Analysis of the hippo transducers TAZ and YAP in cervical cancer and its microenvironment

ABSTRACT

Hippo is a tumor-suppressor pathway that negatively regulates the oncoproteins TAZ and YAP. Moreover, Hippo affects the biology of a variety of non-neoplastic cells in the tumor microenvironment, even including immune cells. We herein assessed the predictive role of TAZ and YAP, assessed by immunohistochemistry, in 50 cervical cancer patients prevalently treated with neoadjuvant chemotherapy. Tumors were classified as positive or negative according to the percentage of tumor-expressing cells and cellular localization. TAZ/YAP were also evaluated in non-neoplastic cells, namely endothelial cells, non-lymphocytic stromal cells and tumor-infiltrating lymphocytes (TILs). TAZ expression in cancer cells (TAZ^pos) was associated with a reduced pathological complete response (pCR) rate (p = 0.041). Conversely, the expression of TAZ and YAP in TILs (TAZ^TIL+ and YAP^TIL+) seemed to be associated with increased pCRs (p = 0.083 and p = 0.018, respectively). When testing the predictive significance of the concomitant expression of TAZ in cancer cells and its absence in TILs (TAZ^pos/TAZ^TIL-), patients with TAZ^pos/TAZ^TIL- showed lower pCR rate (p = 0.001), as confirmed in multivariate analysis (TAZ^pos/TAZ^TIL-: OR 8.67, 95% CI: 2.31–32.52, p = 0.001). Sensitivity analysis carried out in the 41 patients treated with neoadjuvant chemotherapy yielded comparable results (TAZ^pos/TAZ^TIL-: OR 11.0, 95% CI: 2.42–49.91, p = 0.002). Internal validation carried out with two different procedures confirmed the robustness of this model. Overall, we found evidence on the association between TAZ expression in cervical cancer cells and reduced pCR rate. Conversely, the expression of the Hippo transducers in TILs may predict increased treatment efficacy, possibly mirroring the activation of a non-canonical Hippo/MST pathway necessary for T-cells activation and survival.

Abbreviations

CAFs

cancer-associated fibroblasts

pCR

pathological complete response

TAZ

transcriptional co-activator with PDZ-binding motif

TILs

tumor-infiltrating lymphocytes

YAP

Yes-associated protein

Introduction

The Hippo pathway is a regulator of tissue growth conserved in the animal kingdom, which plays important regulatory functions during organ development and regeneration.¹ Similarly to other embryonic pathways,² Hippo was later recognized as aberrantly regulated in neoplastic diseases where it acts as a tumor suppressor signal.¹

Core components of the Hippo pathway are operatively grouped into a regulatory kinase module and a transcriptional module. The first contains the kinases sterile 20-like kinase 1 (MST1) and 2 (MST2), large tumor suppressor 1 (LATS1) and 2 (LATS2), and the adaptor proteins Salvador homolog 1 (SAV1), MOB kinase activator 1A (MOB1A) and 1B (MOB1B). The transcriptional module is composed by two closely related oncoproteins: the transcriptional co-activator with PDZ-binding motif (TAZ) and the Yes-associated protein (YAP).³

The activation of the Hippo kinase cascade leads to the phosphorylation of TAZ and YAP. This inhibitory phosphorylation promotes nuclear exclusion, cytoplasmic retention and proteasomal degradation of TAZ/YAP.³ When the kinase core module is inactive or its regulatory activity on TAZ/YAP is bypassed by other stimuli, TAZ/YAP translocate to the nucleus where, upon interaction with TEA domain-containing sequence-specific transcription factors (TEAD1 to TEAD4), mediate the transcription of various oncogenes.³

The Hippo pathway also mediates important function in non-neoplastic cells commonly residing in the tumor microenvironment. Hippo transducers were tied to the maintenance of cancer-associated fibroblasts (CAFs).⁴ In turn, CAFs promote tumor growth, neo-angiogenesis and distant dissemination. Next, TAZ/YAP are required for survival of endothelial cells.⁵ Finally, a non-canonical, immune-related Hippo/MST pathway has been described as crucial for maintaining immune homeostasis and immunological self-tolerance.⁶

Activation of TAZ/YAP has been connected to a variety of tumor-promoting functions in a number of neoplastic entities. In breast cancer, TAZ/YAP mediate epithelial-to-mesenchymal transition (EMT), acquisition/retention of cancer stem cell (CSCs) features, chemotherapy resistance and metastatic spread.⁷ In this setting, exploratory clinical analyses pointed to TAZ/YAP as potential prognostic and predictive biomarkers.^8-10

Nevertheless, little is known on the role of the Hippo pathway in cervical cancer.^11,12 Recently, YAP activation was found to encourage proliferation and migration of cervical cancer cells through a positive loop involving the epidermal growth factor receptor (EGFR).¹³ These oncogenic activities were further enforced by the Human Papillomavirus (HPV) E6 protein, which protects YAP from proteasome-dependent degradation.¹³

As a part of our research agenda focused on the clinical translation of potential CSC-related biomarkers previously identified at the preclinical level, we herein focused our attention on TAZ/YAP in cervical cancer. To this end, TAZ/YAP expression was evaluated in a cohort of 50 cervical cancer patients mostly treated with neoadjuvant chemotherapy, and analyzed for their impact on pCR. Given the involvement of TAZ/YAP in the biology of a variety of cell types cohabiting the tumor microenvironment, TAZ and YAP were assessed both in tumor cells and in non-neoplastic cells, namely endothelial cells, non-lymphocytic stromal cells and TILs.

Results

Out of the 71 patients screened, 50 met the eligibility criteria for this study. Twenty-one patients were not eligible for the following reasons: lack of sufficient biological materials in diagnostic biopsies (n = 11), lack of surgical samples (n = 8) and incomplete clinical data (n = 2). Baseline characteristics of the 50 patients included in this study are summarized in Table 1. In this cohort, the pCR rate was 36% (18 out of 50 patients). Table 2 summarizes the expression of TAZ and YAP in tumor cells, non-lymphocytic stromal cells, endothelial cells and TILs. Representative immunohistochemical staining patterns are illustrated in Fig. 1.

Table 1.

Baseline characteristics and treatment outcome of cervical cancer patients treated with neoadjuvant chemotherapy or chemoradiation (n = 50).

Characteristics	N (%)
Age
Median (min–max) [IQ range]	47 (27–69) [37.7–57.0]
Stage
I B	12 (24.0)
II A	15 (30.0)
II B	19 (38.0)
III A	2 (4.0)
III B	2 (4.0)
Histology
Squamous cell carcinoma	43 (86.0)
Adenocarcinoma	7 (14.0)
Treatment
Chemotherapy	41 (82.0)
Chemoradiation	9 (18.0)
pCR
No	32 (64.0)
Yes	18 (36.0)

Table 2.

Expression of TAZ and YAP in cancer cells, non-lymphocytic stromal cells, endothelial cells and TILs.

	N (%)
TAZ (tumor)
Neg	11 (22.0)
Pos	39 (78.0)
TAZ (stroma)
Neg	25 (50.0)
Pos	25 (50.0)
TAZ (endothelium)
Neg	8 (16.0)
Pos	42 (84.0)
TAZ (TILs)
Neg	43 (86.0)
Pos	7 (14.0)
YAP (tumor)
Neg	18 (36.0)
Pos	32 (64.0)
YAP (stroma)
Neg	13 (26.0)
Pos	37 (74.0)
YAP (endothelium)
Neg	13 (26.0)
Pos	37 (74.0)
YAP (TILs)
Neg	44 (88.0)
Pos	6 (12.0)

Figure 1.

Representative examples of immunohistochemical expression of TAZ and YAP in cervical cancer patients. Upper panel: a tumor expressing TAZ in tumor cells, endothelial cells and non-lymphocytic stromal cells, but not in TILs (A); a tumor expressing TAZ exclusively in TILs (B). Lower panel: a tumor expressing YAP in tumor cells, endothelial cells and non-lymphocytic stromal cells, but not in TILs (C); a tumor expressing YAP in TILs, tumor cells and endothelial cells (D). Scale bar = 30 µm.

As shown in Table 3, the only molecular variable associated with reduced pCR rate was TAZ^pos (p = 0.041). Conversely, a suggestion for increased pCR rate was noted for TAZ^TILs+ and YAP^TIL+ tumors compared with their negative counterparts (p = 0.083 and p = 0.018 for TAZ^TILs+ and YAP^TIL+, respectively). Univariate and multivariate logistic regression models confirmed the relationship between TAZ^pos and reduced pCR rate (Table 4). Conversely, TAZ expression in TILs had a protective effect (Table 4).

Table 3.

Association between clinical–molecular variables and pCR (n = 50).

	pCR
	No	Yes	Chi squared test
	N (%)	N (%)	p value
Age
≤ 47	19 (73.1)	7 (26.9)	0.164
> 47	13 (54.2)	11 (45.8)
Stage
I B–II A	17 (63.0)	10 (37.0)	0.869
II B–III B	15 (65.2)	8 (34.8)
Histology
Squamous cell carcinoma	25 (58.1)	18 (41.9)	0.040*
Adenocarcinoma	7 (100.0)	0 (0.0)
Treatment
Chemotherapy	27 (65.9)	14 (34.1)	0.705*
Chemoradiation	5 (55.6)	4 (44.4)
TAZ (tumor)
Neg	4 (36.4)	7 (63.6)	0.041*
Pos	28 (71.8)	11 (28.2)
TAZ (stroma)
Neg	16 (64.0)	9 (36.0)	0.999
Pos	16 (64.0)	9 (36.0)
TAZ (endothelium)
Neg	4 (50.0)	4 (50.0)	0.436*
Pos	28 (66.7)	14 (33.3)
TAZ (TILs)
Neg	30 (69.8)	13 (30.2)	0.083*
Pos	2 (28.6)	5 (71.4)
YAP (tumor)
Neg	10 (55.6)	8 (44.4)	0.351
Pos	22 (68.8)	10 (31.2)
YAP (stroma)
Neg	7 (53.8)	6 (46.2)	0.375
Pos	25 (67.6)	12 (32.4)
YAP (endothelium)
Neg	8 (61.5)	5 (38.5)	0.830
Pos	24 (64.9)	13 (35.1)
YAP (TILs)
Neg	31 (70.5)	13 (29.5)	0.018*
Pos	1 (16.7)	5 (83.3)

^*Fisher's exact test.

Table 4.

Univariate and multivariate logistic regression models for pCR (n = 50).

		Univariate logistic regression model		Multivariate logistic regression model§		Multivariate logistic regression model#
		OR (95% CI)	p value	OR (95% CI)	p value	OR (95% CI)	p value
Age	> 47 vs. ≤ 47	0.43 (0.13–1.42)	0.168			0.73 (0.17–3.09)	0.674
Stage	II B–III B vs. I B– II I A	1.10 (0.35–3.52)	0.869			2.16 (0.43–10.97)	0.352
Treatment	Chemoradiation vs. Chemotherapy	0.65 (0.15–2.80)	0.562			0.69 (0.07–6.43)	0.745
TAZ (tumor)	Pos vs. Neg	4.45 (1.08–18.29)	0.038	7.14 (1.46–34.91)	0.015	6.81 (1.32–35.24)	0.022
TAZ (TILs)	Pos vs. Neg	0.17 (0.03–1.01)	0.052	0.11 (0.02–0.80)	0.029	0.09 (0.01–0.82)	0.033
YAP (TILs)	Pos vs. Neg	0.08 (0.01–0.79)	0.030	0.11 (0.01–1.22)	0.072	0.10 (0.01–1.67)	0.109

^#Adjusted for: age, stage and treatment.

^§Forward stepwise inclusion.

On the basis of the detrimental effect of TAZ expression on cancer cells, and the apparent opposite interaction of the TAZ^TILs+ and YAP^TILs+ phenotypes with pCR, we investigated whether the combination of these molecular endpoints better delineated the category of patients that did not achieve a pCR. As reported in Table S1, both the TAZ^pos/TAZ^TILs- and the TAZ^pos/TAZ^TILs-/YAP^TIL- phenotypes were significantly associated with reduced pCR rate (p = 0.001 and p < 0.001, respectively).

Coherently, the TAZ^pos/TAZ^TILs- was the only variable that tested significant at the univariate and multivariate assessment (OR 8.67, 95% CI: 2.31–32.52, p = 0.001), even when adjusting by age, stage and type of treatment (OR 9.13, 95% CI: 2.19–38.09, p = 0.002) (Table 5). The robustness of this model was internally validated with a re-sampling without replacement procedure. The replication rate was 77.5% (155/200 simulations) with statistical significance set at p < 0.01.

Table 5.

Univariate and multivariate logistic regression models evaluating the impact of the TAZ^pos/TAZ^TIL- phenotype on pCR (n = 50).

		Univariate logistic regression model		Multivariate logistic regression model§		Multivariate logistic regression model#
		OR (95% CI)	p value	OR (95% CI)	p value	OR (95% CI)	p value
Age at diagnosis	> 47 vs. ≤ 47	0.43 (0.13–1.42)	0.168			0.74 (0.18–3.00)	0.676
Stage	II B–III vs. I–II A	1.10 (0.35–3.52)	0.869			1.86 (0.40–8.61)	0.425
Treatment	Chemoradiation vs. Chemotherapy	0.65 (0.15–2.80)	0.562			0.39 (0.05–2.82)	0.352
TAZpos/TAZTIL-	Pos vs. Neg	8.67 (2.31–32.52)	0.001	8.67 (2.31–32.52)	0.001	9.13 (2.19–38.09)	0.002

^#Adjusted for: age at diagnosis, stage and treatment.

^§Forward stepwise inclusion.

Comparable results were obtained when the multivariate model was generated by exclusively considering the 41 patients treated with neoadjuvant chemotherapy (OR 11.0, 95% CI: 2.42–49.91, p = 0.002), even when the model was adjusted for clinical variables that were not significant at univariate assessment (OR 15.77, 95% CI: 2.54–98.1, p = 0.003) (Table 6). The replication rate was 69% (138/200 simulations) with statistical significance set at p < 0.01. Moreover, the results of the logistic regression model were confirmed with the bootstrap method (p = 0.004; 95% CI: 1.05–21.09).

Table 6.

Univariate and multivariate logistic regression models for pCR in patients who received chemotherapy (n = 41).

		Univariate logistic regression model		Multivariate logistic regression model§		Multivariate logistic regression model#
		OR (95% CI)	p value	OR (95% CI)	p value	OR (95% CI)	p value
Age at diagnosis	> 47 vs. ≤ 47	0.44 (0.12–1.63)	0.232			0.75 (0.14–4.10)	0.741
Stage	II B–III B vs. I B–II A	1.72 (0.43–6.90)	0.445			4.27 (0.64–28.65)	0.135
TAZpos/TAZTIL-	Pos vs. Neg	11.0 (2.42–49.91)	0.002	11.0 (2.42–49.91)	0.002	15.77 (2.54–98.1)	0.003

^#Adjusted for: age and stage.

^§Forward stepwise inclusion.

Discussion

In the present study, we investigated the relationship between the expression of TAZ/YAP, evaluated both in tumor cells and in the tumor microenvironment, and pCR in a moderately-sized cohort of cervical cancer patients prevalently treated with neoadjuvant chemotherapy. We observed a reduced pCR rate in patients whose tumors expressed TAZ, but not YAP. The predictive ability of TAZ expression in cancer cells was significantly improved when its status in TILs was considered. To our knowledge, this is the first study thoroughly evaluating the expression pattern of both TAZ and YAP in cervical cancer and its supportive environment, and with a clear focus on a clinical outcome.

We acknowledge that the retrospective nature of our study invites caution in the interpretation of the results. Indeed, the findings herein reported are hypothesis-generating and require larger and/or prospective confirmations.

Before discussing our results from a molecular perspective, it is worth anticipating that the use of neoadjuvant chemotherapy for early-stage or locally-advanced cervical cancer patients, and the impact of pCR on long-term survival outcomes, is still controversial.^14-18 Nonetheless, in some countries and especially in Europe, great attention is put toward this approach. Consistently, the European Organization for Research and Treatment of Cancer (EORTC) is carrying out a phase III randomized trial, having overall survival as the primary endpoint, comparing neoadjuvant chemotherapy followed by radical hysterectomy vs. chemotherapy plus radiation therapy in FIGO Ib2-IIb cervical cancer patients (ClinicalTrials.gov Identifier: NCT00039338).

From a molecular standpoint, this study raises important points that deserve consideration. First, TAZ and YAP are two homologous transcriptional co-activators. Nevertheless, in our study only TAZ expression was associated with reduced pCR rate. Despite the analogies between TAZ and YAP, in a neoplastic background they might exert partly different functions. For instance, in breast cancer only TAZ was linked to the acquisition/retention of CSC traits,^19,20 whereas the molecular output elicited by YAP activation remains ambiguous.⁷ Indeed, both tumor-promoting and tumor-suppressive functions were reported upon its activation.^21-27 Conversely, in the domain of tumor-stroma interplay, YAP function was required for CAFs, whereas TAZ was dispensable for CAF-mediated promotion of cancer-cell invasion.⁴ Even though YAP seems to elicit tumor-enhancing functions in cervical cancer, in our study only TAZ expression seemed to predict the outcome of interest.

Second, the Hippo pathway is subject to a variety of regulatory forces that act both at the level of core kinases or directly on TAZ/YAP. Beyond a variety of effectors capable of activating MST1/2 or LATS1/2 kinases,^28-31 a series of other inputs tune TAZ/YAP activation. These include junctional and apicobasal polarity factors, mechanical forces imposed by the extracellular matrix, G-protein-coupled receptors and Rho GTPases, the actin cytoskeleton, metabolic routes such as the mevalonate pathway and aerobic glycolysis, and the Wnt pathway.^5,32-38 Thus, wider pathway analyses, which also involve regulatory branches and target genes, are necessary to provide a clearer picture on how Hippo transducers affect the efficacy of anticancer treatments. On this basis, we have implemented our search for Hippo pathway-related prognostic and predictive biomarkers by taking into account molecular endpoints related to TAZ/YAP activation.

In our opinion, the most intriguing finding of our study relates to the protective role of TAZ and, although to a lower extent, YAP when expressed in TILs. The Hippo/MST pathway was recently recognized as a multifaceted regulator of mammalian adaptive immunity.⁶ In this context, MST kinases are essential for the development, activation, survival, trafficking and homing of T cells.^39-43 Even though the Hippo/MST pathway does not involve the canonical Hippo cascade, we speculate that cytoplasmic expression of TAZ/YAP might be a readout for the activation of the Hippo/MST pathway, thus labeling a subset of  lymphocytes particularly reactive against cancer cells. Coherently, TAZ/YAP in TILs had an exclusive cytoplasmic localization. This observation assumes even more relevance when considering the growing interest surrounding TIL status as predictor of therapeutic and survival outcomes.⁴⁴ When referring to TILs, the attention turns to breast cancer.⁴⁴ Given that we have already assessed TAZ/YAP in HER2-positive and triple-negative breast cancer patients treated with neoadjuvant chemotherapy, [⁸, and personal unpublished data] their evaluation in TILs, together with MST kinases, is ongoing and will provide further ground to this hypothesis.

In summary, we observed an inverse relationship between TAZ expression and pCR in a cohort of cervical cancer patients prevalently treated with neoadjuvant chemotherapy. Conversely, the expression of TAZ/YAP in TILs seemed associated with an increased likelihood of achieving a pCR. The concomitant evaluation of TAZ in tumor cells and in TILs provided the most accurate prediction of reduced pCRs. Thus, Hippo pathway-associated biomarkers deserve increased consideration in cervical cancer, and larger studies are warranted.

Patients and methods

Study participants

Fifty histologically confirmed cervical cancer patients who received neoadjuvant chemotherapy (n = 41) or concurrent chemoradiation (n = 9) followed by radical hysterectomy were included in this retrospective analysis. Nine patients treated with chemoradiation were included in this study given that, upon radiological assessment, they were deemed suitable for surgery with radical intent. Analyses were initially carried out in the entire cohort, and then repeated in a sensitivity analysis conducted by excluding those patients who received chemoradiation (n = 9). Neoadjuvant chemotherapy consisted in the TIP regimen (paclitaxel 175 mg/m² on day 1 + ifosfamide 2500 mg/m² on days 1 and 2 + cisplatin 50 mg/m² on day 2 every 21 d for 3 or 4 cycles). Chemoradiation was delivered with the following schedule: intensity-modulated radiation therapy (IMRT) administered over 6 weeks (60 Gy to the tumor and 49–50 Gy to non-metastatic pelvic nodes) plus concomitant weekly single-agent cisplatin at 40 mg/m2/week. Eligibility was determined according to the following criteria: (i) completeness of data related to clinical features and treatment outcomes, (ii) sufficient biological materials for pathological and molecular analyses in diagnostic biopsies, (iii) availability of surgical samples for the evaluation of pCR and (iv) completion of the planned treatment. pCR was defined as a complete disappearance of tumor in the cervix with negative nodes, as suggested by Buda et al.⁴⁵ This retrospective study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the “Regina Elena” National Cancer Institute of Rome, the coordinating center. Our work was carried out in full compliance with all of the legal requirements pertaining to the institutions involved. Written informed consents were secured before anticancer therapy.

Study procedures

The immunohistochemical assessment of TAZ and YAP was performed in formalin-fixed paraffin-embedded (FFPE) tissues using the monoclonal antibody (MoAb) anti-TAZ (M2-616, BD PharMingen) at the dilution of 1:400  and the MoAb anti-YAP (H-9, Santa Cruz) at the dilution of 1:200.

In cancer cells, TAZ/YAP staining intensity was graded on a 4-grade scale (0: negative, 1+: weak, 2+: moderate, 3+: strong). In order to classify negative (TAZ^neg, YAP^neg) and positive (TAZ^pos, YAP^pos) tumors, we considered cellular localization, staining intensity and percentage of tumor-expressing cells. On this basis, TAZ/YAP positivity was defined as a distinct nuclear immunoreactivity in ≥10% of neoplastic cells.

For the assessment of TAZ/YAP in the tumor microenvironment, the three main cellular components (endothelial cells, non-lymphocytic stromal cells and TILs) were morphologically identified. For each cellular compartment, TAZ/YAP expression was considered positive when cells exhibited a distinct homogeneous/heterogeneous immunoreactivity, irrespectively of the subcellular localization. This was related to the observation that, when expressed, TAZ/YAP were exclusively cytoplasmic in endothelial cells and TILs, and prevalently cytoplasmic in non-lymphocytic stromal cells. In this analysis, we did not include faintly staining cells or positive cells located in the tumor margin or in areas with poor morphology. Two investigators (SB and MM) blinded to treatment outcome independently evaluated immunoreactivity. Discordant cases were discussed and solved at a face-to-face assessment.

Statistical analysis

Descriptive statistics were used to summarize the characteristics of the study participants and the investigated molecular endpoints. The relationships between categorical variables were assessed with the Pearson's Chi-squared test of independence (2-tailed) or the Fisher Exact test, depending upon the size of the groups compared. Univariate logistic regression models helped identify variables potentially impacting pCR. A multivariate logistic regression model was generated using a stepwise regression approach (forward selection) and the related estimates reported as Odds Ratio (OR) and 95% Confident Interval (CI). The enter and remove limits were p = 0.10 and p = 0.15, respectively. A multivariate logistic regression model was also built by adjusting for clinical variables. A sensitivity analysis was carried out by removing the nine patients treated with chemoradiation. We considered statistically significant p values less than 0.05. The consistency of the TAZ^pos/TAZ^TILs- model was assessed through an internal validation procedure envisioning re-sampling without replacement. In greater detail, 200 hundred, less-powered datasets were generated by randomly removing ∼20% from the original sample. For each simulation, the multivariate model was repeated and the replication rate was calculated.⁴⁶ For internal validation, statistical significance was set at p < 0.01. Finally, in order to avoid overfitting bootstrap (re-sampling with replacement) was used as a second procedure for internal validation.⁴⁷ Five independent procedures, each containing 1.000 bootstrap samples, were applied, and the less optimistic simulation was reported in terms of p value and 95% CI.

Statistical analyses were carried out using SPSS software (SPSS version 21, SPSS Inc., Chicago, IL, USA).

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgment

We thank Tania Merlino for technical assistance.

Funding

This work was supported by Consorzio Interuniversitario Nazionale per la Bio-Oncologia (CINBO) (CN).