Differential YAP expression in glioma cells induces cell competition and promotes tumorigenesis

Malignant cells within individual tumors display striking heterogeneity in their cellular morphology, proliferation rate, genetic lesion, epigenetic modification and therapeutic response. Such intratumor heterogeneity is a recurrent feature of most human tumors, and is associated with progression to malignancy (Andor et al., 2016; Heppner, 1984; Marusyk et al., 2012). Intratumor heterogeneity may account for a substantial portion of observed tumor relapses as well as accompanied resistance to initial therapies (Dexter and Leith, 1986). Currently, two major concepts that attempt to explain the origin of this heterogeneity are the cancer stem cell hypothesis and the clonal evolution model (Campbell and Polyak, 2007). The conceptual difference between these two views could have distinct clinical implications. Besides tumor origin, biological interactions between malignant cells within a tumor compose another key issue of intratumor heterogeneity, and these interactions have been implied to be able to significantly modulate tumor progression and therapeutic response (Calbo et al., 2011; Cleary et al., 2014; Costa et al., 2015; Merlo et al., 2006). The nature of these interactions, and how they affect tumor progression as well as therapeutic outcomes are largely unexplored experimentally.

Glioblastoma (GBM) is the most aggressive primary brain tumor. Gene expression analysis has allowed GBM to be classified into several subtypes differing in treatment responses and survival rates (Phillips et al., 2006; Verhaak et al., 2010). Among these subtypes, the mesenchymal group associates with the worst prognosis (Phillips et al., 2006; Wang et al., 2017). The genetic alterations leading to these differential gene expression and prognosis profiles are not fully understood. Transcriptional coactivator with PDZ-binding motif (TAZ, also known as WWTR1) was proposed to be one of the transcriptional regulators driving the gene expression program of mesenchymal differentiation (Bhat et al., 2011). TAZ and its paralog protein, Yes-associated protein (YAP, also known as YAP1), are the two paralogous nuclear effectors of the Hippo signaling pathway (Zanconato et al., 2016). A comprehensive analysis of brain tumor samples by immunohistochemistry found that YAP expression is increased in a variety of human brain tumors, especially in infiltrating astrocytomas, oligodendrogliomas and GBM (Orr et al., 2011). Remarkably, YAP expression is significantly higher in the mesenchymal subtype of GBM, and this higher expression is also found in more aggressive gliomas associated with poor prognosis (Orr et al., 2011). Considering that both TAZ and YAP have been linked to GBM aggressive progression, it is likely that the transcriptional program controlled by them is responsible for driving GBM mesenchymal transformation.

Like other tumors, intratumor heterogeneity commonly exists in GBM (Morokoff et al., 2015; Parker et al., 2015; Soeda et al., 2015). In addition to the heterogeneous properties that have been described above, the GBM subtypes distinct in gene expression also display heterogeneity (Sottoriva et al., 2013). This suggests that the transcriptional regulators controlling these gene expression profiles might have heterogeneous activities (Sottoriva et al., 2013). In this study, we examined the expression of YAP and TAZ in samples collected from different regions within the same individual human GBM tumors (Sottoriva et al., 2013), and found that the expression of YAP, but not TAZ, is heterogeneous. To functionally study the heterogeneity of YAP expression in GBM, we constructed lines of human GBM cells (LN229 glioma cells) expressing YAP at differential levels and used them to develop tumorigenic models. With these models, we studied the interaction of these heterogeneous cells and its impact on tumorigenesis. Our studies suggested that differential YAP expression induces a competitive interaction, which promotes tumorigenesis.

To examine whether the expression of YAP and TAZ is heterogeneous in GBM, we analyzed the gene expression dataset of 27 samples collected from five GBM patient tumors (5 samples/tumor for 4 tumors, and 7 samples/tumor for 1 tumor; Fig. 1A and data not shown) (Sottoriva et al., 2013). Besides intertumorally heterogeneous expression (2.5-fold difference in maximum gene expression, tumors sp42 versus sp52; Fig. 1A), we found that the expression of YAP also displays intratumoral heterogeneity (1.8-, 1.4-, 2.1-, 2.3- and 1.6-fold increase from minimum to maximum expression in tumors sp42, sp49, sp52, sp54 and sp55, respectively, P≤0.01 for each tumor; Fig. 1A). In contrast, the intratumoral heterogeneity of TAZ expression is not significant (1.5-, 1.2-, 1.3-, 1.6- and 1.5-fold increase from minimum to maximum expression in tumors sp42, sp49, sp52, sp54 and sp55, respectively, P>0.05; data not shown). This result indicated that the expression of YAP in GBM is heterogeneous.

To examine the impact of the heterogeneous expression of YAP on tumorigenesis, we stably expressed recombinant YAP in ZsGreen (G)- or mCherry (R)-expressing LN229 human glioma cells. These cells are denoted C_YAP-G (YAP-G in figures) and C_YAP-R (YAP-R in figures) hereafter, with results combined as C_YAP. In parallel, the differentially labeled cells were transduced with vector alone and were denoted C_Vector-G (Vector-G in figures) and C_Vector-R (Vector-R in figures) hereafter, with results combined as C_Vector. By this method, C_YAP cells express higher levels of YAP compared to C_Vector cells (Fig. 1B). Differentially labeled C_YAP and C_Vector were mixed together (1:1) and co-injected subcutaneously into nude mice to form tumors. Two control experiments were conducted in parallel. In one control, C_YAP-G and C_YAP-R alone were premixed (1:1) and co-injected, while in the other control C_Vector-G and C_Vector-R alone were used. Growth of these tumors was assessed by measuring tumor sizes. Interestingly, although tumors containing C_YAP only grow into tumors at a slightly increased speed comparing to those containing C_Vector only, tumors arising from the mixtures of these two populations grow much faster than either population alone (Fig. 1C). We collected these tumors and analyzed their cellular compositions using flow cytometry. In tumors arising from C_YAP or C_Vector cells alone, differentially labeled cells appear to equally contribute the tumor mass. However, in tumors arising from the mixtures of C_YAP and C_Vector, the C_Vector population is consistently smaller than the C_YAP one (Fig. 1D,E). These results suggested that cells expressing YAP at a higher level become dominant over those expressing less YAP during tumorigenesis. Certain interaction during the clonal dominance may promote tumorigenesis.

To further investigate the interaction between glioma cells expressing YAP at different levels, we co-cultured the differentially labeled C_Vector and C_YAP cells under regular two-dimensional (2D) conditions using the four indicated strategies (Fig. 2A), and compared the growth of each population. Before reaching confluence (day 1–6 after seeding), we found that C_YAP cells proliferated at a similar speed to C_Vector when cultured separately (co-1 and co-2, Fig. 2B,C). After day 6, the growth of both C_YAP and C_Vector cells slows down, presumably due to contact inhibition. Under these conditions, C_YAP cells reached a slightly higher density than C_Vector cells before growth stopped (Fig. 2B,C). In contrast, when C_YAP cells were grown together with C_Vector cells (co-3 and co-4), the confluence density of C_YAP could reach to as much as twofold that of C_Vector (Fig. 2D,E). During co-culture of C_YAP and C_Vector cells, the growth dynamics of C_Vector did not change compared to growth of these cells alone (comparing co-3, co-4 with co-2; Fig. 2F). However, C_YAP cells in co-culture with C_Vector cells displayed enhanced growth compared to conditions when they were cultured alone (comparing co-3, co-4 with co-1; Fig. 2G). These results suggest that certain interactions between C_YAP and C_Vector cells promote the growth of C_YAP. This effect appears to be more pronounced when cells reach confluence.

Multicellular tumor spheroids possess many features mimicking tumors (Hirschhaeuser et al., 2010; Sutherland, 1988), and therefore have been suggested to be a valuable tool to model tumors in vitro. To further study the interaction between heterogeneous tumor cells, we developed a hybrid multicellular tumor spheroid model (Fig. 3A). The tumor spheroids from LN229 cells are able to grow from ∼70 to ∼600 μm in diameter within 15 days (Fig. 3B). To trace cell populations in the spheroids, clones of cells were pre-labeled through stable expression of ZsGreen or mCherry fluorescent protein (Fig. 3A,C). Because the spheroids are formed by cell aggregation, the initial population composition can be controlled by seeding certain numbers of cells. With this model, we could conduct spatial and temporal studies of the growth of each cell population.

We seeded a combination of C_YAP-G and C_Vector-R cells (1:1) to form a multicellular spheroid. At day 5 after seeding, both C_YAP-G and C_Vector-R populations were evenly distributed in the spheroid (Fig. 3D), suggesting that differential expression of YAP does not lead to distinct cellular adhesion properties. As the spheroid grew, the C_YAP-G population expanded evenly across the whole spheroid, whereas the C_Vector-R population's expansion appeared to be limited to the internal part of the spheroid. At day 20 after seeding, most C_Vector-R cells were located at the internal regions, leaving the periphery largely occupied by C_YAP-G (Fig. 3D). These results indicated that the proportion of the C_YAP-G population becomes progressively higher than that of C_Vector-R in the spheroid. To confirm the imaging observation, we dissociated the spheroids and quantified the number of cells in each population. The results showed that both populations expand; however, the C_YAP-G one displays a faster speed during the expansion (Fig. 3E). To rule out that the disproportionate expansion of the C_YAP population compared to C_Vector was due to a difference between ZsGreen- and mCherry-expressing cells, we switched the labeling strategies. Consistently, we observed a similar clonal dominance by the C_YAP-R population over C_Vector-G cells (Fig. 3F,G). These results suggest that C_YAP cells possess a certain growth advantage over C_Vector cells during the growth of spheroids.

Our results above showed that C_YAP cells grow into the dominant population when they are seeded with C_Vector cells at a 1:1 ratio. To examine whether such an initial proportion is required for this clonal dominance, we reduced the initial proportion of C_YAP from 50% to 10% by seeding C_YAP-G and C_Vector-R cells at a ratio of 1:9. Although C_YAP-G was a minor population initially (Fig. 4A, day 5 after seeding), it progressively expanded within the spheroid and eventually grew into the major population (Fig. 4A, day 20 after seeding). Similarly, this clonal dominance also occurred when the labeling strategy was switched (Fig. 4B). Therefore, the dominancy is likely determined by a higher level of YAP expression, not the initial population proportion.

To further examine whether differential YAP expression is required for inducing clonal dominance, we seeded C_YAP-G and C_YAP-R cells together in a 1:1 ratio. In the developed spheroids, both populations were equally represented (Fig. 4C,E), indicating no clonal dominance occurs. Similarly, when differentially labeled C_Vector cells were seeded together (1:1), no clonal dominance was detected in the spheroids (Fig. 4D,F). Therefore, it is the differential, but not the intrinsic, expression levels of YAP that lead to clonal dominance.

To examine how clonal dominance was achieved, we first tested whether C_YAP cells autonomously possess a higher proliferation rate than C_Vector cells under 3D culture condition. Interestingly, spheroids containing C_YAP or C_Vector cells alone are similar in size after they are cultured for the same number of days (Fig. 5A), suggesting no apparent difference between the growth speeds of these spheroids. We confirmed the imaging observation by dissociating the spheroids and quantifying cell numbers (Fig. 5B). Therefore, clonal dominance is not likely due to an autonomous difference in growth. We then compared the growth of C_YAP cells when they are co-cultured with either another population of C_YAP or with C_Vector cells. The C_YAP population growing with C_Vector cells expanded at a faster speed than those cultured with additional C_YAP cells (Fig. 5C). In contrast, when comparing the growth of C_Vector cells when they are cultured with different populations (C_YAP versus C_Vector), we found that the C_Vector population growing with C_YAP expands at a lower speed than that cultured with additional C_Vector (Fig. 5D). These results indicate that when growing together, the intrinsic growth properties of C_YAP and C_Vector populations are altered, with this alteration leading to clonal dominance.

The clonal dominance and growth dynamics changes that we observed in the hybrid spheroids are reminiscent of cell competition, in which the cell population with certain growth advantages outcompetes the other cell population. The competitive interaction induces changes in both cell populations. In this process, the ‘winner’ cells become dominant while the ‘loser’ cells are eliminated through various mechanisms such as apoptosis, senescence, extrusion, etc (Baker, 2017; Maruyama and Fujita, 2017). To examine whether apoptosis is involved in the reduced expansion of the C_Vector population, we conducted immunofluorescence staining of cleaved caspase 3 (denoted CC3), a typical marker for apoptotic cells. In spheroids containing C_YAP and C_Vector cells, overall apoptotic signal is increased compared to spheroids containing C_YAP or C_Vector cells alone (Fig. 5E,F). We then examined the identities of these apoptotic cells. In spheroids containing C_YAP and C_Vector cells, a large proportion (65–75%) of CC3⁺ cells are C_Vector (Fig. 5G). Notably, quantification of CC3⁺ cells in the spheroids containing C_Vector cells alone revealed no increased cell death compared to spheroids containing C_YAP cells alone (Fig. 5E,F). These results suggest that C_Vector cells undergo increased apoptosis during competitive interaction with C_YAP cells. Taken together, the combination of increased cell death observed in C_Vector cells and enhanced growth observed in C_YAP cells accounts for clonal dominance.

In our above studies, higher levels of YAP are expressed in C_YAP cells compared to C_Vector cells, due to exogenous expression of YAP in the former (Fig. 1B). To examine if reducing endogenous YAP expression can also lead to similar competitive interaction, we knocked down YAP in LN229 cells stably expressing ZsGreen or mCherry, using two different shRNAs against YAP (Fig. 6A). These cells were denoted as C_sh-YAP#5 or C_sh-YAP#7 (sh-YAP#5-G/R and sh-YAP#7-G/R in figures). Scramble shRNA-transduced control cells were denoted as C_sh-control (sh-control-G/R in figures). Spheroids of C_sh-YAP#5-R or C_sh-YAP#7-R cells grew more slowly than C_sh-control-G cells (Fig. 6B–D), indicating that YAP is important for their growth. When cultured in hybrid spheroids with C_sh-control-G cells (Fig. 6E), expansion of C_sh-YAP#5-R and C_sh-YAP#7-R populations was further markedly inhibited compared to when they grew alone (Fig. 6F,G,H). Similar inhibition also occured when the labeling strategy was switched (Fig. 6J,K,L). In addition, expansion of the C_sh-control population is enhanced when they are grown with C_sh-YAP#5 or C_sh-YAP#7 cells, compared to when grown alone (Fig. 6E,J,I,M). These results suggest that the clonal dominance by C_sh-control cells is a consequence not only of the intrinsic difference in their growth dynamic, but also of altered growth due to a potential competitive interaction. Furthermore, the results support the notion that competitive interaction is determined by relative, but not absolute, expression levels of YAP.

To understand the mechanism underlying the competitive interaction between C_Vector and C_YAP cells, we examined the gene expression profiles of these cells using RNA sequencing (RNA-seq). Each cell population was isolated from the indicated hybrid spheroids through fluorescence-activated cell sorting (FACS) before RNA-seq analysis (Fig. 7A,B). First, we compared the gene expression profile of C_YAP cells to that of C_Vector cells when each of them was grown with their differentially labeled counterparts (Fig. 7A,C). The expression levels of 319 genes are increased, whereas those of 375 genes are decreased, in C_YAP cells compared to C_Vector cells (Table S1) (≥twofold increase, FDR<0.05). The upregulated genes include the well-known YAP target Cyr61. Ingenuity Pathway Analysis of these genes suggested that multiple cellular functions are significantly different between the two cell populations (Fig. 7D). In C_YAP cells, cell invasion and catabolism-related genes are activated (z-score≥2, P<0.05), whereas intercellular junctions-related genes are inhibited (z-score<−2, P<0.05). These differences may contribute the growth advantage displayed by C_YAP cells in the spheroids.

We then compared the gene expression profiles of C_Vector or C_YAP cells isolated from conditions without competition to those of their counterparts isolated from conditions with competition (Fig. 7E–H). Both cell types showed marked changes in gene expression when they are cultured under conditions of competition compared to non-competition conditions (Fig. 7E,G). For C_YAP cells, the expression of 239 genes is upregulated, whereas that of 377 genes is downregulated (Table S2) (≥twofold increase, FDR<0.05). For C_Vector cells, the expression of 294 genes is upregulated, and of 245 genes downregulated (Table S3) (≥twofold increase, FDR<0.05). These changes in gene expression further support that competitive interaction has a marked impact on both C_Vector and C_YAP cells when they grow together in spheroids. Notably, Ingenuity Pathway Analysis of these genes suggested that tumorigenesis-related genes are activated in both C_Vector and C_YAP cells (Fig. 7F,H). In addition, genes related to neural stem cell differentiation and helper T lymphocyte proliferation are inhibited in C_Vector. The changes in gene expression during competitive interaction between C_Vector and C_YAP cells may contribute the enhanced tumorigenesis observed in vivo (Fig. 1C).

The oncogenic capacity of YAP has been well demonstrated (Dong et al., 2007; Zanconato et al., 2016). Depending on the situation, YAP expression can promote cell proliferation, survival and/or stemness (Yu et al., 2015), and these cell-autonomous effects contribute to its tumor-promoting ability. Recent studies have found that YAP can also regulate the immune system, therefore suggesting that remodeling the tumor immune microenvironment is an additional way for YAP to promote tumorigenesis (Taha et al., 2018). In this study, we found that YAP can promote tumor growth by inducing intercellular interaction in the process of clonal dominance. The interaction occurs when tumor cells express YAP at differential levels and is reminiscent of cell competition. During the interaction, the growth of cells expressing YAP at a higher level is enhanced. In addition, the interaction induces the expression of cancer-related genes. Both of these changes may cause enhanced tumor growth. Considering intratumor heterogeneity is commonly found in tumors (Andor et al., 2016; Heppner et al., 1984; Marusyk et al., 2012), our study suggests that induction of the competitive interaction is another underlying mechanism of YAP-driven tumorigenesis.

By analyzing the gene expression dataset of samples collected from the same GBM tumors as detailed in Fig. 1A (Sottoriva et al., 2013), we found that expression of YAP is intratumorally heterogeneous (Fig. 1A). Previous studies have found intratumor heterogeneity of YAP expression in human colon cancers (Zhou et al., 2011) and meningiomas (Tanahashi et al., 2015). Therefore, heterogeneous YAP expression appears to exist in multiple human cancers.

To study the heterogeneity of YAP expression in GBM, we constructed human GBM cell lines (LN229 glioma cells) expressing YAP at differential levels. Interestingly, tumors arising from the co-culture of cell populations with differential YAP expression grow faster than either population alone (Fig. 1C). These results suggest that certain interactions between cells expressing YAP at differential levels might promote tumorigenesis. When analyzing the cellular composition of the resulting tumors, we found that C_YAP cells form the dominant population compared to C_Vector cells (Fig. 1D). The autonomously stronger tumorigenicity of C_YAP than C_Vector cells (Fig. 1C) may cause such clonal dominance. However, the dominance of C_YAP cells in the heterogeneous tumors is unlikely to be solely caused by the autonomously faster growth of C_YAP cells. Because the heterogeneous tumors showed even faster growth than either homogenous tumor, it is likely that enhanced growth was induced in C_YAP cells in the heterogeneous tumors.

Increased expression of YAP in LN229 glioma cells (C_YAP) per se does not promote their proliferation in either 2D or 3D culture conditions. However, these cells demonstrate stronger abilities to form tumors than C_Vector cells. One explanation is that a certain oncogenic property other than proliferation is responsible for promoting tumor growth. Alternatively, since C_Vector cells already have a strong ability to proliferate, a proliferation enhancement in vitro by YAP expression may be below the detectable threshold. Nevertheless, when growing in co-culture with C_Vector cells, both the growth of C_YAP cells in vitro and the oncogenic ability of C_YAP cells are enhanced (Fig 1C, Fig 2G, Fig 5C). These results suggest that C_Vector cells provide certain stimulations for the growth of C_YAP cells. Previous studies in lung cancer and breast cancer have shown that cooperation between subclones of tumor cells could benefit tumor maintenance, growth and metastasis (Calbo et al., 2011; Cleary et al., 2014; Costa et al., 2015). Our results are in line with these previous findings, supporting the notion that heterogeneous cellular composition in tumors benefits tumor growth and progression. Interestingly, the interaction between C_YAP and C_Vector cells is not a simple cooperation. On the contrary, the enhanced growth in C_YAP cells and induced cell death in C_Vector cells observed in the spheroid model recapitulate classical cellular crosstalk in cell competition.

Competition has long been thought to be a biological interaction between malignant cells. Tumor cells might compete for limited resources under selective pressures. In this process, a clone of advantageous cells may thrive while other clones become extinct, thereby leading to clonal dominance (Kerbel et al., 1988; Miller et al., 1988). Although little is known about competition between tumor cells, studies of cell competition in other circumstances have provided important information about this interesting cellular crosstalk (Baker, 2017; Vincent et al., 2013). A definitive feature of cell competition is that the fitness of the cells involved is not determined by the cells themselves, but by their relative competitive status when confronting their neighbors. In addition, the outcome of competition is not a cell-autonomous process but occurs through dynamic interactions between the cells involved. The cells that ‘win’ the competition eventually take up the tissue territory. Compared to winner cells, the cells that ‘lose’ the competition may face diverse negative consequences, such as cell death, senescence, autophagy and extrusion (Maruyama and Fujita, 2017). The behaviors of C_YAP and C_Vector cells in our in vitro models fit these canonical features of cell competition. In tumors containing C_YAP and C_Vector cells, we saw dominance of C_YAP over C_Vector cells. It appears that the competition between C_YAP and C_Vector cells is milder in tumors than in spheroids (compare Fig. 1D,E with Fig. 3D–G). This is consistent with the thought that competitors can coexist in tumors, probably due to suppression of competition by various factors (Merlo et al., 2006). Accumulation of clonal diversity is a principal property of tumor progression (Maley et al., 2006). Therefore, uncontrolled clonal dominance resulting from competition may be deleterious to progressive tumor growth and be suppressed in cancer. Overall, our studies suggest that clonal oncogenic lesions could benefit tumors by recruiting surrounding tumor cells into a regulated competition process. Eliminating this competition may benefit tumor therapies.

LN229 (CRL-2611) human glioblastoma cell lines were from ATCC and cultured in Dulbecco's modified Eagle's medium (DMEM) (Corning, 10-013-CV) supplemented with 10% fetal bovine serum (Gibco, 10437028) and 1% antibiotic-antimycotic solution (Corning, 30-004-CI) at 37°C with 5% CO₂. The cell lines were not authenticated in this study. The cell lines were examined to be mycoplasma-negative before experiments. Unless otherwise indicated, experiments were performed with cells grown to 50% confluency. To generate C_YAP and C_Vector cells lines, LN229 cells were transduced in vitro with a lentivirus vector expressing ZsGreen or mCherry [plvx-ires-zsgreen1 (#632187) and plvx-ires-mCherry (#631237), Clontech]. ZsGreen- and mCherry-expressing LN229 cells were then further transduced with a retrovirus vector only or a vector expressing YAP (pBABEpuro-Flag-YAP2, Addgene 27472, deposited by Marius Sudol) (Oka et al., 2010). Retroviral generation and infection were as described previously (Li et al., 2014). For generation of YAP knockdown cell lines, lentiviral vectors encoding shRNAs targeting YAP (#5: TRCN0000107266; #7: TRCN00000107268) were from Sigma-Aldrich. Control shRNA, pLKO.1-TRC control, Addgene 10879, deposited by David Root (Moffat et al., 2006).

A total of 1×10⁶ C_YAP and/or C_Vector LN229 cells were injected subcutaneously into six- to eight-week-old female nude mice [Nu(NCr)-Foxn1nu, from Charles River, Strain Code 490]. Tumor size was measured by digital caliper twice per week. All experimental protocols were approved by the Penn State University Institutional Animal Care and Use Committee. All methods were performed in accordance with the relevant guidelines and regulations.

LN229 cells were trypsinized, resuspended and counted. Unless otherwise indicated, 500 cells total were seeded in neural sphere medium [DMEM/F12 (Corning, #15-090-CV), 2 mM L-glutamine (Invitrogen, #25030-081), 1× N-2 supplement (Invitrogen, #17502048), 1× B-27 supplement (Invitrogen, #17504044), 50 μg/ml BSA (Invitrogen, #15260037), 20 ng/ml each of EGF and bFGF (Invitrogen, #PHG0311 and #13256029), 1% antibiotic-antimycotic solution (Corning, #30-004-CI)] per well in 96-well ultra low cluster plates (Costar). After 24–48 h, medium was replaced with regular culture medium (10% FBS in DMEM). The spheroids were incubated at 37°C with 5% CO₂ for 2–3 weeks. Spheroid growth was monitored daily based on detected ZsGreen and mCherry signals. To generate hybrid spheroids containing C_YAP and C_Vector cells at different proportions (1:9), 100 C_YAP and 900 C_Vector cells were seeded.

For cells cultured under 2D conditions, 1×10⁵ (total) ZsGreen- or mCherry-expressing LN229 C_YAP or C_Vector cells were seeded into a 6-well plate. The regular culture medium was replaced daily. After detaching the cells, cell number was manually counted using a hemocytometer based on ZsGreen and mCherry signals under a fluorescence microscope. For cells cultured under 3D conditions, hybrid spheroids were cultured according to the protocol described above. At days 5, 10, 15 and 20, ten spheroids were collected and dissociated with Accutase (Innovative Cell Technologies, Inc. #AT-104). Cell number was manually counted using a hemocytometer based on ZsGreen and mCherry signals under a fluorescence microscope. The total fluorescence intensity of each spheroid was quantified using IncuCyte Live-Cell Analysis System (Essen BioScience).

For western blotting, cells were seeded in complete medium on Petri dishes at a density of 4×10⁴/cm² 24 h before collection. Immunoblotting procedure was described previously (Li et al., 2014). Briefly, cells were lysed in SDS-lysis buffer (10 mM Tris pH 7.5, 1% SDS, 50 mM NaF, 1 mM NaVO₄) and subjected to SDS-PAGE on 4–12% Bis-Tris SDS-PAGE gels (Invitrogen) and transferred to Immobilon-P membranes (Millipore). Membranes were incubated in blocking buffer (5% skim milk, 0.1% Tween, 10 mM Tris at pH 7.6, 100 mM NaCl) for 1 h at room temperature and then with anti-YAP (1:1000, #12395, Cell Signaling Technology) and anti-β-actin (1:2000, #A5316, Sigma-Aldrich) primary antibodies diluted in blocking buffer overnight at 4°C. After three washes, the membranes were incubated with goat anti-rabbit HRP-conjugated antibody (1:5000, #7074, Cell Signaling Technology) or goat anti-mouse HRP-conjugated antibody (1:5000, #7076, Cell Signaling Technology) at room temperature for 2 h and subjected to chemiluminescence using ECL (Pierce #1856136).

Spheroids were fixed and permeabilized for 3 h at 4°C in PBS containing 4% paraformaldehyde and 1% Triton X-100. They were then dehydrated in an ascending series of methanol in PBS (25%, 50%, 75% and 95%, 15 min each) at 4°C, rehydrated in the opposite descending series and washed in PBS (3×10 min). Spheroids were then blocked with 5% BSA/PBS at 4°C for 1 h and followed by incubating overnight at 4°C with anti-cleaved caspase-3 (Asp175) (1:100, 9664, Cell Signaling Technology) primary antibody diluted in 2.5% BSA/0.05% Triton X-100/PBS. After washing with 0.1% Triton X-100/PBS, cells were incubated with donkey anti-rabbit Alexa Fluor 647 secondary antibody (1:200, #711-605-152, Jackson ImmunoResearch) diluted in 2.5% BSA/0.05% Triton X-100/PBS for 24 h at 4°C. Cells were washed with 0.1% Triton X-100/PBS, rinsed with PBS, and mounted in ProLong Gold Mountant (Invitrogen #P10144). When indicated, nuclei were stained with DAPI.

A total of 96 tumor spheroids were collected and dissociated with Accutase. After cell sorting based on ZsGreen and mCherry signals using FACS, cells were lysed and total RNA was extracted using TRIzol (Thermo Fisher Scientific) following the manufacturer's instruction. RNA integration number (RIN) was measured using BioAnalyzer (Agilent) RNA 6000 Nano Kit to confirm RIN above 7. The cDNA libraries were prepared using the NEXTflex Illumina Rapid Directional RNA-Seq Library Prep Kit (Bioo Scientific) as per the manufacturer's instructions. Briefly, polyA RNA was purified from 100 ng of total RNA using oligo (dT) beads. The extracted mRNA fraction was subjected to fragmentation, reverse transcription, end repair, 3′-end adenylation, and adaptor ligation, followed by PCR amplification and SPRI bead purification (Beckman Coulter). The unique barcode sequences were incorporated in the adaptors for multiplexed high-throughput sequencing. The final product was assessed for its size distribution and concentration using BioAnalyzer High Sensitivity DNA Kit (Agilent Technologies). Pooled libraries were diluted to 2 nM in EB buffer (Qiagen) and then denatured using the Illumina protocol. The denatured libraries were diluted to 10 pM by pre-chilled hybridization buffer and loaded onto a TruSeq Rapid flow cell on an Illumina HiSeq 2500 and run for 50 cycles using a single-read recipe according to the manufacturer's instructions. Sequencing data were analyzed using Strand NGS. Briefly, reads were aligned to reference human genome and annotation file (GRCh38, build 38, RefSeq genes and transcripts, 2017_01_13). After filtering the reads by minimal 10 in at least one sample, Audic Claverie Test was performed when comparing each pair of samples. During the analysis, Benjamini Hochberg FDR correction was used for multiple testing corrections and the P-value cut-off was set as 0.05. After this processing, fold change was calculated and the threshold was set as ≥twofold. For hierarchical clustering analysis, genes showing changes above twofold were used. For Ingenuity Pathway Analysis, twofold change was used as the cut-off. Direct relationships were chosen. To investigate the expression of YAP and TAZ in human GBM cells, the gene expression dataset was downloaded from Sottoriva et al. (2013) and preprocessed with quantile normalization and log₂ transformation. The significance of the variance was calculated by chi-square test.

For statistical analyses, samples sizes were chosen based on whether the differences between groups were biologically meaningful and statistically significant. No data were excluded from the analyses. For cell experiments, all cells in one experiment were from the same pooled parental cells. All mice were from the same cohort. The mice were randomly picked to implant different types of cells. For data collection relying on objective instruments, such as FACS, microscopy software and western blotting, the investigators were not blinded to group allocation during data collection. For animal studies, the investigators were not blinded to group allocation during data collection. Statistical significance was determined by unpaired two-tailed Student's t-test unless indicated otherwise. All error bars shown are standard error of the mean (s.e.m.). All statistical calculations and plotting were performed using GraphPad Prism 7.

We thank Thomas Abraham and Wade Edris in the Microscopy Imaging Core Facility, Nate Sheaffer and Joe Bednarczyk in the Flow Cytometry & Cell Sorting Core Facility of Penn State College of Medicine for technical support.