIIC^23^) with a nuclear-localized fluorescent tag containing a PEST sequence to promote reporter turnover^24^. Compared with DMSO treatment, LATS inhibition increased TEAD transcriptional activity maximally within 6 h in asynchronously cycling cells, whereas TEAD inhibition repressed it gradually over time (Fig. 1g). Aligning cells by their time of mitosis, we confirmed that cells receiving either inhibitor treatment during G2-phase of the previous cell cycle (3–6 h before mitosis) entered the subsequent G1 with higher or lower TEAD activity. Imaging over 24 h confirmed that the LATS and TEAD inhibitors lead to approximately doubling or halving of TEAD activity, respectively (Supplementary Fig. S1g). We also confirmed that LATS inhibition triggered nuclear enrichment of endogenous YAP while TEAD inhibition promoted cytoplasmic localization, most apparent in the higher YAP levels present in serum conditions (Fig. 1h). Accordingly, LATS inhibition greatly boosted proliferation at both low and high serum stimuli and reduced density dependence (Fig. 1i). Conversely, TEAD inhibition greatly reduced the fraction of proliferating cells for low serum, suggesting that YAP-TEAD activity is particularly rate limiting for inducing proliferation under low mitogen stimulation conditions.
RPE1 cells differ from several common epithelial models of YAP regulation by expressing N-cadherin instead of E-cadherin. We therefore compared how TEAD activation and inhibition controls proliferation in a non-transformed mammary epithelial cell line (MCF10A). Using high-density seeding (~80,000 cells cm^−2^) as previously used to suppress YAP^25^, we confirmed that LATS and TEAD inhibition have opposing effects on MCF10A proliferation (Fig. 1j). Distinct from RPE1 cells, LATS inhibition in serum-starved MCF10A did not restore proliferation, perhaps due to stronger contact inhibition signaling. Further comparison of varied seeding densities for RPE1 (Supplementary Fig. S1h–j) and MCF10A (Supplementary Fig. S1k–m) revealed that RPE1 increased YAP nuclear levels and proliferation with LATS inhibition, whether seeded for confluent or dense final densities, and regardless of serum levels. In contrast, serum-starved MCF10A only proliferated with LATS inhibition if seeded at a lower initial density, suggesting that Hippo-independent mechanisms may restrict YAP activation and the capacity for cell cycle reentry in these cells. Also, RPE1 cells increased proliferation with LATS inhibition at low, intermediate, and high serum levels, whereas MCF10A were most sensitive with serum (Fig. 1k). Thus, density can restrict LATS-dependent activation of YAP, but in a serum and cell-type-dependent manner.
To test whether other cell lines exhibit serum-dependent sensitivity to LATS inhibition, we compared additional non-transformed (human umbilical vein endothelial cells, HUVEC) and cancerous (CACO2, MDA-MB-231, and MDA-MB-468) cell responses to LATS and TEAD inhibition at high cell density (Supplementary Fig. S1n-q). While all the tested cell lines proliferated under low or serum starvation conditions, HUVEC, MDA-MB-231, and MDA-MB-468 demonstrated dependence on YAP activity manipulation at low and higher serum, whereas CACO2 proliferation varied minimally. Thus, several non-transformed and cancerous cells show YAP-dependent proliferation, particularly for low mitogen conditions. These studies also suggest that, due to their sensitivity to YAP, mitogen signaling, and contact inhibition over a wide range of conditions, RPE1 cells are particularly useful for studying the interaction between the Hippo–LATS-TEAD signaling and mitogen signaling (Fig. 1k).
We conclude that contact inhibition of YAP activity and mitogen signaling can compete to respectively suppress or promote proliferation. However, this balance can be broken in two ways: by high mitogen stimuli inducing proliferation in cells with low YAP-TEAD activity and persistent activation of YAP abolishing contact inhibition of proliferation.
YAP-regulated cell cycle entry and exit is dictated by the cyclin D1/p27 threshold
We next sought to identify how YAP activity integrates with contact inhibition and mitogens to control G1 progression. Cells enter the cell cycle during G1 phase by activating cyclin-CDK to hyperphosphorylate Rb and derepress the E2F cell cycle transcription program^16^. In contrast, in cells that exit the cell cycle, cyclin-CDKs stay inactive, and Rb remains dephosphorylated. This bimodal Rb phosphorylation is thought to be controlled primarily by the CDK4/6 activator cyclin D1 and the cyclin-CDK2/4/6 inhibitor p27^15^. Importantly, cyclin D1 is thought to contribute to the YAP-mediated regulation of proliferation^2^ and p27 is similarly a key regulator of cell proliferation in vivo, with p27 deletion resulting in multi-organ hyperplasia^26^.
Using image-based cytometry and multiplexed immunofluorescence^27^ to measure multiple sets of proteins within each cell, we first computationally gated for G1 cells, as both proteins are degraded in S-phase (Supplementary Fig. S2a, b, see also “Methods”). We first compared how local cell density and YAP activity regulate the cyclin D1 and p27 levels. As local density increased, cyclin D1 levels decreased, and p27 levels increased nearly two-fold (Fig. 2a). Density-normalized (Fig. 2a) and absolute (Supplementary Fig. S2b–c) protein levels of cyclin D1 and p27 were both strongly dependent on serum, respectively higher and lower. Markedly, the CA-YAP cells lost the serum and density dependence of both, keeping the levels of cyclin D1 high and p27 low in G1 phase, arguing that YAP activity is not only upregulating the level of cyclin D1 but also downregulating the protein level of p27.Fig. 2YAP-regulated cell cycle entry and exit is dictated by the cyclin D1/p27 threshold.Normalized levels (real fluorescent units,RFU) of cyclin D1 (left) and p27 (right) per local cell density bin for WT (a) and CA-YAP (b) cells with or without serum. Binned data only include G1-gated cells and median levels are normalized to the lowest density bin. N > 15,000 cells per condition, n > 100 cells/bin. Data are representative of 2 independent experiments. c Representative images of multiplexed immunofluorescence staining comparing cyclin D1 (green) and p27 levels (magenta) with aligned p-Rb stain (cyan, with gray nuclear mask overlay) in WT or CA-YAP cells. Scale bar = 50 μm. Representative single-cell scatterplots using image-based cytometry of cells co-stained for cyclin D1 and p27 levels (log2-transformed), colored by p-Rb state (purple, p-Rb-negative; orange, p-Rb-positive) in WT (d) and CA-YAP (e). Top, no serum; bottom, full serum. f Representative cyclin D1/p27 single-cell scatterplots of siRNA knockdown (top: Ctrl, control; bottom: CycD1, cyclin D1) in serum-starved CA-YAP cells, colored by p-Rb state. Data are representative of 3 independent experiments. g Normalized change in cyclin D1/p27 ratio (top) and percent p-Rb (bottom) in serum-starved WT (gray) and CA-YAP (red) cells with 48 h siRNA knockdown of control, cyclin D1, or p27. Data are normalized to control condition, mean ± STD. Paired Student’s t test: normalized cyclin D1/p27 ratio, p = 0.0035 (WT/siCycD1), 0.044 (WT/sip27), 5.02 × 10^−4^ (CA-YAP/siCycD1), 0.095 (CA-YAP/sip27); normalized p-Rb, p = 0.0035 (WT/siCycD1), 0.22 (WT/sip27), 0.0074 (CA-YAP/siCycD1), 0.37 (CA-YAP/sip27). h, i Top (h): CRISPRa schematic for dCas9-based, transient activation of gene expression (48 h). Bottom h, i: Percent p-Rb per local density bin (left) and mean p-Rb (right) for G1-gated cells overexpressing control (sgCtrl, grey), cyclin D1 (sgCycD1, red), or p27 (sgp27, blue) in starved (h) or serum (i) conditions. N = 2 independent experiments. Student’s t test (compared to sgCtrl): −serum, 1.71 × 10^−4^ (sgCycD1), 0.028 (sgp27); +serum, 0.045 (sgCycD1), 0.022 (sgp27). j Representative single-cell scatterplots of co-stained cyclin D1 and p27 levels colored by p-Rb in starved, WT cells with sgRNA-mediated overexpression of p27, treated with DMSO (left) or LATS inhibitor (right, 0.5 µM TDI-011536) for 24 h. Data are representative of 2 independent experiments. k Schematic of mitogen signaling and contact inhibition of YAP-TEAD activity converging on the cyclin D1/p27 ratio to control Rb inactivation and cell cycle entry. p-Rb boundary lines represent a linear fit using a classification Support Vector Machine model, calculated per experiment (see “Methods”). Single-cell data are representative of 2 independent experiments, with n = 5000 random cells/scatter plot and gated for G1-phase (defined as 2 N DNA/EdU-negative status).
Single-cell analysis of Rb phosphorylation in cells co-stained with cyclin D1 and p27 (Fig. 2c), revealed that the level of cyclin D1 relative to p27 was tightly correlated with Rb phosphorylation (Supplementary Fig. S2d). Only cells with relatively more cyclin D1 than p27 had Rb phosphorylated (Fig. 2d). Thus, even for low mitogen levels, a small fraction of cells had sufficiently high cyclin D1 and low p27 to have Rb phosphorylated. Conversely, at high mitogen stimulation, more cells populated the high-cyclin D1/low-p27 state and had Rb phosphorylated, demonstrating a critical threshold cyclin D1/p27 ratio for proliferation.
In CA-YAP and LATS inhibitor-treated cells, Rb phosphorylation remained closely predicted by the relative nuclear cyclin D1 and p27 levels (Fig. 2e, Supplementary Fig. S2e, f), implying that YAP signals through both cyclin D1 and p27 to control proliferation. As a control, we analyzed cells at higher cell density and found that LATS inhibition still increased cyclin D1 and reduced p27 levels (Supplementary Fig. S2e). We also observed a subpopulation of CA-YAP cells that had high cyclin D1 without Rb phosphorylated and identified that these cells had increased protein levels of p21, likely replacing p27 as CDK inhibitor (Supplementary Fig. S2g). Thus, YAP activation results in an increase of the cyclin D1/p27 ratio, coinciding with mitogen independence and loss of contact inhibition.
We next tested whether perturbations that change cyclin D1 and p27 levels indeed alter Rb phosphorylation downstream of YAP activity. First, we showed that siRNA-mediated knockdown of cyclin D1 strongly reduced the proportion of cells with phosphorylated Rb by shifting most of the CA-YAP cells to a high-p27 and quiescent state while preserving control by the relative cyclin D1/p27 ratio (Fig. 2f, g, Supplementary Fig. S2g, h). Conversely, knocking down p27 increased the cyclin D1/p27 ratio but only weakly increased proliferation (Supplementary Fig. S2i). We observed that a subset of CA-YAP cells retained high p21 levels when p27 was knocked down, explaining why average proliferation levels were similar, whereas binned analyses showed a slight increase (Supplementary Fig. S2h–j). The same changes were also observed at higher density (Supplementary Fig. S2k), suggesting a critical dependence on cyclin D1.
Conversely, we used CRISPRa-mediated induction of cyclin D1 and p27 to manipulate the cyclin D1/p27 ratio in WT cells to assess whether we could confer mitogen or density-independent proliferation independently of YAP. Neither perturbation overcame the density-dependence of the endogenous YAP ratio (Supplementary Fig. S2l), but cyclin D1 induction increased proliferation regardless of serum levels and promoted density-independence at high serum (Fig. 2h, i). As expected, p27 overexpression strongly repressed proliferation regardless of serum levels and treatment with LATS inhibitor could only restore proliferation in a subpopulation of cells by exceeding the single-cell cyclin D1/p27 threshold (Fig. 2j, Supplementary Fig. S2m). We further verified that YAP inactivation by TEAD inhibition could only partially reduce proliferation of cells overexpressing cyclin D1 by doxycycline induction (Supplementary Fig. S2n, o, with weaker suppression at high serum). Thus, over a wide range of conditions, YAP-regulated proliferation converges on the cyclin D1/p27 ratio with a threshold controlling Rb phosphorylation.
Finally, we also assayed whether YAP-mediated control of the cyclin D1/p27 ratio and proliferation is conserved in additional cell lines. CACO2 cells displayed a consistent dependence on the cyclin D1/p27 ratio across serum and LATS inhibitor conditions (Supplementary Fig. S3a), as did MCF10A, albeit with a previously reported^28^ dependence primarily on p21 rather than p27 (Supplementary Fig. S3b). Of the two breast cancer cell lines assayed, MDA-MB-231 harbor a cyclin E amplification and did not exhibit a clear cyclin D1/p27 threshold for Rb phosphorylation (Supplementary Fig. S3c). In contrast, MDA-MB-468 cells, despite lacking Rb, showed cyclin D1/p27-dependent EdU incorporation, with the LATS inhibitor promoting greater proliferation in all density bins and reducing both p21 and p27 protein levels (Supplementary Fig. S3d, e). MDA-MB-231 proliferation levels were similarly sensitive to LATS inhibition but more strongly dependent on TEAD activity, in line with reports of this cell line harboring a NF2/Merlin mutation^25^, but with stronger regulation of p21 levels rather than cyclin D1 or p27 (Supplementary Fig. S3f).
Together, these data suggest a common mechanism in cells with functioning Rb, whereby YAP induces cell cycle entry and Rb phosphorylation by regulating cyclin D1, p27, and alternatively p21. Rather than the absolute level of cyclin D1 or p27, the crucial parameter for the YAP-regulated control of cell cycle entry is the nuclear cyclin D1/p27 ratio (Fig. 2k) in G1 phase (or cyclin D1/p21 ratio).
YAP regulation gradually changes the nuclear cyclin D1/p27 ratio
Our analysis of cyclin D1 and p27 protein regulation raised the question of whether YAP activity controls their gene expression. We used RNA sequencing (RNAseq) to assess transcriptional control of cell cycle regulators and distinguish between potential signaling pathways through which YAP activation may secondarily regulate cyclin D1 and p27 levels over time. Previous studies suggested different regulatory mechanisms for cell cycle entry, such as direct induction of cyclin D1^29^, cooperative regulation through the serum-dependent AP1 transcription factors^9,30^ and E2F1^10,31^, indirect induction of mitogen signaling^12^, or broadly reprogramming the cell state^2,32^. Therefore, we harvested cells after increasing durations of CA-YAP induction (Fig. 3a) or acute LATSi treatment (Supplementary Fig. S4a), using CDK4/6 inhibition (palbociclib) to minimize E2F activation and cell cycle entry so that the gene expression changes would not be dominated by the transcriptional changes associated generally with S/G2/M progression.Fig. 3. Delayed YAP-mediated control of the cyclin D1/p27 ratio and cell cycle arrest.a Schematic for RNA-sequencing of doxycycline-induced CA-YAP expression in RPE1 cells. Cells were seeded densely (38,000 cells/well) to induce low endogenous YAP activity and pre-conditioned with starvation or serum media and palbociclib-arrested (CDK4/6i, 1 μM), prior to doxycycline-induction over 36 h (0.5 μg/mL DOX). See also “Methods”. b Up-regulated genes (q < 0.05, Log2 fold-change >1) for 12, 24, and 36 h YAP induction (- serum) showed the Cordenonsi YAP signature (using ToppGene)^33^. Log2 fold-change (Log2FC) induction of c CYR61 (canonical YAP target gene, gray) compared with CCND1 (cyclin D1, red) and d regulation of WWTR1 (suppressed YAP target gene TAZ, gray), CDKN1A (p21, dark blue), and CDKN1B (p27, blue) in starved and serum conditions (dashed or solid line, respectively). Adjusted p-values reported in Supplementary Table S3. e Mean YAP ratio in starved WT cells following LATSi treatment (0.5 µM TDI-011536) or co-treatment with TEADi (1 µM GNE-7883). N = 3 independent experiments, not significant (NS) by ordinary one-way ANOVA (p > 0.10 for all time points). f Cyclin D1/p27 single-cell scatterplots (G1-gated), colored by p-Rb state (purple, p-Rb-negative; orange, p-Rb-positive) for cells treated as in (e) with LATSi for 0 h (DMSO), 6 h, 12 h. g Mean cyclin D1/p27 ratio (left) and percent p-Rb (right) for cells treated as in (e). N = 3 independent experiments. Ordinary one-way ANOVA with Dunnett’s multiple comparisons: cyclin D1/p27 ratio, p = 0.14 (3 h), 0.0017 (6 h), 7 × 10^−4^ (12 h), 0.0084 (24 h), 0.92 (TEADi); p-Rb, p = 0.31 (3 h), 0.0044 (6 h), 7 × 10^−4^ (12 h), 0.017 (24 h), >0.99 (TEADi). h Mean YAP ratio in starved CA-YAP cells following TEADi treatment (1 µM GNE-7883) for 0–36 h. Ordinary one-way ANOVA with Dunnett’s multiple comparisons: p < 1 × 10^−4^, all time points. i Cyclin D1/p27 single-cell scatterplots (G1-gated), colored by p-Rb state for cells treated as in (h) with TEADi for 0 h (DMSO), 6 h, and 24 h. j Mean cyclin D1/p27 ratio (left) and percent p-Rb (right) for G1-gated cells treated as in (h). N = 3 independent experiments. Ordinary one-way ANOVA with Dunnett’s multiple comparisons: cyclin D1/p27 ratio, p > 0.99 (3 h), 0.75 (6 h), 0.36 (12 h), 0.0074 (24 h), < 1 × 10^−4^ (36 h); p-Rb, p = 0.68 (3 h), 0.17 (6 h), 0.043 (12 h), 0.0030 (24 h), 1 × 10^−4^ (36 h). k Schematic for computational alignment of asynchronously cycling cells based on mitosis time (M), followed by analysis of CDK2 activity reporter and its bifurcation into CDK_high_ and CDK_low_ fractions for cells born after variable time in DMSO, LATSi, or TEADi (see also “Methods”). l Mean CDK2 activity in cells aligned by mitosis and gated to have received drug in the first 3 h of G1 (left), or in the preceding G2/M of the mother cell (3–6 h before mitosis, middle; 6–9 h before mitosis, right). Shaded error are 2 × SEM (n > 500 cells/condition). The dashed line at 0.8 indicates the CDK_high_ threshold for G1/S progression. m Violin plot distribution of G1 lengths in WT cells treated with DMSO, LATSi, or TEADi during G1 (G1), or in G2/M of the previous cell cycle (M, in the mother cell). G1 cells were gated for birth prior to drug treatment and S-phase entry following drug addition. M cells were gated for birth 6–9 h after drug treatment (during G2/M in the mother cell). Solid lines are median and dotted lines are interquartile range (IQR), n = 1000 random cells/condition. Mood’s median test: p = 0.19 (DMSO), 1.13 × 10^−84^ (LATSi), 7.11 × 10^−27^ (TEADi). n Proportion of cells entering S-phase (by FUCCI reporters), binned by window of time in drug as in (m). o Fold-change in the fraction of cells activating CDK2 within 3 h of mitosis (short G1), binned by time in LATSi or TEADi (bin size = 3 h). N = 3 independent experiments, n > 5000 cells/condition. All single-cell (n = 5000 cells/scatterplot) and live-cell data are representative of 3 independent experiments.
First, we confirmed that the conserved YAP target gene signature^33^ was significantly enriched after doxycycline induction (Fig. 3b). Canonical YAP target genes like CYR61 and ANKRD1 were induced quickly within 6 h with or without serum (Fig. 3c), followed by G1-related target genes CDK6^34^ and SKP2^35^ (Supplementary Fig. S4b), and TEAD and AP1 target genes^36^ (TEAD1/4, FOSL1/2, JUN, Supplementary Fig. S4c). Compared with the fast induction of the YAP target gene CYR61, we observed CCND1 (cyclin D1) expression increasing more gradually and weakly (1.4-fold), similar to LATS inhibition (Supplementary Fig. S4a). In comparison, the previously-reported suppressed TAZ gene WWTR1^37^ rapidly and significantly dropped, whereas the expression of p27 (CDKN1B) and p21 (CDKN1A) changed only weakly (Fig. 3d).
Since there was only a small fold change (<1.5) in cyclin D1 and p27 expression, we compared whether cyclin D1 and p27 are more strongly regulated at the protein level following acute YAP-TEAD activation. As CA-YAP cells fail to arrest by serum starvation and can induce EMT, we used LATS inhibition (LATSi) of WT cells to monitor cell cycle reentry in a more synchronous population following acute YAP induction (Fig. 3e and Supplementary Fig. S4d). Consistent with the induced expression of CDK6 and SKP2 by both CA-YAP and LATSi, we confirmed the graded, protein-level increase peaked at 12 h (Supplementary Fig. S4e). Likewise, LATS inhibition significantly increased cyclin D1 protein levels within 6 h but failed to significantly reduce p27 or p21 levels across the population (Fig. 3f, Supplementary Fig. S4f). Importantly, cells entering the cell cycle over 3–12 h following LATS inhibition progressed by increasing the cyclin D1/p27 ratio, which was paralleled by increasing Rb phosphorylation (Fig. 3f, g) and S-phase entry (12–24 h, Supplementary Fig. S4g). Supporting that LATS1/2 signal through YAP-TEADs, stacked inhibition of TEAD reversed the LATSi-induced changes of these protein and proliferation levels back to control levels (Fig. 3g and Supplementary Fig. S4e–g).
As YAP is thought to be used transiently during injury repair processes rather than constitutively^38^, we tested the reversibility of YAP-mediated control of the cyclin D1/p27 ratio and the termination of proliferation. We first confirmed that acute TEAD inhibition of starved CA-YAP cells, or of LATSi pre-treated WT cells, rapidly and persistently reduced the YAP ratio (Fig. 3h and Supplementary Fig. S4h), consistent with the steady-state decrease observed in the TEAD transcriptional reporter (Supplementary Fig. S1g). However, TEAD inhibition resulted in only a delayed reduction of the nuclear level of cyclin D1 protein (Fig. 3i and Supplementary Fig. S4i). Moreover, TEAD inhibition caused a strong and slow increase of the nuclear protein levels of p27, and cells quickly accumulated p21 protein (Supplementary Fig. S4i, j), which has been reported to stabilize cyclin D1^28^. Concordantly, the mean cyclin D1/p27 ratio and the G1-phase Rb phosphorylation both decreased gradually after a delay, with most cells arresting by 24–36 h (Fig. 3j and Supplementary Fig. S4k). Single-cell analysis confirmed that the cyclin D1/p27 ratio still predicted whether Rb was phosphorylated in G1 as individual cells reduced cyclin D1 and increased p27 levels (Fig. 3i). Similarly graded kinetics for cell cycle arrest were observed for TEAD inhibitor treatment of CA-YAP and LATS-inhibited WT cells in serum conditions (Supplementary Fig. S4l, m).
To measure the delay between LATS and TEAD inhibition and the decision point in G1 dictating entry into the next cell cycle after mitosis, we used live-cell analysis of asynchronously dividing cells expressing a reporter of CDK2 activity^39^ and FUCCI reporters of S/G2^40^. CDK2 activity in cycling cells bifurcates within a few hours after mitosis, distinguishing cells that enter the next cell cycle from cells that stay in or prolong G1 phase^28^. We first computationally aligned cells by the time they exited mitosis to compare subsequent kinetics of cell cycle entry (Fig. 3k) and gated cells based on the window of the cell cycle that they received the drug: early G1 (0–3 h after mitosis), or in the preceding G2 of their mother cell (0–3, or 6–9 h prior to mitosis/birth, Fig. 3l). Confirming fast-initiating effects of the LATS inhibitor, mean CDK activity was rising for cells receiving YAP activation in G1, but with significant shortening of G1 length only with drug treatment in the prior G2 of the cell cycle (Fig. 3m). Conversely, TEAD inhibition in the prior G2 lengthened the subsequent G1 of daughter cells and progressively reduced the proportion of cells entering S-phase compared to DMSO or LATS inhibitor treatment (Fig. 3m, n).
To further determine these changes over time, we measured the fraction of cells with high CDK2 activity at 3 h into G1 phase (cells progressing with short G1-phase), binning them by the duration of LATS or TEAD inhibition (Fig. 3o and Supplementary Fig. S4n, o). We normalized the fraction of cells with high CDK2 activity in inhibitor-treated cells to the fraction in control cells to compare the relative effect of LATS or TEAD inhibition on the timing of cell cycle entry or exit. LATS inhibition increased the fraction of cells activating CDK2 about 6-fold over 24 h, while TEAD inhibition decreased the fraction of cells about 6-fold after a delay (Fig. 3o). We combined this live-cell tracking approach with immunofluorescence and further confirmed that, as the cyclin D1/p27 ratio predicted Rb phosphorylation, CDK2 activation similarly increased with the ratio (Supplementary Fig. S4p). These data support that the cyclin D1/p27 ratio is an important, early event preceding the G1/S transition, though CDK2 activation and stabilization of SKP2 are subsequent steps that further ensure p27 phosphorylation and degradation prior to S-phase entry^41^.
Thus, rather than inducing acute and uniform effects on cell cycle entry, YAP activation and inactivation regulate the cyclin D1/p27 ratio, Rb phosphorylation, and cell cycle entry gradually over a period of 24 h. The slow kinetics of the regulation of cell cycle exit suggests that YAP regulates the cyclin D1/p27 ratio and proliferation through a mechanism integrated over time. We therefore asked what mechanism may mediate this slow acting component of YAP-regulated cell cycle control.
YAP activation reversibly increases RTK-dependent signaling
In addition to low-magnitude transcriptional regulation of cell cycle regulators, our RNAseq time course analysis supported earlier studies that YAP activation increases factors related to receptor signaling^12,42^ (Fig. 4a; and negative feedback pathways, Supplementary Fig. S5a). These factors included receptor tyrosine kinases (RTKs) and secreted ligands known to regulate cell proliferation^2,43^, including ERBB3 and AXL receptors, along with ligands like FGF1/2, and the EGFR ligands NRG1 and precursor HBEGF (paralog of AREG), providing multiple potential multiple receptor pathways. We were intrigued by this possibility since the slow kinetics of the expression changes could provide a second input enhancing mitogen-independence over long-term YAP signaling and for diverse ligands. Using whole-cell lysates, we confirmed that both AXL and EGFR protein levels were dependent on TEAD activity in CA-YAP cells, decreasing upwards of 50% over a similar time scale as the loss of proliferation (Fig. 4b).Fig. 4YAP activation induces and prolongs proliferation by increasing receptor signaling.a Heatmap representation of differential expression (row z-score) of select up-regulated ligand and receptor genes, grouped by 3 independent replicates and time points. b Western blot analysis of EGFR and AXL from whole-cell lysates of serum-starved CA-YAP cells treated with TEADi for 0–36 h, with GAPDH and vinculin loading controls. Values indicate fold-change difference relative to 0 h control (after normalization to loading control). Data are representative of 2 independent experiments. c Percent p-Rb (G1-gated) per local cell density bin for serum-starved WT (left) and CA-YAP (right) cells treated with gefitinib (EGFRi, 0–20 μM) for 24 h. Data are representative of 2 independent experiments. d Percent p-Rb (G1-gated) for serum-starved WT, LATSi-treated WT, and CA-YAP cells treated with 0–10 μM SHP2i (RMC-4550, left) and RASi (BI-2865, right). N = 2 independent experiments, where matched colors indicate data from the same experiment. Ordinary one-way ANOVA: SHP2i, p = 0.49 (WT), 0.012 (LATSi), 0.0024 (CA-YAP); RASi, p = 0.86 (WT), 0.0052 (LATSi), 0.0016 (CA-YAP). e Left: percent p-Rb (G1-gated) per local density bin for serum-starved WT and CA-YAP cells after 14 h EGF (25 ng/mL) or control (CTRL) treatment. Right: mean percent p-Rb (G1-gated) for serum-starved WT and CA-YAP cells treated with control, EGF, or FBS (fetal bovine serum) for 12 h (n = 2 independent experiments). Student’s t test: WT, p = 0.023 (CTRL vs. EGF), 0.020 (EGF vs. FBS); CA-YAP, p = 0.0029 (CTRL vs. EGF), 0.38 (EGF vs. FBS). Binned data are representative of 2 independent experiments, n > 25,000 cells/condition. f Western blot analysis of AKT and ERK phosphorylation levels in whole-cell lysates of serum-starved WT (left), CA-YAP (middle), and TEADi pre-treated CA-YAP cells (24 h, right) after 0–10 h EGF treatment (25 ng/mL). Values indicate fold-change difference relative to 0 h control (after normalization to loading control). g Top: experimental conditions for serum-starvation of WT cells, pre-treatment with LATSi, and sustained or washout of LATSi during 14 h EGF treatment (25 ng/mL). Bottom: Mean YAP ratio (n/c) for treatment conditions. N = 3 independent experiments, matched colors indicate conditions from the same experimental replicate. Ordinary one-way ANOVA with Dunnett’s multiple comparisons to DMSO/CTRL: DMSO, p > 0.99 (EGF); LATSi sustained, p = 0.011 (CTRL), 0.0082 (EGF); LATSi washout, p = 0.99 (CTRL), 0.95 (EGF). h Mean percent p-Rb (G1-gated) for cells treated as in (g). Ordinary one-way ANOVA with Dunnett’s multiple comparisons to DMSO/CTRL: p > 0.99 (EGF); LATSi sustained, p = 0.0039 (CTRL), <0.0001 (EGF); LATSi washout, p = 0.49 (CTRL), 0.012 (EGF). i Percent p-Rb (G1-gated) per local cell density bin for cells treated with DMSO (left), LATSi sustained (middle, solid lines), or LATSi washout (right, dashed lines), with or without EGF as in (g), n > 20,000 cells/condition. j Histogram distribution of phosphorylated-S6 (p-S6 240/244) for DMSO-treated (top) or LATSi-pretreated cells (bottom, LATSi sustained and washout) treated as in (g). Data are representative of 3 independent experiments. All single-cell data represent values from G1-gated cells.
Building on previous reports of YAP-induced expression of EGFR ligands^12^, we first considered whether the residual proliferation observed in serum starved cells at low local density might be induced by YAP through activation of EGFR signaling. While siRNA-mediated knockdown of EGFR levels did not reduce proliferation of WT or CA-YAP cells (Supplementary Fig. S5b), we observed that the small portion of WT cells proliferating in the absence of mitogens could be inhibited by titration of gefitinib, which inhibits active EGFR conformations^44^ (Fig. 4c). While EGFR inhibition in CA-YAP cells could only partially suppress Rb phosphorylation and EdU-incorporation (Fig. 4c and Supplementary Fig. S5c), it restored some degree of contact inhibition. Using another EGFR inhibitor reported to target Erbb2 heterodimers (erlotinib), we observed near-complete proliferation inhibition in both LATS-inhibited and CA-YAP cells (Supplementary Fig. S5c). These data suggest that RPE1 cells, which express much lower levels of EGFR than strongly dependent cells like MCF10A^45^, may have more complex regulation of EGFR/ERBB-family signaling compared to previous reports in MCF10A^12^, while still exhibiting signs of density-dependent signaling.
To test the potential dependence of CA-YAP cells on additional RTKs, we titrated AXL inhibitor (bemcentinib) and observed suppression of proliferation at high doses (Supplementary Fig. S5d). More broadly, we postulated that if cells with high YAP activity employ RTK signaling they would also be dependent on inhibitors targeting key mediators of RTK output, notably the multi-receptor signal transducer SHP2 as well as downstream Ras. Both LATS inhibitor-treated WT and CA-YAP cells exhibited dose-dependent suppression of proliferation (Fig. 4d and Supplementary Fig. S5e, f).
Since total EGFR protein levels were YAP-dependent, we next determined whether YAP activation increases EGF sensitivity. Consistent with previous reports that RPE1 are poorly proliferative in response to EGF alone^46^, exogenous stimulation induced a negligible increase in Rb phosphorylation in starved WT cells (Fig. 4e). However, CA-YAP cells acquired a novel sensitivity to EGF, with EGF inducing similar levels of proliferation as full serum (Fig. 4e). Moreover, this EGF-induced proliferation of CA-YAP cells was strongly dependent on cyclin D1 and YAP/TAZ (Supplementary Fig. S5g). While serum starvation conferred low levels of AKT and ERK phosphorylation to both WT and CA-YAP cells, EGF treatment induced persistently higher phosphorylation in CA-YAP cells (4–10 h, Fig. 4f), which could be rescued by TEAD inhibition, suggesting a difference in signaling duration.
Leveraging the fast reversibility of LATS inhibition, we next tested the persistence of enhanced EGF signaling after transient LATS inactivation. We used WT cells pre-treated with LATSi and then either sustained or washed out the LATSi prior to a 14 h addition of EGF (Fig. 4g). While LATSi washout reduced the YAP ratio back to control levels (Fig. 4g), cells retained sensitivity to EGF for at least 14 h, as evidenced by increased p-Rb (Fig. 4h, i), S6 phosphorylation (p-S6, a readout of PI3K/mTOR activity, Fig. 4j), and nuclear Fra1 levels (a YAP target and also a readout of integrated MEK-ERK activity^47^, Supplementary Fig. S5h). Furthermore, transiently pre-treating cells with LATS inhibitor permitted more Rb phosphorylation following EGF, HGF, or lysophosphatidic acid (LPA) stimulation (Supplementary Fig. S5i), supporting that YAP promotes a state of higher signaling through multiple receptors (not limited to EGFR).
We observed that this persistence of EGF sensitivity could be inhibited through dense seeding, which restored low proliferation levels to LATSi washout cells and reduced the magnitude of EGF-induced p-Rb, p-S6, and Fra1 (Supplementary Fig. S5j–m). MCF10A cells, which proliferate in response to EGF at both confluent and dense seedings, also exhibited enhanced EGF sensitivity with sustained LATSi treatment (Supplementary Fig. S5n, o). However, regardless of seeding density, LATSi washout in MCF10A restored basal levels of EGF sensitivity (Supplementary Fig. S5n–p). Thus, the relative contribution of EGFR and contact inhibition of proliferation signaling following YAP activation may differ in MCF10A cells, which may have additional contact inhibition mechanisms.
We conclude that YAP activity increases the levels and activity of EGFR and other receptors in the absence of added EGF in a cell-autonomous manner^12,43^. We show that YAP activation also sensitizes RPE1 cells to exogenous EGF stimulation. Furthermore, this includes multiple receptors since RPE1 gained HGF and FGF sensitivity. Finally, we showed that increased EGFR signaling persisted in a density-dependent manner following YAP inactivation, suggesting that transient YAP pulses can impart prolonged signaling responses that can be reverted gradually as contact inhibition is restored.
YAP activation requires MEK-ERK signaling to increase the cyclin D1/p27 ratio
To determine the downstream contribution of mitogen signaling, we next examined which YAP-induced pathways control the cyclin D1/p27 ratio. We focused on PI3K-AKT-mTOR and MEK-ERK signaling, since various functions of YAP activation are mediated through mTOR activation^20,48^, and both mTOR and MEK-ERK can regulate cyclin D1 and p27 levels^49,50^ (Fig. 5a). We measured phosphorylation of AKT and S6 protein (p-S6) to monitor PI3K pathway activity and nuclear levels of the transcription factor Fra1 to monitor integrated ERK signaling. Serum-starved CA-YAP cells displayed both elevated p-AKT(S473) and p-S6(240/244) as well as Fra1 levels relative to WT cells (Fig. 5b, c). This aberrantly high signaling without serum depended on YAP-TEAD activity, as TEAD inhibition restored low levels of phosphorylation and protein levels (Fig. 5b–d).Fig. 5YAP activation signals through ERK and mTOR to increase the cyclin D1/p27 ratio.a Schematic of proposed indirect component of cyclin D1/p27 control by YAP/TEAD via mitogen receptor-activated signaling pathways. b, c Top: experimental conditions for serum-starvation of WT and CA-YAP cells followed by 24 h treatment with TEAD inhibitor for 24 h in starvation media. Bottom: Mean fold-change (FC) in protein levels of p-AKT(S473, left) and p-S6(240/244, right) (b) and Fra1 (c) (normalized to WT/DMSO condition). N = 2 (p-AKT) or n = 3 (p-S6, Fra1) independent experiments. One-way ANOVA with Dunnett’s multiple comparisons test to DMSO: p-AKT, p = 0.97 (WT/TEADi), 0.095 (CA-YAP/DMSO), 0.99 (CA-YAP/TEADi); p-S6, p = 0.79 (WT/TEADi), 0.0050 (CA-YAP/DMSO), 0.98 (CA-YAP/TEADi); Fra1, p = 0.94 (WT/TEADi), 0.0013 (CA-YAP/DMSO), 0.38 (CA-YAP/TEADi). d Representative multiplexed images for Fra1 and p-S6(240/244) staining in CA-YAP cells treated with TEADi as in (b). Scale bar = 50 μm. Mean percent p-Rb (G1-gated) for WT and CA-YAP cells treated with PI3K (LY294002, 10 μM), AKT (MK2206, 200 nM), or mTORC1/2 (Torin2, 100 nM) inhibitors (e) or MEK (PD0325901, 100 nM) inhibitor (f) in starvation (left) or serum conditions (right). N = 3 independent experiments. Ordinary one-way ANOVA with Dunnett’s multiple comparisons to DMSO: e −serum, p = 0.052 (WT/PI3Ki), 0.23 (WT/AKTi), 0.012 (WT/mTORi), 0.72 (CA-YAP/PI3Ki), 0.87 (CA-YAP/AKTi), 0.021 (CA-YAP/mTORi); +serum, p > 0.99 (WT/PI3Ki), 0.77 (WT/AKTi), 0.55 (WT/mTORi), >0.99 (CA-YAP/PI3Ki), >0.99 (CA-YAP/PI3Ki), 0.47 (CA-YAP/mTORi); f −serum, p = 0.90 (WT/MEKi), <1 × 10^−4^ (CA-YAP/DMSO), 0.92 (CA-YAP/MEKi); +serum, p = 0.056 (WT/MEKi), 0.0002 (CA-YAP/DMSO), 0.13 (CA-YAP/MEKi). g Mean FC in cyclin D1 protein levels for WT and CA-YAP cells treated with combined mTORi and MEKi relative to WT/DMSO condition as in (b). N = 2 independent experiments, where matched colors indicate values from the same experiment. h, i Top (h): experimental conditions for mTORi or MEKi time course treatment of serum-starved CA-YAP cells. Bottom (h): fold-change in p-ERK (MEKi-treated) and p-S6(240/244, mTORi-treated) following drug treatment. i Fold-change in protein levels of cyclin D1, p27, and p-Rb for MEKi and mTORi treatment. Mean ± SEM, n = 2 independent experiments, 2–3 replicate wells/experiment. Data are G1-gated.
To assess the functional importance of PI3K-AKT-mTOR signaling axis, we inhibited PI3K (PI3Ki, LY-294002) or AKT (AKTi, MK2206), which had only small inhibitory effects on Rb phosphorylation in CA-YAP cells in the absence or presence of serum (Fig. 5e), despite reducing the p-AKT(S473) and p-S6(240/244 and 235/236) to levels similar as WT cells (Supplementary Fig. S6a, b). Nevertheless, inhibiting mTORC1/2 (mTORi, Torin2) caused an approximately 50% reduction in p-Rb (Fig. 5e), but reduced the fraction of cells entering S-phase (Supplementary Fig. S6c, likely due to lengthening of G1-phase). In contrast, inhibition of the obligate upstream regulator of ERK, MEK (MEKi, PD0325901), strongly suppressed Rb phosphorylation and readouts of ERK activity both in the absence and presence of serum (Fig. 5f and Supplementary Fig. S6d). Accordingly, MEK inhibition reduced levels of cyclin D1 and increased p27, substantively blocking S-phase entry in starvation or serum conditions (Supplementary Fig. S6e, f).
Since YAP activation increased both cyclin D1 transcript and protein levels (Fig. 3c, Supplementary Fig. S4f), we anticipated that ERK and mTOR signaling could be coopted to enhance expression for sustaining high levels of cyclin D1. In support of YAP activity depending on both pathways, mTOR or MEK inhibition alone (Supplementary Fig. S6c, e) or the combined inhibition fully reduced the cyclin D1 level in CA-YAP cells to the level of inhibited WT cells (Fig. 5g and Supplementary Fig. S6g). A time course analysis in CA-YAP cells showed that cyclin D1 levels dropped rapidly after mTOR inhibition, but slower and more completely after MEK inhibition even though both inhibitors respectively suppressed p-S6 and p-ERK with equivalent kinetics (Fig. 5h, i). As mTOR can rapidly control translation of cyclin D1 mRNA^50^, this fast regulation was anticipated. However, mTOR inhibition did not alter p27 levels, whereas MEK inhibition both reduced cyclin D1 levels and strongly increased p27 levels over time, explaining why MEK inhibition suppressed YAP-dependent proliferation more completely than mTOR inhibition. Thus, in addition to a direct transcriptional regulation of cyclin D1, YAP transiently enhances the upstream activity of EGFR and other receptor signals that feed into PI3K and MAPK signaling cascades. Receptor signaling then increases the cyclin D1/p27 ratio through regulating MEK-ERK and, to a lesser degree, mTOR signaling pathways.
Partial activation of YAP from ablating Merlin can be reversed by increased contact inhibition
Previous studies suggested that EGFR signaling and contact inhibition of proliferation are also regulated independently of YAP through an upstream Hippo pathway activator that localizes to cell junctions, Merlin/NF2^51^. We focused on Merlin since there have been longstanding questions about its relevance in the control of YAP versus EGFR signaling^52^. From our RNAseq data, we confirmed that Merlin and LATS2 are both induced by YAP activation, representing a negative feedback loop (Supplementary Fig. S7a). Consistent with a critical role in contact inhibition, knockdown of Merlin in WT cells increased proliferation in starved and cycling cells but did not abolish the density dependence (Fig. 6a). We also tested siRNA-mediated knockdown of actin and adherens junction regulators (gelsolin, capZ, cofilin, α-catenin and p120-catenin) reported to reduce contact inhibition in mammary epithelial cells^25^. Interestingly, while α-catenin and p120-catenin knockdown showed partial effects, Merlin knockdown was the most potent inducer of proliferation (Supplementary Fig. S7b–d).Fig. 6. Partial activation of YAP by Merlin knockout can be reversed by increased contact inhibition.a Percent p-Rb (all-cell gating) per local density bin for WT cells transfected with siRNA (48 h) against control (Ctrl), NF2/Merlin (NF2), and LATS2 (LATS2) in starvation (left) or serum (right) conditions. b Median YAP ratio (n/c) binned by local cell density for Merlin knockout (Merlin KO) cells in 0.1% (left) or 10% serum (right), compared with WT cells (same experiment as in Fig. 1e). Data are representative of 3 independent experiments. Shaded error are 95% confidence intervals. c TEAD-dependent hyperproliferation of Merlin KO cells. Percent p-Rb (G1-gated) per local density bin for Merlin knockout cells in 0.1% (left) or 10% serum (right) treated with DMSO or TEADi (0.5 μM GNE-7883), compared with DMSO-treated WT cells. Data are representative of 3 independent experiments. d Percent p-Rb (G1-gated) per local density bin for serum-starved Merlin KO cells treated with 0–20 μM gefitinib (EGFRi). Data are representative of 2 independent experiments. e Left: percent p-Rb (G1-gated) per local density bin for serum-starved Merlin KO cells with control (CTRL) or EGF (25 ng/mL) treatment for 14 h (data are representative of 2 independent experiments). Right: mean percent p-Rb (G1-gated) for starved Merlin KO cells treated with mock (CTRL), EGF, or serum (FBS) release for 12 h (n = 2 independent experiments). Student’s t test: p = 0.046 (EGF versus CTRL), 0.26 (FBS versus EGF). f Percent p-Rb (all-cell gating) per local density bin for Merlin KO (seeded at 11,000 cells/well) and conditioned in 0.1%, 1%, or 10% serum for 36 h. Data are representative of 3 independent experiments. g Normalized cyclin D1 (left) or p27 (right) protein levels for G1-gated Merlin KO cells in titrated serum. Protein levels are normalized to lowest density bin. Data are representative of 2 independent experiments. h Model for Hippo-pathway inhibition of YAP-TEAD activity, which regulates signaling through and downstream of receptor tyrosine kinases to control the cyclin D1/p27 ratio. i Model proposing transitions between low and high YAP-TEAD activity as contact inhibition is temporarily reduced then recovered after a delay during epithelial sheet regeneration. For all representative binned plots, data are pooled from at least 2 wells, n > 100 cells/bin, n > 10,000 cells/plot.
As Merlin was a promising regulator of RPE1 cell proliferation, we generated clonal Merlin knockout RPE1 cells (Merlin KO) using CRISPR/Cas9. Similar to LATS inhibition, Merlin KO strongly increased proliferation at low serum and increased the final cell density of cells conditioned in low or high serum (Supplementary Fig. S7e). Consistent with Merlin functioning primarily in the Hippo pathway, we observed that Merlin KO cells had increased nuclear localization of YAP that could be suppressed by TEAD inhibition (Supplementary Fig. S7f, g). Under both low and high serum conditions, Merlin KO displayed increased YAP ratios relative to WT cells at the same local density but did not ablate contact inhibition of YAP (Fig. 6b). Instead, a higher cell density was required to inhibit proliferation of Merlin KO cells compared to WT, at low and high serum levels (Fig. 6c). Together, these data support a model whereby Merlin restricts proliferation primarily through YAP-TEAD activity. However, distinct from our LATS1/2 and CA-YAP manipulations, YAP activity in Merlin KO cells could still be as effectively inhibited by contact inhibition as in WT cells albeit at a much higher local cell density.
We next examined whether EGFR sensitivity is enhanced with Merlin KO. Like CA-YAP cells, Merlin KO cells displayed EGFR signaling dependence, as evidenced by inhibition of proliferation by the EGFR inhibitor gefitinib (Fig. 6d and Supplementary Fig. S7h). Likewise, Merlin KO cells were more sensitive to EGF yet retained a strong local cell density dependence (Fig. 6e). Importantly, this increased receptor sensitivity was not restricted to EGFR, since we found a similar increase in proliferation in response to other receptor tyrosine kinase ligands (HGF and FGF, Supplementary Fig. S7i). Overall, these data support that persistent YAP induces a broad upregulation of receptor signaling and enhanced signaling in response to various ligands beyond EGF.
Finally, we assessed whether Merlin KO cells conserved the crucial cyclin D1/p27 regulatory control of proliferation. Compared with WT cells (Fig. 1c), Merlin KO shifted contact inhibition of p-Rb across increasing serum levels so that higher densities were needed to suppress proliferation (Fig. 6f). Accordingly, low serum and increasing density still suppressed cyclin D1 and increased p27 levels at higher local cell density (Fig. 6g), suggesting that these cells suppress proliferation through the same mechanism as WT cells: inhibiting YAP activity and downregulating mitogen signaling and the cyclin D1/p27 ratio (Supplementary Fig. S7j). Because of higher YAP activity, Merlin KO cells also maintained cyclin D1 and p27 levels closer to the threshold controlling p-Rb so that manipulating the cyclin D1/p27 ratio greatly increased or decreased proliferation (Supplementary Fig. S7k–m).
We conclude that even though constitutive YAP activation abolished contact inhibition of proliferation, cells can robustly contact inhibit YAP-TEAD even if Hippo signaling is reduced and YAP activation increased. These studies also support that Merlin regulates YAP to control EGFR and other receptor signaling pathways that together contribute to suppress the cyclin D1/p27 ratio.
Discussion
Our study shows that contact inhibition of YAP and mitogen signaling control the cyclin D1/p27 ratio and cell cycle entry through a balance mechanism that shapes an intrinsically transient YAP-induced proliferation response. Our study supports a model where YAP activation controls proliferation at multiple stages: increasing cyclin D1 transcription, expression of RTKs and receptor ligands, and known regulators of mitogen signaling that mediate signal sensitivity and duration. This regulation explains how YAP activation reduces mitogen dependence while simultaneously sensitizing cells to exogenous stimulation by multiple RTK ligands (Supplementary Figs. S5i and S7i), which is likely relevant for the role of YAP activation in the inflammatory injury response^53^. We further show that YAP activation increases downstream signaling through MEK-ERK signaling and, to a lesser degree, mTOR signaling, to control the intracellular balance of cyclin D1 and p27 protein levels (Fig. 5i). Most importantly, our study reveals a predominant role of both cyclin D1 and p27 in YAP-induced proliferation. We showed that YAP-stimulated cells decide during G1 phase whether to enter the cell cycle by increasing the nuclear cyclin D1/p27 ratio above a critical activity threshold, or to remain quiescent by staying below this threshold (Figs. 2d, e and 3f–j). We found that regulation of cyclin D1 alone is insufficient for YAP-mediated loss of contact inhibition and acquired mitogen independence, but that both cyclins and CDK inhibitor proteins must be coregulated for cell cycle entry, and there is critical dependence on control of these components at the protein level.
Moreover, we show that contact inhibition of YAP-TEAD causes a delayed inhibition of mitogen signaling as RPE1 cells recover from transient YAP activation, which lowers the cyclin D1/p27 ratio to terminate the proliferation response. Thus, two delay mechanisms may prolong proliferation: negative feedback of YAP is partial until cell division has sufficiently increased the local cell density, then the resulting contact inhibition of YAP lowers the cyclin D1/p27 ratio and arrests proliferation after a delay (Fig. 6h, i). YAP signals often fluctuate^54,55^, therefore a slow integration of YAP ensures that the proliferation response is robustly started and ended. Slow integration provides a filter mechanism which prevents short-lived YAP activity changes from triggering unwanted cell cycle entry or premature cell cycle exit.
We also evaluated the control of YAP inactivation by ablation of Merlin/NF2, a suppressor of Hippo signaling to YAP. In support that the termination of YAP signaling and proliferation is robust, we find that contact inhibition is merely shifted but not abolished when YAP activity is increased by ablation of Merlin, and cells can still effectively terminate proliferation but at higher local cell density. Moreover, since the loss of Merlin can boost proliferation without a loss of contact inhibition, this result further suggests that Merlin has a unique advantage as a drug target, since a potential Merlin inhibitor would trigger a self-limiting proliferation response. This could offer a safer alternative strategy to LATS inhibition, which was recently revealed to cause a widespread and uncontrolled proliferation response and de-differentiation in mice^4^.
Finally, we identified a crucial role for MEK-ERK in YAP-induced proliferation, which was surprising since oncogenic YAP mutants can drive resistance to KRAS inhibition and KRAS primarily regulates proliferation through MEK-ERK and PI3K/mTOR activation^10,22,31^. Resistance to KRAS inhibitors may therefore by mediated by YAP-mediated direct regulation of cell cycle entry. We confirmed that YAP activation can induce cyclin D1 as well as SKP2, which can degrade p27^41^. Indeed, we suggest that these increases depend on very high levels of CA-YAP activation, since DOX induction of CA-YAP in high serum conditions can still increase cyclin D1 and proliferation when MEK is inhibited (Supplementary Fig. S7n, o). YAP control of the cyclin D1/p27 ratio may therefore be most relevant in two-component signaling contexts, where YAP/TEAD and parallel ERK signaling inputs impinge on AP1 cofactors to cooperatively induce transcription.
Concluding remarks
YAP activation must be transient to drive the epithelial regenerative response without excessive hyperplasia. Our study identifies a delayed, self-limiting control mechanism that allows YAP to trigger such a commensurate proliferation response in 2D contexts. We show that transient proliferation responses are initiated by YAP activation increasing receptor tyrosine kinases like EGFR and mitogen signaling to increase the cyclin D1/p27 ratio above a critical threshold. The delayed contact inhibition-mediated inactivation of YAP then terminates proliferation by gradually lowering mitogen signaling and the cyclin D1/p27 ratio below this critical threshold (Fig. 6i). In contrast, we demonstrate that persistent inactivation of LATS1/2 or activation of YAP activity abolishes contact inhibition and promotes persistent mitogen signaling that keeps the cyclin D1/p27 ratio high, explaining how oncogenic mutations that constitutively inactivate LATS1/2 or activate YAP elicit unrestricted proliferation.
Methods
Cell culture
Experiments were performed with human retinal pigmented epithelial cells (RPE1-hTERT immortalized, ATCC Cat# CRL-4000) unless specified otherwise. RPE1 were cultured in DMEM/F12 growth media with HEPES (Thermo Scientific Cat#11039047), supplemented with 10% fetal bovine serum (Millipore Sigma Cat# F4135-500ML) and passaged using trypsin-EDTA (0.05%, Gibco Cat# 25300054). Human mammary epithelial cells (MCF10A, ATCC Cat# CRL-10317, RRID:CVCL_0598) were cultured in DMEM/F12 growth media with HEPES supplemented with 5% horse serum (Gibco Cat# 16050122), 20 ng/mL EGF (PeproTech Cat# C8052), 0.5 μg/mL hydrocortisone (Sigma Cat#H0888), 100 ng/mL cholera toxin (Sigma Cat# C8052), and 10 μg/mL insulin (Sigma Cat# I1882). Human umbilical vein endothelial cells (HUVEC, Lonza Cat# C2519A) with hTERT immortalization were cultured in EGM-2 Endothelial Cell Growth Medium-2 BulletKit (Lonza Cat# CC-3162). Low serum media was prepared by diluting complete media in EBM-2 Basal Medium (Lonza Cat# CC-3162). Breast cancer cell lines MDA-MB-231 (ATCC Cat# HTB-26) and MDA-MB-468 cells (ATCC Cat# HTB-132, RB-negative) and colorectal cancer cell line CACO2 (ATCC Cat# HTB-37) were cultured in DMEM growth media (Gibco Cat# 11995065) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Thermo Fisher). HEK-293T cells used for lentivirus production were cultured in DMEM growth media (Gibco Cat# 11995065) supplemented with 10% fetal bovine serum. All cells were cultured at 37 °C and 5% CO_2_. For serum starvation experiments, cells were first washed with serum-free media, then maintained in serum-free (0%) or 0.1% serum-containing media, except for HUVEC, which were starved in complete media diluted 1:5, for 36–48 h. 0.1% serum-containing media had similar proliferation as 0% serum-containing media, but RPE1 cells were slightly more adherent to glass-bottom plates for subsequent immunofluorescence protocols and, therefore used for a subset of experiments. For mitogen-release experiments, mitogens and growth factors (EGF, HGF, and FGF at 25 ng/mL and LPA at 1 μM) were diluted in starvation media.
Plasmid generation
pSpCas9(BB)-2A-Puro (PX459) V2.0 was a gift from Feng Zhang (Addgene # 62988, RRID:Addgene_62988). Phosphorylated and annealed sgRNA oligos against NF2 (see Cell line generation below) were cloned into restriction enzyme-digested PX459^56^.
pQCXIH-Myc-YAP-5SA was a gift from Kun-Liang Guan (Addgene #33093, RRID: Addgene_33093). Doxycycline-inducible, myc-epitope-tagged CA-YAP (YAP5SA) was generated using Gibson assembly of PCR-amplified myc-YAP5SA with restriction enzyme-digested plasmid backbone derived from a bicistronic vector with a TetOn promoter driving target gene expression and PGK-driven rtTA expression, pCW-Cas9 (Addgene #50661, RRID: Addgene_50661, a gift from Eric Lander and David Sabatini).
pTRE-8XGTIIC-DsRED-DD was a gift from Joan Massague (Addgene # 115798, RRID:Addgene_115798). The reporter was modified using Gibson assembly of PCR-amplified 8x-GTIIC and restriction-enzyme digested plasmid backbone containing mScarlet-linker-NLS-PEST (a nuclear-localization sequence (NLS) and PEST-destabilization domain^24^ to promote turnover of the reporter).
Cell line generation
Merlin knockout cells were generated by transient transfection of parental RPE1 cells with pX459 (2 ng/µL, containing gRNA scaffolds targeting Merlin/Nf2) and Lipofectamine 2000 (1:500, Thermo Cat# 11668019) in Opti-MEM (Fisher Scientific Cat#31-985-070) and growth media overnight, then exchanged for growth media with puromycin selection (10 μg/mL, Invivogen Cat#ant-pr-1) for 72 h. After a recovery period in selection-free media, cells were diluted to 0.5 cells/100 µL and seeded to 96-well plates for single-cell clonal expansion. 24 single-cell clones were further expanded for preliminary knockout assessment by western blot for loss of Merlin protein expression. 2 clones (Merlin KO-clone 1 forward: CACCGCACAGTGGCCTGGCTCAAAA, reverse: AAACTTTTGAGCCAGGCCACTGTGC; Merlin KO-clone 2 forward: CACCGTGATTTGGTGTGCCGGACTC, reverse: AAACGAGTCCGGCACACCAAATCAC) were selected for validation by genomic DNA purification (Qiagen CAT#69504), PCR of the Merlin locus and sequencing. After initial characterizations confirming Merlin KO-clone 1 and Merlin KO-clone 2 were similarly proliferative (data not shown), we proceeded to use KO-clone 1 for all subsequent experiments in this study.
Lenti-EF1a-dCas9-VPR-Puro (Addgene #99373, RRID: Addgene_99373) was a gift from Kristen Brennard and stably expressed in RPE1 cells using puromycin selection (10 µg/mL, Invivogen Cat#ant-pr-1). CA-YAP and dCas9-expressing cells were generated using third-generation lentiviral transduction. Lentivirus was generated in HEK-293T cells by transfection with packaging and envelope plasmids pMDLg/pRRE (Addgene #12251, RRID: Addgene_12251), pRSV-rev (Addgene #1225, RRID: Addgene_1225), and pCMV-VSV-G (Addgene #8454, RRID: Addgene_8454) in Lipofectamine 2000 and Opti-MEM. After 72 h transfection, supernatant was passed through a 0.22 µM filter (Millipore Cat# SCGP00525) and concentrated using 100 kDa centrifugal filters (Millipore Cat# UFC910024). Lentivirus was then frozen or directly transduced into cells. Seventy-two hours after lentiviral transduction, cells were split into antibiotic-selection media or, if expressing a fluorescent construct, expanded prior to sorting by a BD Influx cell sorter (Stanford Shared FACS Facility, Weill Cornell FACS Facility). RPE1 reporter cell lines were generated by transduction with pLV-H2B-miRFP670 (nuclear marker), CSII-pEF1a-hDHB(994-1087)-mTurquoise (cyclin E/A-CDK reporter), and the FUCCI cell-cycle reporters CSII-pEF1a-mVenus-hGeminin(1-110) (APC/C^Cdh1^ reporter) and pLV-mCherry-hCDT1(1-100)ΔCy (S-phase start^40^), as described previously^57^.
siRNA, sgRNA, and plasmid transfection
All oligo sequences are listed in Supplementary Table S1. Pooled siRNA (ON-TARGETplus, Horizon Discovery Biosci.) were used for siCtrl (Non-targeting Control, Horizon Cat#D-001810-10-05), siCycD1 (siCCND1, Horizon Cat# L-003210-00-0005), sip27 (siCDKN1B, Horizon Cat# L-040178-00-0005), siCFL1/2 (Horizon Cat# L-012707-00, L-019078-00), siGSN (Horizon Cat# L-007775-00), siLATS1/2 (Horizon Cat# L-004632-00, L-003865-00), siCAPZB (Horizon Cat# L-011990-00), siCTNNA1 (Horizon Cat# L-010505-00), siCTNND1 (Horizon Cat# L-012572-00). Previously validated siRNA (IDT) were used for siNF2 (5’-UGGCCAACGAAGCACUGAUUU-3’)^58^, siYAP (5’-GACAUCUUCUGGUCAGAGA dTdT-3’) and siWWTR1 (siTAZ, 5’-ACGUUGACUUAGGAACUUU dTdT-3’)^59^. For CRISPRa-mediated overexpression of genes through targeting of endogenous promoters, pooled sgRNA (Horizon) were used for sgCtrl (non-targeting control, Horizon Cat# U-009500-10-05), sgCycD1 (sgCCND1, Horizon Cat# P-003210-01-0005), and sgp27 (sgCDKN1B, Horizon Cat# P-003472-01-0005) with tracRNA (Edit-R tracrRNA, Cat# U-002005-05). RPE1 cells were transfected with siRNA and sgRNA using DharmaFECT 1 (1:500 in Opti-MEM, Dharmacon Cat# T-2001-03) according to manufacturer protocol using 20 nM siRNA, or sgRNA and tracrRNA (in 1:1 ratio, according to manufacturer protocol). Cells were incubated in transfection mixture for 6–12 h in growth media, then replaced with media appropriate for the experiment conditions and assessed 24–48 h after transfection.
Seeding densities and treatments
For serum titration studies, RPE1 cells were seeded at 8000, 12,000, or 24,000 cells/well (~25,000, 40,000, and 80,000 cells cm^−2^). After 24 h, cells were switched into serum-free (or 0.1% serum) media, 1%, or 10% serum. After 24 h preconditioning, cells received drug or DMSO in the corresponding serum condition. For MCF10A serum titration studies, cells were seeded at 6000 or 24,000 cells/well (~20,000 and 80,000 cells cm^−2^). After 48 h outgrowth, cells were switched into serum-free (0%), 1%, or 5% horse serum (complete media), then received drug or DMSO in the corresponding serum condition for 24 h.
Drugs and mitogens
Human growth factors and mitogens were dissolved in sterile, ultrapure water (Fisher Scientific Cat#10977023) and used at the following concentrations unless otherwise stated: 25 ng/mL recombinant EGF (PeproTech cat#AF-100-15), 25 ng/mL recombinant HGF (Life Technologies Cat#PHG0254), 25 ng/mL recombinant FGF basic (R&D Systems Cat#233-FB-025), 1 µM LPA (oleoyl-L-alpha-lysophosphatidic acid, Sigma-Aldrich Cat#L7260-1MG).
Drugs were dissolved in DMSO (Sigma-Aldrich Cat# D2650) and used at the following concentrations unless otherwise stated: 0.5 μM TDI-011536 (Selleck Chemicals Cat#E1314, CAS#2687970-96-1, LATSi), 5 μM TRULI (a gift from James Hudspeth, CAS#1424635-83-5), 1 µM GNE-7883 (ChemieTek Cat#CT-GNE7883, CAS# 2648450-42-2,TEADi), 10 μM gefitinib (Selleck Chemicals Cat#S1025, CAS#184475-35-2, EGFRi), 0–20 µM erlotinib (Selleck Chemicals Cat#S1023, CAS#183321-74-6, EGFRi), 0–20 µM bemcentinib (MedChem Express Cat#HY-15150, CAS#1037624-75-1, AXLi), 0–20 µM RMC-4550 (MedChem Express Cat#HY-116009, CAS#2172651-73-7, SHP2i), 0–20 µM BI-2865 (MedChem Express Cat#HY-153724, CAS#2937327-93-8, RASi), 10 µM LY294002 (Cayman Chemical Cat#70920, CAS#154447-36-6, PI3Ki), 200 nM MK2206 (Cayman Chemical Cat#11593, CAS#1032350-13-2, AKTi), 100 nM Torin2 (Cayman Chemical Cat#14185, CAS#1223001-51-1, mTORi), 100 nM PD0325901 (Selleck Chemicals Cat#S1036, CAS#391210-10-9, MEKi), 1 μM palbociclib (PD-0332991, Selleck Chemicals cat#S1116, CAS#827022-32-2, CDK4/6i), 0.5 µg/mL doxycycline hyclate (Sigma-Aldrich cat#D9891, CAS#24390-14-5, DOX), 10 μM 5-ethynyl-2’-deoxyuridine (EdU, Cayman Chemical Cat#20518, CAS# 61135-33-9). For drug treatments and time courses, control conditions were treated with DMSO, except mitogen release controls (mock), which were dissolved in water. For all EdU-incorporation, cells received an EdU spike-in (10 µM, in serum-free media) for 12–15 min prior to fixation.
Western blotting
Cells grown in 6-well plates were washed once in cold, sterile PBS and lysed in chilled, 1× RIPA buffer (Boston BioProducts Cat#BP-115-5X) or non-denaturing lysis buffer (20 mM HEPES, 150 mM NaCl, 2 mM EDTA, 0.1% IGEPAL-CA630) containing HALT protease and phosphatase inhibitor cocktail (1:100, Thermo Fisher Cat#78439). Lysates were harvested on ice using cell scrapers, sheared by syringe (25 G) and centrifuged at 15,000 RCF for 15 min at 4 °C. Supernatant was stored at −20 °C. Protein concentration was calculated by Pierce BCA Protein Assay (Thermo Fisher Cat#23225) according to manufacturer protocol and a BSA standard curve.
10–30 µg protein was loaded onto 7.5%, 12%, or 4–20% Mini-PROTEAN TGX gels for resolution by SDS-PAGE in 1× Laemmli sample buffer (Fisher Scientific, Cat#AAJ60660AC, or Bio-Rad Cat#1610737) alongside Page RulerPlus (Thermo Scientific Cat#26620) molecular ladders in Tris/Glycine/SDS running buffer (Bio-Rad Cat#161-0772). Gels were run at 150 V for 1 h then transferred to 0.2 µm PVDF membranes by semi-dry transfer (Bio-Rad Trans-Blot SD, Cat# 1703940) in Tris/Glycine buffer (Bio-Rad Cat# 1610734) with 10% methanol, or semi-dry transfer (Bio-Rad Trans-Blot Turbo Cat# 1704150) with Trans-Blot Turbo Transfer Packs (Bio-Rad Cat# 1704156). Membranes were washed in TBST (20 mM Tris, pH 7.5, 150 mM NaCl, 0.1% Tween-20), blocked for 30 min in 5% milk/TBST (VWR Cat#10128-602), and incubated at RT for 3 h or overnight at 4 °C in primary antibody dilutions in 5% BSA/TBST + 0.01% NaN_3_. Primary antibodies used were rabbit anti-AXL (1:1000, Cell Signaling Technology Cat# 8661, RRID:AB_11217435), rabbit anti-EGFR (1:1000, Cell Signaling Technology Cat# 4267, RRID:AB_2246311), rabbit anti-GAPDH (1:1000, Cell Signaling Technology Cat# 5174, RRID:AB_10622025), mouse anti-Vinculin (1:1000, Thermo Fisher Scientific Cat# 14-9777-80, RRID:AB_2573027) (see also Supplementary Table S2).
After 2× washes in TBST, membranes were incubated with anti-rabbit or anti-mouse HRP-conjugated secondary antibodies (1:5000, CST#7074, RRID: AB_2099233 or CST#7076, RRID: AB_330924) or fluorophore-conjugated secondary antibodies (1:10,000, LI-COR Biosciences, IRDye 680 LT CAT#925-68020, or IRDye 800CW, CAT#925-32210) in 5% BSA/TBST for 1 h at RT. Membranes treated with enhanced chemiluminescence substrate (Thermo Fisher Cat#PI34080) were detected by film (Thomas Sci. Cat#EK-5130) or Licor Odyssey Fc. Membranes treated with fluorophore-conjugated secondary antibodies were imaged on a Licor Odyssey Fc. Normalized protein level changes were calculated using FIJI^60^ by measuring the integrated intensity of a set box size encompassing the protein band. The background signal was measured by applying the same box to an adjacent region of the blot and subtracted from the median protein intensity. An equivalent box size was used for reference/loading control protein bands as the protein of interest.
Immunofluorescence
Cells were plated on bovine fibronectin-coated (1:75 in PBS, Sigma-Aldrich F1141, CAT#F1141) or collagen-coated (1:100 in PBS, Advanced Biomatrix Cat#5005-B), 96-well, glass-bottom plates (Cellvis CAT# P96-1.5h-N). Cells were fixed in 4% PFA (in PBS, Fisher Scientific Cat#AA433689) for 15 min at room temperature (RT), washed 3x with PBS, permeabilized in 0.2% Triton X-100 for 15 min, and blocked in blocking solution (10% FBS, 5% BSA, 0.1% Triton X-100, 0.01% NaN_3_ in PBS) for 1 h at RT. For conditions with EdU incorporation (10 μM in basal medium, 12–15 min at 37 °C, Cayman Chemical Cat#20518) prior to fixation, cells were washed 1× with PBS after the blocking step and treated with a click chemistry solution 2 mM CuSO4, 20 mg/mL sodium ascorbate, 3 µM AFDye picolyl azide (488, 568, or 647, Click Chemistry Tools Cat# 1276, 1292, or 1300 respectively) in TBS (50 mM Tris, 150 mM NaCl, pH 8.3) for 25 min. Following a 1× PBS wash, cells were incubated with primary antibody for 3 h at room temperature or overnight at 4 °C (see also Supplementary Table S2) in blocking solution, then washed 3× in PBS. Fluorophore-conjugated secondary antibodies (Thermo Fisher Scientific, goat anti-mouse, goat anti-rabbit, 1:1000 in blocking solution) and Hoechst 333342 (Thermo Fisher Scientific Cat#H3570, 1:10,000) were incubated on cells for 1 h at RT, then washed 3× in PBS. Wells were imaged in 1× PBS or 1× N-acetylcysteine (NAC, 700 mM in ddH2O, pH 7.4, Sigma-Aldrich A7250) imaging buffer for iterative immunofluorescence protocol^27^ (4i).
For the 4i protocol, wells were then washed 1× in dH2O, treated with elution buffer (0.5M L-glycine (Sigma-Aldrich Cat#G2879), 3 M urea (Sigma-Aldrich Cat#U5378), 3 M guanidine hydrochloride (Sigma-Aldrich Cat# G3272), 70 mM TCEP-HCl (Goldbio Cat# TCEP50) in ddH2O, pH 2.5) for 20 min on an orbital shaker. After 1× PBS wash, elution was confirmed by fluorescence microscopy. Cells were then blocked with 4i blocking solution (1% BSA + 150 mM maleimide (Sigma-Aldrich Cat# 129585) in PBS) for 30 min, then incubated with the second round of primaries (in blocking solution) for at least 3 h RT or overnight at 4 °C. Secondary staining, imaging in NAC, and subsequent elution and 4i blocking steps were repeated up to 5 times. Control wells with secondary antibody treatment alone were used to confirm proper elution of previous antibody rounds. Antibody list and dilutions used for immunofluorescence are listed below and in Supplementary Table S2.
Immunofluorescence antibodies
The following primary antibodies were used exclusively for round 1 of immunofluorescence staining: rabbit anti-CDK6 (1:1000, Abcam Cat#ab124821, RRID: AB_10999714), rabbit anti-YAP (1:400, Cell Signaling Technology Cat# 14074, RRID:AB_2650491), mouse anti-p27 (1:1000, Cell Signaling Technology Cat#3698, RRID:AB_2077832), mouse anti-p21 (1:500, BD Bioscience Cat#556430, RRID:AB_396414), rabbit phospho-AKT(S473)(1:400, Cell Signaling Technology Cat#4060, RRID:AB_2315049), rabbit phospho-MAPK1/2(Thr202/Tyr204)(1:400, Cell Signaling Technology Cat#4370, RRID:AB_2315112), mouse phospho-S6(235/236)(1:800, Cell Signaling Technology Cat#62016, RRID:AB_2799618).
The following primary antibodies were used for any round of immunofluorescence: mouse anti-cyclin D1 (1:200, Santa Cruz Biotech Cat#sc-8396, RRID:AB_627344), mouse anti-FRA1 (1:200, Santa Cruz Biotech Cat#sc-28310, RRID: AB_ AB_627632), rabbit anti-Myc (1:800, Cell Signaling Technology Cat#5605, RRID:AB_1903938), mouse anti-Myc tag (1:8000, Cell Signaling Technology Cat#2276, RRID:AB_331783), mouse anti-N-cadherin (1:200, Santa Cruz Biotech Cat#sc-393933, RRID: AB_2832921), rabbit anti-p27 (1:1600, Cell Signaling Technology Cat#3686, RRID:AB_2077850), rabbit anti-p21 (1:2500, Cell Signaling Technology Cat#2947, RRID:AB_823586), rabbit anti-phospho-Rb(Ser807/811)(1:2500, Cell Signaling Technology Cat#8516, RRID:AB_1117658), rabbit anti-phospho-S6(240/244)(1:2500, Cell Signaling Technology Cat#5364, RRID:AB_10694233), rabbit anti-Skp2 (1:800, Cell Signaling Technology Cat#2652, RRID:AB_11178941), mouse anti-YAP (1:200, Santa Cruz Biotech. Cat#sc-101199,RRID: AB_1131430). Antibodies were previously validated in RPE1^15^ and for 4i^27^, and in this study by drug treatment (including AKT, mTORC1/2, MEK, and LATS1/2 inhibitors).
The following secondaries were used in variable combination (488, 514 or 568, and 647), at 1:1000: goat anti-mouse IgG Alexa Fluor 488 (Thermo Fisher Scientific Cat#A11029, RRID:AB_2534088), goat anti-mouse IgG Alexa Fluor 514 (Thermo Fisher Scientific Cat#A31558, RRID:AB_1037789), goat anti-mouse IgG Alexa Fluor 568 (Thermo Fisher Scientific Cat#A-11031 RRID:AB_144696), goat anti-rabbit IgG Alexa Fluor 568 (Thermo Fisher Scientific Cat#A11036, RRID:AB_1056366), goat anti-rabbit IgG Alexa Fluor 647 (Thermo Fisher Scientific Cat#A21245, RRID:AB_23813).
Microscopy
For automated epifluorescence microscopy, cells in glass-bottom 96-well plates were imaged on a Ti2-E inverted microscope (Nikon) using penta-band (DAPI/CFP/YFP/TRITC/Cy5, Chroma: 89405) Sedat filter sets with an LED light source (Lumencor Spectra X or Spectra III), Hamamatsu ORCA-fusionBT or ORCA-Flash4.0 V3 sCMOS camera, and 20x objective (Nikon CFI Plan Apo Lambda, 0.75 NA) in a humidified chamber at 37 °C and 5% CO2. Twenty-five to thirty tiled images were acquired per well in 16-bit mode with 2 × 2 binning. Fixed cells were imaged in 1x N-acetylcysteine imaging buffer. Fluorophores for secondary antibodies were previously optimized to minimize bleed-through^57^. For live-cell imaging, 9 sites were imaged per well and stitched into a single image to maximize the number of cells that could be tracked over time (with at least 3 well-replicates per experiment). Images were acquired at 12 min intervals of with the minimum laser intensity and light exposure to limit phototoxic effects and automated shading correction for each channel based on the autofluorescence of media containing 10% serum. For LATS1/2 and TEAD inhibitor experiments, asynchronously cycling cells were imaged in DMSO for 12 h, then acutely treated with inhibitor and imaged for 24 h to assess cell cycle entry and exit.
RNA sequencing experimental conditions
We measured gene expression changes up to 36 h following CA-YAP induction (as these cells would retain target genes without negative feedback from the Hippo pathway) and up to 6 h following acute LATS inhibition in WT cells. WT and doxycycline-inducible YAP5SA cells were seeded at 38,000 cells per well of 96-well, optical-grade plates (Sigma-Aldrich Cat#CLS3904) in 1.0% FBS. After ~2 h to allow time for adherence to the tissue culture surface, cells were washed with 0.1% FBS then pre-arrested with CDK4/6 inhibitor (palbociclib, 1 µM, SelleckChem CAT#S1116) in either 1.0% FBS or 0.1% FBS (±serum) for 28–36 h prior to doxycycline (500 ng/mL, Sigma-Aldrich CAT#D9891) induction, LATS inhibition (5 µM TRULI, 3 or 6 h), or DMSO control treatment for 6, 12, 24, or 36 h in the presence of CDK4/6 inhibitor to maintain cells in G0 and prevent additional transcriptional variation due to E2F activation and cell cycle progression. Each experiment (repeated on three separate occasions) was performed in technical duplicate (with 6 replicate wells per technical duplicate), reserving one replicate for immunofluorescence staining to confirm quality of cell cycle arrest (by DNA content and EdU incorporation) and the other for harvesting RNA.
RNA-seq and analysis
Prior to harvesting, 96-well plates were washed with PBS, aspirated fully and transferred immediately to −80 °C. RNA was later harvested from the plates by briefly thawing prior to using extraction by RNeasy Mini kit (Qiagen Cat# 74106) and QIAshredder (Qiagen Cat#79654), with pooling of six technical replicate wells/condition directly following the lysis step prior to following the manufacturer protocol, including the optional DNase treatment (Qiagen Cat# 79254). Samples were eluted in ultraPure ddH2O and assessed for quality using an Agilent Bioanalyzer TapeStation. Illumina Stranded mRNA library preparation with polyA capture and sequencing was performed on a NovaSeq6000 system (paired-end, 2 × 50) at the Weill Cornell Genome Resources Core Facility over two rounds to get sufficient read depth.
Raw reads were checked for quality using FASTQC, then aligned to the human genome assembly (GRCh38) with HISAT2^61^ and assigned to genomic features with featureCounts^62^. Differential expression analysis was conducted with DESeq2^63^.
Image quantifications and analysis
Tracking of cells for time-lapse experiments, illumination bias correction, nuclear segmentation, image cropping and alignment across iterative rounds of immunofluorescence, and analysis were performed with a custom MATLAB pipeline (2020b, MathWorks, previously published^57^ and deposited on Github: https://github.com/MeyerLab/image-analysis-ratnayeke-2021) incorporating local density measurements^15^. Cell nuclei were automatically segmented using expression of a fluorescently-tagged nuclear marker (H2B-miRFP670 or H2B-mTurquoise) or Hoechst staining, a Gaussian filter, and a curvature-based object splitting algorithm.
Time-lapse analysis
Nuclei were tracked over time using a nearest-neighbor algorithm, as previously reported, with a jitter correction applied between frames in which drugs were added. Mitosis time was called based on the emergence of two nuclei during anaphase, and S phase entry using the CRL4^Cdt2^ Geminin-based APC/C^Cdh1^ reporters^40^ (FUCCI cell cycle reporters degraded upon CRL4^Cdt2^ activation and inactivation of APC/C^Cdh1^, respectively).
For analysis of CDK2 activity following mitosis in cells with variable time spent in LATS1/2 or TEAD inhibitor (Fig. 3k–o), high-quality cell traces were first gated based on cells that were tracked for at least 20% of the total imaging time (36 h) and tracked from the time of birth or mitosis (denoted as a daughter cell). CDK2 activity of the kinase translocation reporter was measured as a ratio of the median intensity of a cytoplasmic ring (0.65–3.25 µm expansion from the nuclear mask) over the median intensity of the nuclear signal. To measure mean CDK2 activity over time since mitosis, daughter cells traces were aligned by time of mitosis (birth) and then gated based on when they received drug relative to mitosis according to representative time windows: whether in G1 (after birth), 0–3 h or 6–9 h prior to mitosis in the mother cell (before birth). G1 length (Fig. 3m) was calculated between cells that were born after drug addition (and only entered S-phase after drug addition, “G1”) and cells that were born 6–9 h after drug addition (i.e., mother cell received the drug in G2/M, “M”). Only cells that entered S phase within 15 h of mitosis were analyzed, to prevent the skew of data from subpopulations of arresting cells.
For calculating the fraction of cells with high CDK2 activity, the asynchronously cycling cells were first binned into 3 h windows by time of mitosis (to ensure enough cells analyzed per time window of mitosis) and then gated by time spent in the inhibitors by selecting cells that were born at variable times following the switch from DMSO to the LATS or TEAD inhibitor. The fraction of cells with CDK2 activity above a standardized threshold (0.8) was calculated 3 h after mitosis as an early indicator of cells bifurcating into low or high activity^28^, but similar results were also seen if assessing CDK activity at 6 h after mitosis (data not shown).
Immunofluorescence analysis and image-based cytometry
Cell nuclei of fixed cells were detected by Hoechst staining and automatically segmented as described above. For illumination bias correction, images of background autofluorescence from immunofluorescence blocking solution in cell-free wells were taken for each experiment and used to flatten raw images. A segmentation mask from the first round of imaging was used to align subsequent imaging rounds and crop the images for the overlapping regions. Image jitter between 4i rounds was typically less than 500 nm.
For nuclear-localized protein stains (EdU, cyclin D1, p27, phospho-Rb, c-Myc, Fra1, SKP2, CDK6, p21), global background subtraction was performed automatically by expanding a 7.8 μm ring from the nuclear mask and subtracting the 25th percentile of pixel signal as estimated background. For protein stains with cytoplasmic and nuclear localization (YAP, phospho-S6, phospho-AKT, phospho-ERK, N-cadherin), median background signal was estimated during analysis by manual averaging of background pixels of representative cell-free regions and subtracting as a flat threshold. For quantification of immunofluorescence signals, DNA content was calculated as the total nuclear Hoechst intensity (median intensity × nuclear area); nuclear fluorescence signals were calculated as the median nuclear intensity; cytoplasmic fluorescence signals were calculated as the median cytoplasmic intensity (expanding the nuclear mask with a ring of inner radius 0.65 µm and outer radius 3.23 µm to minimize the chance of sampling cytoplasm of neighboring cells). To compare experimental replicates, fold-change in immunofluorescence staining was used to account for the variation in staining and real fluorescence unit values, which also varied based on the LED light source of the microscope used.
Cell gating for G1-phase cells was performed by distinguishing G0/G1 and G2 cell populations by 2 N or 4 N DNA context, then selecting cells below the EdU-incorporation threshold (see sample EdU histograms, Supplementary Fig. 1a) with 2 N DNA. Otherwise, for calculating percent EdU-positive or percent Rb phosphorylated (all-cell gating), cells across cell cycle phases were analyzed together. Rb phosphorylation was defined by histogram analysis of the bifurcated phospho-Rb stain and calculating the fraction of cells above the threshold cutoff (see sample p-Rb histograms, Supplementary Fig. 1a).
For calculating representative, experiment-specific cyclin D1/p27 threshold lines, a classification support vector machine algorithm (MATLAB, fitcsvm) with a linear kernel function was trained on cyclin D1 and p27 protein levels, and binary p-Rb status for G1-gated cells. 10,000 randomly sampled cells were pooled across conditions representing both Rb-negative and Rb-positive states to robustly detect the boundary line separating quiescent and proliferating subpopulations. Cells were further gated for cyclin D1 and p27 values within 3 standard deviations of the mean to filter out disproportionately bright fluorescent signal from dust and particulates that can accumulate through the multiday, iterative immunofluorescence protocol. For experiments with p21 staining, cells with high p21 (based on a histogram of all p21 levels) were gated out to account for the co-stabilization of cyclin D1 in p-Rb-low cells^28^. The calculated boundary lines varied for separate experiments, according to previously reported differences in antibody epitope binding, laser source power across microscopes, variability in immunofluorescence staining, as well as real experimental variability^15^. We therefore compared linear thresholds only for cell conditions within the same experiment and used fold-change effects for comparing protein level changes across experimental replicates.
Local cell density analysis
Local cell density analysis was adapted from Fan et al. Briefly, local density was measured by dividing imaging sites (25–30 tiled images/well) into a 3D histogram based on the centroid coordinates of segmented nuclei (10 × 10 grid, MATLAB hist3). The xy location coordinates of each cell centroid were rounded, and the number of cells per block was automatically counted, then scaled back to match the original image size (giving numerical output rather than integer values). The local cell density value assigned for each cell was based on the scaled number of cells in its assigned local density block.
Quantification corrections
While cells were counted prior to seeding, final cell plating density varied across experiments, and absolute proliferation levels varied per independent experiment. We therefore assessed both absolute (% EdU, % p-Rb) and fold-change proliferation metrics for assessing drug and condition effect sizes.
During the iterative steps of multiplexed immunofluorescence, some cells washed off during later rounds of imaging. As a proxy for null cells, cells that were segmented from the first round of imaging were gated out of the final analysis if their Hoechst value was less than 100 RFU for a subsequent imaging round. As all staining and wash steps were conducted with multichannel pipets, occasionally there were sporadic wells which did not receive antibody due to clogged or defective tips. These wells were excluded from analysis based on visual inspection of background staining relative to successfully stained technical replicate wells.
For live-cell imaging experiments, some cells lost expression of the reporters of cell cycle activity and were therefore removed from analysis if expressing less than a threshold value.
Statistics and reproducibility
Statistical analyses were performed using Student’s one-sample or paired (MATLAB ttest) and two-sample (MATLAB ttest2) t tests, and Mood’s median test (MATLAB). Ordinary one-way ANOVA and ANOVA with Dunnett’s multiple comparisons were performed in Prism (GraphPad). Linear regressions for data binned by local cell density were performed using fitlm (MATLAB). 95% confidence intervals and SEM are plotted as indicated for representative experiments and summary data. For violin plots with single-cell data, a random subsampling of 1000 cells/condition were shown. All experiments were performed as 2-3 independent biological replicates with additional well-replicates (>2) for each experiment, which were pooled for analysis of representative data. Sample sizes for each biological replicate of live-cell experiments were at least 500 cells/condition. While cells were counted prior to seeding for each experiment, we observed technical variation in seeding density across biological replicates that contributed to variation in proliferation rates and final cell densities.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Resource availability
Immunofluorescence protocols and cell lines are available upon request to the lead contact. Plasmids generated in this study (pCW-myc-YAP5SA-puro, Addgene # 249725; pLV-8xGTIIC-mScarlet-NLS-PEST, Addgene # 249724) have been deposited at Addgene.
Supplementary information
Supplementary Information Description of Additional Supplementary Materials Supplementary Data 1 Reporting Summary
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Gut, G., Herrmann, M. D. & Pelkmans, L. Multiplexed protein maps link subcellular organization to cellular states. Science 361, eaar 7042 (2018).10.1126/science.aar 704230072512 · doi ↗ · pubmed ↗
- 2Su, T. et al. Two-signal requirement for growth-promoting function of Yap in hepatocytes. e Life 4, e 02948 (2015).10.7554/e Life.02948 PMC 436387825667983 · doi ↗ · pubmed ↗