THRalpha expression according to EOC subtypes
THRα expression showed significant differences within the histological subtype, accounting nuclear as well as cytoplasmic staining. Serous carcinomas showed only faint expression of THRa in the nucleus (median IRS = 2) as well as in the cytoplasm (median IRS = 0; Figure 2a = 10 x lens, Figure 2f = 25 x lens). A more intense staining was observed in the clear cell cases in the nucleus (median IRS = 2) as well as in the cytoplasm (median IRS = 2; Figure 2b = 10 x lens, Figure 2g = 25 x lens). The endometrioid subtype showed similar expression schemas as the serous subtype in the nucleus (median IRS = 2) as well as in the cytoplasm (median IRS = 0; Figure 2c = 10 x lens, Figure 2h = 25 x lens). The lowest expression of THRa was found in the mucinous subtype in the nucleus (median IRS = 1) as well as in the cytoplasm (median IRS = 0; Figure 2d = 10 x lens, Figure 2i = 25 x lens). A summary of the staining results is shown in Figure 2e for the nuclear staining (p = 0.005) and Figure 2j for the cytoplasmic staining (p = 0.037).
THRa1 as well as THRa2 showed no significant different expression according to the histological subtype. The median expression of THRalpha1 in the nucleus was 2 and the median expression in the cytoplasm was 0. The median expression of THRa2 in the nucleus was 6 and therefore much more intense compared to THRa and -a1, respectively. The median expression of THRa2 in the cytoplasm was 0. There was no significant different expression of the three THRa subtypes according to grading, FIGO staging or age at surgery.
Correlation analyses
By using recently published data by our institute, we were able to correlate the expression of all THRa subtypes stained with former investigation results. There are significant correlations with the gonadotropin receptors [24] and the luteinizing hormone (LH)-receptor ligand hCG [25]; specifically THRa staining in the nucleus showed a positive correlation to the follicle stimulating hormone receptor (FSHR) (correlation coefficient (cc) = 0.181; p = 0.027) and a negative correlation to hCG (cc = -0.247, p = 0.003). In opposite, THRa in the cytoplasm showed a positive correlation to the luteinizing hormone/choriogonadotropin receptor (LH/hCGR) (cc = 0.199, p = 0.014) and a positive correlation to hCG (cc = 0.187, p = 0.027). The THRa1 expression in the cytoplasm is positively correlated to hCG (cc = 0.278, p = 0.001). THRa2 in the nucleus showed a positive correlation to FSHR (cc = 0.185, p = 0.024). In addition, there are also positive correlations to the classical steroid hormone receptors, which were analysed by our research group too [26]. THRa staining in the nucleus showed a positive correlation to the ERb (cc = 0.213, p = 0.009) and to the PRA (cc = 0.172, p = 0.035). The THRa1 expression in the cytoplasm is positively correlated to ERb (cc = 0.219, p = 0.006). THRa2 in the nucleus showed positive correlation to ERa (cc = 0.247, p = 0.002) and to PRA (cc = 0.219, p = 0.007). In addition to the classical estrogen receptors, also the GPER [27-29] showed positive correlation to THRa staining in the nucleus (cc = 0.219, p = 0.007) and to THRa2 in the nucleus (cc = 0.252, p = 0.002).
Another positive correlation was found within the tumour associated mucin 1 epitop (TA-MUC1) detected with the Gatipotuzumab antibody formerly known as PankoMab [30, 31] and THRa staining in the nucleus (cc = 0.279, p = 0.001). In contrast, THRa1 expression in the cytoplasm is negatively correlated to TA-MUC1 (cc = -0.195, p = 0.019). TA-MUC1 as membrane bound protein can also be translocated to the cytoplasm of tumour cells [32]. In that case it is negatively correlated to the expression of THRa1 in the nucleus (cc = -0.166, p = 0.048) and THRa2 in the nucleus (cc = -0.268, p = 0.001). An immunosuppressive glycoprotein that is connected to TA-MUC1 is glycodelin and its specific glycoform glycodelin A [33, 34]. Glycodelin A showed a positive correlation to THRa2 in the cytoplasm (cc = 0.170, p = 0.037). Glycodelin showed positive correlation with THRa in the nucleus (cc = 0.241, p = 0.003) as well as in the cytoplasm (cc = 0.231, p = 0.004). THRa2 expression in the nucleus is positively correlated with glycodelin (cc = 0.265, p = 0.001).
Survival analyses
The expression of the general THRa is connected to significantly reduced overall survival in the subgroup of clear cell carcinomas. The median survival for THRa-negative patients is 5.24 years in contrast to only 0.29 years for patients showing THRa expression in the nucleus (Figure 3a, p = 0.006).
The THRa isoforms -a1 and –a2 are in general positive prognosticators if expressed in the nucleus and negative prognosticator if expressed in the cytoplasm, respectively. In detail, THRa1 is a general positive prognosticator if expressed in the nucleus with a median survival of 4.22 years for patients positive for THRa1 and 2.08 years for patients that do not express THRa1 in the nucleus (Figure 3b, p = 0.024). Subgroup analyses of mucinous carcinomas showed that THRa1 is a negative prognosticator if expressed in the cytoplasm. The median survival time is 16.59 years for mucinous carcinoma patients that do not express THRa1 in the cytoplasm and 2.87 years for mucinous carcinoma patients with cytoplasmic THRa1 expression (Figure 3c, p = 0.037).
The THRa2 receptor in general is a negative prognosticator if expressed in the cytoplasm. The median survival time is 3.75 years for patients and 1.37 years for patients with THRa2 in the cytoplasm (Figure 3d, p = 0.001). Nuclear expression of THRa2 is not a general positive prognosticator. This can be found in the subgroup of serous carcinomas. The mean survival time for serous carcinoma patients with nuclear THRa2 expression is 6.21 years in contrast to 2.32 years for patients with no nuclear THRa2 expression (Figure 3e, p = 0.002). It is remarkable that patients with clear cell carcinomas show opposite results. The median survival time for clear cell carcinoma patients with nuclear THRa2 expression is only 1.65 years in contrast to 5.24 years for patients with no nuclear THRa2 expression (Figure 3f, p = 0.034).
Comparison of THRa, -a1 and –a2 expression in low-grade and high-grade serous ovarian cancer
As shown in Figure 4, the expression of all three a-subunits is higher in the nucleus of low grade serous ovarian cancer cases with a trend to significance in the general THRa (p=0.078), no significance for THRa1 and a significantly higher THRa2 expression in low-grade serous cancer cases compared to high-grade subtype.
Cox Regression analyses of survival
Cox regression (Table 1) was performed to identify independent predictors for OS. Pattern of age at surgery failed to remain significant within multivariate testing, while grading, FIGO staging, THRa1 in the nucleus (1A, p = 0.043) and THRa2 in the cytoplasm (1B, p = 0.002) was still predictive in multivariate testing sets regarding all subtypes of the study group. Due to missing clinical data in single cases cox regression analyses was available in 146 out of 156 cases.