Differentiating N-linked glycan structural isomers in metastatic and nonmetastatic tumor cells using sequential mass spectrometry

Abstract

In an effort to understand the role of molecular glycosylation in cancer a murine model has been used to characterize and fingerprint malignancies in established cell lines that manifest all the hallmarks of metastatic disease: spontaneous development, local invasion, intravasation, immune system survival, extravasation, and secondary tumor formation involving liver, kidney, spleen, lung, and brain. Using astrocyte cell controls, we compared N-linked glycosylation from a nonmetastatic brain tumor cell line and two different metastatic brain tumor cells. Selected ions in each profile were disassembled by ion trap mass spectrometry (MSⁿ) which exhibited multiple structural differences between each tissue. These unique structures were identified within isomeric compositions as pendant nonreducing termini of di- and trisaccharide fragments, probably transparent to a tandem MS approach but distinctively not to sequential ion trap MSⁿ detection.

Introduction

Molecular glycosylation is involved in key developmental roles including cell differentiation, innate immunity, and signal transduction. Equally as demonstrable are numerous aspects of malignancy, from cellular proliferation to angiogenesis and metastasis (Ono and Hakomori 2004). Aberrantglycosyltransferase expression and their Golgi localization correlate with tumor transformation and progression (Brockhausen 2006). These modulations result in glycan structures that are often associated with embryonic and fetal development and downregulated in adult tissues. Such glycan changes alter the function of the cell in numerous ways, often by influencing its adhesive and antigenic properties. These features are considered to endow cancer cells with the propensity to invade and metastasize.

While many primary tumors can be treated with conventional therapies, few treatments are effective against metastatic disease (Chambers et al. 2002; Steeg 2006). A new mouse model of metastatic cancer that shares a number of similarities with human glioblastoma multiforme (GBM), the most aggressive human brain tumor, has recently been developed (Huysentruyt et al. forthcoming). These metastatic VM-M2 and VM-M3 tumors arose spontaneously in the brain of the VM mouse strain and manifested all of the hallmarks of invasive/metastatic disease to include local invasion, intravasation, immune system survival, extravasation, and secondary tumor formation involving liver, kidney, spleen, lung, and brain (Huysentruyt et al. forthcoming). While extracranial metastasis for GBM is rare in humans due to early patient death, GBM is highly metastatic if the tumor cells gain access to tissues outside the brain (Rubinstein 1972; Hoffman and Duffner 1985; Taha et al. 2005). The invasive/metastatic behavior of the VM-M2 and VM-M3 tumors, when grown in extraneural sites, is therefore similar to that seen for human GBM when given access to extraneural sites. Another spontaneous VM brain tumor (VM-NM1) with characteristics of neural stem cells was also discovered. This tumor appears to be nonmetastatic intracranially when given access to extraneural sites (Huysentruyt et al. forthcoming). Tumor cell lines have subsequently been established (El-Abbadi et al. 2001; Huysentruyt et al. forthcoming).

Released N-linked glycans were mass profiled and a series of neutral, fucosylated glycans that exhibited differential ion abundances were evaluated. Unfortunately, molecular compositions alone cannot define a glycan structure, nor distinguish structural isomers. Although having the same composition, structural isomers are the products of different enzymatic pathways and each may perform different biological functions. Cancer-related aberrancy in glycan processing may lead to cancer-specific structural isomers. There are numerous ancillary techniques for resolving the inherent complexities of a glycan structure; however most protocols (chromatography, linkage analysis, enzymology, MS/MS) supplement rather than finalize an exacting structure. Analytical transparency to any one feature of glycan's structure exposes the possibility of a missed opportunity to shed light on their role in tumor progression, to uncover a unique cancer biomarker, or to discover novel targets associated for cancer therapeutics. Within the goal of cancer detection, it remains fundamental that glycan structural details must be comprehensively determined.

This problem has challenged researchers for decades, and even today, a congruent strategy for dissecting all facets of structure remains elusive. The most comprehensive strategy at this time remains the ability to isolate ions and through repetitive rounds of disassembly follow all products and pathways until structural comprehension is achieved. This can be observed in the spectra of methylated mono-, di-, or trisaccharides in an ion trap instrument that is within the realm of sensitivity exhibited by most mass spectrometers. This instrument has a history of unraveling the structural details of molecular glycosylation. Recently, a three-paper series was published, summarizing sequential mass spectrometric (MSⁿ) strategies (Ashline et al. 2005), an MSⁿ fragment library (Zhang et al. 2005), and GlySpy, a tool suite for glycan analysis, including the glycan topology assignment algorithm, OSCAR (Lapadula et al. 2005). We continue to apply these protocols and tools to better appreciate their applications, build our mass spectral library, and discern possibilities for automation (Berenson et al. 2005; Hanneman et al. 2006; Sarkar et al. 2006; Ashline et al. 2007; Lau et al. 2007). The ion traps are effective for defining structures de novo and thus allow structural isomer distinction. These strengths are now focused on the challenges of cancer glycosylation. Using a murine astrocyte cell line as a control, the VM-NM1 cells as a nonmetastatic model, and the VM-M2 and VM-M3 as a metastatic model, we extended ion trap MSⁿ applications to investigate modulations of N-linked glycan structure in nonmetastatic and metastatic cancer.

Results

An examination of the glycan profile spectra taken from the four murine cell lines reveals considerable similarities (Figure 1). However, some ion compositions were noticeably more intense and selected for detailed structural characterization. Comparative analysis was performed on 11 permethylated-reduced neutral mono-fucosylated compositions (H₅N₃F₁, H₆N₃F₁, H₇N₃F₁, H₅N₄F₁, H₆N₄F₁, H₇N₄F₁, H₈N₄F₁, H₇N₅F₁, H₈N₅F₁, H₉N₅F₁, and H₆N₆F₁). Multiple rounds of MSⁿ fragmentation and structural characterization were required to reveal differences between the cell lines. Unsurprisingly, all samples demonstrated multiple structural isomers for each molecular composition. However, the presence of unique structures as well as the relative abundance of common isomeric structures was specific to each cell line.

Fig. 1.

Comparative ESI-MS profile analysis of methylated-reduced N-linked glycans obtained from an astrocyte cell line control, nonmetastatic cell line (VM-NM1), and two metastatic cell lines (VM-M2 and VM-M3).

As an example, the ion m/z 1121.1²⁺, corresponding to composition H6N3F1, was isolated from each of the profiles and examined by CID-MS (Figure 2). The product fragment masses suggest the presence of multiple structural isomers; however, the abundance of these ions are clearly cell-line specific. Further examination of these product ions indicated three matched B-ion/Y-ion complements: (1) m/z 486.3/m/z 1755.8; (2) m/z 660.5/m/z 1581.7; and (3) m/z 690.5/m/z 1551.7. The VM-NM1 sample exhibited a unique pair of B-ion and Y-ion complements, m/z 864.5 and m/z 1377.7, a surprising and unique structural isomer.

Fig. 2.

Comparative analysis of glycan ion composition H₆N₃F₁ (m/z 1121.1²⁺) by ESI-MS².

As mentioned, the abundance of many B-ions differed markedly in the samples. In the astrocyte control, the B₃-ion at m/z 660 was only 2% of the base ion while the B₂-ion, m/z 486, was 16% of the base ion, and the B₃-ion, m/z 690, is moderately abundant at 24% of the base ion (Figure 2). Interestingly, the ion at m/z 690 is the base ion for the VM-M2 sample, while m/z 486 and m/z 660 represent 19% and 4%, respectively. The VM-M3 sample exhibits an ion intensity pattern similar to VM-M2, with m/z 690 being the predominant B-ion fragment. Interestingly, the nonmetastatic VM-NM1 displays a distinct B-ion spectrum. The ion at m/z 660 is the predominant B-ion, 68% of the base ion, while m/z 690 and m/z 486 represent 18% and 8%, respectively. The fragment ion, m/z 864.5, is only present in the VM-NM1 sample and has abundance equal to 27% of the base ion. Although the B- and Y-ion pairs suggest the existence of multiple structural isomers, the topology and linkage of these Bion and Y-ion fragments cannot be assured at the MS² level. For example, consider the ion m/z 690, a B₃-ion consisting of two hexoses, and a single N-acetylhexosamine. This ion could have a number of topologies (Hex-Hex-HexNAc-, Hex-HexNAc-Hex, HexNAc-Hex-Hex-, etc.) as well as different interresidue linkages. Assuming sufficient ion current, disassembly must be pursued until all aspects of the structure are defined. This important point is demonstrated by a close analysis of the ion m/z 690 in these VM cell lines.

The MS³ spectra from the astrocyte control and VM-M2 samples provide the B-ion (m/z 445), B/Y-ions (m/z 268 and m/z 472), C-ions (m/z 259 and m/z 463), and B/Y-ion (m/z 227), all consistent with a Hex-Hex-HexNAc topology (Figure 3). The nonmetastatic VM-NM1 and metastatic VM-M3 samples provided spectra identical to VM-M2 (data not shown). Even at the trisaccharide MS³ level, topology and linkage of these glycan structures cannot be fully characterized. The fragmention, m/z 315, can be assigned in at least three ways: (1) a ^3,5 Aion from an internal 4-linked Hex-HexNAc B/Y-ion, or (2) a ^3,5A-ion from a terminal 3-linked Hex-Hex B₂-ion; or (3) a mixture of product ions from both precursors (Figure 4). Additionally, the fragment ion, m/z 472, may represent two isomeric topologies (HexNAc-Hex and Hex-HexNAc), or more if linkage differences are considered. Thus, it must be emphasized that disassembly must proceed until the glycan's topology is resolved, in this case to MS⁴.

Fig. 3.

Comparative disassembly of the product-ion fragment m/z 690 originating from profile ion m/z 1121.1²⁺ and H₆N₃F₁, by ESI-MSⁿ.

Fig. 4.

B₃-ion, m/z 690, product-ion assignments.

The fragment ion, m/z 472, of the MS³ spectra m/z 690 was selected for MS⁴ analysis (Figure 3). The MS⁴ spectrum of m/z 472 from both samples present B/Y-ion (m/z 227 and m/z 268) and C/Y-ion (m/z 245) fragments confirming an internal Hex-HexNAc topology. Fragment ions at m/z 301, m/z 315, and m/z 398 represent ^2,4A-, ^3,5A-, and ^0,4A-ions, respectively (Figure 4). This A-ion fragmentation pattern is consistent with a 4-linked Hex-HexNAc internal B/Y-ion (Ashline et al. 2005). These fragment assignments are consistent with previous reports demonstrating that normalmurine brain neutral N-linked glycan β-lactosamine sequences are almost exclusively 4-linked (Chen et al. 1998). The fragment ion, m/z 445, from the MS³ m/z 690 spectrum is consistent with a Hex-Hex B₂-ion (Figure 4). The MS⁴ spectra of m/z 445 were obtained from the astrocyte control and VM-M2 samples. Both samples present B-ion (m/z 241), Cion (m/z 259), and B/Y-ion (m/z 227) fragments consistent with a Hex-Hex topology as well as the fragment ion atm/z 329 (^3,5Aion), which is consistent with the B₂-ion from the trisaccharide standard globotriositol, Galα1,4Galβ1,4GlcNAcol (Ashline et al. 2005). The VM-M2 sample also has relatively intense fragments at m/z 315 (^2,4A-ion), m/z 343 (^3,5A-ion), and m/z 371 (^0,4A-ion) (Figure 3). This A-ion fragmentation pattern parallels the fragmentation pattern of the standard linear B2 trisaccharide Galα1,3Galβ1,4GlcNAc (Ashline et al. 2005). These fragment patterns suggest that two structural isomers differing in a terminal monomer linkage are present in the metastatic and nonmetastatic cancer cell lines, while only the 4-linked B₂-ion is present in the astrocyte control (Figure 7, structural isomers 4 and 5). It must be emphasized that 1,3-hexosyl capping of lactosamine sequences is not a cancer-specific epitope. It is the entire glycan structure (1,3-hexosyl-extended 1,4-lactosamine hybrid glycan) that is specific to the nonmetastatic and metastatic cell lines (Figure 7, structural isomer 5).Only after sequential rounds of product ion isolation and fragmentation down to disaccharide units, in this case MS⁴, could the differences in these isomeric structures be resolved without ambiguity.

Fig. 7.

Summary of the structural isomers detected with a composition H₆N₃F₁: control, VM-NM1, VM-M2, and VM-M3 cell lines.

Lewis epitopes, particularly Lewis^x (Le^x) structures, are upregulated in various types of cancer (Itzkowitz et al. 1986; Kim et al. 1988). The increased abundance of m/z 660 in the VM-NM1 sample parallels these findings. The ion m/z 660 is consistent with a B₃-ion composed of hexose, fucose, and an N-acetylglucosamine. To establish the topology of this ion it was selected for CID analysis (Figure 5). The MS³ spectra from both the astrocyte and astrocytoma samples showed fragment ions consistent with two topologies, Hex(Fuc-)-HexNAc- and Fuc-HexNAc-Hex. Both metastatic samples generate comparable spectra (data not shown). Fragment ions at m/z 241 (B-ion), m/z 259 (C-ion), m/z 442 (Y-ion), m/z 424 (Z-ion), and m/z 329 (^3,5 A-ion) are all consistent with a Le^X epitope (Figure 5). Interestingly, a second fucosylated isomer was also detected. Fragment ions at m/z 456 (B₂-ion), m/z 474 (C-ion), m/z 227 (Y-ion), and m/z 544 (^3,5A-ion) indicate an isomer with a Fuc-HexNAc-Hex topology. Both isomers have identical fragment ions at m/z 211 (B-ion), m/z 472 (Y-ion), and m/z 229 (C-ion). Due to limitations in ion current, obtaining a MS⁴ spectrum of m/z456 from the MS³ precursor atm/z 660 as well as performing linkage analysis on this isomeric determinant was not possible. As with m/z 690, the detection and confirmation of the isomeric structures of m/z 660 required at least three sequential rounds of disassembly (MS³). These findings demonstrate two isomeric B₃-ion fragments present at m/z 660: (1) a Le^x epitope and (2) an unanticipated epitope containing a terminal fucose attached to a penultimate HexNAc (Figures 5 and 7; structural isomers 3 and 4).

Fig. 5.

Disassembly of m/z 1121.1²⁺ ion exposes the structural details of a terminal nonreducing trisaccharide, m/z 660, and variations in linkage between astrocyte and VM-NM1 cells.

The Lewis^x epitope is highly regulated during the development of the central nervous system (Roberts et al. 1991; Streit et al. 1993, 1996), and they are associated with increased celladhesion activities in cancer tissues (Itzkowitz et al. 1986; Kim et al. 1988; Mannori et al. 1995). Interestingly, primary carcinomas also exhibit enhanced levels of the Le^x epitope, while in metastatic lesions the Le^x content diminishes (Nakagoe et al. 1994). As Figure 2 illustrates, them/z 660 ionwas elevated in the nonmetastatic VM-NM1. The increased abundance of Le^x may contribute to VM-NM1's aggressive nonmetastatic adhesive behavior. Conversely, m/z 660 is suppressed in the metastatic VMM2 and VM-M3 cell lines, and these diminished levels may contribute to the metastatic behavior of VM-M2 and VM-M3.

VM-NM1 expresses an anomalous fragmentation ion at m/z 864.5 as well as its complement at m/z 1377.7 (Figure 2). The MS³ spectra of m/z 864 reveal fragment ions at m/z 241 (B-ion), m/z 259 (C-ion), m/z 445 (B₂-ion), m/z 463 (C₂-ion), m/z 442 (B/Y/Y-ion), and m/z 533 (^3,5A-ion) that are consistent with a B₄-ion representing a hexose-extended Lewis^x epitope (Figure 6). The fragment ion at m/z 445 (B₂-ion) from m/z 864 was isolated for MS⁴ CID analysis which showed ions m/z 241 (B-ion) and m/z 259 (C-ion) consistent with a terminal hexose. Fragments at m/z 315 (^2,4A-ion), m/z 343 (^3,5A-ion), and m/z 371 (^0,4A-ion) are typical of an α1,3-linked Hex-Hex B₂-ion (Ashline et al. 2005). These spectra suggest that VM-NM1 expresses a unique structural isomer bearing a nonfucosylated reducing end GlcNAc and a 1,3-hexose-extended Le^x epitope (Figure 7, structural isomer 6). Murine teratocarcinoma F9 cells have been reported to express this determinant after retinoic acid-induced cell differentiation (Cho et al. 1996). It is unknown whether this isomeric structure reflects cell line pathology or the differentiation status of the cells. Regardless, detailed glycan structural analysis using MSⁿ provides the glycoanalytic tool to begin comprehensively investigating these questions. Although m/z 864 is detectable in the MS² spectrum of VM-NM1, structural identification and linkage assignment could only be elucidated after additional rounds of MSⁿ. To date, all 11 individual neutral mono-fucosylated glycan molecular compositions subjected to ion trap MSⁿ disassembly revealed at least two structural isomers that are only detectable in the cancer samples. These isomers were only detected and confirmed at the MS^3–5 disassembly levels.

Fig. 6.

Disassembly of the unique fragment ion, m/z 864, exposes a nonreducing tetrasaccharide in the VM-NM1 cell line.

Distinguishing the isomeric structure requires an understanding of product–precursor relationships. Table I lists the MSⁿ product–precursor fragmentation pathways followed to elucidate the structural isomers composed of H6N3F1. Interestingly, certain product–precursor fragmentation pathways were unique to the cancer cell lines. To validate structural assignments made by the analyst, the structures and pathways shown in Table I were given as inputs to GlySpy's core algorithm, OSCAR, which then computed the compatibility of each pathway/structure pair. (A pathway and structure are compatible if some sequential disassembly of the structure could yield the observed pathway.) Within 3 s, GlySpy confirmed (and in several instances, corrected) the set of pathways compatible with structural isomers 1 through 6. Manually, these assignments took 9 months. We envision that computational interpretation of mass spectrometric data will eventually replace, rather than merely augment, human analysis. Bioinformatics development toward that goal is ongoing in many labs, including ours.

Table I.

MSⁿ pathways and glycan topologies from the ion m/z 1121.1²⁺, Hex₆HexNAc₃Fuc₁.

Only after the identification and detailed characterization of each structural isomer present under each analyzed mass profile (molecular composition) may an assessment of cancer related changes in glycan abundance be undertaken through mass spectrometric methods. As the various MS² spectra of H₆N₃F₁ demonstrate, the metastatic, nonmetastatic, and astrocyte control cell lines each express a unique cell-type-specific distribution of biologically relevant glycan structures (Figure 2). The MS² spectra of H₇N₄F₁ reveal a B-ion intensity pattern similar to H₆N₃F₁ (Figure 8). Characteristic of VM-M2 B-ion fragmentation, m/z 690 is amajor ion, at 56% of the base ion and elevated relative to the astrocyte control, 14% of the base ion in the MS² spectra of H₆N₄F₁ (m/z 1243.6²⁺). As observedwith the nonmetastatic VM-NM1 sample (fucosylated B₃- and B₄-ions), the m/z 660 and m/z 864 ions were major fragments compared to the metastatic VM-M2 and VM-M3. These fucosylated B-ion fragments are absent in the astrocyte. The subsequent MSⁿ patterns of all MS² B-ions were in agreement with B-ion fragment assignments of H₆N₃F₁ (data not shown). As expected, the fragment ion, m/z 660, in MS² spectra of H₇N₄F₁ (m/z 1345.7²⁺) was elevated in VM-NM1 when compared to the astrocyte control. Interestingly, the hexose-extended fragment ions at m/z 690 and m/z 864 are only present and abundant in the cancer samples. The presence of m/z 864 in the metastatic cell lines reinforces the importance that complete glycan structures, not terminal epitopes or substructural motifs, will be the likely cancer-type-specific candidates. More importantly, the absence of m/z 864 in the normal astrocyte suggests that cancer may produce glycan structures that are normally undetectable in healthy controls.

Fig. 8.

Comparative disassembly of two profiled glycan ions m/z 1243.6²⁺ and m/z 1345.7²⁺, with respective compositions Hex₆HexNAc₄Fuc₁ and Hex₇HexNAc₄Fuc₁.

Table II illustrates the relative abundance of ions m/z 486, m/z 660, m/z 690, and m/z 864 from three separate cell culture experiments and 11 different compositions (H₅N₃F₁, H₆N₃F₁, H₇N₃F₁, H₅N₄F₁, H₆N₄F₁, H₇N₄F₁, H₈N₄F₁, H₇N₅F₁, H₈N₅F₁, H₉N₅F₁, and H₆N₆F₁) in each sample. The mean normalized percent was calculated by dividing the absolute intensity of the individual terminal epitope by the intensity sum of all four terminal epitopes and multiplying by 100. The fragment ion, m/z 486, represents 31.4% of the neutral terminal epitopes in the astrocyte, which decreases to 12.2%, 13.6%, and 14.3% in VMNM1, VM-M2, and VM-M3, respectively. β-Lactosamine is the substrate for terminal fucosyltransferases, sialyltransferases, and galactosyltransferases. The fragment ion representing the Le^x epitope, m/z 660, is elevated in the astrocytoma, at 23.2% of the terminal epitope pool, compared to the astrocytes 15.9%. Interestingly, both metastatic samples (VM-M2 and VM-M3) showed suppressed levels of m/z 660, 5.8% and 9.1% of the terminal epitope pool, respectively. Meanwhile, m/z 690, representing hexose-extended β-lactosamine epitopes, is elevated in VM-M2 and VM-M3, at 71.1% and 63.3% of the total terminal epitope pool, respectively. This epitope was suppressed in VM-NM1, 36.9%, compared to the astrocyte control, 52.4%. The expression of the terminal epitope ion, m/z 864, was greatly elevated in VM-NM1, 27.7% compared to VM-M2 and VM-M3, 9.6% and 13.7%, respectively. Meanwhile, m/z 864 was not detectable in any of 11 compositions analyzed for the astrocyte control. Figure 9 is a graphical representation of the empirical data listed in Table II. These pie charts illustrate the differential expression of B-ion fragments, hence the differential expression of structural isomers, between the nonmetastatic and metastatic tumors, and astrocyte cell lines. Terminal epitopes bearing outer arm fucosylation, Lewis^x, and 1,3-hexose-extended Le^x structural isomers are upregulated in the astrocytoma, meanwhile Le^x epitopes are suppressed and hexose-extended 1,4-lactosamine and 1,3-hexose-extended Le^x structural isomers are upregulated in the metastatic cancers. This collection of MSⁿ spectral data indicates that the nonmetastatic tumors and metastatic tumors express glycan structures in different abundance. These data indicated the presence of distinct oligosaccharide structures in the cancer cells, as well as intriguing differences in oligosaccharide isomer distributions between the metastatic and the nonmetastatic cancer cell lines. Therefore, B-ion fragment patterning may prognosticate tumor's likelihood to metastasize.

Table II.

Relative amounts of the four indicated terminal epitopes. Dates were obtained from three separate cell-culture experiments and 11 different compositions (H₅N₃F₁, H₆N₃F₁, H₇N₃F₁, H₅N₄F₁, H₆N₄F₁, H₇N₄F₁, H₈N₄F₁, H₇N₅F₁, H₈N₅F₁, H₉N₅F₁, and H₆N₆F₁) in each sample (n = 33). The mean normalized percent was calculated by dividing the absolute intensity of the individual terminal epitope by the intensity sum of all four terminal epitopes and multiplying by 100

		Mean normalized percent	Standard deviation	Lower 95% confidence	Higher 95% confidence
Sample		(n = 33)	(n = 33)	interval	interval
m/z 486
	Astrocyte	31.4	5.38	29.5	33.3
	VM-NM1	12.2	9.96	8.7	15.7
	VM-M2	13.6	2.45	12.7	14.5
	VM-M3	14.3	3.89	12.9	15.7
m/z 660
	Astrocyte	15.9	2.74	14.9	16.9
	VM-NM1	23.2	5.68	21.2	25.2
	VM-M2	5.8	0.86	5.5	6.1
	VM-M3	9.1	0.78	8.8	9.4
m/z 690
	Astrocyte	52.4	11.21	48.4	56.4
	VM-NM1	36.9	7.48	34.2	39.6
	VM-M2	71.1	16.16	65.4	76.8
	VM-M3	63	14.74	57.8	68.2
m/z 864
	Astrocyte	Nd	nd	nd	nd
	VM-NM1	27.7	4.38	26.1	29.3
	VM-M2	9.6	1.03	9.2	10.0
	VM-M3	13.7	3.93	12.3	15.1

nd = not detected.

Fig. 9.

Pie-chart representation of the relative amounts of ions m/z 486, m/z 660, m/z 690, and m/z 864 from three separate cell-culture experiments and 11 different compositions (H₅N₃F₁, H₆N₃F₁, H₇N₃F₁, H₅N₄F₁, H₆N₄F₁, H₇N₄F₁, H₈N₄F₁, H₇N₅F₁, H₈N₅F₁, H₉N₅F₁, and H₆N₆F₁) in each sample (n = 33). The mean normalized percent was calculated by dividing the absolute intensity of the individual terminal epitope by the intensity sum of all four terminal epitopes and multiplying by 100.

Discussion

The basic biology of cancer has been studied using a murine model that possesses all the untoward aspects of malignancy (spontaneous tumor generation, local invasion, intravasation, immune system survival, extravasation, secondary tumor formation and metastasis). Since most of these developmental stages can be observed phenotypically, it would appear that some biochemical measure should confirm these observations. Changes in the cell's physical environmental, kinetics of synthesis, and precursor concentrations are all known to alter glycosylation. In that regard, it can effectively be argued that molecular glycosylation and its glycomer and isomer distributions should provide a valuable fingerprint of cellular biology.

Importantly, to elucidate the far-reaching biological consequences of glycosylation a full characterization of the structure is paramount. As a result, data acquired from mass profiles (molecular composition) and tandem MS analysis are frequently insufficient to assign a comprehensively glycan structure. The omission of structural features is a potential missed opportunity to clarify glycan structures involved in tumor progression, to uncover unique cancer biomarkers, and to discover novel targets for cancer therapeutics. As demonstrated by this study, ion trap MSⁿ provides a method for detailed structural characterization of glycosylation and presents the opportunity to discover cancer-specific structural isomers. We are currently working toward the automatic detection of structural isomers. GlySpy will ideally allow for the automated detection of cancer-specific glycan structures and expedite the process of glycan biomarker discovery.

Materials and methods

Propagation of VM mouse tumor cells

VM mouse tumor cell lines were established from spontaneous brain tumors in the cerebrum of three adult mice, two males and a female. Tumors were identified as poorly defined masses (about 3 × 1 × 1 mm) with yellowish soft regions. To preserve in vivo viability, each tumor was resected and implanted intracranially into host VM mice as previously described (El-Abbadi et al. 2001; Huysentruyt et al. forthcoming). As cranial domes appeared in host mice, tumors were resected and implanted intracranially to another set of host VM mice. After a total of three intracranial passages, the tumors were grown subcutaneously and cell lines were prepared from each tumor. An astrocyte C8-D30 (Astrocyte type III clones) cell line was purchased from American Type Culture Collection (Manassas, VA) and used as the astrocyte control.

All cell lines were grown in Dulbecco's Modified Eagle medium (DMEM, Sigma, St. Louis, MO) with high glucose (25 mM) supplemented with 10% fetal bovine serum (Sigma) and 50 mg/mL penicillin–streptomycin (Sigma). The cells were cultured in a CO₂ incubator with a humidified atmosphere containing 95% air and 5% CO₂ at 37°C. The cells were grown in 150 × 10 mm culture dishes until confluent and cells were then removed by a cell scraper. The cells were pelleted by centrifugation, re-suspended in phosphate-buffered saline, and frozen at −80°C.

Isolation and purification of proteins

Cells were lysed from frozen pellets with 1 mL of membrane disruption buffer containing 4% CHAPS, 8 M urea, 40 mM Tris–HCL, pH = 8, 65 mM dithiothreitol (DTT) and mammalian protease inhibitor cocktail (Sigma). After five cycles of sonication of 6 min at 8°C, samples were centrifuged at 12,000 rpm for 20 min in an Eppendorf^® mini-spin. Protein purification was accomplished by dialysis using 7000 MWCO dialysis cassettes (Pierce, Rockford, IL). Dialysis consisted of 3 h in 20 mM ammonium bicarbonate (NH₄HCO₃), 0.05% sodium dodecyl sulfate (SDS) at 4°C; followed by overnight dialysis in 10 mM NH₄HCO₃, 0.02% SDS at 4°C; then for 3 h in 5 mM NH₄HCO₃ at 4°C. The cassette membrane was hydrated prior to use. Equal quantities of protein (0.5 mg) were prepared for each cell type. N-Linked glycans from purified proteins were released enzymatically and purified as previously described (Hanneman et al. 2006; Ashline et al. 2007). N-Linked glycans were reduced with a solution of sodium borohydride (NaBH₄) (200 μL of 10 mg/mL NaBH₄ in 0.01 M NaOH) at room temperature overnight. The reduction reaction was terminated by the dropwise addition of acetic acid until sample pH = 5. Borate esters were removed by successive evaporations under nitrogen gas with the following solutions: (1) ethanol; (2) 3× 1% (v/v) acetic acid in methanol; and (3) 3× toluene. The reduced glycans were reconstituted in 1 mL of water and desalted on a DOWEX AG 50 W X8-400 cation exchange resin (Sigma-Aldrich). A glycan sample was loaded and eluant was immediately collected and dried in a SpeedVac^® Concentrator (Savant Instruments, Holbrook, NY). Permethylation was carried out as described by Ciucanu and Kerek (1984). Reduced permethylated glycans were resuspended in a 75% (v/v) methanol aqueous solution for mass spectrometry analysis.

Glycan analysis

Sequential mass spectra were obtained from an LTQ (Thermo Fisher Scientific, Waltham, MA) instrument equipped with a TriVersa Nanomate^®-automated nanoelectrospray ion source. Spectra were collected using Xcalibur 2.0 software (Thermo). Signal averaging was accomplished by adjusting the number microscans within each scan, generally ranging between 3 and 20 microscans. Collision parameters were left at default values with normalized collision energy set to 35% or to a value leaving a minimal precursor ion peak. Activation Q was set at 0.25, and activation time for 30 ms.

Data analysis

Initial topology and isomer assignments were made manually. The manually derived structures and their pathways were entered into GlySpy, a set of tools for data handling, complete with the glycan topology assignment algorithm, OSCAR. Ion m/z fragmentation pathways used for OSCAR were selected manually by the analyst. OSCAR was then used to determine which of the isomeric structures were consistent with the fragmentation pathways.

Abbreviations

CID

collision induced dissociation

ESI

electrospray ionization

Fuc

Fucose

Gal

galactose

GBM

glioblastoma multiforme

GlcNAc

N-acetylglucosamine

H_n≥1N_n≥1F_n≥1

hexose_n≥1N-acetylhexosamine_n≥1Fucose_n≥1

Hex

hexose

HexNAc

N-acetylhexosamine

Le^x

Lewis X epitope

MS

mass spectrometry

MSⁿ

sequential mass spectrometry