Targeting Cyclooxygenase-2 in Pheochromocytoma and Paraganglioma: Focus on Genetic Background

Cyclooxygenase 2 (COX-2) is a key enzyme of the tumorigenesis-inflammation interface and can be induced by hypoxia. A pseudohypoxic transcriptional signature characterizes pheochromocytomas and paragangliomas (PPGLs) of the cluster I, mainly represented by tumors with mutations in von Hippel–Lindau (VHL), endothelial PAS domain-containing protein 1 (EPAS1), or succinate dehydrogenase (SDH) subunit genes. The aim of this study was to investigate a possible association between underlying tumor driver mutations and COX-2 in PPGLs. COX-2 gene expression and immunoreactivity were examined in clinical specimens with documented mutations, as well as in spheroids and allografts derived from mouse pheochromocytoma (MPC) cells. COX-2 in vivo imaging was performed in allograft mice. We observed significantly higher COX-2 expression in cluster I, especially in VHL-mutant PPGLs, however, no specific association between COX-2 mRNA levels and a hypoxia-related transcriptional signature was found. COX-2 immunoreactivity was present in about 60% of clinical specimens as well as in MPC spheroids and allografts. A selective COX-2 tracer specifically accumulated in MPC allografts. This study demonstrates that, although pseudohypoxia is not the major determinant for high COX-2 levels in PPGLs, COX-2 is a relevant molecular target. This potentially allows for employing selective COX-2 inhibitors as targeted chemotherapeutic agents and radiosensitizers. Moreover, available models are suitable for preclinical testing of these treatments.

Gene expression profiling provided the basis for classifying PPGLs according to their main transcriptional signatures underlying the aforementioned mutations: cluster I presents with activation of pseudohypoxic signaling pathways and includes mainly VHL-, EPAS1-, and SDHx-mutant cases; cluster II is enriched in kinase receptor signaling pathways and is comprised of RET-, NF1-, TMEM127-, MAX-, and HRAS-mutant cases [22,23]. Both pseudohypoxia and kinase receptor signaling are involved in regulating apoptosis, proliferation, invasion, and metastasis, and angiogenesis via different mechanisms, but can also contribute to inflammatory conditions in various tumor entities [24][25][26][27].
The first report on COX-2 gene expression and immunohistochemistry in adrenal pheochromocytomas was published in 2001 suggesting that the enzyme might have a role in malignant transformation of these tumors [41]. Between 2007 and 2011, another four immunohistochemical studies were published supporting the value of COX-2 as surrogate marker that, in association with other factors, could potentially discriminate between benign and metastatic pheochromocytoma [42][43][44][45]. Due to literature showing that COX-2 is induced by hypoxia signaling [46,47], we hypothesized that cluster I PPGLs have a higher COX-2 expression than cluster II. Accordingly, COX-2 may be a promising molecular target for functional imaging and adjuvant treatment, in particular in cluster I PPGLs.
To address the above hypothesis, we evaluated COX-2 status of PPGLs with known mutational status for VHL, SDHx, EPAS1, NF1, RET, and HRAS on both mRNA and protein level. Furthermore, we characterized COX-2 immunoreactivity in tumor spheroids and allografts derived from mouse Statistical analysis taking into account the general clinico-morphologic features of the tumors showed that gender, tumor location, tumor diameter, age at diagnosis, or metastatic behavior had no relevant influences on COX-2 gene expression (Table S1). On the other hand, COX-2 expression was significantly higher in head and neck PPGLs (p = 0.01) compared to other locations. Of note, all head and neck PPGLs in this series were related to an SDHD germline mutation.

Figure 1.
Normalized COX-2 gene expression in PPGLs with regard to genetic background. gene expression array data were derived from Lopez-Jimenez et al. [50] and Qin et al. [51] mRNA expression series. Seventy samples with known genotype were included in this analysis and classified according to the specific gene mutated, origin of mutations, and transcriptional cluster; numbers in parentheses represent the number of samples investigated in each subgroup; see Table S1 for statistical analyses.
To further investigate a possible association between the pseudohypoxic signature and COX-2 in cluster I PPGLs, unsupervised clustering for 97 hypoxia-related genes (see materials and methods for details) showed that VHL-and SDHx-mutant cases clustered together ( Figure S1). Pearson correlation for COX-2 expression and pseudohypoxic signature indicated significant relationships (p < 0.05) for 86 out of 171 probes, representing 65 out of 97 genes related to hypoxia in PPGLs (Table S2). Correlation coefficients ranged from 0.19 to 0.63 with 65 probes having only a weak correlation below 0.4. Exemplary, consistent with high COX-2 expression, the analysis showed a significant positive relationship with mRNA levels of Ca 2+ -dependent phospholipase A2 (r = 0.564, p < 0.001) since the enzyme is required for releasing arachidonic acid from phospholipid membranes as the specific substrate for cyclooxygenases.

COX-2 Immunoreactivity in Clinical PPGL Tissue Samples of a Second Cohort
COX-2 immunoreactivity was assessed in formalin-fixed paraffin-embedded tumor samples from a separate cohort sharing no case with the RNA sample cohort. This series included 96 PPGLs with a clinically documented mutation in tumor susceptibility genes ( Table 2) and reflects the expected age, location, and metastatic disease, according to the mutations involved. Most cases were adrenal pheochromocytomas (52%), followed by thoracic and abdominal paragangliomas (26%), and  Table S1 for statistical analyses.
To further investigate a possible association between the pseudohypoxic signature and COX-2 in cluster I PPGLs, unsupervised clustering for 97 hypoxia-related genes (see materials and methods for details) showed that VHLand SDHx-mutant cases clustered together ( Figure S1). Pearson correlation for COX-2 expression and pseudohypoxic signature indicated significant relationships (p < 0.05) for 86 out of 171 probes, representing 65 out of 97 genes related to hypoxia in PPGLs (Table S2). Correlation coefficients ranged from 0.19 to 0.63 with 65 probes having only a weak correlation below 0.4. Exemplary, consistent with high COX-2 expression, the analysis showed a significant positive relationship with mRNA levels of Ca 2+ -dependent phospholipase A2 (r = 0.564, p < 0.001) since the enzyme is required for releasing arachidonic acid from phospholipid membranes as the specific substrate for cyclooxygenases.

COX-2 Immunoreactivity in Clinical PPGL Tissue Samples of a Second Cohort
COX-2 immunoreactivity was assessed in formalin-fixed paraffin-embedded tumor samples from a separate cohort sharing no case with the RNA sample cohort. This series included 96 PPGLs with a clinically documented mutation in tumor susceptibility genes ( Table 2) and reflects the expected age, location, and metastatic disease, according to the mutations involved. Most cases were adrenal pheochromocytomas (52%), followed by thoracic and abdominal paragangliomas (26%), and head and neck paragangliomas (22%). Germline mutations were documented in 52% of cases. At the time of investigation, 13% of cases showed metastatic disease. Most of the cases carried a mutation in SDHx (41% comprising 18 SDHD, 3 SDHC, 9 SDHB, 5 SDHA, and 4 SDHAF2 cases) followed by NF1 (22%), VHL (15%), RET (9.4%), EPAS1 (7.3%), and HRAS (6.3%). The SDHx subgroup showed the highest proportion of extra-adrenal paragangliomas (39% thoracic and abdominal, and 51% head and neck), whereas all other genetic subgroups included mostly adrenal pheochromocytomas (57−95%). Tumor diameters were significantly smaller in the SDHx subgroup (3.9 cm) and significantly higher in the RET subgroup (7.2 cm) compared to cases with other genetic backgrounds. Metastatic disease was most frequently documented among SDHx cases (21%) compared to all other genetic subgroups (0−17%).
COX-2 immunoreactivity was assessed by three observers using a three-mark score taking into account the percentage of positively stained tumor cells per tissue section ( Figure 2). Sections with 'strong' (>50% of tumor cells stained) or 'moderate' score (20-50% of tumor cells stained) showed cytoplasmic COX-2 immunoreactivity in pheochromocytes and/or interconnected stromal cells. Tumors with 'negative or weak' score (<20% of tumor cells stained) showed COX-2 immunoreactivity predominantly in few stromal cells scatted over the tissue section. Interobserver variation statistics showed good agreement between the first two observers (weighted κ = 0.67). However, there were 14 cases of disagreement (15%) between 'negative or weak' and 'moderate' scores as well as 16 cases of disagreement (17%) between 'moderate' and 'strong' scores that where passed to a third observer for final decision. Cancers 2019, 11, x 6 of 17 COX-2 immunoreactivity was strong in 23 samples (24%) and moderate in 35 samples (36%) whereas 38 samples (40%) showed negative or weak staining. Tumor location, tumor diameter, age at diagnosis, or metastatic behavior had no statistically relevant influences on COX-2 immunoreactivity (Table S1). On the other hand, COX-2 immunoreactivity was significantly higher in samples from male patients compared to females and all genetic subgroups showed different sex  (Table S1). On the other hand, COX-2 immunoreactivity was significantly higher in samples from male patients compared to females and all genetic subgroups showed different sex ratios. However, multiple regression analyses, testing the relationships between COX-2 expression and the two independent variables 'genetic background' and 'sex' simultaneously, showed that trends in COX-2 immunoreactivity of the genetic subgroups were not artifacts of different sex ratios.
Cancers 2019, 11, x 7 of 17 ratios. However, multiple regression analyses, testing the relationships between COX-2 expression and the two independent variables 'genetic background' and 'sex' simultaneously, showed that trends in COX-2 immunoreactivity of the genetic subgroups were not artifacts of different sex ratios.

COX-2 as Molecular Target in Preclinical PPGL Models
We further assessed the COX-2 status of a commonly used preclinical model of mouse pheochromocytoma (MPC) cells with heterozygous Nf1 knockout. In vitro, MPC spheroids showed strong and homogeneous COX-2 immunoreactivity in pheochromocytes involving the most peripheral 8−10 cellular layers, whereas COX-2 immunoreactivity was absent in the necrotic core ( Figure 5A). In vivo, MPC tumors in a subcutaneous allograft model showed strong tumor-specific uptake of a red-fluorescent COX-2 imaging probe ( Figure 5B). Tissue sections from these allografts showed strong and homogenous COX-2 immunoreactivity predominantly involving pheochromocytes.  Table S1 for statistical analyses; scale bars: 0.1 mm.

COX-2 as Molecular Target in Preclinical PPGL Models
We further assessed the COX-2 status of a commonly used preclinical model of mouse pheochromocytoma (MPC) cells with heterozygous Nf1 knockout. In vitro, MPC spheroids showed strong and homogeneous COX-2 immunoreactivity in pheochromocytes involving the most peripheral 8−10 cellular layers, whereas COX-2 immunoreactivity was absent in the necrotic core ( Figure 5A). In vivo, MPC tumors in a subcutaneous allograft model showed strong tumor-specific uptake of a red-fluorescent COX-2 imaging probe ( Figure 5B). Tissue sections from these allografts showed strong and homogenous COX-2 immunoreactivity predominantly involving pheochromocytes.
Expression of COX-2 was assessed on mRNA or protein level in two separate cohorts of PPGL patients with known tumor driver mutations. Both cohorts were comprised of tumors with a similar distribution of clinical features in respect to sex, tumor location, age at diagnosis, and metastatic behavior, reflecting previously described features of PPGLs [52].
Despite a significant increase in COX-2 mRNA levels in cluster I compared to cluster II PPGLs, we did not find a significant relationship between the pseudohypoxic transcriptional signature of cluster I PPGLs and COX-2 in clinical samples. COX-2 protein levels are consistent with these results showing also no significant difference between cluster I and cluster II PPGLs. This may be due to the fact that PPGLs are characterized by a high degree of intertumoral heterogeneity. Many different factors have been described to activate and interfere with COX-2 in cancer [28]. Amongst others, intratumoral differences in normoxic/hypoxic conditions, systemic chemotherapy, oxidative stress, or even tobacco smoking can interfere with COX-2 levels on gene expression and protein level, possibly masking a potential association with the pseudohypoxic signature of cluster I PPGLs. Nevertheless, trends were observed that may at least partially be explained by the underlying genetic background.
In cluster I PPGLs with loss-of-function mutations in VHL and SDHx, pseudohypoxic transcriptional phenotypes may contribute to COX-2 induction on gene expression and protein level. Functional defects of VHL, an E3 ubiquitin ligase, directly impair ubiquitin labeling of hypoxiainducible factors (HIF-α) for regular proteasomal degradation. Functional defects of SDH indirectly impair ubiquitin-labeling of HIF-α caused by intracellular accumulation of succinate, an intrinsic inhibitor of prolyl hydroxylases. Therefore, both VHL and SDH defects are associated with enhanced
Expression of COX-2 was assessed on mRNA or protein level in two separate cohorts of PPGL patients with known tumor driver mutations. Both cohorts were comprised of tumors with a similar distribution of clinical features in respect to sex, tumor location, age at diagnosis, and metastatic behavior, reflecting previously described features of PPGLs [52].
Despite a significant increase in COX-2 mRNA levels in cluster I compared to cluster II PPGLs, we did not find a significant relationship between the pseudohypoxic transcriptional signature of cluster I PPGLs and COX-2 in clinical samples. COX-2 protein levels are consistent with these results showing also no significant difference between cluster I and cluster II PPGLs. This may be due to the fact that PPGLs are characterized by a high degree of intertumoral heterogeneity. Many different factors have been described to activate and interfere with COX-2 in cancer [28]. Amongst others, intratumoral differences in normoxic/hypoxic conditions, systemic chemotherapy, oxidative stress, or even tobacco smoking can interfere with COX-2 levels on gene expression and protein level, possibly masking a potential association with the pseudohypoxic signature of cluster I PPGLs. Nevertheless, trends were observed that may at least partially be explained by the underlying genetic background.
In cluster I PPGLs with loss-of-function mutations in VHL and SDHx, pseudohypoxic transcriptional phenotypes may contribute to COX-2 induction on gene expression and protein level. Functional defects of VHL, an E3 ubiquitin ligase, directly impair ubiquitin labeling of hypoxia-inducible factors (HIF-α) for regular proteasomal degradation. Functional defects of SDH indirectly impair ubiquitin-labeling of HIF-α caused by intracellular accumulation of succinate, an intrinsic inhibitor of prolyl hydroxylases. Therefore, both VHL and SDH defects are associated with enhanced HIF-α signaling even under normoxic conditions, a metabolic state referred to as pseudohypoxia [25]. From investigations on other tumor entities, in particular on colon cancer, it is known that both HIF-α isoforms (1 and 2) are capable of directly upregulating COX-2 expression [26,27].
COX-2 mRNA levels as well as the percentage of moderate and high COX-2 immunoreactivity tended to be lower among cases carrying a gain-of function mutation in EPAS1, encoding the HIF-2α protein, compared to VHLand SDHx-mutant cases. This observation suggests that activation of HIF-2α alone may not be sufficient for COX-2 upregulation in PPGLs.
In cluster II PPGLs with loss-of-function mutations in NF1 gene, the trend for a relatively high COX-2 immunoreactivity is consistent with another report on elevated COX-2 and prostaglandin E2 (PGE 2 ) levels in NF1 malignant peripheral nerve sheath tumors [53]. These observations raise the possibility that functional defects in NF1, a GTPase-activating protein, could indirectly contribute to the upregulation of COX-2 expression. This may at least be partly explained by elevated levels of activated Ras-GTP leading to hyperactivation of mitogen-activated protein kinase (MAPK) pathways. Under normoxic conditions, COX-2 expression can be induced through activation of oncogenic pathways such as Ras-MAPK and can even be further enhanced by HIF-1α during hypoxia [36]. The lack of high COX-2 mRNA in the gene expression cohort might be due to the low number of NF1-mutant cases in this particular set of tumors. On the other hand, all NF1 cases in the RNA sample cohort carried germline mutations, whereas all NF1 cases in the tissue sample cohort carried somatic mutations. Whether there is a relationship between germ line or somatic NF1 mutations and different COX-2 levels in PPGLs remains to be investigated.
In accordance with the observations in clinical PPGL samples, COX-2 immunoreactivity was also high in spheroids and subcutaneous allografts derived from mouse pheochromocytoma (MPC) cells with a heterozygous Nf1 knockout. Therefore, these models are suitable for preclinical testing of COX-2-targeted treatments for the management of PPGLs. Since we did not find a significant relationship between tumor driver mutations and COX-2 in clinical PPGL samples, molecular imaging could be applied in a personalized approach to pre-estimate whether a tumor is susceptible to COX-2-targeted treatment. In our study, specific accumulation of a red-fluorescent COX-2 probe in subcutaneous MPC allografts demonstrates the potential value of COX-2 tracers for assessing the target status in PPGLs non-invasively. In order to translate this approach into clinical practice, studies have been initiated on the use of selective COX-2 inhibitors as PET radiotracers for cancer imaging [54][55][56].
In the case of HRAS and RET mutations, trends for lower COX-2 mRNA levels in PPGLs are in agreement with COX-2 immunoreactivity. These findings suggest that both HRAS and RET have not major role in regulating COX-2 expression in PPGLs. However, there have been reports on fibroblasts transformed with a mutant HRAS responding with a rapid induction of COX-2 on gene expression and protein level [57]. It has also been reported that HRAS expression increases COX-2 expression in intestinal epithelial cells [58]. In thyroid cancer, RET has been shown to activate Ras, and thus it could indirectly lead to COX-2 activation, however, whether RET could activate COX-2 in any other way is a matter of investigation [59].
The observation that tumor diameter and age at diagnosis had no statistically relevant impact on COX-2 levels is in accordance with previous studies [41][42][43][44][45]. In contrast to these reports, we did not detect a statistically relevant increase of COX-2 mRNA and COX-2 immunoreactivity in primary tumors of metastatic PPGLs. This is most likely due to the relatively small number of metastatic cases in our cohorts. Significantly higher COX-2 mRNA levels in head and neck PPGLs compared to other tumor locations is related to the fact that all head and neck PPGLs in this series carried an SDHD mutation. This raises the possibility that COX-2 expression may be regulated by SDHD-related metabolic alterations in particular in head & neck PPGLs. However, due to low sample numbers available in the SDHD subgroup, further studies focusing on COX-2 expression in specific SDH mutation subtypes are necessary to draw a conclusion from these initial observations. In our tissue series, head and neck PPGLs comprised of cases with different tumor driver mutations explaining why similar effects on COX-2 were not detected in this series. Significantly higher COX-2 protein in PPGLs from male patients compared to females is considered a specific characteristic of our tissue sample cohort since we did not detect a similar effect in the RNA samples.
Higher COX-2 immunoreactivity in stromal cells of PPGLs related to SDHx germline mutations compared to other tumor driver mutations indicates a systemic effect of partial SDHx loss on COX-2 levels. This pattern of COX-2 immunoreactivity may be related to structure-supporting sustentacular cells and/or a characteristic monocytic component in PPGLs that has recently been discovered [60]. Further histologic investigations are required to elucidate the specific cell populations of stromal COX-2 immunoreactivity in PPGLs. A report on COX-2 in cervical cancer showed that the ratio between COX-2 in tumor cells and COX-2 in stroma cells was very effective in distinguishing patients with low versus high risk of death from disease. A very strong relationship between both tumor COX-2 expression and tumor-to-stromal COX-2 ratio has been shown to be highly correlated with response to chemotherapy while, high COX-2 expression in the stroma was significantly associated with better survival, but failed to directly correlate with response to treatment [61]. Further studies are required to fully elucidate the role of COX-2 in PPGL tumorigenesis and therapy resistance.

Tumor Samples and Genetic Testing
For immunohistochemical analysis, a series of 96 tumors from patients with confirmed PPGL diagnosis were used for this study. The cohort was recruited in the four participating centers: Tumor and Normal Tissue Bank of the UCC/NCT at the Universal Hospital Carl Gustav Carus, Dresden, Germany, Spanish National Cancer Research Centre, Madrid, Spain, Radboud University Medical Centre, Nijmegen, The Netherlands, and University of Florence, Italy. All patients provided informed consent to collect clinical and genetic data, in accordance with institutional ethical-approved protocols for each center. Metastatic cases were defined based on clinical documentation of metastases or extensive local invasion. Genetic screening was performed in germline and tumor DNA using a next-generation sequencing panel (PheoSeq) as previously described [62].

Gene Expression Profiling and Data Processing
Gene expression array data were extracted from [50,51]. To investigate the association between a pseudohypoxic transcriptional signature and COX-2 in cluster I PPGLs on RNA level, a published list of 782 genes significantly differentially expressed between VHLand SDHB-mutant cases was taken into account [50]. These genes were compared with the hypoxia database including all genes theoretically related to hypoxia [63]. Unsupervised clustering was applied with 97 hypoxia-related genes overlapping from both lists ( Figure S1). Pearson correlation coefficients ® were calculated between COX-2 expression and each of the 97 genes.

Scoring of COX-2 Immunoreactivity
For each case, COX-2 immunoreactivity was analyzed from a series of bright field images (magnification, ×100) contiguously captured along the diameter of one tumor section (images per section > 5). Perinuclear and cytoplasmic red-brown staining was considered positive. PPGLs form dense, reticular to glandular 'zellballen' or intermediate forms [64]. Typically, structure-supporting sustentacular cells are closely associated with tumor cells [65]. Taking into account these specific histologic features of PPGLs, our examination assessed COX-2 immunoreactivity in both inflammatory and sustentacular cells of the stromal compartment and/or pheochromocytes.
The percentage of COX-2-positive tumor cells was assessed using a three-mark score adapted from [41]: negative or weak (<20% of tumor cells); moderate (20-50% of tumor cells); strong (>50% of tumor cells). Samples were evaluated independently by two histologically experienced observers (Martin Ullrich and Verena Seifert) who were blinded to the genetic subtype of tumors. In cases of disagreement, samples were referred to a third observer for final decision (Sandra Hauser). Notably, our scoring system does not report on staining intensities that were observed to vary between samples from different centers most likely due to differences in tissue quality, preservation techniques, and storage time.

Tumor Allograft Models
Animal experiments were carried out at the Helmholtz-Zentrum Dresden-Rossendorf according to the guidelines of German Regulations for Animal Welfare and have been approved by the local Animal Ethics Committee for Animal Experiments (Landesdirektion Dresden, Germany). Subcutaneous tumor allografts were generated through injection of luciferase-expressing MPC LUC/eGFP-ZEO cells (abbreviated as MPC LUC/GZ ) passage 11 into female NMRI-nude mice (Rj:NMRI-Foxn1 nu , homozygous, T cell-deficient, hairless; Janvier Labs, Le Genest-Saint-Isle, France) as described previously. Five weeks after cell injection, optical in vivo imaging was performed (tumor diameters between 0.8 and 1.2 mm). After imaging, mice were sacrificed using CO 2 inhalation and cervical dislocation. Tumors were excised, fixed in paraformaldehyde, and embedded in paraffin according to standard procedures (n = 6).

Optical In Vivo Imaging
Optical tumor imaging in mice was performed on a preclinical In-Vivo Xtreme imaging system (Bruker, Billerica, MA, USA) under general anesthesia with inhalation of 10% (v/v) desflurane (Baxter, Unterschleißheim, Germany) in 30% (v/v) oxygen air. Location and morphology of luciferase-expressing MPC LUC/GZ allografts were assessed using bioluminescence imaging (BLI) as described previously [49]. Functional COX-2 imaging was performed using the RediJect COX-2 Fluorescent Imaging Probe (PerkinElmer, Waltham, MA, USA) injected intraperitoneally according to manufacturer's instructions. Fluorescence imaging (FLI) was performed three hours after injection. Specific fluorescence of the imaging probe was captured at λ Ex/Em = 570/600 nm and non-specific fluorescence was captured at λ Ex/Em = 480/535 nm. Tumor uptake was analyzed in processed images, showing specific fluorescence/non-specific fluorescence ratios.

Statistical Analysis
Graphs were drawn using Prism version 5.02 (GraphPad Software, San Diego, CA, USA). Incidences of COX-2 immunoreactivity within a defined subgroup are presented as percent of cases, n represents the number of cases. If not stated differently, data are presented as means ± standard error of the means. Significance of differences was tested for n ≥ 6 using the Mann-Whitney U test. Multiple linear regression analysis was performed using OriginPro 2017G (OriginLab Corporation, Northhampton, MA, USA). Relationships were described with the regression coefficient r and considered significant at p-values < 0.05. Interobserver variation was calculated using online QuickCalcs κ statistics version 06/2014 (GraphPad Software, www.graphpad.com/quickcalcs/kappa1).

Conclusions
Moderate to high cyclooxygenase 2 (COX-2) gene expression and immunoreactivity in about 60% of PPGLs demonstrates that, for these patients, COX-2 is considered a clinically relevant molecular target for adjuvant, in particular radiosensitizing treatments using selective COX-2 inhibitors, e.g., in combination with 177 Lu-DOTA-TATE endoradiotherapy. However, taking into account the genetic background of the samples, is an indicator but not the major determinant for COX-2 expression in PPGLs. High COX-2 immunoreactivity in tumor spheroids and subcutaneous tumor allografts derived from mouse pheochromocytoma (MPC) cells demonstrates that available PPGL models are suitable for preclinical in vitro and in vivo testing of COX-2-targeting treatments.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/11/6/743/s1, Figure S1: Unsupervised clustering for 97 hypoxia-related genes in PPGLs applied to our mRNA expression series extracted from [50,51]; VHLand SDHx-mutant cases clustered together whereas, other cases showed no homogeneous profile, Table S1: Statistical analyses of COX-2 status with regard to clinical characteristics and genetic background of tumors in two independent series of PPGL samples; data in parentheses were calculated from low sample numbers (n < 7); as Mann-Whitney U test showed significant sex-related differences in COX-2 immunoreactivity in tissue samples, multiple regression analyses was applied to distinguish whether the trends observed in the genetic subgroups were related to genetic background or different sex ratios only; levels of significance: * p < 0.05; † p <0.01, ‡ p < 0.001, Table S2: Pearson correlation between expression of COX-2 and 97 hypoxia-related genes in PPGLs in the mRNA expression series extracted from [50,51].