N-Glycomic Profiling of Pheochromocytomas and Paragangliomas Separates Metastatic and Nonmetastatic Disease

Abstract

Context

No effective methods for separating primary pheochromocytomas and paragangliomas with metastatic potential are currently available. The identification of specific asparagine-linked glycan (N-glycan) structures, which are associated with metastasized pheochromocytomas and paragangliomas, may serve as a diagnostic tool.

Objective

To identify differences in N-glycomic profiles of primary metastasized and nonmetastasized pheochromocytomas and paragangliomas.

Setting

This study was conducted at Helsinki University Hospital, University of Helsinki, and Glykos Finland Ltd. and included 16 pheochromocytomas and paragangliomas: 8 primary metastasized pheochromocytomas or paragangliomas and 8 nonmetastasized tumors.

Methods

N-glycan structures were analyzed with matrix-assisted laser desorption-ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) profiling of formalin-fixed, paraffin-embedded tissue samples.

Main Outcome Measure

N-glycan profile of tumor tissue.

Results

Four groups of neutral N-glycan signals were more abundant in metastasized tumors than in nonmetastasized tumors: complex-type N-glycan signals of cancer-associated terminal N-acetylglucosamine, multifucosylated glycans (complex fucosylation), hybrid-type N-glycans, and fucosylated pauci-mannose-type N-glycans. Three groups of acidic N-glycans were more abundant in metastasized tumors: multifucosylated glycans, acid ester–modified (sulfated or phosphorylated) glycans, and hybrid-type/monoantennary N-glycans. Fucosylation and complex fucosylation were significantly more abundant in metastasized paragangliomas and pheochromocytomas than in nonmetastasized tumors for individual tests but were over the false positivity critical rate, when adjusted for multiplicity testing.

Conclusions

MALDI-TOF MS profiling of primary pheochromocytomas and paragangliomas can identify diseases with metastatic potential based on their different N-glycan profiles. Thus, malignancy-linked N-glycan structures may serve as potential diagnostic tools for pheochromocytomas and paragangliomas.

MALDI-TOF MS profiling of primary metastasized and nonmetastasized pheochromocytomas and paragangliomas identified differences between their N-glycan profiles.

Paragangliomas (PGLs) arise from sympathoadrenal or parasympathetic paraganglia and pheochromocytomas (PHEOs) from the adrenal medulla. These are rare, neural crest–originating neuroendocrine tumors (1). PHEOs and sympathetic PGLs can secrete catechol amines, which can cause morbidity and mortality in patients. Usually, parasympathetic PGLs do not secrete catechol amines (1, 2). According to the World Health Organization, only metastasized tumors are regarded as malignant (3). Patients may have multiple primary tumors; therefore, the metastasis should be on a site where paraganglionic tissue is not normally present (1, 4). Approximately 10% of PHEOs and 15% to 35% of PGLs metastasize (4–6), and the challenge with this tumor group is to predict the metastatic potential of the primary tumor. Morphological scoring systems have been proposed to predict the aggressiveness of the tumor (7–10). None of these morphological scoring systems are well validated. Many studies have attempted to identify markers for predicting the metastatic potential of PHEOs or PGLs. For example, upregulation of human telomerase reverse transcription and heat-shock protein 90 has been associated with malignancy (11). Increased tissue expression of VEGF (12) and high proliferation index (13) has been associated with metastatic disease; however, no single marker unequivocally predicts the metastatic potential of these tumors. Recently, many mutations have been found in these tumors, and ≤40% of patients are estimated to have a germline mutation (14, 15). Germline or somatic mutations of ≥18 genes (NF1, RET, VHL, SDHA, SDHB, SDHC, SDHD, SDHAF2, TMEM127, MAX, HIF2A, KIF1B, PHD1, PHD2/EGLN1, FH, HRAS, BAP1, and MEN1) are involved in the development of PHEOs and PGLs. Tumors with different mutations may have different pathogenesis (14, 15). SDHB mutations are associated with more aggressive disease (16).

Glycans, which cover all human cells, are the carbohydrate units of glycoproteins, glycolipids, and proteoglycans. An estimated 1% of the human genome plays a role in glycan biosynthesis, which is the most complex posttranslational modification of proteins (17). This glycosylation process occurs most abundantly in the Golgi apparatus and the endoplasmic reticulum but also occurs in the cytoplasm and the nucleus (18). Glycans participate in essential cell functions, such as cell adhesion, motility, and intracellular signaling (18). Changes in these functions are crucial steps when cells transform into malignant ones; this transformation is also reflected in changes in the cell’s glycan profile, as observed in many cancers (19, 20). Glycosylation changes are related to the aggressive behavior of tumor cells and can serve as cancer biomarkers (19, 21, 22).

Cancer-associated asparagine-linked glycan (N-glycan) structures have been shown to participate in tumor progression and to promote growth (23, 24), invasion (25, 26), and angiogenesis (27). Changes in the N-glycan profiles have been described in many cancer types, including lung (28, 29), breast (30), and colorectal cancer (31, 32). In colorectal carcinomas, N-glycosylation profiles differ from those of healthy tissues, benign neoplastic lesions, and adenomas (31, 32).

The aim of this study was to investigate the N-glycomic profiles of PHEOs and PGLs by matrix-assisted laser desorption-ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) profiling of asparagine-linked glycans and to identify differences between metastatic and nonmetastatic primary tumors. Possible differences depending on the primary tumor site were also of interest.

Materials and Methods

Patients and tissues

For MS analysis, we chose 16 primary tumors from 16 patients with PHEO or PGL. Half of the patients (8) had metastatic disease (patients numbered 1–8 in Table 1). Metastasis was confirmed either histologically or radiologically by metaiodobenzylguanidine or somatostatin receptor scintigraphy. Of 16 tumors, 8 were PHEOs, and 8 were PGLs. Characteristics of the patients and tumors are presented in Table 1. Tissue samples were obtained from the Department of Pathology, HUSLAB, Helsinki University Hospital. Clinical data were obtained from the patient records of Helsinki University Hospital, and survival data were obtained from the Finnish Population Register Center. The causes of death were collected from Statistics Finland. This study was approved by the local ethics committee (Dnro HUS 226/E6/06, extension TMK02 §66 17.4.2013) and the National Supervisory Authority of Welfare and Health (TEO Dnro 3990/04/046/07).

Table 1.

Characteristics of the Patients

Patient Number	Sex and Age (y) at Diagnosis	PHEO or PGL	Time of First Metastasis (mo)	Location of the Metastases	Mutations (Tested)	SDHB-IHC (+/−)	Follow-Up (y)	Present Statusa
1	M 19	PHEO	92	Bone, peritoneum, mesenterium (H)	NF1	+	30	Died of the disease
2	M 69	PHEO	0	Lymph node (H)	ND	+	0	Died of the disease
3	M 45	PHEO	57	Lung (R), lymph node (H)	ND	+	19	Died of the disease
4	M 28	PHEO	55	Bone (R), liver, peritoneum, omentum (H)	No (VHL, RET, SDHB)	+	16	Died of the disease
5	M 39	PGL	131	Bone (R), probably brain (H)	SDHB	−	15	Died of the disease
6	M 48	PGL	0	Bone (H)	ND	−	8	Died of other causes
7	M 31	PGL	154	In liver hilus between liver and kidney	SDHB	−	21	Alive with the disease
8	F 57	PGL	96	Liver, lymph nodes (R)	No (MAX, VHL, NF1, REPRKAR1A, RET, SDHA, SDHAF2, SDHB, SDHC, SDHD, TMEM127)	+	10	Alive with the disease
9	F 59	PHEO			No (RET)	+	7	Alive with the disease
10	F 57	PHEO			ND	+	10	Alive with the disease
11	M 50	PHEO			No (RET)	+	8	Alive with the disease
12	M 43	PHEO			No (RET)	+	9	Alive with the disease
13	F 34	PGL			SDHB	−	11	Alive with the disease
14	M 77	PGL			ND	+	3	Died of other causes
15	M 48	PGL			SDHB	−	16	Alive with the disease
16	M 23	PGL			ND	−	16	Alive with the disease

Abbreviations: H, histology; ND, not determined; R, radiology; SDHB-IHC, succinate dehydrogenase B immunohistochemistry.

^aStatus 9 May 2016. With negative SDHB-staining, PGL syndromes 1–5 are possible.

Tissue samples for MS

Representative areas of tumor tissue were marked on hematoxylin and eosin slides. From corresponding areas of formalin-fixed, paraffin-embedded tissue blocks, samples were punched with a 3.0-mm needle. The volume of harvested tissue per patient was ≥1 mm³. The samples were deparaffinized with xylene and with an ethanol–water series according to standard procedures.

MS N-glycan profiling

Asparagine-linked glycans were detached from tissue glycoproteins by PNGase F digestion and purified by a series of microscale solid-phase extraction steps similar to a previously described protocol (32). MALDI-TOF MS was performed with a Bruker Ultraflex III TOF/TOF instrument (Bruker Daltonics Inc., Bremen, Germany). Neutral N-glycans were analyzed in positive ion reflector mode as [M+Na]⁺ ions, and acidic N-glycans were analyzed in negative ion linear mode as [M-H]⁻ ions. Relative molar abundances of neutral and acidic glycan components were assigned based on their relative signal intensities. The MS raw data were processed into the present glycan profiles by removing the effects of isotopic pattern overlapping, multiple alkali metal adduct signals, water elimination products from reducing oligosaccharides, and other interfering MS signals not arising from the sample similar to previously described protocols (29, 33). The glycan profiles were normalized to 100% to allow comparison between samples. The glycan signals were then assigned to biosynthetic groups based on their proposed monosaccharide composition (32, 33).

Succinate dehydrogenase analysis by immunohistochemistry

Immunohistochemical staining was done as previously described (34) with SDHB-antibody 21A11 (ABCAM, Cambridge, MA) in a dilution of 1:1000 and SDHA-antibody 5A11 (ABCAM) in a dilution of 1:1000. The staining result was positive, consistent with an intact succinate dehydrogenase (SDH) complex, when granular cytoplasmic staining existed and negative when tumor cells stained negative. Scoring was done independently by two researchers (M.M. and H.L.) without any knowledge of the clinical data.

Statistical analysis of MS data

Before we imported the data to the software, we removed mass/charge (m/z) variables with all values equal to 0 or with the total sum of the m/z values <1.5%. Altogether, 54 neutral m/z variables (initially, 102 variables) and 69 acidic m/z variables (initially, 171 acidic m/z variables) were included. In addition, 32 neutral and 37 acidic glycan class variables were analyzed. Principal component analysis (PCA) was performed using default parameters (analysis based on correlations, casewise missing data deletion). The Mann–Whitney test was used to compare differences in the glycan structures between metastasized and nonmetastasized tumors. When a statistically significant difference was identified by the Mann–Whitney test, we also calculated the mean relative amounts of glycan structures, the standard errors of the mean, and the fold change of the means between groups. Error propagation was used to assess standard error for the fold change. All tests were two-sided. A P value of 0.05 was considered significant. The Benjamini–Hochberg procedure, with the false positivity rate of 0.15, was used for multiplicity testing adjustment. Statistical analyses were performed with Statistica version 12.6 (StatSoft, version 12.6; Tulsa, OK, Dell Software) and SPSS version 20.0 (IBM SPSS Statistics, version 20.0; SPSS, Inc., Chicago, IL).

Results

Asparagine-linked glycan profiles in PHEOs and PGLs

N-glycan profiles of the four tumor types (PHEO with and without metastasis, PGL with and without metastasis) were successfully analyzed by MS. Figure 1 shows examples of N-glycan mass spectra that differed markedly between one metastasized and one nonmetastasized sample. For example, by comparing Figs. 1A and 1B, it was seen that in a metastasized PGL (Fig. 1A, patient no. 8 in Table 1), the major neutral N-glycan signal m/z 1257 (corresponding to the sodium adduct ion of the monosaccharide composition H5N2) was clearly more abundant than the glycan signal m/z 1743 (H8N2), whereas in the nonmetastasized PGL (Fig. 1B, patient no. 15 in Table 1), the glycan signal m/z 1743 was nearly as abundant as m/z 1257. In addition, many other neutral N-glycan signals, such as m/z 1485 (H3N4F1), m/z 1647 (H4N4F1), m/z 1955 (H5N4F2), and m/z 2101 (H5N4F3), were more abundant in the metastasized sample than in the nonmetastasized sample, whereas m/z 1581 (H7N2) and m/z 1905 (H9N2) were more abundant in the latter. Similarly, there were many acidic N-glycan signals that were more abundant in the metastasized PGL (Fig. 1C, patient no. 8 in Table 1), including m/z 1541 (H3N4F1P1), m/z 1711 (S1H4N3F1), m/z 2589 (S1H6N5F3), and m/z 3100 (S1H7N6F4), or in the nonmetastasized PGL (Fig. 1D, patient no. 15 in Table 1), including m/z 1565 (S1H4N3) and m/z 1930 (S1H5N4).

Figure 1.

Exemplary MALDI-TOF mass spectra of N-glycans isolated from one metastatic and one nonmetastatic tumor: Neutral N-glycan spectra from a metastatic PGL (A, patient no. 8 in Table 1) and nonmetastatic PGL (B, patient no. 15 in Table 1), and acidic N-glycan spectra from the same metastatic (C patient no. 8 in Table 1) and nonmetastatic (D patient no. 15 in Table 1) tumors. Major and differing N-glycan signals that are discussed in the text are highlighted with schematic symbols describing putative glycan structures: blue square, N-acetyl-d-glucosamine; green circle, d-mannose; red triangle, l-fucose; yellow circle, d-galactose; purple diamond, N-acetylneuraminic acid; circled P, sulfate or phosphate ester. The glycans are detected as singly charged ions, neutral glycans as [M+Na]⁺ ions, and acidic glycans as [M-H]⁻ ions. The x-axis shows the m/z ratio. The y-axis shows the relative signal intensity in arbitrary units. The difference in the relative abundance of the glycan signal m/z 1930 may be difficult to see only by comparing the height of the signals; however, by measuring the proportion of the signal intensity of m/z 1930 from the total sum of glycan signal intensities, the difference is clear: In C, the relative abundance of the glycan signal m/z 1930 is 10.7%, whereas in D, it is 18.9%. The data are presented in numeric format in Supplemental Table 1. (E and F) Histology from the same tumors. (E) Metastatic PGL (patient no. 8 in Table 1). In the figure is one mitosis. (F) Nonmetastatic PGL (patient no. 15 in Table 1). The tumor cells have nuclear variation. Histology is a poor predictor of prognosis. Scale bar 50 µm. Magnification ×400.

Thus, the malignant and nonmalignant samples clearly differed in their N-glycan profiles. Schematic putative glycan structures are shown in Fig. 1 to illustrate their differences. The histology of these tumor samples is shown in Fig. 1E (metastatic PGL patient no. 8 in Table 1) and Fig. 1F (nonmetastatic PGL patient no. 15 in Table 1).

Four patient samples from each tumor type were analyzed by the MS method to gain knowledge of which N-glycan structures were associated with either metastasized or nonmetastasized samples or with PHEO or PGL. The results of these analyses are presented as bar diagrams in Fig. 2.

Figure 2.

N-glycan profiles of metastatic and nonmetastatic PHEO and PGL tumor samples. The 40 most abundant neutral (A) and 40 acidic (B) glycan signals are shown. The height of the bar represents the average abundance of the glycan as a percentage of all detected glycans. The data are presented in numeric format in Supplemental Table 1. Four samples were analyzed in each tumor type. All glycan signals have been assigned to proposed monosaccharide compositions. Major and differing N-glycans discussed in the text are highlighted with schematic symbols describing putative glycan structures. See the legend of Fig. 1 for glycan symbols and further details. Error bars represent the standard error of the mean. The scale of the y-axis is nonlinear.

Neutral asparagine-linked glycan profiles

Figure 2A shows the MS profiles of neutral N-glycans from the four tumor types. The five most abundant glycan signals with compositions H5N2, H6N2, H7N2, H8N2, and H9N2 were identified as high-mannose type N-glycans based on their typical monosaccharide composition and high prevalence in all human tissues examined thus far. A few glycan signals were more abundant in samples from metastasized tumors compared with nonmetastasized tumors (highlighted with schematic putative glycan structures in Fig. 2). Among these N-glycan signals were H3N4F1, a complex-type N-glycan signal that has been shown to be indicative of cancer-associated terminal N-acetylglucosamine (29); multifucosylated glycans, such as H4N3F2, H5N4F3, and H5N5F3; and hybrid-type N-glycans, such as H6N3F1. Moreover, fucosylated N-glycans such as H3N2F1, identified as pauci-mannose type based on their monosaccharide composition, were more abundant in malignant tumors, whereas nonfucosylated pauci-mannose glycans such as H3N2 were more abundant in nonmalignant tumors.

Acidic asparagine-linked glycan profiles

Figure 2B shows the MS profiles of acidic N-glycans. Here, abundant glycan signals composed of no or only one deoxyhexose residue (fucose), such as S1H5N4, S1H5N4F1, S2H5N4, and S2H5N4F1, were more abundant in the nonmetastasized tumors. In contrast, malignant tumors were characterized by multifucosylated N-glycans, such as S1H5N4F2 and S1H6N5F3; acid ester–modified (sulfated or phosphorylated) glycans, such as H4N3F1P1, H3N4F1P1, and H4N5F2P1; and hybrid-type/monoantennary N-glycans, such as S1H4N3F1. These structural group assignments were evident from the observed monosaccharide composition (Supplemental Tables 1 and 2). However, the exact structures of the cancer-associated N-glycans are a subject of further study, and they are not reported here.

PCA

PCA was performed on the detected glycan signal and structural classification data from the samples. Figure 3 shows projections of individual samples based on different PCA factor planes. Figure 3A shows the analysis of neutral N-glycan structural classes on factor planes 1 and 2. Figure 3B shows PCA results of combined neutral and acidic N-glycan structural classes, which separate the individual samples slightly differently compared with Fig. 3A. The analysis indicates that the four tumor types can be separated based on MS analysis of their N-glycan profile, because both PHEO and PGL (Fig. 3A) and metastatic and nonmetastatic tumors (Fig. 3B) were located in distinct areas of the diagrams (apart from two samples, patient nos. 5 and 13 in Table 1; please see the Discussion). In Fig. 3A, factor 1 appears to segregate metastasized and nonmetastasized samples. Based on factor loadings, factor 1 relates to higher relative amounts of the following glycan classes in samples with metastasis: complex-type N-glycans, hybrid-type N-glycans, fucosylation, complex fucosylation, and terminal N-acetylhexosamine. Factor 2 of neutral glycan classes may relate to complex-type N-glycans with either terminal N-acetylhexosamine (metastasized) or terminal hexose (nonmetastasized). Therefore, based on PCA, the four sample types had consistently distinct N-glycan profiles except for two patients (no. 5 and 13; please see the Discussion).

Figure 3.

PCA visualizes the differences in glycosylation of metastatic and nonmetastatic PHEO and PGL. Visualization of neutral (A) and combined neutral and acidic (B) N-glycan structural classification data as PCA results on factor planes 1 and 2. Light red, metastasized PHEO; light green, nonmetastasized PHEO; dark red, metastasized PGL; dark green, nonmetastasized PGL. Numbers refer to the patient numbering in Table 1. The percentages indicate the amount of variance captured by each of the factors (1 or 2). Dotted lines illustrate the clustering of PHEO and PGL samples in neutral glycan PCA (A) and metastasized and nonmetastasized samples in acidic and neutral glycan PCA (B). Two PGL tumors (5 and 13, marked with striped background) clustered with opposite sample group; please see Discussion in the main text.

Glycosylation differences between tumors

Glycosylation differences between metastasized and nonmetastasized PHEOs and PGLs were statistically significant with regard to the abundances of neutral and acidic N-glycan signals (Fig. 4). Regarding neutral N-glycans, metastasized tumors had significantly more abundant hybrid type N-glycans, fucosylation, and complex fucosylation (Fig. 4A–4C). Additionally, among acidic N-glycans, both fucosylation and complex fucosylation were significantly increased in metastasized tumors (Fig. 4D and 4E). However, the changes were over the false positivity critical rate, when adjusted for multiplicity testing.

Figure 4.

Differences of neutral (A–C) and acidic (D–E) N-glycan structural features between nonmetastasized and metastasized tumors. Distributions and means of structures shown in box and whisker plots, for neutral glycans A (hybrid type), B (fucosylation), C (complex fucosylation) and acidic glycans D (fucosylation), and E (complex fucosylation). Statistical analysis by the Mann–Whitney test reveal that individual P values are <0.05 but are over the false positivity critical rate, when adjusted for multiplicity testing by the Benjamini–Hochberg procedure. Additional statistics in Supplemental Table 4.

SDH status of the tumors

The results of SDHB immunohistochemistry are shown in Table 1. All tumors were SDHA positive.

Discussion

Using MALDI-TOF MS profiling of N-glycans in 16 primary PHEOs and PGLs, we found differences in acidic and neutral N-glycan profiles between metastasized and nonmetastasized tumors, and PCA of the glycan profiling data could separate these tumors. Regarding neutral N-glycans, the five most abundant glycan signals could be identified as high-mannose type N-glycans. The metastasized tumors had more complex-type N-glycan signals indicative of cancer-associated terminal N-acetylglucosamine (29). Additionally, multifucosylated glycans (indicating complex fucosylation), hybrid-type N-glycans, and fucosylated pauci-mannose type N-glycans were more abundant in metastasized tumors. Acidic N-glycans with no or only one fucose residue, such as S1H4N3 and S1H5N4, were more abundant in nonmetastasized tumors. Metastasized tumors were characterized by multifucosylated N-glycans, acid ester–modified (sulfated or phosphorylated) glycans, and hybrid-type/monoantennary N-glycans. The more abundant fucosylation and complex fucosylation in metastasized PGLs and PHEOs compared with nonmetastasized tumors were statistically significant for individual tests but were above the false positive critical rate when adjusted for multiplicity testing.

H3N4F1, a complex-type N-glycan signal, which is associated with metastasized tumors and which is indicative of cancer-associated terminal N-acetylglucosamine, has also been shown to accumulate in lung, kidney, breast, and ovarian cancers (29) and in colon cancer (32). In rectal tumors, the fucosylated pauci-mannose type N-glycans have been shown to be more common in higher disease stages of rectal adenocarcinoma, whereas nonfucosylated structures are more common in stage I disease (32). Monoantennary N-glycans have also been shown to be more abundant in rectal carcinomas than in benign adenomas (32). An increase in monoantennary N-glycans in serum samples of patients with lung cancer has been shown in comparison with controls (28). Metastasized PGLs and PHEOs had more abundant fucosylation and complex fucosylation when compared with nonmetastasized tumors. In contrast, in rectal tumors, complex fucosylation was more common in benign adenomas than in cancer. The smallest high-mannose type glycan H5N2 was increased in rectal carcinomas in comparison with adenomas (32), but we found no significant differences in high-mannose type N-glycans between metastatic and nonmetastatic PHEOs and PGLs. We found greater amounts of sulfated or phosphorylated N-glycans from primary tumors with metastatic disease. In previous studies on sulfated N-glycans, the results have not been consistent. In colorectal carcinoma, greater amounts of sulfated N-glycans have been reported in comparison with normal colon epithelium (31). In contrast, sulfated N-glycans have been shown to be more abundant in benign rectal adenomas than in carcinomas (32).

Using the Cancer Genome Anatomy Project database (www.cgap.nci.nih.gov), we found no reports regarding the relationship between genes involved in the pathogenesis of PHEO and PGL (15) and glycosylation. Additionally, we did not find any reports linking the glycosylation genes related to malignant transformation based on the present results and genes involved in the pathogenesis of PHEO and PGL (Supplemental Table 3). Glycosylation changes have been identified to correlate with occurrence of cancer, and expression of specific glycan structures by cancer cells has been suggested to be involved in metastasis and evasion of immune surveillance (35). However, distinct glycosylation genes have not been widely recognized as contributing to malignancy. This result may reflect the complex nature of glycosylation, which is a posttranslational modification not directly derived from the genomic sequence. For example, protein glycosylation requires concerted action of hundreds of enzymes and transporters, including glycosyltransferases, glycosidases, enzymes of sugar metabolism, and nucleotide sugar transporters, encoded by hundreds of genes with partly overlapping functions, which is necessary to produce the full diversity of glycan structures in the body (36). In addition, glycosylation is influenced by extragenomic factors such as metabolic and oxygenation status of the tissue, which are important environmental factors inside the tumor. It may thus require analysis of a larger group of glycosylation-related genes to pinpoint possible genes important for formation of the present tumor type–associated glycan profiles in PHEOs and PGLs. In recent years knowledge of the genetic background of these tumors has increased, and today genetically different tumors are thought to have a different pathogenesis (15). The genetic background could have some impact on the glycosylation process also. We do not know the genetic background of all the analyzed tumors. Some tumors were operated on many years ago, when the genetic testing was not common. However, we have information on SDHB immunohistochemistry in these tumors. Immunohistochemically SDHB-negative tumors with deficient SDH complex associate with PGL syndromes 1 to 5, and one-third of patients with germline mutations in the SDHB gene have a metastatic PGL or PHEO (37).

In the PCA, nonmetastasized tumors differed from metastasized tumors, except for one tumor with no known metastasis but a malignant profile (patient no. 13 in Table 1) (Fig. 3). This patient has an SDHB mutation. Tumors from patients with SDHB mutations metastasize more often (16), and metastasis can show up late, after many years. This patient has been followed for 11 years, with no evidence of metastasis, but the emergence of metastasis has not been excluded. It is also possible that this tumor had metastatic potential but was removed before any metastasis evolved.

One metastasized tumor (patient no. 5 in Table 1) had a different PCA profile compared with other metastasized tumors. This analyzed tumor was a retroperitoneal PGL from a patient with a SDHB mutation. Neck PGL on the left side was also removed at the same time. Later, a neck PGL on the right side was removed, and large tumor in the base of the skull was diagnosed with some neuroblastoma-like differentiation. This skull tumor was positive upon fluorodeoxyglucose positron emission tomography and metaiodobenzylguanidine scintigraphy. It is possible for patients with an SDHB mutation to develop multiple primary tumors (37). The retroperitoneal tumor, which we analyzed by MALDI-TOF MS profiling, was probably a nonmetastasized primary tumor. Because of the discrepancy, we decided to analyze the glycomic profiles of all four of the patient’s tumors, and they all had N-glycan profiles similar to those of the nonmalignant samples (Supplemental Table 2). From the tumor in the base of the skull, only scant biopsy material was available (two separate samples). The retroperitoneal tumor had a nonmalignant profile in the second analysis, similar to the patient’s other PGLs. We cannot tell with certainty which of the primary tumors had metastasized or whether the tumor in the base of the skull was a metastasis from PGL. This case reflects the diagnostic difficulties of patients with SDHB mutations. These patients can have not only multiple primary tumors but also metastatic disease.

The diagnostic challenge with PHEO and PGL is that we still lack effective methods to separate tumors that will metastasize from those that will stay local. The differences in N-glycan profiles may provide us with a tool to identify aggressive tumors. An advantage of this method is that numerous biomarkers can be analyzed in a single experiment. For example, numerous low-abundance glycan signals in the neutral N-glycan profile carried multiple fucose (deoxyhexose) residues; in other words, their glycan composition was characterized by the F > 1 feature, including H4N3F2, H5N3F2, H4N4F2, H5N4F2, H4N5F2, H5N4F3, H4N5F3, and H5N5F3, all of which had higher abundance in malignant tumors than in nonmalignant tumors (Fig. 2A). Because the cancer-associated glycan feature occurred in many signals instead of only one, in our approach we calculated together all the glycan signals with the feature and subjected the sum to statistical evaluation (Fig. 4). The MS method is able to detect all these individual signals within the complete glycan profile in one analysis, making analysis of low peaks feasible with significant statistical power. The limitation in this study is that the number of tumors analyzed is small (16 tumors) because of the rarity of these tumors and especially of metastatic tumors. Because of the limited number of tumors, one cannot totally exclude the possibility of chance on the results. MALDI-TOF MS profiling is also a complex, tissue- and time-consuming method that requires special expertise. More research with larger tumor numbers and by different laboratories is needed to validate these results and determine the usefulness of N-glycomic profiling of these tumors. Both MS analysis and data analysis can be automated. It is possible that this method could help us to more accurately determine the prognosis of patients with PHEO or PGL in the future.

By identifying malignancy-associated N-glycans, it might be possible to develop serum biomarker tests that are useful in the diagnosis and follow-up of these tumors or to develop antibodies against these structures for immunohistochemistry, which could further assist in diagnosis. Specific antibodies to known N-glycan structures associated with malignancy may be used as drug carriers for targeted therapy in the future.

To the best of our knowledge, this report is the first to analyze the N-glycomic profiles of PHEOs and PGLs. Metastasized and nonmetastasized primary tumors showed differences in their N-glycomic profiles. The differences were partly similar to differences seen previously in other malignancies, but different changes in the glycomic profiles were found. The molecular structures of malignancy-associated glycan signals must be analyzed in more detail in further studies. More research and larger patient series are needed to evaluate the usefulness of this information in clinical practice.

Abbreviations

MALDI-TOF

matrix-assisted laser desorption-ionization time-of-flight

MS

mass spectrometry

m/z

mass/charge

PCA

principal component analysis

PGL

paraganglioma

PHEO

pheochromocytoma

SDH

succinate dehydrogenase.