Stromal collagen type VI associates with features of malignancy and predicts poor prognosis in salivary gland cancer

Abstract

Purpose

Collagen Type VI (COLVI) is an extracellular matrix protein that is upregulated in various solid tumours during tumour progression and has been shown to stimulate proliferation, suppress apoptosis and promote invasion and metastasis. It has also been described as a mediator of chemotherapy resistance and as a therapeutic target in preclinical cancer models. Here, we aimed to analyse the prognostic role of COLVI in salivary gland cancer (SGC).

Methods

Stromal COLVI protein expression was assessed in primary SGC specimens of 91 patients using immunohistochemistry (IHC). The IHC expression patterns obtained were subsequently correlated with various survival and clinicopathological features, including Ki-67 and p53 expression.

Results

We found that COLVI was expressed in all SGC specimens. High expression was found to be associated with features of malignancy such as high histologic grades, advanced and invasive T stages and metastatic lymph node involvement (p < 0.05 for all variables). COLVI expression was also found to correlate with both Ki-67 and p53 expression (p < 0.01). We found that high COLVI expression predicted a significantly inferior 5-year overall survival (38.3%, 55.1% and 93.8%; p = 0.002) and remained a significant predictor of prognosis in a multivariate Cox regression analysis (hazard ratio, 2.62; 95% confidence interval, 1.22–5.61; p = 0.013). In all low-risk subgroups COLVI expression identified patients with an adverse outcome. Patients receiving adjuvant radiotherapy had a poor survival when expressing high levels of COLVI.

Conclusions

Our data indicate that stromal COLVI expression associates with key features of malignancy, represents a novel independent prognostic factor and may affect response to radiotherapy in SGC. Although our results warrant validation in an independent cohort, assessing stromal COLVI expression may be suitable for future diagnostic and therapeutic decision making in patients with SGC.

Electronic supplementary material

The online version of this article (10.1007/s13402-018-0389-1) contains supplementary material, which is available to authorized users.

Introduction

Salivary gland carcinomas (SGC) are rare malignancies of the head and neck, most of which arise in the parotid gland. According to the WHO 2005 classification, it comprises 24 histological subtypes [1]. Published 5-year overall survival rates vary between 46 and 78% [2]. Clinicopathological factors consistently associated with prognosis include age, TNM stage, tumour grade and Ki-67 expression [2–4]. Surgery - complemented by adjuvant radiotherapy in patients with adverse risk factors - remains the primary treatment modality [5–7]. Definitive radiotherapy is an option for patients with non-resectable disease [5]. So far, chemotherapy and targeted agents have not shown any meaningful survival benefit when added to the adjuvant setting, or when used in advanced stages or at relapse [8, 9]. However, responses to systemic therapy have been described in a proportion of patients [8, 9], but the mechanisms governing sensitivity and refractoriness to treatment are still obscure.

Collagen type VI (COLVI) is a multimeric extracellular matrix protein expressed in a variety of human soft tissues [10, 11]. It has previously been shown to be upregulated during tumour progression and to regulate many cellular features that are considered as hallmarks of cancer, including proliferation, angiogenesis, apoptosis, invasion and metastasis [12–18]. As a chemoattractant for macrophages, COLVI is also able to promote tumour inflammation [14, 19, 20]. Furthermore, chemoprotective properties of COLVI have been established in different preclinical models of cancer [21–23]. It has been found that binding of COLVI to β1-integrins and the proteoglycan NG2 can trigger signalling cascades involved in tumour progression and therapy resistance, such as the serine/threonine kinase AKT or the tyrosine kinase FAK pathways [24–30]. Interestingly, COLVI has recently been described as a critical component of the muscle stem cell niche, regulating satellite cell self-renewal and muscle regeneration [31]. In addition, the COLVI cleaved C5A fragment (endotrophin) has recently been shown to increase cancer stem cell self-renewal through activation of the anthrax toxin receptor 1 (ANTXR1) [32]. Here, we report a comprehensive analysis on the role of COLVI in a large cohort of patients with SGC.

Materials and methods

Patients and samples

Protein expression was analysed on tissue microarrays (TMA) containing formalin-fixed paraffin-embedded parotid gland cancer samples obtained from 91 newly diagnosed patients at the time of surgical resection [33]. The treatment modality was definitive surgery and adjuvant radiotherapy when adverse risk factors were present. The median follow-up of surviving patients was 28 months (range 0 to 206 months). The study was conducted according to the Declaration of Helsinki on biomedical research involving human subjects. Written informed consent was obtained from each patient and the scientific protocol was approved by the local research ethics committee (2012–370-f-S). Details on the patient cohort have been published previously [33] and are listed in Table 1.

Table 1.

Clinicopathological characteristics according to COLVI expression status

Variable	COLVI
	low	intermediate	high	p-value *
N	19	42	30
Age, years				0.057§
Median (range)	54 (22–90)	61 (11–88)	73 (36–90)
Sex, no (%)				0.268¶
Male	7 (36.8)	23 (54.8)	18 (60.0)
Female	12 (63.2)	19 (45.2)	12 (40.0)
Parotidectomy, no (%)				0.024 ‡
Lateral	3 (15.8)	2 (4.8)	6 (20.0)
Subtotal	0 (0.0)	0 (0.0)	1 (3.3)
Total	15 (79.0)	23 (54.8)	12 (40.0)
Radical	1 (5.3)	17 (40.5)	11 (36.6)
Neck dissection, no (%)				0.333‡
None	2 (10.5)	3 (7.1)	3 (10.0)
Selective	11 (57.9)	27 (64.3)	13 (43.3)
Radical	2 (10.6)	8 (19.0)	7 (23.3)
Unknown	4 (21.1)	4 (9.5)	7 (23.3)
Radiotherapy adjuvant, no (%)				0.100¶
Yes	9 (47.4)	25 (59.5)	21 (70.0)
No	10 (52.6)	16 (38.1)	6 (20.0)
Unknown	0 (0.0)	1 (2.4)	3 (10.0)
T classification, no (%)				0.024 ‡
pT1	6 (31.6)	14 (33.3)	7 (23.3)
pT2	10 (52.6)	9 (21.4)	5 (16.7)
pT3	1 (5.3)	11 (26.2)	12 (40.0)
pT4	1 (5.3)	7 (16.7)	5 (16.7)
Unknown	1 (5.3)	1 (2.4)	1 (3.3)
N classification, no (%)				0.035 ‡
pN0	13 (68.4)	25 (59.5)	10 (33.3)
pN1	1 (5.3)	1 (2.4)	4 (13.3)
pN2	2 (10.5)	14 (33.4)	14 (46.6)
pN3	0 (0.0)	0 (0.0)	0 (0.0)
unknown	3 (15.8)	2 (4.8)	2 (6.7)
Grading, no (%)				0.015 ‡
1	5 (26.3)	4 (9.5)	2 (6.7)
2	6 (31.6)	7 (16.7)	9 (30.0)
3	1 (5.3)	16 (38.1)	10 (33.3)
4	0 (0.0)	0 (0.0)	2 (6.7)
Unknown	7 (36.8)	15 (35.7)	7 (23.3)
R classification, no (%)				0.057¶
R0	16 (84.2)	23 (54.8)	17 (56.7)
R1	2 (10.5)	17 (40.5)	11 (36.7)
R2	1 (5.3)	2 (4.8)	2 (6.7)
Histology, no (%)				< 0.001 ‡
Acinic cell carcinoma	9 (47.4)	4 (9.5)	0 (0.0)
Mucoepidermoid carcinoma	1 (5.3)	3 (7.1)	3 (10.0)
Adenoid cystic carcinoma	1 (5.3)	6 (14.3)	2 (6.7)
Cystadenocarcinoma	0 (0.0)	1 (2.4)	0 (0.0)
Low-grade cribriform cystadenocarcinoma	0 (0.0)	1 (2.4)	0 (0.0)
Salivary duct carcinoma	0 (0.0)	8 (19.0)	13 (43.3)
Adenocarcinoma NOS	2 (10.5)	4 (9.5)	0 (0.0)
Squamous cell carcinoma	1 (5.3)	5 (11.9)	4 (13.3)
Large cell undifferentiated carcinoma	0 (0.0)	0 (0.0)	2 (6.7)
Carcinoma ex pleomorphic adenoma	2 (10.5)	9 (21.4)	4 (13.3)
Myoepithelial carcinoma	0 (0.0)	1 (2.4)	2 (6.7)
Basal cell carcinoma	2 (10.5)	0 (0.0)	0 (0.0)
Sebaceous carcinoma	1 (5.3)	0 (0.0)	0 (0.0)

*p-values are for the comparison of clinical characteristics between low, intermediate and high COLVI expressers. Bold values indicate significance (p ≤ 0.05)

^§Kruskal-Wallis test; ^¶ Chi square test; ^‡ Fisher’s exact test

Immunohistochemistry

The TMAs were prepared and processed as reported before [33]. Briefly, staining was performed on 3 μm TMA sections using the APAAP method (Dako). Nuclei were counterstained with Mayer’s haemalum (Merck). The following primary antibodies were used for staining: anti-COLVI (Novus Biologicals, NB120–6588, 1:200), anti-Ki-67 (MIB-1 clone; Dako, M7240, 1:100) and anti-p53 (Dako; clone DO-7; 1:3000). Stromal COLVI immunoreactivity was scored by three investigators who were blinded towards the clinical history and patient outcome, with immunoreactivity scores (IRS) of 0 (no staining), 1 (scant staining), 2 (moderate staining) or 3 (extensive staining). P53 was scored by multiplying the nuclear staining intensity (0–3 as above) with the percentage of p53 positive cells.

Statistical analyses

Clinical and pathological data were compared between groups using the χ² or Fisher’s exact test for categorical and the Mann-Whitney-U or Kruskal-Wallis test for continuous variables, where appropriate. Spearman’s rank correlation test was used for assessing correlations between clinicopathological variables and COLVI expression. Survival was estimated according to the Kaplan-Meier method and compared by log-rank test. Overall survival (OS) was defined from initial diagnosis to last follow-up (censored) or death (event). For survival analysis in distinct clinicopathological subgroups, IRS of 1 and 2 were considered as low and IRS of 3 as high COLVI expression. A Ki-67 score above the median and a p53 score above 10 were considered high. Multivariable Cox regression analysis was used to assess independent statistical significance. Clinical relevant variables with a p-value < 0.1 upon univariate analysis were included in the model (Supplementary Table S1, Supplementary material). Two-sided p-values ≤ 0.05 were considered significant. All analyses were performed using IBM SPSS Statistics for Macintosh, version 24.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism version 7.0 for Macintosh (GraphPad Software, La Jolla, CA, USA).

Results

COLVI expression associates with features of malignancy in SGC

Using immunohistochemistry (IHC) expression of COLVI was found in all SGC specimens tested, with 21%, 46 and 33% of the cases exhibiting immunoreactivity scores (IRS) of 1, 2 and 3, respectively (Fig. 1A-C). Association of COLVI expression with clinical and pathological characteristics is shown in Table 1. We found that COLVI expression associated with malignant features and tumour aggressiveness such as grading, invasive tumour stages, metastatic potential, proliferation and loss of tumour suppressor protein p53 (Fig. 1D,E). Overall, we found that high histologic grading according to WHO 2005 classification was associated with increased COLVI scores (3%, 55% and 41% for G3–4 vs 35%, 29% and 35% for G1–2; p = 0.006; Fig. 1D). In addition, we found a significant co-occurrence of higher COLVI expression with advanced and invasive T stages (5%, 49% and 46% in pT3–4 vs 31%, 45% and 24% in pT1–2; p = 0.006; Fig. 1D). Patients with metastatic lymph node involvement had significantly higher COLVI expression scores than those without (8%, 42% and 50% in pN+ vs 29%, 49% and 22% in pN0; p = 0.009; Fig. 1D). We also found that the number of involved lymph nodes increased with progressively higher COLVI expression levels (mean number of metastatic lymph nodes 0.6, 3.3 and 5.3 for IRS 1, 2 and 3, respectively; p = 0.01; Fig. 1E). In fact, COLVI expression positively correlated with the number of involved lymph nodes (r = 0.340; p = 0.002). High COLVI expression was also found to be associated with expression of the proliferation marker Ki-67 (mean Ki-67 7.8%, 19.4 and 23.6% for IRS 1, 2 and 3 respectively; p = 0.001; Fig. 1E) and with increased cellular staining of the tumour suppressor protein p53 (mean p53 score 11.7, 35.0 and 72.8 for IRS 1, 2 and 3, respectively; p = 0.031; Fig. 1E) as an indicator of mutated TP53 [34]. Both Ki-67 and p53 were found to be positively correlated with COLVI expression (r = 0.278 and 0.341; p = 0.008 and 0.001, respectively).

Fig. 1.

Expression of COLVI in distinct clinical and histopathological subgroups. a Representative images of COLVI immunoreactivity scores (IRS) in histologic salivary gland cancer specimens. IRS of 1, 2, and 3 were considered as low, intermediate and high, respectively. Original magnification for all images 200×. b Absolute frequencies of COLVI expression scores in the complete patient cohort. COLVI was expressed in all tumours in variable degrees. Relative frequencies were 21%, 46%, and 33% for IRS of 1, 2, and 3, respectively. c COLVI expression in all salivary gland cancer entities. d COLVI expression in distinct clinicopathological subgroups. High COLVI expression scores were more frequently observed in the context of malignant features such as high tumour grade, invasive T stages and lymph node metastases. P ≤ 0.05 for all groups (e) Mean number of lymph node metastases, mean percentage of Ki-67 positive tumour cells and mean p53 scores in patients with low, intermediate and high COLVI expression levels. Error bars indicate standard error of the mean. f OS according to trichotomized COLVI expression in the entire patient cohort. Patients with a low COLVI expression (IRS 1) had a better outcome than patients with an intermediate (IRS 2) or a high expression (IRS 3)

COLVI expression correlates with survival in SGC patients

We found that patients with a low COLVI expression had a significantly better OS than patients with a high expression (5-year OS 93.8%, 55.1%, and 38.3% for IRS 1, 2, and 3, respectively; p = 0.002; Fig. 1F). In a multivariate analysis including age, gender, T stage, N stage, histology, grading and Ki-67, COLVI expression remained a significant predictor of OS (hazard ratio, 2.62; 95% confidence interval, 1.22–5.61; p = 0.013; Table 2).

Table 2.

Multivariate Cox proportional hazard analysis for overall survival

Variables in the final model	Hazard ratio (95% confidence interval)	p-value
T stage	2.39 (1.12–5.10)	0.024
Grading	2.47 (1.14–5.35)	0.022
COLVI expression	2.62 (1.22–5.61)	0.013

T stage (pT3–4 vs pT1–2), grading (G3–4 vs G1–2), COLVI expression (high vs low); reference category underlined

COLVI expression has a prognostic impact in distinct SGC subgroups

To analyse the impact on OS in distinct clinicopathological subgroups the patient cohort was dichotomized into low (IRS 1 and 2) and high (IRS 3) COLVI expressers. We found that high COLVI expression was significantly associated with decreased survival in patients with otherwise favourable prognostic factors such as low histologic grading (31.1 and 83.1%, for high and low COLVI expression; p = 0.028; Fig. 2A), early T stage (36.4 and 81.3%; p = 0.017; Fig. 2B), pN0 stage (48.0 and 80.4%; p = 0.034; Fig. 2C), low Ki-67 staining (51.9 and 87.8%; p = 0.005; Fig. 2D) and low p53 expression (39.3 and 77.1%; p = 0.002; Fig. 2E). In contrast, we found that COLVI expression had no significant impact on OS in the corresponding adverse prognostic subgroups (high histologic grading, advanced T stage, positive N stage, high Ki-67 and high p53; Supplementary Fig. S1A-E, Supplementary material). Taken together, we found that high COLVI expression abrogated the beneficial prognostic impact of favourable features and identified patients in each subgroup with OS rates that were in the range of the corresponding adverse subgroup (5-year OS for: high grading 25.4%; pN+ 28.8%; advanced T stage 35.9%; high p53 31.5%; high Ki-67 34.4%; Fig. 2A-E, right panel).

Fig. 2.

Impact of COLVI expression on overall survival in favourable risk subgroups. In all favourable risk subgroups, patients with high COLVI expression had an inferior outcome (a-e, left panel), which was in the range of the corresponding adverse risk subgroups (a-e, right panel). Subgroups included: (a) low vs high tumour grading, (b) early vs advanced T stages, (c) absence vs presence of metastatic lymph nodes, (d) low vs high proliferation and (e) low vs high staining of p53

In addition, we found that the prognostic significance remained evident in patients treated with adjuvant radiotherapy (5-year OS 34.2 and 64.8% for high and low COLVI expression; p = 0.005; Fig. 3). No significant effect on survival was observed in the small subgroup of patients not receiving radiation therapy (Supplementary Fig. S1F, Supplementary material). Overall, we found that patients receiving adjuvant radiation therapy showed a trend towards a worse survival when compared to untreated patients (5-year OS 53.6 and 74.8%; p = 0.086) (Fig. 3). Interestingly, we found that the subgroup of patients with a low COLVI expression who received adjuvant treatment had similar outcomes as the favourable risk group of non-treated patients (5-year OS 64.8 and 74.8%), while patients with high COLVI levels had a far worse outcome (34.2%) despite adjuvant radiation (Fig. 3).

Fig. 3.

Impact of COLVI expression on overall survival in patients treated with adjuvant radiotherapy. Patients receiving adjuvant radiotherapy due to adverse risk factors had a significantly inferior outcome when expressing high levels of COLVI (left), whereas treated patients with low COLVI expressing tumours had survival rates comparable to those of favourable risk patients not in need for adjuvant radiation (right)

Discussion

Here, we identified intratumoural COLVI expression, as measured by immunohistochemistry, as a novel predictor of prognosis in SGC, which is independent from established risk factors, including tumour grade. Specifically, in all classical favourable risk subgroups, patients with a significant adverse clinical outcome could be identified by COLVI expression. No additional adverse effect of COLVI expression on survival in the corresponding adverse risk subgroups was noted, but this observation may be biased by low patient numbers. Thus, especially in the context of patients with favourable risk factors, COLVI adds relevant information for a more individualized prediction of prognosis in this disease. Still, our results warrant further validation preferentially within the framework of a prospective trial. Evaluation of circulating PRO-C6, a product specifically generated during COLVI formation, as a prognosticator or tumour marker might be included in such studies [35].

We found that COLVI expression was associated with many features of malignancy including proliferation, invasion, metastasis and loss of p53. In addition, we found that high COLVI expression was consistently associated with an adverse outcome in all favourable, but in none of the adverse, risk subgroups. Considering these findings and its established tumour promoting role in different cancer models [14, 16, 24] it is likely that COLVI itself is contributing to tumour progression in SGC patients. Thus, COLVI receptors or their downstream signalling pathways may constitute potential novel targets for pharmacological interventions using monoclonal antibodies or small molecule inhibitors. In fact, inhibitors of AKT and FAK are currently being investigated in clinical trials in patients with solid tumours [36, 37]. In addition, antibody-based targeting of ANTXR1 or the cleaved C5A fragment of COLVI have shown therapeutic efficacy in various murine cancer models [21, 38]. Finally, it has been shown that even stromal proteins such as COLVI may be exploited for targeted delivery approaches using monoclonal antibodies carrying bioactive payloads, such as cytotoxic drugs, cytokines or radionuclides [13, 39, 40].

Interestingly, we found that in the preselected subgroup of patients with adverse risk factors receiving adjuvant radiotherapy, COLVI expression retained its prognostic impact. Of note, patients with low COLVI levels treated with adjuvant radiotherapy and untreated favourable risk patients had comparable outcomes, while patients with high COLVI levels had a worse prognosis despite adjuvant radiation. Given that COLVI had no additional impact on survival when analysing all adverse risk subgroups alone, it is conceivable that high COLVI expression may, at least in part, contribute to an unfavourable outcome by conveying resistance to radiation. Even though to our knowledge COLVI has not yet been reported to mediate radio-resistance in cancer, this notion would be supported by the well-described anti-apoptotic and chemoprotective properties of COLVI [17, 21, 22]. While previous studies could not establish a meaningful survival benefit for chemotherapy in SGC, responses to chemotherapy have been described in subsets of patients [8, 9]. Unfortunately, specimens obtained from patients treated with chemotherapy were not available in our study. It is tempting, however, to speculate that COLVI may influence responsiveness to chemotherapy in this disease as well. Future clinical trials in SGC should incorporate measurement of COLVI expression, to relate therapy responses to the respective tumour phenotypes.

In summary, we conclude that our work identifies COLVI as a novel independent prognostic factor in SGC and points toward a critical role of COLVI in tumour progression and resistance to radiotherapy. Assessing stromal COLVI expression may complement routine pathologic work-up of SGC for a more accurate prediction of the tumour’s biologic and clinical behaviour and may ultimately help to improve therapeutic decision-making.

Funding

This work was supported by the fund “Innovative Medical Research” of the University of Münster Medical School (grants SC211008 (C.S.), SC111411 (C.S.), and SC221410 (L.A. and C.S.)). W.E.B. is supported by the Deutsche Forschungsgemeinschaft (DFG EXC 1003, Cluster of excellence “Cells in Motion”).

Compliance with ethical standards

Conflict of interest

The authors declare no conflicts of interest.