Expression and clinical value of EGFR in human meningiomas

Introduction

Meningiomas account for approximately 30% of intracranial tumors in adults, and despite being predominantly benign, many cases recur (Backer-Grondahl, Moen & Torp, 2012; Perry et al., 2007; Willis et al., 2005). Biomarkers capable of identifying recurrent cases could provide better treatment options and closer follow-up for such patients. A potential candidate is the epidermal growth factor receptor (EGFR), which has been shown to be overexpressed and/or amplified in many human cancers (Libermann et al., 1984; Rokita et al., 2013; Salomon et al., 1995; Torp et al., 1991). Ligands such as EGF (epidermal growth factor) and TGFα (transforming growth factor alpha) bind to the extracellular domain (ECD) of the receptor, leading to effects on differentiation, growth, migration, adhesion, or apoptosis (Yarden & Sliwkowski, 2001).

EGFR and the ligands EGF and TGFα have been shown to be overexpressed in human meningiomas (Andersson et al., 2004; Baxter et al., 2014; Caltabiano et al., 2013; Camby et al., 1997; Carroll et al., 1997; Diedrich et al., 1995; Guillaudeau et al., 2012; Halper et al., 1999; Johnson et al., 1994; Kuratsu et al., 1994; Laurendeau et al., 2009; Lusis, Chicoine & Perry, 2005; Maiuri et al., 2007; Narla et al., 2014; Reubi et al., 1989; Torp et al., 1992; Wernicke et al., 2010), whereas non-neoplastic meningeal tissue show sparse or no reactivity (Carroll et al., 1997; Johnson et al., 1994; Torp et al., 1992), indicating a potential tumorigenic role of EGFR in these tumors. Both membranous and cytoplasmic EGFR immunoreactivity have been described in meningiomas (Guillaudeau et al., 2012; Halper et al., 1999; Horsfall et al., 1989; Johnson et al., 1994; Jones et al., 1990; Smith et al., 2007; Torp et al., 1992), however, the prognostic relevance is scarcely described (Guillaudeau et al., 2012). The receptor also seems to be activated (phosphorylated) (Carroll et al., 1997; Hilton et al., 2016). Moreover, the literature is conflicting regarding expression levels of EGFR across malignancy grades, as some report increasing expression with higher grades (Caltabiano et al., 2013; Diedrich et al., 1995; Halper et al., 1999) and others the opposite or no connection at all (Baxter et al., 2014; Guillaudeau et al., 2012; Jones et al., 1990; Kuratsu et al., 1994; Narla et al., 2014; Wernicke et al., 2010). Furthermore, the prognostic value of expression of either the intracellular or extracellular domain of the receptor is uncertain (Guillaudeau et al., 2012).

The aim of this study was to investigate the expression of the intracellular and extracellular domains of EGFR, as well as its activated status and ligands in a large population-based series of human meningiomas with long follow-up data and the association with histopathological features, meningioma subtypes, malignancy grade, and risk of recurrence.

Methods

Results

Clinical characteristics are presented in Table S1. Among the 186 meningiomas, 130 (69.9%) were benign, 55 (29.6%) atypical, and one (0.5%) malignant (Table S2). The median age at the date of operation was 59 years (range: 25–86), and the female to male ratio was 2.9:1. Results of immunohistochemical analyses and SI related to histological features are shown in Tables 1 and 2, respectively. Typical immunostainings are shown in Fig. 1. No cases were excluded from the study due to tissue loss or during processing failures. The TMA tissue cores were well preserved after processing, and all three cores were intact for most cases. EMA immunoreactivity was seen in 177/180 meningiomas (98.3%) as well as in normal leptomeninges, which were negative for the other antibodies.

ICD immunoreactivity was found in all meningioma cases at high levels with a median SI of 9.00. Concerning cellular localization, the ICD showed both membranous and cytoplasmic staining, yet with a clear membranous predominance. There was no association between expression of ICD and subtypes (p = 0.765), localization (p = 0.862), and malignancy grade (p = 0.983, Table 2). However, ICD levels were high in tumors lacking psammoma bodies (p = 0.005, Table 2).

Immunoreactivity for ECD was also observed in most cases (99.5%), although the SI was much lower than that observed for ICD with a median SI of 4.00. Membranous immunoreactivity was slightly stronger than cytoplasmic. The ECD was lower in fibrous tumors than in transitional tumors (p = 0.028), atypical (p = 0.044), and meningothelial (p = 0.036). Further, the ECD was higher in tumors presenting sheeting (p = 0.035, Table 2), and in tumors lacking psammoma bodies (p < 0.001, Table 2). There was no association between either malignancy grade or localization and ECD SI (p = 0.787 and p = 0.562, respectively).

Phosphorylated EGFR was found in most cases (99.5%) with a median SI of 9.00. Both membranous and cytoplasmic immunostaining was present in most tumors, although the latter demonstrated higher intensity. There was no association with any histological features. Concerning malignancy, expression was higher in grade II tumors compared to grade I (p = 0.018, Table 2). Further, expression was lower in skull base than convexity (p < 0.001) and falcine tumors (p = 0.038).

Immunoreactivity for EGF and TGFα was strong and seen in all cases with median SIs of 9.00. Statistical analyses concerning histological features and malignancy grades were also performed for grade I and grade II tumors individually (Tables S3 and S4, respectively.)

In univariable Cox regression analyses (Table 3), high ECD expression was significantly associated with reduced survival or recurrence in the time to recurrence (TTR) analysis (HR 1.152, CI (1.036–1.280), p = 0.009). When tumor grades were analyzed individually for this antibody, a similar result was achieved only for benign tumors (n = 38, HR 1.213, CI (1.064–1.383), p = 0.004). In the multivariable survival analysis (Table 4), only Simpson resection grade was significantly associated with worse TTR (HR 2.606, CI (1.558–4.359), p < 0.001), while ECD expression did not show such association (HR 1.099, CI (0.987–1.222), p = 0.084). For ICD, phosphorylated receptor, EGF, and TGFα no associations were found with TTR or OS (Table 3), even when grades were analyzed separately.

Regarding κ-statistics variable scores were established (Table S5).

Discussion

In the present study we have demonstrated that EGFR and its ligands EGF and TGFα are widely expressed in human meningiomas and that the receptor appears to be in an activated state. Increased ECD expression was associated with poorer survival in a univariable analysis. However, it did not prove to be an independent predictor of survival in the multivariable analysis.

Since normal meninges were EGFR immunonegative, our findings support a general upregulation of this receptor in meningioma tissue in accordance with previous results and other studies (Carroll et al., 1997; Di Carlo et al., 1992; Halper et al., 1999; Horsfall et al., 1989; Huisman et al., 1991; Johnson et al., 1994; Kuratsu et al., 1994; Laurendeau et al., 2009; Lusis, Chicoine & Perry, 2005; Maiuri et al., 2007; Narla et al., 2014; Reubi et al., 1989; Torp et al., 1992). These findings are also supported by studies in which EGFR expression in meningiomas was determined by other techniques such as Northern and western blots, polymerase chain reactions (PCR) and ligand-binding studies (Andersson et al., 2004; Carroll et al., 1997; Guillaudeau et al., 2012; Kurihara et al., 1989; Laurendeau et al., 2009; Torp et al., 1992; Weisman, Raguet & Kelly, 1987). As such, immunohistochemistry appears as a reliable and practical method to determine EGFR expression in meningioma tissue. For instance, this may be useful in diagnostics to distinguish between normal and neoplastic meninges as well as to detect meningioma infiltration in soft tissue. The upregulation of EGFR in neoplastic meningeal tissue supports a role of this receptor as a facilitator in the tumorigenesis of these tumors. Whether this is linked to tumor initiation or progression is, however, unclear. The increased levels of EGFR may be a result of enhanced transcription, translation or decreased receptor turnover, but is not likely due to gene amplification, as this is not described in meningiomas (Guillaudeau et al., 2012; Torp et al., 1992).

Few studies have focused on the different expression patterns of the internal and external domains of EGFR in human meningiomas (Guillaudeau et al., 2012). In the present study, nearly all tumors expressed both domains (99.5% for the ECD, 100% for ICD), even though the internal domain had an overall higher intensity. High levels of ICD positivity have also been observed in high-grade astrocytomas and non-small cell lung cancer (Gulati et al., 2010; Mascaux et al., 2011; Torp et al., 2007), although negative immunoreactivity has also been reported (Wickremesekera, Hovens & Kaye, 2010). Concerning the ECD, most studies have shown high expression (Andersson et al., 2004; Baxter et al., 2014; Camby et al., 1997; Diedrich et al., 1995; Guillaudeau et al., 2012), whereas some have recorded lower positivity (Jones et al., 1990). The observed stronger expression of the ICD compared with the ECD may be a result of a fragile ECD damaged during tissue processing or due to various shedding mechanisms (sEGFR) (Perez-Torres et al., 2008). In high-grade astrocytomas such a discrepancy may be due to EGFR gene mutation and truncated receptors, which have not been reported in meningiomas (Aldape et al., 2004; Fan et al., 2013; Guillaudeau et al., 2012; Torp et al., 1992). Contradictory to our findings, Guillaudeau et al. (2012) found higher expression of the extracellular domain compared with the intracellular domain. The mechanisms of these discrepancies remain uncertain, but expression of different EGFR isoforms and various antibodies may be part of the issue (Guillaudeau et al., 2012).

In agreement with others, widespread EGF and TGFα immunoreactivity was observed in all meningiomas (Carroll et al., 1997; Halper et al., 1999; Hsu, Efird & Hedley-Whyte, 1998; Torp, Unsgaard & Dalen, 1993). Both cytoplasmic and membranous localizations were found, with the latter showing lower intensity. However, some did not find any membranous positivity for TGFα (Halper et al., 1999; Hsu, Efird & Hedley-Whyte, 1998). Strong cytoplasmic reactivity may either be due to antibodies binding ligands produced in the tumor cells or internalized ligand–receptor complexes (Tomas, Futter & Eden, 2014). Expression of these EGFR ligands in meningioma tissue has also been demonstrated by other techniques such as Northern and protein blots, PCR, in situ hybridization, and radioimmunoassays (Carroll et al., 1997; Halper et al., 1999; Torp, Unsgaard & Dalen, 1993), supporting the applicability of immunohistochemistry to investigate the presence of these ligands as well. Furthermore, these findings support the existence of autocrine/paracrine growth loops in human meningiomas.

EGFR expression alone does not reveal the activation status, so clarifying whether the receptor is activated or not is relevant. There are sparse data on this issue in meningioma tissue (Carroll et al., 1997; Hilton et al., 2016), and according to our findings the activated receptor is expressed at high levels in most tumors. In their study, Hilton et al. found significantly higher expression of phosphorylated EGFR in tumor tissues compared with non-neoplastic tissues, and high expression of downstream signaling molecules (Hilton et al., 2016). The literature is conflicting concerning EGFR and malignancy, as some have found that high expression is positively related to tumor grade (Caltabiano et al., 2013; Diedrich et al., 1995; Halper et al., 1999) whereas others have found the opposite or no differences (Baxter et al., 2014; Guillaudeau et al., 2012; Jones et al., 1990; Kuratsu et al., 1994; Narla et al., 2014; Wernicke et al., 2010). However, these previous studies have not investigated the activated receptor. Thus, our results suggest that the expression level of the activated receptor can be useful in meningioma grading.

Both membranous and cytoplasmic EGFR immunoreactivity were observed in the meningioma tissue, however, membranous reactivity dominated in most cases for both the ECD and ICD. Distinguishing between these patterns of immunostaining was often problematic, mostly due to high cellularity and fibrous growth. For this reason it seems inappropriate to make such a distinction in human meningiomas, although it may be of clinical relevance in other human malignancies such as renal cell carcinoma (Pu et al., 2009). A study on colorectal cancer, however, demonstrated that cellular EGFR localization was unrelated to clinicopathological parameters or patient outcome (McKay et al., 2002).

Despite the uniform histological appearance of meningiomas, we observed both heterogeneous immunoreactivity in the form of hotspots and homogeneous EGFR immunoreactivity, pointing to meningiomas as heterogeneous in this regard. EGFR expression also varied with regard to meningioma subtypes, suggesting that EGFR may play different roles in tumorigenesis of these variants. Concerning histology, the association between high expression of the internal and external domains and absence of psammoma bodies is interesting as the latter favors poorer survival (Backer-Grondahl et al., 2014). Further, imaging studies suggest that calcified meningiomas have a slower growth rate compared with uncalcified tumors (Nakamura et al., 2003; Nakasu et al., 2005). Accordingly, if there exists a connection between microscopic- (psammoma bodies) and macroscopic calcification, EGFR immunostaining may act as a marker for fast growing tumors.

Concerning survival, only high expression of ECD was significantly associated with decreased TTR, also when benign meningiomas were evaluated separately. Thus, high expression of ECD may be an indicator of recurrence-prone benign meningiomas. This is in contrast to other studies, which report better survival with high levels of ECD and similarily, poorer survival with low expression of ECD (Guillaudeau et al., 2012; Smith et al., 2007). Others report no association between EGFR expression and survival (Caltabiano et al., 2013). Contrary to our findings, Maiuri et al. (2007) found higher EGFR positivity in non-recurrent benign meningiomas. Further, we observed no association between antibody expression and OS. These discrepancies may be due to different antibodies, tumor size, observation time and endpoints. The expression of the ECD, however, was not significant in the multivariable analysis. ECD SI is therefore not an independent predictor of survival for meningioma patients. Thus, the prognostic value of EGFR in human meningiomas appears uncertain and needs to be further clarified.

The key strengths of this study are a population-based study design with long observation time and a large number of patients. Limitations are its retrospective profile and the subjective nature of immunohistochemistry. To compensate for interobserver variability, all cases were evaluated by two observers independently. In addition, the weighted kappa-statistics showed moderate or substantial agreement (Landis & Koch, 1977), suggesting reproducible results. Another uncertainty is the antigenicity of the archived tumor samples (Economou et al., 2014), which may vary between tumors since the inclusion period was over 10 years. The use of TMAs has potential advantages and disadvantages, and our evaluation of the immunostainings might have been interfered by heterogeneous expression patterns. To compensate for this we tried to include TMA cores representing the varying histological patterns. All in all, TMAs have benefits in many aspects such as reproducibility and uniform assay conditions (Camp, Neumeister & Rimm, 2008; Voduc, Kenney & Nielsen, 2008).

Conclusions

EGFR and its ligands are frequently overexpressed in human meningiomas compared with normal meninges, and the receptor is generally in an activated state. These findings support the role of this growth factor system in the tumorigenesis of these tumors. Accordingly, EGFR is a potential candidate for targeted therapy. Regarding clinical value, only overexpression of the external domain of EGFR was associated with prognosis in univariable analyses. It was not, however, significant in the multivariable analysis, and as such EGFR status appears to have minor significance as a prognosticator.

Supplemental Information