Isomer Information from Ion Mobility Separation of High-Mannose Glycan Fragments

Extracted arrival time distributions of negative ion CID-derived fragments produced prior to traveling-wave ion mobility separation were evaluated for their ability to provide structural information on N-linked glycans. Fragmentation of high-mannose glycans released from several glycoproteins, including those from viral sources, provided over 50 fragments, many of which gave unique collisional cross-sections and provided additional information used to assign structural isomers. For example, cross-ring fragments arising from cleavage of the reducing terminal GlcNAc residue on Man8GlcNAc2 isomers have unique collision cross-sections enabling isomers to be differentiated in mixtures. Specific fragment collision cross-sections enabled identification of glycans, the antennae of which terminated in the antigenic α-galactose residue, and ions defining the composition of the 6-antenna of several of the glycans were also found to have different cross-sections from isomeric ions produced in the same spectra. Potential mechanisms for the formation of the various ions are discussed and the estimated collisional cross-sections are tabulated. Graphical Abstract ᅟ Electronic supplementary material The online version of this article (10.1007/s13361-018-1890-5) contains supplementary material, which is available to authorized users.


Introduction
N -Glycans are those carbohydrates attached to asparagine residues [1] in about half of the known proteins and are critical for many of the properties of these compounds, such as cell-cell adhesion and half-life. All contain a chitobiose core attached to three mannose residues to which several glycan chains called antennae are attached. They are biosynthesised [2] in most species, including mammals, by attachment of the glycan Glc 3 Man 9 GlcNAc 2 (15, Scheme 1) to the nascent protein followed by removal of the glucose and α-mannose residues to give Man 5 GlcNAc 2 (2). These compounds and the intermediate glycans such as 3-14 are known as high-mannose glycans. Addition of GlcNAc and galactose residues to the 3-mannose residue (the nomenclature for describing features of the high-mannose glycans is outlined in Figure 1) yields socalled hybrid glycans [16][17][18][19][20] and removal of the 6 and 7 mannose residues from these hybrid glycans, followed by similar attachment of GlcNAc and galactose residues gives compounds known as complex glycans (e.g., the biantennary glycans 21 and 22). Furthermore, these glycans can be decorated with further glycan residues such as N-acetylneuraminic (sialic) acid and fucose (17)(18)(19)(20)(21)(22) at various positions.
The high-mannose glycans are now key targets in vaccine design such as those against the human immunodeficiency virus (HIV) [3,4] where the sole target for antibody neutralisation is the glycoprotein, Env, a heavily glycosylated glycoprotein consisting of a trimer of gp120/gp41 heterodimers. Understanding the detailed structure of the high-mannose glycans, several of which occur as isomers, has prompted increased interest in methods for their analysis.
Structural analysis of these compounds, following their release from the protein, can be accomplished with various techniques, of which mass spectrometry plays a key role. Ion mobility has recently been shown to provide another dimension to the mass spectrometric analysis of carbohydrates (for recent reviews, see [5][6][7]) by, for example, its ability to separate isomers  and to enable glycan ions to be selectively extracted from complex and sometimes contaminated samples [42][43][44][45][46][47]. Combined with negative ion fragmentation [48][49][50][51] of native (underivatized) glycans, this technique is proving to be an ideal method for the analysis of N-linked glycans. A further extension of this technique has been the ability of ion mobility to separate isobaric fragment ions whose different gas-phase conformations has, for example, enabled Lewis epitopes and blood group antigens to be differentiated in positive ion mode   [41]. This paper extends the above methods to an examination of the ability of ion mobility to extract structural information from fragments of high-mannose and other N-glycans generated from negative molecular ions produced in the trap region of a Waters traveling wave ion mobility Synapt G2Si mass spectrometer [52]. Most of the work so far reported on ion mobility of negative ion collision-induced dissociation (CID) fragmentation of Nglycans has been performed in the transfer region of the instrument and has been reported in detail in previous publications [48][49][50][51] . Briefly, major ions arise from 2,4 A cross-ring cleavages of the GlcNAc residues of the chitobiose core and from a B cleavage between these residues. These ions define the presence or absence of core fucose residues. Branching patterns are defined by several specific ions. An ion, termed ion D, formed by loss of the 3-antenna and chitobiose core, accompanied by a further ion formed by loss of water (termed the D-18 ion) and two cross-ring cleavage fragments of the branching mannose ( 0,3 A and 0,4 A fragments) define the composition of the 6-antenna. A pair of ions at m/z 485 and 467 are present in the spectra of high-mannose glycans when a mannose residue (mannose 8) is present on the 6-branch of the 6-antenna, as in the spectra of compounds 5, 6, 8-12, 14, and 15. Triantennary isomers can be differentiated by the masses of these ions and by the presence or absence of an ion at m/z 831 that is diagnostic for the structure 23 [53], and bisected glycans (possessing a GlcNAc residue attached to the 4-position of the branching mannose as in Compound 24) can be identified by the presence of a prominent D-221 fragment. Antenna structure is revealed by cross-ring cleavage ions such as Gal-GlcNAc-O-CH=CH-O -(m/z 424) and, finally, C 1 -Type fragments identify the residues terminating the antennae. Trap fragmentation spectra contain these ions but secondary fragments resulting from glycosidic cleavages of the major 2,4 A fragments from the core GlcNAc residues are more prominent. Most of the glycan fragment peaks are composed of several structures and, as we [41] and others [54,55] (and see the review by Hofmann and Pagel [7]) have shown for positive ion spectra, these structures can often be separated by ion mobility to provide structural information additional to that contained within the CID spectra themselves. Fragment ion structures generated from the high-mannose glycans in negative ion mode have not been studied before and are the topic of this paper. Reference to the hybrid and complex glycans is made when additional structural information can be obtained from their fragment ions.

Mass Spectrometry
Synthetic glycans (1μg/μL) and hydrazine-released glycans (similar but unmeasured concentration) were used without further purification; the PNGase F-released glycans (from about 20 μ of glycoprotein), in 2 μL of water were cleaned with a Nafion membrane [72] prior to analysis. All glycans were dissolved in 1:1 (v:v) water:methanol (about 6 μL) to which had been added about 0.2 mL of an 0.5 mM solution of ammonium phosphate (to form phosphate adducts of the glycans). Samples were infused into the nano-electrospray (nano-ESI) source of a Waters Synapt G2Si traveling wave ion mobility (TWIMS) Q-TOF mass spectrometer (Waters, Manchester, UK) through gold-coated borosilicate capillaries prepared in-house [73]. The instrument was configured as follows: capillary voltage, 0.8-1.0 kV; sample cone, 150 V; extraction cone, 25 V; cone gas, 40 L/h; source temperature, 80°C; trap collision voltage, 50-160 V (dependent on the mass of the target ion); transfer collision voltage, 4 V; trap DC bias, 60 V; ion mobility wave velocity, 450 m/s; ion mobility wave height, 40 V; trap gas flow, 1.6 mL/min; ion mobility gas flow, 80 mL/min. Spectra were acquired for about 2 min for most samples although longer time was required with some of the biological samples (normally 2-5 min although compound 8 required 22 min). Data was acquired and processed with MassLynx ver. 4.1 and Driftscope ver. 2.8 software (Waters). The nomenclature used to describe the fragment ions is that devised by Domon and Costello [74] with the addition of ion D (described above) and the use of the subscript R (for reducing terminus) to describe fragments of the reducing-terminal GlcNAc residue and R-1 to describe fragments from the other core GlcNAc. This nomenclature simplifies description of the fragmentation processes because it avoids the subscript number changing with different antenna chain lengths. Estimated rotationally averaged collision cross-sections (CCSc) were calculated by fitting the arrival time distributions (ATDs) to a single or double Gaussian distribution. A dextran calibration ladder with known absolute CCSs that were measured on a drift tube instrument was used for estimating N-glycan traveling wave CCS values as previously described [75,76].

Results and Discussion
In order to generate the CID spectra, the N-glycans are normally adducted with suitable anions to render them sufficiently stable to be transmitted to the collision cell and thus avoid extensive fragmentation in the ion source as is common with [M-H]ions [49]. Phosphate was chosen in this case because this anion is normally present in material derived from biological sources and provides a convenient stabilizing anion. Chloride is also a common constituent and a satisfactory anion but sensitivity can be compromised by the presence of the chlorine isotopes. Both anions, however, yield identical CID spectra because the first stage of fragmentation involves abstraction of a proton by the anion to leave a selection of [M-H]ions whose further fragmentation appears to be independent of the anion. Figure 1 shows the trap fragmentation of a typical high-mannose glycan (Man 9 GlcNAc 2 9). The inset shows the nomenclature used to describe various parts of the molecule. Most of the ions discussed in this paper arise as fragments of the 2,4 A R and 2,4 A R-1 ions by losses of mannose residues, normally involving additional cross-ring fragmentation. These ions are identified in Figure 1 by red and blue arrows, respectively. Also in Figure 1 are two extracted ATDs of the fragments at m/z 647 and 707, which show separated structures. Ions are discussed below in order of the number of constituent hexose residues.  Table 1. There is the possibility that the ATDs shown in the figures represent the average of unresolved conformers. However, in most cases the peak widths suggest otherwise. If conformers do exist, they would only be able to be resolved with instruments with much higher IMS resolving power. Thus, the CCS values reported here are appropriate to most existing ion mobility instruments. Measurements made on the same ion from different compounds rarely differed by more than 1 Å 2 . Two structures (a and b, Figure 2a) are possible for this ion derived from the high-mannose glycans 1-15. The extracted fragment ATD of m/z 586 from glycans 2-15 gave a single peak (green trace). A single cleavage, namely loss of the 6-antenna (mannoses 2, 6 and 7), would yield ion b from Man 5 GlcNAc 2 (2) . Two cleavages, namely loss of the 6antenna together with d1 mannose residues 3, 4, and 5 (see Figure 1) would produce this ion from the larger glycans. Man 3 GlcNAc 2 (1) where this ion was prominent, on the other hand, produced a second ion at a shorter drift time which, therefore, must have structure a. Complex, biantennary glycans with Gal-GlcNAc-antenna (e.g., 21 and 22) and the hybrid glycan (20) produced a third ion at a longer drift time. This latter ion presumably arose from the antennae and has the structure c ( Figure 2a). In the spectra of the fucosylated biantennary glycans from parotid glycoproteins, the ion with structure c was absent when both antennae contained fucose, confirming the structure and demonstrating the stability of the fucose residues in the negative ion spectra. In positive ion mode, major fragments arose from these antennae-fucosylated glycans by successive losses of fucose.
The negative ion CID spectra of complex glycans containing an α-galactose residue terminating the antennae (e.g., 22 from porcine thyroglobulin) produce a very prominent ion at m/z 586 (pink trace) with the structure d ( Figure 2b) and a drift time intermediate between ions a and c. The presence of α-galactose residues on the termini of the antennae of complex glycans is of interest to investigators developing therapeutics because of its potential antigenic properties and, consequently, the collision cross-section of m/z 586 provides an alternative way of confirming the presence of this feature.
Similar behavior to the above was seen for the corresponding ion lacking the GlcNAc residue (m/z 383) but the separation between the peaks was smaller than that for m/z 586 and is not shown.  Table 1.The extracted fragment ATD from this ion derived from Man 3 GlcNAc 2 (1) where it was the 2,4 A R ion, gave a single peak with structure e (Figure 3a, red trace, CCS = 245.3 Å 2 ± 0.6, n = 14). Man 5 GlcNAc 2 (2, pink trace) also produced this ion by presumably releasing mannoses 6 and 7. The hybrid glycans also produced a single ion at the same drift time confirming the absence of additional mannose residues in the 6-antenna. Man 6 GlcNAc 2 (3) produced two equally abundant well separated peaks (blue trace), the second of which could be formed by a single cleavage (losses of mannoses 2, 6, and 7) to give ion f. This ion is essentially linear, whereas the other is branched, probably explaining its longer drift time. Ions with these two ATDs were produced in varying relative abundance from all of the other high-mannose glycans [green traces from Man 8 GlcNAc 2 (6) Man 9 GlcNAc 2 (9) and Glc 1 Man 9 GlcNAc 2 (12)]. Thus the presence of the well-separated second peak appears to be diagnostic for mannose substitution in the 3-antenna and is useful because no fragment has been found in the CID spectrum that is diagnostic of this substitution. Hybrid glycans produced only ion e. Two equally abundant peaks with the same two   drift times as those from the above glycans were produced from the α-Gal-containing glycan (22). Clearly the second peak could not have the structure f and, thus, it was assigned the structure g. The reference sample of the biantennary glycan 21 gave a single peak corresponding to ion e. However, the same compound from thyroglobulin produced a second peak (light blue trace) with the drift time of ion g. This sample is the source of the α-galactose substituted biantennary glycan 22 and, although the presence of an isomer (structure 23) of the biantennary glycan 21 has been suspected to occur in this sample, no conclusive evidence for its existence has been found. The presence of the mobility peak at the drift time of this ion, therefore, strongly supports its existence. Undoubtedly, this ion and ion f have slightly different drift times but the resolution of current instrumentation is insufficient to resolve them. The corresponding ion lacking the GlcNAc residue (m/z 545), however, showed a slightly different behavior. It produced essentially a single peak (green trace, Figure 3b) from the high-mannose glycans Man 5-7 GlcNAc 2 (2, 3, 4, ion i) but a second one with a slight shift to higher drift times from the d1d1d3 isomer of Man 8 GlcNAc 2 (6) and Man 9 GlcNAc 2 (9, orange trace). This latter peak would, therefore, appear to have structure j. Glc 1 Man 9 GlcNAc 3 (12, pink trace) gave a broader peak consistent with the presence of ions i and j. Man 3 GlcNAc 2 (1) and the hybrid glycan (16) produced an earlier peak (red trace, h), coincident with the shoulder on the main peak (i) from the other high-mannose glycans. The structure (h) is confirmed by its presence as the 2,4 A R-1 ion from Man 3 GlcNAc 2 (1). The three Glc 3 -containing glycans (13, 14, and 15) produced a minor fourth peak at a longer drift time (blue trace from compound 14) attributed to the linear structure k. For each ion the table shows the glycan structure, the ion number (in bold), the mass and the collisional cross-section in Å 2 . Where multiple measurements of the CCS values were obtained, the error (from the STDEV function of Microsoft Excel) and the number (n) of measurements is given.

Fragments of Composition Hex 3 GlcNAc 1 (m/z 688)
An important feature of the negative ion CID spectra of Nglycans is the presence of a series of ions specific to the 6antenna and, thus, revealing its composition. Elimination of the chitobiose core and the 3-antenna yields an ion containing the 6-antenna together with the core branching mannose residue (mannose 1, Structure α in Figure 3, panel d). This ion has been named the D ion and is accompanied by another ion formed by additional loss of water (the D-18 ion) and two cross-ring cleavage ions ( 0,3 A and 0,4 A) of the branching mannose as described in the Introduction. The D ion from biantennary glycans such as   19 and 20, i.e., those glycans lacking the α1→2 mannose residues 8 and 9 gave a single peak (ion p, red traces, Figure 4a), whereas that from the various isomers of Man 8 GlcNAc 2 (6-8) and Man 9 GlcNAc 2 (9) with mannose residues 8 and 9 produced a peak with a slightly greater drift time (blue traces, panel a). This behavior is consistent with the occurrence of structures r and/or s. Assignment of structure r is consistent with the general observation that the presence of mannose 8 leads to longer drift times and evidence for ion s is provided by the presence of an ion at this drift time from the d1d1d2-isomer of Man 8 GlcNAc 2 (7). The broad, green trace from the d1d1-isomer of Man 7 GlcNAc 2 (5) appeared to contain ions p and q, both of which could be formed from a single cleavage from the 2,4 A R fragment. Support for the structure of ion q was the observation that it was seen in the spectra of glycans having mannoses 4 and 5 in the 3-antenna but which lacked the α1→2 mannoses 8 and 9. An ion with the drift time of ion s was also prominent in the spectrum of the d1d1d2 isomer of Man 8 GlcNAc 2 (7). The broadness of the peaks (blue traces) was consistent with some or all of these structures being present. The hybrid glycan, Man 6 GlcNAc 3 Fuc 1 (20), produced a third peak (red trace, Figure 4b) with a slightly shorter drift time than that produced by Man 5 GlcNAc 2 (2) even though it contained the same 6antenna. This peak was coincident with the early part of a broad peak from the biantennary glycan Gal 2 Man 3 GlcNAc 4 Fuc 1 (21, pink trace, Figure 4b), which appeared to consist of the ions u and v. Clearly, this latter compound does not have four attached hexose residues leaving the structures u and v as the only possible ions. The hybrid glycan Man 4 GlcNAc 3 Fuc 1 (17) produced yet another peak (green trace) with an even shorter drift time. Because this ion contains all four mannose residues from this glycan it must have structure t. (6, blue trace), Man 9 GlcNAc 2 (9, green trace, illustrated in Figure 1), and Glc 1 Man 9 GlcNAc 2 (12, not shown) produced a well-separated second peak consistent with structure y. The longer drift time produced by the presence of mannose residue 8 in the 6-antenna was consistent with earlier observations made above and on the parent compounds [60]. Man 9 GlcNAc 2 (9), the d1d1d2 isomer of Man 8 GlcNAc 2 (7, pink trace), and Glc 1 Man 9 GlcNAc 2 (12, not shown) produced a third peak that was just separated from the first to which structure x was assigned. This ion could be produced from the 2,4 A R-1 fragment by cleavage of the 3-antenna (mannoses 3, 4, and 5) together with loss of mannoses 6 and 8. Although the separation between the first (red) and second (pink) peaks was only marginal, it was observed consistently on repeat experiments. All three of the glycans containing three glucose residues in the 3antenna (13-15) produced an additional ion at longer drift times. This ion appears to have the structure z and has been observed as a major diagnostic ion in the transfer fragmentation spectra of these compounds.

Fragments of Composition Man 4 (m/z 647)
This was the only ion of the Hex n series that showed any significant drift time differences among the various possible isomers and is an important ion because it is a potential D ion (structure ac, Figure 4d) from hybrid and complex glycans containing mannoses 2, 6, and 7 in the 6-antenna. However, as with the D ion from the biantennary glycans (m/z 688), discussed above, m/z 647 can also be produced from other regions of the molecules and its positive identification as a D ion by ion mobility would be useful. A major and a minor peak were recorded from this ion from Man 5 GlcNAc 2 (2), Man 6 GlcNAc 2 (3, Figure 4d, red traces), and from the d1d1 isomer of Man 7 GlcNAc 2 (4), which would be expected to produce a D ion at m/z 647. The ion from Man 7 GlcNAc 2 (d1d3 isomer) (5 green trace) and Man 8 GlcNAc 2 (6), which have additional mannose residues in the 6-branch of the 6-antenna and for which m/z 647 is not the D ion, produced peaks with a longer drift time (green trace) consistent with the above observations. These peaks would appear to have structure ad as the result of a single cleavage from the 2,4 A R or 2,4 A R-1 ions. A single peak with a shorter drift time than the other two, and consistent with structure ab, was recorded from the d1d1d2 isomer of Man 8 GlcNAc 2 (7, blue trace). This ion was presumably produced by a similar single cleavage. The ion from Man 9 GlcNAc 2 (9, pink trace and illustrated peak in Figure 1) also produced a peak with a similar drift time, together with a second at a drift time equivalent to that from the other glycans containing the 8-mannose in the 6antenna (ion ad). The minor peaks with shorter drift times on the side of the trace (red) from ion D (structure ac) appear to have structure aa. All the three glucosecontaining glycans (13)(14)(15) produced the same single ATD peak with a drift time (red traces, Figure 4e) slightly shorter than that produced by ion D. The equality of the drift time of m/z 647 from these three glycans suggests that all three ions had the same structure with ion ae being the obvious candidate. The ATD peak from the D-18 ion (m/z 629) from Man 5 GlcNAc 2 (2) and Man 6 GlcNAc 2 (3) and from the corresponding ion in the spectra of the other highmannose glycans was broad and ill-defined, suggesting water loss from several sites.   (2) and Man 6 GlcNAc 2 (3). Ion ac is also known as ion D. Green trace, ion ad from glycans with mannose 8 in the 6-antenna (from Man 7 GlcNAc 2 (5). Blue trace, ion ab from the d1d1d2 isomer of Man 8 GlcNAc 2 (7). Pink trace, doublet from Man 9 GlcNAc 2 (9). (e). Blue and green traces as in panel a together with traces from ion ae from glycans containing Glc 3 groups lacking the GlcNAc residue (m/z 869, Figure 5a) produced separated isomers, Man 5 GlcNAc 2 (2, red trace), Man 6 GlcNAc 2 (3, blue trace), Man 7 GlcNAc 2 (d1d1-isomer, 4, green trace), Man 9 GlcNAc 2 (9), Glc 1 Man 9 GlcNAc 2 (12), and the hybrid glycan Gal 1 Man 5 GlcNAc 2 (20) produced a prominent peak with a CCS of 272.3 Å 2 ± 0.9 (n = 8). This ion is the 2,4 A R-1 ion from Man 5 GlcNAc 2 and, therefore, must have structure ak. Although this structure is consistent with the structures of glycans 3, 4, and 10, the single symmetrical peak from Man 9 GlcNAc 2 (9, light blue) is more likely to have structure af because it can be formed in a single cleavage from the molecular ion by an 0,4 A 3 cleavage of mannose 1. Support for this proposal comes from the spectrum of the monoglucosylated Man 9 -analogue (12), which produced a single peak at the same drift time.
A minor peak appeared on the shorter drift time side of this peak in the spectrum of Man 6 GlcNAc 2 (3) (blue trace) and a larger one was present in the spectrum of the d1d1isomer of Man 7 GlcNAc 2 (4, green trace). This peak was absent from the spectra of Man 9 GlcNAc 2 (9) and the glucosylated analogue (12). Given that the difference in structure between Man 5 GlcNAc 2 (2) and Man 6 GlcNAc 2 (6) is the presence of the mannose at position 4, this indicates that the ion has the structure ag. An ion at the same drift time but with a higher relative abundance was present in the spectrum of the d1d1-isomer of Man 7 GlcNAc 2 (6, green trace). This ion could have the same structure, produced by loss of two single mannose residues from the 2,4 A R-1 ion or structure ah, also the product of two cleavages. The d1d1d2 isomer of Man 8 GlcNAc 2 (7) also produced a major ion at this drift time (purple trace) but in this case the most likely structure is ai produced in a single cleavage from the 2,4 A R-1 ion by loss of the 3-antenna (mannoses 3, 4, and 5). However, structures ag and ah cannot be discounted. Further ions with longer drift times were also present in the spectrum of this glycan but their structures are uncertain. The d1d1d3 isomer of Man 8 GlcNAc 2 (6, black trace) and the d1d3 isomer of Man 7 GlcNAc 2 (5) also produced broad peaks indicating the presence of several structures but with a maximum consistent with structures aj, which could also be formed in a single cleavage from the 2,4 A R-1 ion. An additional peak (orange trace) was observed from the three glycans with three glucose residues in the 3-antenna (11)(12)(13). As with m/z 707 discussed above, this ion is a major diagnostic ion for glycans with this composition and, because it is produced from all three of these glycans, it must have the structure al.

Hexose 6
Fragments of Composition Hex 6 GlcNAc 1 -O-CH=CH-O -(m/z 1234) Multiple structures were also produced from m/z 1234; extracted fragment ATDs are shown in Figure 5b. This ion is the 2,4 A 6 fragment from glycans of composition Man 6 GlcNAc 2 (3) and is present as a single symmetrical peak from the d1d1 isomer (3, ion am, red trace). It is also present as a fragment in the spectra of various isomers of Man 7 GlcNAc 2 (4, 5) and Man 8 GlcNAc 2 (6). In the spectrum of glycans carrying mannose 8 in the 6-antenna, i.e., the d1d3 isomer of Man 7 GlcNAc 2 (5), the d1d1d3 isomer of Man 8 GlcNAc 2 (6, blue trace), and Glc 3 Man 8 GlcNAc 2 (14, pink trace), a second well-separated, abundant peak was present at higher drift times. The position of this peak is consistent with the incorporation of mannose 8 from the 6-antenna indicating structure ao. Peaks from the other compounds were broad but maxima from both Man 9 GlcNAc 2 (9, green trace) and its glucosylated analogues (12 and 15) suggested structure an formed by the loss of the 3-antenna in a single cleavage from the 2,4 A R ion.
Similar profiles were observed from the ion lacking the GlcNAc moiety (m/z 1031) but with less separation between most of the peaks (Figure 5c). However, the ion at m/z 1031 from Man 10 GlcNAc 2 (11) derived from S. Cerevisiae (orange trace) appears to have the structure aq formed by loss of the 6-antenna from the 2,4 A R-1 ion.

Hexose 7
Only marginal separation was seen with various structures of ions containing seven mannose residues and no useful conclusions were drawn regarding their structures.

Hexose 8
Fragments of Composition Hex 8 GlcNAc 1 -O-CH=CH-O -(m/z 1558) ATDs of fragments containing eight mannose residues are shown in Figure 6. The ion at m/z 1558 is the 2,4 A 6 ion from the Man 8 GlcNAc 2 glycans (6-8). The d1d1d3 isomer of Man 8 GlcNAc 2 (6) produced a single peak (Figure 6a, red trace), which has the structure of ion au. It was also produced from Glc 3 Man 8 GlcNAc 2 (14) by loss of the three glucose residues. The corresponding ion as from the d1d1d2 isomer (7, blue trace) had a shorter drift time, consistent with the behavior shown by the parent compounds [60]. These two isomers can be distinguished by their characteristic negative ion CID spectra recorded in the transfer region of the Synapt instrument but the trap fragmentation spectra were virtually identical. In the transfer fragmentation spectrum, both the d1d1d2 (7) and d1d1d3 isomers (6) produce D, D-18, 0,3 A 4 and 0,4 A 4 ions at m/z 809, 791, 737, and 707, respectively. The glycans can be differentiated because the d1d1d3 isomer (6) produces a characteristic pair of ions at m/z 485 and 467, whereas the d1d2 isomer (7) does not. No diagnostic ions have been found to indicate the presence of the d1d1d2 isomer so that when m/z 485 and 467 are present, it is not always possible to say if the spectrum is from a mixture of both isomers, particularly as the relative abundance of m/z 485 and 467 varies with the collision energy. However, observation of the profiles of m/z 1558 in the trap fragmentation spectrum leaves no doubt as to which isomers are present. Figure 6b shows the profile  (6). The blue trace is the corresponding ion (as) from the d1d1d2-isomer of Man 8 GlcNAc 2 (6) and the green trace (at) the ion from the d1d2d3 isomer. Glycans with extended 3antennae (e.g., Glc 3 Man 9 GlcNAc 2 ) also gave this ion (pink trace). The orange trace is a fragment of Glc 3 Man 7 GlcNAc 2 but its structure is uncertain. (b) The blue trace (as) is the 2,5 A R ion from the d1d1d2-isomer of Man 8 GlcNAc 2 (6) and the red trace is the corresponding profile of Man 8 GlcNAc 2 from porcine thyroglobulin showing both isomers 6 and 7. The green trace (av) is from Man 9 GlcNAc 2 showing fragmentation by losses of each of the α-mannose residues from Man 8 GlcNAc 2 obtained from a sample of glycans released from porcine thyroglobulin (red trace) showing the presence of both isomers. Glc 1 Man 9 GlcNAc 2 (12, green trace Figure 6a), Glc 3 Man 9 GlcNAc 2 (15, pink trace), and the d1d2d3 isomer of Man 8 GlcNAc 2 (8, green trace) produced a third ion whose structure must be at. This ion could also simply be formed from the glucose-containing glycans by a single elimination from the 3-antenna. Thus, the difference in drift time of the 2,4 A R from these three isomers of Man 8 GlcNAc 2 (structures as, at, and au) allows them to be identified by their extracted fragment ATDs. A fourth ion was produced from Glc 3 Man 7 GlcNAc 2 (13, orange trace). Several structures can be proposed but no evidence was found to determine which one(s) were present.
Ion mobility in the trap region also provided additional information on the fragmentation pathways shown by these ions. Thus m/z 1558 is formed from Man 9 GlcNAc 2 (9) by loss of a mannose residue from one of the three antennae. The trap fragmentation profile of this ion as a fragment of Man 9 GlcNAc 2 (9) is shown in Figure 6b (av, green trace) superimposed on the profiles from the d1d1d2 and d1d1d3 isomers of Man 8 GlcNAc 2 (7 and 6, respectively). The broad, unresolved profile indicates the presence of several ions at this mass produced by loss of α-mannose residue from each antenna.
The fragments lacking the GlcNAc moiety ( 2,4 A R-1 ion, m/z 1355) showed only marginal separation and were not considered analytically useful.

Hexose 9
Fragments of Composition Hex 9 GlcNAc 1 -O-CH=CH-O -(m/z 1720) Two isomers of Man 9 GlcNAc 2 (9 and 10) were available, giving m/z 1720 as the 2,4 A R ions with different cross-sections as shown in Figure 7a and b, and Table 1 (ions aw and ax, respectively). The extracted fragment ATD of ion aw from Man 9 GlcNAc 2 (9) is shown as the red trace in Figure 7b. Loss of one of the mannoses from the 6-antenna was responsible for the formation of an isomeric ion with a slightly longer drift time than ion aw from Glc 1 Man 9 GlcNAc 2 (12, ion ba, blue trace, Figure 7c) but the particular mannose that was eliminated was not determined. Loss of a hexose residue from Glc 3 Man 7 GlcNAc 2 (13) produced an ion with a considerably longer drift time (ions bb and bc, green trace) but, again, the specific hexose loss was not determined.  Figure 7b. Unlike the situation with the corresponding GlcNAc-containing 2,4 A R ion, the drift time for the ion from the isomer 10 was shorter than that of the ion from the d1d1d3 isomer 9, The profile (blue trace) from the S. cerevisiae-derived glycan (10) was asymmetric, revealing the presence of the d1d1d3-isomer az. Figure 7d shows the extracted fragment ATD of m/z 1517 from Red trace 2,4 A R-1 ion (az) from Man 9 GlcNAc 2 (9). Blue trace, ion bd as a fragment of Glc 1 Man 9 GlcNAc 2 (12) Man 9 GlcNAc 2 (recorded on a different occasion, hence the slightly different drift time), with that from the fragment produced by mannose loss from Glc 1 Man 9 GlcNAc 2 (12, ion bd). Again, the ion (az) from the d1d1d2 isomer of Man 9 GlcNAc 2 (9) had the longer drift time in contrast to the situation with the 2,4 A R ion aw. Ions from the Glc 3 -containing glycans were too weak to determine if significant separation occurred; most had drift times slightly less than that of ion az.

Hexose 10
These fragments were only found in the spectra of the glucosecontaining glycans and are shown in Figure 8.
Fragments of Composition Hex 10 GlcNAc 1 -O-CH=CH-O -(m/z 1882) The extracted fragment ATDs of this ion as the 2,4 A R fragment from Glc 1 Man 9 GlcNAc 2 (12, 428.3) and Glc 3 Man 7 GlcNAc 2 (13, 431.6) produced different drift times (ions bf, red trace and bg, blue trace, respectively, Figure 8a). Again, the glycan with three glucose residues in the 3-antenna produced the longer drift time, reflecting its less compact structure. Glc 3 Man 8 GlcNAc 2 (14, green trace) produced two peaks, one of which had a drift time coincident with that of ion bg and one with a considerably shorter drift time. The ion with the longer drift time could be formed by loss of mannose 8 to give ion bg or by loss of a glucose residue to give be. Glc 3 Man 9 GlcNAc 1 (15, pink trace) also gave a minor ion of undetermined structure at this drift time.
Fragments of Composition H ex 1 0 -O-CH=CH-O -(m/z 1679) The ATDs of the 2 , 4 A R -1 ions from Glc 1 Man 9 GlcNAc 2 (12, ion bj, red trace, CCS 390.2 Å 2 ) and Glc 3 Man 7 GlcNAc 2 (13, ion bh, green trace, 380.0 Å 2 ) were well separated but, as with the corresponding Hex 9 fragments above, their relative drift times were reversed from the ions containing GlcNAc (Figure 8b). Proposed structures are in Figure 8b. Glc 3 Man 9 GlcNAc 2 (15) produced an ion with a drift time identical to that from the corresponding ion from Glc 1 Man 9 GlcNAc 2 , (12) presumably by loss of two glucose residues, whereas Glc 3 Man 8 GlcNAc 2 (14) produced a third ion of indeterminate structure (probably bi).

Conclusions
This work shows that the ATDs of fragment ions in the negative ion CID spectra of high-mannose N-glycans can be used to obtain additional structural and fragmentation information to that present in the CID spectra themselves. Most of the ions that provided useful information fell into two groups; those arising from the 2,4 A R fragments by additional losses from the antennae, or from corresponding fragmentation of the 2,4 A R-1 ions. Many of the ions showed sufficient differences in collision cross-section to enable separation to baseline. Of particular significance is the additional information that this method provided on the presence of isomers. For example, although all three isomers of Man 8 GlcNAc 2 can be identified individually by their negative ion CID spectra, determination of the relative amounts of, for example, the d1d1d2 isomer (7) in the presence of the d1d1d3 isomer (6) is difficult because the former isomer is only identified by the absence of the fragments at m/z 485 and 463. However, the CCSs of the 2,4 A R ions from these isomers are significantly different from each other and from that of the third isomer (8) to allow easy differentiation.
The CID spectra contained ions specific to the composition of the 6-antenna of these compounds. Specifically, an  (12) and Glc 3 Man 7 GlcNAc 2 (13) respectively, Blue trace, fragments from Glc 3 Man 8 GlcNAc 2 (14) ion formed by loss of the chitobiose core and 3-antenna, known as the D ion, is of particular interest. However, in some cases, other ion structures with the same mass can sometimes occur from fragmentation of other parts of the molecules, meaning that positive identification of the D ion is sometimes difficult. In this work, it was found that the D ion at m/z 647 from high-mannose glycans with three mannose residues in the 6-antenna, and m/z 688 from biantennary glycans, produced unique CCSs allowing them to be identified. In general, glycans with mannose 8 in the 6-antenna produced longer drift times than most of the other isomers. The exception was glycans with three glucose residues capping the 3-antenna. These compounds invariably produced cross-ring fragment ions (Glc 3 Man 1 -O-CH=CH-Oor Glc 3 Man 2 -O-CH=CH-O)with high CCS values. These ions are prominent in the CID spectra and are used as diagnostic fragments for this structural feature. In general, linear ions such as these appeared to have higher CCS values than isomeric ions with branched structures. Other structural features such as the presence of α-galactose residues on the non-reducing termini of the antenna also gave specific CCS values. These moieties are of interest to the biopharmaceutical industry because of their antigenic properties.
Although many of the ATDs shown above appeared to be produced by single ions, the possibility that some contained contributions from other minor fragments with similar CCSs cannot be ruled out. This situation is most likely to occur when fragmentation involves more than one additional stage. However, formation of many of the fragments could be rationalized by a single additional cleavage from the 2,4 A R or 2,4 A R-1 ions, and, in these cases, the ATDs probably represented single structures. Another point to be taken into account is that the above data were recorded from a comparatively small selection of glycans, although most possible high-mannose glycans were included. Thus, CCSs, such as those from the D ions, although appearing to be unique, may not, in fact, be so. However, even in these situations, the absence of a peak with the specific CCS of these ions can be taken as positive information on their absence.
The shape of the ATD peaks also provided some information on fragmentation mechanisms. Thus, fragmentation of the larger high-mannose glycans such as Man 9 GlcNAc 2 (9) by loss of a single α-mannose residue can involve any of the three residues present at the nonreducing termini of the three antennae. The very broad, ill-defined peak that was observed spanning the CCS range for the 2,4 A R ions from the three Man 8 GlcNAc 2 isomers observed for this ion showed that the loss from Man 9 GlcNAc 2 involved any of the terminal mannose residues in roughly equal amounts.
Further work in this area will focus on positive ions to ascertain if corresponding structural information can be obtained in a manner similar to recent work that provided structural information on Lewis epitopes.