Assessing the Use of Host-Guest Chemistry in Conjunction with Cyclic Ion Mobility Separations for the Linkage-Specific Characterization of Human Milk Oligosaccharides

Abstract

Human milk oligosaccharides (HMOs) are a class of glycans that are highly abundant in human milk and contribute to the healthy growth of an infant’s immune system. While new advancements in analytical methodologies have been made in glycomics, the high degree of isomeric heterogeneity and lack of authentic standards have made the high-resolution separation and accurate characterization of linkage positioning of all HMO species very challenging. Herein, we present an evaluation of the use of host-guest chemistry in conjunction with cyclic ion mobility spectrometry-mass spectrometry (cIMS-MS)-based separations for the identification of linkage positioning in three pairs of di-, tetra-, and hexasaccharide HMO isomers that only differ in the positioning of one glycosidic linkage (β1,3 versus β1,4). Suitable hosts, such as α/β cyclodextrins, cucurbit[n]urils (n = 5, 7), crown ethers, cyclic peptides, and an ionophore, were used to assess host-guest inclusion complex formation as well as linkage-specific cIMS-MS trends. Our results indicated a linkage-specific trend for the [M + 2α + 2H]²⁺ cyclodextrin-based host-guest inclusion complexes where the β1,3 linkage-containing isomers were always higher mobility than the β1,4 linkage-containing ones as well one for the [M + α + β + 2H]²⁺ complexes where the β1,4 linkage-containing isomers were always higher mobility than the β1,3 linkage-containing ones. We also observed diagnostic mobility fingerprints for the cucurbituril-based complexes. We anticipate that linkage-specific and mobility fingerprint trends can potentially aid in identifying linkage positioning for other HMO isomers as well as in complex human milk samples.

Graphical Abstract

1. Introduction

Human milk oligosaccharides (HMOs) encompass a class of unconjugated glycans and have several biological functions, primarily for infant development in the gut microbiome.^1–6 Although infants cannot digest HMOs, studies have shown that HMOs modulate and enhance the infant’s immune system by improving brain development, serving as prebiotic metabolic substrates, interfering with harmful pathogens (e.g., bacteria and viruses), and offering protection against fatal (e.g., necrotizing enterocolitis) and atopic (e.g., asthma, eczema, allergies) disorders.^{1–4, 7–10} HMO structures are comprised of the following five monosaccharide building blocks: D-glucose (Glc), D-galactose (Gal), N-acetyl glucosamine (GlcNAc), L-fucose (Fuc), and N-acetylneuraminic acid (Neu5Ac; Sia) as the prevalent sialic acid in humans.^1–5 The HMO core is comprised of lactose, Gal(β1,4)Glc, and can be elongated in a linear or branched manner by the addition of acetylated, fucosylated, or sialylated HMOs.^{1, 5, 7, 11} With over 200 uniquely identified structures ranging from 3 to over 20 monosaccharides,^{5, 7, 12} HMOs are difficult to characterize due to their structural complexity, variable glycosidic linkages/anomericities, and unavailability of authentic high-purity analytical standards.^{1, 5–7, 13}

Liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) remains the ‘gold standard’ for HMO analyses.^14–16 However, such approaches are often limited by insufficient LC resolution which leads to difficulty deconvoluting MS/MS spectra using collision-induced dissociation for co-eluting isomeric species, thus precluding the characterization of all key HMOs.^17–19 It is important to mention that other dissociation methods (e.g., ultraviolet photodissociation, electron capture/induced/transfer dissociation, and charge transfer dissociation) have shown some promise for delineating isomeric carbohydrates, but still remain limited by the necessary chromatographic resolution to sample individual HMO species.^20–25 Thus, new complementary and orthogonal analytical separations methods are essential for the rapid and accurate characterization of HMOs. Ion mobility spectrometry coupled to mass spectrometry (IMS-MS) separates ions in the gas phase under the presence of an electric field based on their 3-D structures/shapes and charge (i.e., mobilities).^26–30 In recent years, IMS-MS has seen widespread use in omics-based research for the resolution of isobaric/isomeric components and provides a rapid, orthogonal, alternative to condensed-phase methods.^26–30 Unfortunately, standalone IMS-MS approaches have seen little use in glycomics due to the limited resolution of existing instrumentation to resolve the many possible isomeric species present (i.e., isomeric heterogeneity bottleneck).^{31, 32} It is noteworthy to mention that coupling IMS-MS-based separations to cryogenic infrared spectroscopy has shown promise for distinguishing very structurally similar glycan species.^33–36 Recent developments in IMS-MS technology (e.g., trapped IMS, TIMS; structures for lossless ion manipulations, SLIM; and cyclic IMS, cIMS) have enabled the higher-resolution separations of important isomeric biomolecules (e.g., lipids, isotopomers, D-amino acid containing peptides, glycans, etc.).^36–52

Recently, our group has reported on the cyclic ion mobility spectrometry-mass spectrometry (cIMS-MS)-based separations of key disaccharide and trisaccharide HMO building block isomers.⁴¹ These isomers represent the smaller subunits that comprise larger HMOs consisting of various anomers, linkage positions, and branched/elongated structures. While our prior study did demonstrate that these building block isomers could be resolved with cIMS-MS, as well as further characterized with cIMS/cIMS and cIMS-MS/MS, we did not observe any linkage position-specific trends in their separation (e.g., an α2,3 linkage would always be higher mobility than an α2,6 linkage or one linkage displaying more cIMS peaks than another). Thus, new methodologies are needed to tease out the subtle differences amongst linkage positioning present in key HMO isomers.

Host-guest chemistry, or the formation of host-guest inclusion complexes, refers to non-covalent interactions between a host molecule and the analyte(s) of interest (i.e., guest(s)).^53–56 Host-guest chemistry has seen use for providing diagnostic fragment ions in MS/MS, improving solubility in drug delivery, as well as enabling the separation of biologically-relevant isomers in IMS-MS based on the formation of host-guest inclusion complexes that have different structures/shapes and thus mobilities.^{38, 42, 53, 56–59} However, to the best of our knowledge, the use of host-guest chemistry has never been applied to assess linkage-specific trends in the analyses of HMOs with high-resolution cyclic ion mobility spectrometry-mass spectrometry (cIMS-MS)-based separations. Herein, we present a systematic investigation on the use of host-guest inclusion complexes in cIMS-MS to distinguish between a single glycosidic linkage (β1,3 versus β1,4) in pairs of di-, tetra-, and hexasaccharide HMO isomers. As seen in Figure 1, each of these isomer pairs only structural differences are either a β1,3 or a β1,4 glycosidic linkage between a galactose and N-acetyl glucosamine monosaccharide moiety at the non-reducing end. For the selection of potential hosts for the analyses of these HMO linkage isomer guests, we chose various cyclic molecules that have seen previous utility in the formation of host-guest inclusion complexes (e.g., cyclodextrins, cucurbiturils, crown ethers, cyclic peptides, and an ionophore) as shown in Figure 2.^{38, 42, 54, 55, 57, 58, 60–63} We specifically chose these cyclodextrins because of their amine group, which could potentially aid in the analyses of host-guest inclusion complexes in positive ion mode; unfortunately the amine version of gamma cyclodextrin is not commercially available. We also note that these cucurbiturils were selected based on their solubility in water, whereas the [6] and [8] versions have much poorer solubility.^{53, 55} Our overall hypothesis was that certain host-guest inclusion complexes may enable diagnostic linkage-specific trends to be observed with cIMS-MS. Such trends could either be based on the relative arrival time order observed (e.g., β1,3 linkage host-guest inclusion complex being higher mobility than a β1,4 one) or in terms of mobility fingerprints (e.g., β1,4 linkage host-guest inclusion complex displaying greater or fewer number of cIMS-MS peaks than β1,3 ones). Overall, our motivation was to assess whether the use of host-guest chemistry would provide any additional information when comparing to cIMS-MS-based separations of the commonly used adducts (e.g., protonated and sodiated). If so, we would envision creating a database of the information obtained from each set of experiments to help identify unknown HMOs in future experiments.

Figure 1.

Structures of the monosaccharide constituents in the HMO linkage isomers studied as well as the structural difference between a β1,3 and β1,4 linkage (A). Symbolic nomenclature of the HMO linkage isomers studied (B).⁶⁴

Figure 2.

Structures of the cyclic hosts.

2. Materials and Methods

2.1. Reagents and Sample Preparation

Human milk oligosaccharide standards were purchased from Carbosynth (Berkshire, UK): lacto-N-biose (LNB), N-acetyl lactosamine (LacNAc), lacto-N-tetraose (LNT), lacto-N-neotetraose (LNnT), para-Lacto-N-hexaose (pLNH), and para-Lacto-N-neohexaose (pLNnH). 3-amino 3-deoxy α-cyclodextrin (α-CD) and 3-amino 3-deoxy β-cyclodextrin (β-CD) were purchased from Tokyo Chemical Industry (Tokyo, Japan). Cucurbit[5]uril (CB5), cucurbit[7]uril (CB7), cyclosporin A, valinomycin, polymyxin B, and ammonium ionophore I were purchased from Sigma-Aldrich (Milwaukee, WI USA). 12-Crown-4, 15-Crown-5, and 18-Crown-6 were purchased from Alfa Aesar (Ward Hill, MA USA). LC-MS grade solvents were obtained from Fisher Scientific (Pittsburgh, PA USA). Stock solutions for all HMOs, cyclodextrins, cucurbiturils, crown ethers, and polymyxin B were dissolved in 100% H₂O, while stock solutions for cyclosporin A, valinomycin, and ammonium ionophore I were dissolved in 100% methanol. Each host-guest solution mixture was prepared to 5 μM in 49.75/49.75/0.5 (v/v) water/methanol/acetic acid for each reagent used (i.e., both the host and guest were at 5 μM). Each HMO pair was also run as a mixture (see the solid black traces in Figures S1–S3 of the Supporting Information).

2.2. Cyclic IMS-MS Conditions

All experiments were performed on a commercially-available cIMS−MS instrument (Waters Corporation; Wilmslow, UK).⁶⁵ Each host-guest solution was subjected to direct infusion via positive mode electrospray ionization at a voltage of 2.5 kV in positive ion mode at a flow rate of 2 μL/min. Each host-guest inclusion complex that was observed in the mass spectrum was m/z selected with a quadrupole and then subjected to cIMS-MS separations in nitrogen buffer gas operating at 1.74 mbar before being routed to the time-of-flight (TOF) operated in ‘V’ mode for detection. Ions were cycled around the separation region multiple times until the “wrap-around” effect was observed, where higher mobility species overtake lower mobility ones. We also note that for the 3 m separations, the signal intensity was too low to enable ions to be subjected to further cycles. It should be mentioned that slicing specific mobility regions can overcome the wrap-around effect; however, this work focused on assessing linkage-specific trends amongst the various host-guest inclusion complexes rather than enhancing resolution. Traveling wave conditions were optimized for each host-guest inclusion complex as well as the individually run HMO isomers. Signal averaging for all arrival time distributions was done for 3 min without any additional smoothing. All data acquisition and processing was performed using MassLynx and Quartz software.

3. Results and Discussion

3.1. Host-Guest Inclusion Complex Formation

Initially, we subjected each host-guest pair shown in Figures 1 and 2 to direct infusion positive mode ESI-MS to determine if these non-covalent complexes would even form, let alone form with analytically relevant formation efficiencies (i.e., sufficient signal intensity to permit cIMS separations). Additionally, we made sure to survey all possible host-guest inclusion complex ratios (e.g., 1:1 or 2:1) as well as possible adduct formation (e.g., Na⁺ or K⁺). We observed host-guest inclusion complex formation for both the cucurbituril and cyclodextrin hosts, but unfortunately did not observe formation for the crown ethers, cyclic peptides, or for the ionophore. We hypothesize that the wider and deeper cavity sizes of the cucurbituril and cyclodextrin hosts enable non-covalent complex formation with the HMO linkage isomer guests, while the smaller cavity sizes for the other hosts assessed did not yield complex formation.^{53, 55, 66} For example, the inner diameter cavity size is 4.7–5.3 Å for α-cyclodextrins, 6.0–6.5 Å for β-cyclodextrins, 2.4 Å for cucurbit[5]uril, and 5.4 Å for cucurbit[7]uril.^{55, 60} The cavity sizes of 12-crown-4, 15-crown-5, and 18-crown-6 are 1.2–1.5, 1.7–2.2, and 2.6–3.2 Å, respectively.⁵⁵

Figure 3 highlights the mass spectra for the cyclodextrin and cucurbituril host-guest inclusion complexes with the tetrasaccharide HMOs. These were collected exclusively in “pass-through” mode, where no cIMS separation occurs. Additionally, we would like to note that ESI conditions were optimized for the sensitivity of the complexes, which could potentially affect the signal intensities of the other adducts (e.g., protonated or sodiated). In Figure 3, it is observed that the desired complexes were detected with analytically-relevant formation efficiency for both the cyclodextrin and cucurbituril hosts; however poorer intensity was seen with CB5. Of note, we observed the cyclodextrin hosts to form doubly charged protonated homo- and heterodimers and the cucurbiturils to complex with both a proton and potassium. Based on the formation of these complexes, we subjected each one to cIMS-MS-based separations. We ensured that the cIMS-MS parameters were tuned not only for resolution but also to avoid any unwanted fragmentation/activation. We also note that the cucurbiturils, cyclodextrins, and HMO isomers are all highly water soluble. For their exact masses, please see Table S1 in the Supporting Information.

Figure 3.

Mass spectra of the cyclodextrin (A) and cucurbituril (B) host-guest inclusion complexes with a mixture of LNT and LNnT. HMOs. A 1:1 mixture of each cyclodextrin and each cucurbituril was used in these experiments (i.e., 5 μm for both). ESI parameters were optimized for sensitivity of the complexes and not those of the common sodiated or protonated adducts. It is noted that this data was collected in “pass-through” mode, where no cIMS separation occurs.

3.2. Cyclodextrin-Based Complexes

For all the cyclodextrin-based host-guest inclusion complexes, we observed a 2:1 host-to-guest ratio for all HMO species assessed (as shown in Figure 3), which has also been observed for other analyte species in previous studies.⁵⁹ Based on previous literature, we speculate that each HMO species is “sandwiched” between the two cyclodextrin hosts molecules.⁶⁰ We note that no 1:1 ratio was observed in our experiments, thus indicating that two cyclodextrin hosts are required for complex formation with HMO guests. This was also further validated by detecting the doubly protonated homo- and hetero-dimers of the cyclodextrins as shown in Figure 3. We also note that no sodiated or potassiated complexes were observed; we only detected doubly protonated ones with the proton locations presumably on the amine groups of either the cyclodextrins or the HMOs. Interestingly, since both α and β cyclodextrins were run in the same mixture, three different doubly charged host-guest inclusion complexes were observed: [M + 2α + 2H]²⁺ with two α-CD, [M + α + β + 2H]²⁺ with one α-CD and one β-CD, and [M + 2β + 2H]²⁺ with two β-CD, where M is the HMO isomer. Future work involving theoretical modeling is needed to understand the formation of such complexes. Our separation conditions were optimized for each host-guest inclusion complex, with the experimental details presented in Table 1.

Table 1.

Traveling wave conditions and m/z values for the cyclodextrin-based host-guest inclusion complexes.

HMO species	Host-guest complex	m/z	TW conditions	Pathlength
Disaccharides	[M + 2α + 2H]2+	1163.8	375 m/s at 25 V	5 m
	[M + α + β + 2H]2+	1244.8	375 m/s at 25 V	5 m
	[M + 2β + 2H]2+	1325.8	375 m/s at 25 V	5 m
Tetrasaccharides	[M + 2α + 2H]2+	1325.9	350 m/s at 25 V	5 m
	[M + α + β + 2H]2+	1406.9	350 m/s at 25 V	5 m
	[M + 2β + 2H]2+	1487.9	350 m/s at 25 V	5 m
Hexasaccharides	[M + 2α + 2H]2+	1508.5	450 m/s at 30 V	3 m
	[M + α + β + 2H]2+	1589.5	450 m/s at 30 V	5 m
	[M + 2β + 2H]2+	1670.5	450 m/s at 30 V	3 m

Figure 4 illustrates the cIMS-MS separations for all cyclodextrin-based host-guest inclusion complexes as their individual HMO species; please refer to Figures S1–S3 in the Supporting Information for the arrival time distributions of the isomeric mixtures. We reiterate that each HMO isomer pair only differs in one glycosidic linkage (β1,3 versus β1,4) as highlighted in Figure 1 and the legend in Figure 4. For the [M + 2α + 2H]²⁺ complexes (Figures 4A–C), the disaccharide isomers (Figure 4A) were nearly baseline resolved after 5 m of separation with the β1,3 linkage-containing disaccharide being higher mobility than that of its β1,4 counterpart. We also observed two cIMS peaks for each individual disaccharide isomer, potentially from their α/β anomers.^{43, 67} For the tetrasaccharides (Figure 4B), we observed lower resolution than that of the disaccharide isomers, but their arrival time order matched that for the disaccharides (i.e., the major feature of the β1,3 linkage-containing LNT was higher mobility than the major feature for the β1,4 linkage-containing LNnT). In Figure 4C, a similar trend was observed for the hexasaccharides where the pLNH with a β1,3 linkage was higher mobility than pLNnH with a β1,4 linkage. We do note that the isomeric resolution (see Figures S1–S3 in the Supporting Information) for the tetra- and hexasaccharides as their [M + 2α + 2H]²⁺ complexes was rather poor, thus potentially precluding the ability to resolve these in a biological sample. We do note that the presence or absence of cIMS peaks (e.g., multiple peaks present for pLNH but not for pLNnH) could be a diagnostic mobility fingerprint that may enable unknown identification, but the poorer signal intensity of pLNH could preclude analyses of a complex HMO mixture. Nonetheless, we observed an overall arrival time order trend of these HMO linkage isomers as their [M + 2α + 2H]²⁺ complexes where the major features for the β1,3 linkage-containing HMO isomers were higher in mobility than the β1,4 linkage-containing ones. We attribute these multiple peaks to either be from their α/β anomers or also from the possibility of having different orientations/conformations of the host-guest inclusion complexes. Molecular dynamics-based theoretical modeling would be required to better understand this and is a subject for future work.

Figure 4.

cIMS-MS separations of HMO linkage isomers as their [M + 2α + 2H]²⁺ (A–C), [M + α + β + 2H]²⁺ (D–F), and [M + 2β + 2H]²⁺ (G–I) host-guest inclusion complexes. Please see Table 1 for the TW and separation pathlengths used for each complex. Both the cyclodextrin hosts (α and β) and HMO guests were at concentrations of 5 μM. All arrival time distributions are relative to the intensity of the most abundant species.

For the [M + α + β + 2H]²⁺ complexes (Figures 4D–F), improved resolution was observed as compared to the [M + 2α + 2H]²⁺ complexes for each of the three HMO linkage isomer pairs. Of note, the arrival time order was flipped for these host-guest inclusion complexes, where the β1,4 linkage-containing species were higher mobility than the β1,3 linkage-containing ones. This indicates that an orthogonal approach that couples cIMS-MS separation information from both the [M + α + β + 2H]²⁺ and the [M + 2α + 2H]²⁺ host-guest inclusion complexes could resolve each of the HMO isomer pairs assessed in this study. We once again note that the presence of multiple cIMS peaks can also act as a method to distinguish amongst some species as their [M + α + β + 2H]²⁺ complexes. Additionally, we once again observed poorer signal intensity for pLNH relative to pLNnH as their host-guest inclusion complexes. We hypothesize that the structure of the pLNH hexasaccharide could preclude efficient complex formation as evident by its lower signal intensity. We also envision that the use of machine learning or chemometrics could enable an automated approach to match these fingerprints in an unknown sample. Additionally, we hypothesize these multiple peaks could be from their anomers or other conformations for the host-guest inclusion complexes. Lastly, for the [M + 2β + 2H]²⁺ complexes highlighted in Figures 4G–I, we observed poorer isomer resolution, but distinct mobility fingerprints in some cases (e.g., two cIMS peaks for LNB but only one for LacNAc). We also observed a similar arrival time order trend for the [M + 2β + 2H]²⁺ complexes as for the [M + 2α + 2H]²⁺ ones, where the β1,3 linkage-containing tetrasaccharides were higher mobility than the β1,4 linkage-containing ones. Clearly future work, potentially involving theoretical modeling through molecular dynamics, is needed to understand the arrival time order flip for the [M + α + β + 2H]²⁺ host-guest inclusion complexes. We hypothesize that this trend could be due to the combination of two different cyclodextrins, and thus two different sized cavities for the host molecules, but would need to be further verified through molecular modeling. Globally, we note that improved isomeric resolution was observed for the disaccharide HMO linkage isomers as compared to the tetra/hexasaccharides, regardless of host size. For the unresolved species, this indicates that the host-guest inclusion complexes have similar mobilities, and thus very similar gas-phase structures/shapes. This potentially implies that the cavity size of the cyclodextrins is not sufficient enough to improve isomeric resolution as the guest size increases; however, future studies are needed to determine the scope and limitations.

3.3. Cucurbituril-Based Complexes

As previously mentioned, cucurbit[n]uril complexes where n = 5, 7 were chosen due to their solubilities in water,^{55, 61–63} which could potentially enable coupling of host-guest inclusion complexation with post-column addition chromatographic strategies. Despite different binding properties, 3D structures, and composition, cucurbiturils have often been compared to cyclodextrins due to their similar cavity sizes.^{53, 55, 68} Surprisingly, exclusively 1:1 host-to-guest ratios were observed for the cucurbituril-based inclusion complexes. Specifically, doubly charged host-guest inclusion complexes were observed as their [M + CB5 + H + K]²⁺ and [M + CB7 + H + K]²⁺ forms. The incorporation of both a proton and potassium is noteworthy to observe since no added metal salts were used in our sample preparation; however, this is unsurprising given that previous literature has seen similar proton-potassium bound doubly charged complexes in the analyses of carbohydrates with IMS-MS, as well as our observation of this adduct with the host alone in Figure 3.^{31, 37} Presumably, the potassium ion is present in our solvents. We optimized our cIMS-MS conditions for each host-guest inclusion complex, with the experimental conditions shown in Table 2.

Table 2.

Traveling wave conditions and m/z values for the cucurbit[n]uril-based host-guest inclusion complexes.

HMO species	Host-guest complex	m/z	TW conditions	Pathlength
Disaccharides	[M + CB5 + H + K]2+	626.6	400 m/s at 20 V	3 m
	[M + CB7 + H + K]2+	792.7	400 m/s at 22 V	3 m
Tetrasaccharides	[M + CB5 + H + K]2+	788.7	450 m/s at 25 V	5 m
	[M + CB7 + H + K]2+	954.8	450 m/s at 25 V	5 m
Hexasaccharides	[M + CB5 + H + K]2+	971.2	450 m/s at 25 V	5 m
	[M + CB7 + H + K]2+	1137.3	450 m/s at 25 V	5 m

Figure 5 depicts the cIMS-MS separations for the cucurbituril-based host-guest inclusion complexes. For the disaccharide HMOs, we observed no resolution for the linkage isomers with the CB5 host, but did observe a diagnostic fingerprint with the CB7 host (i.e., 3 peaks for LacNAc and only 2 for LNB). Of note, the intensity of the LacNAc was significantly lower than that of LNB with both CB hosts; please see the zoomed in versions of LacNAc run individually in Figure 5. The presence of three peaks could be attributed to the α/β anomers of the HMOs as well as the possibility for multiple conformations of the host-guest inclusion complexes. For the tetrasaccharide linkage isomers, improved resolution was seen with the CB5 host as compared to the CB7 one. Additionally, a greater number of cIMS peaks were observed for the LNT-based host-guest inclusion complexes as compared to LNnT (i.e., 2 peaks versus 1 with CB5 and 3 peaks versus 2 with CB7). For the hexasaccharides, a similar trend was observed as with the tetrasaccharides: improved resolution with the CB5 host and the β1,3-linkage containing isomer (i.e., pLNH) displaying a greater number of cIMS peaks with both the CB5 and CB7 hosts as compared to the β1,4 linkage-containing isomer (i.e., pLNnH). In comparing our results between the cyclodextrin and cucurbituril-based host-guest inclusion complexes, the tetra- and hexasaccharide linkage isomers displayed the best cIMS resolution with the CB5 hosts, while the disaccharide linkage isomers were best resolved with the [M + α + β + 2H]²⁺ complex. Additionally, while there were linkage-specific trends observed with the cyclodextrin-based complexes (e.g., β1,4 linkage-containing species were higher mobility than the β1,3 linkage-containing ones for the [M + α + β + 2H]²⁺ complexes), no such trends were globally observed for the cucurbituril-based host-guest inclusion complexes. Thus, we envision that both cucurbiturils and cyclodextrins could potentially have broad utility in aiding the identification of unknown HMO linkage isomers, especially ones that only differ amongst 1 linkage such as in our results (i.e., β1,3 versus β1,4 at the non-reducing end). Future work is also clearly needed to see if our presented methodology using host-guest inclusion complexes can be applied to other molecule classes (e.g., identification of double bond positioning in lipids) as well as for other key linkages in HMOs (e.g., α2,3 versus α2,6 in sialylated ones).

Figure 5.

cIMS-MS separations of HMO linkage isomers as their [M + CB5 + H + K]²⁺ (A–C) and their [M + CB7 + H + K]²⁺ (D–F) host-guest inclusion complexes. Zoomed-in cIMS-MS separations of LacNAc with both the CB5 and CB7 hosts (G, H). Please see Table 2 for the TW conditions and separation pathlengths used for each complex. Both the cucurbituril hosts (CB5 and CB7) and HMO guests were at concentrations of 5 μM. All arrival time distributions are relative to the intensity of the most abundant species.

3.4. Commonly Used Adducts

In order to determine if the use of cyclodextrin and cucurbituril-based host-guest inclusion complexes do indeed provide additional and new information, we were interested in comparing those results with ones using individual adducts (e.g., H⁺, Na⁺, K⁺). Specifically, we wanted to assess the separation of each of the three pairs of HMO linkage isomers as their [M + Na]⁺ adducts since sodium adduction is widely used in IMS-MS analyses of carbohydrates, as well as [M + H]⁺ and [M + H + K]²⁺ adducts since they were similar to the ones used in our host-guest inclusion complexes. Table 3 shows the optimized conditions for each of the adducts and Figure 6 shows their respective cIMS-MS separations. We do note that the [M + H + K]²⁺ adduct was not observed for the disaccharide linkage isomers; we hypothesize that their smaller overall size precludes the ability for this doubly charged complex to form.

Table 3.

Traveling wave conditions and m/z values for the [M + H]⁺, [M + Na]⁺, and [M + H + K]⁺ adducts.

HMO species	Adducts	m/z	TW conditions	Pathlength
Disaccharides	[M + H]+	384.1	350 m/s at 17 V	5 m
	[M + Na]+	406.1	350 m/s at 17 V	5 m
	[M + H + K]2+	211.5	N/A	N/A
Tetrasaccharides	[M + H]+	708.2	350 m/s at 25 V	3 m
	[M + Na]+	730.2	350 m/s at 25 V	5 m
	[M + H + K]2+	373.6	700 m/s at 20 V	5 m
Hexasaccharides	[M + H]+	1073.3	450 m/s at 30 V	3 m
	[M + Na]+	1095.3	450 m/s at 35 V	5 m
	[M + H + K]2+	556.1	500 m/s at 20 V	5 m

Figure 6.

cIMS-MS separations of HMO linkage isomers as their [M + H]⁺ (A–C), [M + Na]⁺ (D–F), [M + H + K]²⁺ (H–I) adducts. Please see Table 3 for the TW conditions and separation pathlengths used for each complex. Note: the [M + H + K]⁺ adduct was not observed for the HMO disaccharides. All arrival time distributions are relative to the intensity of the most abundant species.

For the protonated adducts (Figure 6A–C), all three pairs of HMO linkage isomers were globally better resolved as compared to both the cyclodextrin and cucurbituril-based host-guest inclusion complexes. However, no distinct mobility fingerprints were observed for these [M + H]⁺ adducts (i.e., multiple cIMS peaks), as well as no obvious linkage-specific arrival time order trend. Specifically, for the protonated adducts, the β1,3 linkage-containing isomer was higher mobility in the di- and tetrasaccharides, but was lowest mobility in the hexasaccharides. Conversely, the β1,3 linkage-containing isomer was highest mobility in all cases for the [M + 2α + 2H]²⁺ complexes and lowest mobility in all cases for the [M + α + β + 2H]²⁺ complexes (i.e., β1,4 linkage-containing isomer was highest mobility). Poor resolution of the tetrasaccharides was observed for the sodiated adducts, while the disaccharide and hexasaccharide HMO linkage isomers were well resolved (Figures 6D–F). We speculate that at longer cIMS separation pathlengths, the sodiated tetrasaccharides may also begin to separate. Lastly, for the [M + H + K]²⁺ adduct, there was no resolution for the tetrasaccharide isomers (Figure 6H), though the hexasaccharide isomers (Figure 6I) were partially resolved. As previously mentioned, we did not observe this adduct for the disaccharide species, presumably from their smaller size not permitting adduction. Overall, when comparing the use of host-guest inclusion complexes versus commonly used adducts for the analyses of three pairs of HMO linkage isomers, we observed linkage-specific trends with the [M + 2α + 2H]²⁺ and [M + α + β + 2H]²⁺ complexes as well as mobility fingerprints with both cyclodextrin and cucurbituril-based complexes in other cases. In separate experiments, we assessed the relative mobilities for LNnT as its common adducts and cyclodextrin-based host-guest inclusion complexes as well as pLNH as its common adducts and cucurbituril-based host-guest inclusion complexes. To be able to compare the mobilities for each HMO species adducted or complexed, we made sure to use the same traveling wave conditions and subject each to a single cycle, 1 m, cIMS-MS separation (see Figure S4 in the Supporting Information). From these separations, we observed that the sodiated adduct for LNnT was higher mobility than its protonated one, which presumably can be attributed to sodium adduction causing structural compaction. Unsurprisingly, the doubly charged cyclodextrin-based host-guest inclusion complexes were similar in mobility (i.e., arrival time) to the sodiated and protonated adducts since charge greatly impacts the mobility of an ion. Additionally, we observed that as host size increased, mobility was lowered (i.e., the [M + 2α + 2H]²⁺ was fastest while the [M + 2β + 2H]²⁺ was slowest). For pLNH, we observed the sodiated adduct to be lower in mobility than the protonated one, while both of these commonly used adducts were lower in mobility than the doubly charged cucurbituril-based host-guest inclusion complexes. We envision future work that couples the information obtained from host-guest chemistry with that obtained from the protonated/sodiated adducts to be useful in analytical measurements. Specifically, our future goal is to develop a database of cIMS based collision cross section (CCS) values that will enable the better identification of unknown HMOs in human milk.

4. Conclusions

Herein, we have presented the use of host-guest inclusion complexes to assess the separation of select pairs of di-, tetra-, and hexasaccharide HMO linkage isomers that vary in only a single (β1,3 versus β1,4) glycosidic linkage by cIMS-MS-based separations. To the best of our knowledge, this is the first ever demonstration of coupling cIMS-MS separations with host-guest chemistry for HMO analyses. With the hosts that were chosen, we observed complex formation for cyclodextrin and cucurbituril hosts, but none for crown ethers, cyclic peptides, and ammonium ionophore. For the cyclodextrin hosts, the [M + 2α + 2H]²⁺ host-guest inclusion complex exhibited a linkage-specific trend (i.e., the β1,3 linkage-containing isomer was always higher mobility than its β1,4 counterpart). A linkage-specific trend was also observed for the [M + α + β + 2H]²⁺ complexes, where the β1,4 linkage-containing isomer was always higher mobility than its β1,3 counterpart. For the cucurbiturils, no such linkage-specific trends were observed. However, diagnostic mobility fingerprints were observed for disaccharides (i.e., LNB, LacNAc) with the [M + 2α + 2H]²⁺ and CB7 inclusion complexes, tetrasaccharides (i.e., LNT, LNnT) with CB5 and CB7 complexes, and hexasaccharides (i.e., pLNH, pLNnH) with [M + α + β + 2H]²⁺, CB5, and CB7 complexes where multiple peaks were observed for several of the isomeric pairs. Potentially the use of machine learning or chemometrics could provide an automated approach to identify these mobility fingerprints in unknown samples. In comparing the results with the host-guest inclusion complexes versus the commonly used adducts, we observed varying arrival time order trends. We believe this demonstrates the orthogonality of information that can be obtained when combining data from both experiments. Ultimately, we envision that by developing a database of calculated cIMS-based collision cross section (CCS) values for each complex and adduct assessed, we could potentially enable the better identification of unknown HMOs in the complex matrix that is human milk. We hypothesize that our observed results are due to the varying cavity sizes of the host molecules enabling non-covalent complexation, potentially through hydrogen bonding. Clearly, theoretical modeling through molecular dynamics is an avenue of future direction to better understand the formation of these host-guest inclusion complexes. Overall, this work highlights how host-guest chemistry can enable additional information to be obtained that is otherwise not present in the analyses of HMOs as their commonly used protonated or sodiated adducts. This demonstrates how host-guest chemistry can be an added piece to the analytical toolbox related to the ultimate goal of characterizing all possible HMOs in the human milk glycome. In the context of an LC-cIMS-MS experiment, we would plan on adding the host molecules through a post-column addition strategy prior to cIMS-MS.⁶⁹ Additionally, we envision that such an approach can also have broad utility toward other molecular classes (e.g., lipids, steroids, bile acids, etc.) as well as for other HMO species (e.g., fucosylated and sialylated ones). We would like to note that the graphical abstract is a cartoon depiction and is not at all intended to display topology of the host-guest inclusion complexes.

Highlights for: Assessing the Use of Host-Guest Chemistry in Conjunction with Cyclic Ion Mobility Separations for the Linkage-Specific Characterization of Human Milk Oligosaccharides.

• Human milk oligosaccharides are challenging to characterize through conventional analytical approaches

• Non-covalent complex formation between human milk oligosaccharide isomers and cyclodextrins and cucurbiturils was observed

• Host-guest chemistry coupled with cyclic ion mobility separations enables linkage-specific trends in human milk oligosaccharides