Christian
Manz
ab,
Márkó
Grabarics
ab,
Friederike
Hoberg
b,
Michele
Pugini
ab,
Alexandra
Stuckmann
a,
Weston B.
Struwe
*c and
Kevin
Pagel
*ab
aInstitute of Chemistry and Biochemistry, Freie Universität Berlin, Takustrasse 3, 14195 Berlin, Germany. E-mail: kevin.pagel@fu-berlin.de
bFritz Haber Institute of the Max Planck Society, Department of Molecular Physics, Faradayweg 4-6, 14195 Berlin, Germany
cOxford Glycobiology Institute, Department of Biochemistry, University of Oxford, Oxford OX1 3QU, UK. E-mail: weston.struwe@chem.ox.ac.uk
First published on 26th July 2019
The analysis of complex oligosaccharides is traditionally based on multidimensional workflows where liquid chromatography is coupled to tandem mass spectrometry (LC-MS/MS). Due to the presence of multiple isomers, which cannot be distinguished easily using tandem MS, a detailed structural elucidation is still challenging in many cases. Recently, ion mobility spectrometry (IMS) showed great potential as an additional structural parameter in glycan analysis. While the time-scale of the IMS separation is fully compatible to that of LC-MS-based workflows, there are very few reports in which both techniques have been directly coupled for glycan analysis. As a result, there is little knowledge on how the derivatization with fluorescent labels as common in glycan LC-MS affects the mobility and, as a result, the selectivity of IMS separations. Here, we address this problem by systematically analyzing six isomeric glycans derivatized with the most common fluorescent tags using ion mobility spectrometry. We report >150 collision cross-sections (CCS) acquired in positive and negative ion mode and compare the quality of the separation for each derivatization strategy. Our results show that isomer separation strongly depends on the chosen label, as well as on the type of adduct ion. In some cases, fluorescent labels significantly enhance peak-to-peak resolution which can help to distinguish isomeric species.
A common way to resolve and separate glycan isomers is liquid chromatography (LC). Reversed phase (RP) chromatography, which is commonly used for protein analysis, often struggles with glycans due to their inherently high polarity. Instead, other stationary phases such as hydrophilic interaction chromatography (HILIC) and porous graphitic carbon (PGC) are often applied as an alternative, powerful way to separate glycan isomers.6 However, as glycans naturally do not contain chromophores or fluorophores, it is often necessary to derivatize them with fluorescent labels to facilitate a sufficient detection and enable quantification.7
Another emerging and promising technique capable of separating glycan isomers is ion mobility spectrometry (IMS).5,8,9 Here, ions travel through a drift cell filled with an inert buffer gas under the influence of a weak electric field and undergo low-energy collisions with the buffer gas. Compact ions collide less frequently with the buffer gas than more extended ions, which leads to a separation based on size, shape and charge. This enables the separation of isomeric species as shown for small molecules,10 oligosaccharides as well as for glycoconjugates.11,12 In addition, the resulting drift times can be converted into the rotationally averaged collision cross-section (CCS). When measured under controlled conditions, CCSs can be universally compared, which enables an efficient incorporation into databases to allow for structural elucidation.13–17 While over the last several years both LC and IMS, showed their individual capabilities to resolve glycan isomers, very few attempts have been made to combine both methods into a consistent LC-IM-MS workflow for glycan analysis.18 Importantly, the impact of derivatization, in particular with fluorescence labels, on the mobility separation of isomeric glycans is poorly understood. To close this gap, we present a systematical analysis using a set of isomeric glycans derivatized with different common glycan fluorophores as well as native and reduced species. Our data indicate that labelling can significantly affect the ability to separate individual glycan isomers via IM-MS. Depending on the label, this can diminish or improve selectivity, and therefore, labels should be specifically selected for a given glycan analysis.
For nano-electrospray ionization (nano-ESI) typically 5 μL of sample was loaded to a capillary and electrosprayed by applying a capillary voltage of 0.6–1.1 kV. Typical parameters in positive ion mode were: 60 V sampling cone voltage, 1 V source offset voltage, 30 °C source temperature, 0 V trap CE (MS) up to 30 V trap CE (MSMS), 2 V transfer CE, 3 mL min−1 trap gas flow. Ion mobility parameters were: 2.2 Torr helium IMS gas, 27–30 °C IMS temperature, 5.0 V trap DC entrance voltage, 5.0 V trap DC bias voltage, −10.0 V trap DC voltage, 2.0 V trap DC exit voltage, −25.0 V IMS DC entrance voltage, 50–180 V helium cell DC voltage, −40.0 V helium exit voltage, 50–150 V IMS bias voltage, 0 V IMS DC exit voltage, 5.0 V transfer DC entrance voltage, 15.0 V transfer DC exit voltage, 150 m s−1 trap wave velocity, 1.0 V trap wave height voltage, 200 m s−1 transfer wave velocity, 5.0 V transfer wave height voltage.
In negative ion mode typical parameters were: 90 V sampling cone voltage, 10 V source offset voltage, 30 °C source temperature, 0 V trap CE (MS) up to 30 V trap CE (MSMS), 2 V transfer CE, 3 mL min−1 trap gas flow. Ion mobility parameters were: 2.2 Torr helium IMS gas, 27–30 °C IMS temperature, 1.0 V trap DC entrance voltage, 2.0 V trap DC bias voltage, −1.0 V trap DC voltage, 1.5 V trap DC exit voltage, −25.0 V IMS DC entrance voltage, 50–150 V helium cell DC voltage, −40.0 V helium exit voltage, 50–150 V IMS bias voltage, 0 V IMS DC exit voltage, 5.0 V transfer DC entrance voltage, 15.0 V transfer DC exit voltage, 200 m s−1 trap wave velocity, 10.0 V trap wave height voltage, 250 m s−1 transfer wave velocity, 3.0 V transfer wave height voltage. The resulting drift times were converted to rotationally-averaged collision cross-sections (CCS) using the Mason–Schamp equation.23
![]() | ||
Fig. 1 The investigated set of isomeric blood group epitopes. (a) Tetrasaccharide Lewis B (LeB) and the corresponding trisaccharide fragments/motifs Lewis A (LeA) and blood group H1 (BG H1). (b) Tetrasaccharide Lewis Y (LeY) and the corresponding trisaccharide fragments/motifs Lewis X (LeX) and blood group H2 (BG H2). (c) Glycan structures are depicted using the SNFG nomenclature.25 |
Recently, we investigated these isomeric tri- and tetrasaccharides in an underivatized form using IM-MS and showed that fragment CCS can be used as fingerprints to systematically differentiate between the epitopes.28 The intact tetrasaccharide precursors exhibit very similar drift times and CCS, which makes it difficult to distinguish them using IMS. However, fragmentation of the tetrasaccharide precursors with collision induced dissociation (CID) yields trisaccharide fragments that can be used to identify specific terminal fucose motifs. Some of those isomeric fragments such as LeX and BG H2 can be readily distinguished by IMS, while LeA and BG H1 are difficult to differentiate in underivatized form. In the present study, we focus on the IMS separation of derivatized forms of these epitopes.
After reductive amination with fluorescent labels such as 2-AB, 2-AA or ProA, the reducing end monosaccharide will exhibit an open ring structure. To study the influence of this ring opening, we reduced glycans to open-ring alditols (Red) to compare them with the predominantly closed-ring native structures. The reduction of glycans often precedes permethylation, but alditols themselves are also often used as standalone modification for various stationary phases.39
![]() | ||
Fig. 3 ATDs and DTCCSHe of the isomeric blood group epitopes LeA vs. BG H1 (left panel) and LeX vs. BG H2 (right panel) as sodium adducts in He drift gas. |
As native glycans, LeA and BG H1 show minor isomer separation compared to LeX and BG H2, which are almost baseline separated. Compared to the native closed-ring structure, the ring opening during reduction to alditols does not seem to have a significant impact on the separation of LeA and BG H1. In contrast, the drift-time difference of LeX and BG H2 decreases significantly after reduction. This effect is further amplified after introduction of the chromophore labels ProA, 2-AA and 2-AB. 2-AA and 2-AB labelled glycans show the largest isomer separation for the LeA and BG H1 isomers, while LeX and BG H2 are basically indistinguishable.
Similarly, all native and derivatized isomers were measured as alkali metal adducts, which are known to significantly alter isomer separation in IMS.40 Negatively charged adducts such as chloride and nitrate complexes predominantly lead to the formation of deprotonated ions, which are therefore the only ions with negative polarity studied here.41,42 In Table 1, the CCSs of protonated and deprotonated glycans, as well as for three commonly observed typical alkali adducts (Li+, Na+, K+) measured in helium (DTCCSHe) are shown. The upper part of Table 1 shows the CCSs of all derivatives and metal adducts of the LeB series, while the lower part shows the CCSs of all modifications of the LeY series. The exact masses of all labelled glycans are shown in Table S1.†
DTCCSHe (Å2) | Type | Native | Red | 2-AB | 2-AA | ProA |
---|---|---|---|---|---|---|
LeB | [M + H]+ | 167 | 167 | 189 | 189 | 209 |
[M + Li]+ | 165 | 159 | 179 | 178 | 209 | |
[M + Na]+ | 166 | 160 | 180 | 178 | 210 | |
[M + K]+ | 166 | 164 | 182 | 180 | 214 | |
[M − H]− | 165 | 161 | 184 | 179 | 218 | |
LeA | [M + H]+ | 144 | 144 | 168 | 167 | 183 |
[M + Li]+ | 143 | 138 | 159 | 158 | 188 | |
[M + Na]+ | 145 | 141 | 161 | 159 | 190 | |
[M + K]+ | 145 | 146 | 163 | 163 | 194 | |
[M − H]− | 143 | 141 | 162 | 160 | 190 | |
BG H1 | [M + H]+ | 144 | 144 | 170 | 168 | 188 |
[M + Li]+ | 146 | 142 | 167 | 167 | 195 | |
[M + Na]+ | 147 | 143 | 169 | 168 | 195 | |
[M + K]+ | 145 | 146 | 169 | 170 | 193 | |
[M − H]− | 143 | 143 | 163 | 157 | 192 | |
LeY | [M + H]+ | 171 | 169 | 190 | 191 | 209 |
[M + Li]+ | 163 | 163 | 190 | 190 | 219 | |
[M + Na]+ | 164 | 164 | 192 | 191 | 218 | |
[M + K]+ | 165 | 166 | 192 | 191 | 217 | |
[M − H]− | 167 | 164 | 185 | 183 | 213 | |
LeX | [M + H]+ | 144 | 145 | 169 | 170 | 188 |
[M + Li]+ | 138 | 141 | 169 | 165 | 202 | |
[M + Na]+ | 140 | 142 | 171 | 168 | 202 | |
[M + K]+ | 141 | 144 | 171 | 169 | 198 | |
[M − H]− | 146 | 141 | 161 | 159 | 189 | |
BG H2 | [M + H]+ | 144 | 145 | 172 | 166 | 185 |
[M + Li]+ | 148 | 145 | 170 | 169 | 201 | |
[M + Na]+ | 148 | 146 | 171 | 170 | 201 | |
[M + K]+ | 149 | 148 | 172 | 172 | 197 | |
[M − H]− | 144 | 139 | 163 | 155 | 190 |
Generally, there is a clear trend of increasing CCS from the native glycans up to the ProA-labelled glycans. The CCSs for all species is growing proportionally to the size of the added fluorescent label, and is therefore correlated to the increase in molecular mass.40 A similar trend is observed with the addition of alkali metals, which generally lead to larger CCSs in the order of H+ < Li+ < Na+ < K+. Deprotonated species on the other hand behave similarly counterparts. However, there are some exceptions to this behaviour. Especially alditols (Red) seem to have their largest CCS when protonated or adducted with potassium, while sodiated and lithiated species show significantly smaller CCSs. Another example are glycans labelled with 2-AA, whose protonated species show larger CCSs than metal adducted species, which indicates a compaction of the gas-phase structure with the addition of small alkali metal ions. This behaviour is a result of the structure of the oligosaccharide-metal complex, which is dictated by the solvation of the metal cation.
![]() | (1) |
Here, 2σ is the temporal peak width measured between the two inflection points of a peak of Gaussian profile and t the drift time. In case of separating isomers of the same charge state by linear drift tube ion mobility spectrometry (DTIMS), eqn (1) can be rewritten in the following form:
![]() | (2) |
In eqn (2)N and p are the average plate number and average resolving power, respectively (the latter one should not be confused with peak-to-peak resolution). The relation between the two figures of merit is given by definition as
, the origin of the constant in the formula being the difference between the standard deviation and the full-width-at-half-maximum (FWHM) of Gaussian distributions. The fraction
is the relative difference between the collision cross sections of the ions that are to be separated, i.e. it is a measure of selectivity. In following, we used this term to describe the difference in CCS between the tetrasaccharide isomer pair LeY/LeB as well as the trisaccharide isomer pairs LeX/BG H2 and LeA/BG H1 and visualize this difference as heat map in Table 2.
The upper part of Table 2 shows the isomer separation for the tetrasaccharides LeY/LeB. Here two general trends are observed: (1) isomer separation is increased when a fluorescent label is introduced and (2) isomer separation is improved upon adduct formation with alkali metal adducts. There are, however, significant differences between each individual modification. While native tetrasaccharides only show up to 2.4% CCS difference, 2-AA and 2-AB labelled species separate much better with a difference of up to 6.9% for sodium adducts. ProA-labelled isomers, on the other hand, are separated as lithiated species with a difference of 4.9%. Thus, specific labels can increase isomer separation, which in some cases makes them beneficial for IMS separation.
A very similar behaviour of improved separation is observed for the LeB submotifs LeA vs. BG H1. Native structures of these isomers do practically not separate in IMS independent of the charge carrier; only lithiated and sodiated species do show minor differences up to 2.1%. In contrast, a functionalization with fluorescent labels yields considerably different CCSs, which differ up to 5.5%.
However, as shown for the LeY submotifs LeX vs. BG H2, the above-mentioned trends cannot be generalized and may in some cases even be reversed. Here, the native, underivatized form of the glycan show a difference of 7% in CCS for lithium adducts. Upon modification of the reducing end, the quality of the separation suffers drastically. With up to 3% difference alditol structures may be resolvable on some instruments, while 2-AA, 2-AB and ProA labelled ions cannot be distinguished (<2%). Remarkably, although the trend is reversed for the trisaccharide isomer pairs LeA/BG H1 and LeX vs. BG H2, the average CCS difference is similar at ∼2%. For the tetrasaccharides, the CCS difference is even larger with almost 3%.
To evaluate which CCS difference is sufficient to identify two isomeric species in a mixture, the resolving power from eqn (2) has to be considered. Besides showing the most important factors that influence and, ultimately, determine Rs in DTIMS, eqn (2) also provides a means to calculate the required resolving power (Rp) to achieve a specified peak-to-peak resolution for a given pair of ions. If the relative CCS difference of two ions is 2%, a resolving power of 64 (corresponding to a plate number of 22500) is required to distinguish them (i.e. separation with a peak-to-peak resolution of Rs = 0.75). To achieve baseline resolution for the same two peaks (Rs = 1.5), the resolving power has to be substantially higher, approximately 127 (corresponding to a plate number of 90
000). This is already achievable with state-of-the-art custom-built and commercial instruments and shows that IMS can be readily applied for isomer separations, as shown in this study for fucose-containing isomers.
The results of the isomer pairs LeA/BG H1 and LeX/BG H2 obtained here fully agree with those of previous reports. LeA/BG H1 as well as LeX/BG H2 yield very similar CCSs as protonated ions and the ATDs overlap perfectly, which indicates migration into a similar structure. A similar behaviour can by hypothesized for protonated reduced glycans, which show very similar CCSs that are well within the error of the measurement (relative standard deviation (RSD) of 0.5%).45 However, based on the present data a clear conclusion cannot be drawn. In strong contrast, most of the metal adducts differ substantially in CCS, which clearly contradicts a rearrangement reaction. Labelling with 2-AA, 2-AB and ProA not only changes the UV and fluorescence activity of the glycan, but also introduces apparent basic sites, which reduces or inhibits proton mobility. As a result, different structures leading to distinct CCSs are retained. Fluorescence labelling can therefore not only help to increase isomer separation in IMS as shown above, but can also inhibit fucose migration in protonated glycan ions.
Seen from a broader perspective, the presented data show the great potential of an LC-IM-MS coupling for glycomics. Both methods have previously shown their individual strengths and weaknesses in glycan analysis. LC can resolve and quantify certain isomers and retention indices (i.e. glucose units46) can be used for the structural identification of known components. IM-MS on the other hand is more sensitive and can also resolve isomers with an amphiphilic character such as synthetic glycans or glycolipids, which due to their mixed polarity can often not be separated by LC.11 In addition, fragmentation and subsequent IMS analysis enables the rapid identification of unknown components based on database CCSs of small fragments.12,13,28 Regarding time scale, LC and IM-MS are furthermore highly complementary and data can be obtained simultaneously on a high-throughput scale.47,48 A combination of LC and IM-MS is therefore highly synergistic and more than the sum of its parts. When combined with suitable software tools to annotate tandem MS spectra and calculate glucose units49 and CCSs,50 LC-IM-MS has the potential to serve as the future core technology in glycomics.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9an00937j |
This journal is © The Royal Society of Chemistry 2019 |