Rationale and experiment design
The efficiency and selectivity of glycopeptide enrichment by ZIC-HILIC SPE were first investigated using well-defined synthetic glycopeptides spiked into different concentrations of tryptic peptides derived from BSA. Following these initial experiments, HILIC was performed on proteolytic digests of standard glycoproteins (IgG and A1PI) and finally applied on a complex sample derived from human serum (Fig. 1). The influence of the mobile phase on glycopeptide enrichment efficiency was tested using four different solvents: (i) acetonitrile, (ii) methanol, (iii) ethanol and (iv) isopropanol were used at a concentration of 80% in water containing 1% TFA as an acidic modifier. In the course of this work, we also evaluated a simplified HILIC enrichment technique (“Drop-HILIC”) and compared it to the traditional micro-spin HILIC approach.
Salt removal is crucial for efficient HILIC enrichment
Initial experiments applying the micro-spin method to enrich IgG glycopeptides resulted in no/insufficient enrichment when performed subsequently following in-solution trypsin digestion (data not shown). Applying this workflow directly on in-gel digested (glyco)peptides, however, provided the expected results. We found that the residual presence of higher salt concentrations derived from the alkylating reagents resulted in electrostatic (ionic) interactions that were compromising glycopeptide enrichment when performed subsequently after an in-solution proteolytic digest. A simple reversed phase-based desalting step could be introduced prior HILIC. This step, however, can also lead to the loss of very hydrophilic glycopeptides . To avoid such losses, we modified our in-solution sample preparation protocol by introducing a simple chloroform–methanol precipitation step immediately after reduction and alkylation to remove any excess DTT and IAA. This simple, fast and efficient step allowed us to successfully enrich glycopeptides, greatly minimise sample losses and eliminate any molecules affecting the enrichment efficiency. These optimised proteolytic sample preparation conditions were then applied for all following experiments.
Same, same but different: to spin or to “Drop-HILIC”
Using acetonitrile loading solvent conditions in our hands, the sample required approximately 3–5 min to pass through the ZIC-HILIC micro-spin column. However, when applying more viscous solvents such as isopropanol up to 20 min were required. As we intended to evaluate the influence of various organic mobile phases on the glycopeptide enrichment, we established a simplified and accelerated technique by simply co-incubating the sample with the HILIC beads, termed “Drop-HILIC”. This approach provided the opportunity to normalise incubation times and evaluate any solvent effect, which could not reasonably be achieved by the micro-spin HILIC method. Besides being easier and quicker to perform, Drop-HILIC provided additional advantages: (i) the amount of HILIC material could be optimised to accommodate variable sample amounts and (ii) it was more cost and time effective as no C18 Zip-tip columns or custom made tips were required.
We first compared the micro-spin and Drop-HILIC approaches using two well-characterised standard glycoproteins, human IgG and A1PI using 80% acetonitrile loading solvent conditions. The glycopeptide enriched fractions were subsequently analysed by nanoLC-ESI-MSMS (Fig. 2, ESM Tables S2 and S3). Both techniques yielded comparable IgG glycopeptide profiles (Fig. 2a). IgG2 provided the most abundant signals followed by the glycopeptides derived from IgG1 and IgG4. However, a slightly different trend was observed in the case of A1PI. The drop approach enriched glycopeptides A1PI-GP3 (268YLGNATAIFFLPDEGK283) significantly better. On the other hand, glycopeptides containing a larger peptide backbone (A1PI-GP4 carrying H5N4F0Na2 and H3N3F0Na1 as well as A1PI-GP1) were not enriched with similar efficiency (Fig. 2b). In order to evaluate whether the one-minute incubation time affects Drop-HILIC enrichment efficiency for these larger glycopeptides, possibly due to inadequate phase partitioning, we additionally tested various incubation times.
Incubation Time Does Not Influence Drop-HILIC Enrichment Efficiency
Various incubation times (1, 3 and 5 min) did not show any significant changes in the enrichment efficiency of IgG glycopeptides under standard conditions. Longer incubation times did also not improve the enrichment of the larger hydrophobic A1PI glycopeptides, indicating that the Drop-HILIC approach exhibited some limitations for efficiently enriching such large, >25 amino acid long, comparably hydrophobic glycopeptides. Despite these observed limitations the Drop-HILIC approach provided glyco-profile results comparable to the classical micro-spin HILIC method for IgG and A1PI-GP3 (ESM Fig. S1). Drop-HILIC, however, came with the advantages of being significantly quicker and cheaper to perform. As the incubation time did not show any influence on the enrichment efficiency all further experiments were performed using 1 min incubation times for the evaluation of any effect the organic mobile phase has on glycopeptide enrichment efficiency.
It is all about the solvent—influence of the solvent system on ZIC-HILIC glycopeptide enrichment
The underlying mechanism of analyte retention in ZIC-HILIC is originating from hydrophilic partitioning in addition to contributions derived from minor electrostatic interactions. The use of TFA in the mobile phase nullifies any possible electrostatic interactions making hydrophilic partitioning the only cause for analyte retention . For an efficient enrichment, the ideal mobile phase should be water miscible, but not contribute any hydrogen donor or acceptor functionalities. A completely aprotic solvent such as acetonitrile embraces this particular feature and thus is often used for glycopeptide enrichment by HILIC SPE. However, a systematic evaluation of any mobile phase effect on glycopeptide enrichment is still lacking. The optimised sample preparation step in combination with our simple, fast and equally efficient Drop-HILIC approach allowed us to evaluate the influence of various MS compatible mobile phase solvents on ZIC-HILIC glycopeptide enrichment efficiency.
First, we evaluated the solvent effect in a defined system using glycoform-specific synthetic N-glycopeptides spiked into the background of tryptically digested BSA (please refer ESM – Glycopeptide synthesis section). The synthetic glycopeptides corresponded to the tryptic glycopeptide sequences present in IgG1, IgG2 and IgG3 and carried a bianntenary, disialylated N-glycan (Fig. 1). These compounds were mixed with tryptic BSA derived peptides in molar ratios 3:1:1:3 (IgG1/IgG2/IgG3/BSA), and glycopeptide enrichment was performed using the Drop-HILIC technique. Our initial results indicated that methanol is a non-favoured mobile phase for this purpose (data not shown). Due to its strong tendencies to form hydrogen bonds, methanol effectively competes for the active sites on the stationary phase and is thereby perturbing hydrophilic partitioning. This subsequently resulted in strongly reduced glycopeptide retention and thus was not further evaluated . The glycopeptide-enriched fractions from the other solvents (ACN, EtOH, IPA), however, were analysed by RP-nano LC-ESI-IT-MSMS.
We observed a mobile phase solvent dependency in the selectivity and efficiency for glycopeptide enrichment (Fig. 3a), which was also considerably influenced by the hydrophilicity of the peptide backbone (ESM Table S2). The synthetic IgG1 glycopeptide was efficiently enriched in a similar manner by all three solvents, while a strong mobile phase dependency was observed for IgG2 and IgG3 synthetic glycopeptides. Ethanol significantly enriched the synthetic IgG3 (75.43 ± 8.42%) and IgG2 (88.58 ± 6.76%) glycopeptides better than ACN or IPA while the IgG2 glycopeptide could not be enriched at all using IPA in the background of tryptic BSA peptides (Fig. 3a). Interestingly, this peptide is also the most hydrophobic of the three synthetic N-glycopeptides (GRAVY score of −1.60, see also ESM Table S4). The data suggested that either the hydrophilic BSA peptides outcompeted the IgG2 synthetic N-glycopeptide or suppressed its ionisation, making it not detectable under the used conditions. When excess molar ratios of BSA were applied (IgG1/IgG2/IgG3/BSA = 3:1:1:6 and 3:1:1:10), glycopeptide enrichment efficiency was compromised especially in the case of EtOH and IPA as the abundances of co-enriched peptides clearly increased (ESM Fig. S4). Under the tested conditions, ACN provided the best compromise for the retention of all three synthetic glycopeptides while keeping the number of co-enriched BSA peptides at a low level. Nevertheless, with all three mobile phases, numerous peptides were co-enriched in a mobile phase dependent manner (Fig. 3b, ESM Table S5).
Inspired from these results, we next studied the loading solvent influence on ZIC-HILIC glycopeptide enrichment using individual standard glycoprotein digests of IgG and A1PI. Also for these low-complex samples, a loading solvent-dependant selectivity was found. The human IgG glycopeptides showed similar results as obtained for the synthetic ones, with the exception that IPA was a suitable solvent to enrich the IgG2 glycopeptide (Fig. 3c), indicating that the BSA tryptic peptide background was interfering with its enrichment in earlier experiments (Fig. 2a). In the case of A1PI, however, isopropanol provided the best compromise for the simultaneous enrichment of hydrophilic and hydrophobic glycopeptides (Fig. 3d). This can possibly be explained by the fact that not all (glyco)peptides were equally soluble under 80% organic mobile phase conditions . This hypothesis was also supported by the different identified co-enriched unmodified peptides that were found for the individual solvents (ESM Table S6A-
S6C, Fig. S5). As a consequence of this insolubility, an insufficient enrichment of certain glycopeptide species was observed when using EtOH. Our results indicated that analyte retention in HILIC was not just controlled by hydrophilic partition but that more complex mechanisms occurring prior sample loading and at the interface of the stationary polar and organic mobile phase during sample solvation were also influencing enrichment efficiency.
Glycopeptide enrichment from human serum using Drop-HILIC
Finally, glycopeptide enrichment from human serum before and after depletion of the four most abundant proteins (albumin, A1PI, transferrin and haptoglobin) was evaluated. As observed for the purified standard glycoproteins, the number of enriched glycopeptides varied in a solvent dependant manner (Fig. 4a–c; ESM Table S7). Glycopeptide enrichment efficiencies were determined taking the presence of identified co-enriched, non-glycosylated peptides as an indicator, while a simple automated glycopeptide classification feature available in the ProteinScape software tool was used to establish the number of enriched glycopeptides. After manual verification of the MSMS spectra for oxonium ions, just hits with a minimum oxonium ion intensity score of ≥60 were considered as glycopeptides.
A significant number of relatively low molecular weight, non-glycosylated peptides (between 1000 and 2200 Da) were frequently co-enriched. The degree of non-specific enrichment of higher molecular weight peptides was found to largely dependent on the used solvent (ESM Table S8). A high number of human serum albumin-derived peptides was co-enriched by all loading solvents, but each solvent co-enriched an individual peptide subset (Fig. 4d–f). The enriched glycopeptide fractions were also treated with PNGase F and analysed by RP-nano LC-ESI-MSMS to simply identify enriched, previously glycosylated peptides. The use of acetonitrile, ethanol and isopropanol, respectively, as loading solvent resulted in the identification of 26, 28 and 33 non-redundant, previously glycosylated peptides carrying the deamidated N-glycosylation sequence motif. So despite the fact that isopropanol also co-enriched the highest number of unmodified peptides, it also provided the highest number of glycopeptides. These results clearly emphasise that besides (glyco)peptide hydrophilicity sample solvation plays an important role in ZIC-HILIC glycopeptide enrichment.