YAP1-TEAD activation is highly prevalent in MPM patients
We started by assessing the prevalence of Hippo pathway-related genetic alterations in 87 MPM samples published in the TCGA MESO collection. Genetic alteration collectively refers to homozygous deletion (GISTIC value = -2; [33], amplification (GISTIC = 2), or non-synonymous sequence mutation of a gene. The TCGA MESO collection comprises 23 cases of biphasic type and 54 epithelioid tumors. We interrogated 85 genes (listed in the Supplementary Information) which are either reported as Hippo pathway components or Hippo-regulating genes in the Kyoto Encyclopedia of Genes and Genomes (KEGG; Kanehisa Laboratories, Kyoto, Japan) or in MetaCore™ (Clarivate, Philadelphia, PA), or are established oncogenes (KRAS, EGFR), tumor suppressors (CDKN2A, CDKN2B, RB1), or otherwise mentioned in the referenced literature. In total, 56 of the 85 interrogated genes (49/74 with a documented Hippo relation) had genetic alterations, in TCGA MESO. The total prevalence of alterations in the 85 genes was 89% in the epithelioid group and 86% in the biphasic group. At least one Hippo-related alteration was found in 84% of the tumors. The most frequently altered genes were BAP1 (38%), CDKN2A (45%), CDKN2B (41%), NF2 (34%), TP53 (17%), and LATS2 (11%). The core Hippo genes NF2 and LATS2 were altered in 55% of the biphasic and 36% of the epithelioid tumors.
Using public data from experiments where the TEAD transcription co-factors or their Hippo regulators were knocked out, or in, we generated a transcriptional signature that measures the activity of TEAD. MPM was the indication with the highest scores for this signature among all TCGA tumors, and pleural mesothelioma cell lines had the highest scores among the cell lines of the CCLE collection, on average (Fig. S1A). Pooled TCGA tumors with mutations in the Hippo regulators BAP1, LATS2, and NF2 scored significantly higher than wildtype tumors (Fig. S1B). In the TCGA MESO cohort, tumors with NF2 mutation or deletion scored significantly higher than tumors with diploid, wildtype NF2 (Fig. S1C). In fact, NF2 mutation or deletion was the best predictor of the YAP1-TEAD-activity score among all tested genetic events in all the queried genes (data not shown). The YAP1-TEAD activity score was much lower in other tumor indications, e.g., cervical cancer, where YAP1-TEAD activation was also strongly associated with HIPPO-YAP1 genomic alterations [34].
Evaluation of MPM cell lines for in vivo studies
To select an in vivo xenograft model that would be representative of the TCGA patient dataset, 13 MPM cell lines were characterized for Hippo pathway regulator alterations, TEAD1-4 expression pattern and YAP1 activation in cellular assays, and for which we subsequently determined the in vivo tumor growth in mice (supplementary data Table S1). All cell lines, except NCI-H28, carried Hippo pathway regulator alterations and showed YAP1 activation according to RNA signature status. In addition, all cell lines, except NCI-H28, responded to YAP1 down-regulation using YAP1 siRNA treatment in vitro (data not shown). Among the 13 cell lines tested, only three cell lines formed tumors when implanted as subcutaneous xenografts in mice. Tumor formation was most optimal for the MSTO-211H model with no associated body weight loss, so this model was chosen for in vivo target validation experiments.
TEAD2 dominant negative (DN) expression transiently inhibits YAP1 and tumor growth in vivo
The effect of YAP1 downregulation on tumor growth was first tested in vitro using a genetic approach based on a TEAD dominant-negative (TEAD2-DN) construct. This construct was previously reported as an efficient inhibitor of YAP1 activity [16]. It is based on a truncated version of TEAD2, which lacks the DNA binding domain but retains its ability to associate with YAP1. It thereby acts as a non-functional competitor of endogenous TEADs for binding to YAP1. We engineered this construct behind a doxycycline-inducible promoter and used it for the stable transfection of MSTO-211H cells. Upon doxycycline addition, TEAD2-DN expression resulted in 66% inhibition of tumor cell growth in vitro, 96 h post doxycycline induction, Fig. 1A-B, accompanied by the downregulation of target genes CYR61 and CTGF, Fig. 1C, two frequently used and direct biomarkers of YAP1-TEAD activity [35,36,37].
To study the effects of YAP1 downregulation in vivo, we grafted the MSTO-211H-TEAD2-DN cell line on SCID mice and established a subcutaneous xenograft model. Induction of TEAD2-DN by doxycycline administration in vivo induced a significant tumor growth inhibition (TGI) of 44% (p < 0,0001) when doxycycline was administered right after tumor cell implantation. When doxycycline was administered later, on already established tumors ranging from 171 to 414 mm3, the effect was a transient tumor stasis with a TGI of 50% (p < 0,0001) Fig. 2A-B.
In parallel, in an in vivo pharmacodynamic study, we evaluated biomarker modulation post-TEAD2-DN induction by doxycycline at 24 h, 96 h, and 216 h Fig. 3A. TEAD2-DN was highly expressed at 24 h and 96 h following induction by doxycycline. However, 216 h post continuous doxycycline administration, TEAD2-DN was no longer present Fig. 3B. The transient TEAD2-DN induction correlated with the inhibition of the downstream effectors CTGF (gene name CNN2) and CYR61(gene name CNN1) at the mRNA level for which a maximum inhibition was observed 24 h post doxycycline induction (45% and 39% respectively) and for which inhibition was lost at 216 h post doxycycline induction Fig. 3C. The transient inhibition was also detected at the protein level for CYR61 Fig. S2. Along the same lines, maximal YAP1 inhibition based on the YAP1 gene transcription signature score was observed 24 h post doxycycline induction and was gradually lost with no inhibition at all detectable at 216 h post doxycycline treatment Fig. 3D. We suspected that the transient nature of the TEAD2-DN effect stemmed from an escape mechanism and derivation of the cell population expressing TEAD2-DN in vivo. This assumption is supported by the loss of CYR61 and CTGF inhibition as well as the decrease in the YAP1 activation score observed at the very end of the in vivo tumor growth study 34 days post in vivo tumor cell implantation and 15 days post doxycycline treatment Fig. S3.
Nevertheless, the TEAD dominant negative approach showed that even transient downregulation of YAP1 and YAP1-TEAD target genes in vivo can lead to tumor growth inhibition in established tumors. To confirm these observations in a more stable setting over time, we decided to extend the in vivo target validation experiments to an orthogonal genetic model.
YAP1 shRNA knockdown induced in vivo tumor regression and was specific for YAP1 activated MPM
We generated a construct expressing an shRNA against YAP1 behind a doxycycline-inducible promoter, stably transfected it into the MSTO-211H cell line and evaluated its effect in vitro. Doxycycline induction of YAP1shRNA led to 80% downregulation of YAP1 at the mRNA level and resulted in approximately 50% inhibition of tumor cell growth, 96 h post doxycycline induction Fig. 4A-B. YAP1 downregulation was accompanied by the inhibition of gene expression for the YAP1-TEAD target genes CYR61 (60%) and CTGF (45%) as determined by RT-qPCR. In addition, YAP1 knockdown led to significant induction of apoptosis Fig. 4C-D, and supplemental information for Incucyte® video.
We engrafted the MSTO-211H-SH-YAP1 cell line subcutaneously in a mouse model and evaluated its activity in a series of efficacy studies. Downregulation of YAP1 by the doxycycline inducible YAP1 shRNA prevented tumor initiation in 10 out of 10 mice Fig. 5A. Furthermore, knockdown of YAP1 following doxycycline supplementation in mice bearing established tumors with sizes ranging from 160 to 360 mm3, induced tumor regression in five out of five mice with a median regression of 80% as early as 12 days post doxycycline treatment and a maximum of 96% median regression at 39 days post doxycycline treatment Fig. 5B. Reinforcing the role of YAP1 in mesothelioma tumor growth, tumor regressions were also observed upon YAP1 knockdown in larger tumors, ranging from 476–1157 mm3. However, regressions were followed by a prompt regrowth of the tumors Fig S4.
We then evaluated biomarker modulation in an in vivo pharmacodynamic study 24 h and 96 h post doxycycline induction and at the time point of 96 h post doxycycline induction followed by 6 days without doxycycline treatment Fig. 6A. The last time point was intended to evaluate PD modulation upon tumor regrowth once doxycycline treatment was stopped, but surprisingly no tumor regrowth was observed after doxycycline removal. While YAP1 shRNA induction was associated with 62% and 72% knockdown of YAP1 mRNA 24 h and 96 h post doxycycline supplementation, respectively, 6 days post doxycycline removal, 41% inhibition of YAP1 mRNA was still present, possibly explaining the absence of tumor regrowth observed. Once again, YAP1 downregulation 24 h and 96 h post shRNA YAP1 induction was associated with a significant decrease of YAP1-TEAD dependent transcription as determined by RT-qPCR experiments for CTGF and CYR61, as well as by the inhibition of the signature scores. Furthermore, CTGF and CYR61 RNA biomarkers and the signature scores continued to drop even after doxycycline removal, which correlated with the observed continued inhibition of the tumor growth phenotype. Figure 6 B-C.
Finally, to demonstrate the specificity of the YAP1 shRNA effects, we evaluated the same doxycycline inducible YAP1 shRNA knockdown construct in a Hippo pathway independent colon cancer model, HCT116. In cell culture, doxycycline treatment led to 87% knockdown of YAP1 mRNA and downregulated the expression of YAP1-TEAD target genes CYR61 and CTGF, but contrary to what we observed with the MSTO-211H cell model, we did not detect any effect on tumor cell growth for HCT-116, indicating that HCT116 cell growth does not depend on YAP1 Fig. S5 A-B-C. In full alignment with these in vitro data, the subsequent in vivo studies with YAP1 shRNA HCT-116 xenograft models demonstrated that YAP1 knockdown had no effect on tumor growth and prevented neither tumor initiation in 6 out of 6 mice nor did it induce tumor regression on already established tumors (165 to 561 mm3) in 6 out of 6 mice tested Fig. S5 D-E. These data strongly suggest that the growth inhibitory effect observed with YAP1 KD in the MPM MSTO-211H in vivo model was specific and not a general cytotoxicity phenomenon.
Interestingly, tumoral regressions achieving 96% were obtained upon doxycycline treatment in mice bearing MSTO-211H-SH-YAP1, but a tumor escape was noticed following extended exposure to YAP1 shRNA Fig. S4**. To further investigate this tumor escape, a new batch of mice was engrafted with one of the re-growing tumors and evaluated for sensitivity to YAP1 shRNA expression upon doxycycline supplementation. We observed that shRNA induction still led to the downregulation of YAP1 protein levels as determined by Western blot in this long-term treatment model. However, we did not detect any effect on tumor growth anymore, indicating that this tumor model had become independent from exclusive regulation by YAP1 Fig. S6. We consider this model as an in vivo model of acquired resistance to YAP1 inhibition and believe that it merits further comprehensive analyses as it can be a valuable tool for studying escape of treatment with YAP1 pathway inhibitors. The mechanism of resistance in this tumor model will be the subject of our future studies.
Pharmacological inhibition by allosteric TEAD inhibitor K-975 mimics genetic studies in MPM in vivo models
Recently, therapeutic approaches directed at targeting aberrant YAP1 activation in tumors have advanced. In particular, allosteric inhibitors blocking TEAD palmitoylation and thereby inhibiting YAP1-TEAD-dependent gene transcription have been developed [15, 38,39,40,41,42]. K-975 is one of these allosteric TEAD inhibitors [15], and we compared the activity of this compound to our genetic studies in vivo.
We first assessed the in vivo efficacy of K-975 in the MSTO-211H xenograft model. Robust antitumor activity was observed with 49% median regression after treatment with K-975 at the highest dose tested of 200 mg/kg twice a day (BID) for 18 consecutive days. The regressions were transient since tumors started to regrow under treatment. At lower doses, 100 mg/kg BID resulted in tumor growth delay, and K-975 was inactive at 30 mg/kg BID Fig. 7A. In accordance with the level of in vivo antitumor activity, K-975 was able to decrease CTGF and CYR61 protein levels at the active doses of 200 and 100 mg/kg, but not at the inactive dose of 30 mg/kg, 24 h after single administration in a corresponding pharmacodynamic study Fig. 7B.
We then continued the evaluation of K-975 in the NCI-H226 xenograft model, an MPM model harboring an NF2 deletion. Robust antitumor activity was also observed in this model with 34% median regression of tumors after 200 mg/kg BID treatment for 18 consecutive days. The treatment dose of 100 mg/kg BID was also active, but with tumor stasis, and no efficacy was obtained at the dose of 30 mg/kg BID Fig. 7C. In line with the NCI-H226 in vivo efficacy study and similar to what we had previously seen in the MSTO-211H MPM xenograft model, K-975 was able to modulate CYR61 protein levels 24 h after a single administration at the active doses of 200 and 100 mg/kg, but not at the inactive dose of 30 mg/kg in the corresponding pharmacodynamic study Fig. 7D. Since K-975 is a covalent inhibitor forming a stable bond with the cysteine at the entry of the TEAD allosteric lipid pocket [15, 38, 40, 43] we also evaluated the degree of TEAD target occupancy by this molecule by determining the Target Occupancy Ratio (TOR) [44]. This study aimed to relate the on-target binding of TEAD with PD biomarker modulation and efficacy results. Out of the four members of the TEAD protein family, the NCI-H226 model predominantly expresses TEAD1 and TEAD4, so TOR was determined for these two TEAD proteins. Significant time and dose effects were observed for K-975 on TOR measured on TEAD1 and TEAD4, with a maximal target engagement of 67% on TEAD1 and a lower TOR of 38% on TEAD4 at 6 h after a single administration of 200 mg/kg Fig. 7E. Regarding biomarker modulation, significant dose and time-dependent inhibition were observed for both the YAP1-TEAD transcription score Fig. 7F and the downregulation of CYR61 protein levels (60% inhibition), 24 h after a single administration of 200 mg/kg. Hence, based on this data, 67% TOR on TEAD1, 30–40% decrease in the YAP1-TEAD transcription score, and 60% downregulation of CYR61 protein levels are sufficient to achieve tumor regression in the MPM NCI-H226 model.
In summary, the TEAD allosteric inhibitor K-975 led to the regression of established MPM tumors in vivo. K-975 efficacy and dose–response could be associated with the degree of TEAD target occupancy and with the modulation of YAP1 biomarkers such as the YAP1-TEAD dependent gene transcription score and CYR61 protein levels. While the in vivo tumor growth inhibition and biomarker modulation effects of the small molecule compound qualitatively compared to the effects we obtained with the YAP1 shRNA, the maximal tumor growth inhibition for K-975 was inferior to what we had observed with the genetic YAP1 shRNA knockdown. Whether this is due to the pharmacological properties of the K-975 compound or to its mechanism of action (TEAD binding versus YAP1 downregulation) will need to be further explored.