Glycans are assemblies of linear and branched monosaccharide chains that govern molecular interactions and, therefore, cell communication, signal transduction, pathogen recognition, and immune responses1-3. In living mammalian cells, glycosylation (catalyzed by glycosyltransferases and glycosidases) produces a highly complex and vast repertoire of cellular glycans, with a colossal structural diversity1. Mammalian cells are covered with a dense layer of glycans and the proteins and lipids they are attached to, termed the glycocalyx, which is involved in various vital cellular processes4-6. The type, size, structure, and charge of cell surface glycans may affect the biological properties of the cells and their susceptibility to potential viral infections2,4,7. Mammalian glycoconjugates, located on the cell membrane and extra- or intracellular space, play crucial roles in physiological and pathological events2,4,5,7,8. Alterations in the glycomic profiles of glycoproteins, e.g., overexpression of sialylated or core fucosylated glycans, or increased levels of complex-type branched glycans, may promote the acquisition of cellular features required for the malignant transformation of cells1,9-12. Recent studies have also reported altered glycosylation patterns in patients with Alzheimer's disease13,14 Glycosylation abnormalities represent an overt source of potential biomarkers for the diagnostic, prognostic, and treatment monitoring of various human diseases, including autoimmune, congenital, oncological, and neurodegenerative pathologies5,7,12. In the emerging field of glycomedicine, novel therapeutic drugs or vaccines targeting tumor-associated carbohydrate antigens, like Lewis antigens and polysialic acids, are also currently in clinical evaluation5,15,16.
The analysis of single cells may result in the characterization of distinct cell subpopulations and rare cells, and in the differentiation of various cell states, which are often overlooked in bulk sample analyses17,18. Single-cell analysis is also beneficial to certain applications (e.g., minimally invasive liquid biopsy/microbiopsy-based cancer diagnosis and monitoring), where the amounts of original biological material may be limited for downstream molecular profiling. Single-cell omics encompasses a broad spectrum of analytical techniques that rely on vastly different technological principles, including genomics, transcriptomics, metabolomics, lipidomics, proteomics, and glycomics17,20,21. In contrast to single-cell genomics and transcriptomics, single-cell proteomics and glycomics remain in their infancy. The analysis of proteomes and glycomes at the single-cell level is limited by the minute amounts of starting biological material and the inability to directly amplify protein and glycan species. For single-cell proteomics, ultra-sensitive mass spectrometry (MS)-based analytical platforms are under development22-25. The Slavov group developed Single Cell ProtEomics by Mass Spectrometry (SCoPE-MS) techniques that rely on liquid chromatography (LC)-MS/MS methods with tandem mass tag (TMT)-labeling of peptides and a carrier channel for quantifying protein covariation across hundreds and potentially thousands of single cells26,27. The Kelly28,29 and Mechtler30 groups succeeded in developing reliable platforms for label-free single-cell proteomics using in-house or commercial nanoliter-liquid handlers. Gebreyesus et al. implemented a fully automated workflow combining microchips and MS for sample preparation and bottom-up analysis, which resulted in the detection of >1,500 protein groups from one single mammalian PC-9 cell31. Some of the present inventors have recently reported a capillary electrophoresis (CE)-MS method for top-down single-cell proteomic profiling32. This proof-of-concept approach allowed identification of >60 unique proteoforms in single HeLa cells33.
Contrary to the rapidly growing field of single-cell proteomics, single-cell glycomics (SCG) has shown lagging progress. This may be explained by i) the substantially lowered amounts of biological material available (glycosylation may account for only ˜1-10% of the total mass of one human glycoprotein1), and ii) the necessity to label the native glycans in the most commonly used positive electrospray ionization (ESI)-MS mode for increased ionization efficiency and detectability of released glycans, which obviously induces additional and substantial sample losses during the sample processing. Analytical platforms were implemented for total cellular glycome analysis of various mammalian cells, but these studies required large amounts of cells (˜105-107 cells)34,35. To date, only a few analytical technologies have been developed for SCG. The Johnston group developed the SUrface-protein Glycan And RNA-seq (SUGAR-seq) method for the analysis of the transcriptome, extracellular epitopes, and N-linked glycosylation at the single-cell level, using biotinylated lectins and anti-biotin antibodies combined with multimodal RNA-seq technology36. Oligonucleotide-labeled lectins were used by the Tateno group for sequencing-based glycomic profiling of single cells, which enabled the acquisition of 39 lectin-binding signals per single cell and provided an informative picture of cell surface glycosylation37. Roan and co-workers implemented a cytometry time-of-flight-lectin (CyTOF-Lec) technique for the simultaneous detection and quantification of proteins and glycans on the surface of human cells38. In this approach, lanthanide-conjugated lectins were combined with traditional CyTOF mass cytometry using lanthanide-conjugated antibodies to specifically bind glycans and proteins and quantify them (through the lanthanide metal quantification) using inductively-coupled-plasma-MS (ICP-MS). The developed technique showed that CD4+ T cell surface glycosylation could influence the susceptibility of CD4+ T cells to viral infection. Nevertheless, the above-described approaches involve tedious, expensive, and time-consuming analytical workflows with sophisticated instrumentation and, most importantly, do not allow the direct analysis, quantitation, and accurate structural characterization of the glycans. Furthermore, some cell surface glycans may not interact with the lectins selected in the developed methodologies. These last few years, computational modeling software tools to predict the glycome at the single-cell level were also developed, based, for example, on single-cell RNA-seq transcriptomics data 39. Yet, so far, methods enabling the direct analysis, characterization, and quantitation of cell glycomes at the single-cell level have not been reported.
The present technology provides an integrated platform coupling online in-capillary sample processing with high-sensitivity label-free capillary electrophoresis-mass spectrometry for N-glycan profiling of single mammalian cells and nL-volumes of amount-limited biomedically-relevant samples. Direct and unbiased characterization and quantification of single-cell surface N-glycomes were demonstrated for HeLa and U87 cells, with the detection of up to 100 N-glycans per single cell. N-glycome alterations were unequivocally detected at the single-cell level in HeLa and U87 cells stimulated with lipopolysaccharide. The developed method was also applied to the profiling of ng-level amounts of blood-derived protein, extracellular vesicles, and total plasma isolates, resulting in over 170, 220, and 370 quantitated N-glycans, respectively.
An aspect of the present technology is a method of glycan analysis including the following steps: (a) providing an open tube capillary electrophoresis instrument whose output provides an electrospray or other ionization-based input for a mass spectrometer, a glycan release agent solution disposed within a capillary tube of the open tube capillary electrophoresis instrument, and a sample comprising a glycoprotein or glycolipid in an aqueous medium; (b) introducing the sample into an inlet of a capillary tube of the open tube capillary electrophoresis instrument or the channel of a microfluidic capillary electrophoresis instrument, whereby the glycoprotein or glycolipid contacts the glycan release agent solution; (c) allowing the glycan release agent to release one or more glycan moieties from the glycoprotein or glycolipid without modification of glycan structure or composition; (d) separating the released glycan moieties within the capillary tube or channel using the open tube capillary electrophoresis instrument or the microfluidic capillary electrophoresis instrument based on charge and hydrodynamic volume of the released glycan moieties to form a plurality of separated glycan moieties within the capillary tube; (e) injecting the separated glycan moieties from an outlet of the capillary tube or channel into the mass spectrometer, whereby the separated glycan moieties are ionized and fragmented to form a plurality of charged glycan fragments; (f) separating and detecting the charged glycan fragments based on mass-to-charge ratio using the mass spectrometer; and (g) analyzing the separated and detected charged glycan fragments, whereby one or more structural characteristics of said plurality of glycan moieties are determined. The glycan release agent can be a single enzyme, such as PNGase F, or a combination of enzymes, or a chemical reagent. In the present technology, glycans are released from glycoproteins or glycolipids, preferably as an intact unit with its native structure preserved and without chemical derivatization or labeling, through an enzymatic reaction carried out in the capillary tube of a capillary electrophoresis instrument. The method allows separation, identification, and quantification of glycans in a single run and avoids sample manipulations that lead to sample losses or alteration.
The present technology can also be summarized in the following list of features.
1. A method of glycan analysis, the method comprising the steps of:
The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.
The inventors developed an integrative platform coupling on-line in-capillary sample processing with high-sensitivity label-free capillary electrophoresis-mass spectrometry (CE-MS) for N-glycan profiling of single mammalian cells and limited amounts of biomedically-relevant specimens (e.g., blood-derived isolates). The biological samples (e.g., individual cells, serum proteins, total plasma, and extracellular vesicles (EVs) were tested) are injected into the CE capillary by hydrodynamic pressure. Then, underivatized and native N-glycans are released within the capillary with a digestion solution of PNGase F enzyme. Finally, the released N-glycans are analyzed by high-sensitivity CE-MS in their non-labeled state without any further sample preparation. The developed technique allows, for the first time, direct and unbiased characterization and quantification of single mammalian cell N-glycomes.
To date, only a few analytical technologies, based on carbohydrate-binding lectins, have been developed for single-cell glycomics. These technologies, which require sophisticated instrumentation, are very expensive and time-consuming, and, most importantly, result in undirect and biased profiling of cell-surface glycans (i.e., the glycans are not released from the cell surface for their direct detection and quantification, and the glycan detectability is dependent on lectin binding affinity). Furthermore, some cell surface glycans may not interact with the lectins selected in the developed methodologies. The present technology offers a vastly simplified in-capillary sample preparation approach coupled on-line with label-free CE-MS. With this method, the glycans are analyzed by CE-MS in their native non-labeled state, and the substantial sample losses associated with sample handling and transfer steps in the off-line approach are eliminated. This present innovative workflow allows straightforward and effective glycan profiling of small populations of cells (1-10 cells) as well as nL down to pL levels of biomedically-relevant samples. CE-MS analysis of N-glycans in their native non-labeled state enables the preservation of the integrity and endogenous structural features of glycans, especially fucosylation and sialylation. In contrast, the most commonly used glycan labeling approaches can substantially alter the biological sample and result in partial disintegration of glycans during the process of labeling and sample preparation. Besides, the label-free CE-MS method, the present technology provides accurate and unambiguous structural characterization of the detected glycans, and provides crucial information on glycan antenna-branching and glycosidic linkages, without using selective enzymatic digestions (a tedious and time-consuming approach commonly used in glycomic analysis, which may leave some ambiguity in the assignment of glycosidic linkages).
The mild conditions used for N-glycan release allowed preservation of cell membrane integrity and specifically liberated cell surface N-glycans. N-glycans were analyzed in their native underivatized state to preserve their endogenous glycan features and eliminate the drawbacks associated with any labeling procedures, including incomplete derivatization, side-products, sample losses during cleanup steps, and high levels of defucosylation/desialylation during sample preparation and MS analysis. For glycan analysis of intact mammalian cells (1-10 cells), a manual hydrodynamic cell loading procedure32 was further optimized, not only to increase the robustness and throughput of cell loading, but also to improve the detectability and separation of the released N-glycans during CE-MS analysis. The vastly simplified in-capillary sample preparation approach coupled online with CE-MS eliminated sample losses associated with sample handling and transfer steps in the offline approach and allowed us to analyze sub-ng-levels of model proteins and pL-levels of total plasma, as well as single mammalian cells. Such an approach enabling direct analysis and quantification of N-glycans derived from one single cell has not been reported previously. In addition, biochemical stimulation of mammalian cells induced significant qualitative and quantitative changes in cell glycosylation profiles, which confirmed the potential of the method to detect cell glycome alterations in biological and biomedical applications at the single-cell level.
The present methods utilize in-line, in-instrument sample processing, separation, and analysis, all in a single continuous process. A key aspect is the release of glycan moieties from glycoproteins and glycolipids at the outset of the method, via one or more enzyme reactions carried out within the capillary tube or microfluidic channel of a capillary electrophoresis instrument. In a preferred embodiment, a single reaction is carried out in a liquid “plug” in the capillary containing the sample or a portion thereof and one or more glycan release agents. A glycan release agent is capable of releasing glycan moieties, preferably in there entirety and in their native state, as found in a cell or in the biological source of the glycoprotein or glycolipid. Also preferred is that a single glycan release agent is used that releases complete glycan moieties in a single step. However, also suitable can be two or more enzymes that work together to release portions of glycans either in a single step (i.e., performed in a single plug in the capillary, or in a series of sequential steps performed in a sequential series of plugs, each containing a different enzyme (e.g., endoglycosidases and/or exoglycosidases).
Any suitable glycan release agent can be used in the present technology. A preferred glycan release agent for N-glycans is PNGase F, which cleaves off N-glycans in their entirety and in their native structure from N-linked glycoproteins. For O-glycans, released from O-linked glycoproteins, monosaccharides can be sequentially released by a series of exoglycosidases until only the Gal-β(1→3)-GalNAc core remains. O-Glycosidase can then remove the intact core structure with no modification of the serine or threonine residues. Chemical reagents that release glycan moieties are known and also can be used. See Song, et al., 2016, Nat. Methods 13(6):528-534.
N-glycan profiling of human serum IgM. Human immunoglobulin M (IgM), a heavily glycosylated multimeric protein (Mrth 970 kDa and 1,080 kDa for pentameric and hexameric forms, respectively), was selected to develop and optimize the in-capillary sample processing method for N-glycan release coupled online to label-free CE-MS analysis. N-glycans account for ˜10% of the total mass of IgM 4, and the serum level of IgM is in the range 0.4-2.5 mg/mL45,46. CE-MS analysis of IgM-derived in-capillary released N-glycans resulted in the identification of 173±6 (n=3) non-redundant N-glycan compositions in human serum IgM isolate for injected amounts of 5 ng (i.e., 5 fmol) of protein, corresponding to ˜500 pg of N-glycans and equivalent to the amount of IgM isolated from ˜3 nL of human serum (
The optimized workflow described above was well adapted to the analysis of 5 ng and sub-ng amounts of serum IgM. Using these tiny sample amounts and intensive rinsing steps between runs, no significant carryover derived from the analysis of preceding IgM samples was observed, based on the analysis of the water blank control sample. The injection of larger amounts of protein (e.g., 25-100 ng), which could potentially increase the glycan coverage of IgM, would require the re-optimization of several parameters, including the glycosidase: protein substrate ratio, the incubation time, and the rinsing steps between runs to efficiently clean the capillary. This scale-up workflow would obviously increase the sample processing and total analysis times. Since the goal of our study was to develop an effective and quick CE-MS-based workflow applicable to single-cell analysis, It was estimated that glycan amounts released from the digestion of 0.1 to 5 ng of model glycoprotein within the CE capillary should reflect well the amounts of glycans released from one to ten mammalian cells.
N-glycan profiling of human serum IgG. The newly developed N-glycan profiling workflow was tested using CE-MS analysis of human IgG (Mrth 150 kDa), a class of immunoglobulin less glycosylated than IgM. N-glycans account for ˜2% of the total mass of IgG, and the human serum level of IgG is in the range 7-16 mg/mL45,46. The developed CE-MS method allowed identification of 142±9 (n=3) non-redundant N-glycan compositions in human serum IgG isolate for injected amounts of 5 ng (i.e., 33 fmol) of protein, corresponding to ˜100 pg of N-glycans and equivalent to the amount of IgG isolated from −500 pL of serum (
N-glycan profiling of total human plasma. The above described method allowed analysis of sub-nL volumes of total human plasma isolate, and enabled direct online N-glycan profiling of plasma volumes as small as 5 pL, which was not reported before. Data processing of CE-MS analyses resulted in the identification of 375±12, 234±10, and 152±21 (n=3) non-redundant N-glycan compositions in whole blood plasma for injected amounts of 500 pL, 50 pL, and 5 pL of plasma (i.e., ˜1,500, 150, and 15 pL of human blood), respectively (
N-glycan profiling of blood-derived extracellular vesicles. The developed in-capillary workflow was applied to the analysis of N-glycans released from human plasma-derived extracellular vesicles (EVs), another attractive source of disease biomarkers51,52. Experiments were carried out with the injection of a purified EV isolate, containing ˜1×104 EV particles/nL (see Methods section). CE-MS analysis resulted in the detection and identification of 127±14 and 226±7 N-glycans in the total EV isolate, using 1 nL and 50 nL of EV isolate injection volumes, respectively (containing approximately 1×104 EVs and 5×105 EVs, respectively) (
Differential N-glycan profiling of IgM, IgG, whole plasma, and EV isolates from blood. A qualitative and quantitative comparative analysis of N-glycans detected in the four types of analyzed blood-derived isolates (IgM, IgG, total plasma, and EVs) was conducted with an exhaustive list of 679 glycans, encompassing all the non-redundant N-glycan compositions identified in the four blood isolates. This differential analysis further demonstrated the uniqueness and high complexity of the four examined N-glycomes (
Single-cell loading and in-capillary N-glycan release. Individual mammalian cells were introduced into the CE capillary in a controlled manner using a hydrodynamic injection mode (
The cells loaded into the CE capillary were sandwiched between two plugs of a PNGase F digestion solution, and two short CE voltage pulses (30 s each) were applied in normal and reverse polarity to effectively mix the cells with the glycosidase. No lysis buffer was employed and/or injected to preserve the cell integrity and release only the N-glycans from the cell surface. Ideally, to preserve cellular integrity, the cells should be maintained in a buffer solution that closely mimics physiological pH and osmolarity. However, such buffers are typically incompatible with MS or CE analysis and may result in ionization suppression, adduct formation, and decreased separation performance phenomena. In this study, a stacking strategy was used to enhance the peak shape and intensity and improve the separation of the detected glycans. To enable this strategy, the cells were resuspended in 1 mM ammonium acetate pH 6.7, immediately prior to their loading into the CE capillary, and the commercial PNGase F enzyme was diluted 7-fold in water to highly decrease the salt concentration (see Methods section). Given that the cells were exposed to a low osmolarity environment during the deglycosylation step for N-glycan release, an assessment of the post-incubation cell integrity was conducted through the offline incubation of single cells for 1 h, using the conditions employed in the in-capillary sample processing workflow (i.e., the cells were sandwiched between two PNGase F plugs). Fluorescence imaging of the single cells prior to and after offline incubation in a small piece of capillary did not reveal discernible alterations in the cell morphology, size, or membrane integrity (
N-glycan profiling of single, five, and ˜ten HeLa cells. To assess the capability of the present method for direct and unbiased N-glycan profiling of mammalian cells, sets of five repetitive experiments were performed with the injection of one, five, and about ten (i.e., 10±3 cells, referred to as “bulk sample”) HeLa cells. Characteristic ion density maps acquired in CE-MS analysis of HeLa cell-derived N-glycans are presented in
Higher levels of fucosylation (up to 6 fucose residues) and sialylation (5-12 SiA residues) were detected in 5-10 HeLa cells, compared to single HeLa cells, for which the degrees of fucosylation and sialylation of identified N-glycans did not exceed 2 and 4, respectively (
Due to cell-to-cell heterogeneity, high variations in the absolute glycan abundances measured in the five repetitive analyses were observed for one, five, and ˜ten injected HeLa cells (e.g., for single HeLa cell measurements, the relative standard deviations (RSD) of peak areas could be as high as 65%). It was hypothesized that such significant variation might be mostly attributed to the cell size, surface area, and cell cycle state. Nevertheless, a substantial increase in the glycan abundance levels was demonstrated with increased loaded cell numbers. A linear relationship was demonstrated between the injected cell number and the total cellular glycan amount for the eight selected glycans, based on peak area measurements.
CE-MS2 analyses of HeLa cell-derived N-glycans were performed to confirm the N-glycan composition identification results and provide information on structural features of the detected glycans (e.g., antenna-branching, fucose position, and SiA linkage). CE-MS2 experiments performed with ˜ten HeLa cells resulted in accurate and unambiguous structural characterization of 53 N-glycans in HeLa cells, among which acidic (i.e., sialylated) and neutral glycans (
Differential N-glycome analysis of single HeLa and U87 cells. Next, the in-capillary workflow was applied to the CE-MS analysis of single U87 cells to assess, as a proof-of-concept, whether qualitative and/or quantitative differences in cell surface N-glycomes of different cell types could be detected at the single-cell level. In comparison to single HeLa cells, a significantly higher number (˜5-fold) of N-glycans were detected and identified in single U87 cells. Five repetitive experiments carried out with single U87 cells resulted in the detection of 62±20 N-glycans per single cell (
89% of the glycans identified in single HeLa cells were also detected in single U87 cells (
Noticeable differences in the abundances of the N-glycans detected in HeLa and U87 single cells were also observed, based on peak area measurements. As an illustration,
Finally, proof-of-concept CE-MS2 experiments were performed with ˜ten U87 cells and resulted in accurate and unambiguous structural characterization of 29 N-glycans, including sialylated and neutral N-glycans.
Single-cell N-glycome alterations induced by LPS stimulation. Previous studies reported that THP-1 mammalian cells treated with lipopolysaccharide (LPS) exhibited increased60 or decreased61 levels of sialylation. Downregulation of glycan fucosylation was also reported for LPS-stimulated brain cells62. To test whether the developed CE-MS-based workflow could detect glycome alterations at the single-cell level, HeLa and U87 cells were stimulated with LPS. Interestingly, N-glycan profiling of single HeLa cells, after stimulation of HeLa cells with LPS, resulted in an ˜3-fold increased number of detected N-glycans, compared to the untreated HeLa cells (
Significant alterations of U87 cell N-glycome profiles were also observed at the single-cell level when U87 cells were treated with LPS, compared to the untreated U87 cells. CE-MS analysis of single U87 cells after LPS treatment resulted in the detection of 55±30 N-glycans per single U87 cell (n=5), and in the assignment of 161 non-redundant N-glycan compositions in total. 68% of the N-glycans identified in LPS-treated U87 cells were fucosylated, and the fractional distributions were 36%, 15%, 11%, 1%, 1%, and 4% for mono-, di-, tri-, tetra-, penta-, and hexafucosylated N-glycans, respectively (
The present technology offers several novel and unusual features. The method enables straightforward, unbiased, accurate, and deep qualitative and quantitative glycomic profiling of single mammalian cells. The method is based on innovative in-capillary sample processing coupled online to ultra-high sensitivity label-free CE-MS analysis of released glycans. The method also enables deep and highly informative glycan profiling of sub-0.5 ng-levels of model proteins and nL to pL levels of plasma volume equivalents (e.g., 5 pL of total plasma can be loaded in the CE capillary for straightforward glycan profiling). The numbers of N-glycans identified in the four types of analyzed blood-derived isolates (IgM, IgG, total plasma, and EVs) described here exceed by ˜7-fold those reported in other N-glycan profiling studies of similar complexity blood-derived isolates. Further, optimization of ionization, desolvation, and CE-MS conditions achieved the highest sensitivity levels available for glycan analysis. Further, even highly fucosylated (5-7 fucose residues) and highly sialylated (5-14 salic acid residues) N-glycans, which are difficult to detect using the prior glycan labeling-based methodologies, can be detected and structurally characterized using the present label-free CE-MS-based workflow. Such peculiar N-glycans can now be included in biopharmaceutical and clinical research and development efforts as new classes of targets. Moreover, unmatched separation performance of the present technology was achieved with the present CE-MS method, which allows separation and analysis of positional and linkage glycan isomers in a single CE-MS analysis. Such isomeric differentiation is a challenge for other separation approaches.
The present technology has several advantages over related technology of the prior art. (i) The developed technique offers a vastly simplified, in-capillary sample preparation that eliminates the sample losses associated with sample handling and transfer steps in off-line approaches, and it allows straightforward and in-depth glycan profiling of extremely small amounts of sample, such as nL or pL equivalent amounts of blood isolates as well as small populations of cells or even single cells. Up to now, there are no established methods for single-cell glycome analysis due to the inability to amplify glycan sequences and sample losses associated with any sample processing and glycan labeling. To date, only a few lectin-based glycan profiling technologies have been developed for single-cell glycomics. These approaches involve tedious, expensive, and time-consuming analytical workflows with sophisticated instrumentation and do not allow the direct analysis, quantitation, and accurate structural characterization of the glycans. Furthermore, some cell surface glycans may not interact with the lectins selected in the developed methodologies. (ii) The present glycan profiling technique is quick, requiring about 3 hours per single-cell or blood isolate analysis, including sample loading, in-capillary glycan release, and CE-MS analysis. It allows a well-controlled injection of individual cells, and requires only affordable analytical instruments. (iii) N-glycans are analyzed in their native underivatized state, preserving their endogenous glycan features and eliminating the drawbacks associated with labeling procedures, including incomplete derivatization, over-labeling, formation of side-products, sample losses during sample labeling and cleanup steps, ionization suppression and MS signal interference by the labeling reagent, and high levels of defucosylation/desialylation during sample preparation and MS analysis. With the present label-free approach higher levels of sialylation and fucosylation are detected because of lower levels of electrospray ionization and source-induced decay during MS analysis. (iv) The present method allows increased depth of glycan profiling, yielding higher numbers and varieties of glycans. (v) The high sensitivity and high dynamic range of detection (over 5 orders of magnitude) of the present technique enables the detection of highly fucosylated (5-7 fucose residues) and highly/heavily sialylated (5-14 salic acid residues) N-glycans, most of which were not reported before.
Uses of the present methodology include glycan profiling of small populations of cells (≤20 cells) and single cells, extracellular vesicles and single microvesicles, and limited amounts of biological or clinical samples, such as minimally-invasive liquid microbiopsies, and tissue isolates. The methods also can be used for discovery of novel diagnostic, prognostic, and treatment-monitoring of specific disease biomarkers based on glycosylation profile alterations associated with diseases and medical conditions. The present methods also can be integrated into development of a vast single-cell glycomics or multi-omics platform, which could provide crucial information on biological mechanisms underlying complex diseases, unachievable by merging data sets obtained from mono-omics studies of different cells. Characterization of glycosylation levels in biopharmaceuticals (e.g., therapeutic proteins and vaccines) can be obtained using the present technology.
Deionized water, methanol (99.9%, LC/MS Grade), Gibco F12-K medium, Gibco DMEM medium, Gibco 0.25% Trypsin-EDTA, Gibco penicillin-streptomycin (P/S), phosphate-buffered saline (PBS), CellMask™ plasma membrane stain (green), LIVE/DEAD™ fixable green dead cell stain kit, and lipopolysaccharide (LPS) were obtained from Thermo Fisher Scientific (Waltham, MA). 1 N NaOH, 1 N HCl, 5 N ammonium hydroxide, glacial acetic acid (99.99%), ultra-high purity ammonium acetate (99.999%), trypan blue, and total human serum IgM and human serum IgG isolates (purity ≥95%, based on non-reduced SDS-PAGE and verified by nanoLC-MS/MS of tryptic digests) were purchased from Sigma-Aldrich (St. Louis, MO). PNGase F enzyme was from New England Biolabs (Ipswich, MA). Fetal bovine serum (FBS) was purchased from R&D Systems (Minneapolis, MN). Platelet-free anticoagulated with EDTA pooled total human plasma (from blood donated by self-declared healthy male donors of 23-67 years old) was kindly provided by Prof. Ghiran's laboratory (BIDMC, Boston, MA). All bare fused silica (BFS) capillaries (91 cm×30 μm i.d.×150 μm o.d.) with sheathless CESI-MS emitters in OptiMS™ cartridges were from SCIEX (Redwood City, CA). Aquapel® was purchased at Pittsburgh Glass Works (Pittsburgh, PA).
HeLa-S3 and U87-MG cell lines (called HeLa and U87 cells thereafter) were from ATCC (Manassas, VA). HeLa cells were cultured in suspension at 37° C. in a complete F-12K medium supplemented with 10% FBS, 1% P/S, and 5% CO2. The cell density was maintained within a range of 2×105-1×106 cells/mL. Adherent U87 cells were cultured at 37° C. in DMEM medium supplemented with 10% FBS, 1% P/S, and 5% CO2. Upon reaching confluence, one flask of U87 cells was split into five flasks. Cells were stained with trypan blue and counted using a 2-chip disposable hemocytometer (Bulldog Bio, Portsmouth, NH) to estimate the cell density and viability.
2 or 4 μL of 2.5 mg/mL LPS were added to the 5 or 10 mL culture media in each HeLa or U87 cell culture flask, to get a final LPS concentration of 1 pg/mL. The HeLa and U87 cells were exposed to LPS for 24 h before being harvested and analyzed.
HeLa and U87 cells were collected, washed, counted, and subsequently centrifuged into pellets prior to CE-MS analysis. The HeLa cell pellets were obtained by direct centrifugation of the HeLa cell culture suspension at 300×g for 5 min. To detach the U87 cells from the flask bottom, 0.25% trypsin-EDTA was added, followed by incubation at 37° C. for 5 min. The digestion was stopped by adding complete DMEM medium, and the detached U87 cells were centrifuged at 300×g for 5 min to obtain the cell pellets. HeLa and U87 cell pellets were washed three times with 1×PBS, and their viability and density were assessed using a 2-chip disposable hemocytometer prior to the final centrifugation at 300×g for 5 min. The cell viability was typically >90%. The cell pellets were kept on ice until their use.
Offline Cell Loading into the CE Capillary
One or five cells were loaded offline into the silica surface OptiMS Cartridge, following the protocol described in our previous work32, with modifications. The cell loading process was visualized and monitored under an IX83 microscope (Olympus, Center Valley, PA), using a 10× magnification. First, the inlet of the CE capillary separation line was immobilized on a glass slide (pretreated with Aquapel®) placed under the microscope. Then, the capillary inlet was immersed in a 40 μL droplet of 1 mM ammonium acetate pH 6.7. A hydrodynamic flow was generated by manually lifting or lowering by ˜45 cm the electrospray emitter tip of the capillary (separation line outlet) to generate an ultra-low flow rate of ˜140 pL/s and enable precise control of the cell influx. Flow towards the separation line inlet was created by lifting the emitter tip to expel air bubbles before cell loading or dislodge unwanted cells after cell loading. For cell loading, 5 μL of a cell suspension at ˜5 cells/nL was mixed with the droplet in which the separation line inlet was immersed, while the emitter tip of the capillary was held at the same height as the separation line inlet to prevent any forward or backward flow. The cell-containing droplet was gently agitated with a pipet tip until a target single cell (e.g., with the desired size and morphology) was observed in close proximity to the inlet. Then, the emitter tip of the capillary was lowered to introduce the cell into the capillary, and the flow was maintained until the cell was located approximately 500 μm away from the capillary inlet for targeted cell injection. The same procedure was repeated several times to load manually the desired number of cells.
To record the cell morphology and size distribution, 5 μL of a suspension of unstained cells in 1×PBS (with a cell density of ˜1×106 cells/mL) were deposited on a glass slide and imaged with the microscope under bright field at 10× magnification. For improved visualization of the cell morphology and membrane integrity, the cells were stained with CellMask™ plasma membrane green stain, following a procedure adapted from the manufacturer's protocol. The 1,000× concentrated stain solution was diluted to 1× working solution with PBS. Subsequently, the cell pellet was resuspended with the working solution to an approximate cell density of 1×106 cells/mL. Then, the cells were incubated in the dark for 30 min, followed by three washes with PBS to remove the excess stain. For fluorescence microscopy imaging of stained cells loaded within the capillary, the polyimide coating was removed before the experiments to avoid interference. To determine the cell viability and membrane integrity under the selected in-capillary sample processing conditions (i.e., after 60 min of incubation with the PNGase F enzyme in 1 mM ammonium acetate pH 6.7 buffer), the cells were stained with LIVE/DEAD™ fixable green dead cell stain. For this, one vial of the fluorescent dye was resuspended with 50 μL of DMSO. HeLa cells were harvested, washed, and resuspended with 1 mM ammonium acetate pH 6.7 to an approximate cell density of 5×105 cells/mL. Then, 1 μL of the resuspended dye was added to 1 mL of the cell suspension. Finally, 20 μL of the stained cells were mixed with 15 mlU of PNGase F. Bright field and fluorescence-based microscopic images were acquired at different time points to evaluate the cell viability and morphology during the deglycosylation step with PNGase F.
Plasma-derived EVs were isolated using a size exclusion chromatography (SEC) column with a Sepharose CL-2B stationary phase. Briefly, 100 μL of platelet-free anticoagulated with EDTA pooled human plasma (from blood donated by self-declared healthy male donors of 23-67 years old) were loaded on the SEC column. EVs were eluted from the SEC column with 0.1×dPBS, and the EV-containing fractions were pooled. The pooled EV fractions were then concentrated using Amicon® 30 kDa MWCO ultrafiltration devices to a final volume of ˜33 μL, and stored at 4° C. until their analysis. The approximate EV particle concentration was estimated to be 1×1010 EV particles/mL, based on a combination of EV counting, using tunable resistive pulse sensing (TRPS), nano-flow cytometry, and immunoaffinity-based interferometry.
The CE capillary was interfaced with an Orbitrap™ Fusion Lumos™ mass spectrometer using a Nanospray Flex ion source (both Thermo Fisher Scientific, Bremen, Germany). All analyses were carried out in negative ESI mode. The nanoelectrospray potential was set to ˜1.8 kV. The ion transfer tube (ITT) temperature was set to 150° C. (the distance between the electrospray emitter tip and the MS ITT was set to ˜5 mm). The CE-MS analyses were performed with automatic gain control (AGC) of 1×106 or 250%, a maximum injection time of 250 ms, 5 microscans, an S-lens voltage set to 65 eV, the nominal resolving power of 120,000 at 200 m/z, and in-source collision-induced dissociation (ISCID) at 70 eV. For CE-MS2 experiments, instrument resolving power was set at 60,000 at 200 m/z with 1 microscan. AGC was set to 2×105 with a maximum injection time of 1,000 ms. An isolation window of 2 m/z was selected, and 32 eV was determined to provide the optimum normalized collision energy.
For data acquisition and processing, Xcalibur™ (v. 3.1) software was used. CE-MS data were processed with GlycReSoft (v. 3.10) software (Boston University, Boston, MA, USA). Analyses of CE-MS2 data were performed with SimGlycan (v. 5.91) software (Premier Biosoft, Palo Alto, CA, USA). The generated results were based on the processing of three replicate (model proteins, total plasma, and EVs) and five repetitive (mammalian cells) analyses. For CE-MS1 processing with GlycReSoft, a mass matching error tolerance of 20 ppm was used in all searches. Up to six charge states, and sodium and ammonium adducts were included in the search. Other parameters were the same as described in our previous reports40,41. The glycan identification analysis of the CE-MS data was conducted using database searches against in-house built mammalian database (version of December 2020) encompassing 27,335 N-glycan compositions (the mammalian database provided with the GlycReSoft software package encompasses 1,766 N-glycan compositions). For CE-MS2 processing with SimGlycan, a 20 ppm precursor mass tolerance and a 10 ppm fragment mass tolerance were used in all searches. Non-labeled glycans (unmodified or with sodium adduction) were searched selecting the options “Underivatized” and “Free” in the chemical derivatization and reducing terminal windows, respectively. Other parameters were as described in our previous studies40,41. The glycan composition identification results were mainly based on CE-MS data processing using GlycReSoft. As additional verification of the plausible glycan identifications made using GlycReSoft, several supplementary levels of manual data examination were applied according to our recent studies40,41. In brief, this verification included 1. Predictable trends in CE-MS migration patterns, 2. Charge state and isotopic distributions characteristic to glycan ions, 3. Detection of neutral losses, and 4. Manual examination of CE-MS2 data for low intensity parent ions. The relative quantitation of the detected N-glycans was based on the single-stage MS signal intensities or peak areas of the detected N-glycans that were normalized with respect to the summed MS signal intensities or peak areas of all the N-glycans detected in the sample. In addition, a qualitative comparison was performed based on the fractional distributions corresponding to the number of specific species (e.g., diasialylated glycans) out of the total number of N-glycans detected and identified in the analyzed biological specimens.
The bar charts with individual data points, mean values, and error bars were plotted using the R language and ggplot2 package in the rStudio development environment (2023.03.0+386 “Cherry Blossom” Release). The R language in the rStudio development environment was also used to perform statistical ANOVA and paired t-tests. The open-access tBtools-II (v1.120) software42 was employed to generate heatmap clustering, utilizing the Euclidean distance-based clustering method and the complete cluster approach. For data clustering, N-glycan abundances (based on peak intensities) were normalized with respect to the summed abundances of all the N-glycans detected in the analyzed biological sample, and the clustering was done using the normalized abundance values after imputing 10% of minimum abundance for missing values followed by log2 transformation. The PCA plots were created with the open-access version of ClustVis (https://biit.cs.ut.ee/clustvis/software43. The average cell diameters of HeLa and U87 cells were measured using the open-access ImageJ (v1.53k) software. The glycan structures were designed with the open-access version of GlycoWorkBench (v2.0). Other schematic images (e.g., cell structure illustration) were built with the BioRender graphical tool.
For the Euclidean distance-based hierarchical clustering of single HeLa and single U87 cells before LPS treatment, glycans that were detected in at least three CE-MS analyses out of the ten total repetitive analyses (i.e., five repetitive analyses for single HeLa cells and five repetitive analyses for single U87 cells) were selected. This selection generated a set of 47 glycans highly representative of single HeLa and single U87 cells. For the Euclidean distance-based hierarchical clustering of single HeLa and single U87 cells after LPS treatment, glycans that were detected in at least two CE-MS analyses out of the five repetitive analyses of LPS-treated HeLa cells, and glycans that were detected in at least two CE-MS analyses out of the five repetitive analyses of LPS-treated U87 cells were selected and added to the above-described 47 glycans that are highly representative of HeLa and U87 untreated cells. For the PCA analysis of the glycans identified in single HeLa and single U87 cells after LPS treatment, more stringent parameters were used. In this case, glycans that were detected in at least three CE-MS analyses out of the five repetitive analyses of LPS-treated HeLa cells, and glycans that were detected in at least three CE-MS analyses out of the five repetitive analyses of LPS-treated U87 cells were selected and added to the above-described 47 glycans highly representative of HeLa and U87 untreated cells.
IgM and IgG were isolated from blood serum by size-exclusion chromatography (IgM) and ion-exchange chromatography (IgG), sample injections were performed at 1 or 5 psi for 6 s, corresponding to 1 and 5 nL injection volumes, respectively (i.e., 0.16 and 0.8% of the capillary volume, respectively). Three replicate analyses were performed with the injection of 0.1 ng and 5 ng of IgM, corresponding to ˜60 pL and ˜3,000 pL of human serum, respectively. Three replicate analyses were performed with the injection of 0.5 ng and 5 ng of IgG, corresponding to ˜50 pL and ˜500 pL of human serum, respectively.
1 mL of whole blood plasma isolate was centrifuged at 16,000×g for 20 min at 4° C., and the supernatant was carefully pipetted to avoid collecting the lipid layer. For CE-MS analysis of total plasma, sample injections were performed at 1 psi for 6 s, corresponding to 1 nL injection volumes, and 5, 50, or 500 pL of plasma, depending on the dilution of the “pre-processed” plasma sample in water.
Sample injections were performed at 1 psi for 6 s (1 nL injected) or 5 psi for 60 s (50 nL injected), corresponding to injected amounts equivalent to ˜3 nL and ˜150 nL of plasma, respectively.
Cell pellets were resuspended in 200 μL of 1 mM ammonium acetate pH 6.7 to get a final cell density of ˜5 cells/nL. For online cell loading of ˜10 cells, 2 nL of a cell suspension at ˜5 cells/nL were injected, applying 1 psi for 12 s. In the offline cell loading mode, 1 to 5 cells were selected and injected manually as described above, and the 1-5 cell-containing plugs corresponded to ˜4-6 nL injection volumes, based on microscope visualization. Owing to cell size variations, sets of five repetitive analyses were systematically performed with one (offline injection), five (offline injection), and ˜ten (online “bulk sample” injection) mammalian cells for each cell type (HeLa and U87 cells). CE-MS analyses of a blank sample of water were performed systematically to confirm insignificant levels of carryover derived from the analysis of preceding biological samples (blood-derived isolates and mammalian cells). For single-cell analysis, a water blank sample was analyzed between each single-cell injection. CE-MS analyses of the cell suspension medium (i.e., 1 mM ammonium acetate cell suspension buffer) were also performed. For these control analyses, 2-4 nL of water or cell suspension medium were injected inside the capillary and processed using the developed workflow, including the digestion step with PNGase F.
Off-Line Cell Loading into the CE Capillary
One to five cells were loaded off-line into the silica surface OptiMS Cartridge. The cell loading process was visualized and monitored under an IX83 microscope (Olympus, Center Valley, PA), using a 10× magnification. First, the inlet of the CE capillary separation line was immobilized on a glass slide (pretreated with Aquapel®) placed under the microscope. Then, the capillary inlet was immersed in a 40 μL droplet of 1 mM ammonium acetate pH 6.7. A hydrodynamic flow was generated by manually lifting or lowering by ˜45 cm the electrospray emitter tip of the capillary (separation line outlet) to generate an ultra-low flow rate of ˜140 pL/s and enable precise control of the cell influx. Flow towards the separation line inlet was created by lifting the emitter tip to expel air bubbles before cell loading or dislodge unwanted cells after cell loading. For cell loading, 5 μL of a cell suspension at ˜5 cell/nL was mixed with the droplet in which the separation line inlet was immersed, while the emitter tip of the capillary was held at the same height as the separation line inlet to prevent any forward or backward flow. The cell-containing droplet was agitated with a pipet tip until a single cell with the desired size and morphology was close to the inlet. Then, the emitter tip of the capillary was lowered to introduce the cell into the capillary and the flow was maintained until the cell was located approximately 500 μm away from the capillary inlet to prevent the cell dislodging. The same procedure was repeated several times to load manually the desired number of cells.
In-capillary sample processing for N-glycan release with PNGase F and CE-MS experiments were conducted using a CESI 8000™ instrument (SCIEX). In all experiments, bare fused silica (BFS) OptiMS capillaries (91 cm×30 μm i.d.×150 μm o.d.) were used. Prior to each online or offline sample injection, a series of rinses of the separation and conductive lines were performed. For the separation capillary, these rinses included: MeOH (100 psi, 10 min), 0.1 M NaOH (100 psi, 3 min), 0.1 M HCl (100 psi, 3 min), and Milli-Q water (100 psi, 5 min), followed by the background electrolyte (BGE) (100 psi, 7 min). The conductive line was rinsed with the BGE (100 psi, 2 min). Before and after online (model glycoproteins, whole plasma, EVs, and ˜ten cells (referred to as “bulk cells” in this study)) or offline (1-5 cells) sample loading into the CE capillary inlet, a plug (1 or 2 nL applying 1 psi for 6 or 12 s) of a PNGase F digestion solution at 1.1 mlU/pL in 7 mM NaCl, 3 mM Tris-HCl, and 0.7 mM Na2EDTA was injected into the capillary using the CESI 8000 instrument. For offline cell loading, the CE cartridge was removed from the CESI instrument after the injection of the first PNGase F plug for manual cell loading, as described above, and placed back in the CESI instrument for subsequent in-capillary sample processing. After online or offline sample loading, a short plug of water (1 nL) was injected before the injection of the second PNGase F plug, followed by a short plug (0.5 or 1 nL) of 50 mM ammonium acetate pH 6.7. Then, two voltage pulses of 20 kV were applied in normal and reverse polarity for 30 s with the BGE composed of 10 mM (ionic strength) ammonium acetate pH 4.5 with 10% isopropanol, before incubating the capillary inlet in a vial containing 50 mM ammonium acetate pH 6.7 for either 30 min (model glycoproteins, whole plasma, and EVs) or 60 min (mammalian cells). After the in-capillary incubation step (performed at −12° C.) for N-glycan release with PNGase F, a BGE plug (10 psi for 10 s (model glycoproteins and whole plasma) or 10 psi for 60 s (mammalian cells)) was injected in the capillary prior to label-free CE-MS analysis of released N-glycans performed as described below. All CE methods employed 20 kV in reverse polarity with a voltage ramp time of 1 min. The CE-MS experiments were carried out with a BGE of 10 mM (ionic strength) ammonium acetate pH 4.5 with 10% isopropanol. This BGE generated a relatively low cathodic EOF (μEOF 2.02×10−8 m2/V/s) based on the detection of a neutral marker (acetaminophen). All CE-MS analyses were performed with a CE supplemental pressure (SP) of 5 psi, which was applied 18 min after switching on the CE voltage at the beginning of the CE run. Due to the variability in the manually injected cell plugs (performed offline), the migration time ranges in CE-MS analysis of mammalian cells were normalized, based on the most abundant detected glycan species.
The CE capillary was interfaced with an Orbitrap™ Fusion Lumos™ mass spectrometer using a Nanospray Flex ion source (both Thermo Fisher Scientific, Bremen, Germany). All analyses were carried out in negative ESI mode. The nanoelectrospray potential was set to ˜1.8 kV. The ion transfer tube (ITT) temperature was set to 150° C. (the distance between the electrospray emitter tip and the MS ITT was set to 5 mm). The CE-MS analyses were performed with automatic gain control (AGC) of 1×106 or 250%, a maximum injection time of 250 ms, 5 microscans, a S-lens voltage set to 65 eV, the nominal resolving power of 120,000 at 200 m/z, and in-source collision-induced dissociation (ISCID) at 70 eV. For CE-MS2 experiments, instrument resolving power was set at 60,000 at 200 m/z with 1 microscan. AGC was set to 2×105 with a maximum injection time of 1,000 ms. An isolation window of 2 m/z was selected, and 32 eV was determined to provide the optimum normalized collision energy. CE-MS1 was performed as described above.
As used herein, “consisting essentially of” allows the inclusion of materials or steps that do not materially affect the basic and novel characteristics of the claim. Any recitation herein of the term “comprising”, particularly in a description of components of a composition or in a description of elements of a device, can be exchanged with “consisting essentially of” or “consisting of”.
While the present invention has been described in conjunction with certain preferred embodiments, one of ordinary skill, after reading the foregoing specification, will be able to effect various changes, substitutions of equivalents, and other alterations to the compositions and methods set forth herein.
This application claims the priority of U.S. Provisional Application No. 63/432,953, filed 15 Dec. 2022 and entitled “Ultra-Sensitive Label-Free Deep Profiling of Glycans Released from Single-Cells and nL-Volumes of Amount-Limited Biomedically-Relevant Samples”, the whole of which is hereby incorporated by reference.
This invention was made with government support under Grant Nos. 1R01CA218500 and 1R35GM136421 awarded by the National Institutes of Health. The government has certain rights in the invention.
Number | Date | Country | |
---|---|---|---|
63432953 | Dec 2022 | US |