Introduction.

Glycans play a critical role in many biological functions because of their structural and modulatory properties and their specific interaction with lectins [1, 2]. Obtaining the intact structural information of glycans is vital to understanding these biological functions. Mass spectrometry (MS) is applied in the structural analysis of glycans because of its high sensitivity [3]. MS requires a small amount of sample compared with other methods (e.g., nuclear magnetic resonance spectroscopy). Since their introduction nearly three decades ago, ultraviolet matrix–assisted laser desorption/ionization (UV-MALDI) [4, 5] and electrospray ionization (ESI) [6, 7] have been the main ionization methods used in MS for biomolecules. However, the applications of MALDI to glycans have not been as successful as they have been to proteins [8] mainly because glycans have low ionization efficiency [9] and easily fragments in MS [10, 11]. Fragmentation among various matrices used in MALDI has been extensively and comparatively investigated [12,13,14,15,16,17,18,19]. Matrix α-cyano-4-hydroxybenzoic acid (CHCA) tends to produce more fragments than 2,5-dihydroxybenzoic acid (2,5-DHB) in post-source decay (PSD) [15,16,17,18]. By contrast, the CHCA matrix produces fewer fragments in in-source decay (ISD) [19]. Although conventional MALDI is considered a type of soft ionization, studies have shown that fragile molecules, such as glycans containing sialic acid, fucose residues, and sulfate and phosphate functional groups, easily dissociate into fragment ions during the MALDI process [10].

In recent years, many research groups have proposed modified methods or new matrices to reduce the ISD and PSD in MALDI as well as to increase the sensitivity of intact molecules. Huang et al. [20] combined surface-assisted laser desorption/ionization (SALDI) with HgTe nanostructures and revealed higher sensitivity and less fragmentation for large molecular weight carbohydrates, including pullulan 6k, pullulan 10k, and dextran 12k, compared to the conventional MALDI matrix THAP. Wu et al. [21] used diamond nanoparticle (DNP) trilayer samples to enhance the ion abundance of carbohydrates. DNPs were incorporated into MALDI samples to optimize the sample morphology and to stabilize the samples for thermally labile compounds. The signal of dextran 1.5k obtained from trilayer samples was approximately 79 times stronger than that obtained from conventional dried-droplet MALDI methods.

Frozen solution has been used in MALDI to reduce the fragmentation of DNA, proteins, and peptides in previous studies [22,23,24]. Liang et al. [25] modified the conventional UV-MALDI method by using droplets of frozen solution for the study of oligosaccharides. Specifically, droplets of an aqueous acetonitrile solution containing oligosaccharides and 2,5-DHB were frozen at 100K for mass analysis using UV-MALDI. Compared with conventional UV-MALDI, droplets of frozen solution generate more oligosaccharide ions and fewer fragments from PSD. Furthermore, the ion signal from droplets of frozen solution lasts longer than that in conventional MALDI.

Gangliosides, which are glycosphingolipids with one or more sialic acid at the oligosaccharide chain, are one of the major constituents of plasma membrane. More than 180 gangliosides, differing in their positions and number of sialic acid residues, have been reported [26]. Gangliosides GM1, GD1, GT1, and GQ1 are related to cancer [27], Alzheimer’s disease [28, 29], and Parkinson’s disease [29]. The measurement of the relative abundance of GM1, GD1, GT1, and GQ1 is crucial for research on these diseases [30,31,32]. However, loss of one sialic acid in the ionization process changes the gangliosides from GM1, GD1, GT1, and GQ1 to asialo GM1, GM1, GD1, and GT1, respectively, which makes the quantitative analysis of these gangliosides difficult. In previous MALDI studies, gangliosides have been derivatized for signal enhancement and for reducing the loss of sialic acid [33, 34]. More recent studies have presented various new matrices [30, 35,36,37] and modified MALDI methods [38, 39] to reduce the loss of sialic acid for underivatized gangliosides.

Heparin is a heterogeneous sulfated glycosaminoglycan polymer and is used as an anticoagulant. Containing a sulfate functional group, heparin represents another category of glycans in which intact ions cannot be easily obtained using MS. Ionic liquid matrices for UV-MALDI [40] and water ice for IR-MALDI [41] have been reported to reduce the fragmentation of sulfated saccharides.

Oligosaccharides and polysaccharides have previously been used to investigate the softness of MALDI in droplets of frozen solution [25], trilayer [21], and dried-droplet HgTe samples [20], whereas gangliosides and sulfated saccharides have been used to examine the softness of the matrices, 5-methoxysalicylic acid (MSA) [37], and ionic liquid [40] in MALDI. Because different analytes and instruments were used in these studies, comparing the ionization softness of these matrices is difficult. In this study, these matrices were compared for their ionization softness by using the same instrument for the analytes, gangliosides, heparin, and pullulan.

Experiment

Materials

2,5-DHB, pullulan 10k (average molecular weight of approximately 10,000 Da), and heparin disaccharide I-S sodium salt (C12H15NO19S3Na4, MW = 665.40) were purchased from Sigma-Aldrich (St. Louis, MO, USA). MSA and butylamine were purchased from Acros Organics (Geel, Belgium). The gangliosides asialo GM1, GM1, GD1a, GT1b, and GQ1b were purchased from Cayman (Ann Arbor, MI, USA). DNPs were purchased from Element Six (Luxembourg City, Luxembourg). HgTe nanostructures were synthesized using the method described in a previous study [20].

Time-Of-Flight Mass Spectrometer

The mass spectrometer used in this study was a modified commercial MALDI-time of flight (MALDI-TOF; Autoflex III, Bruker Daltonik). The modification and experimental processes are described in our previous paper [25]. Briefly, the modifications involved (1) replacing the sample-holding plate with a vacuum gate valve, a sample preparation vacuum chamber, and a homemade sample holder and (2) addition of a temperature-controlled (100–300 K) cold finger in the main vacuum chamber. The cold finger, which was in thermal contact with the sample holder, was maintained at 100 and 300 K for the samples measured at 100 and 300 K, respectively. A third-harmonic (355 nm) laser beam from a built-in Nd:YAG laser operated at 20 Hz was used in this study. Heparin disaccharide I-S, ganglioside GD1a, and the mixture of gangliosides (asialo GM1, GM1, GD1a, GT1b, and GQ1b) were measured in reflectron mode at an accelerating voltage of ± 19 kV in positive and negative modes; pullulans—the molecular weights for which are beyond the mass range of the reflectron mode—were measured in linear mode at an accelerating voltage of 20 kV in the positive mode. Ions were accelerated after 380 ns of laser irradiation through delay extraction.

Sample Preparation

  1. (a)

    Droplets of Frozen Solution Sample

The droplets of frozen solution were prepared by mixing 2 μL of 2 M 2,5-DHB solution and 2 μL of analyte solution in a vial (Eppendorf, Germany). Then, 4 μL of the mixed solution was deposited onto the sample holder and was immediately sprayed with low-temperature N2 gas (approximately 100 K). After cooling the sample and part of the sample holder to 100 K, the sample preparation chamber was pumped down to a pressure of 1 × 10−6 Torr. The solid sample and sample holder were transferred to the main chamber before the melting of the sample. During the mass analysis, the sample holder was in thermal contact with a cold finger maintained at 100 K.

Various solvents were used to determine the optimal performance of the droplets of frozen solution in mass spectra. The solvents used in the 2,5-DHB solutions included 1:3 water and acetonitrile mixture, 1:3 water and methanol mixture, and 1:1 water and isopropanol mixture with 1% NH4OH. The pullulans and gangliosides were prepared in aqueous solutions. Heparin disaccharide I-S sodium salt was prepared in water, 1:3 water and methanol mixture, or 1:1 water and isopropanol mixture with 1% NH4OH. The matrix and analyte solutions were mixed to a final 2,5-DHB concentration of 1 M. In addition, the concentration of pullulan and ganglioside GD1a was 5 × 10−6 M, that of heparin I-S was 5 × 10−5 M, and that of the ganglioside mixture was 1 × 10−6 M.

  1. (b)

    Conventional MALDI Sample

A solid matrix (2,5-DHB or MSA) was dissolved in a solvent containing acetonitrile and water (3:1) to prepare a 0.1-M matrix stock solution. The analytes were dissolved in deionized water to generate 10−5 M (for pullulan and GD1a), 10−4 M (for heparin disaccharide I-S sodium salt), and 2 × 10−6 M (for ganglioside mixture) stock solutions. The mixture of 2-μL matrix stock solution and 2-μL analyte stock solution in a vial was deposited onto the sample holder into two separate spots and vacuum-dried in the sample preparation chamber before being transferred to a mass spectrometer for mass analysis.

  1. (c)

    DNP Trilayer Sample

The procedure for preparing DNPs and trilayer samples is as described in a previous study [21]. Briefly, DNP was heated in an oven at 400 °C for 2 h and washed with acetone several times until the solution became whitish. The DNP trilayer sample contained three layers. Every layer was vacuum-dried for 5–20 min immediately after deposition. The first layer was prepared by depositing 0.5-μL matrix solution on the sample holder. A conventional MALDI matrix, 2,5-DHB (0.1 M in a 1:3 water and acetonitrile mixture), or ionic liquid matrix, 2,5-dihydroxybenzoic acid butylamine (DHBB), was prepared by dissolving butylamine in a DHB solution with a concentration identical to that of DHB. The second layer was prepared by depositing 1-μL DNP solution (1–6 mg/mL) on top of the first dried layer. The third layer contained 0.5 μL of the analyte solution. The analyte solutions prepared for conventional MALDI samples were used here.

  1. (d)

    SALDI

HgTe nanostructures were used as a matrix for SALDI. The SALDI samples were prepared as described in a previous study [20]. Analyte solutions were prepared in ammonium citrate buffer solution (pH 9.0). The concentration of pullulan and ganglioside GD1a was 5 × 10−6 M and that of heparin I-S was 5 × 10−5 M. The ammonium citrate buffer solution was prepared by mixing 48 mg of citric acid and 70 μL of ammonium hydroxide solution (35%) in 430 μL of water. After the 1-μL analyte solution was mixed with 1-μL HgTe (where the synthesized solution was concentrated by a factor of three), the mixture was air-dried on a stainless sample plate.

Results and Discussion

  1. (a)

    Ganglioside GD1a

Ganglioside GD1a contains two sialic acids (Neu5Ac) (Figure 1a), which are eliminated easily during the MALDI process. The ganglioside GD1a used in this study was a mixture of two species, GD1a (d18:1/d18:0) and GD1a (d20:1/d18:0), which are different in terms of the length of fatty chains in the glycosphingolipid. Both species have two sialic acids and are annotated as M1 and M2, respectively. The corresponding molecular formulas are C84H148N4O39 (M1, MW = 1838.1 Da) and C86H152N4O39 (M2, MW = 1866.1 Da).

Figure 1
figure 1

Analytes used in this study: (a) Structures of ganglioside GD1a (M1 and M2), (b) symbol nomenclature (Nomenclature Committee Consortium for Functional Glycomics) of gangliosides asialo GM1a, GM1a, GD1a, GT1b, and GQ1b, (c) structure of heparin disaccharide I-S, and (d) structure of pullulan

Because gangliosides contain acidic glycans, negative ions are easily observed in the MALDI mass spectra. Figure 2a–f illustrates the mass spectra of gangliosides obtained from samples prepared using different matrices in negative ion mode. The intact ions at m/z 1837.07 ([M1-H]) and 1865.13 ([M2-H] ) were observed in the mass spectra of the droplets of frozen solution and practically, no fragment ions were found. By contrast, the mass spectra obtained from the remaining matrices showed both intact and fragment ions. Details of the assignments of mass spectra are listed in Table 1 of the supplementary material. These fragment ions mainly resulted from the loss of one sialic acid or one sialic acid and one galactose. The softness of the matrices derived from ionization/desorption occurred in the following sequence: droplets of frozen solution, trilayer with DHBB, trilayer with DHB, dried droplets of MSA, dried droplets of DHB, and dried droplets of HgTe sample. The mass spectra of ganglioside GD1a in positive ion mode, as illustrated in Fig. S1 of supplementary material, show the similar properties of these matrices.

Figure 2
figure 2

MALDI-TOF mass spectra of ganglioside GD1a obtained from samples prepared using different methods. Each spectrum was accumulated for 200 laser shots. (af) and (gl) are in negative and positive ion modes, respectively: (a) droplets of frozen solution 1 M 2,5-DHB in 37.5% acetonitrile aqueous solution (650 J/m2), (b) trilayer with DHBB (500 J/m2), (c) trilayer with DHB (270 J/m2), (d) dried-droplet DHB (270 J/m2), (e) dried-droplet MSA (210 J/m2), and (f) dried-droplet HgTe (130 J/m2). No fragment ion was observed in the mass spectra of the droplets of the frozen solution, indicating that droplet of frozen solution is the softest matrix

Because droplets of frozen solution samples provided the optimal results for GD1a, we further investigated the laser fluence dependence of GD1a. Figure 3 shows that the intensity of GD1a intact negative ions increased as the laser fluence increased. No fragmentation was observed except for low intensity of fragments corresponding to the loss of one sialic acid at a very large laser fluence (790 J/m2). The mass spectra of ganglioside GD1a in positive ion mode, as illustrated in Fig. S2 of the supplementary material, show similar property.

Figure 3
figure 3

Laser fluence dependence of ganglioside GD1a in MALDI-TOF mass spectra obtained from the droplets of frozen solution (1 M 2,5-DHB in 37.5% acetonitrile aqueous solution) in negative ion mode. Each spectrum was accumulated for 200 laser shots. No fragmentation was observed except for low intensity of fragments corresponding to the loss of one sialic acid at a very large laser fluence (790 J/m2)

  1. (b)

    Gangliosides Asialo GM1, GM1, GD1a, GT1b, and GQ1b

The relative abundance of gangliosides asialo GM1, GM1, GD1a, GT1b, and GQ1b is crucial in the research on various diseases [27,28,29]. Because the loss of one sialic acid from gangliosides changes GM1, GD1a, GT1b, and GQ1b to asialo GM1, GM1, GD1a, and GT1b, respectively, the loss of sialic acid during the conventional MALDI process makes the relative quantitative analysis of these gangliosides difficult. The MALDI mass spectra of gangliosides GT1b and GQ1b, which have three and four sialic acids, respectively, showed no elimination of sialic acid from the droplets of the frozen solution (Figures 4 a, b). No elimination of sialic acid from the droplets of frozen solution for GD1a (Figure 2), GT1b, and GQ1b (Figure 4) suggests that samples prepared using droplets of frozen solution are suitable for quantitative analysis of these gangliosides. To further investigate the capability of quantitative analysis using droplets of frozen solution, the mass spectra from the mixture of four gangliosides (GM1, GD1a, GT1b, and GQ1b) were measured. The relative ion abundances of these four gangliosides (Figure 4c) represent a combination of relative initial concentration and ionization efficiency. For the same concentration of GM1, GD1a, GT1b, and GQ1b (as shown by the orange bars in Figure 4c), GM1 showed the largest ionization efficiency, and GD1a and GT1b had similar ionization efficiencies, although they were lower than that of GM1. GQ1b had the lowest ionization efficiency. Figure 4d shows the relative ion intensity of GQ1b to GM1 as a function of different relative concentrations of the GQ1b and GM1 mixture. The linear relationship between the relative concentration and the relative ion intensity demonstrates the capability of gangliosides for quantitative analysis using droplets of frozen solution.

Figure 4
figure 4

Negative-ion MALDI-TOF mass spectra of (a) GT1b and (b) GQ1b obtained from droplets of frozen solution samples using 1 M 2,5-DHB as a matrix dissolved in 37.5% acetonitrile aqueous solution at a laser fluence of 650 J/m2. Each spectrum was accumulated for 100 laser shots. (c) Relative ion intensities of GM1, GD1a, GT1b, and GQ1b for different concentrations of the mixture. (d) Linear dynamic ranges of the relative intensities of GQ1b and GM1a are plotted as a function of relative concentration. Each data point represents an average of 600 laser shots, including six mass spectra and 100 laser shots for each spectrum

Elimination of sialic acid from droplets of frozen solution does not occur simply because of the low initial temperature. Frozen solvent molecules play a critical role in limiting the elimination of sialic acid from gangliosides. Figures 5 a and b present the mass spectra of the dried-droplet MALDI sample at an initial temperature of 100 K and the droplets of frozen solution at the same initial temperature for the same mixture of five gangliosides (asialo GM1, GM1, GD1a, GT1b, and GQ1b) in negative-ion mode. Severe fragmentation (loss of sialic acid) from the dried-droplet MALDI sample at 100 K was observed, indicating that the solvent (acetonitrile and water) is crucial in desorption and ionization. Both acetonitrile and water have low boiling and melting points compared with 2,5-DHB, which is critical for making gangliosides easy to be desorbed and ionized without losing sialic acid during the MALDI process. The mass spectra in positive ion mode, as illustrated in Fig. S3 of supplementary material, show similar property.

Figure 5
figure 5

MALDI-TOF mass spectra of the ganglioside mixture (asialo GM1, GM1, GD1a, GT1b, and GQ1b at equal concentrations) obtained from the droplets of frozen solution at 100 K using (a) 2,5-DHB as a matrix dissolved in 37.5% acetonitrile aqueous solution, and (b) the dried-droplet sample using 2,5-DHB at 100 K at a laser fluence 650 J/m2. Each spectrum was accumulated for 100 laser shots

  1. (c)

    Heparin Disaccharide I-S

Heparin represents another category of glycans for which intact ions are not easily obtained using MS. Heparin disaccharide I-S (Figure 1c) is the repeating unit in heparin. It contains three sulfate groups, which tend to be lost during the MALDI process. Figure 6 shows a comparison of heparin disaccharide I-S mass spectra obtained from the samples prepared using different matrices in negative ion mode. Only a few fragments corresponding to the loss of one or two sulfate groups were observed from the mass spectrum of the droplets of frozen solution when 50% isopropanol in water with 1% ammonium hydroxide was used as a solvent (Figure 6a); this is a commonly used solvent in ESI operated in negative ion mode. When the solvent of the frozen solution was changed to 75% methanol in water (which is a solvent with lower melting and boiling points), intact ion abundance increased, and practically no fragmentation was observed in the mass spectrum of the droplets of frozen solution sample (Figure 6b). The mass spectra of the trilayer with DHBB and DHB samples showed significant loss of one or two sulfate groups (Figure 6c, d), respectively. A very low intensity of intact ions was found in the dried-droplet DHB sample (Figure 6e), and no intact ions were observed in the dried-droplet MSA and HgTe samples (Figures 6f and g). Ionic liquid DHBB produced large fragments with loss of one or two sulfate groups (Figure 6h). The droplets of frozen solution sample with the proper choice of solvent yielded the optimal performance regarding ionization/desorption of heparin disaccharide I-S. The mass spectra in positive ion mode, as illustrated in Fig. S4 of supplementary material, show the similar properties of these matrices. In general, negative ions of heparin had less fragmentation than the positive ions; this is likely because heparin ions are stabilized by the deprotonation. Similar influence of deprotonation on the ion stability of sulfate groups has been reported [42].

Figure 6
figure 6

Negative-ion MALDI-TOF mass spectra of heparin I-S obtained from samples using different matrices: (a) droplets of frozen solution 2,5-DHB in 50% isopropanol and 1% ammonium hydroxide aqueous solution (500 J/m2), (b) droplets of frozen solution DHB with 75% methanol aqueous solution (650 J/m2), (c) trilayer with DHBB (500 J/m2), (d) trilayer with DHB (270 J/m2), (e) dried-droplet 2,5-DHB (270 J/m2), (f) dried-droplet MSA (210 J/m2), (g) dried-droplet HgTe (130 J/m2), and (h) ionic liquid DHBB (500 J/m2). Each spectrum was accumulated for 200 laser shots. The results show that the droplets of frozen solution sample with 75% methanol aqueous solution is the softest matrix for heparin disaccharide I-S. Ions, resulting from the matrix or background, are labeled with asterisks

  1. (d)

    Pullulan

Because few matrices are suitable for detecting polysaccharides with molecular weights greater than 3 kDa, detecting underivatized polysaccharides remain challenging with MALDI. Here, we used pullulan as an analyte of polysaccharide for the comparison of various MALDI matrices. Pullulan is a polymer with the repeating unit maltotriose [α-Glc-(1-4)-α-Glc-(1-4)-α-Glc] whereas the repeating units are connected by an α-1-6 glycosidic bond. Figure 7 shows the mass spectra of pullulan 10k obtained from different matrices. When the droplets of frozen solution were used as the matrix, two distinct distributions were observed (Figure 7a). Ions at m/z > 6000 (with a peak at 8000 Da) were primarily produced from intact pullulan. The mass difference between the peaks with large intensity is m = 486 Da, corresponding to a maltotriose. These peaks represent the different polymerization degrees of pullulan. The peaks with low intensity above each primary peak by mass 162 Da, or equivalent to below each primary peak by mass m = 326 Da, represent additional glucose (m = 162 Da) in polysaccharide, or the loss of a glucose disaccharide (m = 326 Da) from polysaccharide. They can result from the defeat of polymerization or fragmentation of pullulan upon ionization, respectively. The other distribution is ions at m/z < 6000 which were produced from fragments. The mass difference between these peaks is m = 162 Da, representing the loss of different numbers of glucose. These peaks suggest fragmentation occurs at every glycosidic bond. Only one distribution (m/z > 6000) was observed from the trilayer with DHBB (Figure 7b). When the trilayer with DHB or 2,5-DHB dried droplet was used as the matrix (Figure 7c, d), two distinct distributions were also observed. However, the intensities of the ions distributed at m/z < 6000 (mainly produced from fragments) were large, whereas those of the ions at m/z > 6000 (primarily from intact pullulan) were low, and the peak of the distribution shifted from 8000 to 6500 Da. This indicates that the droplets of frozen solution and the trilayer with DHBB were softer than 2,5-DHB and the trilayer with DHB. No pullulan signal (intact or fragment pullulan) was observed when MSA or HgTe was used as the matrix (Figure 7e, f). When ionic liquid DHBB or a trilayer with DHBB was used as a matrix, only one ion distribution (m/z > 6000) was observed. The ion distributions from these two matrices were very similar in intensity, and the peak of the distribution was located at a higher position (8000 Da) than that of the distribution of other matrices. Because both matrices contained ionic liquid DHBB, their softness likely resulted from the ionic liquid DHBB, suggesting that DHBB is the optimal matrix for pullulan. One possible reason why ionic liquid works well for pullulan is the solvation effect. Ionic liquids have been shown to achieve good dissolution of polysaccharides [43, 44] compared with water and organic solvents. The solvation effect likely is critical for the desorption and ionization processes of pullulan.

Figure 7
figure 7

Positive-ion MALDI-TOF mass spectra of pullulan 10k obtained from samples using different matrices: (a) droplets of frozen solution of 2,5-DHB in 37.5% acetonitrile aqueous solution (650 J/m2), (b) trilayer with DHBB (620 J/m2), (c) trilayer with DHB (360 J/m2), (d) dried-droplet DHB (260 J/m2), (e) dried-droplet MSA (160 J/m2), (f) dried-droplet HgTe (180 J/m2), and (g) ionic liquid DHBB (620 J/m2). Each spectrum was accumulated for 200 laser shots

In summary, we demonstrated that droplets of frozen solution with the proper choice of solvent are the softest matrices for gangliosides and heparin. Ionic liquid 2,5-dihydroxybenzoic acid butylamine is the most suitable matrix for pullulan. The desorption and ionization of intact labile glycoconjugates without fragmentation depends on the softness of matrices and the properties of the analyte ions. Solvents in droplets of frozen solution with low boiling and melting points can easily desorb at low temperature, and this greatly reduces the potential fragmentation of labile glycoconjugates. Desorption at low temperature is one of the major contributors to the softness of matrix. Solvation of glycoconjugates by matrix molecules, which makes desorption of glycoconjugates easy, is another contributor to the matrix softness. Rapid temperature decrease of solution droplets by spraying with low-temperature (100 K) N2 gas prevents the growth of large ice crystals, thereby facilitating the solvation of glycoconjugates by solvent molecules. Good solubility of pullulan in ionic liquid is another example of the importance of the solvation effect. The stability of ions further contributes to the reduction of fragmentation. Heparin negative ions have less fragmentation than positive ions, indicating that the proper choice of ionic species also contributes to the reduction of fragmentation. These three contributors can help inform the choice of soft matrix for labile biomolecules in MALDI.