Perioperative changes in osteopontin and TGFβ1 plasma levels and their prognostic impact for radiotherapy in head and neck cancer

Patients with newly diagnosed squamous cell carcinoma of the head and neck (HNSCC) were consecutively enrolled in two prospective trials (A and B1, 2). In group A we included patients with locally confined tumors which were eligible for primary resection. After giving their written informed consent, blood samples were taken at different time points: T0) before surgery, T1) 1 day after surgery, T2) 1 week and T3) 3 to 4 weeks after surgery. Blood samples were immediately centrifuged and plasma was stored at -80 °C. Group B1 consisted of patients who were medically or technically not eligible for surgical interventions or who refused surgery. In group B2 we recruited patients who were treated by primary surgery and were referred to adjuvant treatment according to their final tumor stage. Clinico-pathological patient characteristics are summarized in Table  1. Patients in group A were treated with primary surgery. According to national guidelines these patients received adjuvant treatment when appropriate. No adjuvant treatment was started before time point T3. In group B plasma samples of patients were analyzed before and during radio-(chemo) therapy (definitive treatment n = 41 (B1), postoperative treatment n = 89 (B2)). Patients from group B were enrolled before the start of the second trial (group A). The study was approved by the local clinical ethics committee. For a better understanding of the trial a scheme is shown in Fig.  1.

Blood was anticoagulated with EDTA and subsequently centrifuged (4000 rpm) at room temperature for 10 min. Plasma was removed, aliquoted and stored at -80 °C until use. For comparison of OPN we used archived plasma samples collected from group B which had been prepared in the same way. These samples were collected just before the start of radiotherapy (T0).

Aliquots of each sample were analyzed in duplicate using the Human Osteopontin Assay Kit-IBL (Immuno-Biological Laboratories Co., Ltd., Japan) according to the manufacturer’s instructions.

The same aliquots were analysed in duplicate using a commercially available kit (ELISA Pro Kit for Human Latent TGFβ1, Mabtech, Sweden) according to the manufacturer’s instructions. Absolute plasma concentrations for osteopontin and TGFβ1 are given in ng/ml.

All statistical analyses were done with SPSS for Windows version 23.0 (IBM SPSS, Inc.). Statistical significance was set at p < 0.05. All reported p values were two-sided. For comparison of patient characteristics Fischer’s Exact test was used. Student’s t-test was used for comparison of plasma concentrations between groups. To test for correlations between plasma osteopontin and TGFβ1 we used Pearson product-moment correlation coefficient. Analysis of variance (ANOVA) was used for evaluation of OPN and TGFβ1 distribution among the different clinical parameters. For comparison of OPN and TGFβ1 values at different time points we employed a general linear model for repeated measures for each plasma marker. Kaplan-Meier analysis using log-rank statistics were used for comparing overall survival. As done in the DAHANCA 5 OPN sub-study [ 18] and TROG 02.02 study [ 20] groups were divided according to tertiles and median values of OPN and TGFβ1 concentrations.

Table  1 describes the patient groups. Age and gender were comparable. Control patients were significantly younger. Patients from group B1 had more advanced tumor stages compared to group A and B2.

There was no association of OPN and TGFβ1 with clinical tumor parameters (e.g., histology, TNM- or UICC stage; data not shown).

Mean (±SD) osteopontin plasma concentration (ng/ml) was higher in patient groups compared to healthy controls (group A: 630.8 ± 353.0 ng/ml, group B1: 811.5 ± 365.1 ng/ml, group B2: 734.7 ± 310.1 ng/ml, controls: 478.9 ± 155.0 ng/ml; p = 0.028, p = 0.008 and p = 0.04 for group A, B1 and B2 vs controls, respectively, Fig.  2a). TGFβ1 plasma levels differed significantly between group A (15.23 ± 11.6 ng/ml) and group B1 (25.5 ± 11.8 ng/ml), p = 0.002 and between group B1 and controls (18.2 ± 10.1 ng/ml), p = 0.046 (Fig.  2b).

Mean osteopontin plasma concentrations (ng/ml) for the different time points T0 to T3 (mean ± SD) was 631 ± 353, 1363 ± 660, 936 ± 526 and 649 ± 374, p < 0.01 (Fig.  3a). The most prominent difference was seen directly after surgery between time points T0 and T1. Three to four weeks after surgery OPN concentration reached base line levels again (T0 and T3). Patients with higher OPN concentrations (> median) at the time of surgery showed also higher values 3–4 weeks postoperatively (Fig.  3c, p < 0.05).

No significant changes were observed in the time course of TGFβ1 concentrations (Fig.  3b) with the highest TGFβ1 values at time points T2 and T3 (as we would expect it in wound healing).

Pretherapeutic plasma concentrations of osteopontin and TGFβ1 values were analysed by the Pearson correlation coefficient test. We observed a significant positive correlation between both parameters, R = 0.619, p = 0.001 (Fig.  4).

Both osteopontin and TGFβ1 at the start of treatment correlated with patient overall survival (Figs.  5a-d). Higher OPN values were associated with a shorter overall survival. Median survival times were 11.5 and 33.0 months, p = 0.017 in patients with definitive radiochemotherapy (group B1). Median survival was 33.4 months for patients with higher OPN values and was not reached for lower OPN values ( p = 0.031) in patients treated with postoperative RT (group B2). In group A (patients with earlier tumor stage partly with no adjuvant treatment) survival was also worse for the high OPN group but the difference was not statistically significant (survival at 3 years was 76 and 95%, p = 0.13). Patients with TGFβ1 values in the upper tertile showed a worse outcome with median survival times of 10.7 and 33.0 months, p = 0.003 (group B1).

We propose the hypothesis that the transitory rise in OPN plasma levels in the postoperative setting is associated with wound healing and not caused by OPN secretion or expression from cancer cells since its increase was seen within 24 h. It is well known that OPN is not a tumor specific protein and can also originate from immune cells like macrophages or from endothelial cells [ 23, 24]. In wound healing there is a wide range of cells and cytokines which are differentially expressed [ 25]. Therefore we chose TGFβ1 as a representative marker and looked for changes in its expression patterns. We observed an increase of its plasma concentration peaking at 1 week after surgery which is in line with data from the literature [ 26, 27]. Changes of OPN and TGFβ1 levels were correlated (R = 0.62). From preclinical studies there is good evidence for an OPN mediated TGFβ1 expression [ 28– 30]. This is in agreement with the kinetics observed in this study, peak concentration of TGFβ1 lagged behind.

Transforming growth factor beta 1 is both expressed by tumor cells and adjacent stroma [ 31– 33]. Prognostic impact of plasma levels is therefore controversial [ 34– 38]. In this patient cohort we observed a significant negative correlation of pre-therapeutic TGFβ1 with overall survival.

The prognostic significance of osteopontin in head and neck cancer has been reported in patients treated by definite radiotherapy [ 15, 16, 18, 39] and is thought to relate to an association with tumor hypoxia and malignant phenotype. A hypoxic sensitizer (nimorazole) was of benefit in the high osteopontin tertile in the Dahanca 5 study. In contrast, data from TROG 02.02 did not find an association with survival parameters [ 20] and no predictive correlation with tirapazamine treatment.

Our data support a role of OPN as a prognostic biomarker for inoperable patients (treated with definite radiochemotherapy) and extend this observation to patients with combined surgery and radiotherapy.

Limitations of this and other single center studies are caused by the limited sample size. Furthermore, despite the fact that there is a large body of data on OPN detection there is still not a generally validated and certified test, making cross study comparisons more difficult. Most groups have been using an ELISA based system. But still there is also no standard ELISA kit, which would make at least these data more comparable. As shown by Vordermark et al. OPN values differed significantly when different ELISA systems were applied [ 40]. Also different OPN values are generated when using plasma or serum samples. For TGFβ1 the described ELISA system can only detect the total latent form and not the functionally active form of TGFβ1 which could also lead to some bias.